diff options
Diffstat (limited to 'xlators/mgmt')
39 files changed, 35849 insertions, 7132 deletions
diff --git a/xlators/mgmt/glusterd/src/Makefile.am b/xlators/mgmt/glusterd/src/Makefile.am index 817f02735..933c44019 100644 --- a/xlators/mgmt/glusterd/src/Makefile.am +++ b/xlators/mgmt/glusterd/src/Makefile.am @@ -1,27 +1,50 @@ xlator_LTLIBRARIES = glusterd.la xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/mgmt -glusterd_la_CPPFLAGS = "-DFILTERDIR=\"$(libdir)/glusterfs/$(PACKAGE_VERSION)/filter\"" -glusterd_la_LDFLAGS = -module -avoidversion $(LIBXML2_LIBS) -glusterd_la_SOURCES = glusterd.c glusterd-handler.c glusterd-sm.c glusterd-op-sm.c \ - glusterd-utils.c glusterd-rpc-ops.c glusterd-store.c glusterd-handshake.c \ - glusterd-pmap.c glusterd-volgen.c glusterd-rebalance.c glusterd-quota.c \ +glusterd_la_CPPFLAGS = $(AM_CPPFLAGS) "-DFILTERDIR=\"$(libdir)/glusterfs/$(PACKAGE_VERSION)/filter\"" +glusterd_la_LDFLAGS = -module -avoid-version +if ENABLE_BD_XLATOR +glusterd_la_LDFLAGS += -llvm2app +endif +glusterd_la_SOURCES = glusterd.c glusterd-handler.c glusterd-sm.c \ + glusterd-op-sm.c glusterd-utils.c glusterd-rpc-ops.c \ + glusterd-store.c glusterd-handshake.c glusterd-pmap.c \ + glusterd-volgen.c glusterd-rebalance.c glusterd-quota.c \ glusterd-geo-rep.c glusterd-replace-brick.c glusterd-log-ops.c \ - glusterd-volume-ops.c glusterd-brick-ops.c glusterd-mountbroker.c + glusterd-volume-ops.c glusterd-brick-ops.c glusterd-mountbroker.c \ + glusterd-syncop.c glusterd-hooks.c glusterd-volume-set.c \ + glusterd-locks.c glusterd-snapshot.c glusterd-mgmt-handler.c \ + glusterd-mgmt.c glusterd_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \ $(top_builddir)/rpc/xdr/src/libgfxdr.la \ - $(top_builddir)/rpc/rpc-lib/src/libgfrpc.la + $(top_builddir)/rpc/rpc-lib/src/libgfrpc.la \ + $(XML_LIBS) -lcrypto -noinst_HEADERS = glusterd.h glusterd-utils.h glusterd-op-sm.h glusterd-sm.h \ - glusterd-store.h glusterd-mem-types.h glusterd-pmap.h glusterd-volgen.h \ - glusterd-mountbroker.h +noinst_HEADERS = glusterd.h glusterd-utils.h glusterd-op-sm.h \ + glusterd-sm.h glusterd-store.h glusterd-mem-types.h \ + glusterd-pmap.h glusterd-volgen.h glusterd-mountbroker.h \ + glusterd-syncop.h glusterd-hooks.h glusterd-locks.h \ + glusterd-mgmt.h -AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\ - -I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)\ - -I$(rpclibdir) -L$(xlatordir)/ -I$(CONTRIBDIR)/rbtree -I$(top_srcdir)/rpc/xdr/src\ - -I$(top_srcdir)/rpc/rpc-lib/src -I$(CONTRIBDIR)/uuid -I$(top_srcdir)/contrib/md5 -DSBIN_DIR=\"$(sbindir)\"\ - -DDATADIR=\"$(localstatedir)\" -DGSYNCD_PREFIX=\"$(libexecdir)/glusterfs\"\ - -DSYNCDAEMON_COMPILE=$(SYNCDAEMON_COMPILE) $(LIBXML2_CFLAGS) +AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \ + -I$(rpclibdir) -I$(CONTRIBDIR)/rbtree \ + -I$(top_srcdir)/rpc/xdr/src -I$(top_srcdir)/rpc/rpc-lib/src \ + -I$(CONTRIBDIR)/uuid \ + -DSBIN_DIR=\"$(sbindir)\" -DDATADIR=\"$(localstatedir)\" \ + -DGSYNCD_PREFIX=\"$(libexecdir)/glusterfs\"\ + -DSYNCDAEMON_COMPILE=$(SYNCDAEMON_COMPILE) $(XML_CPPFLAGS) +AM_CFLAGS = -Wall $(GF_CFLAGS) + +AM_LDFLAGS = -L$(xlatordir) CLEANFILES = + +install-data-hook: + +if GF_INSTALL_VAR_LIB_GLUSTERD + $(mkdir_p) $(localstatedir)/lib/ + (stat $(sysconfdir)/glusterd && \ + mv $(sysconfdir)/glusterd $(localstatedir)/lib/) || true; + (ln -sf $(localstatedir)/lib/glusterd $(sysconfdir)/glusterd) || true; +endif diff --git a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c index 5a4bbab77..596503c21 100644 --- a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c +++ b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" @@ -31,6 +21,7 @@ #include "glusterd-utils.h" #include "glusterd-volgen.h" #include "run.h" +#include <sys/signal.h> /* misc */ @@ -107,12 +98,19 @@ gd_addbr_validate_stripe_count (glusterd_volinfo_t *volinfo, int stripe_count, case GF_CLUSTER_TYPE_NONE: if ((volinfo->brick_count * stripe_count) == total_bricks) { /* Change the volume type */ + *type = GF_CLUSTER_TYPE_STRIPE; gf_log (THIS->name, GF_LOG_INFO, "Changing the type of volume %s from " - "None to 'stripe'", volinfo->volname); - *type = GF_CLUSTER_TYPE_STRIPE; + "'distribute' to 'stripe'", volinfo->volname); ret = 0; goto out; + } else { + snprintf (err_str, err_len, "Incorrect number of " + "bricks (%d) supplied for stripe count (%d).", + (total_bricks - volinfo->brick_count), + stripe_count); + gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str); + goto out; } break; case GF_CLUSTER_TYPE_REPLICATE: @@ -125,16 +123,25 @@ gd_addbr_validate_stripe_count (glusterd_volinfo_t *volinfo, int stripe_count, volinfo->volname); ret = 0; goto out; + } else { + snprintf (err_str, err_len, "Incorrect number of " + "bricks (%d) supplied for changing volume's " + "stripe count to %d, need at least %d bricks", + (total_bricks - volinfo->brick_count), + stripe_count, + (volinfo->replica_count * stripe_count)); + gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str); + goto out; } break; case GF_CLUSTER_TYPE_STRIPE: case GF_CLUSTER_TYPE_STRIPE_REPLICATE: if (stripe_count < volinfo->stripe_count) { snprintf (err_str, err_len, - "wrong stripe count (%d) given. " - "already have %d", + "Incorrect stripe count (%d) supplied. " + "Volume already has stripe count (%d)", stripe_count, volinfo->stripe_count); - gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str); + gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str); goto out; } if (stripe_count == volinfo->stripe_count) { @@ -170,7 +177,8 @@ out: static int gd_addbr_validate_replica_count (glusterd_volinfo_t *volinfo, int replica_count, - int total_bricks, int *type, char *err_str, int err_len) + int total_bricks, int *type, char *err_str, + int err_len) { int ret = -1; @@ -179,12 +187,20 @@ gd_addbr_validate_replica_count (glusterd_volinfo_t *volinfo, int replica_count, case GF_CLUSTER_TYPE_NONE: if ((volinfo->brick_count * replica_count) == total_bricks) { /* Change the volume type */ + *type = GF_CLUSTER_TYPE_REPLICATE; gf_log (THIS->name, GF_LOG_INFO, "Changing the type of volume %s from " - "None to 'replica'", volinfo->volname); - *type = GF_CLUSTER_TYPE_REPLICATE; + "'distribute' to 'replica'", volinfo->volname); ret = 0; goto out; + + } else { + snprintf (err_str, err_len, "Incorrect number of " + "bricks (%d) supplied for replica count (%d).", + (total_bricks - volinfo->brick_count), + replica_count); + gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str); + goto out; } break; case GF_CLUSTER_TYPE_STRIPE: @@ -197,16 +213,26 @@ gd_addbr_validate_replica_count (glusterd_volinfo_t *volinfo, int replica_count, volinfo->volname); ret = 0; goto out; + } else { + snprintf (err_str, err_len, "Incorrect number of " + "bricks (%d) supplied for changing volume's " + "replica count to %d, need at least %d " + "bricks", + (total_bricks - volinfo->brick_count), + replica_count, (volinfo->dist_leaf_count * + replica_count)); + gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str); + goto out; } break; case GF_CLUSTER_TYPE_REPLICATE: case GF_CLUSTER_TYPE_STRIPE_REPLICATE: if (replica_count < volinfo->replica_count) { snprintf (err_str, err_len, - "wrong replica count (%d) given. " - "already have %d", + "Incorrect replica count (%d) supplied. " + "Volume already has (%d)", replica_count, volinfo->replica_count); - gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str); + gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str); goto out; } if (replica_count == volinfo->replica_count) { @@ -239,8 +265,10 @@ out: } static int -gd_rmbr_validate_replica_count (glusterd_volinfo_t *volinfo, int32_t replica_count, - int32_t brick_count, char *err_str) +gd_rmbr_validate_replica_count (glusterd_volinfo_t *volinfo, + int32_t replica_count, + int32_t brick_count, char *err_str, + size_t err_len) { int ret = -1; int replica_nodes = 0; @@ -248,7 +276,7 @@ gd_rmbr_validate_replica_count (glusterd_volinfo_t *volinfo, int32_t replica_cou switch (volinfo->type) { case GF_CLUSTER_TYPE_NONE: case GF_CLUSTER_TYPE_STRIPE: - snprintf (err_str, 2048, + snprintf (err_str, err_len, "replica count (%d) option given for non replicate " "volume %s", replica_count, volinfo->volname); gf_log (THIS->name, GF_LOG_WARNING, "%s", err_str); @@ -258,7 +286,7 @@ gd_rmbr_validate_replica_count (glusterd_volinfo_t *volinfo, int32_t replica_cou case GF_CLUSTER_TYPE_STRIPE_REPLICATE: /* in remove brick, you can only reduce the replica count */ if (replica_count > volinfo->replica_count) { - snprintf (err_str, 2048, + snprintf (err_str, err_len, "given replica count (%d) option is more " "than volume %s's replica count (%d)", replica_count, volinfo->volname, @@ -267,15 +295,30 @@ gd_rmbr_validate_replica_count (glusterd_volinfo_t *volinfo, int32_t replica_cou goto out; } if (replica_count == volinfo->replica_count) { + /* This means the 'replica N' option on CLI was + redundant. Check if the total number of bricks given + for removal is same as 'dist_leaf_count' */ + if (brick_count % volinfo->dist_leaf_count) { + snprintf (err_str, err_len, + "number of bricks provided (%d) is " + "not valid. need at least %d " + "(or %dxN)", brick_count, + volinfo->dist_leaf_count, + volinfo->dist_leaf_count); + gf_log (THIS->name, GF_LOG_WARNING, "%s", + err_str); + goto out; + } ret = 1; goto out; } - replica_nodes = ((volinfo->brick_count / volinfo->replica_count) * + replica_nodes = ((volinfo->brick_count / + volinfo->replica_count) * (volinfo->replica_count - replica_count)); if (brick_count % replica_nodes) { - snprintf (err_str, 2048, + snprintf (err_str, err_len, "need %d(xN) bricks for reducing replica " "count of the volume from %d to %d", replica_nodes, volinfo->replica_count, @@ -290,11 +333,9 @@ out: return ret; } - - /* Handler functions */ int -glusterd_handle_add_brick (rpcsvc_request_t *req) +__glusterd_handle_add_brick (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; @@ -302,7 +343,6 @@ glusterd_handle_add_brick (rpcsvc_request_t *req) char *bricks = NULL; char *volname = NULL; int brick_count = 0; - char *brick_list = NULL; void *cli_rsp = NULL; char err_str[2048] = {0,}; gf_cli_rsp rsp = {0,}; @@ -318,15 +358,16 @@ glusterd_handle_add_brick (rpcsvc_request_t *req) GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, + (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; snprintf (err_str, sizeof (err_str), "Garbage args received"); goto out; } - gf_log ("glusterd", GF_LOG_INFO, "Received add brick req"); + gf_log (this->name, GF_LOG_INFO, "Received add brick req"); if (cli_req.dict.dict_len) { /* Unserialize the dictionary */ @@ -336,59 +377,62 @@ glusterd_handle_add_brick (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); snprintf (err_str, sizeof (err_str), "Unable to decode " - "the buffer"); + "the command"); goto out; } } ret = dict_get_str (dict, "volname", &volname); - gf_cmd_log ("Volume add-brick", "on volname: %s attempted", - volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); snprintf (err_str, sizeof (err_str), "Unable to get volume " "name"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } if (!(ret = glusterd_check_volume_exists (volname))) { ret = -1; - snprintf(err_str, 2048, "Volume %s does not exist", volname); - gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str); + snprintf (err_str, sizeof (err_str), "Volume %s does not exist", + volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } ret = dict_get_int32 (dict, "count", &brick_count); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get count"); snprintf (err_str, sizeof (err_str), "Unable to get volume " "brick count"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } ret = dict_get_int32 (dict, "replica-count", &replica_count); if (!ret) { - gf_log (THIS->name, GF_LOG_INFO, "replica-count is %d", + gf_log (this->name, GF_LOG_INFO, "replica-count is %d", replica_count); } ret = dict_get_int32 (dict, "stripe-count", &stripe_count); if (!ret) { - gf_log (THIS->name, GF_LOG_INFO, "stripe-count is %d", + gf_log (this->name, GF_LOG_INFO, "stripe-count is %d", stripe_count); } + if (!dict_get (dict, "force")) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get flag"); + goto out; + } ret = glusterd_volinfo_find (volname, &volinfo); if (ret) { snprintf (err_str, sizeof (err_str), "Unable to get volinfo " "for volume name %s", volname); - gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } @@ -404,10 +448,10 @@ glusterd_handle_add_brick (rpcsvc_request_t *req) goto brick_val; if ((brick_count % volinfo->dist_leaf_count) != 0) { - snprintf(err_str, 2048, "Incorrect number of bricks" - " supplied %d with count %d", + snprintf (err_str, sizeof (err_str), "Incorrect number " + "of bricks supplied %d with count %d", brick_count, volinfo->dist_leaf_count); - gf_log("glusterd", GF_LOG_ERROR, "%s", err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); ret = -1; goto out; } @@ -423,7 +467,7 @@ glusterd_handle_add_brick (rpcsvc_request_t *req) err_str, sizeof (err_str)); if (ret == -1) { - gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } @@ -433,7 +477,7 @@ glusterd_handle_add_brick (rpcsvc_request_t *req) ret = dict_set_int32 (dict, "stripe-count", stripe_count); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to set the stripe-count in dict"); goto out; } @@ -445,7 +489,7 @@ glusterd_handle_add_brick (rpcsvc_request_t *req) &type, err_str, sizeof (err_str)); if (ret == -1) { - gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } @@ -455,7 +499,7 @@ glusterd_handle_add_brick (rpcsvc_request_t *req) ret = dict_set_int32 (dict, "replica-count", replica_count); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to set the replica-count in dict"); goto out; } @@ -465,51 +509,111 @@ brick_val: if (ret) { snprintf (err_str, sizeof (err_str), "Unable to get volume " "bricks"); - gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } - gf_cmd_log ("Volume add-brick", "volname: %s type %d count:%d bricks:%s" - ,volname, volinfo->type, brick_count, brick_list); - if (type != volinfo->type) { ret = dict_set_int32 (dict, "type", type); if (ret) - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to set the new type in dict"); } - ret = glusterd_op_begin (req, GD_OP_ADD_BRICK, dict); + ret = glusterd_op_begin_synctask (req, GD_OP_ADD_BRICK, dict); out: - gf_cmd_log ("Volume add-brick","on volname: %s %s", volname, - (ret != 0)? "FAILED" : "SUCCESS"); if (ret) { - if (dict) - dict_unref (dict); rsp.op_ret = -1; rsp.op_errno = 0; if (err_str[0] == '\0') snprintf (err_str, sizeof (err_str), "Operation failed"); rsp.op_errstr = err_str; cli_rsp = &rsp; - glusterd_submit_reply(req, cli_rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf_cli_rsp); + glusterd_to_cli (req, cli_rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_cli_rsp, dict); ret = 0; //sent error to cli, prevent second reply } - glusterd_friend_sm (); - glusterd_op_sm (); + free (cli_req.dict.dict_val); //its malloced by xdr + + return ret; +} + +int +glusterd_handle_add_brick (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_add_brick); +} + +static int +subvol_matcher_init (int **subvols, int count) +{ + int ret = -1; + + *subvols = GF_CALLOC (count, sizeof(int), gf_gld_mt_int); + if (*subvols) + ret = 0; + + return ret; +} + +static void +subvol_matcher_update (int *subvols, glusterd_volinfo_t *volinfo, + glusterd_brickinfo_t *brickinfo) +{ + glusterd_brickinfo_t *tmp = NULL; + int32_t sub_volume = 0; + int pos = 0; + + list_for_each_entry (tmp, &volinfo->bricks, brick_list) { + + if (strcmp (tmp->hostname, brickinfo->hostname) || + strcmp (tmp->path, brickinfo->path)) { + pos++; + continue; + } + gf_log (THIS->name, GF_LOG_DEBUG, LOGSTR_FOUND_BRICK, + brickinfo->hostname, brickinfo->path, + volinfo->volname); + sub_volume = (pos / volinfo->dist_leaf_count); + subvols[sub_volume]++; + break; + } + +} - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val); //its malloced by xdr +static int +subvol_matcher_verify (int *subvols, glusterd_volinfo_t *volinfo, char *err_str, + size_t err_len, char *vol_type) +{ + int i = 0; + int ret = 0; + + do { + + if (subvols[i] % volinfo->dist_leaf_count == 0) { + continue; + } else { + ret = -1; + snprintf (err_str, err_len, + "Bricks not from same subvol for %s", vol_type); + gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str); + break; + } + } while (++i < volinfo->subvol_count); return ret; } +static void +subvol_matcher_destroy (int *subvols) +{ + GF_FREE (subvols); +} int -glusterd_handle_remove_brick (rpcsvc_request_t *req) +__glusterd_handle_remove_brick (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; @@ -521,10 +625,7 @@ glusterd_handle_remove_brick (rpcsvc_request_t *req) int i = 1; glusterd_volinfo_t *volinfo = NULL; glusterd_brickinfo_t *brickinfo = NULL; - int32_t pos = 0; - int32_t sub_volume = 0; - int32_t sub_volume_start = 0; - int32_t sub_volume_end = 0; + int *subvols = NULL; glusterd_brickinfo_t *tmp = NULL; char err_str[2048] = {0}; gf_cli_rsp rsp = {0,}; @@ -536,18 +637,23 @@ glusterd_handle_remove_brick (rpcsvc_request_t *req) int found = 0; int diff_count = 0; char *volname = 0; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, + (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; + snprintf (err_str, sizeof (err_str), "Received garbage args"); goto out; } - gf_log ("glusterd", GF_LOG_INFO, "Received rem brick req"); + gf_log (this->name, GF_LOG_INFO, "Received rem brick req"); if (cli_req.dict.dict_len) { /* Unserialize the dictionary */ @@ -557,52 +663,61 @@ glusterd_handle_remove_brick (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); goto out; } } ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Unable to get volname"); + snprintf (err_str, sizeof (err_str), "Unable to get volume " + "name"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } - gf_cmd_log ("Volume remove-brick","on volname: %s attempted", volname); ret = dict_get_int32 (dict, "count", &count); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get count"); + snprintf (err_str, sizeof (err_str), "Unable to get brick " + "count"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } ret = glusterd_volinfo_find (volname, &volinfo); if (ret) { - snprintf (err_str, 2048, "Volume %s does not exist", - volname); - gf_log ("", GF_LOG_ERROR, "%s", err_str); + snprintf (err_str, sizeof (err_str),"Volume %s does not exist", + volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } ret = dict_get_int32 (dict, "replica-count", &replica_count); if (!ret) { - gf_log (THIS->name, GF_LOG_INFO, + gf_log (this->name, GF_LOG_INFO, "request to change replica-count to %d", replica_count); ret = gd_rmbr_validate_replica_count (volinfo, replica_count, - count, err_str); + count, err_str, + sizeof (err_str)); if (ret < 0) { - /* logging and error msg are done in above function itself */ + /* logging and error msg are done in above function + itself */ goto out; } dict_del (dict, "replica-count"); if (ret) { replica_count = 0; } else { - ret = dict_set_int32 (dict, "replica-count", replica_count); + ret = dict_set_int32 (dict, "replica-count", + replica_count); if (ret) { - gf_log (THIS->name, GF_LOG_WARNING, - "failed to set the replica_count in dict"); + gf_log (this->name, GF_LOG_WARNING, + "failed to set the replica_count " + "in dict"); goto out; } } @@ -619,41 +734,54 @@ glusterd_handle_remove_brick (rpcsvc_request_t *req) strcpy (vol_type, "distribute"); } - /* Do not allow remove-brick if the volume is plain stripe */ + /* Do not allow remove-brick if the volume is a stripe volume*/ if ((volinfo->type == GF_CLUSTER_TYPE_STRIPE) && (volinfo->brick_count == volinfo->stripe_count)) { - snprintf (err_str, 2048, "Removing brick from a plain stripe is not allowed"); - gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str); + snprintf (err_str, sizeof (err_str), + "Removing brick from a stripe volume is not allowed"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); ret = -1; goto out; } - /* Do not allow remove-brick if the bricks given is less than the replica count - or stripe count */ - if (!replica_count && (volinfo->type != GF_CLUSTER_TYPE_NONE) && - !(volinfo->brick_count <= volinfo->dist_leaf_count)) { - if (volinfo->dist_leaf_count && - (count % volinfo->dist_leaf_count)) { - snprintf (err_str, 2048, "Remove brick incorrect" - " brick count of %d for %s %d", - count, vol_type, volinfo->dist_leaf_count); - gf_log ("", GF_LOG_ERROR, "%s", err_str); - ret = -1; - goto out; - } - } - if (!replica_count && (volinfo->type == GF_CLUSTER_TYPE_STRIPE_REPLICATE) && (volinfo->brick_count == volinfo->dist_leaf_count)) { - snprintf (err_str, 2048, "Removing bricks from stripe-replicate" + snprintf (err_str, sizeof(err_str), + "Removing bricks from stripe-replicate" " configuration is not allowed without reducing " "replica or stripe count explicitly."); - gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); ret = -1; goto out; } + if (!replica_count && + (volinfo->type == GF_CLUSTER_TYPE_REPLICATE) && + (volinfo->brick_count == volinfo->dist_leaf_count)) { + snprintf (err_str, sizeof (err_str), + "Removing bricks from replicate configuration " + "is not allowed without reducing replica count " + "explicitly."); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + ret = -1; + goto out; + } + + /* Do not allow remove-brick if the bricks given is less than + the replica count or stripe count */ + if (!replica_count && (volinfo->type != GF_CLUSTER_TYPE_NONE)) { + if (volinfo->dist_leaf_count && + (count % volinfo->dist_leaf_count)) { + snprintf (err_str, sizeof (err_str), "Remove brick " + "incorrect brick count of %d for %s %d", + count, vol_type, volinfo->dist_leaf_count); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + ret = -1; + goto out; + } + } + brick_list = GF_MALLOC (120000 * sizeof(*brick_list),gf_common_mt_char); if (!brick_list) { @@ -662,23 +790,32 @@ glusterd_handle_remove_brick (rpcsvc_request_t *req) } strcpy (brick_list, " "); + + if ((volinfo->type != GF_CLUSTER_TYPE_NONE) && + (volinfo->subvol_count > 1)) { + ret = subvol_matcher_init (&subvols, volinfo->subvol_count); + if (ret) + goto out; + } + while ( i <= count) { - snprintf (key, 256, "brick%d", i); + snprintf (key, sizeof (key), "brick%d", i); ret = dict_get_str (dict, key, &brick); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get %s", key); + snprintf (err_str, sizeof (err_str), "Unable to get %s", + key); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } - gf_log ("", GF_LOG_DEBUG, "Remove brick count %d brick: %s", - i, brick); + gf_log (this->name, GF_LOG_DEBUG, "Remove brick count %d brick:" + " %s", i, brick); ret = glusterd_volume_brickinfo_get_by_brick(brick, volinfo, - &brickinfo, - GF_PATH_COMPLETE); + &brickinfo); if (ret) { - snprintf(err_str, 2048,"Incorrect brick %s for volume" - " %s", brick, volname); - gf_log ("", GF_LOG_ERROR, "%s", err_str); + snprintf (err_str, sizeof (err_str), "Incorrect brick " + "%s for volume %s", brick, volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } strcat(brick_list, brick); @@ -699,7 +836,7 @@ glusterd_handle_remove_brick (rpcsvc_request_t *req) found = 0; list_for_each_entry (tmp, &volinfo->bricks, brick_list) { tmp_brick_idx++; - gf_log (THIS->name, GF_LOG_TRACE, + gf_log (this->name, GF_LOG_TRACE, "validate brick %s:%s (%d %d %d)", tmp->hostname, tmp->path, tmp_brick_idx, brick_index, volinfo->replica_count); @@ -717,8 +854,9 @@ glusterd_handle_remove_brick (rpcsvc_request_t *req) if (found) continue; - snprintf(err_str, 2048,"Bricks are from same subvol"); - gf_log (THIS->name, GF_LOG_INFO, + snprintf (err_str, sizeof (err_str), "Bricks are from " + "same subvol"); + gf_log (this->name, GF_LOG_INFO, "failed to validate brick %s:%s (%d %d %d)", tmp->hostname, tmp->path, tmp_brick_idx, brick_index, volinfo->replica_count); @@ -727,72 +865,146 @@ glusterd_handle_remove_brick (rpcsvc_request_t *req) goto out; } - pos = 0; - list_for_each_entry (tmp, &volinfo->bricks, brick_list) { - - if (strcmp (tmp->hostname,brickinfo->hostname) || - strcmp (tmp->path, brickinfo->path)) { - pos++; - continue; - } + /* Find which subvolume the brick belongs to */ + subvol_matcher_update (subvols, volinfo, brickinfo); + } - gf_log ("", GF_LOG_INFO, "Found brick"); - if (!sub_volume && (volinfo->dist_leaf_count > 1)) { - sub_volume = (pos / volinfo->dist_leaf_count) + 1; - sub_volume_start = (volinfo->dist_leaf_count * - (sub_volume - 1)); - sub_volume_end = (volinfo->dist_leaf_count * - sub_volume) - 1; - } else { - if (pos < sub_volume_start || - pos >sub_volume_end) { - ret = -1; - snprintf(err_str, 2048,"Bricks not from" - " same subvol for %s", - vol_type); - gf_log ("", GF_LOG_ERROR, - "%s", err_str); - goto out; - } - } - break; - } + /* Check if the bricks belong to the same subvolumes.*/ + if ((volinfo->type != GF_CLUSTER_TYPE_NONE) && + (volinfo->subvol_count > 1)) { + ret = subvol_matcher_verify (subvols, volinfo, + err_str, sizeof(err_str), + vol_type); + if (ret) + goto out; } - gf_cmd_log ("Volume remove-brick","volname: %s count:%d bricks:%s", - volname, count, brick_list); - ret = glusterd_op_begin (req, GD_OP_REMOVE_BRICK, dict); - gf_cmd_log ("Volume remove-brick","on volname: %s %s", volname, - (ret) ? "FAILED" : "SUCCESS"); + ret = glusterd_op_begin_synctask (req, GD_OP_REMOVE_BRICK, dict); out: if (ret) { - if (dict) - dict_unref (dict); rsp.op_ret = -1; rsp.op_errno = 0; if (err_str[0] == '\0') - snprintf (err_str, sizeof (err_str), "Operation failed"); - gf_log ("", GF_LOG_ERROR, "%s", err_str); + snprintf (err_str, sizeof (err_str), + "Operation failed"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); rsp.op_errstr = err_str; cli_rsp = &rsp; - glusterd_submit_reply(req, cli_rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf_cli_rsp); + glusterd_to_cli (req, cli_rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_cli_rsp, dict); ret = 0; //sent error to cli, prevent second reply } - if (brick_list) - GF_FREE (brick_list); - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val); //its malloced by xdr - glusterd_friend_sm (); - glusterd_op_sm (); + GF_FREE (brick_list); + subvol_matcher_destroy (subvols); + free (cli_req.dict.dict_val); //its malloced by xdr return ret; } +int +glusterd_handle_remove_brick (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_remove_brick); +} + +static int +_glusterd_restart_gsync_session (dict_t *this, char *key, + data_t *value, void *data) +{ + char *slave = NULL; + char *slave_buf = NULL; + char *path_list = NULL; + char *slave_vol = NULL; + char *slave_ip = NULL; + char *conf_path = NULL; + char **errmsg = NULL; + int ret = -1; + glusterd_gsync_status_temp_t *param = NULL; + gf_boolean_t is_running = _gf_false; + + param = (glusterd_gsync_status_temp_t *)data; + + GF_ASSERT (param); + GF_ASSERT (param->volinfo); + + slave = strchr(value->data, ':'); + if (slave) { + slave++; + slave_buf = gf_strdup (slave); + if (!slave_buf) { + gf_log ("", GF_LOG_ERROR, + "Failed to gf_strdup"); + ret = -1; + goto out; + } + } + else + return 0; + + ret = dict_set_dynstr (param->rsp_dict, "slave", slave_buf); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store slave"); + if (slave_buf) + GF_FREE(slave_buf); + goto out; + } + + ret = glusterd_get_slave_details_confpath (param->volinfo, + param->rsp_dict, + &slave_ip, &slave_vol, + &conf_path, errmsg); + if (ret) { + if (*errmsg) + gf_log ("", GF_LOG_ERROR, "%s", *errmsg); + else + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave or confpath details."); + goto out; + } + + /* In cases that gsyncd is not running, we will not invoke it + * because of add-brick. */ + ret = glusterd_check_gsync_running_local (param->volinfo->volname, + slave, conf_path, + &is_running); + if (ret) { + gf_log ("", GF_LOG_ERROR, "gsync running validation failed."); + goto out; + } + if (_gf_false == is_running) { + gf_log ("", GF_LOG_DEBUG, "gsync session for %s and %s is" + " not running on this node. Hence not restarting.", + param->volinfo->volname, slave); + ret = 0; + goto out; + } + + ret = glusterd_get_local_brickpaths (param->volinfo, &path_list); + if (!path_list) { + gf_log ("", GF_LOG_DEBUG, "This node not being part of" + " volume should not be running gsyncd. Hence" + " no gsyncd process to restart."); + ret = 0; + goto out; + } + + ret = glusterd_check_restart_gsync_session (param->volinfo, slave, + param->rsp_dict, path_list, + conf_path, 0); + if (ret) + gf_log ("", GF_LOG_ERROR, + "Unable to restart gsync session."); + +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d.", ret); + return ret; +} /* op-sm */ @@ -800,17 +1012,21 @@ int glusterd_op_perform_add_bricks (glusterd_volinfo_t *volinfo, int32_t count, char *bricks, dict_t *dict) { - glusterd_brickinfo_t *brickinfo = NULL; - char *brick = NULL; - int32_t i = 1; - char *brick_list = NULL; - char *free_ptr1 = NULL; - char *free_ptr2 = NULL; - char *saveptr = NULL; - int32_t ret = -1; - int32_t stripe_count = 0; - int32_t replica_count = 0; - int32_t type = 0; + char *brick = NULL; + int32_t i = 1; + char *brick_list = NULL; + char *free_ptr1 = NULL; + char *free_ptr2 = NULL; + char *saveptr = NULL; + int32_t ret = -1; + int32_t stripe_count = 0; + int32_t replica_count = 0; + int32_t type = 0; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_gsync_status_temp_t param = {0, }; + gf_boolean_t restart_needed = 0; + char msg[1024] __attribute__((unused)) = {0, }; + int caps = 0; GF_ASSERT (volinfo); @@ -839,7 +1055,7 @@ glusterd_op_perform_add_bricks (glusterd_volinfo_t *volinfo, int32_t count, } while ( i <= count) { - ret = glusterd_brickinfo_from_brick (brick, &brickinfo); + ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo); if (ret) goto out; @@ -868,46 +1084,94 @@ glusterd_op_perform_add_bricks (glusterd_volinfo_t *volinfo, int32_t count, if (stripe_count) { volinfo->stripe_count = stripe_count; } - volinfo->dist_leaf_count = (volinfo->stripe_count * - volinfo->replica_count); + volinfo->dist_leaf_count = glusterd_get_dist_leaf_count (volinfo); /* backward compatibility */ volinfo->sub_count = ((volinfo->dist_leaf_count == 1) ? 0: volinfo->dist_leaf_count); + volinfo->subvol_count = (volinfo->brick_count / + volinfo->dist_leaf_count); + + ret = glusterd_create_volfiles_and_notify_services (volinfo); + if (ret) + goto out; + + ret = 0; + if (GLUSTERD_STATUS_STARTED != volinfo->status) + goto out; + brick_list = gf_strdup (bricks); free_ptr2 = brick_list; i = 1; if (count) brick = strtok_r (brick_list+1, " \n", &saveptr); - - ret = glusterd_create_volfiles_and_notify_services (volinfo); - if (ret) - goto out; +#ifdef HAVE_BD_XLATOR + if (brickinfo->vg[0]) + caps = CAPS_BD | CAPS_THIN | + CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT; +#endif while (i <= count) { - ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, - &brickinfo, - GF_PATH_PARTIAL); + &brickinfo); if (ret) goto out; +#ifdef HAVE_BD_XLATOR + /* Check for VG/thin pool if its BD volume */ + if (brickinfo->vg[0]) { + ret = glusterd_is_valid_vg (brickinfo, 0, msg); + if (ret) { + gf_log (THIS->name, GF_LOG_CRITICAL, "%s", msg); + goto out; + } + /* if anyone of the brick does not have thin support, + disable it for entire volume */ + caps &= brickinfo->caps; + } else + caps = 0; +#endif - if (GLUSTERD_STATUS_STARTED == volinfo->status) { - ret = glusterd_brick_start (volinfo, brickinfo); - if (ret) + if (uuid_is_null (brickinfo->uuid)) { + ret = glusterd_resolve_brick (brickinfo); + if (ret) { + gf_log ("", GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK, + brickinfo->hostname, brickinfo->path); goto out; + } } + + ret = glusterd_brick_start (volinfo, brickinfo, + _gf_true); + if (ret) + goto out; i++; brick = strtok_r (NULL, " \n", &saveptr); + + /* Check if the brick is added in this node, and set + * the restart_needed flag. */ + if ((!uuid_compare (brickinfo->uuid, MY_UUID)) && + !restart_needed) { + restart_needed = 1; + gf_log ("", GF_LOG_DEBUG, + "Restart gsyncd session, if it's already " + "running."); + } } + /* If the restart_needed flag is set, restart gsyncd sessions for that + * particular master with all the slaves. */ + if (restart_needed) { + param.rsp_dict = dict; + param.volinfo = volinfo; + dict_foreach (volinfo->gsync_slaves, + _glusterd_restart_gsync_session, ¶m); + } + volinfo->caps = caps; out: - if (free_ptr1) - GF_FREE (free_ptr1); - if (free_ptr2) - GF_FREE (free_ptr2); + GF_FREE (free_ptr1); + GF_FREE (free_ptr2); gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; @@ -919,7 +1183,6 @@ glusterd_op_perform_remove_brick (glusterd_volinfo_t *volinfo, char *brick, int force, int *need_migrate) { glusterd_brickinfo_t *brickinfo = NULL; - char *dup_brick = NULL; int32_t ret = -1; glusterd_conf_t *priv = NULL; @@ -929,12 +1192,8 @@ glusterd_op_perform_remove_brick (glusterd_volinfo_t *volinfo, char *brick, priv = THIS->private; GF_ASSERT (priv); - dup_brick = gf_strdup (brick); - if (!dup_brick) - goto out; - - ret = glusterd_volume_brickinfo_get_by_brick (dup_brick, volinfo, - &brickinfo, GF_PATH_COMPLETE); + ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, + &brickinfo); if (ret) goto out; @@ -942,30 +1201,27 @@ glusterd_op_perform_remove_brick (glusterd_volinfo_t *volinfo, char *brick, if (ret) goto out; - if (!uuid_compare (brickinfo->uuid, priv->uuid)) { + glusterd_volinfo_reset_defrag_stats (volinfo); + + if (!uuid_compare (brickinfo->uuid, MY_UUID)) { /* Only if the brick is in this glusterd, do the rebalance */ if (need_migrate) *need_migrate = 1; } if (force) { - if (GLUSTERD_STATUS_STARTED == volinfo->status) { - ret = glusterd_brick_stop (volinfo, brickinfo); - if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Unable to stop " - "glusterfs, ret: %d", ret); - goto out; - } + ret = glusterd_brick_stop (volinfo, brickinfo, + _gf_true); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Unable to stop " + "glusterfs, ret: %d", ret); } - glusterd_delete_brick (volinfo, brickinfo); goto out; } brickinfo->decommissioned = 1; + ret = 0; out: - if (dup_brick) - GF_FREE (dup_brick); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -989,6 +1245,7 @@ glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr) gf_boolean_t brick_alloc = _gf_false; char *all_bricks = NULL; char *str_ret = NULL; + gf_boolean_t is_force = _gf_false; priv = THIS->private; if (!priv) @@ -1008,6 +1265,20 @@ glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr) goto out; } + ret = glusterd_validate_volume_id (dict, volinfo); + if (ret) + goto out; + + if (glusterd_is_rb_ongoing (volinfo)) { + snprintf (msg, sizeof (msg), "Replace brick is in progress on " + "volume %s. Please retry after replace-brick " + "operation is committed or aborted", volname); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + ret = -1; + goto out; + } + if (glusterd_is_defrag_on(volinfo)) { snprintf (msg, sizeof(msg), "Volume name %s rebalance is in " "progress. Please retry after completion", volname); @@ -1028,23 +1299,14 @@ glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr) goto out; } + is_force = dict_get_str_boolean (dict, "force", _gf_false); + if (bricks) { brick_list = gf_strdup (bricks); all_bricks = gf_strdup (bricks); free_ptr = brick_list; } - /* Check whether any of the bricks given is the destination brick of the - replace brick running */ - - str_ret = glusterd_check_brick_rb_part (all_bricks, count, volinfo); - if (str_ret) { - gf_log (THIS->name, GF_LOG_ERROR, "%s", str_ret); - *op_errstr = gf_strdup (str_ret); - ret = -1; - goto out; - } - if (count) brick = strtok_r (brick_list+1, " \n", &saveptr); @@ -1062,24 +1324,14 @@ glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr) } - ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, - &brickinfo, - GF_PATH_PARTIAL); - if (!ret) { + ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo); + if (ret) { gf_log (THIS->name, GF_LOG_ERROR, - "Adding duplicate brick: %s", brick); - ret = -1; + "Add-brick: Unable" + " to get brickinfo"); goto out; - } else { - ret = glusterd_brickinfo_from_brick (brick, &brickinfo); - if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, - "Add-brick: Unable" - " to get brickinfo"); - goto out; - } - brick_alloc = _gf_true; } + brick_alloc = _gf_true; ret = glusterd_new_brick_validate (brick, brickinfo, msg, sizeof (msg)); @@ -1089,11 +1341,22 @@ glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr) goto out; } - if (!uuid_compare (brickinfo->uuid, priv->uuid)) { - ret = glusterd_brick_create_path (brickinfo->hostname, - brickinfo->path, + if (!uuid_compare (brickinfo->uuid, MY_UUID)) { +#ifdef HAVE_BD_XLATOR + if (brickinfo->vg[0]) { + ret = glusterd_is_valid_vg (brickinfo, 1, msg); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "%s", + msg); + *op_errstr = gf_strdup (msg); + goto out; + } + } +#endif + + ret = glusterd_validate_and_create_brickpath (brickinfo, volinfo->volume_id, - 0777, op_errstr); + op_errstr, is_force); if (ret) goto out; } @@ -1106,14 +1369,11 @@ glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr) } out: - if (free_ptr) - GF_FREE (free_ptr); + GF_FREE (free_ptr); if (brick_alloc && brickinfo) glusterd_brickinfo_delete (brickinfo); - if (str_ret) - GF_FREE (str_ret); - if (all_bricks) - GF_FREE (all_bricks); + GF_FREE (str_ret); + GF_FREE (all_bricks); gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -1131,23 +1391,42 @@ glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr) char msg[2048] = {0,}; int32_t flag = 0; gf1_op_commands cmd = GF_OP_CMD_NONE; + char *task_id_str = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } ret = glusterd_volinfo_find (volname, &volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "Volume %s does not exist", volname); + gf_log (this->name, GF_LOG_ERROR, "Volume %s does not exist", volname); + goto out; + } + + ret = glusterd_validate_volume_id (dict, volinfo); + if (ret) + goto out; + + if (glusterd_is_rb_ongoing (volinfo)) { + snprintf (msg, sizeof (msg), "Replace brick is in progress on " + "volume %s. Please retry after replace-brick " + "operation is committed or aborted", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + ret = -1; goto out; } ret = dict_get_int32 (dict, "command", &flag); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get brick count"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get brick count"); goto out; } cmd = flag; @@ -1164,35 +1443,67 @@ glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr) case GF_OP_CMD_START: { + if ((volinfo->type == GF_CLUSTER_TYPE_REPLICATE) && + dict_get (dict, "replica-count")) { + snprintf (msg, sizeof(msg), "Migration of data is not " + "needed when reducing replica count. Use the" + " 'force' option"); + errstr = gf_strdup (msg); + gf_log (this->name, GF_LOG_ERROR, "%s", errstr); + goto out; + } + if (GLUSTERD_STATUS_STARTED != volinfo->status) { - snprintf (msg, sizeof (msg), "Volume %s needs to be started " - "before remove-brick (you can use 'force' or " - "'commit' to override this behavior)", + snprintf (msg, sizeof (msg), "Volume %s needs to be " + "started before remove-brick (you can use " + "'force' or 'commit' to override this " + "behavior)", volinfo->volname); + errstr = gf_strdup (msg); + gf_log (this->name, GF_LOG_ERROR, "%s", errstr); + goto out; + } + if (!gd_is_remove_brick_committed (volinfo)) { + snprintf (msg, sizeof (msg), "An earlier remove-brick " + "task exists for volume %s. Either commit it" + " or stop it before starting a new task.", volinfo->volname); errstr = gf_strdup (msg); - gf_log (THIS->name, GF_LOG_ERROR, "%s", errstr); + gf_log (this->name, GF_LOG_ERROR, "Earlier remove-brick" + " task exists for volume %s.", + volinfo->volname); goto out; } if (glusterd_is_defrag_on(volinfo)) { - errstr = gf_strdup("Rebalance is in progress. Please retry" - " after completion"); - gf_log ("glusterd", GF_LOG_ERROR, "%s", errstr); + errstr = gf_strdup("Rebalance is in progress. Please " + "retry after completion"); + gf_log (this->name, GF_LOG_ERROR, "%s", errstr); goto out; } - break; - } - case GF_OP_CMD_PAUSE: - case GF_OP_CMD_ABORT: - { - if (!volinfo->decommission_in_progress) { - errstr = gf_strdup("remove-brick is not in progress"); - gf_log ("glusterd", GF_LOG_ERROR, "%s", errstr); - goto out; + if (is_origin_glusterd (dict)) { + ret = glusterd_generate_and_set_task_id + (dict, GF_REMOVE_BRICK_TID_KEY); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to generate task-id"); + goto out; + } + } else { + ret = dict_get_str (dict, GF_REMOVE_BRICK_TID_KEY, + &task_id_str); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Missing remove-brick-id"); + ret = 0; + } } break; } + case GF_OP_CMD_STOP: + ret = 0; + break; + case GF_OP_CMD_COMMIT: if (volinfo->decommission_in_progress) { errstr = gf_strdup ("use 'force' option as migration " @@ -1207,7 +1518,7 @@ glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr) ret = dict_get_int32 (dict, "count", &brick_count); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get brick count"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get brick count"); goto out; } @@ -1220,7 +1531,7 @@ glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr) } out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); if (ret && errstr) { if (op_errstr) *op_errstr = errstr; @@ -1265,6 +1576,7 @@ glusterd_remove_brick_migrate_cbk (glusterd_volinfo_t *volinfo, brickinfo->path); brickinfo->decommissioned = 0; if (GLUSTERD_STATUS_STARTED == volinfo->status) { + /*TODO: use the 'atomic' flavour of brick_stop*/ ret = glusterd_brick_stop (volinfo, brickinfo); if (ret) { gf_log (THIS->name, GF_LOG_ERROR, @@ -1355,17 +1667,6 @@ glusterd_op_add_brick (dict_t *dict, char **op_errstr) goto out; } - /* Need to reset the defrag/rebalance status accordingly */ - switch (volinfo->defrag_status) { - case GF_DEFRAG_STATUS_FAILED: - case GF_DEFRAG_STATUS_COMPLETE: - case GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE: - case GF_DEFRAG_STATUS_MIGRATE_DATA_COMPLETE: - volinfo->defrag_status = 0; - default: - break; - } - ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); if (ret) goto out; @@ -1380,42 +1681,74 @@ out: int glusterd_op_remove_brick (dict_t *dict, char **op_errstr) { - int ret = -1; - char *volname = NULL; - glusterd_volinfo_t *volinfo = NULL; - char *brick = NULL; - int32_t count = 0; - int32_t i = 1; - char key[256] = {0,}; - int32_t flag = 0; - char err_str[4096] = {0,}; - int need_rebalance = 0; - int force = 0; - gf1_op_commands cmd = 0; - int32_t replica_count = 0; - glusterd_brickinfo_t *brickinfo = NULL; - glusterd_brickinfo_t *tmp = NULL; + int ret = -1; + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + char *brick = NULL; + int32_t count = 0; + int32_t i = 1; + char key[256] = {0,}; + int32_t flag = 0; + char err_str[4096] = {0,}; + int need_rebalance = 0; + int force = 0; + gf1_op_commands cmd = 0; + int32_t replica_count = 0; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_brickinfo_t *tmp = NULL; + char *task_id_str = NULL; + xlator_t *this = NULL; + dict_t *bricks_dict = NULL; + char *brick_tmpstr = NULL; + + this = THIS; + GF_ASSERT (this); ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } ret = glusterd_volinfo_find (volname, &volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to allocate memory"); + gf_log (this->name, GF_LOG_ERROR, "Unable to allocate memory"); goto out; } ret = dict_get_int32 (dict, "command", &flag); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get brick count"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get command"); goto out; } cmd = flag; + /* Set task-id, if available, in ctx dict for operations other than + * start + */ + if (is_origin_glusterd (dict) && (cmd != GF_OP_CMD_START)) { + if (!uuid_is_null (volinfo->rebal.rebalance_id)) { + ret = glusterd_copy_uuid_to_dict + (volinfo->rebal.rebalance_id, dict, + GF_REMOVE_BRICK_TID_KEY); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set remove-brick-id"); + goto out; + } + } + } + + /* Clear task-id, rebal.op and stored bricks on commmitting/stopping + * remove-brick */ + if ((cmd != GF_OP_CMD_START) || (cmd != GF_OP_CMD_STATUS)) { + uuid_clear (volinfo->rebal.rebalance_id); + volinfo->rebal.op = GD_OP_NONE; + dict_unref (volinfo->rebal.dict); + volinfo->rebal.dict = NULL; + } + ret = -1; switch (cmd) { case GF_OP_CMD_NONE: @@ -1425,46 +1758,44 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) ret = 0; goto out; - case GF_OP_CMD_PAUSE: - { - if (volinfo->decommission_in_progress) { - if (volinfo->defrag) { - LOCK (&volinfo->defrag->lock); - - volinfo->defrag_status = GF_DEFRAG_STATUS_PAUSED; - - UNLOCK (&volinfo->defrag->lock); - } - } - - /* no need to update anything */ - ret = 0; - goto out; - } - - case GF_OP_CMD_ABORT: + case GF_OP_CMD_STOP: { - if (volinfo->decommission_in_progress) { - if (volinfo->defrag) { - LOCK (&volinfo->defrag->lock); - - volinfo->defrag_status = GF_DEFRAG_STATUS_STOPPED; - - UNLOCK (&volinfo->defrag->lock); - } - } - /* Fall back to the old volume file */ - list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks, brick_list) { + list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks, + brick_list) { if (!brickinfo->decommissioned) continue; brickinfo->decommissioned = 0; } + ret = glusterd_create_volfiles_and_notify_services (volinfo); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to create volfiles"); + goto out; + } + + ret = glusterd_store_volinfo (volinfo, + GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to store volinfo"); + goto out; + } + ret = 0; - break; + goto out; } case GF_OP_CMD_START: + ret = dict_get_str (dict, GF_REMOVE_BRICK_TID_KEY, &task_id_str); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Missing remove-brick-id"); + ret = 0; + } else { + uuid_parse (task_id_str, volinfo->rebal.rebalance_id) ; + volinfo->rebal.op = GD_OP_REMOVE_BRICK; + } force = 0; break; @@ -1475,13 +1806,14 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) case GF_OP_CMD_COMMIT_FORCE: if (volinfo->decommission_in_progress) { - if (volinfo->defrag) { - LOCK (&volinfo->defrag->lock); + if (volinfo->rebal.defrag) { + LOCK (&volinfo->rebal.defrag->lock); /* Fake 'rebalance-complete' so the graph change happens right away */ - volinfo->defrag_status = GF_DEFRAG_STATUS_COMPLETE; + volinfo->rebal.defrag_status = + GF_DEFRAG_STATUS_COMPLETE; - UNLOCK (&volinfo->defrag->lock); + UNLOCK (&volinfo->rebal.defrag->lock); } /* Graph change happens in rebalance _cbk function, no need to do anything here */ @@ -1499,30 +1831,62 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) goto out; } - + /* Save the list of bricks for later usage. Right now this is required + * for displaying the task parameters with task status in volume status. + */ + bricks_dict = dict_new (); + if (!bricks_dict) { + ret = -1; + goto out; + } + ret = dict_set_int32 (bricks_dict, "count", count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to save remove-brick count"); + goto out; + } while ( i <= count) { snprintf (key, 256, "brick%d", i); ret = dict_get_str (dict, key, &brick); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get %s", key); + gf_log (this->name, GF_LOG_ERROR, "Unable to get %s", + key); + goto out; + } + + brick_tmpstr = gf_strdup (brick); + if (!brick_tmpstr) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "Failed to duplicate brick name"); + goto out; + } + ret = dict_set_dynstr (bricks_dict, key, brick_tmpstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add brick to dict"); goto out; } + brick_tmpstr = NULL; ret = glusterd_op_perform_remove_brick (volinfo, brick, force, - (i == 1) ? &need_rebalance : NULL); + &need_rebalance); if (ret) goto out; i++; } ret = dict_get_int32 (dict, "replica-count", &replica_count); if (!ret) { - gf_log (THIS->name, GF_LOG_INFO, + gf_log (this->name, GF_LOG_INFO, "changing replica count %d to %d on volume %s", volinfo->replica_count, replica_count, volinfo->volname); volinfo->replica_count = replica_count; - volinfo->dist_leaf_count = (volinfo->stripe_count * - replica_count); + volinfo->sub_count = replica_count; + volinfo->dist_leaf_count = glusterd_get_dist_leaf_count (volinfo); + volinfo->subvol_count = (volinfo->brick_count / + volinfo->dist_leaf_count); + if (replica_count == 1) { if (volinfo->type == GF_CLUSTER_TYPE_REPLICATE) { volinfo->type = GF_CLUSTER_TYPE_NONE; @@ -1535,39 +1899,41 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) } } } + volinfo->rebal.dict = bricks_dict; + bricks_dict = NULL; ret = glusterd_create_volfiles_and_notify_services (volinfo); if (ret) { - gf_log (THIS->name, GF_LOG_WARNING, "failed to create volfiles"); + gf_log (this->name, GF_LOG_WARNING, "failed to create volfiles"); goto out; } ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); if (ret) { - gf_log (THIS->name, GF_LOG_WARNING, "failed to store volinfo"); + gf_log (this->name, GF_LOG_WARNING, "failed to store volinfo"); goto out; } /* Need to reset the defrag/rebalance status accordingly */ - switch (volinfo->defrag_status) { + switch (volinfo->rebal.defrag_status) { case GF_DEFRAG_STATUS_FAILED: case GF_DEFRAG_STATUS_COMPLETE: - case GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE: - case GF_DEFRAG_STATUS_MIGRATE_DATA_COMPLETE: - volinfo->defrag_status = 0; + volinfo->rebal.defrag_status = 0; default: break; } if (!force && need_rebalance) { /* perform the rebalance operations */ - ret = glusterd_handle_defrag_start (volinfo, err_str, 4096, - GF_DEFRAG_CMD_START_FORCE, - glusterd_remove_brick_migrate_cbk); + ret = glusterd_handle_defrag_start + (volinfo, err_str, sizeof (err_str), + GF_DEFRAG_CMD_START_FORCE, + glusterd_remove_brick_migrate_cbk, GD_OP_REMOVE_BRICK); + if (!ret) volinfo->decommission_in_progress = 1; if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to start the rebalance"); } } else { @@ -1579,5 +1945,9 @@ out: if (ret && err_str[0] && op_errstr) *op_errstr = gf_strdup (err_str); + GF_FREE (brick_tmpstr); + if (bricks_dict) + dict_unref (bricks_dict); + return ret; } diff --git a/xlators/mgmt/glusterd/src/glusterd-geo-rep.c b/xlators/mgmt/glusterd/src/glusterd-geo-rep.c index db43ecb4d..5786694bd 100644 --- a/xlators/mgmt/glusterd/src/glusterd-geo-rep.c +++ b/xlators/mgmt/glusterd/src/glusterd-geo-rep.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" @@ -31,19 +21,203 @@ #include "glusterd-utils.h" #include "glusterd-volgen.h" #include "run.h" +#include "syscall.h" #include <signal.h> +static int +dict_get_param (dict_t *dict, char *key, char **param); + +static int +glusterd_get_statefile_name (glusterd_volinfo_t *volinfo, char *slave, + char *conf_path, char **statefile); + +static int +glusterd_get_slave_info (char *slave, char **slave_ip, + char **slave_vol, char **op_errstr); + +static int +glusterd_gsync_read_frm_status (char *path, char *buf, size_t blen); + +struct gsync_config_opt_vals_ gsync_confopt_vals[] = { + {.op_name = "change_detector", + .no_of_pos_vals = 2, + .case_sensitive = _gf_true, + .values = {"xsync", "changelog"}, + }, + {.op_name = "special_sync_mode", + .no_of_pos_vals = 2, + .case_sensitive = _gf_true, + .values = {"partial", "recover"} + }, + {.op_name = "log-level", + .no_of_pos_vals = 5, + .case_sensitive = _gf_false, + .values = {"critical", "error", "warning", "info", "debug"} + }, + {.op_name = NULL, + }, +}; + static char *gsync_reserved_opts[] = { - "gluster-command", + "gluster-command-dir", "pid-file", + "remote-gsyncd" "state-file", "session-owner", + "state-socket-unencoded", + "socketdir", + "ignore-deletes", + "local-id", + "local-path", + "slave-id", NULL }; int -glusterd_handle_gsync_set (rpcsvc_request_t *req) +__glusterd_handle_sys_exec (rpcsvc_request_t *req) +{ + int32_t ret = 0; + dict_t *dict = NULL; + gf_cli_req cli_req = {{0},}; + glusterd_op_t cli_op = GD_OP_SYS_EXEC; + glusterd_conf_t *priv = NULL; + char *host_uuid = NULL; + char err_str[2048] = {0,}; + xlator_t *this = NULL; + + GF_ASSERT (req); + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + ret = xdr_to_generic (req->msg[0], &cli_req, + (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (cli_req.dict.dict_len) { + dict = dict_new (); + if (!dict) + goto out; + + + ret = dict_unserialize (cli_req.dict.dict_val, + cli_req.dict.dict_len, + &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); + goto out; + } else { + dict->extra_stdfree = cli_req.dict.dict_val; + } + + host_uuid = gf_strdup (uuid_utoa(MY_UUID)); + if (host_uuid == NULL) { + snprintf (err_str, sizeof (err_str), "Failed to get " + "the uuid of local glusterd"); + ret = -1; + goto out; + } + + ret = dict_set_dynstr (dict, "host-uuid", host_uuid); + if (ret) + goto out; + } + + ret = glusterd_op_begin_synctask (req, cli_op, dict); + +out: + if (ret) { + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); + ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, + dict, err_str); + } + return ret; +} + +int +__glusterd_handle_copy_file (rpcsvc_request_t *req) +{ + int32_t ret = 0; + dict_t *dict = NULL; + gf_cli_req cli_req = {{0},}; + glusterd_op_t cli_op = GD_OP_COPY_FILE; + glusterd_conf_t *priv = NULL; + char *host_uuid = NULL; + char err_str[2048] = {0,}; + xlator_t *this = NULL; + + GF_ASSERT (req); + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + ret = xdr_to_generic (req->msg[0], &cli_req, + (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (cli_req.dict.dict_len) { + dict = dict_new (); + if (!dict) + goto out; + + + ret = dict_unserialize (cli_req.dict.dict_val, + cli_req.dict.dict_len, + &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); + goto out; + } else { + dict->extra_stdfree = cli_req.dict.dict_val; + } + + host_uuid = gf_strdup (uuid_utoa(MY_UUID)); + if (host_uuid == NULL) { + snprintf (err_str, sizeof (err_str), "Failed to get " + "the uuid of local glusterd"); + ret = -1; + goto out; + } + + ret = dict_set_dynstr (dict, "host-uuid", host_uuid); + if (ret) + goto out; + } + + ret = glusterd_op_begin_synctask (req, cli_op, dict); + +out: + if (ret) { + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); + ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, + dict, err_str); + } + return ret; +} + +int +__glusterd_handle_gsync_set (rpcsvc_request_t *req) { int32_t ret = 0; dict_t *dict = NULL; @@ -55,15 +229,19 @@ glusterd_handle_gsync_set (rpcsvc_request_t *req) int type = 0; glusterd_conf_t *priv = NULL; char *host_uuid = NULL; + char err_str[2048] = {0,}; + xlator_t *this = NULL; GF_ASSERT (req); - GF_ASSERT (THIS); - GF_ASSERT (THIS->private); - priv = THIS->private; + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, + (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { req->rpc_err = GARBAGE_ARGS; goto out; } @@ -77,17 +255,19 @@ glusterd_handle_gsync_set (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, "failed to " + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); goto out; } else { dict->extra_stdfree = cli_req.dict.dict_val; } - host_uuid = gf_strdup (uuid_utoa(priv->uuid)); + host_uuid = gf_strdup (uuid_utoa(MY_UUID)); if (host_uuid == NULL) { - gf_log ("glusterd", GF_LOG_ERROR, "failed to get" - "the uuid of the host machine"); + snprintf (err_str, sizeof (err_str), "Failed to get " + "the uuid of local glusterd"); ret = -1; goto out; } @@ -99,26 +279,31 @@ glusterd_handle_gsync_set (rpcsvc_request_t *req) ret = dict_get_str (dict, "master", &master); if (ret < 0) { - gf_log ("", GF_LOG_INFO, "master not found, while handling" - GEOREP" options"); + gf_log (this->name, GF_LOG_INFO, "master not found, while " + "handling "GEOREP" options"); master = "(No Master)"; } ret = dict_get_str (dict, "slave", &slave); if (ret < 0) { - gf_log ("", GF_LOG_INFO, "slave not not found, while" + gf_log (this->name, GF_LOG_INFO, "slave not found, while " "handling "GEOREP" options"); slave = "(No Slave)"; } ret = dict_get_int32 (dict, "type", &type); if (ret < 0) { - gf_log ("", GF_LOG_WARNING, "command type not found, while" - "handling "GEOREP" options"); + snprintf (err_str, sizeof (err_str), "Command type not found " + "while handling "GEOREP" options"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } switch (type) { + case GF_GSYNC_OPTION_TYPE_CREATE: + strncpy (operation, "create", sizeof (operation)); + cli_op = GD_OP_GSYNC_CREATE; + break; case GF_GSYNC_OPTION_TYPE_START: strncpy (operation, "start", sizeof (operation)); @@ -135,30 +320,38 @@ glusterd_handle_gsync_set (rpcsvc_request_t *req) case GF_GSYNC_OPTION_TYPE_STATUS: strncpy (operation, "status", sizeof (operation)); break; - case GF_GSYNC_OPTION_TYPE_ROTATE: - strncpy (operation, "rotate", sizeof(operation)); - break; } - gf_cmd_log ("volume "GEOREP, " %s command on %s,%s", operation, master, - slave); - ret = glusterd_op_begin (req, GD_OP_GSYNC_SET, dict); - gf_cmd_log ("volume "GEOREP, " %s command on %s,%s %s ", operation, - master, slave, (ret != 0)? "FAILED" : "SUCCEEDED"); + ret = glusterd_op_begin_synctask (req, cli_op, dict); out: - glusterd_friend_sm (); - glusterd_op_sm (); - if (ret) { - if (dict) - dict_unref (dict); + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, err_str); } return ret; } +int +glusterd_handle_sys_exec (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_sys_exec); +} + +int +glusterd_handle_copy_file (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_copy_file); +} + +int +glusterd_handle_gsync_set (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_gsync_set); +} /***** * @@ -180,7 +373,7 @@ glusterd_urltransform_add (runner_t *runner, const char *url) runner_add_arg (runner, url); } -static void +static int _glusterd_urltransform_add_iter (dict_t *dict, char *key, data_t *value, void *data) { runner_t *runner = (runner_t *)data; @@ -190,6 +383,8 @@ _glusterd_urltransform_add_iter (dict_t *dict, char *key, data_t *value, void *d GF_ASSERT (slave); slave++; runner_add_arg (runner, slave); + + return 0; } static void @@ -228,6 +423,7 @@ glusterd_urltransform (runner_t *runner, char ***linearrp) arr_idx = 0; for (;;) { + size_t len; line = GF_MALLOC (1024, gf_gld_mt_linebuf); if (!line) { error = _gf_true; @@ -238,21 +434,24 @@ glusterd_urltransform (runner_t *runner, char ***linearrp) NULL) break; - if (line[strlen (line) - 1] != '\n') { + len = strlen (line); + if (len == 0 || line[len - 1] != '\n') { GF_FREE (line); error = _gf_true; goto out; } - line[strlen (line) - 1] = '\0'; + line[len - 1] = '\0'; if (arr_idx == arr_len) { + void *p = linearr; arr_len <<= 1; - linearr = GF_REALLOC (linearr, arr_len); - if (!linearr) { + p = GF_REALLOC (linearr, arr_len); + if (!p) { GF_FREE (line); error = _gf_true; goto out; } + linearr = p; } linearr[arr_idx] = line; @@ -300,7 +499,7 @@ struct dictidxmark { char *ikey; }; -static void +static int _dict_mark_atindex (dict_t *dict, char *key, data_t *value, void *data) { struct dictidxmark *dim = data; @@ -309,6 +508,7 @@ _dict_mark_atindex (dict_t *dict, char *key, data_t *value, void *data) dim->ikey = key; dim->ithis++; + return 0; } static char * @@ -354,9 +554,9 @@ glusterd_get_slave (glusterd_volinfo_t *vol, const char *slaveurl, char **slavek static int -glusterd_query_extutil (char *resbuf, runner_t *runner) +glusterd_query_extutil_generic (char *resbuf, size_t blen, runner_t *runner, void *data, + int (*fcbk)(char *resbuf, size_t blen, FILE *fp, void *data)) { - char *ptr = NULL; int ret = 0; runner_redir (runner, STDOUT_FILENO, RUN_PIPE); @@ -366,92 +566,117 @@ glusterd_query_extutil (char *resbuf, runner_t *runner) return -1; } - ptr = fgets(resbuf, PATH_MAX, runner_chio (runner, STDOUT_FILENO)); - if (ptr) - resbuf[strlen(resbuf)-1] = '\0'; //strip off \n + ret = fcbk (resbuf, blen, runner_chio (runner, STDOUT_FILENO), data); - ret = runner_end (runner); + ret |= runner_end (runner); if (ret) gf_log ("", GF_LOG_ERROR, "reading data from child failed"); return ret ? -1 : 0; } -int -glusterd_gsync_get_param_file (char *prmfile, const char *param, char *master, - char *slave, char *gl_workdir) +static int +_fcbk_singleline(char *resbuf, size_t blen, FILE *fp, void *data) { - runner_t runner = {0,}; + char *ptr = NULL; + + errno = 0; + ptr = fgets (resbuf, blen, fp); + if (ptr) { + size_t len = strlen(resbuf); + if (len && resbuf[len-1] == '\n') + resbuf[len-1] = '\0'; //strip off \n + } - runinit (&runner); - runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL); - runner_argprintf (&runner, "%s/"GSYNC_CONF, gl_workdir); - runner_argprintf (&runner, ":%s", master); - runner_add_args (&runner, slave, "--config-get", NULL); - runner_argprintf (&runner, "%s-file", param); + return errno ? -1 : 0; +} - return glusterd_query_extutil (prmfile, &runner); +static int +glusterd_query_extutil (char *resbuf, runner_t *runner) +{ + return glusterd_query_extutil_generic (resbuf, PATH_MAX, runner, NULL, + _fcbk_singleline); } -int -glusterd_gsync_get_session_owner (char *master, char *slave, char *session_owner, - char *gl_workdir) +static int +_fcbk_conftodict (char *resbuf, size_t blen, FILE *fp, void *data) { - runner_t runner = {0,}; + char *ptr = NULL; + dict_t *dict = data; + char *v = NULL; - runinit(&runner); - runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL); - runner_argprintf (&runner, "%s/"GSYNC_CONF, gl_workdir); - runner_argprintf (&runner, ":%s", master); - runner_add_args (&runner, slave, "--config-get", "session-owner", - NULL); + for (;;) { + errno = 0; + ptr = fgets (resbuf, blen, fp); + if (!ptr) + break; + v = resbuf + strlen(resbuf) - 1; + while (isspace (*v)) + /* strip trailing space */ + *v-- = '\0'; + if (v == resbuf) + /* skip empty line */ + continue; + v = strchr (resbuf, ':'); + if (!v) + return -1; + *v++ = '\0'; + while (isspace (*v)) + v++; + v = gf_strdup (v); + if (!v) + return -1; + if (dict_set_dynstr (dict, resbuf, v) != 0) { + GF_FREE (v); + return -1; + } + } - return glusterd_query_extutil (session_owner, &runner); + return errno ? -1 : 0; } -int -glusterd_gsync_get_slave_log_file (char *master, char *slave, char *log_file) +static int +glusterd_gsync_get_config (char *master, char *slave, char *conf_path, dict_t *dict) { - int ret = -1; - runner_t runner = {0,}; - char uuid_str[64] = {0,}; - glusterd_conf_t *priv = NULL; - char *gl_workdir = NULL; - - GF_ASSERT(THIS); - GF_ASSERT(THIS->private); - - priv = THIS->private; + /* key + value, where value must be able to accommodate a path */ + char resbuf[256 + PATH_MAX] = {0,}; + runner_t runner = {0,}; - GF_VALIDATE_OR_GOTO("gsyncd", master, out); - GF_VALIDATE_OR_GOTO("gsyncd", slave, out); + runinit (&runner); + runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL); + runner_argprintf (&runner, "%s", conf_path); + runner_argprintf (&runner, ":%s", master); + runner_add_args (&runner, slave, "--config-get-all", NULL); - gl_workdir = priv->workdir; + return glusterd_query_extutil_generic (resbuf, sizeof (resbuf), + &runner, dict, _fcbk_conftodict); +} - /* get the session owner for the master-slave session */ - ret = glusterd_gsync_get_session_owner (master, slave, uuid_str, - gl_workdir); - if (ret) - goto out; +static int +glusterd_gsync_get_param_file (char *prmfile, const char *param, char *master, + char *slave, char *conf_path) +{ + runner_t runner = {0,}; - /* get the log file for the slave */ - runinit(&runner); + runinit (&runner); runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL); - runner_argprintf (&runner, "%s/"GSYNC_CONF, gl_workdir); - runner_argprintf (&runner, "--session-owner=%s", uuid_str); - runner_add_args (&runner, slave, "--config-get", "log-file", NULL); - - ret = glusterd_query_extutil (log_file, &runner); + runner_argprintf (&runner, "%s", conf_path); + runner_argprintf (&runner, ":%s", master); + runner_add_args (&runner, slave, "--config-get", NULL); + runner_argprintf (&runner, "%s-file", param); - out: - return ret; + return glusterd_query_extutil (prmfile, &runner); } static int -gsyncd_getpidfile (char *master, char *slave, char *pidfile) +gsyncd_getpidfile (char *master, char *slave, char *pidfile, char *conf_path) { int ret = -1; glusterd_conf_t *priv = NULL; + char *confpath = NULL; + char conf_buf[PATH_MAX] = ""; + struct stat stbuf = {0,}; + GF_ASSERT (THIS); GF_ASSERT (THIS->private); @@ -461,8 +686,22 @@ gsyncd_getpidfile (char *master, char *slave, char *pidfile) GF_VALIDATE_OR_GOTO ("gsync", master, out); GF_VALIDATE_OR_GOTO ("gsync", slave, out); + ret = lstat (conf_path, &stbuf); + if (!ret) { + gf_log ("", GF_LOG_DEBUG, "Using passed config template(%s).", + conf_path); + confpath = conf_path; + } else { + ret = snprintf (conf_buf, sizeof(conf_buf) - 1, + "%s/"GSYNC_CONF_TEMPLATE, priv->workdir); + conf_buf[ret] = '\0'; + confpath = conf_buf; + gf_log ("", GF_LOG_DEBUG, "Using default config template(%s).", + confpath); + } + ret = glusterd_gsync_get_param_file (pidfile, "pid", master, - slave, priv->workdir); + slave, confpath); if (ret == -1) { ret = -2; gf_log ("", GF_LOG_WARNING, "failed to create the pidfile string"); @@ -476,32 +715,6 @@ gsyncd_getpidfile (char *master, char *slave, char *pidfile) } static int -glusterd_gsyncd_getlogfile (char *master, char *slave, char *log_file) -{ - int ret = -1; - glusterd_conf_t *priv = NULL; - - GF_ASSERT (THIS); - GF_ASSERT (THIS->private); - - priv = THIS->private; - - GF_VALIDATE_OR_GOTO ("gsync", master, out); - GF_VALIDATE_OR_GOTO ("gsync", slave, out); - - ret = glusterd_gsync_get_param_file (log_file, "log", master, - slave, priv->workdir); - if (ret == -1) { - ret = -2; - gf_log ("", GF_LOG_WARNING, "failed to gsyncd logfile"); - goto out; - } - - out: - return ret; -} - -static int gsync_status_byfd (int fd) { GF_ASSERT (fd >= -1); @@ -518,23 +731,24 @@ gsync_status_byfd (int fd) * return -1 when not running */ int -gsync_status (char *master, char *slave, int *status) +gsync_status (char *master, char *slave, char *conf_path, int *status) { char pidfile[PATH_MAX] = {0,}; int fd = -1; - fd = gsyncd_getpidfile (master, slave, pidfile); + fd = gsyncd_getpidfile (master, slave, pidfile, conf_path); if (fd == -2) return -1; *status = gsync_status_byfd (fd); - close (fd); + sys_close (fd); + return 0; } -int32_t +static int32_t glusterd_gsync_volinfo_dict_set (glusterd_volinfo_t *volinfo, char *key, char *value) { @@ -558,17 +772,49 @@ out: return 0; } -int -gsync_verify_config_options (dict_t *dict, char **op_errstr) +static int +glusterd_verify_gsyncd_spawn (char *master, char *slave) +{ + int ret = 0; + runner_t runner = {0,}; + + runinit (&runner); + runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", + "--verify", "spawning", NULL); + runner_argprintf (&runner, ":%s", master); + runner_add_args (&runner, slave, NULL); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + ret = runner_start (&runner); + if (ret) { + gf_log ("", GF_LOG_ERROR, "spawning child failed"); + ret = -1; + goto out; + } + + if (runner_end (&runner) != 0) + ret = -1; + +out: + gf_log ("", GF_LOG_DEBUG, "returning %d", ret); + return ret; +} + +static int +gsync_verify_config_options (dict_t *dict, char **op_errstr, char *volname) { - char **resopt = NULL; - int i = 0; - char *subop = NULL; - char *slave = NULL; - char *op_name = NULL; - char *op_value = NULL; - char *t = NULL; - gf_boolean_t banned = _gf_true; + char **resopt = NULL; + int i = 0; + int ret = -1; + char *subop = NULL; + char *slave = NULL; + char *op_name = NULL; + char *op_value = NULL; + char *t = NULL; + char errmsg[PATH_MAX] = ""; + gf_boolean_t banned = _gf_true; + gf_boolean_t op_match = _gf_true; + gf_boolean_t val_match = _gf_true; + struct gsync_config_opt_vals_ *conf_vals = NULL; if (dict_get_str (dict, "subop", &subop) != 0) { gf_log ("", GF_LOG_WARNING, "missing subop"); @@ -592,6 +838,12 @@ gsync_verify_config_options (dict_t *dict, char **op_errstr) } if (runcmd (GSYNCD_PREFIX"/gsyncd", "--config-check", op_name, NULL)) { + ret = glusterd_verify_gsyncd_spawn (volname, slave); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to spawn gsyncd"); + return 0; + } + gf_log ("", GF_LOG_WARNING, "Invalid option %s", op_name); *op_errstr = gf_strdup ("Invalid option"); @@ -636,34 +888,113 @@ gsync_verify_config_options (dict_t *dict, char **op_errstr) } } + /* Check options in gsync_confopt_vals for invalid values */ + for (conf_vals = gsync_confopt_vals; conf_vals->op_name; conf_vals++) { + op_match = _gf_true; + for (i = 0; conf_vals->op_name[i] && op_name[i]; i++) { + if (conf_vals->op_name[i] == op_name[i] || + (conf_vals->op_name[i] == '_' && op_name[i] == '-')) + continue; + op_match = _gf_false; + } + + if (op_match) { + val_match = _gf_false; + for (i = 0; i < conf_vals->no_of_pos_vals; i++) { + if(conf_vals->case_sensitive){ + if (!strcmp (conf_vals->values[i], op_value)) + val_match = _gf_true; + } else { + if (!strcasecmp (conf_vals->values[i], op_value)) + val_match = _gf_true; + } + } + + if (!val_match) { + ret = snprintf (errmsg, sizeof(errmsg) - 1, + "Invalid values (%s) for" + " option %s", op_value, + op_name); + errmsg[ret] = '\0'; + + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + *op_errstr = gf_strdup (errmsg); + return -1; + } + } + } + return 0; } -static void +static int +glusterd_get_gsync_status_mst_slv (glusterd_volinfo_t *volinfo, + char *slave, char *conf_path, + dict_t *rsp_dict, char *node); + +static int _get_status_mst_slv (dict_t *this, char *key, data_t *value, void *data) { glusterd_gsync_status_temp_t *param = NULL; char *slave = NULL; - int ret = 0; + char *slave_buf = NULL; + char *slave_ip = NULL; + char *slave_vol = NULL; + char *errmsg = NULL; + char conf_path[PATH_MAX] = ""; + int ret = -1; + glusterd_conf_t *priv = NULL; param = (glusterd_gsync_status_temp_t *)data; GF_ASSERT (param); GF_ASSERT (param->volinfo); + if (THIS) + priv = THIS->private; + if (priv == NULL) { + gf_log ("", GF_LOG_ERROR, "priv of glusterd not present"); + goto out; + } + slave = strchr(value->data, ':'); - if (slave) - slave ++; - else - return; + if (!slave) + return 0; + slave++; + + ret = glusterd_get_slave_info (slave, &slave_ip, &slave_vol, &errmsg); + if (ret) { + if (errmsg) + gf_log ("", GF_LOG_ERROR, "Unable to fetch " + "slave details. Error: %s", errmsg); + else + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave details."); + ret = -1; + goto out; + } + + ret = snprintf (conf_path, sizeof(conf_path) - 1, + "%s/"GEOREP"/%s_%s_%s/gsyncd.conf", + priv->workdir, param->volinfo->volname, + slave_ip, slave_vol); + conf_path[ret] = '\0'; ret = glusterd_get_gsync_status_mst_slv(param->volinfo, - slave, param->rsp_dict); + slave, conf_path, + param->rsp_dict, + param->node); +out: + + if (slave_buf) + GF_FREE(slave_buf); + gf_log ("", GF_LOG_DEBUG, "Returning %d.", ret); + return ret; } -static void +static int _get_max_gsync_slave_num (dict_t *this, char *key, data_t *value, void *data) { int tmp_slvnum = 0; @@ -672,25 +1003,30 @@ _get_max_gsync_slave_num (dict_t *this, char *key, data_t *value, void *data) sscanf (key, "slave%d", &tmp_slvnum); if (tmp_slvnum > *slvnum) *slvnum = tmp_slvnum; + + return 0; } static int glusterd_remove_slave_in_info (glusterd_volinfo_t *volinfo, char *slave, char **op_errstr) { + int zero_slave_entries = _gf_true; int ret = 0; char *slavekey = NULL; GF_ASSERT (volinfo); GF_ASSERT (slave); - ret = glusterd_get_slave (volinfo, slave, &slavekey); - if (ret < 0) { - ret++; - goto out; - } - - dict_del (volinfo->gsync_slaves, slavekey); + do { + ret = glusterd_get_slave (volinfo, slave, &slavekey); + if (ret < 0 && zero_slave_entries) { + ret++; + goto out; + } + zero_slave_entries = _gf_false; + dict_del (volinfo->gsync_slaves, slavekey); + } while (ret >= 0); ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); @@ -743,8 +1079,9 @@ glusterd_gsync_get_uuid (char *slave, glusterd_volinfo_t *vol, return ret; } -static int +int glusterd_check_gsync_running_local (char *master, char *slave, + char *conf_path, gf_boolean_t *is_run) { int ret = -1; @@ -755,7 +1092,7 @@ glusterd_check_gsync_running_local (char *master, char *slave, GF_ASSERT (is_run); *is_run = _gf_false; - ret = gsync_status (master, slave, &ret_status); + ret = gsync_status (master, slave, conf_path, &ret_status); if (ret == 0 && ret_status == 0) { *is_run = _gf_true; } else if (ret == -1) { @@ -772,7 +1109,8 @@ glusterd_check_gsync_running_local (char *master, char *slave, static int glusterd_store_slave_in_info (glusterd_volinfo_t *volinfo, char *slave, - char *host_uuid, char **op_errstr) + char *host_uuid, char **op_errstr, + gf_boolean_t is_force) { int ret = 0; int maxslv = 0; @@ -795,7 +1133,8 @@ glusterd_store_slave_in_info (glusterd_volinfo_t *volinfo, char *slave, case -1: break; default: - GF_ASSERT (ret > 0); + if (!is_force) + GF_ASSERT (ret > 0); ret = dict_get_str (volinfo->gsync_slaves, slavekey, &slaveentry); GF_ASSERT (ret == 0); @@ -804,13 +1143,23 @@ glusterd_store_slave_in_info (glusterd_volinfo_t *volinfo, char *slave, * assert an uuid mismatch */ t = strtail (slaveentry, host_uuid); - GF_ASSERT (!t || *t != ':') + if (!is_force) + GF_ASSERT (!t || *t != ':'); + + if (is_force) { + gf_log ("", GF_LOG_DEBUG, GEOREP" has already been " + "invoked for the %s (master) and %s (slave)." + " Allowing without saving info again due to" + " force command.", volinfo->volname, slave); + ret = 0; + goto out; + } gf_log ("", GF_LOG_ERROR, GEOREP" has already been invoked for " "the %s (master) and %s (slave) " "from a different machine", volinfo->volname, slave); - *op_errstr = gf_strdup (GEOREP" already running in an an" + *op_errstr = gf_strdup (GEOREP" already running in " "another machine"); ret = -1; goto out; @@ -843,23 +1192,26 @@ glusterd_store_slave_in_info (glusterd_volinfo_t *volinfo, char *slave, return ret; } - static int glusterd_op_verify_gsync_start_options (glusterd_volinfo_t *volinfo, - char *slave, char **op_errstr) + char *slave, char *conf_path, + char *statefile, char **op_errstr, + gf_boolean_t is_force) { int ret = -1; gf_boolean_t is_running = _gf_false; char msg[2048] = {0}; uuid_t uuid = {0}; - glusterd_conf_t *priv = NULL; - xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + struct stat stbuf = {0,}; this = THIS; GF_ASSERT (volinfo); GF_ASSERT (slave); GF_ASSERT (op_errstr); + GF_ASSERT (conf_path); GF_ASSERT (this && this->private); priv = this->private; @@ -869,26 +1221,56 @@ glusterd_op_verify_gsync_start_options (glusterd_volinfo_t *volinfo, "before "GEOREP" start", volinfo->volname); goto out; } + + ret = lstat (statefile, &stbuf); + if (ret) { + snprintf (msg, sizeof (msg), "Session between %s and %s has" + " not been created. Please create session and retry.", + volinfo->volname, slave); + gf_log ("", GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + goto out; + } + + /* Check if the gsync slave info is stored. If not + * session has not been created */ + ret = glusterd_gsync_get_uuid (slave, volinfo, uuid); + if (ret) { + snprintf (msg, sizeof (msg), "Session between %s and %s has" + " not been created. Please create session and retry.", + volinfo->volname, slave); + gf_log ("", GF_LOG_ERROR, "%s", msg); + goto out; + } + + if (is_force) { + ret = 0; + goto out; + } + /*Check if the gsync is already started in cmd. inited host * If so initiate add it into the glusterd's priv*/ - ret = glusterd_gsync_get_uuid (slave, volinfo, uuid); - if ((ret == 0) && (uuid_compare (priv->uuid, uuid) == 0)) { - ret = glusterd_check_gsync_running_local (volinfo->volname, - slave, &is_running); - if (ret) { - snprintf (msg, sizeof (msg), GEOREP" start option " - "validation failed "); - goto out; - } - if (_gf_true == is_running) { - snprintf (msg, sizeof (msg), GEOREP " session between" - " %s & %s already started", volinfo->volname, - slave); - ret = -1; - goto out; - } + ret = glusterd_check_gsync_running_local (volinfo->volname, + slave, conf_path, + &is_running); + if (ret) { + snprintf (msg, sizeof (msg), GEOREP" start option " + "validation failed "); + goto out; + } + if (_gf_true == is_running) { + snprintf (msg, sizeof (msg), GEOREP " session between" + " %s & %s already started", volinfo->volname, + slave); + ret = -1; + goto out; + } + + ret = glusterd_verify_gsyncd_spawn (volinfo->volname, slave); + if (ret) { + snprintf (msg, sizeof (msg), "Unable to spawn gsyncd"); + gf_log ("", GF_LOG_ERROR, "%s", msg); } - ret = 0; out: if (ret && (msg[0] != '\0')) { *op_errstr = gf_strdup (msg); @@ -914,11 +1296,13 @@ glusterd_check_gsync_running (glusterd_volinfo_t *volinfo, gf_boolean_t *flag) static int glusterd_op_verify_gsync_running (glusterd_volinfo_t *volinfo, - char *slave, char **op_errstr) + char *slave, char *conf_path, + char **op_errstr) { - int ret = -1; - char msg[2048] = {0}; - uuid_t uuid = {0}; + int pfd = -1; + int ret = -1; + char msg[2048] = {0}; + char pidfile[PATH_MAX] = {0,}; GF_ASSERT (THIS && THIS->private); GF_ASSERT (volinfo); @@ -931,13 +1315,26 @@ glusterd_op_verify_gsync_running (glusterd_volinfo_t *volinfo, goto out; } - ret = glusterd_gsync_get_uuid (slave, volinfo, uuid); - if (ret == -1) { - snprintf (msg, sizeof (msg), GEOREP" session between %s & %s" - " not active", volinfo->volname, slave); + + pfd = gsyncd_getpidfile (volinfo->volname, slave, pidfile, conf_path); + if (pfd == -2) { + gf_log ("", GF_LOG_ERROR, GEOREP" stop validation " + "failed for %s & %s", volinfo->volname, slave); + ret = -1; + goto out; + } + if (gsync_status_byfd (pfd) == -1) { + snprintf (msg, sizeof (msg), GEOREP" session b/w %s & %s is not" + " running on this node.", volinfo->volname, slave); + gf_log ("", GF_LOG_ERROR, "%s", msg); + ret = -1; + /* monitor gsyncd already dead */ goto out; } + if (pfd < 0) + goto out; + ret = 0; out: if (ret && (msg[0] != '\0')) { @@ -956,6 +1353,18 @@ glusterd_verify_gsync_status_opts (dict_t *dict, char **op_errstr) gf_boolean_t exists = _gf_false; glusterd_volinfo_t *volinfo = NULL; int ret = 0; + char *conf_path = NULL; + char *slave_ip = NULL; + char *slave_vol = NULL; + glusterd_conf_t *priv = NULL; + + if (THIS) + priv = THIS->private; + if (priv == NULL) { + gf_log ("", GF_LOG_ERROR, "priv of glusterd not present"); + *op_errstr = gf_strdup ("glusterd defunct"); + goto out; + } ret = dict_get_str (dict, "master", &volname); if (ret < 0) { @@ -980,38 +1389,447 @@ glusterd_verify_gsync_status_opts (dict_t *dict, char **op_errstr) goto out; } - out: + ret = glusterd_get_slave_details_confpath (volinfo, dict, &slave_ip, + &slave_vol, &conf_path, + op_errstr); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave or confpath details."); + ret = -1; + goto out; + } + +out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; - } -static int +int glusterd_op_gsync_args_get (dict_t *dict, char **op_errstr, - char **master, char **slave) + char **master, char **slave, char **host_uuid) { int ret = -1; GF_ASSERT (dict); GF_ASSERT (op_errstr); - GF_ASSERT (master); - GF_ASSERT (slave); - ret = dict_get_str (dict, "master", master); - if (ret < 0) { - gf_log ("", GF_LOG_WARNING, "master not found"); - *op_errstr = gf_strdup ("master not found"); + if (master) { + ret = dict_get_str (dict, "master", master); + if (ret < 0) { + gf_log ("", GF_LOG_WARNING, "master not found"); + *op_errstr = gf_strdup ("master not found"); + goto out; + } + } + + if (slave) { + ret = dict_get_str (dict, "slave", slave); + if (ret < 0) { + gf_log ("", GF_LOG_WARNING, "slave not found"); + *op_errstr = gf_strdup ("slave not found"); + goto out; + } + } + + if (host_uuid) { + ret = dict_get_str (dict, "host-uuid", host_uuid); + if (ret < 0) { + gf_log ("", GF_LOG_WARNING, "host_uuid not found"); + *op_errstr = gf_strdup ("host_uuid not found"); + goto out; + } + } + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterd_op_stage_sys_exec (dict_t *dict, char **op_errstr) +{ + char errmsg[PATH_MAX] = ""; + char *command = NULL; + char command_path[PATH_MAX] = ""; + struct stat st = {0,}; + int ret = -1; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + GF_ASSERT (conf); + + if (conf->op_version < 2) { + gf_log ("", GF_LOG_ERROR, "Op Version not supported."); + snprintf (errmsg, sizeof(errmsg), "One or more nodes do not" + " support the required op version."); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + + ret = dict_get_str (dict, "command", &command); + if (ret) { + strcpy (errmsg, "internal error"); + gf_log ("", GF_LOG_ERROR, + "Unable to get command from dict"); + goto out; + } + + /* enforce local occurrence of the command */ + if (strchr (command, '/')) { + strcpy (errmsg, "invalid command name"); + ret = -1; + goto out; + } + + sprintf (command_path, GSYNCD_PREFIX"/peer_%s", command); + /* check if it's executable */ + ret = access (command_path, X_OK); + if (!ret) + /* check if it's a regular file */ + ret = stat (command_path, &st); + if (!ret && !S_ISREG (st.st_mode)) + ret = -1; + +out: + if (ret) { + if (errmsg[0] == '\0') + snprintf (errmsg, sizeof (errmsg), "%s not found.", + command); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + } + + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterd_op_stage_copy_file (dict_t *dict, char **op_errstr) +{ + char abs_filename[PATH_MAX] = ""; + char errmsg[PATH_MAX] = ""; + char *filename = NULL; + char *host_uuid = NULL; + char uuid_str [64] = {0}; + int ret = -1; + glusterd_conf_t *priv = NULL; + struct stat stbuf = {0,}; + + if (THIS) + priv = THIS->private; + if (priv == NULL) { + gf_log ("", GF_LOG_ERROR, "priv of glusterd not present"); + *op_errstr = gf_strdup ("glusterd defunct"); goto out; } - ret = dict_get_str (dict, "slave", slave); + if (priv->op_version < 2) { + gf_log ("", GF_LOG_ERROR, "Op Version not supported."); + snprintf (errmsg, sizeof(errmsg), "One or more nodes do not" + " support the required op version."); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + + ret = dict_get_str (dict, "host-uuid", &host_uuid); if (ret < 0) { - gf_log ("", GF_LOG_WARNING, "slave not found"); - *op_errstr = gf_strdup ("slave not found"); + gf_log ("", GF_LOG_ERROR, "Unable to fetch" + " host-uuid from dict."); + goto out; + } + + uuid_utoa_r (MY_UUID, uuid_str); + if (!strcmp (uuid_str, host_uuid)) { + ret = dict_get_str (dict, "source", &filename); + if (ret < 0) { + gf_log ("", GF_LOG_ERROR, "Unable to fetch" + " filename from dict."); + *op_errstr = gf_strdup ("command unsuccessful"); + goto out; + } + snprintf (abs_filename, sizeof(abs_filename), + "%s/%s", priv->workdir, filename); + + ret = lstat (abs_filename, &stbuf); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Source file" + " does not exist in %s", priv->workdir); + *op_errstr = gf_strdup (errmsg); + goto out; + } + + if (!S_ISREG(stbuf.st_mode)) { + snprintf (errmsg, sizeof (errmsg), "Source file" + " is not a regular file."); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } + } + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +static int +glusterd_get_statefile_name (glusterd_volinfo_t *volinfo, char *slave, + char *conf_path, char **statefile) +{ + glusterd_conf_t *priv = NULL; + int ret = -1; + char *master = NULL; + char *buf = NULL; + dict_t *confd = NULL; + char *confpath = NULL; + char conf_buf[PATH_MAX] = ""; + struct stat stbuf = {0,}; + + GF_ASSERT (THIS); + GF_ASSERT (THIS->private); + GF_ASSERT (volinfo); + + master = volinfo->volname; + + confd = dict_new (); + if (!confd) { + gf_log ("", GF_LOG_ERROR, "Unable to create new dict"); + goto out; + } + + priv = THIS->private; + + ret = lstat (conf_path, &stbuf); + if (!ret) { + gf_log ("", GF_LOG_INFO, "Using passed config template(%s).", + conf_path); + confpath = conf_path; + } else { + ret = snprintf (conf_buf, sizeof(conf_buf) - 1, + "%s/"GSYNC_CONF_TEMPLATE, priv->workdir); + conf_buf[ret] = '\0'; + confpath = conf_buf; + gf_log ("", GF_LOG_INFO, "Using default config template(%s).", + confpath); + } + + ret = glusterd_gsync_get_config (master, slave, confpath, + confd); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get configuration data" + "for %s(master), %s(slave)", master, slave); + goto out; + + } + + ret = dict_get_param (confd, "state_file", &buf); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get state_file's name."); + goto out; + } + + *statefile = gf_strdup(buf); + if (!*statefile) { + gf_log ("", GF_LOG_ERROR, "Unable to gf_strdup."); + ret = -1; + goto out; + } + + ret = 0; + out: + if (confd) + dict_destroy (confd); + + gf_log ("", GF_LOG_DEBUG, "Returning %d ", ret); + return ret; +} + +static int +glusterd_create_status_file (char *master, char *slave, char *slave_ip, + char *slave_vol, char *status) +{ + int ret = -1; + runner_t runner = {0,}; + glusterd_conf_t *priv = NULL; + + if (THIS) + priv = THIS->private; + if (priv == NULL) { + gf_log ("", GF_LOG_ERROR, "priv of glusterd not present"); + goto out; + } + + if (!status) { + gf_log ("", GF_LOG_ERROR, "Status Empty"); + goto out; + } + gf_log ("", GF_LOG_DEBUG, "slave = %s", slave); + + runinit (&runner); + runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "--create", + status, "-c", NULL); + runner_argprintf (&runner, "%s/"GEOREP"/%s_%s_%s/gsyncd.conf", + priv->workdir, master, slave_ip, slave_vol); + runner_argprintf (&runner, ":%s", master); + runner_add_args (&runner, slave, NULL); + synclock_unlock (&priv->big_lock); + ret = runner_run (&runner); + synclock_lock (&priv->big_lock); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Creating status file failed."); + ret = -1; + goto out; + } + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "returning %d", ret); + return ret; +} + +static int +glusterd_verify_slave (char *volname, char *slave_ip, char *slave, + char **op_errstr, gf_boolean_t *is_force_blocker) +{ + int32_t ret = -1; + runner_t runner = {0,}; + char log_file_path[PATH_MAX] = ""; + char buf[PATH_MAX] = ""; + char *tmp = NULL; + char *save_ptr = NULL; + glusterd_conf_t *priv = NULL; + + GF_ASSERT (volname); + GF_ASSERT (slave_ip); + GF_ASSERT (slave); + + if (THIS) + priv = THIS->private; + if (priv == NULL) { + gf_log ("", GF_LOG_ERROR, "priv of glusterd not present"); + goto out; + } + + snprintf (log_file_path, sizeof(log_file_path), + DEFAULT_LOG_FILE_DIRECTORY"/create_verify_log"); + + runinit (&runner); + runner_add_args (&runner, GSYNCD_PREFIX"/gverify.sh", NULL); + runner_argprintf (&runner, "%s", volname); + runner_argprintf (&runner, "%s", slave_ip); + runner_argprintf (&runner, "%s", slave); + runner_argprintf (&runner, "%s", log_file_path); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + synclock_unlock (&priv->big_lock); + ret = runner_run (&runner); + synclock_lock (&priv->big_lock); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Not a valid slave"); + ret = glusterd_gsync_read_frm_status (log_file_path, + buf, sizeof(buf)); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to read from %s", + log_file_path); + goto out; + } + + /* Tokenize the error message from gverify.sh to figure out + * if the error is a force blocker or not. */ + tmp = strtok_r (buf, "|", &save_ptr); + if (!strcmp (tmp, "FORCE_BLOCKER")) + *is_force_blocker = 1; + else { + /* No FORCE_BLOCKER flag present so all that is + * present is the error message. */ + *is_force_blocker = 0; + if (tmp) + *op_errstr = gf_strdup (tmp); + ret = -1; + goto out; + } + + /* Copy rest of the error message to op_errstr */ + tmp = strtok_r (NULL, "|", &save_ptr); + if (tmp) + *op_errstr = gf_strdup (tmp); + ret = -1; goto out; } + ret = 0; +out: + unlink (log_file_path); + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterd_mountbroker_check (char **slave_ip, char **op_errstr) +{ + int ret = -1; + char *tmp = NULL; + char *save_ptr = NULL; + char *username = NULL; + char *host = NULL; + char errmsg[PATH_MAX] = ""; + + GF_ASSERT (slave_ip); + GF_ASSERT (*slave_ip); + + /* Checking if hostname has user specified */ + host = strstr (*slave_ip, "@"); + if (!host) { + gf_log ("", GF_LOG_DEBUG, "No username provided."); + ret = 0; + goto out; + } else { + /* Moving the host past the '@' and checking if the + * actual hostname also has '@' */ + host++; + if (strstr (host, "@")) { + gf_log ("", GF_LOG_DEBUG, "host = %s", host); + ret = snprintf (errmsg, sizeof(errmsg) - 1, + "Invalid Hostname (%s).", host); + errmsg[ret] = '\0'; + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + if (op_errstr) + *op_errstr = gf_strdup (errmsg); + goto out; + } + /* Fetching the username and hostname + * and checking if the username is non-root */ + username = strtok_r (*slave_ip, "@", &save_ptr); + tmp = strtok_r (NULL, "@", &save_ptr); + if (strcmp (username, "root")) { + ret = snprintf (errmsg, sizeof(errmsg) - 1, + "Non-root username (%s@%s) not allowed.", + username, tmp); + errmsg[ret] = '\0'; + if (op_errstr) + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, + "Non-Root username not allowed."); + ret = -1; + goto out; + } + + *slave_ip = gf_strdup (tmp); + if (!*slave_ip) { + gf_log ("", GF_LOG_ERROR, "Out of memory"); + ret = -1; + goto out; + } + } ret = 0; out: @@ -1020,17 +1838,257 @@ out: } int +glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr) +{ + char *down_peerstr = NULL; + char *slave = NULL; + char *volname = NULL; + char *host_uuid = NULL; + char *statefile = NULL; + char *slave_ip = NULL; + char *slave_vol = NULL; + char *conf_path = NULL; + char errmsg[PATH_MAX] = ""; + char common_pem_file[PATH_MAX] = ""; + char hook_script[PATH_MAX] = ""; + char uuid_str [64] = ""; + int ret = -1; + int is_pem_push = -1; + gf_boolean_t is_force = -1; + gf_boolean_t is_force_blocker = -1; + gf_boolean_t exists = _gf_false; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + struct stat stbuf = {0,}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + GF_ASSERT (conf); + + ret = glusterd_op_gsync_args_get (dict, op_errstr, &volname, + &slave, &host_uuid); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to fetch arguments"); + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return -1; + } + + if (conf->op_version < 2) { + gf_log ("", GF_LOG_ERROR, "Op Version not supported."); + snprintf (errmsg, sizeof(errmsg), "One or more nodes do not" + " support the required op version."); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + + exists = glusterd_check_volume_exists (volname); + ret = glusterd_volinfo_find (volname, &volinfo); + if ((ret) || (!exists)) { + gf_log ("", GF_LOG_WARNING, "volume name does not exist"); + snprintf (errmsg, sizeof(errmsg), "Volume name %s does not" + " exist", volname); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return -1; + } + + ret = glusterd_get_slave_details_confpath (volinfo, dict, &slave_ip, + &slave_vol, &conf_path, + op_errstr); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave or confpath details."); + ret = -1; + goto out; + } + + is_force = dict_get_str_boolean (dict, "force", _gf_false); + + uuid_utoa_r (MY_UUID, uuid_str); + if (!strcmp (uuid_str, host_uuid)) { + ret = glusterd_are_vol_all_peers_up (volinfo, + &conf->peers, + &down_peerstr); + if ((ret == _gf_false) && !is_force) { + snprintf (errmsg, sizeof (errmsg), "Peer %s," + " which is a part of %s volume, is" + " down. Please bring up the peer and" + " retry.", down_peerstr, + volinfo->volname); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + *op_errstr = gf_strdup (errmsg); + GF_FREE (down_peerstr); + down_peerstr = NULL; + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return -1; + } else if (ret == _gf_false) { + gf_log ("", GF_LOG_INFO, "Peer %s," + " which is a part of %s volume, is" + " down. Force creating geo-rep session." + " On bringing up the peer, re-run" + " \"gluster system:: execute" + " gsec_create\" and \"gluster volume" + " geo-replication %s %s create push-pem" + " force\"", down_peerstr, volinfo->volname, + volinfo->volname, slave); + } + + /* Checking if slave host is pingable, has proper passwordless + * ssh login setup, slave volume is created, slave vol is empty, + * and if it has enough memory and bypass in case of force if + * the error is not a force blocker */ + ret = glusterd_verify_slave (volname, slave_ip, slave_vol, + op_errstr, &is_force_blocker); + if (ret) { + if (is_force && !is_force_blocker) { + gf_log ("", GF_LOG_INFO, "%s is not a valid slave" + " volume. Error: %s. Force creating geo-rep" + " session.", slave, *op_errstr); + } else { + gf_log ("", GF_LOG_ERROR, + "%s is not a valid slave volume. Error: %s", + slave, *op_errstr); + ret = -1; + goto out; + } + } + + ret = dict_get_int32 (dict, "push_pem", &is_pem_push); + if (!ret && is_pem_push) { + ret = snprintf (common_pem_file, + sizeof(common_pem_file) - 1, + "%s"GLUSTERD_COMMON_PEM_PUB_FILE, + conf->workdir); + common_pem_file[ret] = '\0'; + + ret = snprintf (hook_script, sizeof(hook_script) - 1, + "%s"GLUSTERD_CREATE_HOOK_SCRIPT, + conf->workdir); + hook_script[ret] = '\0'; + + ret = lstat (common_pem_file, &stbuf); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "%s" + " required for push-pem is" + " not present. Please run" + " \"gluster system:: execute" + " gsec_create\"", common_pem_file); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + + ret = lstat (hook_script, &stbuf); + if (ret) { + snprintf (errmsg, sizeof (errmsg), + "The hook-script (%s) required " + "for push-pem is not present. " + "Please install the hook-script " + "and retry", hook_script); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + + if (!S_ISREG(stbuf.st_mode)) { + snprintf (errmsg, sizeof (errmsg), "%s" + " required for push-pem is" + " not a regular file. Please run" + " \"gluster system:: execute" + " gsec_create\"", common_pem_file); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } + } + } + + ret = glusterd_get_statefile_name (volinfo, slave, conf_path, &statefile); + if (ret) { + if (!strstr(slave, "::")) + snprintf (errmsg, sizeof (errmsg), + "%s is not a valid slave url.", slave); + else + snprintf (errmsg, sizeof (errmsg), "Please check gsync " + "config file. Unable to get statefile's name"); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } + + ret = dict_set_str (dict, "statefile", statefile); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store statefile path"); + goto out; + } + + ret = lstat (statefile, &stbuf); + if (!ret && !is_force) { + snprintf (errmsg, sizeof (errmsg), "Session between %s" + " and %s is already created.", + volinfo->volname, slave); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } else if (!ret) + gf_log ("", GF_LOG_INFO, "Session between %s" + " and %s is already created. Force" + " creating again.", volinfo->volname, slave); + + ret = glusterd_verify_gsyncd_spawn (volinfo->volname, slave); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Unable to spawn gsyncd."); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } + + ret = 0; +out: + + if (ret && errmsg[0] != '\0') + *op_errstr = gf_strdup (errmsg); + + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr) { int ret = 0; int type = 0; char *volname = NULL; char *slave = NULL; + char *slave_ip = NULL; + char *slave_vol = NULL; + char *down_peerstr = NULL; + char *statefile = NULL; + char *path_list = NULL; + char *conf_path = NULL; gf_boolean_t exists = _gf_false; glusterd_volinfo_t *volinfo = NULL; char errmsg[PATH_MAX] = {0,}; dict_t *ctx = NULL; + gf_boolean_t is_force = 0; + gf_boolean_t is_force_blocker = -1; + gf_boolean_t is_running = _gf_false; + uuid_t uuid = {0}; + char uuid_str [64] = {0}; + char *host_uuid = NULL; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + struct stat stbuf = {0,}; + this = THIS; + GF_ASSERT (this); + conf = this->private; + GF_ASSERT (conf); ret = dict_get_int32 (dict, "type", &type); if (ret < 0) { @@ -1039,25 +2097,26 @@ glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr) goto out; } - switch (type) { - case GF_GSYNC_OPTION_TYPE_STATUS: + if (type == GF_GSYNC_OPTION_TYPE_STATUS) { ret = glusterd_verify_gsync_status_opts (dict, op_errstr); - goto out; - case GF_GSYNC_OPTION_TYPE_CONFIG: - ret = gsync_verify_config_options (dict, op_errstr); + } + ret = glusterd_op_gsync_args_get (dict, op_errstr, + &volname, &slave, &host_uuid); + if (ret) goto out; - case GF_GSYNC_OPTION_TYPE_ROTATE: - /* checks same as status mode */ - ret = glusterd_verify_gsync_status_opts(dict, op_errstr); - goto out; - } + uuid_utoa_r (MY_UUID, uuid_str); - ret = glusterd_op_gsync_args_get (dict, op_errstr, &volname, &slave); - if (ret) + if (conf->op_version < 2) { + gf_log ("", GF_LOG_ERROR, "Op Version not supported."); + snprintf (errmsg, sizeof(errmsg), "One or more nodes do not" + " support the required op version."); + *op_errstr = gf_strdup (errmsg); + ret = -1; goto out; + } exists = glusterd_check_volume_exists (volname); ret = glusterd_volinfo_find (volname, &volinfo); @@ -1070,19 +2129,115 @@ glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr) goto out; } + ret = glusterd_get_slave_details_confpath (volinfo, dict, &slave_ip, + &slave_vol, &conf_path, + op_errstr); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave or confpath details."); + ret = -1; + goto out; + } + + ret = glusterd_get_statefile_name (volinfo, slave, conf_path, &statefile); + if (ret) { + /* Checking if slave host is pingable, has proper passwordless + * ssh login setup */ + ret = glusterd_verify_slave (volname, slave_ip, slave_vol, + op_errstr, &is_force_blocker); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "%s is not a valid slave volume. Error: %s", + slave, *op_errstr); + goto out; + } + + if (!strstr(slave, "::")) + snprintf (errmsg, sizeof (errmsg), + "%s is not a valid slave url.", slave); + else + snprintf (errmsg, sizeof (errmsg), + "Unable to get statefile's name"); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } + + ret = dict_set_str (dict, "statefile", statefile); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store statefile path"); + goto out; + } + + is_force = dict_get_str_boolean (dict, "force", _gf_false); + + /* Allowing stop force to bypass the statefile check + * as this command acts as a fail safe method to stop geo-rep + * session. */ + if ((type == GF_GSYNC_OPTION_TYPE_CONFIG) || + ((type == GF_GSYNC_OPTION_TYPE_STOP) && !is_force) || + (type == GF_GSYNC_OPTION_TYPE_DELETE)) { + ret = lstat (statefile, &stbuf); + if (ret) { + snprintf (errmsg, sizeof(errmsg), "Geo-replication" + " session between %s and %s does not exist.", + volinfo->volname, slave); + gf_log ("", GF_LOG_ERROR, "%s. statefile = %s", + errmsg, statefile); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + } + + /* Check if all peers that are a part of the volume are up or not */ + if ((type == GF_GSYNC_OPTION_TYPE_DELETE) || + ((type == GF_GSYNC_OPTION_TYPE_STOP) && !is_force)) { + if (!strcmp (uuid_str, host_uuid)) { + ret = glusterd_are_vol_all_peers_up (volinfo, + &conf->peers, + &down_peerstr); + if (ret == _gf_false) { + snprintf (errmsg, sizeof (errmsg), "Peer %s," + " which is a part of %s volume, is" + " down. Please bring up the peer and" + " retry.", down_peerstr, + volinfo->volname); + *op_errstr = gf_strdup (errmsg); + ret = -1; + GF_FREE (down_peerstr); + down_peerstr = NULL; + goto out; + } + } + } + switch (type) { case GF_GSYNC_OPTION_TYPE_START: + /* don't attempt to start gsync if replace-brick is + * in progress */ + if (glusterd_is_rb_ongoing (volinfo)) { + snprintf (errmsg, sizeof(errmsg), "replace-brick is in" + " progress, not starting geo-replication"); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + ret = glusterd_op_verify_gsync_start_options (volinfo, slave, - op_errstr); + conf_path, statefile, + op_errstr, is_force); if (ret) goto out; ctx = glusterd_op_get_ctx(); if (ctx) { - /*gsyncd does a fuse mount to start the geo-rep session*/ + /* gsyncd does a fuse mount to start + * the geo-rep session */ if (!glusterd_is_fuse_available ()) { - gf_log ("glusterd", GF_LOG_ERROR, "Unable to open" - " /dev/fuse (%s), geo-replication start" - " failed", strerror (errno)); + gf_log ("glusterd", GF_LOG_ERROR, "Unable to " + "open /dev/fuse (%s), geo-replication " + "start failed", strerror (errno)); snprintf (errmsg, sizeof(errmsg), "fuse unvailable"); *op_errstr = gf_strdup (errmsg); @@ -1093,17 +2248,72 @@ glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr) break; case GF_GSYNC_OPTION_TYPE_STOP: - ret = glusterd_op_verify_gsync_running (volinfo, slave, - op_errstr); + if (!is_force) { + ret = glusterd_op_verify_gsync_running (volinfo, slave, + conf_path, + op_errstr); + if (ret) { + ret = glusterd_get_local_brickpaths (volinfo, + &path_list); + if (path_list) + ret = -1; + } + } + break; + + case GF_GSYNC_OPTION_TYPE_CONFIG: + ret = gsync_verify_config_options (dict, op_errstr, volname); + goto out; + break; + + case GF_GSYNC_OPTION_TYPE_DELETE: + /* Check if the gsync session is still running + * If so ask the user to stop geo-replication first.*/ + ret = glusterd_gsync_get_uuid (slave, volinfo, uuid); + if (ret) { + snprintf (errmsg, sizeof(errmsg), "Geo-replication" + " session between %s and %s does not exist.", + volinfo->volname, slave); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } else { + ret = glusterd_check_gsync_running_local (volinfo->volname, + slave, conf_path, + &is_running); + if (_gf_true == is_running) { + snprintf (errmsg, sizeof (errmsg), GEOREP + " session between %s & %s is " + "still active. Please stop the " + "session and retry.", + volinfo->volname, slave); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + } + + ret = glusterd_verify_gsyncd_spawn (volinfo->volname, slave); + if (ret) { + snprintf (errmsg, sizeof (errmsg), + "Unable to spawn gsyncd"); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + } + break; } out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } -int -stop_gsync (char *master, char *slave, char **msg) +static int +stop_gsync (char *master, char *slave, char **msg, + char *conf_path, gf_boolean_t is_force) { int32_t ret = 0; int pfd = -1; @@ -1115,23 +2325,23 @@ stop_gsync (char *master, char *slave, char **msg) GF_ASSERT (THIS); GF_ASSERT (THIS->private); - pfd = gsyncd_getpidfile (master, slave, pidfile); - if (pfd == -2) { + pfd = gsyncd_getpidfile (master, slave, pidfile, conf_path); + if (pfd == -2 && !is_force) { gf_log ("", GF_LOG_ERROR, GEOREP" stop validation " " failed for %s & %s", master, slave); ret = -1; goto out; } - if (gsync_status_byfd (pfd) == -1) { + if (gsync_status_byfd (pfd) == -1 && !is_force) { gf_log ("", GF_LOG_ERROR, "gsyncd b/w %s & %s is not" " running", master, slave); - if (msg) - *msg = gf_strdup ("Warning: "GEOREP" session was in " - "corrupt state"); /* monitor gsyncd already dead */ goto out; } + if (pfd < 0) + goto out; + ret = read (pfd, buf, 1024); if (ret > 0) { pid = strtol (buf, NULL, 10); @@ -1158,17 +2368,17 @@ stop_gsync (char *master, char *slave, char **msg) ret = 0; out: - close (pfd); + sys_close (pfd); + + if (is_force) + ret = 0; return ret; } -int -glusterd_check_restart_gsync_session (glusterd_volinfo_t *volinfo, char *slave, - dict_t *resp_dict); - -int +static int glusterd_gsync_configure (glusterd_volinfo_t *volinfo, char *slave, - dict_t *dict, dict_t *resp_dict, char **op_errstr) + char *path_list, dict_t *dict, + dict_t *resp_dict, char **op_errstr) { int32_t ret = -1; char *op_name = NULL; @@ -1177,6 +2387,10 @@ glusterd_gsync_configure (glusterd_volinfo_t *volinfo, char *slave, glusterd_conf_t *priv = NULL; char *subop = NULL; char *master = NULL; + char *conf_path = NULL; + char *slave_ip = NULL; + char *slave_vol = NULL; + struct stat stbuf = {0, }; GF_ASSERT (slave); GF_ASSERT (op_errstr); @@ -1211,10 +2425,17 @@ glusterd_gsync_configure (glusterd_volinfo_t *volinfo, char *slave, goto out; } + ret = dict_get_str (dict, "conf_path", &conf_path); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch conf file path."); + goto out; + } + master = ""; runinit (&runner); runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL); - runner_argprintf (&runner, "%s/"GSYNC_CONF, priv->workdir); + runner_argprintf (&runner, "%s", conf_path); if (volinfo) { master = volinfo->volname; runner_argprintf (&runner, ":%s", master); @@ -1224,7 +2445,9 @@ glusterd_gsync_configure (glusterd_volinfo_t *volinfo, char *slave, runner_add_arg (&runner, op_name); if (op_value) runner_add_arg (&runner, op_value); + synclock_unlock (&priv->big_lock); ret = runner_run (&runner); + synclock_lock (&priv->big_lock); if (ret) { gf_log ("", GF_LOG_WARNING, "gsyncd failed to " "%s %s option for %s %s peers", @@ -1235,13 +2458,46 @@ glusterd_gsync_configure (glusterd_volinfo_t *volinfo, char *slave, goto out; } + + if (!strcmp (op_name, "state_file")) { + + ret = lstat (op_value, &stbuf); + if (ret) { + ret = dict_get_str (dict, "slave_ip", &slave_ip); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave IP."); + goto out; + } + + ret = dict_get_str (dict, "slave_vol", &slave_vol); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave volume name."); + goto out; + } + + ret = glusterd_create_status_file (volinfo->volname, slave, + slave_ip, slave_vol, + "Switching Status File"); + if (ret || lstat (op_value, &stbuf)) { + gf_log ("", GF_LOG_ERROR, "Unable to create %s" + ". Error : %s", op_value, + strerror (errno)); + ret = -1; + goto out; + } + } + } + ret = 0; gf_asprintf (op_errstr, "config-%s successful", subop); out: if (!ret && volinfo) { ret = glusterd_check_restart_gsync_session (volinfo, slave, - resp_dict); + resp_dict, path_list, + conf_path, 0); if (ret) *op_errstr = gf_strdup ("internal error"); } @@ -1250,74 +2506,233 @@ out: return ret; } -int -glusterd_gsync_read_frm_status (char *path, char *data) +static int +glusterd_gsync_read_frm_status (char *path, char *buf, size_t blen) { int ret = 0; - FILE *status_file = NULL; + int status_fd = -1; GF_ASSERT (path); - GF_ASSERT (data); - status_file = fopen (path, "r"); - if (status_file == NULL) { - gf_log ("", GF_LOG_WARNING, "Unable to read gsyncd status" + GF_ASSERT (buf); + status_fd = open (path, O_RDONLY); + if (status_fd == -1) { + gf_log ("", GF_LOG_ERROR, "Unable to read gsyncd status" " file"); return -1; } - ret = fread (data, PATH_MAX, 1, status_file); - if (ret < 0) { - gf_log ("", GF_LOG_WARNING, "Status file of gsyncd is corrupt"); + ret = read (status_fd, buf, blen - 1); + if (ret > 0) { + size_t len = strnlen (buf, ret); + /* Ensure there is a NUL byte and that it's not the first. */ + if (len == 0 || len == blen - 1) { + ret = -1; + } else { + char *p = buf + len - 1; + while (isspace (*p)) + *p-- = '\0'; + ret = 0; + } + } else if (ret < 0) + gf_log ("", GF_LOG_ERROR, "Status file of gsyncd is corrupt"); + + close (status_fd); + return ret; +} + +static int +glusterd_gsync_fetch_status_extra (char *path, char *buf, size_t blen) +{ + char sockpath[PATH_MAX] = {0,}; + struct sockaddr_un sa = {0,}; + size_t l = 0; + int s = -1; + struct pollfd pfd = {0,}; + int ret = 0; + + l = strlen (buf); + /* seek to end of data in buf */ + buf += l; + blen -= l; + + glusterd_set_socket_filepath (path, sockpath, sizeof (sockpath)); + + strncpy(sa.sun_path, sockpath, sizeof(sa.sun_path)); + if (sa.sun_path[sizeof (sa.sun_path) - 1]) return -1; + sa.sun_family = AF_UNIX; + + s = socket(AF_UNIX, SOCK_STREAM, 0); + if (s == -1) + return -1; + ret = fcntl (s, F_GETFL); + if (ret != -1) + ret = fcntl (s, F_SETFL, ret | O_NONBLOCK); + if (ret == -1) + goto out; + + ret = connect (s, (struct sockaddr *)&sa, sizeof (sa)); + if (ret == -1) + goto out; + pfd.fd = s; + pfd.events = POLLIN; + /* we don't want to hang on gsyncd */ + if (poll (&pfd, 1, 5000) < 1 || + !(pfd.revents & POLLIN)) { + ret = -1; + goto out; } + ret = read(s, buf, blen); + /* we expect a terminating 0 byte */ + if (ret == 0 || (ret > 0 && buf[ret - 1])) + ret = -1; + if (ret > 0) + ret = 0; + + out: + close (s); + return ret; +} - data[strlen(data)-1] = '\0'; +static int +dict_get_param (dict_t *dict, char *key, char **param) +{ + char *dk = NULL; + char *s = NULL; + char x = '\0'; + int ret = 0; - return 0; + if (dict_get_str (dict, key, param) == 0) + return 0; + + dk = gf_strdup (key); + if (!key) + return -1; + + s = strpbrk (dk, "-_"); + if (!s) + return -1; + x = (*s == '-') ? '_' : '-'; + *s++ = x; + while ((s = strpbrk (s, "-_"))) + *s++ = x; + + ret = dict_get_str (dict, dk, param); + + GF_FREE (dk); + return ret; } -int -glusterd_read_status_file (char *master, char *slave, - dict_t *dict) +static int +glusterd_read_status_file (glusterd_volinfo_t *volinfo, char *slave, + char *conf_path, dict_t *dict, char *node) { - glusterd_conf_t *priv = NULL; + glusterd_conf_t *priv = NULL; int ret = 0; - char statusfile[PATH_MAX] = {0, }; - char buff[PATH_MAX] = {0, }; - char mst[PATH_MAX] = {0, }; - char slv[PATH_MAX] = {0, }; - char sts[PATH_MAX] = {0, }; + char *statefile = NULL; + char *master = NULL; + char buf[1024] = "defunct"; + char nds[1024] = {0, }; + char mst[1024] = {0, }; + char slv[1024] = {0, }; + char sts[1024] = {0, }; + char *bufp = NULL; + dict_t *confd = NULL; int gsync_count = 0; int status = 0; + char *dyn_node = NULL; + char *path_list = NULL; GF_ASSERT (THIS); GF_ASSERT (THIS->private); + GF_ASSERT (volinfo); + + master = volinfo->volname; + + confd = dict_new (); + if (!dict) { + gf_log ("", GF_LOG_ERROR, "Not able to create dict."); + return -1; + } priv = THIS->private; - ret = glusterd_gsync_get_param_file (statusfile, "state", master, - slave, priv->workdir); + + ret = glusterd_gsync_get_config (master, slave, conf_path, + confd); if (ret) { - gf_log ("", GF_LOG_WARNING, "Unable to get the name of status" - "file for %s(master), %s(slave)", master, slave); - goto out; + gf_log ("", GF_LOG_ERROR, "Unable to get configuration data" + "for %s(master), %s(slave)", master, slave); + goto done; } - ret = gsync_status (master, slave, &status); - if (ret == 0 && status == -1) { - strncpy (buff, "corrupt", sizeof (buff)); + ret = dict_get_param (confd, "state_file", &statefile); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get state_file's name " + "for %s(master), %s(slave). Please check gsync " + "config file.", master, slave); goto done; - } else if (ret == -1) - goto out; - - ret = glusterd_gsync_read_frm_status (statusfile, buff); + } + ret = glusterd_gsync_read_frm_status (statefile, buf, sizeof (buf)); if (ret) { - gf_log ("", GF_LOG_WARNING, "Unable to read the status" + gf_log ("", GF_LOG_ERROR, "Unable to read the status" "file for %s(master), %s(slave)", master, slave); - goto out; + strncpy (buf, "defunct", sizeof (buf)); + goto done; + } + ret = gsync_status (master, slave, conf_path, &status); + if (ret == 0 && status == -1) { + if ((strcmp (buf, "Not Started")) && + (strcmp (buf, "Stopped"))) + strncpy (buf, "defunct", sizeof (buf)); + goto done; + } else if (ret == -1) { + gf_log ("", GF_LOG_ERROR, "Unable to get gsync status"); + goto done; + } + + if (strcmp (buf, "Stable") != 0) + goto done; + + ret = dict_get_param (confd, "state_socket_unencoded", &statefile); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get state_socket_unencoded" + " filepath. Please check gsync config file."); + goto done; + } + ret = glusterd_gsync_fetch_status_extra (statefile, buf, sizeof (buf)); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to fetch extra status" + "for %s(master), %s(slave)", master, slave); + /* there is a slight chance that this occurs due to race + * -- in that case, the following options all seem bad: + * + * - suppress irregurlar behavior by just leaving status + * on "OK" + * - freak out users with a misleading "defunct" + * - overload the meaning of the regular error signal + * mechanism of gsyncd, that is, when status is "faulty" + * + * -- so we just come up with something new... + */ + strncpy (buf, "N/A", sizeof (buf)); + goto done; } done: + if ((!strcmp (buf, "defunct")) || + (!strcmp (buf, "Not Started")) || + (!strcmp (buf, "Stopped"))) { + ret = glusterd_get_local_brickpaths (volinfo, &path_list); + if (!path_list) { + gf_log ("", GF_LOG_DEBUG, "This node not being part of" + " volume should not be running gsyncd. Hence" + " shouldn't display status for this node."); + ret = 0; + goto out; + } + } + ret = dict_get_int32 (dict, "gsync-count", &gsync_count); if (ret) @@ -1325,39 +2740,65 @@ glusterd_read_status_file (char *master, char *slave, else gsync_count++; + (void) snprintf (nds, sizeof (nds), "node%d", gsync_count); + dyn_node = gf_strdup (node); + if (!dyn_node) + goto out; + ret = dict_set_dynstr (dict, nds, dyn_node); + if (ret) { + GF_FREE (dyn_node); + goto out; + } + snprintf (mst, sizeof (mst), "master%d", gsync_count); - ret = dict_set_dynstr (dict, mst, gf_strdup (master)); - if (ret) + master = gf_strdup (master); + if (!master) goto out; + ret = dict_set_dynstr (dict, mst, master); + if (ret) { + GF_FREE (master); + goto out; + } snprintf (slv, sizeof (slv), "slave%d", gsync_count); - ret = dict_set_dynstr (dict, slv, gf_strdup (slave)); - if (ret) + slave = gf_strdup (slave); + if (!slave) + goto out; + ret = dict_set_dynstr (dict, slv, slave); + if (ret) { + GF_FREE (slave); goto out; + } snprintf (sts, sizeof (slv), "status%d", gsync_count); - ret = dict_set_dynstr (dict, sts, gf_strdup (buff)); - if (ret) + bufp = gf_strdup (buf); + if (!bufp) goto out; + ret = dict_set_dynstr (dict, sts, bufp); + if (ret) { + GF_FREE (bufp); + goto out; + } ret = dict_set_int32 (dict, "gsync-count", gsync_count); if (ret) goto out; - ret = 0; out: - gf_log ("", GF_LOG_DEBUG, "Returning %d ", ret); - return ret; + dict_destroy (confd); + + return 0; } int glusterd_check_restart_gsync_session (glusterd_volinfo_t *volinfo, char *slave, - dict_t *resp_dict) + dict_t *resp_dict, char *path_list, + char *conf_path, gf_boolean_t is_force) { int ret = 0; - uuid_t uuid = {0, }; glusterd_conf_t *priv = NULL; char *status_msg = NULL; + gf_boolean_t is_running = _gf_false; GF_ASSERT (volinfo); GF_ASSERT (slave); @@ -1366,26 +2807,30 @@ glusterd_check_restart_gsync_session (glusterd_volinfo_t *volinfo, char *slave, priv = THIS->private; - if (glusterd_gsync_get_uuid (slave, volinfo, uuid)) - /* session does not exist, nothing to do */ + ret = glusterd_check_gsync_running_local (volinfo->volname, + slave, conf_path, + &is_running); + if (!ret && (_gf_true != is_running)) + /* gsynd not running, nothing to do */ goto out; - if (uuid_compare (priv->uuid, uuid) == 0) { - ret = stop_gsync (volinfo->volname, slave, &status_msg); - if (ret == 0 && status_msg) - ret = dict_set_str (resp_dict, "gsync-status", - status_msg); - if (ret == 0) - ret = glusterd_start_gsync (volinfo, slave, - uuid_utoa(priv->uuid), NULL); - } + + ret = stop_gsync (volinfo->volname, slave, &status_msg, + conf_path, is_force); + if (ret == 0 && status_msg) + ret = dict_set_str (resp_dict, "gsync-status", + status_msg); + if (ret == 0) + ret = glusterd_start_gsync (volinfo, slave, path_list, + conf_path, uuid_utoa(MY_UUID), + NULL); out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } -int32_t -glusterd_marker_create_volfile (glusterd_volinfo_t *volinfo) +static int32_t +glusterd_marker_changelog_create_volfile (glusterd_volinfo_t *volinfo) { int32_t ret = 0; @@ -1408,58 +2853,83 @@ out: return ret; } -int -glusterd_set_marker_gsync (glusterd_volinfo_t *volinfo) +static int +glusterd_set_gsync_knob (glusterd_volinfo_t *volinfo, char *key, int *vc) { - int ret = -1; - int marker_set = _gf_false; - char *gsync_status = NULL; + int ret = -1; + int conf_enabled = _gf_false; + char *knob_on = NULL; GF_ASSERT (THIS); GF_ASSERT (THIS->private); - marker_set = glusterd_volinfo_get_boolean (volinfo, VKEY_MARKER_XTIME); - if (marker_set == -1) { - gf_log ("", GF_LOG_ERROR, "failed to get the marker status"); - ret = -1; + conf_enabled = glusterd_volinfo_get_boolean (volinfo, key); + if (conf_enabled == -1) { + gf_log ("", GF_LOG_ERROR, + "failed to get key %s from volinfo", key); goto out; } - if (marker_set == _gf_false) { - gsync_status = gf_strdup ("on"); - if (gsync_status == NULL) { + ret = 0; + if (conf_enabled == _gf_false) { + *vc = 1; + knob_on = gf_strdup ("on"); + if (knob_on == NULL) { ret = -1; goto out; } ret = glusterd_gsync_volinfo_dict_set (volinfo, - VKEY_MARKER_XTIME, gsync_status); - if (ret < 0) - goto out; - - ret = glusterd_marker_create_volfile (volinfo); - if (ret) { - gf_log ("", GF_LOG_ERROR, "Setting dict failed"); - goto out; - } + key, knob_on); } - ret = 0; -out: + out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } +static int +glusterd_set_gsync_confs (glusterd_volinfo_t *volinfo) +{ + int ret = -1; + int volfile_changed = 0; + ret = glusterd_set_gsync_knob (volinfo, + VKEY_MARKER_XTIME, &volfile_changed); + if (ret) + goto out; + /** + * enable ignore-pid-check blindly as it could be needed for + * cascading setups. + */ + ret = glusterd_set_gsync_knob (volinfo, VKEY_MARKER_XTIME_FORCE, + &volfile_changed); + if (ret) + goto out; -int -glusterd_get_gsync_status_mst_slv( glusterd_volinfo_t *volinfo, - char *slave, dict_t *rsp_dict) + ret = glusterd_set_gsync_knob (volinfo, + VKEY_CHANGELOG, &volfile_changed); + if (ret) + goto out; + + if (volfile_changed) + ret = glusterd_marker_changelog_create_volfile (volinfo); + + out: + return ret; +} + +static int +glusterd_get_gsync_status_mst_slv (glusterd_volinfo_t *volinfo, + char *slave, char *conf_path, + dict_t *rsp_dict, char *node) { + char *statefile = NULL; uuid_t uuid = {0, }; glusterd_conf_t *priv = NULL; int ret = 0; + struct stat stbuf = {0, }; GF_ASSERT (volinfo); GF_ASSERT (slave); @@ -1469,24 +2939,45 @@ glusterd_get_gsync_status_mst_slv( glusterd_volinfo_t *volinfo, priv = THIS->private; ret = glusterd_gsync_get_uuid (slave, volinfo, uuid); - if ((ret == 0) && (uuid_compare (priv->uuid, uuid) != 0)) - goto out; - if (ret) { - ret = 0; gf_log ("", GF_LOG_INFO, "geo-replication status %s %s :" "session is not active", volinfo->volname, slave); - goto out; + + ret = glusterd_get_statefile_name (volinfo, slave, + conf_path, &statefile); + if (ret) { + if (!strstr(slave, "::")) + gf_log ("", GF_LOG_INFO, + "%s is not a valid slave url.", slave); + else + gf_log ("", GF_LOG_INFO, "Unable to get" + " statefile's name"); + ret = 0; + goto out; + } + + ret = lstat (statefile, &stbuf); + if (ret) { + gf_log ("", GF_LOG_INFO, "%s statefile not present.", + statefile); + ret = 0; + goto out; + } } - ret = glusterd_read_status_file (volinfo->volname, slave, rsp_dict); - out: + ret = glusterd_read_status_file (volinfo, slave, conf_path, + rsp_dict, node); +out: + if (statefile) + GF_FREE (statefile); + gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); return ret; } static int -glusterd_get_gsync_status_mst (glusterd_volinfo_t *volinfo, dict_t *rsp_dict) +glusterd_get_gsync_status_mst (glusterd_volinfo_t *volinfo, dict_t *rsp_dict, + char *node) { glusterd_gsync_status_temp_t param = {0, }; @@ -1494,13 +2985,14 @@ glusterd_get_gsync_status_mst (glusterd_volinfo_t *volinfo, dict_t *rsp_dict) param.rsp_dict = rsp_dict; param.volinfo = volinfo; + param.node = node; dict_foreach (volinfo->gsync_slaves, _get_status_mst_slv, ¶m); return 0; } static int -glusterd_get_gsync_status_all ( dict_t *rsp_dict) +glusterd_get_gsync_status_all (dict_t *rsp_dict, char *node) { int32_t ret = 0; @@ -1513,7 +3005,7 @@ glusterd_get_gsync_status_all ( dict_t *rsp_dict) GF_ASSERT (priv); list_for_each_entry (volinfo, &priv->volumes, vol_list) { - ret = glusterd_get_gsync_status_mst (volinfo, rsp_dict); + ret = glusterd_get_gsync_status_mst (volinfo, rsp_dict, node); if (ret) goto out; } @@ -1529,15 +3021,22 @@ glusterd_get_gsync_status (dict_t *dict, char **op_errstr, dict_t *rsp_dict) { char *slave = NULL; char *volname = NULL; + char *conf_path = NULL; char errmsg[PATH_MAX] = {0, }; gf_boolean_t exists = _gf_false; glusterd_volinfo_t *volinfo = NULL; int ret = 0; + char my_hostname[256] = {0,}; + ret = gethostname(my_hostname, 256); + if (ret) { + /* stick to N/A */ + (void) strcpy (my_hostname, "N/A"); + } ret = dict_get_str (dict, "master", &volname); if (ret < 0){ - ret = glusterd_get_gsync_status_all (rsp_dict); + ret = glusterd_get_gsync_status_all (rsp_dict, my_hostname); goto out; } @@ -1555,297 +3054,427 @@ glusterd_get_gsync_status (dict_t *dict, char **op_errstr, dict_t *rsp_dict) ret = dict_get_str (dict, "slave", &slave); if (ret < 0) { - ret = glusterd_get_gsync_status_mst (volinfo, rsp_dict); + ret = glusterd_get_gsync_status_mst (volinfo, + rsp_dict, my_hostname); goto out; } - ret = glusterd_get_gsync_status_mst_slv (volinfo, slave, rsp_dict); + ret = dict_get_str (dict, "conf_path", &conf_path); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch conf file path."); + goto out; + } + + ret = glusterd_get_gsync_status_mst_slv (volinfo, slave, conf_path, + rsp_dict, my_hostname); out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } -int -glusterd_send_sigstop (pid_t pid) -{ - int ret = 0; - ret = kill (pid, SIGSTOP); - if (ret) - gf_log ("", GF_LOG_ERROR, GEOREP"failed to send SIGSTOP signal"); - return ret; -} - -int -glusterd_send_sigcont (pid_t pid) -{ - int ret = 0; - ret = kill (pid, SIGCONT); - if (ret) - gf_log ("", GF_LOG_ERROR, GEOREP"failed to send SIGCONT signal"); - return ret; -} - -/* - * Log rotations flow is something like this: - * - Send SIGSTOP to process group (this will stop monitor/worker process - * and also the slave if it's local) - * - Rotate log file for monitor/worker - * - Rotate log file for slave if it's local - * - Send SIGCONT to the process group. Monitor wakes up, kills the worker - * (this is done in the SIGCONT handler), which results in the termination - * of the slave (local/remote). After returning from signal handler, - * monitor detects absence of worker and starts it again, which in-turn - * starts the slave. - */ -int -glusterd_send_log_rotate_signal (pid_t pid, char *logfile1, char *logfile2) +static int +glusterd_gsync_delete (glusterd_volinfo_t *volinfo, char *slave, char *slave_ip, + char *slave_vol, char *path_list, dict_t *dict, + dict_t *resp_dict, char **op_errstr) { - int ret = 0; - struct stat stbuf = {0,}; - char rlogfile[PATH_MAX] = {0,}; - time_t rottime = 0; - - ret = glusterd_send_sigstop (-pid); - rottime = time (NULL); + int32_t ret = -1; + runner_t runner = {0,}; + glusterd_conf_t *priv = NULL; + char *master = NULL; + char *gl_workdir = NULL; + char geo_rep_dir[PATH_MAX] = ""; + char *conf_path = NULL; - snprintf (rlogfile, sizeof (rlogfile), "%s.%"PRIu64, logfile1, - (uint64_t) rottime); - ret = rename (logfile1, rlogfile); - if (ret) - gf_log ("", GF_LOG_ERROR, "rename failed for geo-rep log file"); + GF_ASSERT (slave); + GF_ASSERT (slave_ip); + GF_ASSERT (slave_vol); + GF_ASSERT (op_errstr); + GF_ASSERT (dict); + GF_ASSERT (resp_dict); + if (THIS) + priv = THIS->private; + if (priv == NULL) { + gf_log ("", GF_LOG_ERROR, "priv of glusterd not present"); + *op_errstr = gf_strdup ("glusterd defunct"); + goto out; + } - snprintf (rlogfile, sizeof (rlogfile), "%s.%"PRIu64, logfile2, - (uint64_t) rottime); - ret = stat (logfile2, &stbuf); + ret = dict_get_str (dict, "conf_path", &conf_path); if (ret) { - if (errno != ENOENT) - gf_log("", GF_LOG_ERROR, "stat failed for slave log" - " file: %s", logfile2); - else { - gf_log ("", GF_LOG_DEBUG, "Slave is not local, skipping rotation"); - ret = 0; - } + gf_log ("", GF_LOG_ERROR, + "Unable to fetch conf file path."); goto out; } - ret = rename (logfile2, rlogfile); - if (ret) - gf_log ("", GF_LOG_ERROR, "rename failed for geo-rep slave log file"); - - out: - ret = glusterd_send_sigcont (-pid); - - return ret; -} + gl_workdir = priv->workdir; + master = ""; + runinit (&runner); + runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", + "--delete", "-c", NULL); + runner_argprintf (&runner, "%s", conf_path); -int -glusterd_get_pid_from_file (char *master, char *slave, pid_t *pid) -{ - int ret = -1; - int pfd = 0; - char pidfile[PATH_MAX] = {0,}; - char buff[1024] = {0,}; + if (volinfo) { + master = volinfo->volname; + runner_argprintf (&runner, ":%s", master); + } + runner_add_arg (&runner, slave); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + synclock_unlock (&priv->big_lock); + ret = runner_run (&runner); + synclock_lock (&priv->big_lock); + if (ret) { + gf_log ("", GF_LOG_ERROR, "gsyncd failed to " + "delete session info for %s and %s peers", + master, slave); - pfd = gsyncd_getpidfile (master, slave, pidfile); + gf_asprintf (op_errstr, "gsyncd failed to " + "delete session info for %s and %s peers", + master, slave); - if (pfd == -2) { - gf_log ("", GF_LOG_ERROR, GEOREP" log-rotate validation " - " failed for %s & %s", master, slave); - goto out; - } - if (gsync_status_byfd (pfd) == -1) { - gf_log ("", GF_LOG_ERROR, "gsyncd b/w %s & %s is not" - " running", master, slave); goto out; } - ret = read (pfd, buff, 1024); - if (ret < 0) { - gf_log ("", GF_LOG_ERROR, GEOREP" cannot read pid from pid-file"); - goto out; - } + ret = snprintf (geo_rep_dir, sizeof(geo_rep_dir) - 1, + "%s/"GEOREP"/%s_%s_%s", gl_workdir, + volinfo->volname, slave_ip, slave_vol); + geo_rep_dir[ret] = '\0'; - close(pfd); + ret = rmdir (geo_rep_dir); + if (ret) { + if (errno == ENOENT) + gf_log ("", GF_LOG_DEBUG, "Geo Rep Dir(%s) Not Present.", + geo_rep_dir); + else { + gf_log ("", GF_LOG_ERROR, "Unable to delete " + "Geo Rep Dir(%s). Error: %s", geo_rep_dir, + strerror (errno)); + goto out; + } + } - *pid = strtol (buff, NULL, 10); ret = 0; - out: + gf_asprintf (op_errstr, "delete successful"); + +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } int -glusterd_do_gsync_log_rotate (char *master, char *slave, uuid_t *uuid, char **op_errstr) +glusterd_op_sys_exec (dict_t *dict, char **op_errstr, dict_t *rsp_dict) { - int ret = 0; - glusterd_conf_t *priv = NULL; - pid_t pid = 0; - char log_file1[PATH_MAX] = {0,}; - char log_file2[PATH_MAX] = {0,}; - - GF_ASSERT (THIS); - GF_ASSERT (THIS->private); + char buf[PATH_MAX] = ""; + char cmd_arg_name[PATH_MAX] = ""; + char output_name[PATH_MAX] = ""; + char errmsg[PATH_MAX] = ""; + char *ptr = NULL; + char *bufp = NULL; + char *command = NULL; + char **cmd_args = NULL; + int ret = -1; + int i = -1; + int cmd_args_count = 0; + int output_count = 0; + glusterd_conf_t *priv = NULL; + runner_t runner = {0,}; - priv = THIS->private; + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); - ret = glusterd_get_pid_from_file (master, slave, &pid); - if (ret) + if (THIS) + priv = THIS->private; + if (priv == NULL) { + gf_log ("", GF_LOG_ERROR, "priv of glusterd not present"); + *op_errstr = gf_strdup ("glusterd defunct"); goto out; + } - /* log file */ - ret = glusterd_gsyncd_getlogfile (master, slave, log_file1); - if (ret) + ret = dict_get_str (dict, "command", &command); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get command from dict"); goto out; + } - /* slave log file */ - ret = glusterd_gsync_get_slave_log_file (master, slave, log_file2); + ret = dict_get_int32 (dict, "cmd_args_count", &cmd_args_count); if (ret) - goto out; - - ret = glusterd_send_log_rotate_signal (pid, log_file1, log_file2); - - out: - if (ret && op_errstr) - *op_errstr = gf_strdup("Error rotating log file"); - return ret; -} - -int -glusterd_do_gsync_log_rotation_mst_slv (glusterd_volinfo_t *volinfo, char *slave, - char **op_errstr) -{ - uuid_t uuid = {0, }; - glusterd_conf_t *priv = NULL; - int ret = 0; - char errmsg[1024] = {0,}; - - GF_ASSERT (volinfo); - GF_ASSERT (slave); - GF_ASSERT (THIS); - GF_ASSERT (THIS->private); + gf_log ("", GF_LOG_INFO, "No cmd_args_count"); + + if (cmd_args_count) { + cmd_args = GF_CALLOC (cmd_args_count, sizeof (char*), + gf_common_mt_char); + if (!cmd_args) { + gf_log ("", GF_LOG_ERROR, "Unable to calloc. " + "Errno = %s", strerror(errno)); + goto out; + } - priv = THIS->private; + for (i=1; i <= cmd_args_count; i++) { + memset (cmd_arg_name, '\0', sizeof(cmd_arg_name)); + snprintf (cmd_arg_name, sizeof(cmd_arg_name), + "cmd_arg_%d", i); + ret = dict_get_str (dict, cmd_arg_name, &cmd_args[i-1]); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get %s in dict", + cmd_arg_name); + goto out; + } + } + } - ret = glusterd_gsync_get_uuid (slave, volinfo, uuid); - if ((ret == 0) && (uuid_compare (priv->uuid, uuid) != 0)) + runinit (&runner); + runner_argprintf (&runner, GSYNCD_PREFIX"/peer_%s", command); + for (i=0; i < cmd_args_count; i++) + runner_add_arg (&runner, cmd_args[i]); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + synclock_unlock (&priv->big_lock); + ret = runner_start (&runner); + if (ret == -1) { + snprintf (errmsg, sizeof (errmsg), "Unable to " + "execute command. Error : %s", + strerror (errno)); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + synclock_lock (&priv->big_lock); goto out; + } + ptr = fgets(buf, sizeof(buf), runner_chio (&runner, STDOUT_FILENO)); + if (ptr) { + ret = dict_get_int32 (rsp_dict, "output_count", &output_count); + if (ret) + output_count = 1; + else + output_count++; + memset (output_name, '\0', sizeof (output_name)); + snprintf (output_name, sizeof (output_name), + "output_%d", output_count); + if (buf[strlen(buf) - 1] == '\n') + buf[strlen(buf) - 1] = '\0'; + bufp = gf_strdup (buf); + if (!bufp) + gf_log ("", GF_LOG_ERROR, "gf_strdup failed."); + ret = dict_set_dynstr (rsp_dict, output_name, bufp); + if (ret) { + GF_FREE (bufp); + gf_log ("", GF_LOG_ERROR, "output set failed."); + } + ret = dict_set_int32 (rsp_dict, "output_count", output_count); + if (ret) + gf_log ("", GF_LOG_ERROR, "output_count set failed."); + } + + ret = runner_end (&runner); if (ret) { - snprintf(errmsg, sizeof(errmsg), "geo-replication session b/w %s %s not active", - volinfo->volname, slave); - gf_log ("", GF_LOG_WARNING, errmsg); - if (op_errstr) - *op_errstr = gf_strdup(errmsg); + snprintf (errmsg, sizeof (errmsg), "Unable to " + "end. Error : %s", + strerror (errno)); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + synclock_lock (&priv->big_lock); goto out; } + synclock_lock (&priv->big_lock); - ret = glusterd_do_gsync_log_rotate (volinfo->volname, slave, &uuid, op_errstr); + ret = 0; +out: + if (cmd_args) { + GF_FREE (cmd_args); + cmd_args = NULL; + } - out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } -static void -_iterate_log_rotate_mst_slv (dict_t *this, char *key, data_t *value, void *data) +int +glusterd_op_copy_file (dict_t *dict, char **op_errstr) { - glusterd_gsync_status_temp_t *param = NULL; - char *slave = NULL; + char abs_filename[PATH_MAX] = ""; + char errmsg[PATH_MAX] = ""; + char *filename = NULL; + char *host_uuid = NULL; + char uuid_str [64] = {0}; + char *contents = NULL; + char buf[1024] = ""; + int ret = -1; + int fd = -1; + int bytes_writen = 0; + int bytes_read = 0; + int contents_size = -1; + int file_mode = -1; + glusterd_conf_t *priv = NULL; + struct stat stbuf = {0,}; - param = (glusterd_gsync_status_temp_t *) data; - GF_ASSERT (param); - GF_ASSERT (param->volinfo); + if (THIS) + priv = THIS->private; + if (priv == NULL) { + gf_log ("", GF_LOG_ERROR, "priv of glusterd not present"); + *op_errstr = gf_strdup ("glusterd defunct"); + goto out; + } - slave = strchr (value->data, ':'); - if (slave) - slave++; - else { - gf_log ("", GF_LOG_ERROR, "geo-replication log-rotate: slave (%s) " - "not conforming to format", slave); - return; + ret = dict_get_str (dict, "host-uuid", &host_uuid); + if (ret < 0) + goto out; + + ret = dict_get_str (dict, "source", &filename); + if (ret < 0) { + gf_log ("", GF_LOG_ERROR, "Unable to fetch" + " filename from dict."); + *op_errstr = gf_strdup ("command unsuccessful"); + goto out; } + snprintf (abs_filename, sizeof(abs_filename), + "%s/%s", priv->workdir, filename); - (void) glusterd_do_gsync_log_rotation_mst_slv (param->volinfo, slave, NULL); -} + uuid_utoa_r (MY_UUID, uuid_str); + if (!strcmp (uuid_str, host_uuid)) { + ret = lstat (abs_filename, &stbuf); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Source file" + " does not exist in %s", priv->workdir); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } -int -glusterd_do_gsync_log_rotation_mst (glusterd_volinfo_t *volinfo) -{ - glusterd_gsync_status_temp_t param = {0, }; + contents = GF_CALLOC(1, stbuf.st_size+1, gf_common_mt_char); + if (!contents) { + snprintf (errmsg, sizeof (errmsg), + "Unable to allocate memory"); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } - GF_ASSERT (volinfo); + fd = open (abs_filename, O_RDONLY); + if (fd < 0) { + snprintf (errmsg, sizeof (errmsg), "Unable to open %s", + abs_filename); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } - param.volinfo = volinfo; - dict_foreach (volinfo->gsync_slaves, _iterate_log_rotate_mst_slv, ¶m); - return 0; -} + do { + ret = read (fd, buf, sizeof(buf)); + if (ret > 0) { + memcpy (contents+bytes_read, buf, ret); + bytes_read += ret; + memset (buf, '\0', sizeof(buf)); + } + } while (ret > 0); -static int -glusterd_rotate_gsync_all () -{ - int32_t ret = 0; - glusterd_conf_t *priv = NULL; - glusterd_volinfo_t *volinfo = NULL; + if (bytes_read != stbuf.st_size) { + snprintf (errmsg, sizeof (errmsg), "Unable to read all " + "the data from %s", abs_filename); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } - GF_ASSERT (THIS); - priv = THIS->private; + ret = dict_set_int32 (dict, "contents_size", stbuf.st_size); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Unable to set" + " contents size in dict."); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } - GF_ASSERT (priv); + ret = dict_set_int32 (dict, "file_mode", + (int32_t)stbuf.st_mode); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Unable to set" + " file mode in dict."); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } - list_for_each_entry (volinfo, &priv->volumes, vol_list) { - ret = glusterd_do_gsync_log_rotation_mst (volinfo); - if (ret) + ret = dict_set_bin (dict, "common_pem_contents", + contents, stbuf.st_size); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Unable to set" + " pem contents in dict."); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); goto out; - } + } + close (fd); + } else { + ret = dict_get_bin (dict, "common_pem_contents", + (void **) &contents); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Unable to get" + " pem contents in dict."); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } - out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); - return ret; -} + ret = dict_get_int32 (dict, "contents_size", &contents_size); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Unable to set" + " contents size in dict."); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } -static int -glusterd_rotate_gsync_logs (dict_t *dict, char **op_errstr, dict_t *rsp_dict) -{ - char *slave = NULL; - char *volname = NULL; - char errmsg[1024] = {0,}; - gf_boolean_t exists = _gf_false; - glusterd_volinfo_t *volinfo = NULL; - int ret = 0; + ret = dict_get_int32 (dict, "file_mode", &file_mode); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Unable to get" + " file mode in dict."); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } - ret = dict_get_str (dict, "master", &volname); - if (ret < 0) { - ret = glusterd_rotate_gsync_all (); - goto out; - } + fd = open (abs_filename, O_WRONLY | O_TRUNC | O_CREAT, 0600); + if (fd < 0) { + snprintf (errmsg, sizeof (errmsg), "Unable to open %s", + abs_filename); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } - exists = glusterd_check_volume_exists (volname); - ret = glusterd_volinfo_find (volname, &volinfo); - if ((ret) || (!exists)) { - snprintf (errmsg, sizeof(errmsg), "Volume %s does not" - " exist", volname); - gf_log ("", GF_LOG_WARNING, errmsg); - *op_errstr = gf_strdup (errmsg); - ret = -1; - goto out; - } + bytes_writen = write (fd, contents, contents_size); - ret = dict_get_str (dict, "slave", &slave); - if (ret < 0) { - ret = glusterd_do_gsync_log_rotation_mst (volinfo); - goto out; - } + if (bytes_writen != contents_size) { + snprintf (errmsg, sizeof (errmsg), "Failed to write" + " to %s", abs_filename); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } - ret = glusterd_do_gsync_log_rotation_mst_slv (volinfo, slave, op_errstr); + fchmod (fd, file_mode); + close (fd); + } - out: + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } - int glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict) { @@ -1855,11 +3484,16 @@ glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict) dict_t *resp_dict = NULL; char *host_uuid = NULL; char *slave = NULL; + char *slave_ip = NULL; + char *slave_vol = NULL; char *volname = NULL; + char *path_list = NULL; glusterd_volinfo_t *volinfo = NULL; glusterd_conf_t *priv = NULL; + gf_boolean_t is_force = _gf_false; char *status_msg = NULL; - uuid_t uuid = {0, }; + gf_boolean_t is_running = _gf_false; + char *conf_path = NULL; GF_ASSERT (THIS); GF_ASSERT (THIS->private); @@ -1885,15 +3519,28 @@ glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict) goto out; } - if (type == GF_GSYNC_OPTION_TYPE_ROTATE) { - ret = glusterd_rotate_gsync_logs (dict, op_errstr, resp_dict); + ret = dict_get_str (dict, "slave", &slave); + if (ret < 0) goto out; + ret = dict_get_str (dict, "slave_ip", &slave_ip); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to fetch slave volume name."); + goto out; } - ret = dict_get_str (dict, "slave", &slave); - if (ret < 0) + ret = dict_get_str (dict, "slave_vol", &slave_vol); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to fetch slave volume name."); goto out; + } + + ret = dict_get_str (dict, "conf_path", &conf_path); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch conf file path."); + goto out; + } if (dict_get_str (dict, "master", &volname) == 0) { ret = glusterd_volinfo_find (volname, &volinfo); @@ -1902,11 +3549,31 @@ glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict) volname); goto out; } + + ret = glusterd_get_local_brickpaths (volinfo, &path_list); } if (type == GF_GSYNC_OPTION_TYPE_CONFIG) { - ret = glusterd_gsync_configure (volinfo, slave, dict, resp_dict, - op_errstr); + ret = glusterd_gsync_configure (volinfo, slave, path_list, + dict, resp_dict, op_errstr); + + ret = dict_set_str (resp_dict, "conf_path", conf_path); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store conf_file_path."); + goto out; + } + goto out; + } + + if (type == GF_GSYNC_OPTION_TYPE_DELETE) { + ret = glusterd_remove_slave_in_info(volinfo, slave, op_errstr); + if (ret && !is_force && path_list) + goto out; + + ret = glusterd_gsync_delete (volinfo, slave, slave_ip, + slave_vol, path_list, dict, + resp_dict, op_errstr); goto out; } @@ -1915,48 +3582,652 @@ glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict) goto out; } + is_force = dict_get_str_boolean (dict, "force", _gf_false); + if (type == GF_GSYNC_OPTION_TYPE_START) { - ret = glusterd_set_marker_gsync (volinfo); + ret = glusterd_set_gsync_confs (volinfo); if (ret != 0) { - gf_log ("", GF_LOG_WARNING, "marker start failed"); + gf_log ("", GF_LOG_WARNING, "marker/changelog start failed"); *op_errstr = gf_strdup ("failed to initialize indexing"); ret = -1; goto out; } - ret = glusterd_store_slave_in_info(volinfo, slave, - host_uuid, op_errstr); - if (ret) - goto out; - ret = glusterd_start_gsync (volinfo, slave, host_uuid, - op_errstr); + ret = glusterd_start_gsync (volinfo, slave, path_list, + conf_path, host_uuid, op_errstr); } if (type == GF_GSYNC_OPTION_TYPE_STOP) { - - ret = glusterd_gsync_get_uuid (slave, volinfo, uuid); - if (ret) { + ret = glusterd_check_gsync_running_local (volinfo->volname, + slave, conf_path, + &is_running); + if (!ret && !is_force && path_list && + (_gf_true != is_running)) { gf_log ("", GF_LOG_WARNING, GEOREP" is not set up for" "%s(master) and %s(slave)", volname, slave); *op_errstr = strdup (GEOREP" is not set up"); goto out; } - ret = glusterd_remove_slave_in_info(volinfo, slave, op_errstr); - if (ret) + ret = stop_gsync (volname, slave, &status_msg, conf_path, is_force); + if (ret == 0 && status_msg) + ret = dict_set_str (resp_dict, "gsync-status", + status_msg); + if (ret != 0 && !is_force && path_list) + *op_errstr = gf_strdup ("internal error"); + + if (!ret) { + ret = glusterd_create_status_file (volinfo->volname, + slave, slave_ip, + slave_vol, "Stopped"); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to update" + "state_file. Error : %s", + strerror (errno)); + } + } + } + +out: + if (path_list) { + GF_FREE (path_list); + path_list = NULL; + } + + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterd_get_slave_details_confpath (glusterd_volinfo_t *volinfo, dict_t *dict, + char **slave_ip, char **slave_vol, + char **conf_path, char **op_errstr) +{ + int ret = -1; + char confpath[PATH_MAX] = ""; + glusterd_conf_t *priv = NULL; + char *slave = NULL; + + GF_ASSERT (THIS); + priv = THIS->private; + GF_ASSERT (priv); + + ret = dict_get_str (dict, "slave", &slave); + if (ret || !slave) { + gf_log ("", GF_LOG_ERROR, "Unable to fetch slave from dict"); + ret = -1; + goto out; + } + + ret = glusterd_get_slave_info (slave, slave_ip, slave_vol, op_errstr); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave details."); + ret = -1; + goto out; + } + + ret = dict_set_str (dict, "slave_ip", *slave_ip); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store slave IP."); + goto out; + } + + ret = dict_set_str (dict, "slave_vol", *slave_vol); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store slave volume name."); + goto out; + } + + ret = snprintf (confpath, sizeof(confpath) - 1, + "%s/"GEOREP"/%s_%s_%s/gsyncd.conf", + priv->workdir, volinfo->volname, + *slave_ip, *slave_vol); + confpath[ret] = '\0'; + *conf_path = gf_strdup (confpath); + if (!(*conf_path)) { + gf_log ("", GF_LOG_ERROR, + "Unable to gf_strdup. Error: %s", strerror (errno)); + ret = -1; + goto out; + } + + ret = dict_set_str (dict, "conf_path", *conf_path); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store conf_path"); + goto out; + } + +out: + gf_log ("", GF_LOG_DEBUG,"Returning %d", ret); + return ret; + +} + +static int +glusterd_get_slave_info (char *slave, char **slave_ip, + char **slave_vol, char **op_errstr) +{ + char *tmp = NULL; + char *save_ptr = NULL; + char **linearr = NULL; + int32_t ret = -1; + char errmsg[PATH_MAX] = ""; + + ret = glusterd_urltransform_single (slave, "normalize", + &linearr); + if (ret == -1) { + ret = snprintf (errmsg, sizeof(errmsg) - 1, + "Invalid Url: %s", slave); + errmsg[ret] = '\0'; + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "Failed to normalize url"); + goto out; + } + + tmp = strtok_r (linearr[0], "/", &save_ptr); + tmp = strtok_r (NULL, "/", &save_ptr); + slave = strtok_r (tmp, ":", &save_ptr); + if (slave) { + ret = glusterd_mountbroker_check (&slave, op_errstr); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Invalid slave url: %s", *op_errstr); goto out; + } - if (uuid_compare (priv->uuid, uuid) != 0) { + *slave_ip = gf_strdup (slave); + if (!*slave_ip) { + gf_log ("", GF_LOG_ERROR, + "Failed to gf_strdup"); + ret = -1; goto out; } + gf_log ("", GF_LOG_DEBUG, "Slave IP : %s", *slave_ip); + ret = 0; + } else { + gf_log ("", GF_LOG_ERROR, "Invalid slave name"); + goto out; + } - ret = stop_gsync (volname, slave, &status_msg); - if (ret == 0 && status_msg) - ret = dict_set_str (resp_dict, "gsync-status", - status_msg); - if (ret != 0) - *op_errstr = gf_strdup ("internal error"); + slave = strtok_r (NULL, ":", &save_ptr); + if (slave) { + *slave_vol = gf_strdup (slave); + if (!*slave_vol) { + gf_log ("", GF_LOG_ERROR, + "Failed to gf_strdup"); + ret = -1; + goto out; + } + gf_log ("", GF_LOG_DEBUG, "Slave Vol : %s", *slave_vol); + ret = 0; + } else { + gf_log ("", GF_LOG_ERROR, "Invalid slave name"); + goto out; + } + +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +static void +runinit_gsyncd_setrx (runner_t *runner, char *conf_path) +{ + runinit (runner); + runner_add_args (runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL); + runner_argprintf (runner, "%s", conf_path); + runner_add_arg (runner, "--config-set-rx"); +} + +static int +glusterd_check_gsync_present (int *valid_state) +{ + char buff[PATH_MAX] = {0, }; + runner_t runner = {0,}; + char *ptr = NULL; + int ret = 0; + + runinit (&runner); + runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "--version", NULL); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + ret = runner_start (&runner); + if (ret == -1) { + if (errno == ENOENT) { + gf_log ("glusterd", GF_LOG_INFO, GEOREP + " module not installed in the system"); + *valid_state = 0; + } + else { + gf_log ("glusterd", GF_LOG_ERROR, GEOREP + " module not working as desired"); + *valid_state = -1; + } + goto out; + } + + ptr = fgets(buff, sizeof(buff), runner_chio (&runner, STDOUT_FILENO)); + if (ptr) { + if (!strstr (buff, "gsyncd")) { + ret = -1; + gf_log ("glusterd", GF_LOG_ERROR, GEOREP" module not " + "working as desired"); + *valid_state = -1; + goto out; + } + } else { + ret = -1; + gf_log ("glusterd", GF_LOG_ERROR, GEOREP" module not " + "working as desired"); + *valid_state = -1; + goto out; + } + + ret = 0; + out: + + runner_end (&runner); + + gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + return ret; + +} + +static int +create_conf_file (glusterd_conf_t *conf, char *conf_path) +#define RUN_GSYNCD_CMD do { \ + ret = runner_run_reuse (&runner); \ + if (ret == -1) { \ + runner_log (&runner, "glusterd", GF_LOG_ERROR, "command failed"); \ + runner_end (&runner); \ + goto out; \ + } \ + runner_end (&runner); \ +} while (0) +{ + int ret = 0; + runner_t runner = {0,}; + char georepdir[PATH_MAX] = {0,}; + int valid_state = 0; + + valid_state = -1; + ret = glusterd_check_gsync_present (&valid_state); + if (-1 == ret) { + ret = valid_state; + goto out; + } + + ret = snprintf (georepdir, sizeof(georepdir) - 1, "%s/"GEOREP, + conf->workdir); + georepdir[ret] = '\0'; + + /************ + * master pre-configuration + ************/ + + /* remote-gsyncd */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, "remote-gsyncd", GSYNCD_PREFIX"/gsyncd", ".", ".", NULL); + RUN_GSYNCD_CMD; + + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, "remote-gsyncd", "/nonexistent/gsyncd", + ".", "^ssh:", NULL); + RUN_GSYNCD_CMD; + + /* gluster-command-dir */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, "gluster-command-dir", SBIN_DIR"/", + ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* gluster-params */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, "gluster-params", + "aux-gfid-mount xlator-option=*-dht.assert-no-child-down=true", + ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* ssh-command */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_arg (&runner, "ssh-command"); + runner_argprintf (&runner, + "ssh -oPasswordAuthentication=no " + "-oStrictHostKeyChecking=no " + "-i %s/secret.pem", georepdir); + runner_add_args (&runner, ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* pid-file */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_arg (&runner, "pid-file"); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.pid", georepdir); + runner_add_args (&runner, ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* state-file */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_arg (&runner, "state-file"); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.status", georepdir); + runner_add_args (&runner, ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* state-detail-file */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_arg (&runner, "state-detail-file"); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}-detail.status", georepdir); + runner_add_args (&runner, ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* state-socket */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_arg (&runner, "state-socket-unencoded"); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.socket", georepdir); + runner_add_args (&runner, ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* socketdir */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, "socketdir", GLUSTERD_SOCK_DIR, ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* log-file */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, + "log-file", + DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/${mastervol}/${eSlave}.log", + ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* gluster-log-file */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, + "gluster-log-file", + DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/${mastervol}/${eSlave}${local_id}.gluster.log", + ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* ignore-deletes */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, "ignore-deletes", "true", ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* special-sync-mode */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, "special-sync-mode", "partial", ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* change-detector == changelog */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args(&runner, "change-detector", "changelog", ".", ".", NULL); + RUN_GSYNCD_CMD; + + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_arg(&runner, "working-dir"); + runner_argprintf(&runner, "%s/${mastervol}/${eSlave}", + DEFAULT_VAR_RUN_DIRECTORY); + runner_add_args (&runner, ".", ".", NULL); + RUN_GSYNCD_CMD; + + /************ + * slave pre-configuration + ************/ + + /* gluster-command-dir */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, "gluster-command-dir", SBIN_DIR"/", + ".", NULL); + RUN_GSYNCD_CMD; + + /* gluster-params */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, "gluster-params", + "aux-gfid-mount xlator-option=*-dht.assert-no-child-down=true", + ".", NULL); + RUN_GSYNCD_CMD; + + /* log-file */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, + "log-file", + DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/${session_owner}:${eSlave}.log", + ".", NULL); + RUN_GSYNCD_CMD; + + /* MountBroker log-file */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, + "log-file-mbr", + DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/mbr/${session_owner}:${eSlave}.log", + ".", NULL); + RUN_GSYNCD_CMD; + + /* gluster-log-file */ + runinit_gsyncd_setrx (&runner, conf_path); + runner_add_args (&runner, + "gluster-log-file", + DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/${session_owner}:${eSlave}.gluster.log", + ".", NULL); + RUN_GSYNCD_CMD; + + out: + return ret ? -1 : 0; +} + +static int +glusterd_create_essential_dir_files (glusterd_volinfo_t *volinfo, dict_t *dict, + char *slave, char *slave_ip, + char *slave_vol, char **op_errstr) +{ + int ret = -1; + char *conf_path = NULL; + char *statefile = NULL; + char buf[PATH_MAX] = ""; + char errmsg[PATH_MAX] = ""; + glusterd_conf_t *conf = NULL; + struct stat stbuf = {0,}; + + GF_ASSERT (THIS); + conf = THIS->private; + + ret = dict_get_str (dict, "conf_path", &conf_path); + if (ret) { + snprintf (errmsg, sizeof (errmsg), + "Unable to fetch conf file path."); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } + + ret = dict_get_str (dict, "statefile", &statefile); + if (ret) { + snprintf (errmsg, sizeof (errmsg), + "Unable to fetch statefile path."); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } + + ret = snprintf (buf, sizeof(buf) - 1, "%s/"GEOREP"/%s_%s_%s", + conf->workdir, volinfo->volname, slave_ip, slave_vol); + buf[ret] = '\0'; + ret = mkdir_p (buf, 0777, _gf_true); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Unable to create %s" + ". Error : %s", buf, strerror (errno)); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } + + ret = snprintf (buf, PATH_MAX, DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/%s", + volinfo->volname); + buf[ret] = '\0'; + ret = mkdir_p (buf, 0777, _gf_true); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Unable to create %s" + ". Error : %s", buf, strerror (errno)); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } + + ret = lstat (conf_path, &stbuf); + if (!ret) { + gf_log ("", GF_LOG_DEBUG, "Session already running." + " Not creating config file again."); + } else { + ret = create_conf_file (conf, conf_path); + if (ret || lstat (conf_path, &stbuf)) { + snprintf (errmsg, sizeof (errmsg), "Failed to create" + " config file(%s).", conf_path); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } + } + + ret = lstat (statefile, &stbuf); + if (!ret) { + gf_log ("", GF_LOG_DEBUG, "Session already running." + " Not creating status file again."); + goto out; + } else { + ret = glusterd_create_status_file (volinfo->volname, slave, + slave_ip, slave_vol, + "Not Started"); + if (ret || lstat (statefile, &stbuf)) { + snprintf (errmsg, sizeof (errmsg), "Unable to create %s" + ". Error : %s", statefile, strerror (errno)); + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } + } + +out: + gf_log ("", GF_LOG_DEBUG,"Returning %d", ret); + return ret; +} + +int +glusterd_op_gsync_create (dict_t *dict, char **op_errstr, dict_t *rsp_dict) +{ + char common_pem_file[PATH_MAX] = ""; + char errmsg[PATH_MAX] = ""; + char hooks_args[PATH_MAX] = ""; + char uuid_str [64] = ""; + char *host_uuid = NULL; + char *slave_ip = NULL; + char *slave_vol = NULL; + char *arg_buf = NULL; + char *volname = NULL; + char *slave = NULL; + int32_t ret = -1; + int32_t is_pem_push = -1; + gf_boolean_t is_force = -1; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + + GF_ASSERT (THIS); + conf = THIS->private; + GF_ASSERT (conf); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + ret = glusterd_op_gsync_args_get (dict, op_errstr, + &volname, &slave, &host_uuid); + if (ret) + goto out; + + snprintf (common_pem_file, sizeof(common_pem_file), + "%s"GLUSTERD_COMMON_PEM_PUB_FILE, conf->workdir); + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Volinfo for %s" + " (master) not found", volname); + goto out; + } + + ret = dict_get_str (dict, "slave_vol", &slave_vol); + if (ret) { + snprintf (errmsg, sizeof (errmsg), + "Unable to fetch slave volume name."); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; + } + + ret = dict_get_str (dict, "slave_ip", &slave_ip); + if (ret) { + snprintf (errmsg, sizeof (errmsg), + "Unable to fetch slave IP."); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } + + is_force = dict_get_str_boolean (dict, "force", _gf_false); + + uuid_utoa_r (MY_UUID, uuid_str); + if (!strcmp (uuid_str, host_uuid)) { + ret = dict_get_int32 (dict, "push_pem", &is_pem_push); + if (!ret && is_pem_push) { + gf_log ("", GF_LOG_DEBUG, "Trying to setup" + " pem files in slave"); + is_pem_push = 1; + } else + is_pem_push = 0; + + snprintf(hooks_args, sizeof(hooks_args), + "is_push_pem=%d pub_file=%s slave_ip=%s", + is_pem_push, common_pem_file, slave_ip); + + } else + snprintf(hooks_args, sizeof(hooks_args), + "This argument will stop the hooks script"); + + arg_buf = gf_strdup (hooks_args); + if (!arg_buf) { + gf_log ("", GF_LOG_ERROR, "Failed to" + " gf_strdup"); + if (is_force) { + ret = 0; + goto create_essentials; + } + ret = -1; + goto out; + } + + ret = dict_set_str (dict, "hooks_args", arg_buf); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Failed to set" + " hooks_args in dict."); + if (is_force) { + ret = 0; + goto create_essentials; + } + goto out; + } + +create_essentials: + + ret = glusterd_create_essential_dir_files (volinfo, dict, slave, + slave_ip, slave_vol, + op_errstr); + if (ret) + goto out; + + ret = glusterd_store_slave_in_info (volinfo, slave, + host_uuid, op_errstr, + is_force); + if (ret) { + snprintf (errmsg, sizeof (errmsg), "Unable to store" + " slave info."); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + goto out; } out: diff --git a/xlators/mgmt/glusterd/src/glusterd-handler.c b/xlators/mgmt/glusterd/src/glusterd-handler.c index 93900acc3..71d076624 100644 --- a/xlators/mgmt/glusterd/src/glusterd-handler.c +++ b/xlators/mgmt/glusterd/src/glusterd-handler.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" @@ -43,6 +33,7 @@ #include "glusterd-op-sm.h" #include "glusterd-utils.h" #include "glusterd-store.h" +#include "glusterd-locks.h" #include "glusterd1-xdr.h" #include "cli1-xdr.h" @@ -57,6 +48,39 @@ #include "defaults.c" #include "common-utils.h" +#include "globals.h" +#include "glusterd-syncop.h" + +#ifdef HAVE_BD_XLATOR +#include <lvm2app.h> +#endif + +extern uuid_t global_txn_id; + +int glusterd_big_locked_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, + void *data, rpc_clnt_notify_t notify_fn) +{ + glusterd_conf_t *priv = THIS->private; + int ret = -1; + synclock_lock (&priv->big_lock); + ret = notify_fn (rpc, mydata, event, data); + synclock_unlock (&priv->big_lock); + return ret; +} + +int glusterd_big_locked_handler (rpcsvc_request_t *req, rpcsvc_actor actor_fn) +{ + glusterd_conf_t *priv = THIS->private; + int ret = -1; + + synclock_lock (&priv->big_lock); + ret = actor_fn (req); + synclock_unlock (&priv->big_lock); + + return ret; +} + static int glusterd_handle_friend_req (rpcsvc_request_t *req, uuid_t uuid, char *hostname, int port, @@ -78,10 +102,12 @@ glusterd_handle_friend_req (rpcsvc_request_t *req, uuid_t uuid, ret = glusterd_friend_find (uuid, rhost, &peerinfo); if (ret) { - ret = glusterd_xfer_friend_add_resp (req, rhost, port, -1, - GF_PROBE_UNKNOWN_PEER); - if (friend_req->vols.vols_val) + ret = glusterd_xfer_friend_add_resp (req, hostname, rhost, port, + -1, GF_PROBE_UNKNOWN_PEER); + if (friend_req->vols.vols_val) { free (friend_req->vols.vols_val); + friend_req->vols.vols_val = NULL; + } goto out; } @@ -139,19 +165,16 @@ out: if (0 != ret) { if (ctx && ctx->hostname) GF_FREE (ctx->hostname); - if (ctx) - GF_FREE (ctx); + GF_FREE (ctx); if (dict) { if ((!dict->extra_stdfree) && friend_req->vols.vols_val) free (friend_req->vols.vols_val); dict_unref (dict); } else { - if (friend_req->vols.vols_val) - free (friend_req->vols.vols_val); + free (friend_req->vols.vols_val); } - if (event) - GF_FREE (event); + GF_FREE (event); } else { if (peerinfo && (0 == peerinfo->connected)) ret = GLUSTERD_CONNECTION_AWAITED; @@ -221,8 +244,7 @@ out: if (0 != ret) { if (ctx && ctx->hostname) GF_FREE (ctx->hostname); - if (ctx) - GF_FREE (ctx); + GF_FREE (ctx); } return ret; @@ -235,13 +257,14 @@ glusterd_add_peer_detail_to_dict (glusterd_peerinfo_t *peerinfo, int ret = -1; char key[256] = {0, }; + char *peer_uuid_str = NULL; GF_ASSERT (peerinfo); GF_ASSERT (friends); snprintf (key, 256, "friend%d.uuid", count); - uuid_utoa_r (peerinfo->uuid, peerinfo->uuid_str); - ret = dict_set_str (friends, key, peerinfo->uuid_str); + peer_uuid_str = gd_peer_uuid_str (peerinfo); + ret = dict_set_str (friends, key, peer_uuid_str); if (ret) goto out; @@ -255,6 +278,11 @@ glusterd_add_peer_detail_to_dict (glusterd_peerinfo_t *peerinfo, if (ret) goto out; + snprintf (key, 256, "friend%d.stateId", count); + ret = dict_set_int32 (friends, key, peerinfo->state.state); + if (ret) + goto out; + snprintf (key, 256, "friend%d.state", count); ret = dict_set_str (friends, key, glusterd_friend_sm_state_name_get(peerinfo->state.state)); @@ -270,10 +298,34 @@ out: return ret; } +struct args_pack { + dict_t *dict; + int vol_count; + int opt_count; +}; + +static int +_build_option_key (dict_t *d, char *k, data_t *v, void *tmp) +{ + char reconfig_key[256] = {0, }; + struct args_pack *pack = NULL; + int ret = -1; + + pack = tmp; + if (strcmp (k, GLUSTERD_GLOBAL_OPT_VERSION) == 0) + return 0; + snprintf (reconfig_key, 256, "volume%d.option.%s", + pack->vol_count, k); + ret = dict_set_str (pack->dict, reconfig_key, v->data); + if (0 == ret) + pack->opt_count++; + + return 0; +} int glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo, - dict_t *volumes, int count) + dict_t *volumes, int count) { int ret = -1; @@ -281,19 +333,20 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo = NULL; char *buf = NULL; int i = 1; - data_pair_t *pairs = NULL; - char reconfig_key[256] = {0, }; dict_t *dict = NULL; - data_t *value = NULL; - int opt_count = 0; glusterd_conf_t *priv = NULL; char *volume_id_str = NULL; - + struct args_pack pack = {0,}; + xlator_t *this = NULL; +#ifdef HAVE_BD_XLATOR + int caps = 0; +#endif GF_ASSERT (volinfo); GF_ASSERT (volumes); - priv = THIS->private; + this = THIS; + priv = this->private; GF_ASSERT (priv); @@ -312,6 +365,21 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; + /* As of now, the snap volumes are also displayed as part of + volume info command. So this change is to display whether + the volume is original volume or the snap_volume. If + displaying of snap volumes in volume info o/p is not needed + this should be removed. + */ + snprintf (key, 256, "volume%d.snap_volume", count); + ret = dict_set_int32 (volumes, key, volinfo->is_snap_volume); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "failed to set whether " + "the volume is a snap volume or actual volume (%s)", + volinfo->volname); + goto out; + } + snprintf (key, 256, "volume%d.brick_count", count); ret = dict_set_int32 (volumes, key, volinfo->brick_count); if (ret) @@ -346,8 +414,82 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; + snprintf (key, 256, "volume%d.rebalance", count); + ret = dict_set_int32 (volumes, key, volinfo->rebal.defrag_cmd); + if (ret) + goto out; + +#ifdef HAVE_BD_XLATOR + if (volinfo->caps) { + caps = 0; + snprintf (key, 256, "volume%d.xlator0", count); + buf = GF_MALLOC (256, gf_common_mt_char); + if (!buf) { + ret = ENOMEM; + goto out; + } + if (volinfo->caps & CAPS_BD) + snprintf (buf, 256, "BD"); + ret = dict_set_dynstr (volumes, key, buf); + if (ret) { + GF_FREE (buf); + goto out; + } + + if (volinfo->caps & CAPS_THIN) { + snprintf (key, 256, "volume%d.xlator0.caps%d", count, + caps++); + buf = GF_MALLOC (256, gf_common_mt_char); + if (!buf) { + ret = ENOMEM; + goto out; + } + snprintf (buf, 256, "thin"); + ret = dict_set_dynstr (volumes, key, buf); + if (ret) { + GF_FREE (buf); + goto out; + } + } + + if (volinfo->caps & CAPS_OFFLOAD_COPY) { + snprintf (key, 256, "volume%d.xlator0.caps%d", count, + caps++); + buf = GF_MALLOC (256, gf_common_mt_char); + if (!buf) { + ret = ENOMEM; + goto out; + } + snprintf (buf, 256, "offload_copy"); + ret = dict_set_dynstr (volumes, key, buf); + if (ret) { + GF_FREE (buf); + goto out; + } + } + + if (volinfo->caps & CAPS_OFFLOAD_SNAPSHOT) { + snprintf (key, 256, "volume%d.xlator0.caps%d", count, + caps++); + buf = GF_MALLOC (256, gf_common_mt_char); + if (!buf) { + ret = ENOMEM; + goto out; + } + snprintf (buf, 256, "offload_snapshot"); + ret = dict_set_dynstr (volumes, key, buf); + if (ret) { + GF_FREE (buf); + goto out; + } + } + + } +#endif + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { char brick[1024] = {0,}; + char brick_uuid[64] = {0,}; snprintf (key, 256, "volume%d.brick%d", count, i); snprintf (brick, 1024, "%s:%s", brickinfo->hostname, brickinfo->path); @@ -355,6 +497,25 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo, ret = dict_set_dynstr (volumes, key, buf); if (ret) goto out; + snprintf (key, 256, "volume%d.brick%d.uuid", count, i); + snprintf (brick_uuid, 64, "%s", uuid_utoa (brickinfo->uuid)); + buf = gf_strdup (brick_uuid); + if (!buf) + goto out; + ret = dict_set_dynstr (volumes, key, buf); + if (ret) + goto out; + +#ifdef HAVE_BD_XLATOR + if (volinfo->caps & CAPS_BD) { + snprintf (key, 256, "volume%d.vg%d", count, i); + snprintf (brick, 1024, "%s", brickinfo->vg); + buf = gf_strdup (brick); + ret = dict_set_dynstr (volumes, key, buf); + if (ret) + goto out; + } +#endif i++; } @@ -364,25 +525,14 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo, goto out; } - pairs = dict->members_list; - - while (pairs) { - if (1 == glusterd_check_option_exists (pairs->key, NULL)) { - value = pairs->value; - if (!value) - continue; + pack.dict = volumes; + pack.vol_count = count; + pack.opt_count = 0; + dict_foreach (dict, _build_option_key, (void *) &pack); + dict_foreach (priv->opts, _build_option_key, &pack); - snprintf (reconfig_key, 256, "volume%d.option.%s", count, - pairs->key); - ret = dict_set_str (volumes, reconfig_key, value->data); - if (!ret) - opt_count++; - } - pairs = pairs->next; - } - - snprintf (key, 256, "volume%d.opt_count", count); - ret = dict_set_int32 (volumes, key, opt_count); + snprintf (key, 256, "volume%d.opt_count", pack.vol_count); + ret = dict_set_int32 (volumes, key, pack.opt_count); out: return ret; } @@ -392,13 +542,18 @@ glusterd_friend_find (uuid_t uuid, char *hostname, glusterd_peerinfo_t **peerinfo) { int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); if (uuid) { ret = glusterd_friend_find_by_uuid (uuid, peerinfo); if (ret) { - gf_log ("glusterd", GF_LOG_INFO, - "Unable to find peer by uuid"); + gf_log (this->name, GF_LOG_DEBUG, + "Unable to find peer by uuid: %s", + uuid_utoa (uuid)); } else { goto out; } @@ -409,7 +564,7 @@ glusterd_friend_find (uuid_t uuid, char *hostname, ret = glusterd_friend_find_by_hostname (hostname, peerinfo); if (ret) { - gf_log ("glusterd", GF_LOG_INFO, + gf_log (this->name, GF_LOG_DEBUG, "Unable to find hostname: %s", hostname); } else { goto out; @@ -421,73 +576,200 @@ out: } int32_t -glusterd_op_txn_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx) +glusterd_op_txn_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx, + char *err_str, size_t err_len) { - int32_t ret = -1; - xlator_t *this = NULL; - glusterd_conf_t *priv = NULL; - int32_t locked = 0; + int32_t ret = -1; + dict_t *dict = NULL; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int32_t locked = 0; + char *tmp = NULL; + char *volname = NULL; + uuid_t *txn_id = NULL; + uuid_t *originator_uuid = NULL; + glusterd_op_info_t txn_op_info = {{0},}; + glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; GF_ASSERT (req); GF_ASSERT ((op > GD_OP_NONE) && (op < GD_OP_MAX)); GF_ASSERT (NULL != ctx); this = THIS; + GF_ASSERT (this); priv = this->private; GF_ASSERT (priv); - ret = glusterd_lock (priv->uuid); + dict = ctx; + + /* Generate a transaction-id for this operation and + * save it in the dict. This transaction id distinguishes + * each transaction, and helps separate opinfos in the + * op state machine. */ + txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + if (!txn_id) + goto out; + + uuid_generate (*txn_id); + + ret = dict_set_bin (dict, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set transaction id."); + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, + "Transaction_id = %s", uuid_utoa (*txn_id)); + + /* Save the MY_UUID as the originator_uuid. This originator_uuid + * will be used by is_origin_glusterd() to determine if a node + * is the originator node for a command. */ + originator_uuid = GF_CALLOC (1, sizeof(uuid_t), + gf_common_mt_uuid_t); + if (!originator_uuid) { + ret = -1; + goto out; + } + uuid_copy (*originator_uuid, MY_UUID); + ret = dict_set_bin (dict, "originator_uuid", + originator_uuid, sizeof (uuid_t)); if (ret) { gf_log (this->name, GF_LOG_ERROR, - "Unable to acquire local lock, ret: %d", ret); + "Failed to set originator uuid."); goto out; } + /* Based on the op_version, acquire a cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + ret = glusterd_lock (MY_UUID); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock on localhost, ret: %d", + ret); + snprintf (err_str, err_len, + "Another transaction is in progress. " + "Please try again after sometime."); + goto out; + } + } else { + /* If no volname is given as a part of the command, locks will + * not be held */ + ret = dict_get_str (dict, "volname", &tmp); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Failed to get volume " + "name"); + goto local_locking_done; + } else { + /* Use a copy of volname, as cli response will be + * sent before the unlock, and the volname in the + * dict, might be removed */ + volname = gf_strdup (tmp); + if (!volname) + goto out; + } + + ret = glusterd_mgmt_v3_lock (volname, MY_UUID, "vol"); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock for %s", volname); + snprintf (err_str, err_len, + "Another transaction is in progress for %s. " + "Please try again after sometime.", volname); + goto out; + } + } + locked = 1; - gf_log (this->name, GF_LOG_INFO, "Acquired local lock"); + gf_log (this->name, GF_LOG_DEBUG, "Acquired lock on localhost"); + +local_locking_done: + + /* If no volname is given as a part of the command, locks will + * not be held, hence sending stage event. */ + if (volname) + event_type = GD_OP_EVENT_START_LOCK; + else { + txn_op_info.state.state = GD_OP_STATE_LOCK_SENT; + event_type = GD_OP_EVENT_ALL_ACC; + } + + /* Save opinfo for this transaction with the transaction id */ + glusterd_txn_opinfo_init (&txn_op_info, NULL, &op, ctx, req); + + ret = glusterd_set_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set transaction's opinfo"); + if (ctx) + dict_unref (ctx); + goto out; + } - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_START_LOCK, NULL); + ret = glusterd_op_sm_inject_event (event_type, txn_id, ctx); if (ret) { gf_log (this->name, GF_LOG_ERROR, "Failed to acquire cluster" " lock."); goto out; } - glusterd_op_set_op (op); - glusterd_op_set_ctx (ctx); - glusterd_op_set_req (req); - - out: - if (locked && ret) - glusterd_unlock (priv->uuid); + if (locked && ret) { + /* Based on the op-version, we release the + * cluster or mgmt_v3 lock */ + if (priv->op_version < 3) + glusterd_unlock (MY_UUID); + else { + ret = glusterd_mgmt_v3_unlock (volname, MY_UUID, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to release lock for %s", + volname); + ret = -1; + } + } + + if (volname) + GF_FREE (volname); gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } int -glusterd_handle_cluster_lock (rpcsvc_request_t *req) +__glusterd_handle_cluster_lock (rpcsvc_request_t *req) { - gd1_mgmt_cluster_lock_req lock_req = {{0},}; - int32_t ret = -1; - glusterd_op_lock_ctx_t *ctx = NULL; - glusterd_peerinfo_t *peerinfo = NULL; + dict_t *op_ctx = NULL; + int32_t ret = -1; + gd1_mgmt_cluster_lock_req lock_req = {{0},}; + glusterd_op_lock_ctx_t *ctx = NULL; + glusterd_op_t op = GD_OP_EVENT_LOCK; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_op_info_t txn_op_info = {{0},}; + uuid_t *txn_id = &global_txn_id; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &lock_req, (xdrproc_t)xdr_gd1_mgmt_cluster_lock_req)) { - //failed to decode msg; + ret = xdr_to_generic (req->msg[0], &lock_req, + (xdrproc_t)xdr_gd1_mgmt_cluster_lock_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode lock " + "request received from peer"); req->rpc_err = GARBAGE_ARGS; goto out; } - gf_log ("glusterd", GF_LOG_INFO, - "Received LOCK from uuid: %s", uuid_utoa (lock_req.uuid)); + gf_log (this->name, GF_LOG_DEBUG, "Received LOCK from uuid: %s", + uuid_utoa (lock_req.uuid)); if (glusterd_friend_find_by_uuid (lock_req.uuid, &peerinfo)) { - gf_log (THIS->name, GF_LOG_WARNING, "%s doesn't " + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " "belong to the cluster. Ignoring request.", uuid_utoa (lock_req.uuid)); ret = -1; @@ -503,11 +785,32 @@ glusterd_handle_cluster_lock (rpcsvc_request_t *req) uuid_copy (ctx->uuid, lock_req.uuid); ctx->req = req; + ctx->dict = NULL; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_LOCK, ctx); + op_ctx = dict_new (); + if (!op_ctx) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set new dict"); + goto out; + } + + glusterd_txn_opinfo_init (&txn_op_info, NULL, &op, op_ctx, req); + + ret = glusterd_set_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set transaction's opinfo"); + dict_unref (txn_op_info.op_ctx); + goto out; + } + + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_LOCK, txn_id, ctx); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to inject event GD_OP_EVENT_LOCK"); out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); glusterd_friend_sm (); glusterd_op_sm (); @@ -516,6 +819,13 @@ out: } int +glusterd_handle_cluster_lock (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cluster_lock); +} + +int glusterd_req_ctx_create (rpcsvc_request_t *rpc_req, glusterd_op_t op, uuid_t uuid, char *buf_val, size_t buf_len, @@ -526,10 +836,13 @@ glusterd_req_ctx_create (rpcsvc_request_t *rpc_req, char str[50] = {0,}; glusterd_req_ctx_t *req_ctx = NULL; dict_t *dict = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); uuid_unparse (uuid, str); - gf_log ("glusterd", GF_LOG_INFO, - "Received op from uuid: %s", str); + gf_log (this->name, GF_LOG_DEBUG, "Received op from uuid %s", str); dict = dict_new (); if (!dict) @@ -544,7 +857,7 @@ glusterd_req_ctx_create (rpcsvc_request_t *rpc_req, req_ctx->op = op; ret = dict_unserialize (buf_val, buf_len, &dict); if (ret) { - gf_log ("", GF_LOG_WARNING, + gf_log (this->name, GF_LOG_WARNING, "failed to unserialize the dictionary"); goto out; } @@ -557,29 +870,38 @@ out: if (ret) { if (dict) dict_unref (dict); - if (req_ctx) - GF_FREE (req_ctx); + GF_FREE (req_ctx); } return ret; } int -glusterd_handle_stage_op (rpcsvc_request_t *req) +__glusterd_handle_stage_op (rpcsvc_request_t *req) { int32_t ret = -1; glusterd_req_ctx_t *req_ctx = NULL; gd1_mgmt_stage_op_req op_req = {{0},}; glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; + glusterd_op_info_t txn_op_info = {{0},}; + glusterd_op_sm_state_info_t state; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &op_req, (xdrproc_t)xdr_gd1_mgmt_stage_op_req)) { - //failed to decode msg; + + ret = xdr_to_generic (req->msg[0], &op_req, + (xdrproc_t)xdr_gd1_mgmt_stage_op_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode stage " + "request received from peer"); req->rpc_err = GARBAGE_ARGS; goto out; } if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) { - gf_log (THIS->name, GF_LOG_WARNING, "%s doesn't " + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " "belong to the cluster. Ignoring request.", uuid_utoa (op_req.uuid)); ret = -1; @@ -592,34 +914,76 @@ glusterd_handle_stage_op (rpcsvc_request_t *req) if (ret) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_OP, req_ctx); + ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + + /* In cases where there is no volname, the receivers won't have a + * transaction opinfo created, as for those operations, the locking + * phase where the transaction opinfos are created, won't be called. */ + ret = glusterd_get_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "No transaction's opinfo set"); + + state.state = GD_OP_STATE_LOCKED; + glusterd_txn_opinfo_init (&txn_op_info, &state, + &op_req.op, req_ctx->dict, req); + + ret = glusterd_set_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set transaction's opinfo"); + dict_unref (req_ctx->dict); + goto out; + } + } + + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_OP, + txn_id, req_ctx); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to inject event GD_OP_EVENT_STAGE_OP"); out: - if (op_req.buf.buf_val) - free (op_req.buf.buf_val);//malloced by xdr + free (op_req.buf.buf_val);//malloced by xdr glusterd_friend_sm (); glusterd_op_sm (); return ret; } int -glusterd_handle_commit_op (rpcsvc_request_t *req) +glusterd_handle_stage_op (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_stage_op); +} + + +int +__glusterd_handle_commit_op (rpcsvc_request_t *req) { int32_t ret = -1; glusterd_req_ctx_t *req_ctx = NULL; gd1_mgmt_commit_op_req op_req = {{0},}; glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &op_req, (xdrproc_t)xdr_gd1_mgmt_commit_op_req)) { - //failed to decode msg; + ret = xdr_to_generic (req->msg[0], &op_req, + (xdrproc_t)xdr_gd1_mgmt_commit_op_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode commit " + "request received from peer"); req->rpc_err = GARBAGE_ARGS; goto out; } if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) { - gf_log (THIS->name, GF_LOG_WARNING, "%s doesn't " + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " "belong to the cluster. Ignoring request.", uuid_utoa (op_req.uuid)); ret = -1; @@ -634,69 +998,124 @@ glusterd_handle_commit_op (rpcsvc_request_t *req) if (ret) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_OP, req_ctx); - if (ret) - goto out; - ret = glusterd_op_init_ctx (op_req.op); + ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_OP, + txn_id, req_ctx); out: - if (op_req.buf.buf_val) - free (op_req.buf.buf_val);//malloced by xdr + free (op_req.buf.buf_val);//malloced by xdr glusterd_friend_sm (); glusterd_op_sm (); return ret; } + int -glusterd_handle_cli_probe (rpcsvc_request_t *req) +glusterd_handle_commit_op (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_commit_op); +} + +int +__glusterd_handle_cli_probe (rpcsvc_request_t *req) { int32_t ret = -1; - gf1_cli_probe_req cli_req = {0,}; - glusterd_peerinfo_t *peerinfo = NULL; - gf_boolean_t run_fsm = _gf_true; + gf_cli_req cli_req = {{0,},}; + glusterd_peerinfo_t *peerinfo = NULL; + gf_boolean_t run_fsm = _gf_true; + xlator_t *this = NULL; + char *bind_name = NULL; + dict_t *dict = NULL; + char *hostname = NULL; + int port = 0; + GF_ASSERT (req); + this = THIS; - if (!xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf1_cli_probe_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; gf_log ("", GF_LOG_ERROR, "xdr decoding error"); req->rpc_err = GARBAGE_ARGS; goto out; } - gf_cmd_log ("peer probe", " on host %s:%d", cli_req.hostname, - cli_req.port); + if (cli_req.dict.dict_len) { + dict = dict_new (); + + ret = dict_unserialize (cli_req.dict.dict_val, + cli_req.dict.dict_len, &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "unserialize req-buffer to dictionary"); + goto out; + } + } + + ret = dict_get_str (dict, "hostname", &hostname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get hostname"); + goto out; + } + + ret = dict_get_int32 (dict, "port", &port); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get port"); + goto out; + } + + if (glusterd_is_any_volume_in_server_quorum (this) && + !does_gd_meet_server_quorum (this)) { + glusterd_xfer_cli_probe_resp (req, -1, GF_PROBE_QUORUM_NOT_MET, + NULL, hostname, port, dict); + gf_log (this->name, GF_LOG_ERROR, "Quorum does not meet, " + "rejecting operation"); + ret = 0; + goto out; + } + gf_log ("glusterd", GF_LOG_INFO, "Received CLI probe req %s %d", - cli_req.hostname, cli_req.port); + hostname, port); - if (!(ret = glusterd_is_local_addr(cli_req.hostname))) { + if (dict_get_str(this->options,"transport.socket.bind-address", + &bind_name) == 0) { + gf_log ("glusterd", GF_LOG_DEBUG, + "only checking probe address vs. bind address"); + ret = gf_is_same_address (bind_name, hostname); + } + else { + ret = gf_is_local_addr (hostname); + } + if (ret) { glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_LOCALHOST, - cli_req.hostname, cli_req.port); + NULL, hostname, port, dict); + ret = 0; goto out; } - if (!(ret = glusterd_friend_find_by_hostname(cli_req.hostname, - &peerinfo))) { - if (strcmp (peerinfo->hostname, cli_req.hostname) == 0) { + if (!(ret = glusterd_friend_find_by_hostname (hostname, &peerinfo))) { + if (strcmp (peerinfo->hostname, hostname) == 0) { - gf_log ("glusterd", GF_LOG_DEBUG, "Probe host %s port %d" - " already a peer", cli_req.hostname, cli_req.port); + gf_log ("glusterd", GF_LOG_DEBUG, "Probe host %s port " + "%d already a peer", hostname, port); glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_FRIEND, - cli_req.hostname, cli_req.port); + NULL, hostname, port, + dict); goto out; } } - ret = glusterd_probe_begin (req, cli_req.hostname, cli_req.port); - - gf_cmd_log ("peer probe","on host %s:%d %s",cli_req.hostname, cli_req.port, - (ret) ? "FAILED" : "SUCCESS"); + ret = glusterd_probe_begin (req, hostname, port, dict); if (ret == GLUSTERD_CONNECTION_AWAITED) { //fsm should be run after connection establishes run_fsm = _gf_false; ret = 0; } + out: - if (cli_req.hostname) - free (cli_req.hostname);//its malloced by xdr + free (cli_req.dict.dict_val); if (run_fsm) { glusterd_friend_sm (); @@ -707,14 +1126,24 @@ out: } int -glusterd_handle_cli_deprobe (rpcsvc_request_t *req) +glusterd_handle_cli_probe (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_cli_probe); +} + +int +__glusterd_handle_cli_deprobe (rpcsvc_request_t *req) { int32_t ret = -1; - gf1_cli_deprobe_req cli_req = {0,}; - uuid_t uuid = {0}; - int op_errno = 0; - xlator_t *this = NULL; - glusterd_conf_t *priv = NULL; + gf_cli_req cli_req = {{0,},}; + uuid_t uuid = {0}; + int op_errno = 0; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + dict_t *dict = NULL; + char *hostname = NULL; + int port = 0; + int flags = 0; this = THIS; GF_ASSERT (this); @@ -722,62 +1151,99 @@ glusterd_handle_cli_deprobe (rpcsvc_request_t *req) GF_ASSERT (priv); GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf1_cli_deprobe_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, + (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; } + if (cli_req.dict.dict_len) { + dict = dict_new (); + + ret = dict_unserialize (cli_req.dict.dict_val, + cli_req.dict.dict_len, &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "unserialize req-buffer to dictionary"); + goto out; + } + } + gf_log ("glusterd", GF_LOG_INFO, "Received CLI deprobe req"); - ret = glusterd_hostname_to_uuid (cli_req.hostname, uuid); + ret = dict_get_str (dict, "hostname", &hostname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get hostname"); + goto out; + } + + ret = dict_get_int32 (dict, "port", &port); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get port"); + goto out; + } + + ret = dict_get_int32 (dict, "flags", &flags); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get flags"); + goto out; + } + + ret = glusterd_hostname_to_uuid (hostname, uuid); if (ret) { op_errno = GF_DEPROBE_NOT_FRIEND; goto out; } - if (!uuid_compare (uuid, priv->uuid)) { + if (!uuid_compare (uuid, MY_UUID)) { op_errno = GF_DEPROBE_LOCALHOST; ret = -1; goto out; } - if (!uuid_is_null (uuid) && !(cli_req.flags & GF_CLI_FLAG_OP_FORCE)) { - /* Check if peers are connected, except peer being detached*/ - if (!glusterd_chk_peers_connected_befriended (uuid)) { - ret = -1; - op_errno = GF_DEPROBE_FRIEND_DOWN; - goto out; + if (!(flags & GF_CLI_FLAG_OP_FORCE)) { + if (!uuid_is_null (uuid)) { + /* Check if peers are connected, except peer being detached*/ + if (!glusterd_chk_peers_connected_befriended (uuid)) { + ret = -1; + op_errno = GF_DEPROBE_FRIEND_DOWN; + goto out; + } + ret = glusterd_all_volume_cond_check ( + glusterd_friend_brick_belongs, + -1, &uuid); + if (ret) { + op_errno = GF_DEPROBE_BRICK_EXIST; + goto out; + } } - ret = glusterd_all_volume_cond_check ( - glusterd_friend_brick_belongs, - -1, &uuid); - if (ret) { - op_errno = GF_DEPROBE_BRICK_EXIST; + + if (glusterd_is_any_volume_in_server_quorum (this) && + !does_gd_meet_server_quorum (this)) { + gf_log (this->name, GF_LOG_ERROR, "Quorum does not " + "meet, rejecting operation"); + ret = -1; + op_errno = GF_DEPROBE_QUORUM_NOT_MET; goto out; } } if (!uuid_is_null (uuid)) { - ret = glusterd_deprobe_begin (req, cli_req.hostname, - cli_req.port, uuid); + ret = glusterd_deprobe_begin (req, hostname, port, uuid, dict); } else { - ret = glusterd_deprobe_begin (req, cli_req.hostname, - cli_req.port, NULL); + ret = glusterd_deprobe_begin (req, hostname, port, NULL, dict); } - gf_cmd_log ("peer deprobe", "on host %s:%d %s", cli_req.hostname, - cli_req.port, (ret) ? "FAILED" : "SUCCESS"); out: + free (cli_req.dict.dict_val); + if (ret) { - ret = glusterd_xfer_cli_deprobe_resp (req, ret, op_errno, - cli_req.hostname); + ret = glusterd_xfer_cli_deprobe_resp (req, ret, op_errno, NULL, + hostname, dict); } - if (cli_req.hostname) - free (cli_req.hostname);//malloced by xdr - glusterd_friend_sm (); glusterd_op_sm (); @@ -785,7 +1251,13 @@ out: } int -glusterd_handle_cli_list_friends (rpcsvc_request_t *req) +glusterd_handle_cli_deprobe (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_cli_deprobe); +} + +int +__glusterd_handle_cli_list_friends (rpcsvc_request_t *req) { int32_t ret = -1; gf1_cli_peer_list_req cli_req = {0,}; @@ -793,7 +1265,9 @@ glusterd_handle_cli_list_friends (rpcsvc_request_t *req) GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf1_cli_peer_list_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, + (xdrproc_t)xdr_gf1_cli_peer_list_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -831,7 +1305,14 @@ out: } int -glusterd_handle_cli_get_volume (rpcsvc_request_t *req) +glusterd_handle_cli_list_friends (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_list_friends); +} + +int +__glusterd_handle_cli_get_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; @@ -840,7 +1321,8 @@ glusterd_handle_cli_get_volume (rpcsvc_request_t *req) GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -883,31 +1365,313 @@ out: return ret; } -int32_t -glusterd_op_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx) +int +glusterd_handle_cli_get_volume (rpcsvc_request_t *req) { - int ret = -1; + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_get_volume); +} + +int +__glusterd_handle_cli_uuid_reset (rpcsvc_request_t *req) +{ + int ret = -1; + dict_t *dict = NULL; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + uuid_t uuid = {0}; + gf_cli_rsp rsp = {0,}; + gf_cli_req cli_req = {{0,}}; + char msg_str[2048] = {0,}; + + GF_ASSERT (req); + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + //failed to decode msg; + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + gf_log ("glusterd", GF_LOG_DEBUG, "Received uuid reset req"); + + if (cli_req.dict.dict_len) { + /* Unserialize the dictionary */ + dict = dict_new (); + + ret = dict_unserialize (cli_req.dict.dict_val, + cli_req.dict.dict_len, + &dict); + if (ret < 0) { + gf_log ("glusterd", GF_LOG_ERROR, + "failed to " + "unserialize req-buffer to dictionary"); + snprintf (msg_str, sizeof (msg_str), "Unable to decode " + "the buffer"); + goto out; + } else { + dict->extra_stdfree = cli_req.dict.dict_val; + } + } + + /* In the above section if dict_unserialize is successful, ret is set + * to zero. + */ + ret = -1; + // Do not allow peer reset if there are any volumes in the cluster + if (!list_empty (&priv->volumes)) { + snprintf (msg_str, sizeof (msg_str), "volumes are already " + "present in the cluster. Resetting uuid is not " + "allowed"); + gf_log (this->name, GF_LOG_WARNING, "%s", msg_str); + goto out; + } + + // Do not allow peer reset if trusted storage pool is already formed + if (!list_empty (&priv->peers)) { + snprintf (msg_str, sizeof (msg_str),"trusted storage pool " + "has been already formed. Please detach this peer " + "from the pool and reset its uuid."); + gf_log (this->name, GF_LOG_WARNING, "%s", msg_str); + goto out; + } + + uuid_copy (uuid, priv->uuid); + ret = glusterd_uuid_generate_save (); + + if (!uuid_compare (uuid, MY_UUID)) { + snprintf (msg_str, sizeof (msg_str), "old uuid and the new uuid" + " are same. Try gluster peer reset again"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg_str); + ret = -1; + goto out; + } + +out: + if (ret) { + rsp.op_ret = -1; + if (msg_str[0] == '\0') + snprintf (msg_str, sizeof (msg_str), "Operation " + "failed"); + rsp.op_errstr = msg_str; + ret = 0; + } else { + rsp.op_errstr = ""; + } - ret = glusterd_op_txn_begin (req, op, ctx); + glusterd_to_cli (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_cli_rsp, dict); return ret; } +int +glusterd_handle_cli_uuid_reset (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_uuid_reset); +} + +int +__glusterd_handle_cli_uuid_get (rpcsvc_request_t *req) +{ + int ret = -1; + dict_t *dict = NULL; + dict_t *rsp_dict = NULL; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + gf_cli_rsp rsp = {0,}; + gf_cli_req cli_req = {{0,}}; + char msg_str[2048] = {0,}; + char uuid_str[64] = {0,}; + + GF_ASSERT (req); + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + gf_log ("glusterd", GF_LOG_DEBUG, "Received uuid get req"); + + if (cli_req.dict.dict_len) { + dict = dict_new (); + if (!dict) { + ret = -1; + goto out; + } + + ret = dict_unserialize (cli_req.dict.dict_val, + cli_req.dict.dict_len, + &dict); + if (ret < 0) { + gf_log ("glusterd", GF_LOG_ERROR, + "failed to " + "unserialize req-buffer to dictionary"); + snprintf (msg_str, sizeof (msg_str), "Unable to decode " + "the buffer"); + goto out; + + } else { + dict->extra_stdfree = cli_req.dict.dict_val; + + } + } + + rsp_dict = dict_new (); + if (!rsp_dict) { + ret = -1; + goto out; + } + + uuid_utoa_r (MY_UUID, uuid_str); + ret = dict_set_str (rsp_dict, "uuid", uuid_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set uuid in " + "dictionary."); + goto out; + } + + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to serialize " + "dictionary."); + goto out; + } + ret = 0; +out: + if (ret) { + rsp.op_ret = -1; + if (msg_str[0] == '\0') + snprintf (msg_str, sizeof (msg_str), "Operation " + "failed"); + rsp.op_errstr = msg_str; + + } else { + rsp.op_errstr = ""; + + } + glusterd_to_cli (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_cli_rsp, dict); + return 0; +} int -glusterd_handle_reset_volume (rpcsvc_request_t *req) +glusterd_handle_cli_uuid_get (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_uuid_get); +} + +int +__glusterd_handle_cli_list_volume (rpcsvc_request_t *req) +{ + int ret = -1; + dict_t *dict = NULL; + glusterd_conf_t *priv = NULL; + glusterd_volinfo_t *volinfo = NULL; + int count = 0; + char key[1024] = {0,}; + gf_cli_rsp rsp = {0,}; + + GF_ASSERT (req); + + priv = THIS->private; + GF_ASSERT (priv); + + dict = dict_new (); + if (!dict) + goto out; + + list_for_each_entry (volinfo, &priv->volumes, vol_list) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d", count); + ret = dict_set_str (dict, key, volinfo->volname); + if (ret) + goto out; + count++; + } + + ret = dict_set_int32 (dict, "count", count); + if (ret) + goto out; + + ret = dict_allocate_and_serialize (dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); + if (ret) + goto out; + + ret = 0; + +out: + rsp.op_ret = ret; + if (ret) + rsp.op_errstr = "Error listing volumes"; + else + rsp.op_errstr = ""; + + glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_cli_rsp); + ret = 0; + + if (dict) + dict_unref (dict); + + glusterd_friend_sm (); + glusterd_op_sm (); + + return ret; +} + +int +glusterd_handle_cli_list_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_list_volume); +} + +int32_t +glusterd_op_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx, + char *err_str, size_t err_len) +{ + int ret = -1; + + ret = glusterd_op_txn_begin (req, op, ctx, err_str, err_len); + + return ret; +} + +int +__glusterd_handle_reset_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; dict_t *dict = NULL; glusterd_op_t cli_op = GD_OP_RESET_VOLUME; char *volname = NULL; + char err_str[2048] = {0,}; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); - if (!xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req)) { - //failed to decode msg; + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + snprintf (err_str, sizeof (err_str), "Failed to decode request " + "received from cli"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); req->rpc_err = GARBAGE_ARGS; goto out; } @@ -920,8 +1684,10 @@ glusterd_handle_reset_volume (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, "failed to " + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); goto out; } else { dict->extra_stdfree = cli_req.dict.dict_val; @@ -930,31 +1696,37 @@ glusterd_handle_reset_volume (rpcsvc_request_t *req) ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to get volname"); + snprintf (err_str, sizeof (err_str), "Failed to get volume " + "name"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } + gf_log (this->name, GF_LOG_DEBUG, "Received volume reset request for " + "volume %s", volname); - gf_cmd_log ("Volume reset", "volume : %s", volname); - ret = glusterd_op_begin (req, GD_OP_RESET_VOLUME, dict); - gf_cmd_log ("Volume reset", " on volume %s %s ", volname, - ((ret == 0)? " SUCCEDED":" FAILED")); + ret = glusterd_op_begin_synctask (req, GD_OP_RESET_VOLUME, dict); out: - glusterd_friend_sm (); - glusterd_op_sm (); if (ret) { - if (dict) - dict_unref (dict); + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, err_str); } return ret; } +int +glusterd_handle_reset_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_reset_volume); +} int -glusterd_handle_set_volume (rpcsvc_request_t *req) +__glusterd_handle_set_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; @@ -963,11 +1735,21 @@ glusterd_handle_set_volume (rpcsvc_request_t *req) char *key = NULL; char *value = NULL; char *volname = NULL; + char *op_errstr = NULL; + gf_boolean_t help = _gf_false; + char err_str[2048] = {0,}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req)) { - //failed to decode msg; + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + snprintf (err_str, sizeof (err_str), "Failed to decode " + "request received from cli"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); req->rpc_err = GARBAGE_ARGS; goto out; } @@ -980,9 +1762,11 @@ glusterd_handle_set_volume (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); goto out; } else { dict->extra_stdfree = cli_req.dict.dict_val; @@ -991,65 +1775,81 @@ glusterd_handle_set_volume (rpcsvc_request_t *req) ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_WARNING, "Unable to get volume name, while" - "handling volume set command"); + snprintf (err_str, sizeof (err_str), "Failed to get volume " + "name while handling volume set command"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + goto out; + } + + if (strcmp (volname, "help") == 0 || + strcmp (volname, "help-xml") == 0) { + ret = glusterd_volset_help (dict, &op_errstr); + help = _gf_true; goto out; } ret = dict_get_str (dict, "key1", &key); if (ret) { - if (strcmp (volname, "help-xml") && strcmp (volname, "help")) { - gf_log ("", GF_LOG_WARNING, "Unable to get key, while " - "handling volume set for %s",volname); - goto out; - } + snprintf (err_str, sizeof (err_str), "Failed to get key while" + " handling volume set for %s", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + goto out; } ret = dict_get_str (dict, "value1", &value); if (ret) { - if (strcmp (volname, "help-xml") && strcmp (volname, "help")) { - gf_log ("", GF_LOG_WARNING, "Unable to get value, while" - "handling volume set for %s",volname); - goto out; - } + snprintf (err_str, sizeof (err_str), "Failed to get value while" + " handling volume set for %s", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + goto out; } + gf_log (this->name, GF_LOG_DEBUG, "Received volume set request for " + "volume %s", volname); + ret = glusterd_op_begin_synctask (req, GD_OP_SET_VOLUME, dict); - gf_cmd_log ("volume set", "volume-name:%s: key:%s, value:%s",volname, - key, value); - ret = glusterd_op_begin (req, GD_OP_SET_VOLUME, dict); - gf_cmd_log ("volume set", "volume-name:%s: key:%s, value:%s %s", - volname, key, value, (ret == 0)? "SUCCEDED" : "FAILED" ); out: - - glusterd_friend_sm (); - glusterd_op_sm (); - - if (ret) { - if (dict) - dict_unref (dict); + if (help) + ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, dict, + (op_errstr)? op_errstr:""); + else if (ret) { + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, err_str); } + if (op_errstr) + GF_FREE (op_errstr); + return ret; } int -glusterd_handle_sync_volume (rpcsvc_request_t *req) +glusterd_handle_set_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_set_volume); +} + +int +__glusterd_handle_sync_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; dict_t *dict = NULL; gf_cli_rsp cli_rsp = {0.}; char msg[2048] = {0,}; - glusterd_volinfo_t *volinfo = NULL; char *volname = NULL; gf1_cli_sync_volume flags = 0; char *hostname = NULL; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); - if (!xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -1063,9 +1863,11 @@ glusterd_handle_sync_volume (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (msg, sizeof (msg), "Unable to decode the " + "command"); goto out; } else { dict->extra_stdfree = cli_req.dict.dict_val; @@ -1074,7 +1876,8 @@ glusterd_handle_sync_volume (rpcsvc_request_t *req) ret = dict_get_str (dict, "hostname", &hostname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to get hostname"); + snprintf (msg, sizeof (msg), "Failed to get hostname"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); goto out; } @@ -1082,48 +1885,25 @@ glusterd_handle_sync_volume (rpcsvc_request_t *req) if (ret) { ret = dict_get_int32 (dict, "flags", (int32_t*)&flags); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Unable to get volume" - "name, or flags"); + snprintf (msg, sizeof (msg), "Failed to get volume name" + " or flags"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); goto out; } } - gf_log ("glusterd", GF_LOG_INFO, "Received volume sync req " - "for volume %s", - (flags & GF_CLI_SYNC_ALL) ? "all" : volname); + gf_log (this->name, GF_LOG_INFO, "Received volume sync req " + "for volume %s", (flags & GF_CLI_SYNC_ALL) ? "all" : volname); - if (!glusterd_is_local_addr (hostname)) { + if (gf_is_local_addr (hostname)) { ret = -1; snprintf (msg, sizeof (msg), "sync from localhost" " not allowed"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); goto out; } - if (!flags) { - ret = glusterd_volinfo_find (volname, &volinfo); - if (!ret) { - snprintf (msg, sizeof (msg), "please delete the " - "volume: %s before sync", volname); - ret = -1; - goto out; - } - - ret = dict_set_dynmstr (dict, "volname", volname); - if (ret) { - gf_log ("", GF_LOG_ERROR, "volume name set failed"); - snprintf (msg, sizeof (msg), "volume name set failed"); - goto out; - } - } else { - if (glusterd_volume_count_get ()) { - snprintf (msg, sizeof (msg), "please delete all the " - "volumes before full sync"); - ret = -1; - goto out; - } - } - - ret = glusterd_op_begin (req, GD_OP_SYNC_VOLUME, dict); + ret = glusterd_op_begin_synctask (req, GD_OP_SYNC_VOLUME, dict); out: if (ret) { @@ -1131,22 +1911,22 @@ out: cli_rsp.op_errstr = msg; if (msg[0] == '\0') snprintf (msg, sizeof (msg), "Operation failed"); - glusterd_submit_reply(req, &cli_rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf_cli_rsp); - if (dict) - dict_unref (dict); - + glusterd_to_cli (req, &cli_rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_cli_rsp, dict); ret = 0; //sent error to cli, prevent second reply } - glusterd_friend_sm (); - glusterd_op_sm (); - return ret; } int +glusterd_handle_sync_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_sync_volume); +} + +int glusterd_fsm_log_send_resp (rpcsvc_request_t *req, int op_ret, char *op_errstr, dict_t *dict) { @@ -1161,12 +1941,11 @@ glusterd_fsm_log_send_resp (rpcsvc_request_t *req, int op_ret, rsp.op_errstr = op_errstr; if (rsp.op_ret == 0) ret = dict_allocate_and_serialize (dict, &rsp.fsm_log.fsm_log_val, - (size_t *)&rsp.fsm_log.fsm_log_len); + &rsp.fsm_log.fsm_log_len); ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_gf1_cli_fsm_log_rsp); - if (rsp.fsm_log.fsm_log_val) - GF_FREE (rsp.fsm_log.fsm_log_val); + GF_FREE (rsp.fsm_log.fsm_log_val); gf_log ("glusterd", GF_LOG_DEBUG, "Responded, ret: %d", ret); @@ -1174,7 +1953,7 @@ glusterd_fsm_log_send_resp (rpcsvc_request_t *req, int op_ret, } int -glusterd_handle_fsm_log (rpcsvc_request_t *req) +__glusterd_handle_fsm_log (rpcsvc_request_t *req) { int32_t ret = -1; gf1_cli_fsm_log_req cli_req = {0,}; @@ -1187,7 +1966,9 @@ glusterd_handle_fsm_log (rpcsvc_request_t *req) GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf1_cli_fsm_log_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, + (xdrproc_t)xdr_gf1_cli_fsm_log_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; snprintf (msg, sizeof (msg), "Garbage request"); @@ -1218,8 +1999,7 @@ glusterd_handle_fsm_log (rpcsvc_request_t *req) ret = glusterd_sm_tr_log_add_to_dict (dict, log); out: (void)glusterd_fsm_log_send_resp (req, ret, msg, dict); - if (cli_req.name) - free (cli_req.name);//malloced by xdr + free (cli_req.name);//malloced by xdr if (dict) dict_unref (dict); @@ -1230,6 +2010,12 @@ out: } int +glusterd_handle_fsm_log (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_fsm_log); +} + +int glusterd_op_lock_send_resp (rpcsvc_request_t *req, int32_t status) { @@ -1243,8 +2029,7 @@ glusterd_op_lock_send_resp (rpcsvc_request_t *req, int32_t status) ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_gd1_mgmt_cluster_lock_rsp); - gf_log ("glusterd", GF_LOG_INFO, - "Responded, ret: %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Responded to lock, ret: %d", ret); return 0; } @@ -1263,35 +2048,91 @@ glusterd_op_unlock_send_resp (rpcsvc_request_t *req, int32_t status) ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_rsp); - gf_log ("glusterd", GF_LOG_INFO, - "Responded to unlock, ret: %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Responded to unlock, ret: %d", ret); return ret; } int -glusterd_handle_cluster_unlock (rpcsvc_request_t *req) +glusterd_op_mgmt_v3_lock_send_resp (rpcsvc_request_t *req, uuid_t *txn_id, + int32_t status) +{ + + gd1_mgmt_v3_lock_rsp rsp = {{0},}; + int ret = -1; + + GF_ASSERT (req); + GF_ASSERT (txn_id); + glusterd_get_uuid (&rsp.uuid); + rsp.op_ret = status; + if (rsp.op_ret) + rsp.op_errno = errno; + uuid_copy (rsp.txn_id, *txn_id); + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp); + + gf_log (THIS->name, GF_LOG_DEBUG, "Responded to mgmt_v3 lock, ret: %d", + ret); + + return ret; +} + +int +glusterd_op_mgmt_v3_unlock_send_resp (rpcsvc_request_t *req, uuid_t *txn_id, + int32_t status) +{ + + gd1_mgmt_v3_unlock_rsp rsp = {{0},}; + int ret = -1; + + GF_ASSERT (req); + GF_ASSERT (txn_id); + rsp.op_ret = status; + if (rsp.op_ret) + rsp.op_errno = errno; + glusterd_get_uuid (&rsp.uuid); + uuid_copy (rsp.txn_id, *txn_id); + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp); + + gf_log (THIS->name, GF_LOG_DEBUG, + "Responded to mgmt_v3 unlock, ret: %d", + ret); + + return ret; +} + +int +__glusterd_handle_cluster_unlock (rpcsvc_request_t *req) { gd1_mgmt_cluster_unlock_req unlock_req = {{0}, }; int32_t ret = -1; glusterd_op_lock_ctx_t *ctx = NULL; glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &unlock_req, - (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_req)) { - //failed to decode msg; + ret = xdr_to_generic (req->msg[0], &unlock_req, + (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode unlock " + "request received from peer"); req->rpc_err = GARBAGE_ARGS; goto out; } - gf_log ("glusterd", GF_LOG_INFO, + gf_log (this->name, GF_LOG_DEBUG, "Received UNLOCK from uuid: %s", uuid_utoa (unlock_req.uuid)); if (glusterd_friend_find_by_uuid (unlock_req.uuid, &peerinfo)) { - gf_log (THIS->name, GF_LOG_WARNING, "%s doesn't " + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " "belong to the cluster. Ignoring request.", uuid_utoa (unlock_req.uuid)); ret = -1; @@ -1306,8 +2147,9 @@ glusterd_handle_cluster_unlock (rpcsvc_request_t *req) } uuid_copy (ctx->uuid, unlock_req.uuid); ctx->req = req; + ctx->dict = NULL; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_UNLOCK, ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_UNLOCK, txn_id, ctx); out: glusterd_friend_sm (); @@ -1317,14 +2159,25 @@ out: } int +glusterd_handle_cluster_unlock (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cluster_unlock); +} + +int glusterd_op_stage_send_resp (rpcsvc_request_t *req, int32_t op, int32_t status, char *op_errstr, dict_t *rsp_dict) { gd1_mgmt_stage_op_rsp rsp = {{0},}; int ret = -1; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); + rsp.op_ret = status; glusterd_get_uuid (&rsp.uuid); rsp.op = op; @@ -1333,11 +2186,10 @@ glusterd_op_stage_send_resp (rpcsvc_request_t *req, else rsp.op_errstr = ""; - ret = dict_allocate_and_serialize (rsp_dict, - &rsp.dict.dict_val, - (size_t *)&rsp.dict.dict_len); + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); if (ret < 0) { - gf_log ("", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_ERROR, "failed to get serialized length of dict"); return ret; } @@ -1345,10 +2197,8 @@ glusterd_op_stage_send_resp (rpcsvc_request_t *req, ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_gd1_mgmt_stage_op_rsp); - gf_log ("glusterd", GF_LOG_INFO, - "Responded to stage, ret: %d", ret); - if (rsp.dict.dict_val) - GF_FREE (rsp.dict.dict_val); + gf_log (this->name, GF_LOG_DEBUG, "Responded to stage, ret: %d", ret); + GF_FREE (rsp.dict.dict_val); return ret; } @@ -1360,7 +2210,10 @@ glusterd_op_commit_send_resp (rpcsvc_request_t *req, { gd1_mgmt_commit_op_rsp rsp = {{0}, }; int ret = -1; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); rsp.op_ret = status; glusterd_get_uuid (&rsp.uuid); @@ -1372,11 +2225,10 @@ glusterd_op_commit_send_resp (rpcsvc_request_t *req, rsp.op_errstr = ""; if (rsp_dict) { - ret = dict_allocate_and_serialize (rsp_dict, - &rsp.dict.dict_val, - (size_t *)&rsp.dict.dict_len); + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); if (ret < 0) { - gf_log ("", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_ERROR, "failed to get serialized length of dict"); goto out; } @@ -1386,24 +2238,24 @@ glusterd_op_commit_send_resp (rpcsvc_request_t *req, ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_gd1_mgmt_commit_op_rsp); - gf_log ("glusterd", GF_LOG_INFO, - "Responded to commit, ret: %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Responded to commit, ret: %d", ret); out: - if (rsp.dict.dict_val) - GF_FREE (rsp.dict.dict_val); + GF_FREE (rsp.dict.dict_val); return ret; } int -glusterd_handle_incoming_friend_req (rpcsvc_request_t *req) +__glusterd_handle_incoming_friend_req (rpcsvc_request_t *req) { int32_t ret = -1; gd1_mgmt_friend_req friend_req = {{0},}; gf_boolean_t run_fsm = _gf_true; GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &friend_req, (xdrproc_t)xdr_gd1_mgmt_friend_req)) { + ret = xdr_to_generic (req->msg[0], &friend_req, + (xdrproc_t)xdr_gd1_mgmt_friend_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -1422,8 +2274,7 @@ glusterd_handle_incoming_friend_req (rpcsvc_request_t *req) } out: - if (friend_req.hostname) - free (friend_req.hostname);//malloced by xdr + free (friend_req.hostname);//malloced by xdr if (run_fsm) { glusterd_friend_sm (); @@ -1434,14 +2285,23 @@ out: } int -glusterd_handle_incoming_unfriend_req (rpcsvc_request_t *req) +glusterd_handle_incoming_friend_req (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_incoming_friend_req); +} + +int +__glusterd_handle_incoming_unfriend_req (rpcsvc_request_t *req) { int32_t ret = -1; gd1_mgmt_friend_req friend_req = {{0},}; char remote_hostname[UNIX_PATH_MAX + 1] = {0,}; GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &friend_req, (xdrproc_t)xdr_gd1_mgmt_friend_req)) { + ret = xdr_to_generic (req->msg[0], &friend_req, + (xdrproc_t)xdr_gd1_mgmt_friend_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -1460,10 +2320,8 @@ glusterd_handle_incoming_unfriend_req (rpcsvc_request_t *req) remote_hostname, friend_req.port); out: - if (friend_req.hostname) - free (friend_req.hostname);//malloced by xdr - if (friend_req.vols.vols_val) - free (friend_req.vols.vols_val);//malloced by xdr + free (friend_req.hostname);//malloced by xdr + free (friend_req.vols.vols_val);//malloced by xdr glusterd_friend_sm (); glusterd_op_sm (); @@ -1472,6 +2330,14 @@ out: } int +glusterd_handle_incoming_unfriend_req (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_incoming_unfriend_req); + +} + +int glusterd_handle_friend_update_delete (dict_t *dict) { char *hostname = NULL; @@ -1517,7 +2383,7 @@ out: } int -glusterd_handle_friend_update (rpcsvc_request_t *req) +__glusterd_handle_friend_update (rpcsvc_request_t *req) { int32_t ret = -1; gd1_mgmt_friend_update friend_req = {{0},}; @@ -1543,7 +2409,9 @@ glusterd_handle_friend_update (rpcsvc_request_t *req) priv = this->private; GF_ASSERT (priv); - if (!xdr_to_generic (req->msg[0], &friend_req, (xdrproc_t)xdr_gd1_mgmt_friend_update)) { + ret = xdr_to_generic (req->msg[0], &friend_req, + (xdrproc_t)xdr_gd1_mgmt_friend_update); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -1603,7 +2471,13 @@ glusterd_handle_friend_update (rpcsvc_request_t *req) gf_log ("", GF_LOG_INFO, "Received uuid: %s, hostname:%s", uuid_buf, hostname); - if (!uuid_compare (uuid, priv->uuid)) { + if (uuid_is_null (uuid)) { + gf_log (this->name, GF_LOG_WARNING, "Updates mustn't " + "contain peer with 'null' uuid"); + continue; + } + + if (!uuid_compare (uuid, MY_UUID)) { gf_log ("", GF_LOG_INFO, "Received my uuid as Friend"); i++; continue; @@ -1622,13 +2496,13 @@ glusterd_handle_friend_update (rpcsvc_request_t *req) ret = glusterd_friend_add (hostname, friend_req.port, GD_FRIEND_STATE_BEFRIENDED, - &uuid, NULL, &peerinfo, 0, &args); + &uuid, &peerinfo, 0, &args); i++; } out: - uuid_copy (rsp.uuid, priv->uuid); + uuid_copy (rsp.uuid, MY_UUID); ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_gd1_mgmt_friend_update_rsp); if (dict) { @@ -1636,8 +2510,7 @@ out: free (friend_req.friends.friends_val);//malloced by xdr dict_unref (dict); } else { - if (friend_req.friends.friends_val) - free (friend_req.friends.friends_val);//malloced by xdr + free (friend_req.friends.friends_val);//malloced by xdr } glusterd_friend_sm (); @@ -1647,7 +2520,14 @@ out: } int -glusterd_handle_probe_query (rpcsvc_request_t *req) +glusterd_handle_friend_update (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_friend_update); +} + +int +__glusterd_handle_probe_query (rpcsvc_request_t *req) { int32_t ret = -1; xlator_t *this = NULL; @@ -1661,7 +2541,9 @@ glusterd_handle_probe_query (rpcsvc_request_t *req) GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &probe_req, (xdrproc_t)xdr_gd1_mgmt_probe_req)) { + ret = xdr_to_generic (req->msg[0], &probe_req, + (xdrproc_t)xdr_gd1_mgmt_probe_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -1678,6 +2560,20 @@ glusterd_handle_probe_query (rpcsvc_request_t *req) gf_log ("glusterd", GF_LOG_INFO, "Received probe from uuid: %s", uuid_utoa (probe_req.uuid)); + /* Check for uuid collision and handle it in a user friendly way by + * sending the error. + */ + if (!uuid_compare (probe_req.uuid, MY_UUID)) { + gf_log (THIS->name, GF_LOG_ERROR, "Peer uuid %s is same as " + "local uuid. Please check the uuid of both the peers " + "from %s/%s", uuid_utoa (probe_req.uuid), + GLUSTERD_DEFAULT_WORKDIR, GLUSTERD_INFO_FILE); + rsp.op_ret = -1; + rsp.op_errno = GF_PROBE_SAME_UUID; + rsp.port = port; + goto respond; + } + ret = glusterd_remote_hostname_get (req, remote_hostname, sizeof (remote_hostname)); if (ret) { @@ -1694,7 +2590,7 @@ glusterd_handle_probe_query (rpcsvc_request_t *req) args.mode = GD_MODE_ON; ret = glusterd_friend_add (remote_hostname, port, GD_FRIEND_STATE_PROBE_RCVD, - NULL, NULL, &peerinfo, 0, &args); + NULL, &peerinfo, 0, &args); if (ret) { gf_log ("", GF_LOG_ERROR, "Failed to add peer %s", remote_hostname); @@ -1702,20 +2598,22 @@ glusterd_handle_probe_query (rpcsvc_request_t *req) } } - uuid_copy (rsp.uuid, conf->uuid); +respond: + uuid_copy (rsp.uuid, MY_UUID); rsp.hostname = probe_req.hostname; + rsp.op_errstr = ""; - ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gd1_mgmt_probe_rsp); + glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_probe_rsp); + ret = 0; gf_log ("glusterd", GF_LOG_INFO, "Responded to %s, op_ret: %d, " - "op_errno: %d, ret: %d", probe_req.hostname, + "op_errno: %d, ret: %d", remote_hostname, rsp.op_ret, rsp.op_errno, ret); out: - if (probe_req.hostname) - free (probe_req.hostname);//malloced by xdr + free (probe_req.hostname);//malloced by xdr glusterd_friend_sm (); glusterd_op_sm (); @@ -1723,8 +2621,13 @@ out: return ret; } +int glusterd_handle_probe_query (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_probe_query); +} + int -glusterd_handle_cli_profile_volume (rpcsvc_request_t *req) +__glusterd_handle_cli_profile_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; @@ -1732,10 +2635,15 @@ glusterd_handle_cli_profile_volume (rpcsvc_request_t *req) glusterd_op_t cli_op = GD_OP_PROFILE_VOLUME; char *volname = NULL; int32_t op = 0; + char err_str[2048] = {0,}; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); - if (!xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -1751,42 +2659,50 @@ glusterd_handle_cli_profile_volume (rpcsvc_request_t *req) ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to get volname"); + snprintf (err_str, sizeof (err_str), "Unable to get volume " + "name"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } - gf_log (THIS->name, GF_LOG_INFO, "Received volume profile req " + gf_log (this->name, GF_LOG_INFO, "Received volume profile req " "for volume %s", volname); ret = dict_get_int32 (dict, "op", &op); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to get op"); + snprintf (err_str, sizeof (err_str), "Unable to get operation"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } - gf_cmd_log ("Volume stats", "volume : %s, op: %d", volname, op); - ret = glusterd_op_begin (req, cli_op, dict); - gf_cmd_log ("Volume stats", " on volume %s, op: %d %s ", - volname, op, - ((ret == 0)? " SUCCEDED":" FAILED")); + ret = glusterd_op_begin (req, cli_op, dict, err_str, sizeof (err_str)); out: glusterd_friend_sm (); glusterd_op_sm (); - if (ret && dict) - dict_unref (dict); - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val); - if (ret) + free (cli_req.dict.dict_val); + + if (ret) { + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, err_str); + } - gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } int -glusterd_handle_getwd (rpcsvc_request_t *req) +glusterd_handle_cli_profile_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_profile_volume); +} + +int +__glusterd_handle_getwd (rpcsvc_request_t *req) { int32_t ret = -1; gf1_cli_getwd_rsp rsp = {0,}; @@ -1801,8 +2717,9 @@ glusterd_handle_getwd (rpcsvc_request_t *req) rsp.wd = priv->workdir; - ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf1_cli_getwd_rsp); + glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf1_cli_getwd_rsp); + ret = 0; glusterd_friend_sm (); glusterd_op_sm (); @@ -1810,18 +2727,27 @@ glusterd_handle_getwd (rpcsvc_request_t *req) return ret; } +int +glusterd_handle_getwd (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_getwd); +} int -glusterd_handle_mount (rpcsvc_request_t *req) +__glusterd_handle_mount (rpcsvc_request_t *req) { gf1_cli_mount_req mnt_req = {0,}; gf1_cli_mount_rsp rsp = {0,}; dict_t *dict = NULL; int ret = 0; + glusterd_conf_t *priv = NULL; GF_ASSERT (req); + priv = THIS->private; - if (!xdr_to_generic (req->msg[0], &mnt_req, (xdrproc_t)xdr_gf1_cli_mount_req)) { + ret = xdr_to_generic (req->msg[0], &mnt_req, + (xdrproc_t)xdr_gf1_cli_mount_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; rsp.op_ret = -1; @@ -1850,15 +2776,18 @@ glusterd_handle_mount (rpcsvc_request_t *req) } } + synclock_unlock (&priv->big_lock); rsp.op_ret = glusterd_do_mount (mnt_req.label, dict, &rsp.path, &rsp.op_errno); + synclock_lock (&priv->big_lock); out: if (!rsp.path) rsp.path = ""; - ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf1_cli_mount_rsp); + glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf1_cli_mount_rsp); + ret = 0; if (dict) dict_unref (dict); @@ -1872,7 +2801,13 @@ glusterd_handle_mount (rpcsvc_request_t *req) } int -glusterd_handle_umount (rpcsvc_request_t *req) +glusterd_handle_mount (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_mount); +} + +int +__glusterd_handle_umount (rpcsvc_request_t *req) { gf1_cli_umount_req umnt_req = {0,}; gf1_cli_umount_rsp rsp = {0,}; @@ -1885,11 +2820,15 @@ glusterd_handle_umount (rpcsvc_request_t *req) gf_boolean_t dir_ok = _gf_false; char *pdir = NULL; char *t = NULL; + glusterd_conf_t *priv = NULL; GF_ASSERT (req); GF_ASSERT (this); + priv = this->private; - if (!xdr_to_generic (req->msg[0], &umnt_req, (xdrproc_t)xdr_gf1_cli_umount_req)) { + ret = xdr_to_generic (req->msg[0], &umnt_req, + (xdrproc_t)xdr_gf1_cli_umount_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; rsp.op_ret = -1; @@ -1928,7 +2867,9 @@ glusterd_handle_umount (rpcsvc_request_t *req) runner_add_args (&runner, "umount", umnt_req.path, NULL); if (umnt_req.lazy) runner_add_arg (&runner, "-l"); + synclock_unlock (&priv->big_lock); rsp.op_ret = runner_run (&runner); + synclock_lock (&priv->big_lock); if (rsp.op_ret == 0) { if (realpath (umnt_req.path, mntp)) rmdir (mntp); @@ -1946,8 +2887,9 @@ glusterd_handle_umount (rpcsvc_request_t *req) if (rsp.op_errno) rsp.op_ret = -1; - ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf1_cli_umount_rsp); + glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf1_cli_umount_rsp); + ret = 0; glusterd_friend_sm (); glusterd_op_sm (); @@ -1956,6 +2898,12 @@ glusterd_handle_umount (rpcsvc_request_t *req) } int +glusterd_handle_umount (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_umount); +} + +int glusterd_friend_remove (uuid_t uuid, char *hostname) { int ret = 0; @@ -1988,8 +2936,9 @@ glusterd_rpc_create (struct rpc_clnt **rpc, GF_ASSERT (this); GF_ASSERT (options); - new_rpc = rpc_clnt_new (options, this->ctx, this->name); + /* TODO: is 32 enough? or more ? */ + new_rpc = rpc_clnt_new (options, this->ctx, this->name, 16); if (!new_rpc) goto out; @@ -2005,7 +2954,7 @@ out: } } - gf_log ("", GF_LOG_DEBUG, "returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret); return ret; } @@ -2028,8 +2977,8 @@ glusterd_transport_keepalive_options_get (int *interval, int *time) } int -glusterd_transport_inet_keepalive_options_build (dict_t **options, - const char *hostname, int port) +glusterd_transport_inet_options_build (dict_t **options, const char *hostname, + int port) { dict_t *dict = NULL; int32_t interval = -1; @@ -2041,10 +2990,25 @@ glusterd_transport_inet_keepalive_options_build (dict_t **options, if (!port) port = GLUSTERD_DEFAULT_PORT; + + /* Build default transport options */ ret = rpc_transport_inet_options_build (&dict, hostname, port); if (ret) goto out; + /* Set frame-timeout to 10mins. Default timeout of 30 mins is too long + * when compared to 2 mins for cli timeout. This ensures users don't + * wait too long after cli timesout before being able to resume normal + * operations + */ + ret = dict_set_int32 (dict, "frame-timeout", 600); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, + "Failed to set frame-timeout"); + goto out; + } + + /* Set keepalive options */ glusterd_transport_keepalive_options_get (&interval, &time); if ((interval > 0) || (time > 0)) @@ -2056,91 +3020,117 @@ out: } int -glusterd_friend_add (const char *hoststr, int port, - glusterd_friend_sm_state_t state, - uuid_t *uuid, - struct rpc_clnt *rpc, - glusterd_peerinfo_t **friend, - gf_boolean_t restore, - glusterd_peerctx_args_t *args) +glusterd_friend_rpc_create (xlator_t *this, glusterd_peerinfo_t *peerinfo, + glusterd_peerctx_args_t *args) { - int ret = 0; - glusterd_conf_t *conf = NULL; - glusterd_peerinfo_t *peerinfo = NULL; - glusterd_peerctx_t *peerctx = NULL; - gf_boolean_t is_allocated = _gf_false; dict_t *options = NULL; - - conf = THIS->private; - GF_ASSERT (conf) - GF_ASSERT (hoststr); + int ret = -1; + glusterd_peerctx_t *peerctx = NULL; + data_t *data = NULL; peerctx = GF_CALLOC (1, sizeof (*peerctx), gf_gld_mt_peerctx_t); - if (!peerctx) { - ret = -1; + if (!peerctx) goto out; - } if (args) peerctx->args = *args; - ret = glusterd_peerinfo_new (&peerinfo, state, uuid, hoststr); + peerctx->peerinfo = peerinfo; + + ret = glusterd_transport_inet_options_build (&options, + peerinfo->hostname, + peerinfo->port); if (ret) goto out; - peerctx->peerinfo = peerinfo; - if (friend) - *friend = peerinfo; - if (!rpc) { - ret = glusterd_transport_inet_keepalive_options_build (&options, - hoststr, port); - if (ret) - goto out; - ret = glusterd_rpc_create (&rpc, options, - glusterd_peer_rpc_notify, - peerctx); - if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "failed to create rpc for" - " peer %s", (char*)hoststr); - goto out; + /* + * For simulated multi-node testing, we need to make sure that we + * create our RPC endpoint with the same address that the peer would + * use to reach us. + */ + if (this->options) { + data = dict_get(this->options,"transport.socket.bind-address"); + if (data) { + ret = dict_set(options, + "transport.socket.source-addr",data); } - is_allocated = _gf_true; } - /* If peer is unreachable when in DEFAULT state, we cleanup peerinfo - * via the friend state machine. ie, peerinfo could have been freed. - * peer_rpc_notify sets peerctx->peerinfo to NULL to indicate the - * same*/ - peerinfo = peerctx->peerinfo; + ret = glusterd_rpc_create (&peerinfo->rpc, options, + glusterd_peer_rpc_notify, peerctx); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to create rpc for" + " peer %s", peerinfo->hostname); + goto out; + } + peerctx = NULL; + ret = 0; +out: + GF_FREE (peerctx); + return ret; +} - if (peerinfo) { - peerinfo->rpc = rpc; +int +glusterd_friend_add (const char *hoststr, int port, + glusterd_friend_sm_state_t state, + uuid_t *uuid, + glusterd_peerinfo_t **friend, + gf_boolean_t restore, + glusterd_peerctx_args_t *args) +{ + int ret = 0; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; - if (!restore) - ret = glusterd_store_peerinfo (peerinfo); + this = THIS; + conf = this->private; + GF_ASSERT (conf); + GF_ASSERT (hoststr); - list_add_tail (&peerinfo->uuid_list, &conf->peers); + ret = glusterd_peerinfo_new (friend, state, uuid, hoststr, port); + if (ret) { + goto out; } -out: - if (ret) { - if (peerctx) - GF_FREE (peerctx); - if (is_allocated && rpc) { - (void) rpc_clnt_unref (rpc); + /* + * We can't add to the list after calling glusterd_friend_rpc_create, + * even if it succeeds, because by then the callback to take it back + * off and free might have happened already (notably in the case of an + * invalid peer name). That would mean we're adding something that had + * just been free, and we're likely to crash later. + */ + list_add_tail (&(*friend)->uuid_list, &conf->peers); + + //restore needs to first create the list of peers, then create rpcs + //to keep track of quorum in race-free manner. In restore for each peer + //rpc-create calls rpc_notify when the friend-list is partially + //constructed, leading to wrong quorum calculations. + if (!restore) { + ret = glusterd_store_peerinfo (*friend); + if (ret == 0) { + synclock_unlock (&conf->big_lock); + ret = glusterd_friend_rpc_create (this, *friend, args); + synclock_lock (&conf->big_lock); } - if (peerinfo) { - peerinfo->rpc = NULL; - (void) glusterd_friend_cleanup (peerinfo); + else { + gf_log (this->name, GF_LOG_ERROR, + "Failed to store peerinfo"); } } - gf_log ("glusterd", GF_LOG_INFO, "connect returned %d", ret); + if (ret) { + (void) glusterd_friend_cleanup (*friend); + *friend = NULL; + } + +out: + gf_log (this->name, GF_LOG_INFO, "connect returned %d", ret); return ret; } int -glusterd_probe_begin (rpcsvc_request_t *req, const char *hoststr, int port) +glusterd_probe_begin (rpcsvc_request_t *req, const char *hoststr, int port, + dict_t *dict) { int ret = -1; glusterd_peerinfo_t *peerinfo = NULL; @@ -2156,9 +3146,10 @@ glusterd_probe_begin (rpcsvc_request_t *req, const char *hoststr, int port) " for host: %s (%d)", hoststr, port); args.mode = GD_MODE_ON; args.req = req; + args.dict = dict; ret = glusterd_friend_add ((char *)hoststr, port, GD_FRIEND_STATE_DEFAULT, - NULL, NULL, &peerinfo, 0, &args); + NULL, &peerinfo, 0, &args); if ((!ret) && (!peerinfo->connected)) { ret = GLUSTERD_CONNECTION_AWAITED; } @@ -2176,11 +3167,12 @@ glusterd_probe_begin (rpcsvc_request_t *req, const char *hoststr, int port) event->peerinfo = peerinfo; ret = glusterd_friend_sm_inject_event (event); glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_SUCCESS, - (char*)hoststr, port); + NULL, (char*)hoststr, + port, dict); } } else { - glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_FRIEND, - (char*)hoststr, port); + glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_FRIEND, NULL, + (char*)hoststr, port, dict); } out: @@ -2190,7 +3182,7 @@ out: int glusterd_deprobe_begin (rpcsvc_request_t *req, const char *hoststr, int port, - uuid_t uuid) + uuid_t uuid, dict_t *dict) { int ret = -1; glusterd_peerinfo_t *peerinfo = NULL; @@ -2231,6 +3223,7 @@ glusterd_deprobe_begin (rpcsvc_request_t *req, const char *hoststr, int port, ctx->hostname = gf_strdup (hoststr); ctx->port = port; ctx->req = req; + ctx->dict = dict; event->ctx = ctx; @@ -2265,7 +3258,7 @@ glusterd_xfer_friend_remove_resp (rpcsvc_request_t *req, char *hostname, int por conf = this->private; - uuid_copy (rsp.uuid, conf->uuid); + uuid_copy (rsp.uuid, MY_UUID); rsp.hostname = hostname; rsp.port = port; ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, @@ -2278,76 +3271,236 @@ glusterd_xfer_friend_remove_resp (rpcsvc_request_t *req, char *hostname, int por int -glusterd_xfer_friend_add_resp (rpcsvc_request_t *req, char *hostname, int port, - int32_t op_ret, int32_t op_errno) +glusterd_xfer_friend_add_resp (rpcsvc_request_t *req, char *myhostname, + char *remote_hostname, int port, int32_t op_ret, + int32_t op_errno) { gd1_mgmt_friend_rsp rsp = {{0}, }; int32_t ret = -1; xlator_t *this = NULL; glusterd_conf_t *conf = NULL; - GF_ASSERT (hostname); + GF_ASSERT (myhostname); this = THIS; GF_ASSERT (this); conf = this->private; - uuid_copy (rsp.uuid, conf->uuid); + uuid_copy (rsp.uuid, MY_UUID); rsp.op_ret = op_ret; rsp.op_errno = op_errno; - rsp.hostname = gf_strdup (hostname); + rsp.hostname = gf_strdup (myhostname); rsp.port = port; ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_gd1_mgmt_friend_rsp); gf_log ("glusterd", GF_LOG_INFO, - "Responded to %s (%d), ret: %d", hostname, port, ret); - if (rsp.hostname) - GF_FREE (rsp.hostname) + "Responded to %s (%d), ret: %d", remote_hostname, port, ret); + GF_FREE (rsp.hostname); return ret; } +static void +set_probe_error_str (int op_ret, int op_errno, char *op_errstr, char *errstr, + size_t len, char *hostname, int port) +{ + if ((op_errstr) && (strcmp (op_errstr, ""))) { + snprintf (errstr, len, "%s", op_errstr); + return; + } + + if (!op_ret) { + switch (op_errno) { + case GF_PROBE_LOCALHOST: + snprintf (errstr, len, "Probe on localhost not " + "needed"); + break; + + case GF_PROBE_FRIEND: + snprintf (errstr, len, "Host %s port %d already" + " in peer list", hostname, port); + break; + + default: + if (op_errno != 0) + snprintf (errstr, len, "Probe returned " + "with unknown errno %d", + op_errno); + break; + } + } else { + switch (op_errno) { + case GF_PROBE_ANOTHER_CLUSTER: + snprintf (errstr, len, "%s is already part of " + "another cluster", hostname); + break; + + case GF_PROBE_VOLUME_CONFLICT: + snprintf (errstr, len, "Atleast one volume on " + "%s conflicts with existing volumes " + "in the cluster", hostname); + break; + + case GF_PROBE_UNKNOWN_PEER: + snprintf (errstr, len, "%s responded with " + "'unknown peer' error, this could " + "happen if %s doesn't have localhost " + "in its peer database", hostname, + hostname); + break; + + case GF_PROBE_ADD_FAILED: + snprintf (errstr, len, "Failed to add peer " + "information on %s", hostname); + break; + + case GF_PROBE_SAME_UUID: + snprintf (errstr, len, "Peer uuid (host %s) is " + "same as local uuid", hostname); + break; + + case GF_PROBE_QUORUM_NOT_MET: + snprintf (errstr, len, "Cluster quorum is not " + "met. Changing peers is not allowed " + "in this state"); + break; + + default: + snprintf (errstr, len, "Probe returned with " + "unknown errno %d", op_errno); + break; + } + } +} + int glusterd_xfer_cli_probe_resp (rpcsvc_request_t *req, int32_t op_ret, - int32_t op_errno, char *hostname, int port) + int32_t op_errno, char *op_errstr, char *hostname, + int port, dict_t *dict) { - gf1_cli_probe_rsp rsp = {0, }; + gf_cli_rsp rsp = {0,}; int32_t ret = -1; + char errstr[2048] = {0,}; + char *cmd_str = NULL; + xlator_t *this = THIS; GF_ASSERT (req); + GF_ASSERT (this); + + (void) set_probe_error_str (op_ret, op_errno, op_errstr, errstr, + sizeof (errstr), hostname, port); + + if (dict) { + ret = dict_get_str (dict, "cmd-str", &cmd_str); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "command string"); + } rsp.op_ret = op_ret; rsp.op_errno = op_errno; - rsp.hostname = hostname; - rsp.port = port; + rsp.op_errstr = (errstr[0] != '\0') ? errstr : ""; + + gf_cmd_log ("", "%s : %s %s %s", cmd_str, + (op_ret) ? "FAILED" : "SUCCESS", + (errstr[0] != '\0') ? ":" : " ", + (errstr[0] != '\0') ? errstr : " "); ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf1_cli_probe_rsp); + (xdrproc_t)xdr_gf_cli_rsp); - gf_log ("glusterd", GF_LOG_INFO, "Responded to CLI, ret: %d",ret); + if (dict) + dict_unref (dict); + gf_log (this->name, GF_LOG_DEBUG, "Responded to CLI, ret: %d",ret); return ret; } +static void +set_deprobe_error_str (int op_ret, int op_errno, char *op_errstr, char *errstr, + size_t len, char *hostname) +{ + if ((op_errstr) && (strcmp (op_errstr, ""))) { + snprintf (errstr, len, "%s", op_errstr); + return; + } + + if (op_ret) { + switch (op_errno) { + case GF_DEPROBE_LOCALHOST: + snprintf (errstr, len, "%s is localhost", + hostname); + break; + + case GF_DEPROBE_NOT_FRIEND: + snprintf (errstr, len, "%s is not part of " + "cluster", hostname); + break; + + case GF_DEPROBE_BRICK_EXIST: + snprintf (errstr, len, "Brick(s) with the peer " + "%s exist in cluster", hostname); + break; + + case GF_DEPROBE_FRIEND_DOWN: + snprintf (errstr, len, "One of the peers is " + "probably down. Check with " + "'peer status'"); + break; + + case GF_DEPROBE_QUORUM_NOT_MET: + snprintf (errstr, len, "Cluster quorum is not " + "met. Changing peers is not allowed " + "in this state"); + break; + + default: + snprintf (errstr, len, "Detach returned with " + "unknown errno %d", op_errno); + break; + + } + } +} + + int glusterd_xfer_cli_deprobe_resp (rpcsvc_request_t *req, int32_t op_ret, - int32_t op_errno, char *hostname) + int32_t op_errno, char *op_errstr, + char *hostname, dict_t *dict) { - gf1_cli_deprobe_rsp rsp = {0, }; + gf_cli_rsp rsp = {0,}; int32_t ret = -1; + char *cmd_str = NULL; + char errstr[2048] = {0,}; GF_ASSERT (req); + (void) set_deprobe_error_str (op_ret, op_errno, op_errstr, errstr, + sizeof (errstr), hostname); + + if (dict) { + ret = dict_get_str (dict, "cmd-str", &cmd_str); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, "Failed to get " + "command string"); + } + rsp.op_ret = op_ret; rsp.op_errno = op_errno; - rsp.hostname = hostname; + rsp.op_errstr = (errstr[0] != '\0') ? errstr : ""; + + gf_cmd_log ("", "%s : %s %s %s", cmd_str, + (op_ret) ? "FAILED" : "SUCCESS", + (errstr[0] != '\0') ? ":" : " ", + (errstr[0] != '\0') ? errstr : " "); ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf1_cli_deprobe_rsp); + (xdrproc_t)xdr_gf_cli_rsp); - gf_log ("glusterd", GF_LOG_INFO, "Responded to CLI, ret: %d",ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Responded to CLI, ret: %d",ret); return ret; } @@ -2361,39 +3514,52 @@ glusterd_list_friends (rpcsvc_request_t *req, dict_t *dict, int32_t flags) int32_t count = 0; dict_t *friends = NULL; gf1_cli_peer_list_rsp rsp = {0,}; + char my_uuid_str[64] = {0,}; + char key[256] = {0,}; priv = THIS->private; GF_ASSERT (priv); - if (!list_empty (&priv->peers)) { - friends = dict_new (); - if (!friends) { - gf_log ("", GF_LOG_WARNING, "Out of Memory"); - goto out; - } - } else { - ret = 0; + friends = dict_new (); + if (!friends) { + gf_log ("", GF_LOG_WARNING, "Out of Memory"); goto out; } - - if (flags == GF_CLI_LIST_ALL) { - list_for_each_entry (entry, &priv->peers, uuid_list) { - count++; - ret = glusterd_add_peer_detail_to_dict (entry, + if (!list_empty (&priv->peers)) { + list_for_each_entry (entry, &priv->peers, uuid_list) { + count++; + ret = glusterd_add_peer_detail_to_dict (entry, friends, count); - if (ret) - goto out; + if (ret) + goto out; + } + } - } + if (flags == GF_CLI_LIST_POOL_NODES) { + count++; + snprintf (key, 256, "friend%d.uuid", count); + uuid_utoa_r (MY_UUID, my_uuid_str); + ret = dict_set_str (friends, key, my_uuid_str); + if (ret) + goto out; - ret = dict_set_int32 (friends, "count", count); + snprintf (key, 256, "friend%d.hostname", count); + ret = dict_set_str (friends, key, "localhost"); + if (ret) + goto out; - if (ret) - goto out; + snprintf (key, 256, "friend%d.connected", count); + ret = dict_set_int32 (friends, key, 1); + if (ret) + goto out; } + ret = dict_set_int32 (friends, "count", count); + if (ret) + goto out; + ret = dict_allocate_and_serialize (friends, &rsp.friends.friends_val, - (size_t *)&rsp.friends.friends_len); + &rsp.friends.friends_len); if (ret) goto out; @@ -2406,10 +3572,10 @@ out: rsp.op_ret = ret; - ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf1_cli_peer_list_rsp); - if (rsp.friends.friends_val) - GF_FREE (rsp.friends.friends_val); + glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf1_cli_peer_list_rsp); + ret = 0; + GF_FREE (rsp.friends.friends_val); return ret; } @@ -2500,7 +3666,7 @@ respond: if (ret) goto out; ret = dict_allocate_and_serialize (volumes, &rsp.dict.dict_val, - (size_t *)&rsp.dict.dict_len); + &rsp.dict.dict_len); if (ret) goto out; @@ -2510,19 +3676,19 @@ out: rsp.op_ret = ret; rsp.op_errstr = ""; - ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf_cli_rsp); + glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_cli_rsp); + ret = 0; if (volumes) dict_unref (volumes); - if (rsp.dict.dict_val) - GF_FREE (rsp.dict.dict_val); + GF_FREE (rsp.dict.dict_val); return ret; } int -glusterd_handle_status_volume (rpcsvc_request_t *req) +__glusterd_handle_status_volume (rpcsvc_request_t *req) { int32_t ret = -1; uint32_t cmd = 0; @@ -2530,11 +3696,15 @@ glusterd_handle_status_volume (rpcsvc_request_t *req) char *volname = 0; gf_cli_req cli_req = {{0,}}; glusterd_op_t cli_op = GD_OP_STATUS_VOLUME; + char err_str[2048] = {0,}; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -2547,8 +3717,10 @@ glusterd_handle_status_volume (rpcsvc_request_t *req) ret = dict_unserialize (cli_req.dict.dict_val, cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to " + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize buffer"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); goto out; } @@ -2561,46 +3733,172 @@ glusterd_handle_status_volume (rpcsvc_request_t *req) if (!(cmd & GF_CLI_STATUS_ALL)) { ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, - "failed to get volname"); + snprintf (err_str, sizeof (err_str), "Unable to get " + "volume name"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } - gf_log (THIS->name, GF_LOG_INFO, - "Received status volume req " - "for volume %s", volname); + gf_log (this->name, GF_LOG_INFO, + "Received status volume req for volume %s", volname); } - ret = glusterd_op_begin (req, GD_OP_STATUS_VOLUME, dict); + ret = glusterd_op_begin_synctask (req, GD_OP_STATUS_VOLUME, dict); out: - if (ret && dict) - dict_unref (dict); - glusterd_friend_sm (); - glusterd_op_sm (); + if (ret) { + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); + ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, + dict, err_str); + } + free (cli_req.dict.dict_val); - if (ret) + return ret; +} + +int +glusterd_handle_status_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_status_volume); +} + +int +__glusterd_handle_cli_clearlocks_volume (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gf_cli_req cli_req = {{0,}}; + glusterd_op_t cli_op = GD_OP_CLEARLOCKS_VOLUME; + char *volname = NULL; + dict_t *dict = NULL; + char err_str[2048] = {0,}; + xlator_t *this = NULL; + + GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); + + ret = -1; + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (cli_req.dict.dict_len) { + dict = dict_new (); + + ret = dict_unserialize (cli_req.dict.dict_val, + cli_req.dict.dict_len, + &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to unserialize req-buffer to" + " dictionary"); + snprintf (err_str, sizeof (err_str), "unable to decode " + "the command"); + goto out; + } + + } else { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Empty cli request."); + goto out; + } + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + snprintf (err_str, sizeof (err_str), "Unable to get volume " + "name"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + goto out; + } + + gf_log (this->name, GF_LOG_INFO, "Received clear-locks volume req " + "for volume %s", volname); + + ret = glusterd_op_begin_synctask (req, GD_OP_CLEARLOCKS_VOLUME, dict); + +out: + if (ret) { + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val); + dict, err_str); + } + free (cli_req.dict.dict_val); return ret; } int -glusterd_brick_rpc_notify (struct rpc_clnt *rpc, void *mydata, - rpc_clnt_event_t event, - void *data) +glusterd_handle_cli_clearlocks_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_clearlocks_volume); +} + +static int +get_brickinfo_from_brickid (char *brickid, glusterd_brickinfo_t **brickinfo) +{ + glusterd_volinfo_t *volinfo = NULL; + char *volid_str = NULL; + char *brick = NULL; + char *brickid_dup = NULL; + uuid_t volid = {0}; + int ret = -1; + + brickid_dup = gf_strdup (brickid); + if (!brickid_dup) + goto out; + + volid_str = brickid_dup; + brick = strchr (brickid_dup, ':'); + *brick = '\0'; + brick++; + if (!volid_str || !brick) + goto out; + + uuid_parse (volid_str, volid); + ret = glusterd_volinfo_find_by_volume_id (volid, &volinfo); + if (ret) { + /* Check if it a snapshot volume */ + ret = glusterd_snap_volinfo_find_by_volume_id (volid, &volinfo); + if (ret) + goto out; + } + + ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, + brickinfo); + if (ret) + goto out; + + ret = 0; +out: + GF_FREE (brickid_dup); + return ret; +} + +int +__glusterd_brick_rpc_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, void *data) { xlator_t *this = NULL; glusterd_conf_t *conf = NULL; int ret = 0; + char *brickid = NULL; glusterd_brickinfo_t *brickinfo = NULL; - brickinfo = mydata; - if (!brickinfo) + brickid = mydata; + if (!brickid) + return 0; + + ret = get_brickinfo_from_brickid (brickid, &brickinfo); + if (ret) return 0; this = THIS; @@ -2610,15 +3908,21 @@ glusterd_brick_rpc_notify (struct rpc_clnt *rpc, void *mydata, switch (event) { case RPC_CLNT_CONNECT: - gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_CONNECT"); + gf_log (this->name, GF_LOG_DEBUG, "Connected to %s:%s", + brickinfo->hostname, brickinfo->path); glusterd_set_brick_status (brickinfo, GF_BRICK_STARTED); ret = default_notify (this, GF_EVENT_CHILD_UP, NULL); break; case RPC_CLNT_DISCONNECT: - gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_DISCONNECT"); + if (GF_BRICK_STARTED == brickinfo->status) + gf_log (this->name, GF_LOG_INFO, "Disconnected from " + "%s:%s", brickinfo->hostname, brickinfo->path); + glusterd_set_brick_status (brickinfo, GF_BRICK_STOPPED); + if (rpc_clnt_is_disabled (rpc)) + GF_FREE (brickid); break; default: @@ -2631,12 +3935,20 @@ glusterd_brick_rpc_notify (struct rpc_clnt *rpc, void *mydata, } int -glusterd_shd_rpc_notify (struct rpc_clnt *rpc, void *mydata, - rpc_clnt_event_t event, - void *data) +glusterd_brick_rpc_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, void *data) +{ + return glusterd_big_locked_notify (rpc, mydata, event, data, + __glusterd_brick_rpc_notify); +} + +int +__glusterd_nodesvc_rpc_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, void *data) { xlator_t *this = NULL; glusterd_conf_t *conf = NULL; + char *server = NULL; int ret = 0; this = THIS; @@ -2644,17 +3956,21 @@ glusterd_shd_rpc_notify (struct rpc_clnt *rpc, void *mydata, conf = this->private; GF_ASSERT (conf); + server = mydata; + if (!server) + return 0; + switch (event) { case RPC_CLNT_CONNECT: gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_CONNECT"); - (void) glusterd_shd_set_running (_gf_true); + (void) glusterd_nodesvc_set_online_status (server, _gf_true); ret = default_notify (this, GF_EVENT_CHILD_UP, NULL); break; case RPC_CLNT_DISCONNECT: gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_DISCONNECT"); - (void) glusterd_shd_set_running (_gf_false); + (void) glusterd_nodesvc_set_online_status (server, _gf_false); break; default: @@ -2667,10 +3983,29 @@ glusterd_shd_rpc_notify (struct rpc_clnt *rpc, void *mydata, } int -glusterd_friend_remove_notify (glusterd_peerinfo_t *peerinfo, rpcsvc_request_t *req) +glusterd_nodesvc_rpc_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, void *data) +{ + return glusterd_big_locked_notify (rpc, mydata, event, data, + __glusterd_nodesvc_rpc_notify); +} + +int +glusterd_friend_remove_notify (glusterd_peerctx_t *peerctx) { - int ret = -1; - glusterd_friend_sm_event_t *new_event = NULL; + int ret = -1; + glusterd_friend_sm_event_t *new_event = NULL; + glusterd_peerinfo_t *peerinfo = peerctx->peerinfo; + rpcsvc_request_t *req = peerctx->args.req; + char *errstr = peerctx->errstr; + dict_t *dict = NULL; + + GF_ASSERT (peerctx); + + peerinfo = peerctx->peerinfo; + req = peerctx->args.req; + dict = peerctx->args.dict; + errstr = peerctx->errstr; ret = glusterd_friend_sm_new_event (GD_FRIEND_EVENT_REMOVE_FRIEND, &new_event); @@ -2682,12 +4017,12 @@ glusterd_friend_remove_notify (glusterd_peerinfo_t *peerinfo, rpcsvc_request_t * goto out; } - glusterd_xfer_cli_probe_resp (req, -1, ENOTCONN, - peerinfo->hostname, peerinfo->port); + glusterd_xfer_cli_probe_resp (req, -1, ENOTCONN, errstr, + peerinfo->hostname, + peerinfo->port, dict); new_event->peerinfo = peerinfo; ret = glusterd_friend_sm_inject_event (new_event); - glusterd_friend_sm (); } else { gf_log ("glusterd", GF_LOG_ERROR, @@ -2700,17 +4035,16 @@ out: } int -glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata, - rpc_clnt_event_t event, - void *data) +__glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, void *data) { xlator_t *this = NULL; glusterd_conf_t *conf = NULL; int ret = 0; glusterd_peerinfo_t *peerinfo = NULL; glusterd_peerctx_t *peerctx = NULL; - uuid_t owner = {0,}; - uuid_t *peer_uuid = NULL; + gf_boolean_t quorum_action = _gf_false; + glusterd_volinfo_t *volinfo = NULL; peerctx = mydata; if (!peerctx) @@ -2725,8 +4059,9 @@ glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata, { gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_CONNECT"); peerinfo->connected = 1; + peerinfo->quorum_action = _gf_true; - ret = glusterd_peer_handshake (this, rpc, peerctx); + ret = glusterd_peer_dump_version (this, rpc, peerctx); if (ret) gf_log ("", GF_LOG_ERROR, "glusterd handshake failed"); break; @@ -2737,48 +4072,36 @@ glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata, gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_DISCONNECT %d", peerinfo->state.state); - peerinfo->connected = 0; + if (peerinfo->connected) { + list_for_each_entry (volinfo, &conf->volumes, vol_list) { + ret = glusterd_mgmt_v3_unlock (volinfo->volname, + peerinfo->uuid, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_TRACE, + "Lock not released for %s", + volinfo->volname); + } - /* - local glusterd (thinks that it) is the owner of the cluster - lock and 'fails' the operation on the first disconnect from - a peer. - */ - glusterd_get_lock_owner (&owner); - if (!uuid_compare (conf->uuid, owner)) { - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_START_UNLOCK, - NULL); - if (ret) - gf_log (this->name, GF_LOG_ERROR, "Unable" - " to enqueue cluster unlock event"); - break; + ret = 0; } - peer_uuid = GF_CALLOC (1, sizeof (*peer_uuid), gf_common_mt_char); - if (!peer_uuid) { - ret = -1; - break; + if ((peerinfo->quorum_contrib != QUORUM_DOWN) && + (peerinfo->state.state == GD_FRIEND_STATE_BEFRIENDED)) { + peerinfo->quorum_contrib = QUORUM_DOWN; + quorum_action = _gf_true; + peerinfo->quorum_action = _gf_false; } - uuid_copy (*peer_uuid, peerinfo->uuid); - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_LOCAL_UNLOCK_NO_RESP, - peer_uuid); - if (ret) - gf_log (this->name, GF_LOG_ERROR, "Unable" - " to enque local lock flush event."); - - //Inject friend disconnected here - if (peerinfo->state.state == GD_FRIEND_STATE_DEFAULT) { - /* Remove the friend as it was the newly requested - 'peer' and connection with this peer didn't - succeed. we have opportunity to notify user - */ - peerctx->peerinfo = NULL; - glusterd_friend_remove_notify (peerinfo, - peerctx->args.req); + /* Remove peer if it is not a friend and connection/handshake + * fails, and notify cli. Happens only during probe. + */ + if (peerinfo->state.state == GD_FRIEND_STATE_DEFAULT) { + glusterd_friend_remove_notify (peerctx); + goto out; } - //default_notify (this, GF_EVENT_CHILD_DOWN, NULL); + peerinfo->connected = 0; break; } default: @@ -2788,12 +4111,23 @@ glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata, break; } +out: glusterd_friend_sm (); glusterd_op_sm (); + if (quorum_action) + glusterd_do_quorum_action (); return ret; } int +glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, void *data) +{ + return glusterd_big_locked_notify (rpc, mydata, event, data, + __glusterd_peer_rpc_notify); +} + +int glusterd_null (rpcsvc_request_t *req) { @@ -2801,11 +4135,11 @@ glusterd_null (rpcsvc_request_t *req) } rpcsvc_actor_t gd_svc_mgmt_actors[] = { - [GLUSTERD_MGMT_NULL] = { "NULL", GLUSTERD_MGMT_NULL, glusterd_null, NULL, NULL, 0}, - [GLUSTERD_MGMT_CLUSTER_LOCK] = { "CLUSTER_LOCK", GLUSTERD_MGMT_CLUSTER_LOCK, glusterd_handle_cluster_lock, NULL, NULL, 0}, - [GLUSTERD_MGMT_CLUSTER_UNLOCK] = { "CLUSTER_UNLOCK", GLUSTERD_MGMT_CLUSTER_UNLOCK, glusterd_handle_cluster_unlock, NULL, NULL, 0}, - [GLUSTERD_MGMT_STAGE_OP] = { "STAGE_OP", GLUSTERD_MGMT_STAGE_OP, glusterd_handle_stage_op, NULL, NULL, 0}, - [GLUSTERD_MGMT_COMMIT_OP] = { "COMMIT_OP", GLUSTERD_MGMT_COMMIT_OP, glusterd_handle_commit_op, NULL, NULL, 0}, + [GLUSTERD_MGMT_NULL] = { "NULL", GLUSTERD_MGMT_NULL, glusterd_null, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_CLUSTER_LOCK] = { "CLUSTER_LOCK", GLUSTERD_MGMT_CLUSTER_LOCK, glusterd_handle_cluster_lock, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_CLUSTER_UNLOCK] = { "CLUSTER_UNLOCK", GLUSTERD_MGMT_CLUSTER_UNLOCK, glusterd_handle_cluster_unlock, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_STAGE_OP] = { "STAGE_OP", GLUSTERD_MGMT_STAGE_OP, glusterd_handle_stage_op, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_COMMIT_OP] = { "COMMIT_OP", GLUSTERD_MGMT_COMMIT_OP, glusterd_handle_commit_op, NULL, 0, DRC_NA}, }; struct rpcsvc_program gd_svc_mgmt_prog = { @@ -2814,14 +4148,15 @@ struct rpcsvc_program gd_svc_mgmt_prog = { .progver = GD_MGMT_VERSION, .numactors = GLUSTERD_MGMT_MAXVALUE, .actors = gd_svc_mgmt_actors, + .synctask = _gf_true, }; rpcsvc_actor_t gd_svc_peer_actors[] = { - [GLUSTERD_FRIEND_NULL] = { "NULL", GLUSTERD_MGMT_NULL, glusterd_null, NULL, NULL, 0}, - [GLUSTERD_PROBE_QUERY] = { "PROBE_QUERY", GLUSTERD_PROBE_QUERY, glusterd_handle_probe_query, NULL, NULL, 0}, - [GLUSTERD_FRIEND_ADD] = { "FRIEND_ADD", GLUSTERD_FRIEND_ADD, glusterd_handle_incoming_friend_req, NULL, NULL, 0}, - [GLUSTERD_FRIEND_REMOVE] = { "FRIEND_REMOVE", GLUSTERD_FRIEND_REMOVE, glusterd_handle_incoming_unfriend_req, NULL, NULL, 0}, - [GLUSTERD_FRIEND_UPDATE] = { "FRIEND_UPDATE", GLUSTERD_FRIEND_UPDATE, glusterd_handle_friend_update, NULL, NULL, 0}, + [GLUSTERD_FRIEND_NULL] = { "NULL", GLUSTERD_MGMT_NULL, glusterd_null, NULL, 0, DRC_NA}, + [GLUSTERD_PROBE_QUERY] = { "PROBE_QUERY", GLUSTERD_PROBE_QUERY, glusterd_handle_probe_query, NULL, 0, DRC_NA}, + [GLUSTERD_FRIEND_ADD] = { "FRIEND_ADD", GLUSTERD_FRIEND_ADD, glusterd_handle_incoming_friend_req, NULL, 0, DRC_NA}, + [GLUSTERD_FRIEND_REMOVE] = { "FRIEND_REMOVE", GLUSTERD_FRIEND_REMOVE, glusterd_handle_incoming_unfriend_req, NULL, 0, DRC_NA}, + [GLUSTERD_FRIEND_UPDATE] = { "FRIEND_UPDATE", GLUSTERD_FRIEND_UPDATE, glusterd_handle_friend_update, NULL, 0, DRC_NA}, }; struct rpcsvc_program gd_svc_peer_prog = { @@ -2830,37 +4165,45 @@ struct rpcsvc_program gd_svc_peer_prog = { .progver = GD_FRIEND_VERSION, .numactors = GLUSTERD_FRIEND_MAXVALUE, .actors = gd_svc_peer_actors, + .synctask = _gf_false, }; rpcsvc_actor_t gd_svc_cli_actors[] = { - [GLUSTER_CLI_PROBE] = { "CLI_PROBE", GLUSTER_CLI_PROBE, glusterd_handle_cli_probe, NULL, NULL, 0}, - [GLUSTER_CLI_CREATE_VOLUME] = { "CLI_CREATE_VOLUME", GLUSTER_CLI_CREATE_VOLUME, glusterd_handle_create_volume, NULL,NULL, 0}, - [GLUSTER_CLI_DEFRAG_VOLUME] = { "CLI_DEFRAG_VOLUME", GLUSTER_CLI_DEFRAG_VOLUME, glusterd_handle_defrag_volume, NULL,NULL, 0}, - [GLUSTER_CLI_DEPROBE] = { "FRIEND_REMOVE", GLUSTER_CLI_DEPROBE, glusterd_handle_cli_deprobe, NULL, NULL, 0}, - [GLUSTER_CLI_LIST_FRIENDS] = { "LIST_FRIENDS", GLUSTER_CLI_LIST_FRIENDS, glusterd_handle_cli_list_friends, NULL, NULL, 0}, - [GLUSTER_CLI_START_VOLUME] = { "START_VOLUME", GLUSTER_CLI_START_VOLUME, glusterd_handle_cli_start_volume, NULL, NULL, 0}, - [GLUSTER_CLI_STOP_VOLUME] = { "STOP_VOLUME", GLUSTER_CLI_STOP_VOLUME, glusterd_handle_cli_stop_volume, NULL, NULL, 0}, - [GLUSTER_CLI_DELETE_VOLUME] = { "DELETE_VOLUME", GLUSTER_CLI_DELETE_VOLUME, glusterd_handle_cli_delete_volume, NULL, NULL, 0}, - [GLUSTER_CLI_GET_VOLUME] = { "GET_VOLUME", GLUSTER_CLI_GET_VOLUME, glusterd_handle_cli_get_volume, NULL, NULL, 0}, - [GLUSTER_CLI_ADD_BRICK] = { "ADD_BRICK", GLUSTER_CLI_ADD_BRICK, glusterd_handle_add_brick, NULL, NULL, 0}, - [GLUSTER_CLI_REPLACE_BRICK] = { "REPLACE_BRICK", GLUSTER_CLI_REPLACE_BRICK, glusterd_handle_replace_brick, NULL, NULL, 0}, - [GLUSTER_CLI_REMOVE_BRICK] = { "REMOVE_BRICK", GLUSTER_CLI_REMOVE_BRICK, glusterd_handle_remove_brick, NULL, NULL, 0}, - [GLUSTER_CLI_LOG_ROTATE] = { "LOG FILENAME", GLUSTER_CLI_LOG_ROTATE, glusterd_handle_log_rotate, NULL, NULL, 0}, - [GLUSTER_CLI_SET_VOLUME] = { "SET_VOLUME", GLUSTER_CLI_SET_VOLUME, glusterd_handle_set_volume, NULL, NULL, 0}, - [GLUSTER_CLI_SYNC_VOLUME] = { "SYNC_VOLUME", GLUSTER_CLI_SYNC_VOLUME, glusterd_handle_sync_volume, NULL, NULL, 0}, - [GLUSTER_CLI_RESET_VOLUME] = { "RESET_VOLUME", GLUSTER_CLI_RESET_VOLUME, glusterd_handle_reset_volume, NULL, NULL, 0}, - [GLUSTER_CLI_FSM_LOG] = { "FSM_LOG", GLUSTER_CLI_FSM_LOG, glusterd_handle_fsm_log, NULL, NULL, 0}, - [GLUSTER_CLI_GSYNC_SET] = { "GSYNC_SET", GLUSTER_CLI_GSYNC_SET, glusterd_handle_gsync_set, NULL, NULL, 0}, - [GLUSTER_CLI_PROFILE_VOLUME] = { "STATS_VOLUME", GLUSTER_CLI_PROFILE_VOLUME, glusterd_handle_cli_profile_volume, NULL, NULL, 0}, - [GLUSTER_CLI_QUOTA] = { "QUOTA", GLUSTER_CLI_QUOTA, glusterd_handle_quota, NULL, NULL, 0}, - [GLUSTER_CLI_GETWD] = { "GETWD", GLUSTER_CLI_GETWD, glusterd_handle_getwd, NULL, NULL, 1}, - [GLUSTER_CLI_STATUS_VOLUME] = {"STATUS_VOLUME", GLUSTER_CLI_STATUS_VOLUME, glusterd_handle_status_volume, NULL, NULL, 0}, - [GLUSTER_CLI_MOUNT] = { "MOUNT", GLUSTER_CLI_MOUNT, glusterd_handle_mount, NULL, NULL, 1}, - [GLUSTER_CLI_UMOUNT] = { "UMOUNT", GLUSTER_CLI_UMOUNT, glusterd_handle_umount, NULL, NULL, 1}, - [GLUSTER_CLI_HEAL_VOLUME] = { "HEAL_VOLUME", GLUSTER_CLI_HEAL_VOLUME, glusterd_handle_cli_heal_volume, NULL, NULL, 0}, - [GLUSTER_CLI_STATEDUMP_VOLUME] = {"STATEDUMP_VOLUME", GLUSTER_CLI_STATEDUMP_VOLUME, glusterd_handle_cli_statedump_volume, NULL, NULL, 0}, + [GLUSTER_CLI_PROBE] = { "CLI_PROBE", GLUSTER_CLI_PROBE, glusterd_handle_cli_probe, NULL, 0, DRC_NA}, + [GLUSTER_CLI_CREATE_VOLUME] = { "CLI_CREATE_VOLUME", GLUSTER_CLI_CREATE_VOLUME, glusterd_handle_create_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_DEFRAG_VOLUME] = { "CLI_DEFRAG_VOLUME", GLUSTER_CLI_DEFRAG_VOLUME, glusterd_handle_defrag_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_DEPROBE] = { "FRIEND_REMOVE", GLUSTER_CLI_DEPROBE, glusterd_handle_cli_deprobe, NULL, 0, DRC_NA}, + [GLUSTER_CLI_LIST_FRIENDS] = { "LIST_FRIENDS", GLUSTER_CLI_LIST_FRIENDS, glusterd_handle_cli_list_friends, NULL, 0, DRC_NA}, + [GLUSTER_CLI_UUID_RESET] = { "UUID_RESET", GLUSTER_CLI_UUID_RESET, glusterd_handle_cli_uuid_reset, NULL, 0, DRC_NA}, + [GLUSTER_CLI_UUID_GET] = { "UUID_GET", GLUSTER_CLI_UUID_GET, glusterd_handle_cli_uuid_get, NULL, 0, DRC_NA}, + [GLUSTER_CLI_START_VOLUME] = { "START_VOLUME", GLUSTER_CLI_START_VOLUME, glusterd_handle_cli_start_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_STOP_VOLUME] = { "STOP_VOLUME", GLUSTER_CLI_STOP_VOLUME, glusterd_handle_cli_stop_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_DELETE_VOLUME] = { "DELETE_VOLUME", GLUSTER_CLI_DELETE_VOLUME, glusterd_handle_cli_delete_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_GET_VOLUME] = { "GET_VOLUME", GLUSTER_CLI_GET_VOLUME, glusterd_handle_cli_get_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_ADD_BRICK] = { "ADD_BRICK", GLUSTER_CLI_ADD_BRICK, glusterd_handle_add_brick, NULL, 0, DRC_NA}, + [GLUSTER_CLI_REPLACE_BRICK] = { "REPLACE_BRICK", GLUSTER_CLI_REPLACE_BRICK, glusterd_handle_replace_brick, NULL, 0, DRC_NA}, + [GLUSTER_CLI_REMOVE_BRICK] = { "REMOVE_BRICK", GLUSTER_CLI_REMOVE_BRICK, glusterd_handle_remove_brick, NULL, 0, DRC_NA}, + [GLUSTER_CLI_LOG_ROTATE] = { "LOG FILENAME", GLUSTER_CLI_LOG_ROTATE, glusterd_handle_log_rotate, NULL, 0, DRC_NA}, + [GLUSTER_CLI_SET_VOLUME] = { "SET_VOLUME", GLUSTER_CLI_SET_VOLUME, glusterd_handle_set_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_SYNC_VOLUME] = { "SYNC_VOLUME", GLUSTER_CLI_SYNC_VOLUME, glusterd_handle_sync_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_RESET_VOLUME] = { "RESET_VOLUME", GLUSTER_CLI_RESET_VOLUME, glusterd_handle_reset_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_FSM_LOG] = { "FSM_LOG", GLUSTER_CLI_FSM_LOG, glusterd_handle_fsm_log, NULL, 0, DRC_NA}, + [GLUSTER_CLI_GSYNC_SET] = { "GSYNC_SET", GLUSTER_CLI_GSYNC_SET, glusterd_handle_gsync_set, NULL, 0, DRC_NA}, + [GLUSTER_CLI_PROFILE_VOLUME] = { "STATS_VOLUME", GLUSTER_CLI_PROFILE_VOLUME, glusterd_handle_cli_profile_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_QUOTA] = { "QUOTA", GLUSTER_CLI_QUOTA, glusterd_handle_quota, NULL, 0, DRC_NA}, + [GLUSTER_CLI_GETWD] = { "GETWD", GLUSTER_CLI_GETWD, glusterd_handle_getwd, NULL, 1, DRC_NA}, + [GLUSTER_CLI_STATUS_VOLUME] = {"STATUS_VOLUME", GLUSTER_CLI_STATUS_VOLUME, glusterd_handle_status_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_MOUNT] = { "MOUNT", GLUSTER_CLI_MOUNT, glusterd_handle_mount, NULL, 1, DRC_NA}, + [GLUSTER_CLI_UMOUNT] = { "UMOUNT", GLUSTER_CLI_UMOUNT, glusterd_handle_umount, NULL, 1, DRC_NA}, + [GLUSTER_CLI_HEAL_VOLUME] = { "HEAL_VOLUME", GLUSTER_CLI_HEAL_VOLUME, glusterd_handle_cli_heal_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_STATEDUMP_VOLUME] = {"STATEDUMP_VOLUME", GLUSTER_CLI_STATEDUMP_VOLUME, glusterd_handle_cli_statedump_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_LIST_VOLUME] = {"LIST_VOLUME", GLUSTER_CLI_LIST_VOLUME, glusterd_handle_cli_list_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_CLRLOCKS_VOLUME] = {"CLEARLOCKS_VOLUME", GLUSTER_CLI_CLRLOCKS_VOLUME, glusterd_handle_cli_clearlocks_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_COPY_FILE] = {"COPY_FILE", GLUSTER_CLI_COPY_FILE, glusterd_handle_copy_file, NULL, 0, DRC_NA}, + [GLUSTER_CLI_SYS_EXEC] = {"SYS_EXEC", GLUSTER_CLI_SYS_EXEC, glusterd_handle_sys_exec, NULL, 0, DRC_NA}, + [GLUSTER_CLI_SNAP] = {"SNAP", GLUSTER_CLI_SNAP, glusterd_handle_snapshot, NULL, 0, DRC_NA}, }; struct rpcsvc_program gd_svc_cli_prog = { @@ -2869,4 +4212,26 @@ struct rpcsvc_program gd_svc_cli_prog = { .progver = GLUSTER_CLI_VERSION, .numactors = GLUSTER_CLI_MAXVALUE, .actors = gd_svc_cli_actors, + .synctask = _gf_true, +}; + +/* This is a minimal RPC prog, which contains only the readonly RPC procs from + * the cli rpcsvc + */ +rpcsvc_actor_t gd_svc_cli_actors_ro[] = { + [GLUSTER_CLI_LIST_FRIENDS] = { "LIST_FRIENDS", GLUSTER_CLI_LIST_FRIENDS, glusterd_handle_cli_list_friends, NULL, 0, DRC_NA}, + [GLUSTER_CLI_UUID_GET] = { "UUID_GET", GLUSTER_CLI_UUID_GET, glusterd_handle_cli_uuid_get, NULL, 0, DRC_NA}, + [GLUSTER_CLI_GET_VOLUME] = { "GET_VOLUME", GLUSTER_CLI_GET_VOLUME, glusterd_handle_cli_get_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_GETWD] = { "GETWD", GLUSTER_CLI_GETWD, glusterd_handle_getwd, NULL, 1, DRC_NA}, + [GLUSTER_CLI_STATUS_VOLUME] = {"STATUS_VOLUME", GLUSTER_CLI_STATUS_VOLUME, glusterd_handle_status_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_LIST_VOLUME] = {"LIST_VOLUME", GLUSTER_CLI_LIST_VOLUME, glusterd_handle_cli_list_volume, NULL, 0, DRC_NA}, +}; + +struct rpcsvc_program gd_svc_cli_prog_ro = { + .progname = "GlusterD svc cli read-only", + .prognum = GLUSTER_CLI_PROGRAM, + .progver = GLUSTER_CLI_VERSION, + .numactors = GLUSTER_CLI_MAXVALUE, + .actors = gd_svc_cli_actors_ro, + .synctask = _gf_true, }; diff --git a/xlators/mgmt/glusterd/src/glusterd-handshake.c b/xlators/mgmt/glusterd/src/glusterd-handshake.c index 04e5fae7b..0f0357c4c 100644 --- a/xlators/mgmt/glusterd/src/glusterd-handshake.c +++ b/xlators/mgmt/glusterd/src/glusterd-handshake.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H @@ -31,6 +21,7 @@ #include "glusterd.h" #include "glusterd-utils.h" #include "glusterd-op-sm.h" +#include "glusterd-store.h" #include "glusterfs3.h" #include "protocol-common.h" @@ -39,31 +30,170 @@ extern struct rpc_clnt_program gd_peer_prog; extern struct rpc_clnt_program gd_mgmt_prog; +extern struct rpc_clnt_program gd_mgmt_v3_prog; +extern struct rpc_clnt_program gd_mgmt_v3_prog; + + +#define TRUSTED_PREFIX "trusted-" typedef ssize_t (*gfs_serialize_t) (struct iovec outmsg, void *data); +static int +get_snap_volname_and_volinfo (const char *volpath, char **volname, + glusterd_volinfo_t **volinfo) +{ + int ret = -1; + char *save_ptr = NULL; + char *str_token = NULL; + char *snapname = NULL; + char *volname_token = NULL; + char *vol = NULL; + glusterd_snap_t *snap = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (volpath); + GF_ASSERT (volinfo); + + str_token = gf_strdup (volpath); + if (NULL == str_token) { + goto out; + } + + /* Input volname will have below formats: + * /snaps/<snapname>/<volname>.<hostname> + * or + * /snaps/<snapname>/<parent-volname> + * We need to extract snapname and parent_volname */ + + /*split string by "/" */ + strtok_r (str_token, "/", &save_ptr); + snapname = strtok_r(NULL, "/", &save_ptr); + if (!snapname) { + gf_log(this->name, GF_LOG_ERROR, "Invalid path: %s", volpath); + goto out; + } + + volname_token = strtok_r(NULL, "/", &save_ptr); + if (!volname_token) { + gf_log(this->name, GF_LOG_ERROR, "Invalid path: %s", volpath); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + gf_log(this->name, GF_LOG_ERROR, "Failed to " + "fetch snap %s", snapname); + goto out; + } + + /* Find if its a parent volume name or snap volume + * name. This function will succeed if volname_token + * is a parent volname + */ + ret = glusterd_volinfo_find (volname_token, volinfo); + if (ret) { + *volname = gf_strdup (volname_token); + if (NULL == *volname) { + ret = -1; + goto out; + } + + ret = glusterd_snap_volinfo_find (volname_token, snap, + volinfo); + if (ret) { + /* Split the volume name */ + vol = strtok_r (volname_token, ".", &save_ptr); + if (!vol) { + gf_log(this->name, GF_LOG_ERROR, "Invalid " + "volname (%s)", volname_token); + goto out; + } + + ret = glusterd_snap_volinfo_find (vol, snap, volinfo); + if (ret) { + gf_log(this->name, GF_LOG_ERROR, "Failed to " + "fetch snap volume from volname (%s)", + vol); + goto out; + } + } + } else { + /*volname_token is parent volname*/ + ret = glusterd_snap_volinfo_find_from_parent_volname ( + volname_token, snap, volinfo); + if (ret) { + gf_log(this->name, GF_LOG_ERROR, "Failed to " + "fetch snap volume from parent " + "volname (%s)", volname_token); + goto out; + } + + /* Since volname_token is a parent volname we should + * get the snap volname here*/ + *volname = gf_strdup ((*volinfo)->volname); + if (NULL == *volname) { + ret = -1; + goto out; + } + } + +out: + if (ret && NULL != *volname) { + GF_FREE (*volname); + *volname = NULL; + } + return ret; +} + static size_t build_volfile_path (const char *volname, char *path, - size_t path_len) + size_t path_len, char *trusted_str) { - struct stat stbuf = {0,}; - int32_t ret = -1; - glusterd_conf_t *priv = NULL; - char *vol = NULL; - char *dup_volname = NULL; - char *free_ptr = NULL; - char *tmp = NULL; - glusterd_volinfo_t *volinfo = NULL; - char *server = NULL; - - priv = THIS->private; + struct stat stbuf = {0,}; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; + char *vol = NULL; + char *dup_volname = NULL; + char *free_ptr = NULL; + char *save_ptr = NULL; + char *str_token = NULL; + glusterd_volinfo_t *volinfo = NULL; + char *server = NULL; + const char *volname_ptr = NULL; + char path_prefix [PATH_MAX] = {0,}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (volname); + GF_ASSERT (path); if (strstr (volname, "gluster/")) { server = strchr (volname, '/') + 1; glusterd_get_nodesvc_volfile (server, priv->workdir, - path, path_len); + path, path_len); ret = 1; goto out; + } else if ((str_token = strstr (volname, "/snaps/"))) { + ret = get_snap_volname_and_volinfo (str_token, &dup_volname, + &volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get snap" + " volinfo from path (%s)", volname); + ret = -1; + goto out; + } + + snprintf (path_prefix, sizeof (path_prefix), "%s/snaps/%s", + priv->workdir, volinfo->snapshot->snapname); + + free_ptr = dup_volname; + volname_ptr = dup_volname; + goto gotvolinfo; } else if (volname[0] != '/') { /* Normal behavior */ dup_volname = gf_strdup (volname); @@ -74,66 +204,221 @@ build_volfile_path (const char *volname, char *path, dup_volname = gf_strdup (&volname[1]); } + if (!dup_volname) { + gf_log(THIS->name, GF_LOG_ERROR, "strdup failed"); + ret = -1; + goto out; + } free_ptr = dup_volname; + volname_ptr = volname; + + snprintf (path_prefix, sizeof (path_prefix), "%s/vols", + priv->workdir); ret = glusterd_volinfo_find (dup_volname, &volinfo); + if (ret) { /* Split the volume name */ - vol = strtok_r (dup_volname, ".", &tmp); + vol = strtok_r (dup_volname, ".", &save_ptr); if (!vol) goto out; + ret = glusterd_volinfo_find (vol, &volinfo); if (ret) goto out; } - ret = snprintf (path, path_len, "%s/vols/%s/%s.vol", - priv->workdir, volinfo->volname, volname); + +gotvolinfo: + if (!glusterd_auth_get_username (volinfo)) + trusted_str = NULL; + + ret = snprintf (path, path_len, "%s/%s/%s.vol", path_prefix, + volinfo->volname, volname_ptr); if (ret == -1) goto out; ret = stat (path, &stbuf); + if ((ret == -1) && (errno == ENOENT)) { - ret = snprintf (path, path_len, "%s/vols/%s/%s-fuse.vol", - priv->workdir, volinfo->volname, volname); + snprintf (path, path_len, "%s/%s/%s%s-fuse.vol", + path_prefix, volinfo->volname, + (trusted_str ? trusted_str : ""), + dup_volname); + ret = stat (path, &stbuf); } + if ((ret == -1) && (errno == ENOENT)) { - ret = snprintf (path, path_len, "%s/vols/%s/%s-tcp.vol", - priv->workdir, volinfo->volname, volname); + snprintf (path, path_len, "%s/%s/%s-tcp.vol", + path_prefix, volinfo->volname, volname_ptr); } ret = 1; out: - if (free_ptr) - GF_FREE (free_ptr); + GF_FREE (free_ptr); return ret; } -int -server_getspec (rpcsvc_request_t *req) +/* Get and store op-versions of the clients sending the getspec request + * Clients of versions <= 3.3, don't send op-versions, their op-versions are + * defaulted to 1 + */ +static int +_get_client_op_versions (gf_getspec_req *args, peer_info_t *peerinfo) { - int32_t ret = -1; - int32_t op_errno = 0; - int32_t spec_fd = -1; - size_t file_len = 0; - char filename[ZR_PATH_MAX] = {0,}; - struct stat stbuf = {0,}; - char *volume = NULL; - int cookie = 0; + int ret = 0; + int client_max_op_version = 1; + int client_min_op_version = 1; + dict_t *dict = NULL; + + GF_ASSERT (args); + GF_ASSERT (peerinfo); + + if (args->xdata.xdata_len) { + dict = dict_new (); + if (!dict) { + ret = -1; + goto out; + } - gf_getspec_req args = {0,}; - gf_getspec_rsp rsp = {0,}; + ret = dict_unserialize (args->xdata.xdata_val, + args->xdata.xdata_len, &dict); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, + "Failed to unserialize request dictionary"); + goto out; + } + ret = dict_get_int32 (dict, "min-op-version", + &client_min_op_version); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, + "Failed to get client-min-op-version"); + goto out; + } - if (!xdr_to_generic (req->msg[0], &args, (xdrproc_t)xdr_gf_getspec_req)) { + ret = dict_get_int32 (dict, "max-op-version", + &client_max_op_version); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, + "Failed to get client-max-op-version"); + goto out; + } + } + + peerinfo->max_op_version = client_max_op_version; + peerinfo->min_op_version = client_min_op_version; + +out: + return ret; +} + +/* Checks if the client supports the volume, ie. client can understand all the + * options in the volfile + */ +static gf_boolean_t +_client_supports_volume (peer_info_t *peerinfo, int32_t *op_errno) +{ + gf_boolean_t ret = _gf_true; + glusterd_volinfo_t *volinfo = NULL; + + GF_ASSERT (peerinfo); + GF_ASSERT (op_errno); + + + /* Only check when the volfile being requested is a volume. Not finding + * a volinfo implies that the volfile requested for is not of a gluster + * volume. A non volume volfile is requested by the local gluster + * services like shd and nfs-server. These need not be checked as they + * will be running at the same op-version as glusterd and will be able + * to support all the features + */ + if ((glusterd_volinfo_find (peerinfo->volname, &volinfo) == 0) && + ((peerinfo->min_op_version > volinfo->client_op_version) || + (peerinfo->max_op_version < volinfo->client_op_version))) { + ret = _gf_false; + *op_errno = ENOTSUP; + gf_log ("glusterd", GF_LOG_INFO, + "Client %s (%d -> %d) doesn't support required " + "op-version (%d). Rejecting volfile request.", + peerinfo->identifier, peerinfo->min_op_version, + peerinfo->max_op_version, volinfo->client_op_version); + } + + return ret; +} + +int +__server_getspec (rpcsvc_request_t *req) +{ + int32_t ret = -1; + int32_t op_errno = 0; + int32_t spec_fd = -1; + size_t file_len = 0; + char filename[PATH_MAX] = {0,}; + struct stat stbuf = {0,}; + char *volume = NULL; + char *tmp = NULL; + int cookie = 0; + rpc_transport_t *trans = NULL; + gf_getspec_req args = {0,}; + gf_getspec_rsp rsp = {0,}; + char addrstr[RPCSVC_PEER_STRLEN] = {0}; + peer_info_t *peerinfo = NULL; + + ret = xdr_to_generic (req->msg[0], &args, + (xdrproc_t)xdr_gf_getspec_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto fail; } + peerinfo = &req->trans->peerinfo; + volume = args.key; + /* Need to strip leading '/' from volnames. This was introduced to + * support nfs style mount parameters for native gluster mount + */ + if (volume[0] == '/') + strncpy (peerinfo->volname, &volume[1], strlen(&volume[1])); + else + strncpy (peerinfo->volname, volume, strlen(volume)); + + ret = _get_client_op_versions (&args, peerinfo); + if (ret) + goto fail; + + if (!_client_supports_volume (peerinfo, &op_errno)) { + ret = -1; + goto fail; + } - ret = build_volfile_path (volume, filename, sizeof (filename)); + trans = req->trans; + /* addrstr will be empty for cli socket connections */ + ret = rpcsvc_transport_peername (trans, (char *)&addrstr, + sizeof (addrstr)); + if (ret) + goto fail; + + tmp = strrchr (addrstr, ':'); + if (tmp) + *tmp = '\0'; + + /* The trusted volfiles are given to the glusterd owned process like NFS + * server, self-heal daemon etc., so that they are not inadvertently + * blocked by a auth.{allow,reject} setting. The trusted volfile is not + * meant for external users. + */ + if (strlen (addrstr) && gf_is_local_addr (addrstr)) { + + ret = build_volfile_path (volume, filename, + sizeof (filename), + TRUSTED_PREFIX); + } else { + ret = build_volfile_path (volume, filename, + sizeof (filename), NULL); + } if (ret > 0) { /* to allocate the proper buffer to hold the file data */ @@ -179,45 +464,360 @@ fail: rsp.op_errno = cookie; if (!rsp.spec) - rsp.spec = ""; + rsp.spec = strdup (""); glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_gf_getspec_rsp); - if (args.key) - free (args.key);//malloced by xdr - if (rsp.spec && (strcmp (rsp.spec, ""))) - free (rsp.spec); + free (args.key);//malloced by xdr + free (rsp.spec); return 0; } +int +server_getspec (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __server_getspec); +} + +int32_t +__server_event_notify (rpcsvc_request_t *req) +{ + int32_t ret = -1; + int32_t op_errno = 0; + gf_event_notify_req args = {0,}; + gf_event_notify_rsp rsp = {0,}; + dict_t *dict = NULL; + gf_boolean_t need_rsp = _gf_true; + + ret = xdr_to_generic (req->msg[0], &args, + (xdrproc_t)xdr_gf_event_notify_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + goto fail; + } + + if (args.dict.dict_len) { + dict = dict_new (); + if (!dict) + return ret; + ret = dict_unserialize (args.dict.dict_val, + args.dict.dict_len, &dict); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Failed to unserialize req"); + goto fail; + } + } + + switch (args.op) { + case GF_EN_DEFRAG_STATUS: + gf_log ("", GF_LOG_INFO, + "received defrag status updated"); + if (dict) { + glusterd_defrag_event_notify_handle (dict); + need_rsp = _gf_false; + } + break; + default: + gf_log ("", GF_LOG_ERROR, "Unknown op received in event " + "notify"); + ret = -1; + break; + } + +fail: + rsp.op_ret = ret; + + if (op_errno) + rsp.op_errno = gf_errno_to_error (op_errno); + + if (need_rsp) + glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_event_notify_rsp); + if (dict) + dict_unref (dict); + free (args.dict.dict_val);//malloced by xdr + + return 0; +} + +int32_t +server_event_notify (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __server_event_notify); +} + +int +gd_validate_cluster_op_version (xlator_t *this, int cluster_op_version, + char *peerid) +{ + int ret = -1; + glusterd_conf_t *conf = NULL; + + conf = this->private; + + if (cluster_op_version > GD_OP_VERSION_MAX) { + gf_log (this->name, GF_LOG_ERROR, + "operating version %d is more than the maximum " + "supported (%d) on the machine (as per peer request " + "from %s)", cluster_op_version, GD_OP_VERSION_MAX, + peerid); + goto out; + } + + /* The peer can only reduce its op-version when it doesn't have any + * volumes. Reducing op-version when it already contains volumes can + * lead to inconsistencies in the cluster + */ + if ((cluster_op_version < conf->op_version) && + !list_empty (&conf->volumes)) { + gf_log (this->name, GF_LOG_ERROR, + "cannot reduce operating version to %d from current " + "version %d as volumes exist (as per peer request from " + "%s)", cluster_op_version, conf->op_version, peerid); + goto out; + } + + ret = 0; +out: + return ret; +} + +int +__glusterd_mgmt_hndsk_versions (rpcsvc_request_t *req) +{ + dict_t *dict = NULL; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + int ret = -1; + int op_errno = EINVAL; + gf_mgmt_hndsk_req args = {{0,},}; + gf_mgmt_hndsk_rsp rsp = {0,}; + + this = THIS; + conf = this->private; + + ret = xdr_to_generic (req->msg[0], &args, + (xdrproc_t)xdr_gf_mgmt_hndsk_req); + if (ret < 0) { + //failed to decode msg; + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_set_int32 (dict, GD_OP_VERSION_KEY, conf->op_version); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to set operating version"); + rsp.op_ret = ret; + goto out; + } + + ret = dict_set_int32 (dict, GD_MIN_OP_VERSION_KEY, GD_OP_VERSION_MIN); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to set %s", GD_MIN_OP_VERSION_KEY); + rsp.op_ret = ret; + goto out; + } + + ret = dict_set_int32 (dict, GD_MAX_OP_VERSION_KEY, GD_OP_VERSION_MAX); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to set %s", GD_MAX_OP_VERSION_KEY); + rsp.op_ret = ret; + goto out; + } + + ret = 0; + + GF_PROTOCOL_DICT_SERIALIZE (this, dict, (&rsp.hndsk.hndsk_val), + rsp.hndsk.hndsk_len, op_errno, out); +out: + + rsp.op_ret = ret; + rsp.op_errno = op_errno; + + glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_mgmt_hndsk_rsp); + + ret = 0; + + if (dict) + dict_unref (dict); + + if (args.hndsk.hndsk_val) + free (args.hndsk.hndsk_val); + + if (rsp.hndsk.hndsk_val) + GF_FREE (rsp.hndsk.hndsk_val); + + return ret; +} + +int +glusterd_mgmt_hndsk_versions (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_mgmt_hndsk_versions); +} + +int +__glusterd_mgmt_hndsk_versions_ack (rpcsvc_request_t *req) +{ + dict_t *clnt_dict = NULL; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + int ret = -1; + int op_errno = EINVAL; + int peer_op_version = 0; + gf_mgmt_hndsk_req args = {{0,},}; + gf_mgmt_hndsk_rsp rsp = {0,}; + + this = THIS; + conf = this->private; + + ret = xdr_to_generic (req->msg[0], &args, + (xdrproc_t)xdr_gf_mgmt_hndsk_req); + if (ret < 0) { + //failed to decode msg; + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + GF_PROTOCOL_DICT_UNSERIALIZE (this, clnt_dict, args.hndsk.hndsk_val, + (args.hndsk.hndsk_len), ret, op_errno, + out); + + ret = dict_get_int32 (clnt_dict, GD_OP_VERSION_KEY, &peer_op_version); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to get the op-version key peer=%s", + req->trans->peerinfo.identifier); + goto out; + } + + ret = gd_validate_cluster_op_version (this, peer_op_version, + req->trans->peerinfo.identifier); + if (ret) + goto out; + + + /* As this is ACK from the Cluster for the versions supported, + can set the op-version of 'this' glusterd to the one + received. */ + gf_log (this->name, GF_LOG_INFO, "using the op-version %d", + peer_op_version); + conf->op_version = peer_op_version; + ret = glusterd_store_global_info (this); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "Failed to store op-version"); + +out: + rsp.op_ret = ret; + rsp.op_errno = op_errno; + + glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_mgmt_hndsk_rsp); + + ret = 0; + + if (clnt_dict) + dict_unref (clnt_dict); + + if (args.hndsk.hndsk_val) + free (args.hndsk.hndsk_val); + + return ret; +} + +int +glusterd_mgmt_hndsk_versions_ack (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_mgmt_hndsk_versions_ack); +} rpcsvc_actor_t gluster_handshake_actors[] = { - [GF_HNDSK_NULL] = {"NULL", GF_HNDSK_NULL, NULL, NULL, NULL, 0}, - [GF_HNDSK_GETSPEC] = {"GETSPEC", GF_HNDSK_GETSPEC, server_getspec, NULL, NULL, 0}, + [GF_HNDSK_NULL] = {"NULL", GF_HNDSK_NULL, NULL, NULL, 0, DRC_NA}, + [GF_HNDSK_GETSPEC] = {"GETSPEC", GF_HNDSK_GETSPEC, server_getspec, NULL, 0, DRC_NA}, + [GF_HNDSK_EVENT_NOTIFY] = {"EVENTNOTIFY", GF_HNDSK_EVENT_NOTIFY, server_event_notify, NULL, 0, DRC_NA}, }; struct rpcsvc_program gluster_handshake_prog = { - .progname = "GlusterFS Handshake", + .progname = "Gluster Handshake", .prognum = GLUSTER_HNDSK_PROGRAM, .progver = GLUSTER_HNDSK_VERSION, .actors = gluster_handshake_actors, .numactors = GF_HNDSK_MAXVALUE, }; +/* A minimal RPC program just for the cli getspec command */ +rpcsvc_actor_t gluster_cli_getspec_actors[] = { + [GF_HNDSK_GETSPEC] = {"GETSPEC", GF_HNDSK_GETSPEC, server_getspec, NULL, 0, DRC_NA}, +}; + +struct rpcsvc_program gluster_cli_getspec_prog = { + .progname = "Gluster Handshake (CLI Getspec)", + .prognum = GLUSTER_HNDSK_PROGRAM, + .progver = GLUSTER_HNDSK_VERSION, + .actors = gluster_cli_getspec_actors, + .numactors = GF_HNDSK_MAXVALUE, +}; + + char *glusterd_dump_proc[GF_DUMP_MAXVALUE] = { [GF_DUMP_NULL] = "NULL", [GF_DUMP_DUMP] = "DUMP", }; rpc_clnt_prog_t glusterd_dump_prog = { - .progname = "GLUSTERD-DUMP", - .prognum = GLUSTER_DUMP_PROGRAM, - .progver = GLUSTER_DUMP_VERSION, - .procnames = glusterd_dump_proc, + .progname = "GLUSTERD-DUMP", + .prognum = GLUSTER_DUMP_PROGRAM, + .progver = GLUSTER_DUMP_VERSION, + .procnames = glusterd_dump_proc, +}; + + +rpcsvc_actor_t glusterd_mgmt_hndsk_actors[] = { + [GD_MGMT_HNDSK_NULL] = {"NULL", GD_MGMT_HNDSK_NULL, NULL, + NULL, 0}, + [GD_MGMT_HNDSK_VERSIONS] = {"MGMT-VERS", GD_MGMT_HNDSK_VERSIONS, + glusterd_mgmt_hndsk_versions, NULL, + 0}, + [GD_MGMT_HNDSK_VERSIONS_ACK] = {"MGMT-VERS-ACK", + GD_MGMT_HNDSK_VERSIONS_ACK, + glusterd_mgmt_hndsk_versions_ack, + NULL, 0}, }; +struct rpcsvc_program glusterd_mgmt_hndsk_prog = { + .progname = "Gluster MGMT Handshake", + .prognum = GD_MGMT_HNDSK_PROGRAM, + .progver = GD_MGMT_HNDSK_VERSION, + .actors = glusterd_mgmt_hndsk_actors, + .numactors = GD_MGMT_HNDSK_MAXVALUE, +}; + +char *glusterd_mgmt_hndsk_proc[GD_MGMT_HNDSK_MAXVALUE] = { + [GD_MGMT_HNDSK_NULL] = "NULL", + [GD_MGMT_HNDSK_VERSIONS] = "MGMT-VERS", + [GD_MGMT_HNDSK_VERSIONS_ACK] = "MGMT-VERS-ACK", +}; + +rpc_clnt_prog_t gd_clnt_mgmt_hndsk_prog = { + .progname = "Gluster MGMT Handshake", + .prognum = GD_MGMT_HNDSK_PROGRAM, + .progver = GD_MGMT_HNDSK_VERSION, + .procnames = glusterd_mgmt_hndsk_proc, +}; + + static int glusterd_event_connected_inject (glusterd_peerctx_t *peerctx) { @@ -249,6 +849,7 @@ glusterd_event_connected_inject (glusterd_peerctx_t *peerctx) ctx->hostname = gf_strdup (peerinfo->hostname); ctx->port = peerinfo->port; ctx->req = peerctx->args.req; + ctx->dict = peerctx->args.dict; event->peerinfo = peerinfo; event->ctx = ctx; @@ -266,12 +867,286 @@ out: return ret; } + +int +gd_validate_peer_op_version (xlator_t *this, glusterd_peerinfo_t *peerinfo, + dict_t *dict, char **errstr) +{ + int ret = -1; + glusterd_conf_t *conf = NULL; + int32_t peer_op_version = 0; + int32_t peer_min_op_version = 0; + int32_t peer_max_op_version = 0; + + if (!dict && !this && !peerinfo) + goto out; + + conf = this->private; + + ret = dict_get_int32 (dict, GD_OP_VERSION_KEY, &peer_op_version); + if (ret) + goto out; + + ret = dict_get_int32 (dict, GD_MAX_OP_VERSION_KEY, + &peer_max_op_version); + if (ret) + goto out; + + ret = dict_get_int32 (dict, GD_MIN_OP_VERSION_KEY, + &peer_min_op_version); + if (ret) + goto out; + + ret = -1; + /* Check if peer can support our op_version */ + if ((peer_max_op_version < conf->op_version) || + (peer_min_op_version > conf->op_version)) { + ret = gf_asprintf (errstr, "Peer %s does not support required " + "op-version", peerinfo->hostname); + ret = -1; + goto out; + } + + ret = 0; +out: + gf_log (this->name , GF_LOG_DEBUG, "Peer %s %s", peerinfo->hostname, + ((ret < 0) ? "rejected" : "accepted")); + return ret; +} + +int +__glusterd_mgmt_hndsk_version_ack_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int ret = -1; + int op_errno = EINVAL; + gf_mgmt_hndsk_rsp rsp = {0,}; + xlator_t *this = NULL; + call_frame_t *frame = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_peerctx_t *peerctx = NULL; + char msg[1024] = {0,}; + + this = THIS; + frame = myframe; + peerctx = frame->local; + peerinfo = peerctx->peerinfo; + + if (-1 == req->rpc_status) { + snprintf (msg, sizeof (msg), + "Error through RPC layer, retry again later"); + gf_log ("", GF_LOG_ERROR, "%s", msg); + peerctx->errstr = gf_strdup (msg); + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_mgmt_hndsk_rsp); + if (ret < 0) { + snprintf (msg, sizeof (msg), "Failed to decode XDR"); + gf_log ("", GF_LOG_ERROR, "%s", msg); + peerctx->errstr = gf_strdup (msg); + goto out; + } + + op_errno = rsp.op_errno; + if (-1 == rsp.op_ret) { + ret = -1; + snprintf (msg, sizeof (msg), + "Failed to get handshake ack from remote server"); + gf_log (frame->this->name, GF_LOG_ERROR, "%s", msg); + peerctx->errstr = gf_strdup (msg); + goto out; + } + + /* TODO: this is hardcoded as of now, but I don't forsee any problems + * with this as long as we are properly handshaking operating versions + */ + peerinfo->mgmt = &gd_mgmt_prog; + peerinfo->peer = &gd_peer_prog; + peerinfo->mgmt_v3 = &gd_mgmt_v3_prog; + + ret = default_notify (this, GF_EVENT_CHILD_UP, NULL); + + if (GD_MODE_ON == peerctx->args.mode) { + ret = glusterd_event_connected_inject (peerctx); + peerctx->args.req = NULL; + } else if (GD_MODE_SWITCH_ON == peerctx->args.mode) { + peerctx->args.mode = GD_MODE_ON; + } else { + gf_log (this->name, GF_LOG_WARNING, "unknown mode %d", + peerctx->args.mode); + } + + glusterd_friend_sm (); + + ret = 0; +out: + + frame->local = NULL; + STACK_DESTROY (frame->root); + + if (ret != 0) + rpc_transport_disconnect (peerinfo->rpc->conn.trans); + + if (rsp.hndsk.hndsk_val) + free (rsp.hndsk.hndsk_val); + + return 0; +} + +int +glusterd_mgmt_hndsk_version_ack_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_mgmt_hndsk_version_ack_cbk); +} + +int +__glusterd_mgmt_hndsk_version_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int ret = -1; + int op_errno = EINVAL; + gf_mgmt_hndsk_rsp rsp = {0,}; + gf_mgmt_hndsk_req arg = {{0,}}; + xlator_t *this = NULL; + call_frame_t *frame = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_peerctx_t *peerctx = NULL; + dict_t *dict = NULL; + dict_t *rsp_dict = NULL; + glusterd_conf_t *conf = NULL; + char msg[1024] = {0,}; + + this = THIS; + conf = this->private; + frame = myframe; + peerctx = frame->local; + peerinfo = peerctx->peerinfo; + + if (-1 == req->rpc_status) { + ret = -1; + snprintf (msg, sizeof (msg), + "Error through RPC layer, retry again later"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); + peerctx->errstr = gf_strdup (msg); + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_mgmt_hndsk_rsp); + if (ret < 0) { + snprintf (msg, sizeof (msg), "Failed to decode management " + "handshake response"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); + peerctx->errstr = gf_strdup (msg); + goto out; + } + + GF_PROTOCOL_DICT_UNSERIALIZE (this, dict, rsp.hndsk.hndsk_val, + rsp.hndsk.hndsk_len, ret, op_errno, + out); + + op_errno = rsp.op_errno; + if (-1 == rsp.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get the 'versions' from peer (%s)", + req->conn->trans->peerinfo.identifier); + goto out; + } + + /* Check if peer can be part of cluster */ + ret = gd_validate_peer_op_version (this, peerinfo, dict, + &peerctx->errstr); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to validate the operating version of peer (%s)", + peerinfo->hostname); + goto out; + } + + rsp_dict = dict_new (); + if (!rsp_dict) + goto out; + + ret = dict_set_int32 (rsp_dict, GD_OP_VERSION_KEY, conf->op_version); + if (ret) { + gf_log(this->name, GF_LOG_ERROR, + "failed to set operating version in dict"); + goto out; + } + + GF_PROTOCOL_DICT_SERIALIZE (this, rsp_dict, (&arg.hndsk.hndsk_val), + arg.hndsk.hndsk_len, op_errno, out); + + ret = glusterd_submit_request (peerctx->peerinfo->rpc, &arg, frame, + &gd_clnt_mgmt_hndsk_prog, + GD_MGMT_HNDSK_VERSIONS_ACK, NULL, this, + glusterd_mgmt_hndsk_version_ack_cbk, + (xdrproc_t)xdr_gf_mgmt_hndsk_req); + +out: + if (ret) { + frame->local = NULL; + STACK_DESTROY (frame->root); + rpc_transport_disconnect (peerinfo->rpc->conn.trans); + } + + if (rsp.hndsk.hndsk_val) + free (rsp.hndsk.hndsk_val); + + if (arg.hndsk.hndsk_val) + GF_FREE (arg.hndsk.hndsk_val); + + if (dict) + dict_unref (dict); + + if (rsp_dict) + dict_unref (rsp_dict); + + return 0; +} + +int +glusterd_mgmt_hndsk_version_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_mgmt_hndsk_version_cbk); +} + +int +glusterd_mgmt_handshake (xlator_t *this, glusterd_peerctx_t *peerctx) +{ + call_frame_t *frame = NULL; + gf_mgmt_hndsk_req req = {{0,},}; + int ret = -1; + + frame = create_frame (this, this->ctx->pool); + if (!frame) + goto out; + + frame->local = peerctx; + + ret = glusterd_submit_request (peerctx->peerinfo->rpc, &req, frame, + &gd_clnt_mgmt_hndsk_prog, + GD_MGMT_HNDSK_VERSIONS, NULL, this, + glusterd_mgmt_hndsk_version_cbk, + (xdrproc_t)xdr_gf_mgmt_hndsk_req); + ret = 0; +out: + if (ret && frame) + STACK_DESTROY (frame->root); + + return ret; +} + int glusterd_set_clnt_mgmt_program (glusterd_peerinfo_t *peerinfo, gf_prog_detail *prog) { - gf_prog_detail *trav = NULL; - int ret = -1; + gf_prog_detail *trav = NULL; + int ret = -1; if (!peerinfo || !prog) goto out; @@ -279,45 +1154,79 @@ glusterd_set_clnt_mgmt_program (glusterd_peerinfo_t *peerinfo, trav = prog; while (trav) { - /* Select 'programs' */ + ret = -1; if ((gd_mgmt_prog.prognum == trav->prognum) && (gd_mgmt_prog.progver == trav->progver)) { peerinfo->mgmt = &gd_mgmt_prog; ret = 0; } + if ((gd_peer_prog.prognum == trav->prognum) && (gd_peer_prog.progver == trav->progver)) { peerinfo->peer = &gd_peer_prog; ret = 0; } + if (ret) { gf_log ("", GF_LOG_DEBUG, "%s (%"PRId64":%"PRId64") not supported", trav->progname, trav->prognum, trav->progver); } + trav = trav->next; } if (peerinfo->mgmt) { - gf_log ("", GF_LOG_INFO, - "Using Program %s, Num (%d), Version (%d)", - peerinfo->mgmt->progname, peerinfo->mgmt->prognum, - peerinfo->mgmt->progver); + gf_log ("", GF_LOG_INFO, + "Using Program %s, Num (%d), Version (%d)", + peerinfo->mgmt->progname, peerinfo->mgmt->prognum, + peerinfo->mgmt->progver); } + if (peerinfo->peer) { - gf_log ("", GF_LOG_INFO, - "Using Program %s, Num (%d), Version (%d)", - peerinfo->peer->progname, peerinfo->peer->prognum, - peerinfo->peer->progver); + gf_log ("", GF_LOG_INFO, + "Using Program %s, Num (%d), Version (%d)", + peerinfo->peer->progname, peerinfo->peer->prognum, + peerinfo->peer->progver); } + if (peerinfo->mgmt_v3) { + gf_log ("", GF_LOG_INFO, + "Using Program %s, Num (%d), Version (%d)", + peerinfo->mgmt_v3->progname, peerinfo->mgmt_v3->prognum, + peerinfo->mgmt_v3->progver); + } + + ret = 0; +out: + return ret; + +} + +static gf_boolean_t +_mgmt_hndsk_prog_present (gf_prog_detail *prog) { + gf_boolean_t ret = _gf_false; + gf_prog_detail *trav = NULL; + + GF_ASSERT (prog); + + trav = prog; + + while (trav) { + if ((trav->prognum == GD_MGMT_HNDSK_PROGRAM) && + (trav->progver == GD_MGMT_HNDSK_VERSION)) { + ret = _gf_true; + goto out; + } + trav = trav->next; + } out: return ret; } int -glusterd_peer_dump_version_cbk (struct rpc_req *req, struct iovec *iov, +__glusterd_peer_dump_version_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) { int ret = -1; @@ -328,26 +1237,51 @@ glusterd_peer_dump_version_cbk (struct rpc_req *req, struct iovec *iov, call_frame_t *frame = NULL; glusterd_peerinfo_t *peerinfo = NULL; glusterd_peerctx_t *peerctx = NULL; + glusterd_conf_t *conf = NULL; + char msg[1024] = {0,}; this = THIS; + conf = this->private; frame = myframe; peerctx = frame->local; peerinfo = peerctx->peerinfo; if (-1 == req->rpc_status) { - gf_log ("", GF_LOG_ERROR, - "error through RPC layer, retry again later"); + snprintf (msg, sizeof (msg), + "Error through RPC layer, retry again later"); + gf_log ("", GF_LOG_ERROR, "%s", msg); + peerctx->errstr = gf_strdup (msg); goto out; } ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_dump_rsp); if (ret < 0) { - gf_log ("", GF_LOG_ERROR, "failed to decode XDR"); + snprintf (msg, sizeof (msg), "Failed to decode XDR"); + gf_log ("", GF_LOG_ERROR, "%s", msg); + peerctx->errstr = gf_strdup (msg); goto out; } if (-1 == rsp.op_ret) { - gf_log (frame->this->name, GF_LOG_ERROR, - "failed to get the 'versions' from remote server"); + snprintf (msg, sizeof (msg), + "Failed to get the 'versions' from remote server"); + gf_log (frame->this->name, GF_LOG_ERROR, "%s", msg); + peerctx->errstr = gf_strdup (msg); + goto out; + } + + if (_mgmt_hndsk_prog_present (rsp.prog)) { + gf_log (this->name, GF_LOG_DEBUG, + "Proceeding to op-version handshake with peer %s", + peerinfo->hostname); + ret = glusterd_mgmt_handshake (this, peerctx); + goto out; + } else if (conf->op_version > 1) { + ret = -1; + snprintf (msg, sizeof (msg), + "Peer %s does not support required op-version", + peerinfo->hostname); + peerctx->errstr = gf_strdup (msg); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); goto out; } @@ -370,10 +1304,11 @@ glusterd_peer_dump_version_cbk (struct rpc_req *req, struct iovec *iov, peerctx->args.mode); } - glusterd_friend_sm (); - glusterd_op_sm (); + glusterd_friend_sm(); + glusterd_op_sm(); ret = 0; + out: /* don't use GF_FREE, buffer was allocated by libc */ @@ -398,8 +1333,16 @@ out: int -glusterd_peer_handshake (xlator_t *this, struct rpc_clnt *rpc, - glusterd_peerctx_t *peerctx) +glusterd_peer_dump_version_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_peer_dump_version_cbk); +} + +int +glusterd_peer_dump_version (xlator_t *this, struct rpc_clnt *rpc, + glusterd_peerctx_t *peerctx) { call_frame_t *frame = NULL; gf_dump_req req = {0,}; diff --git a/xlators/mgmt/glusterd/src/glusterd-hooks.c b/xlators/mgmt/glusterd/src/glusterd-hooks.c new file mode 100644 index 000000000..2b43a452e --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-hooks.c @@ -0,0 +1,531 @@ +/* + Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "globals.h" +#include "glusterfs.h" +#include "dict.h" +#include "xlator.h" +#include "logging.h" +#include "run.h" +#include "defaults.h" +#include "compat.h" +#include "compat-errno.h" +#include "glusterd.h" +#include "glusterd-sm.h" +#include "glusterd-op-sm.h" +#include "glusterd-utils.h" +#include "glusterd-store.h" +#include "glusterd-hooks.h" + +#include <fnmatch.h> + +#define EMPTY "" +char glusterd_hook_dirnames[GD_OP_MAX][256] = +{ + [GD_OP_NONE] = EMPTY, + [GD_OP_CREATE_VOLUME] = "create", + [GD_OP_START_BRICK] = EMPTY, + [GD_OP_STOP_BRICK] = EMPTY, + [GD_OP_DELETE_VOLUME] = "delete", + [GD_OP_START_VOLUME] = "start", + [GD_OP_STOP_VOLUME] = "stop", + [GD_OP_DEFRAG_VOLUME] = EMPTY, + [GD_OP_ADD_BRICK] = "add-brick", + [GD_OP_REMOVE_BRICK] = "remove-brick", + [GD_OP_REPLACE_BRICK] = EMPTY, + [GD_OP_SET_VOLUME] = "set", + [GD_OP_RESET_VOLUME] = EMPTY, + [GD_OP_SYNC_VOLUME] = EMPTY, + [GD_OP_LOG_ROTATE] = EMPTY, + [GD_OP_GSYNC_CREATE] = "gsync-create", + [GD_OP_GSYNC_SET] = EMPTY, + [GD_OP_PROFILE_VOLUME] = EMPTY, + [GD_OP_QUOTA] = EMPTY, + [GD_OP_STATUS_VOLUME] = EMPTY, + [GD_OP_REBALANCE] = EMPTY, + [GD_OP_HEAL_VOLUME] = EMPTY, + [GD_OP_STATEDUMP_VOLUME] = EMPTY, + [GD_OP_LIST_VOLUME] = EMPTY, + [GD_OP_CLEARLOCKS_VOLUME] = EMPTY, + [GD_OP_DEFRAG_BRICK_VOLUME] = EMPTY, +}; +#undef EMPTY + +static inline gf_boolean_t +glusterd_is_hook_enabled (char *script) +{ + return (script[0] == 'S'); +} + +int +glusterd_hooks_create_hooks_directory (char *basedir) +{ + int ret = -1; + int op = GD_OP_NONE; + int type = GD_COMMIT_HOOK_NONE; + char version_dir[PATH_MAX] = {0, }; + char path[PATH_MAX] = {0, }; + char *cmd_subdir = NULL; + char type_subdir[GD_COMMIT_HOOK_MAX][256] = {{0, }, + "pre", + "post"}; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + + snprintf (path, sizeof (path), "%s/hooks", basedir); + ret = mkdir_p (path, 0777, _gf_true); + if (ret) { + gf_log (THIS->name, GF_LOG_CRITICAL, "Unable to create %s due" + "to %s", path, strerror (errno)); + goto out; + } + + GLUSTERD_GET_HOOKS_DIR (version_dir, GLUSTERD_HOOK_VER, priv); + ret = mkdir_p (version_dir, 0777, _gf_true); + if (ret) { + gf_log (THIS->name, GF_LOG_CRITICAL, "Unable to create %s due " + "to %s", version_dir, strerror (errno)); + goto out; + } + + for (op = GD_OP_NONE+1; op < GD_OP_MAX; op++) { + cmd_subdir = glusterd_hooks_get_hooks_cmd_subdir (op); + if (strlen (cmd_subdir) == 0) + continue; + + snprintf (path, sizeof (path), "%s/%s", version_dir, + cmd_subdir); + ret = mkdir_p (path, 0777, _gf_true); + if (ret) { + gf_log (THIS->name, GF_LOG_CRITICAL, + "Unable to create %s due to %s", + path, strerror (errno)); + goto out; + } + + for (type = GD_COMMIT_HOOK_PRE; type < GD_COMMIT_HOOK_MAX; + type++) { + snprintf (path, sizeof (path), "%s/%s/%s", + version_dir, cmd_subdir, type_subdir[type]); + ret = mkdir_p (path, 0777, _gf_true); + if (ret) { + gf_log (THIS->name, GF_LOG_CRITICAL, + "Unable to create %s due to %s", + path, strerror (errno)); + goto out; + } + } + } + + ret = 0; +out: + return ret; +} + +char* +glusterd_hooks_get_hooks_cmd_subdir (glusterd_op_t op) +{ + GF_ASSERT ((op > GD_OP_NONE) && (op < GD_OP_MAX)); + + return glusterd_hook_dirnames[op]; +} + +int +glusterd_hooks_set_volume_args (dict_t *dict, runner_t *runner) +{ + int i = 0; + int count = 0; + int ret = -1; + char query[1024] = {0,}; + char *key = NULL; + char *value = NULL; + + ret = dict_get_int32 (dict, "count", &count); + if (ret) + goto out; + + /* This will not happen unless op_ctx + * is corrupted*/ + if (!count) + goto out; + + runner_add_arg (runner, "-o"); + for (i = 1; (ret == 0); i++) { + snprintf (query, sizeof (query), "key%d", i); + ret = dict_get_str (dict, query, &key); + if (ret) + continue; + + snprintf (query, sizeof (query), "value%d", i); + ret = dict_get_str (dict, query, &value); + if (ret) + continue; + + runner_argprintf (runner, "%s=%s", key, value); + } + + ret = 0; +out: + return ret; +} + +static int +glusterd_hooks_add_op_args (runner_t *runner, glusterd_op_t op, + dict_t *op_ctx, glusterd_commit_hook_type_t type) +{ + char *hooks_args = NULL; + int vol_count = 0; + gf_boolean_t truth = _gf_false; + glusterd_volinfo_t *voliter = NULL; + glusterd_conf_t *priv = NULL; + int ret = -1; + + priv = THIS->private; + list_for_each_entry (voliter, &priv->volumes, + vol_list) { + if (glusterd_is_volume_started (voliter)) + vol_count++; + } + + ret = 0; + switch (op) { + case GD_OP_START_VOLUME: + if (type == GD_COMMIT_HOOK_PRE && + vol_count == 0) + truth = _gf_true; + + else if (type == GD_COMMIT_HOOK_POST && + vol_count == 1) + truth = _gf_true; + + else + truth = _gf_false; + + runner_argprintf (runner, "--first=%s", + truth? "yes":"no"); + break; + + case GD_OP_STOP_VOLUME: + if (type == GD_COMMIT_HOOK_PRE && + vol_count == 1) + truth = _gf_true; + + else if (type == GD_COMMIT_HOOK_POST && + vol_count == 0) + truth = _gf_true; + + else + truth = _gf_false; + + runner_argprintf (runner, "--last=%s", + truth? "yes":"no"); + break; + + case GD_OP_SET_VOLUME: + ret = glusterd_hooks_set_volume_args (op_ctx, runner); + break; + + case GD_OP_GSYNC_CREATE: + ret = dict_get_str (op_ctx, "hooks_args", &hooks_args); + if (ret) + gf_log ("", GF_LOG_DEBUG, + "No Hooks Arguments."); + else + gf_log ("", GF_LOG_DEBUG, + "Hooks Args = %s", hooks_args); + if (hooks_args) + runner_argprintf (runner, "%s", hooks_args); + break; + + default: + break; + + } + + return ret; +} + +int +glusterd_hooks_run_hooks (char *hooks_path, glusterd_op_t op, dict_t *op_ctx, + glusterd_commit_hook_type_t type) +{ + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + runner_t runner = {0, }; + struct dirent *entry = NULL; + DIR *hookdir = NULL; + char *volname = NULL; + char **lines = NULL; + int N = 8; /*arbitrary*/ + int lineno = 0; + int line_count = 0; + int ret = -1; + + this = THIS; + priv = this->private; + + ret = dict_get_str (op_ctx, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, "Failed to get volname " + "from operation context"); + goto out; + } + + hookdir = opendir (hooks_path); + if (!hookdir) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Failed to open dir %s, due " + "to %s", hooks_path, strerror (errno)); + goto out; + } + + lines = GF_CALLOC (1, N * sizeof (*lines), gf_gld_mt_charptr); + if (!lines) { + ret = -1; + goto out; + } + + ret = -1; + line_count = 0; + glusterd_for_each_entry (entry, hookdir); + while (entry) { + if (line_count == N-1) { + N *= 2; + lines = GF_REALLOC (lines, N * sizeof (char *)); + if (!lines) + goto out; + } + + if (glusterd_is_hook_enabled (entry->d_name)) { + lines[line_count] = gf_strdup (entry->d_name); + line_count++; + } + + glusterd_for_each_entry (entry, hookdir); + } + + lines[line_count] = NULL; + lines = GF_REALLOC (lines, (line_count + 1) * sizeof (char *)); + if (!lines) + goto out; + + qsort (lines, line_count, sizeof (*lines), glusterd_compare_lines); + + for (lineno = 0; lineno < line_count; lineno++) { + + runinit (&runner); + runner_argprintf (&runner, "%s/%s", hooks_path, lines[lineno]); + /*Add future command line arguments to hook scripts below*/ + runner_argprintf (&runner, "--volname=%s", volname); + ret = glusterd_hooks_add_op_args (&runner, op, op_ctx, type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to add " + "command specific arguments"); + goto out; + } + + ret = runner_run_reuse (&runner); + if (ret) { + runner_log (&runner, this->name, GF_LOG_ERROR, + "Failed to execute script"); + } else { + runner_log (&runner, this->name, GF_LOG_INFO, + "Ran script"); + } + runner_end (&runner); + } + + ret = 0; +out: + if (lines) { + for (lineno = 0; lineno < line_count+1; lineno++) + GF_FREE (lines[lineno]); + + GF_FREE (lines); + } + + if (hookdir) + closedir (hookdir); + + return ret; +} + +int +glusterd_hooks_post_stub_enqueue (char *scriptdir, glusterd_op_t op, + dict_t *op_ctx) +{ + int ret = -1; + glusterd_hooks_stub_t *stub = NULL; + glusterd_hooks_private_t *hooks_priv = NULL; + glusterd_conf_t *conf = NULL; + + conf = THIS->private; + hooks_priv = conf->hooks_priv; + + ret = glusterd_hooks_stub_init (&stub, scriptdir, op, op_ctx); + if (ret) + goto out; + + pthread_mutex_lock (&hooks_priv->mutex); + { + hooks_priv->waitcount++; + list_add_tail (&stub->all_hooks, &hooks_priv->list); + pthread_cond_signal (&hooks_priv->cond); + } + pthread_mutex_unlock (&hooks_priv->mutex); + + ret = 0; +out: + return ret; +} + +int +glusterd_hooks_stub_init (glusterd_hooks_stub_t **stub, char *scriptdir, + glusterd_op_t op, dict_t *op_ctx) +{ + int ret = -1; + glusterd_hooks_stub_t *hooks_stub = NULL; + + GF_ASSERT (stub); + if (!stub) + goto out; + + hooks_stub = GF_CALLOC (1, sizeof (*hooks_stub), + gf_gld_mt_hooks_stub_t); + if (!hooks_stub) + goto out; + + INIT_LIST_HEAD (&hooks_stub->all_hooks); + hooks_stub->op = op; + hooks_stub->scriptdir = gf_strdup (scriptdir); + if (!hooks_stub->scriptdir) + goto out; + + hooks_stub->op_ctx = dict_copy_with_ref (op_ctx, hooks_stub->op_ctx); + if (!hooks_stub->op_ctx) + goto out; + + *stub = hooks_stub; + ret = 0; +out: + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to initialize " + "post hooks stub"); + glusterd_hooks_stub_cleanup (hooks_stub); + } + + return ret; +} + +void +glusterd_hooks_stub_cleanup (glusterd_hooks_stub_t *stub) +{ + if (!stub) { + gf_log_callingfn (THIS->name, GF_LOG_WARNING, + "hooks_stub is NULL"); + return; + } + + if (stub->op_ctx) + dict_unref (stub->op_ctx); + + GF_FREE (stub->scriptdir); + + GF_FREE (stub); +} + +static void* +hooks_worker (void *args) +{ + glusterd_conf_t *conf = NULL; + glusterd_hooks_private_t *hooks_priv = NULL; + glusterd_hooks_stub_t *stub = NULL; + + THIS = args; + conf = THIS->private; + hooks_priv = conf->hooks_priv; + + for (;;) { + pthread_mutex_lock (&hooks_priv->mutex); + { + while (list_empty (&hooks_priv->list)) { + pthread_cond_wait (&hooks_priv->cond, + &hooks_priv->mutex); + } + stub = list_entry (hooks_priv->list.next, + glusterd_hooks_stub_t, + all_hooks); + list_del_init (&stub->all_hooks); + hooks_priv->waitcount--; + + } + pthread_mutex_unlock (&hooks_priv->mutex); + + glusterd_hooks_run_hooks (stub->scriptdir, stub->op, + stub->op_ctx, GD_COMMIT_HOOK_POST); + glusterd_hooks_stub_cleanup (stub); + } + + return NULL; +} + +int +glusterd_hooks_priv_init (glusterd_hooks_private_t **new) +{ + int ret = -1; + glusterd_hooks_private_t *hooks_priv = NULL; + + if (!new) + goto out; + + hooks_priv = GF_CALLOC (1, sizeof (*hooks_priv), + gf_gld_mt_hooks_priv_t); + if (!hooks_priv) + goto out; + + pthread_mutex_init (&hooks_priv->mutex, NULL); + pthread_cond_init (&hooks_priv->cond, NULL); + INIT_LIST_HEAD (&hooks_priv->list); + hooks_priv->waitcount = 0; + + *new = hooks_priv; + ret = 0; +out: + return ret; +} + +int +glusterd_hooks_spawn_worker (xlator_t *this) +{ + int ret = -1; + glusterd_conf_t *conf = NULL; + glusterd_hooks_private_t *hooks_priv = NULL; + + + ret = glusterd_hooks_priv_init (&hooks_priv); + if (ret) + goto out; + + conf = this->private; + conf->hooks_priv = hooks_priv; + ret = pthread_create (&hooks_priv->worker, NULL, hooks_worker, + (void *)this); + if (ret) + gf_log (this->name, GF_LOG_CRITICAL, "Failed to spawn post " + "hooks worker thread"); +out: + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-hooks.h b/xlators/mgmt/glusterd/src/glusterd-hooks.h new file mode 100644 index 000000000..c597ddd2a --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-hooks.h @@ -0,0 +1,89 @@ +/* + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +#ifndef _GLUSTERD_HOOKS_H_ +#define _GLUSTERD_HOOKS_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include <fnmatch.h> + +#define GLUSTERD_GET_HOOKS_DIR(path, version, priv) \ + snprintf (path, PATH_MAX, "%s/hooks/%d", priv->workdir,\ + version); + +#define GLUSTERD_HOOK_VER 1 + +#define GD_HOOKS_SPECIFIC_KEY "user.*" + +typedef enum glusterd_commit_hook_type { + GD_COMMIT_HOOK_NONE = 0, + GD_COMMIT_HOOK_PRE, + GD_COMMIT_HOOK_POST, + GD_COMMIT_HOOK_MAX +} glusterd_commit_hook_type_t; + +typedef struct hooks_private { + struct list_head list; + int waitcount; //debug purposes + pthread_mutex_t mutex; + pthread_cond_t cond; + pthread_t worker; +} glusterd_hooks_private_t; + +typedef struct hooks_stub { + struct list_head all_hooks; + char *scriptdir; + glusterd_op_t op; + dict_t *op_ctx; + +} glusterd_hooks_stub_t; + + +static inline gf_boolean_t +is_key_glusterd_hooks_friendly (char *key) +{ + gf_boolean_t is_friendly = _gf_false; + + /* This is very specific to hooks friendly behavior */ + if (fnmatch (GD_HOOKS_SPECIFIC_KEY, key, FNM_NOESCAPE) == 0) { + gf_log (THIS->name, GF_LOG_DEBUG, "user namespace key %s", key); + is_friendly = _gf_true; + } + + return is_friendly; +} + +int +glusterd_hooks_create_hooks_directory (char *basedir); + +char * +glusterd_hooks_get_hooks_cmd_subdir (glusterd_op_t op); + +int +glusterd_hooks_run_hooks (char *hooks_path, glusterd_op_t op, dict_t *op_ctx, + glusterd_commit_hook_type_t type); +int +glusterd_hooks_spawn_worker (xlator_t *this); + +int +glusterd_hooks_stub_init (glusterd_hooks_stub_t **stub, char *scriptdir, + glusterd_op_t op, dict_t *op_ctx); +void +glusterd_hooks_stub_cleanup (glusterd_hooks_stub_t *stub); + +int +glusterd_hooks_post_stub_enqueue (char *scriptdir, glusterd_op_t op, + dict_t *op_ctx); +int +glusterd_hooks_priv_init (glusterd_hooks_private_t **new); +#endif diff --git a/xlators/mgmt/glusterd/src/glusterd-locks.c b/xlators/mgmt/glusterd/src/glusterd-locks.c new file mode 100644 index 000000000..0af2a186f --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-locks.c @@ -0,0 +1,637 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "common-utils.h" +#include "cli1-xdr.h" +#include "xdr-generic.h" +#include "glusterd.h" +#include "glusterd-op-sm.h" +#include "glusterd-store.h" +#include "glusterd-utils.h" +#include "glusterd-volgen.h" +#include "glusterd-locks.h" +#include "run.h" +#include "syscall.h" + +#include <signal.h> + +#define GF_MAX_LOCKING_ENTITIES 2 + +/* Valid entities that the mgmt_v3 lock can hold locks upon * + * To add newer entities to be locked, we can just add more * + * entries to this table along with the type and default value */ +valid_entities valid_types[] = { + { "vol", _gf_true }, + { "snap", _gf_false }, + { NULL }, +}; + +static dict_t *mgmt_v3_lock; + +/* Checks if the lock request is for a valid entity */ +gf_boolean_t +glusterd_mgmt_v3_is_type_valid (char *type) +{ + int32_t i = 0; + gf_boolean_t ret = _gf_false; + + GF_ASSERT (type); + + for (i = 0; valid_types[i].type; i++) { + if (!strcmp (type, valid_types[i].type)) { + ret = _gf_true; + break; + } + } + + return ret; +} + +/* Initialize the global mgmt_v3 lock list(dict) when + * glusterd is spawned */ +int32_t +glusterd_mgmt_v3_lock_init () +{ + int32_t ret = -1; + + mgmt_v3_lock = dict_new (); + if (!mgmt_v3_lock) { + ret = -1; + goto out; + } + + ret = 0; +out: + return ret; +} + +/* Destroy the global mgmt_v3 lock list(dict) when + * glusterd cleanup is performed */ +void +glusterd_mgmt_v3_lock_fini () +{ + if (mgmt_v3_lock) + dict_destroy (mgmt_v3_lock); +} + +int32_t +glusterd_get_mgmt_v3_lock_owner (char *key, uuid_t *uuid) +{ + int32_t ret = -1; + mgmt_v3_lock_obj *lock_obj = NULL; + uuid_t no_owner = {"\0"}; + xlator_t *this = NULL; + + GF_ASSERT(THIS); + this = THIS; + + if (!key || !uuid) { + gf_log (this->name, GF_LOG_ERROR, "key or uuid is null."); + ret = -1; + goto out; + } + + ret = dict_get_bin(mgmt_v3_lock, key, (void **) &lock_obj); + if (!ret) + uuid_copy (*uuid, lock_obj->lock_owner); + else + uuid_copy (*uuid, no_owner); + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* This function is called with the locked_count and type, to * + * release all the acquired locks. */ +static int32_t +glusterd_release_multiple_locks_per_entity (dict_t *dict, uuid_t uuid, + int32_t locked_count, + char *type) +{ + char name_buf[PATH_MAX] = ""; + char *name = NULL; + int32_t i = -1; + int32_t op_ret = 0; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT (dict); + GF_ASSERT (type); + + if (locked_count == 0) { + gf_log (this->name, GF_LOG_DEBUG, + "No %s locked as part of this transaction", + type); + goto out; + } + + /* Release all the locks held */ + for (i = 0; i < locked_count; i++) { + snprintf (name_buf, sizeof(name_buf), + "%sname%d", type, i+1); + + /* Looking for volname1, volname2 or snapname1, * + * as key in the dict snapname2 */ + ret = dict_get_str (dict, name_buf, &name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get %s locked_count = %d", + name_buf, locked_count); + op_ret = ret; + continue; + } + + ret = glusterd_mgmt_v3_unlock (name, uuid, type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release lock for %s.", + name); + op_ret = ret; + } + } + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", op_ret); + return op_ret; +} + +/* Given the count and type of the entity this function acquires * + * locks on multiple elements of the same entity. For example: * + * If type is "vol" this function tries to acquire locks on multiple * + * volumes */ +static int32_t +glusterd_acquire_multiple_locks_per_entity (dict_t *dict, uuid_t uuid, + int32_t count, char *type) +{ + char name_buf[PATH_MAX] = ""; + char *name = NULL; + int32_t i = -1; + int32_t ret = -1; + int32_t locked_count = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT (dict); + GF_ASSERT (type); + + /* Locking one element after other */ + for (i = 0; i < count; i++) { + snprintf (name_buf, sizeof(name_buf), + "%sname%d", type, i+1); + + /* Looking for volname1, volname2 or snapname1, * + * as key in the dict snapname2 */ + ret = dict_get_str (dict, name_buf, &name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get %s count = %d", + name_buf, count); + break; + } + + ret = glusterd_mgmt_v3_lock (name, uuid, type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire lock for %s %s " + "on behalf of %s. Reversing " + "this transaction", type, name, + uuid_utoa(uuid)); + break; + } + locked_count++; + } + + if (count == locked_count) { + /* If all locking ops went successfuly, return as success */ + ret = 0; + goto out; + } + + /* If we failed to lock one element, unlock others and return failure */ + ret = glusterd_release_multiple_locks_per_entity (dict, uuid, + locked_count, + type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release multiple %s locks", + type); + } + ret = -1; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Given the type of entity, this function figures out if it should unlock a * + * single element of multiple elements of the said entity. For example: * + * if the type is "vol", this function will accordingly unlock a single volume * + * or multiple volumes */ +static int32_t +glusterd_mgmt_v3_unlock_entity (dict_t *dict, uuid_t uuid, char *type, + gf_boolean_t default_value) +{ + char name_buf[PATH_MAX] = ""; + char *name = NULL; + int32_t count = -1; + int32_t ret = -1; + gf_boolean_t hold_locks = _gf_false; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT (dict); + GF_ASSERT (type); + + snprintf (name_buf, sizeof(name_buf), "hold_%s_locks", type); + hold_locks = dict_get_str_boolean (dict, name_buf, default_value); + + if (hold_locks == _gf_false) { + /* Locks were not held for this particular entity * + * Hence nothing to release */ + ret = 0; + goto out; + } + + /* Looking for volcount or snapcount in the dict */ + snprintf (name_buf, sizeof(name_buf), "%scount", type); + ret = dict_get_int32 (dict, name_buf, &count); + if (ret) { + /* count is not present. Only one * + * element name needs to be unlocked */ + snprintf (name_buf, sizeof(name_buf), "%sname", + type); + ret = dict_get_str (dict, name_buf, &name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch %sname", type); + goto out; + } + + ret = glusterd_mgmt_v3_unlock (name, uuid, type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release lock for %s %s " + "on behalf of %s.", type, name, + uuid_utoa(uuid)); + goto out; + } + } else { + /* Unlocking one element name after another */ + ret = glusterd_release_multiple_locks_per_entity (dict, + uuid, + count, + type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release all %s locks", type); + goto out; + } + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Given the type of entity, this function figures out if it should lock a * + * single element or multiple elements of the said entity. For example: * + * if the type is "vol", this function will accordingly lock a single volume * + * or multiple volumes */ +static int32_t +glusterd_mgmt_v3_lock_entity (dict_t *dict, uuid_t uuid, char *type, + gf_boolean_t default_value) +{ + char name_buf[PATH_MAX] = ""; + char *name = NULL; + int32_t count = -1; + int32_t ret = -1; + gf_boolean_t hold_locks = _gf_false; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT (dict); + GF_ASSERT (type); + + snprintf (name_buf, sizeof(name_buf), "hold_%s_locks", type); + hold_locks = dict_get_str_boolean (dict, name_buf, default_value); + + if (hold_locks == _gf_false) { + /* Not holding locks for this particular entity */ + ret = 0; + goto out; + } + + /* Looking for volcount or snapcount in the dict */ + snprintf (name_buf, sizeof(name_buf), "%scount", type); + ret = dict_get_int32 (dict, name_buf, &count); + if (ret) { + /* count is not present. Only one * + * element name needs to be locked */ + snprintf (name_buf, sizeof(name_buf), "%sname", + type); + ret = dict_get_str (dict, name_buf, &name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch %sname", type); + goto out; + } + + ret = glusterd_mgmt_v3_lock (name, uuid, type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire lock for %s %s " + "on behalf of %s.", type, name, + uuid_utoa(uuid)); + goto out; + } + } else { + /* Locking one element name after another */ + ret = glusterd_acquire_multiple_locks_per_entity (dict, + uuid, + count, + type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire all %s locks", type); + goto out; + } + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Try to release locks of multiple entities like * + * volume, snaps etc. */ +int32_t +glusterd_multiple_mgmt_v3_unlock (dict_t *dict, uuid_t uuid) +{ + int32_t i = -1; + int32_t ret = -1; + int32_t op_ret = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + + if (!dict) { + gf_log (this->name, GF_LOG_ERROR, "dict is null."); + ret = -1; + goto out; + } + + for (i = 0; valid_types[i].type; i++) { + ret = glusterd_mgmt_v3_unlock_entity + (dict, uuid, + valid_types[i].type, + valid_types[i].default_value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to unlock all %s", + valid_types[i].type); + op_ret = ret; + } + } + + ret = op_ret; +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +/* Try to acquire locks on multiple entities like * + * volume, snaps etc. */ +int32_t +glusterd_multiple_mgmt_v3_lock (dict_t *dict, uuid_t uuid) +{ + int32_t i = -1; + int32_t ret = -1; + int32_t locked_count = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + + if (!dict) { + gf_log (this->name, GF_LOG_ERROR, "dict is null."); + ret = -1; + goto out; + } + + /* Locking one entity after other */ + for (i = 0; valid_types[i].type; i++) { + ret = glusterd_mgmt_v3_lock_entity + (dict, uuid, + valid_types[i].type, + valid_types[i].default_value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to lock all %s", + valid_types[i].type); + break; + } + locked_count++; + } + + if (locked_count == GF_MAX_LOCKING_ENTITIES) { + /* If all locking ops went successfuly, return as success */ + ret = 0; + goto out; + } + + /* If we failed to lock one entity, unlock others and return failure */ + for (i = 0; i < locked_count; i++) { + ret = glusterd_mgmt_v3_unlock_entity + (dict, uuid, + valid_types[i].type, + valid_types[i].default_value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to unlock all %s", + valid_types[i].type); + } + } + ret = -1; +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_mgmt_v3_lock (const char *name, uuid_t uuid, char *type) +{ + char key[PATH_MAX] = ""; + int32_t ret = -1; + mgmt_v3_lock_obj *lock_obj = NULL; + gf_boolean_t is_valid = _gf_true; + uuid_t owner = {0}; + xlator_t *this = NULL; + + GF_ASSERT(THIS); + this = THIS; + + if (!name || !type) { + gf_log (this->name, GF_LOG_ERROR, "name or type is null."); + ret = -1; + goto out; + } + + is_valid = glusterd_mgmt_v3_is_type_valid (type); + if (is_valid != _gf_true) { + gf_log (this->name, GF_LOG_ERROR, + "Invalid entity. Cannot perform locking " + "operation on %s types", type); + ret = -1; + goto out; + } + + ret = snprintf (key, sizeof(key), "%s_%s", name, type); + if (ret != strlen(name) + 1 + strlen(type)) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Unable to create key"); + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, + "Trying to acquire lock of %s %s for %s as %s", + type, name, uuid_utoa (uuid), key); + + ret = glusterd_get_mgmt_v3_lock_owner (key, &owner); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Unable to get mgmt_v3 lock owner"); + goto out; + } + + /* If the lock has already been held for the given volume + * we fail */ + if (!uuid_is_null (owner)) { + gf_log (this->name, GF_LOG_ERROR, "Lock for %s held by %s", + name, uuid_utoa (owner)); + ret = -1; + goto out; + } + + lock_obj = GF_CALLOC (1, sizeof(mgmt_v3_lock_obj), + gf_common_mt_mgmt_v3_lock_obj_t); + if (!lock_obj) { + ret = -1; + goto out; + } + + uuid_copy (lock_obj->lock_owner, uuid); + + ret = dict_set_bin (mgmt_v3_lock, key, lock_obj, + sizeof(*lock_obj)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set lock owner in mgmt_v3 lock"); + if (lock_obj) + GF_FREE (lock_obj); + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, + "Lock for %s %s successfully held by %s", + type, name, uuid_utoa (uuid)); + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_mgmt_v3_unlock (const char *name, uuid_t uuid, char *type) +{ + char key[PATH_MAX] = ""; + int32_t ret = -1; + gf_boolean_t is_valid = _gf_true; + uuid_t owner = {0}; + xlator_t *this = NULL; + + GF_ASSERT(THIS); + this = THIS; + + if (!name || !type) { + gf_log (this->name, GF_LOG_ERROR, "name is null."); + ret = -1; + goto out; + } + + is_valid = glusterd_mgmt_v3_is_type_valid (type); + if (is_valid != _gf_true) { + gf_log (this->name, GF_LOG_ERROR, + "Invalid entity. Cannot perform unlocking " + "operation on %s types", type); + ret = -1; + goto out; + } + + ret = snprintf (key, sizeof(key), "%s_%s", + name, type); + if (ret != strlen(name) + 1 + strlen(type)) { + gf_log (this->name, GF_LOG_ERROR, "Unable to create key"); + ret = -1; + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, + "Trying to release lock of %s %s for %s as %s", + type, name, uuid_utoa (uuid), key); + + ret = glusterd_get_mgmt_v3_lock_owner (key, &owner); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Unable to get mgmt_v3 lock owner"); + goto out; + } + + if (uuid_is_null (owner)) { + gf_log (this->name, GF_LOG_ERROR, + "Lock for %s %s not held", type, name); + ret = -1; + goto out; + } + + ret = uuid_compare (uuid, owner); + if (ret) { + + gf_log (this->name, GF_LOG_ERROR, "Lock owner mismatch. " + "Lock for %s %s held by %s", + type, name, uuid_utoa (owner)); + goto out; + } + + /* Removing the mgmt_v3 lock from the global list */ + dict_del (mgmt_v3_lock, key); + + gf_log (this->name, GF_LOG_DEBUG, + "Lock for %s %s successfully released", + type, name); + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-locks.h b/xlators/mgmt/glusterd/src/glusterd-locks.h new file mode 100644 index 000000000..83eb8c997 --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-locks.h @@ -0,0 +1,51 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +#ifndef _GLUSTERD_LOCKS_H_ +#define _GLUSTERD_LOCKS_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +typedef struct mgmt_v3_lock_object_ { + uuid_t lock_owner; +} mgmt_v3_lock_obj; + +typedef struct mgmt_v3_lock_valid_entities { + char *type; /* Entity type like vol, snap */ + gf_boolean_t default_value; /* The default value that * + * determines if the locks * + * should be held for that * + * entity */ +} valid_entities; + +int32_t +glusterd_mgmt_v3_lock_init (); + +void +glusterd_mgmt_v3_lock_fini (); + +int32_t +glusterd_get_mgmt_v3_lock_owner (char *volname, uuid_t *uuid); + +int32_t +glusterd_mgmt_v3_lock (const char *key, uuid_t uuid, char *type); + +int32_t +glusterd_mgmt_v3_unlock (const char *key, uuid_t uuid, char *type); + +int32_t +glusterd_multiple_mgmt_v3_lock (dict_t *dict, uuid_t uuid); + +int32_t +glusterd_multiple_mgmt_v3_unlock (dict_t *dict, uuid_t uuid); + +#endif diff --git a/xlators/mgmt/glusterd/src/glusterd-log-ops.c b/xlators/mgmt/glusterd/src/glusterd-log-ops.c index ec8889670..33bd95c03 100644 --- a/xlators/mgmt/glusterd/src/glusterd-log-ops.c +++ b/xlators/mgmt/glusterd/src/glusterd-log-ops.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" @@ -34,18 +24,22 @@ #include <signal.h> int -glusterd_handle_log_rotate (rpcsvc_request_t *req) +__glusterd_handle_log_rotate (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; dict_t *dict = NULL; glusterd_op_t cli_op = GD_OP_LOG_ROTATE; char *volname = NULL; + char msg[2048] = {0,}; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -59,44 +53,50 @@ glusterd_handle_log_rotate (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (msg, sizeof (msg), "Unable to decode the " + "command"); goto out; } } ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to get volname"); + snprintf (msg, sizeof (msg), "Failed to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); goto out; } - gf_log ("glusterd", GF_LOG_INFO, "Received log rotate req " + gf_log (this->name, GF_LOG_INFO, "Received log rotate req " "for volume %s", volname); ret = dict_set_uint64 (dict, "rotate-key", (uint64_t)time (NULL)); if (ret) goto out; - ret = glusterd_op_begin (req, GD_OP_LOG_ROTATE, dict); + ret = glusterd_op_begin_synctask (req, GD_OP_LOG_ROTATE, dict); out: - if (ret && dict) - dict_unref (dict); - - glusterd_friend_sm (); - glusterd_op_sm (); - - if (ret) + if (ret) { + if (msg[0] == '\0') + snprintf (msg, sizeof (msg), "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, msg); + } - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val); + free (cli_req.dict.dict_val); return ret; } +int +glusterd_handle_log_rotate (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_log_rotate); +} + /* op-sm */ int glusterd_op_stage_log_rotate (dict_t *dict, char **op_errstr) @@ -135,19 +135,20 @@ glusterd_op_stage_log_rotate (dict_t *dict, char **op_errstr) } ret = dict_get_str (dict, "brick", &brick); - if (ret) + /* If no brick is specified, do log-rotate for + all the bricks in the volume */ + if (ret) { + ret = 0; goto out; + } - if (strchr (brick, ':')) { - ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, NULL, - GF_PATH_COMPLETE); - if (ret) { - snprintf (msg, sizeof (msg), "Incorrect brick %s " - "for volume %s", brick, volname); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - goto out; - } + ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, NULL); + if (ret) { + snprintf (msg, sizeof (msg), "Incorrect brick %s " + "for volume %s", brick, volname); + gf_log ("", GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + goto out; } out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); @@ -166,7 +167,6 @@ glusterd_op_log_rotate (dict_t *dict) xlator_t *this = NULL; char *volname = NULL; char *brick = NULL; - char path[PATH_MAX] = {0,}; char logfile[PATH_MAX] = {0,}; char pidfile[PATH_MAX] = {0,}; FILE *file = NULL; @@ -193,27 +193,26 @@ glusterd_op_log_rotate (dict_t *dict) } ret = dict_get_str (dict, "brick", &brick); + /* If no brick is specified, do log-rotate for + all the bricks in the volume */ if (ret) - goto out; + goto cont; - if (!strchr (brick, ':')) - brick = NULL; - else { - ret = glusterd_brickinfo_from_brick (brick, &tmpbrkinfo); - if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, - "cannot get brickinfo from brick"); - goto out; - } + ret = glusterd_brickinfo_new_from_brick (brick, &tmpbrkinfo); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, + "cannot get brickinfo from brick"); + goto out; } +cont: ret = glusterd_volinfo_find (volname, &volinfo); if (ret) goto out; ret = -1; list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { - if (uuid_compare (brickinfo->uuid, priv->uuid)) + if (uuid_compare (brickinfo->uuid, MY_UUID)) continue; if (brick && @@ -223,10 +222,7 @@ glusterd_op_log_rotate (dict_t *dict) valid_brick = 1; - GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv); - GLUSTERD_GET_BRICK_PIDFILE (pidfile, path, brickinfo->hostname, - brickinfo->path); - + GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, brickinfo, priv); file = fopen (pidfile, "r+"); if (!file) { gf_log ("", GF_LOG_ERROR, "Unable to open pidfile: %s", diff --git a/xlators/mgmt/glusterd/src/glusterd-mem-types.h b/xlators/mgmt/glusterd/src/glusterd-mem-types.h index 196f5f50b..e6f6a0333 100644 --- a/xlators/mgmt/glusterd/src/glusterd-mem-types.h +++ b/xlators/mgmt/glusterd/src/glusterd-mem-types.h @@ -1,23 +1,13 @@ /* - Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com> + Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com> This file is part of GlusterFS. - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. */ - #ifndef __GLUSTERD_MEM_TYPES_H__ #define __GLUSTERD_MEM_TYPES_H__ @@ -70,8 +60,16 @@ typedef enum gf_gld_mem_types_ { gf_gld_mt_mount_comp_container = gf_common_mt_end + 44, gf_gld_mt_mount_component = gf_common_mt_end + 45, gf_gld_mt_mount_spec = gf_common_mt_end + 46, - gf_gld_mt_nodesrv_t = gf_common_mt_end + 47, - gf_gld_mt_end = gf_common_mt_end + 48, + gf_gld_mt_georep_meet_spec = gf_common_mt_end + 47, + gf_gld_mt_nodesrv_t = gf_common_mt_end + 48, + gf_gld_mt_charptr = gf_common_mt_end + 49, + gf_gld_mt_hooks_stub_t = gf_common_mt_end + 50, + gf_gld_mt_hooks_priv_t = gf_common_mt_end + 51, + gf_gld_mt_mop_commit_req_t = gf_common_mt_end + 52, + gf_gld_mt_int = gf_common_mt_end + 53, + gf_gld_mt_snap_t = gf_common_mt_end + 54, + gf_gld_mt_missed_snapinfo_t = gf_common_mt_end + 55, + gf_gld_mt_end = gf_common_mt_end + 56, } gf_gld_mem_types_t; #endif diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c b/xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c new file mode 100644 index 000000000..a5f38ce9c --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c @@ -0,0 +1,924 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +/* rpc related syncops */ +#include "rpc-clnt.h" +#include "protocol-common.h" +#include "xdr-generic.h" +#include "glusterd1-xdr.h" +#include "glusterd-syncop.h" + +#include "glusterd.h" +#include "glusterd-utils.h" +#include "glusterd-locks.h" +#include "glusterd-mgmt.h" +#include "glusterd-op-sm.h" + +static int +glusterd_mgmt_v3_null (rpcsvc_request_t *req) +{ + return 0; +} + +static int +glusterd_mgmt_v3_lock_send_resp (rpcsvc_request_t *req, int32_t status) +{ + + gd1_mgmt_v3_lock_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + if (rsp.op_ret) + rsp.op_errno = errno; + + glusterd_get_uuid (&rsp.uuid); + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp); + + gf_log (this->name, GF_LOG_DEBUG, + "Responded to mgmt_v3 lock, ret: %d", ret); + + return ret; +} + +static int +glusterd_synctasked_mgmt_v3_lock (rpcsvc_request_t *req, + gd1_mgmt_v3_lock_req *lock_req, + glusterd_op_lock_ctx_t *ctx) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (ctx); + GF_ASSERT (ctx->dict); + + /* Trying to acquire multiple mgmt_v3 locks */ + ret = glusterd_multiple_mgmt_v3_lock (ctx->dict, ctx->uuid); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire mgmt_v3 locks for %s", + uuid_utoa (ctx->uuid)); + + ret = glusterd_mgmt_v3_lock_send_resp (req, ret); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_op_state_machine_mgmt_v3_lock (rpcsvc_request_t *req, + gd1_mgmt_v3_lock_req *lock_req, + glusterd_op_lock_ctx_t *ctx) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_op_info_t txn_op_info = {{0},}; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + glusterd_txn_opinfo_init (&txn_op_info, NULL, &lock_req->op, + ctx->dict, req); + + ret = glusterd_set_txn_opinfo (&lock_req->txn_id, &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set transaction's opinfo"); + goto out; + } + + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_LOCK, + &lock_req->txn_id, ctx); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to inject event GD_OP_EVENT_LOCK"); + +out: + glusterd_friend_sm (); + glusterd_op_sm (); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_handle_mgmt_v3_lock_fn (rpcsvc_request_t *req) +{ + gd1_mgmt_v3_lock_req lock_req = {{0},}; + int32_t ret = -1; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_op_lock_ctx_t *ctx = NULL; + xlator_t *this = NULL; + gf_boolean_t is_synctasked = _gf_false; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &lock_req, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode lock " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, "Received mgmt_v3 lock req " + "from uuid: %s", uuid_utoa (lock_req.uuid)); + + if (glusterd_friend_find_by_uuid (lock_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (lock_req.uuid)); + ret = -1; + goto out; + } + + ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_lock_ctx_t); + if (!ctx) { + ret = -1; + goto out; + } + + uuid_copy (ctx->uuid, lock_req.uuid); + ctx->req = req; + + ctx->dict = dict_new (); + if (!ctx->dict) { + ret = -1; + goto out; + } + + ret = dict_unserialize (lock_req.dict.dict_val, + lock_req.dict.dict_len, &ctx->dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + is_synctasked = dict_get_str_boolean (ctx->dict, + "is_synctasked", _gf_false); + if (is_synctasked) + ret = glusterd_synctasked_mgmt_v3_lock (req, &lock_req, ctx); + else + ret = glusterd_op_state_machine_mgmt_v3_lock (req, &lock_req, + ctx); + +out: + + if (ret) { + if (ctx->dict) + dict_unref (ctx->dict); + if (ctx) + GF_FREE (ctx); + } + + free (lock_req.dict.dict_val); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_mgmt_v3_pre_validate_send_resp (rpcsvc_request_t *req, + int32_t op, int32_t status, + char *op_errstr, dict_t *rsp_dict) +{ + gd1_mgmt_v3_pre_val_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + glusterd_get_uuid (&rsp.uuid); + rsp.op = op; + if (op_errstr) + rsp.op_errstr = op_errstr; + else + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get serialized length of dict"); + goto out; + } + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_pre_val_rsp); + + GF_FREE (rsp.dict.dict_val); +out: + gf_log (this->name, GF_LOG_DEBUG, + "Responded to pre validation, ret: %d", ret); + return ret; +} + +static int +glusterd_handle_pre_validate_fn (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_v3_pre_val_req op_req = {{0},}; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + char *op_errstr = NULL; + dict_t *dict = NULL; + dict_t *rsp_dict = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &op_req, + (xdrproc_t)xdr_gd1_mgmt_v3_pre_val_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to decode pre validation " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (op_req.uuid)); + ret = -1; + goto out; + } + + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (op_req.dict.dict_val, + op_req.dict.dict_len, &dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get new dictionary"); + return -1; + } + + ret = gd_mgmt_v3_pre_validate_fn (op_req.op, dict, &op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Pre Validation failed on operation %s", + gd_op_list[op_req.op]); + } + + ret = glusterd_mgmt_v3_pre_validate_send_resp (req, op_req.op, + ret, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to send Pre Validation " + "response for operation %s", + gd_op_list[op_req.op]); + goto out; + } + +out: + if (op_errstr && (strcmp (op_errstr, ""))) + GF_FREE (op_errstr); + + free (op_req.dict.dict_val); + + if (dict) + dict_unref (dict); + + if (rsp_dict) + dict_unref (rsp_dict); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_mgmt_v3_brick_op_send_resp (rpcsvc_request_t *req, + int32_t op, int32_t status, + char *op_errstr, dict_t *rsp_dict) +{ + gd1_mgmt_v3_brick_op_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + glusterd_get_uuid (&rsp.uuid); + rsp.op = op; + if (op_errstr) + rsp.op_errstr = op_errstr; + else + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get serialized length of dict"); + goto out; + } + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_brick_op_rsp); + + GF_FREE (rsp.dict.dict_val); +out: + gf_log (this->name, GF_LOG_DEBUG, + "Responded to brick op, ret: %d", ret); + return ret; +} + +static int +glusterd_handle_brick_op_fn (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_v3_brick_op_req op_req = {{0},}; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + char *op_errstr = NULL; + dict_t *dict = NULL; + dict_t *rsp_dict = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &op_req, + (xdrproc_t)xdr_gd1_mgmt_v3_brick_op_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode brick op " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (op_req.uuid)); + ret = -1; + goto out; + } + + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (op_req.dict.dict_val, + op_req.dict.dict_len, &dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get new dictionary"); + return -1; + } + + ret = gd_mgmt_v3_brick_op_fn (op_req.op, dict, &op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Brick Op failed on operation %s", + gd_op_list[op_req.op]); + } + + ret = glusterd_mgmt_v3_brick_op_send_resp (req, op_req.op, + ret, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to send brick op " + "response for operation %s", + gd_op_list[op_req.op]); + goto out; + } + +out: + if (op_errstr && (strcmp (op_errstr, ""))) + GF_FREE (op_errstr); + + free (op_req.dict.dict_val); + + if (dict) + dict_unref (dict); + + if (rsp_dict) + dict_unref (rsp_dict); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_mgmt_v3_commit_send_resp (rpcsvc_request_t *req, + int32_t op, int32_t status, + char *op_errstr, dict_t *rsp_dict) +{ + gd1_mgmt_v3_commit_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + glusterd_get_uuid (&rsp.uuid); + rsp.op = op; + if (op_errstr) + rsp.op_errstr = op_errstr; + else + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get serialized length of dict"); + goto out; + } + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_commit_rsp); + + GF_FREE (rsp.dict.dict_val); +out: + gf_log (this->name, GF_LOG_DEBUG, "Responded to commit, ret: %d", ret); + return ret; +} + +static int +glusterd_handle_commit_fn (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_v3_commit_req op_req = {{0},}; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + char *op_errstr = NULL; + dict_t *dict = NULL; + dict_t *rsp_dict = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &op_req, + (xdrproc_t)xdr_gd1_mgmt_v3_commit_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode commit " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (op_req.uuid)); + ret = -1; + goto out; + } + + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (op_req.dict.dict_val, + op_req.dict.dict_len, &dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get new dictionary"); + return -1; + } + + ret = gd_mgmt_v3_commit_fn (op_req.op, dict, &op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "commit failed on operation %s", + gd_op_list[op_req.op]); + } + + ret = glusterd_mgmt_v3_commit_send_resp (req, op_req.op, + ret, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to send commit " + "response for operation %s", + gd_op_list[op_req.op]); + goto out; + } + +out: + if (op_errstr && (strcmp (op_errstr, ""))) + GF_FREE (op_errstr); + + free (op_req.dict.dict_val); + + if (dict) + dict_unref (dict); + + if (rsp_dict) + dict_unref (rsp_dict); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_mgmt_v3_post_validate_send_resp (rpcsvc_request_t *req, + int32_t op, int32_t status, + char *op_errstr, dict_t *rsp_dict) +{ + gd1_mgmt_v3_post_val_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + glusterd_get_uuid (&rsp.uuid); + rsp.op = op; + if (op_errstr) + rsp.op_errstr = op_errstr; + else + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get serialized length of dict"); + goto out; + } + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_post_val_rsp); + + GF_FREE (rsp.dict.dict_val); +out: + gf_log (this->name, GF_LOG_DEBUG, + "Responded to post validation, ret: %d", ret); + return ret; +} + +static int +glusterd_handle_post_validate_fn (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_v3_post_val_req op_req = {{0},}; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + char *op_errstr = NULL; + dict_t *dict = NULL; + dict_t *rsp_dict = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &op_req, + (xdrproc_t)xdr_gd1_mgmt_v3_post_val_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to decode post validation " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (op_req.uuid)); + ret = -1; + goto out; + } + + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (op_req.dict.dict_val, + op_req.dict.dict_len, &dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get new dictionary"); + return -1; + } + + ret = gd_mgmt_v3_post_validate_fn (op_req.op, op_req.op_ret, dict, + &op_errstr, rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Post Validation failed on operation %s", + gd_op_list[op_req.op]); + } + + ret = glusterd_mgmt_v3_post_validate_send_resp (req, op_req.op, + ret, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to send Post Validation " + "response for operation %s", + gd_op_list[op_req.op]); + goto out; + } + +out: + if (op_errstr && (strcmp (op_errstr, ""))) + GF_FREE (op_errstr); + + free (op_req.dict.dict_val); + + if (dict) + dict_unref (dict); + + if (rsp_dict) + dict_unref (rsp_dict); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_mgmt_v3_unlock_send_resp (rpcsvc_request_t *req, int32_t status) +{ + + gd1_mgmt_v3_unlock_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + if (rsp.op_ret) + rsp.op_errno = errno; + + glusterd_get_uuid (&rsp.uuid); + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp); + + gf_log (this->name, GF_LOG_DEBUG, + "Responded to mgmt_v3 unlock, ret: %d", ret); + + return ret; +} + +static int +glusterd_synctasked_mgmt_v3_unlock (rpcsvc_request_t *req, + gd1_mgmt_v3_unlock_req *unlock_req, + glusterd_op_lock_ctx_t *ctx) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (ctx); + + /* Trying to release multiple mgmt_v3 locks */ + ret = glusterd_multiple_mgmt_v3_unlock (ctx->dict, ctx->uuid); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release mgmt_v3 locks for %s", + uuid_utoa(ctx->uuid)); + } + + ret = glusterd_mgmt_v3_unlock_send_resp (req, ret); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + + +static int +glusterd_op_state_machine_mgmt_v3_unlock (rpcsvc_request_t *req, + gd1_mgmt_v3_unlock_req *lock_req, + glusterd_op_lock_ctx_t *ctx) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_UNLOCK, + &lock_req->txn_id, ctx); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to inject event GD_OP_EVENT_UNLOCK"); + + glusterd_friend_sm (); + glusterd_op_sm (); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_handle_mgmt_v3_unlock_fn (rpcsvc_request_t *req) +{ + gd1_mgmt_v3_unlock_req lock_req = {{0},}; + int32_t ret = -1; + glusterd_op_lock_ctx_t *ctx = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + gf_boolean_t is_synctasked = _gf_false; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &lock_req, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode unlock " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, "Received mgmt_v3 unlock req " + "from uuid: %s", uuid_utoa (lock_req.uuid)); + + if (glusterd_friend_find_by_uuid (lock_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (lock_req.uuid)); + ret = -1; + goto out; + } + + ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_lock_ctx_t); + if (!ctx) { + ret = -1; + goto out; + } + + uuid_copy (ctx->uuid, lock_req.uuid); + ctx->req = req; + + ctx->dict = dict_new (); + if (!ctx->dict) { + ret = -1; + goto out; + } + + ret = dict_unserialize (lock_req.dict.dict_val, + lock_req.dict.dict_len, &ctx->dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + is_synctasked = dict_get_str_boolean (ctx->dict, + "is_synctasked", _gf_false); + if (is_synctasked) + ret = glusterd_synctasked_mgmt_v3_unlock (req, &lock_req, ctx); + else + ret = glusterd_op_state_machine_mgmt_v3_unlock (req, &lock_req, + ctx); + +out: + + if (ret) { + if (ctx->dict) + dict_unref (ctx->dict); + if (ctx) + GF_FREE (ctx); + } + + free (lock_req.dict.dict_val); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_handle_mgmt_v3_lock (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_mgmt_v3_lock_fn); +} + +static int +glusterd_handle_pre_validate (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_pre_validate_fn); +} + +static int +glusterd_handle_brick_op (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_brick_op_fn); +} + +static int +glusterd_handle_commit (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_commit_fn); +} + +static int +glusterd_handle_post_validate (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_post_validate_fn); +} + +int +glusterd_handle_mgmt_v3_unlock (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_mgmt_v3_unlock_fn); +} + +rpcsvc_actor_t gd_svc_mgmt_v3_actors[] = { + [GLUSTERD_MGMT_V3_NULL] = { "NULL", GLUSTERD_MGMT_V3_NULL, glusterd_mgmt_v3_null, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_LOCK] = { "MGMT_V3_LOCK", GLUSTERD_MGMT_V3_LOCK, glusterd_handle_mgmt_v3_lock, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_PRE_VALIDATE] = { "PRE_VAL", GLUSTERD_MGMT_V3_PRE_VALIDATE, glusterd_handle_pre_validate, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_BRICK_OP] = { "BRCK_OP", GLUSTERD_MGMT_V3_BRICK_OP, glusterd_handle_brick_op, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_COMMIT] = { "COMMIT", GLUSTERD_MGMT_V3_COMMIT, glusterd_handle_commit, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_POST_VALIDATE] = { "POST_VAL", GLUSTERD_MGMT_V3_POST_VALIDATE, glusterd_handle_post_validate, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_UNLOCK] = { "MGMT_V3_UNLOCK", GLUSTERD_MGMT_V3_UNLOCK, glusterd_handle_mgmt_v3_unlock, NULL, 0, DRC_NA}, +}; + +struct rpcsvc_program gd_svc_mgmt_v3_prog = { + .progname = "GlusterD svc mgmt v3", + .prognum = GD_MGMT_V3_PROGRAM, + .progver = GD_MGMT_V3_VERSION, + .numactors = GLUSTERD_MGMT_V3_MAXVALUE, + .actors = gd_svc_mgmt_v3_actors, + .synctask = _gf_true, +}; diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt.c b/xlators/mgmt/glusterd/src/glusterd-mgmt.c new file mode 100644 index 000000000..d52532e54 --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-mgmt.c @@ -0,0 +1,1893 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +/* rpc related syncops */ +#include "rpc-clnt.h" +#include "protocol-common.h" +#include "xdr-generic.h" +#include "glusterd1-xdr.h" +#include "glusterd-syncop.h" + +#include "glusterd.h" +#include "glusterd-utils.h" +#include "glusterd-locks.h" +#include "glusterd-mgmt.h" +#include "glusterd-op-sm.h" + +extern struct rpc_clnt_program gd_mgmt_v3_prog; + + +static void +gd_mgmt_v3_collate_errors (struct syncargs *args, int op_ret, int op_errno, + char *op_errstr, int op_code, + glusterd_peerinfo_t *peerinfo, u_char *uuid) +{ + char *peer_str = NULL; + char err_str[PATH_MAX] = "Please check log file for details."; + char op_err[PATH_MAX] = ""; + int32_t len = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (args); + GF_ASSERT (uuid); + + if (op_ret) { + args->op_ret = op_ret; + args->op_errno = op_errno; + + if (peerinfo) + peer_str = peerinfo->hostname; + else + peer_str = uuid_utoa (uuid); + + if (op_errstr && strcmp (op_errstr, "")) { + len = snprintf (err_str, sizeof(err_str) - 1, + "Error: %s", op_errstr); + err_str[len] = '\0'; + } + + switch (op_code){ + case GLUSTERD_MGMT_V3_LOCK: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Locking failed " + "on %s. %s", peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_PRE_VALIDATE: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Pre Validation failed " + "on %s. %s", peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_BRICK_OP: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Brick ops failed " + "on %s. %s", peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_COMMIT: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Commit failed on %s. %s", + peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_POST_VALIDATE: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Post Validation failed " + "on %s. %s", peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_UNLOCK: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Unlocking failed " + "on %s. %s", peer_str, err_str); + break; + } + } + op_err[len] = '\0'; + + if (args->errstr) { + len = snprintf (err_str, sizeof(err_str) - 1, + "%s\n%s", args->errstr, + op_err); + GF_FREE (args->errstr); + args->errstr = NULL; + } else + len = snprintf (err_str, sizeof(err_str) - 1, + "%s", op_err); + err_str[len] = '\0'; + + gf_log (this->name, GF_LOG_ERROR, "%s", op_err); + args->errstr = gf_strdup (err_str); + } + + return; +} + +int32_t +gd_mgmt_v3_pre_validate_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + switch (op) { + case GD_OP_SNAP: + { + ret = glusterd_snapshot_prevalidate (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Snapshot Prevalidate Failed"); + goto out; + } + + break; + } + default: + break; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_DEBUG, "OP = %d. Returning %d", op, ret); + return ret; +} + +int32_t +gd_mgmt_v3_brick_op_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + switch (op) { + case GD_OP_SNAP: + { + ret = glusterd_snapshot_brickop (dict, op_errstr, rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "snapshot brickop " + "failed"); + goto out; + } + break; + } + default: + break; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "OP = %d. Returning %d", op, ret); + return ret; +} + +int32_t +gd_mgmt_v3_commit_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + switch (op) { + case GD_OP_SNAP: + { + ret = glusterd_snapshot (dict, op_errstr, rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Snapshot Commit Failed"); + goto out; + } + break; + } + default: + break; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_DEBUG, "OP = %d. Returning %d", op, ret); + return ret; +} + +int32_t +gd_mgmt_v3_post_validate_fn (glusterd_op_t op, int32_t op_ret, dict_t *dict, + char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + switch (op) { + case GD_OP_SNAP: + { + ret = glusterd_snapshot_postvalidate (dict, op_ret, + op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "postvalidate operation failed"); + goto out; + } + break; + } + default: + break; + } + + ret = 0; + +out: + gf_log (this->name, GF_LOG_TRACE, "OP = %d. Returning %d", op, ret); + return ret; +} + +int32_t +gd_mgmt_v3_lock_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_lock_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + /* Even though the lock command has failed, while collating the errors + (gd_mgmt_v3_collate_errors), args->op_ret and args->op_errno will be + used. @args is obtained from frame->local. So before checking the + status of the request and going out if its a failure, args should be + set to frame->local. Otherwise, while collating args will be NULL. + This applies to other phases such as prevalidate, brickop, commit and + postvalidate also. + */ + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_LOCK, + peerinfo, rsp.uuid); + free (rsp.dict.dict_val); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_lock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_lock_cbk_fn); +} + +int +gd_mgmt_v3_lock (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + gd1_mgmt_v3_lock_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_LOCK, + gd_mgmt_v3_lock_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_lock_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_initiate_lockdown (glusterd_conf_t *conf, glusterd_op_t op, + dict_t *dict, char **op_errstr, int npeers, + gf_boolean_t *is_acquired) +{ + char *volname = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + int32_t ret = -1; + int32_t peer_cnt = 0; + struct syncargs args = {0}; + struct list_head *peers = NULL; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (is_acquired); + + peers = &conf->xaction_peers; + + /* Trying to acquire multiple mgmt_v3 locks on local node */ + ret = glusterd_multiple_mgmt_v3_lock (dict, MY_UUID); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire mgmt_v3 locks on localhost"); + goto out; + } + + *is_acquired = _gf_true; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending mgmt_v3 lock req to other nodes in the cluster */ + gd_syncargs_init (&args, NULL); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_lock (op, dict, peerinfo, &args, + MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent lock op req for %s " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + if (ret) { + if (*op_errstr) + gf_log (this->name, GF_LOG_ERROR, "%s", + *op_errstr); + + if (volname) + ret = gf_asprintf (op_errstr, + "Another transaction is in progress " + "for %s. Please try again after " + "sometime.", volname); + else + ret = gf_asprintf (op_errstr, + "Another transaction is in progress " + "Please try again after sometime."); + + if (ret == -1) + *op_errstr = NULL; + + ret = -1; + } + + return ret; +} + +int +glusterd_pre_validate_aggr_rsp_dict (glusterd_op_t op, dict_t *aggr, dict_t *rsp) +{ + int32_t ret = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (aggr); + GF_ASSERT (rsp); + + switch (op) { + case GD_OP_SNAP: + ret = glusterd_snap_pre_validate_use_rsp_dict (aggr, rsp); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to aggregate prevalidate " + "response dictionaries."); + goto out; + } + break; + default: + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Invalid op (%s)", gd_op_list[op]); + + break; + } +out: + return ret; +} + +int32_t +gd_mgmt_v3_pre_validate_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_pre_val_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + dict_t *rsp_dict = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_pre_val_rsp); + if (ret < 0) + goto out; + + if (rsp.dict.dict_len) { + /* Unserialize the dictionary */ + rsp_dict = dict_new (); + + ret = dict_unserialize (rsp.dict.dict_val, + rsp.dict.dict_len, + &rsp_dict); + if (ret < 0) { + GF_FREE (rsp.dict.dict_val); + goto out; + } else { + rsp_dict->extra_stdfree = rsp.dict.dict_val; + } + } + + uuid_copy (args->uuid, rsp.uuid); + pthread_mutex_lock (&args->lock_dict); + { + ret = glusterd_pre_validate_aggr_rsp_dict (rsp.op, args->dict, + rsp_dict); + } + pthread_mutex_unlock (&args->lock_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from " + " node/brick"); + if (!rsp.op_ret) + op_ret = ret; + else { + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + } + } else { + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + } + +out: + if (rsp_dict) + dict_unref (rsp_dict); + + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_PRE_VALIDATE, + peerinfo, rsp.uuid); + + if (rsp.op_errstr) + free (rsp.op_errstr); + + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_pre_validate_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_pre_validate_cbk_fn); +} + +int +gd_mgmt_v3_pre_validate_req (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + int32_t ret = -1; + gd1_mgmt_v3_pre_val_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_PRE_VALIDATE, + gd_mgmt_v3_pre_validate_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_pre_val_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_pre_validate (glusterd_conf_t *conf, glusterd_op_t op, + dict_t *req_dict, char **op_errstr, int npeers) +{ + int32_t ret = -1; + int32_t peer_cnt = 0; + dict_t *rsp_dict = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (req_dict); + GF_ASSERT (op_errstr); + + peers = &conf->xaction_peers; + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create response dictionary"); + goto out; + } + + /* Pre Validation on local node */ + ret = gd_mgmt_v3_pre_validate_fn (op, req_dict, op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Pre Validation failed for " + "operation %s on local node", + gd_op_list[op]); + + if (*op_errstr == NULL) { + ret = gf_asprintf (op_errstr, + "Pre-validation failed " + "on localhost. Please " + "check log file for details"); + if (ret == -1) + *op_errstr = NULL; + + ret = -1; + } + goto out; + } + + ret = glusterd_pre_validate_aggr_rsp_dict (op, req_dict, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from " + " node/brick"); + goto out; + } + + dict_unref (rsp_dict); + rsp_dict = NULL; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending Pre Validation req to other nodes in the cluster */ + gd_syncargs_init (&args, req_dict); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_pre_validate_req (op, req_dict, peerinfo, &args, + MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "Pre Validation failed on peers"); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent pre valaidation req for %s " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + return ret; +} + +int +glusterd_mgmt_v3_build_payload (dict_t **req, char **op_errstr, dict_t *dict, + glusterd_op_t op) +{ + int32_t ret = -1; + dict_t *req_dict = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (op_errstr); + GF_ASSERT (dict); + + req_dict = dict_new (); + if (!req_dict) + goto out; + + switch (op) { + case GD_OP_SNAP: + dict_copy (dict, req_dict); + break; + default: + break; + } + + *req = req_dict; + ret = 0; +out: + return ret; +} + +int32_t +gd_mgmt_v3_brick_op_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_brick_op_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + /* If the operation failed, then iov can be NULL. So better check the + status of the operation and then worry about iov (if the status of + the command is success) + */ + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_brick_op_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_BRICK_OP, + peerinfo, rsp.uuid); + + if (rsp.op_errstr) + free (rsp.op_errstr); + + free (rsp.dict.dict_val); + + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_brick_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_brick_op_cbk_fn); +} + +int +gd_mgmt_v3_brick_op_req (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + int32_t ret = -1; + gd1_mgmt_v3_brick_op_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_BRICK_OP, + gd_mgmt_v3_brick_op_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_brick_op_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_brick_op (glusterd_conf_t *conf, glusterd_op_t op, + dict_t *req_dict, char **op_errstr, int npeers) +{ + int32_t ret = -1; + int32_t peer_cnt = 0; + dict_t *rsp_dict = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (req_dict); + GF_ASSERT (op_errstr); + + peers = &conf->xaction_peers; + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create response dictionary"); + goto out; + } + + /* Perform brick op on local node */ + ret = gd_mgmt_v3_brick_op_fn (op, req_dict, op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Brick ops failed for " + "operation %s on local node", + gd_op_list[op]); + + if (*op_errstr == NULL) { + ret = gf_asprintf (op_errstr, + "Brick ops failed " + "on localhost. Please " + "check log file for details"); + if (ret == -1) + *op_errstr = NULL; + + ret = -1; + } + goto out; + } + + dict_unref (rsp_dict); + rsp_dict = NULL; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending brick op req to other nodes in the cluster */ + gd_syncargs_init (&args, NULL); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_brick_op_req (op, req_dict, peerinfo, &args, + MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "Brick ops failed on peers"); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent brick op req for %s " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + return ret; +} + +int32_t +gd_mgmt_v3_commit_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_commit_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + dict_t *rsp_dict = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_commit_rsp); + if (ret < 0) + goto out; + + if (rsp.dict.dict_len) { + /* Unserialize the dictionary */ + rsp_dict = dict_new (); + + ret = dict_unserialize (rsp.dict.dict_val, + rsp.dict.dict_len, + &rsp_dict); + if (ret < 0) { + GF_FREE (rsp.dict.dict_val); + goto out; + } else { + rsp_dict->extra_stdfree = rsp.dict.dict_val; + } + } + + uuid_copy (args->uuid, rsp.uuid); + pthread_mutex_lock (&args->lock_dict); + { + ret = glusterd_syncop_aggr_rsp_dict (rsp.op, args->dict, + rsp_dict); + } + pthread_mutex_unlock (&args->lock_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from " + " node/brick"); + if (!rsp.op_ret) + op_ret = ret; + else { + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + } + } else { + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + } + +out: + if (rsp_dict) + dict_unref (rsp_dict); + + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_COMMIT, + peerinfo, rsp.uuid); + + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_commit_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_commit_cbk_fn); +} + +int +gd_mgmt_v3_commit_req (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + int32_t ret = -1; + gd1_mgmt_v3_commit_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_COMMIT, + gd_mgmt_v3_commit_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_commit_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_commit (glusterd_conf_t *conf, glusterd_op_t op, + dict_t *op_ctx, dict_t *req_dict, + char **op_errstr, int npeers) +{ + int32_t ret = -1; + int32_t peer_cnt = 0; + dict_t *rsp_dict = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (op_ctx); + GF_ASSERT (req_dict); + GF_ASSERT (op_errstr); + + peers = &conf->xaction_peers; + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create response dictionary"); + goto out; + } + + /* Commit on local node */ + ret = gd_mgmt_v3_commit_fn (op, req_dict, op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Commit failed for " + "operation %s on local node", + gd_op_list[op]); + + if (*op_errstr == NULL) { + ret = gf_asprintf (op_errstr, + "Commit failed " + "on localhost. Please " + "check log file for details."); + if (ret == -1) + *op_errstr = NULL; + + ret = -1; + } + goto out; + } + + ret = glusterd_syncop_aggr_rsp_dict (op, op_ctx, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from " + " node/brick"); + goto out; + } + + dict_unref (rsp_dict); + rsp_dict = NULL; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending commit req to other nodes in the cluster */ + gd_syncargs_init (&args, op_ctx); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_commit_req (op, req_dict, peerinfo, &args, + MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "Commit failed on peers"); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent commit req for %s to %d " + "peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + return ret; +} + +int32_t +gd_mgmt_v3_post_validate_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_post_val_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_post_val_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_POST_VALIDATE, + peerinfo, rsp.uuid); + if (rsp.op_errstr) + free (rsp.op_errstr); + + free (rsp.dict.dict_val); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_post_validate_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_post_validate_cbk_fn); +} + +int +gd_mgmt_v3_post_validate_req (glusterd_op_t op, int32_t op_ret, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + int32_t ret = -1; + gd1_mgmt_v3_post_val_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + req.op_ret = op_ret; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_POST_VALIDATE, + gd_mgmt_v3_post_validate_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_post_val_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_post_validate (glusterd_conf_t *conf, glusterd_op_t op, + int32_t op_ret, dict_t *dict, dict_t *req_dict, + char **op_errstr, int npeers) +{ + int32_t ret = -1; + int32_t peer_cnt = 0; + dict_t *rsp_dict = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (dict); + GF_ASSERT (req_dict); + GF_ASSERT (op_errstr); + + peers = &conf->xaction_peers; + GF_ASSERT (peers); + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create response dictionary"); + goto out; + } + + /* Copy the contents of dict like missed snaps info to req_dict */ + dict_copy (dict, req_dict); + + /* Post Validation on local node */ + ret = gd_mgmt_v3_post_validate_fn (op, op_ret, req_dict, op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Post Validation failed for " + "operation %s on local node", + gd_op_list[op]); + + if (*op_errstr == NULL) { + ret = gf_asprintf (op_errstr, + "Post-validation failed " + "on localhost. Please check " + "log file for details"); + if (ret == -1) + *op_errstr = NULL; + + ret = -1; + } + goto out; + } + + dict_unref (rsp_dict); + rsp_dict = NULL; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending Post Validation req to other nodes in the cluster */ + gd_syncargs_init (&args, req_dict); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_post_validate_req (op, op_ret, req_dict, peerinfo, + &args, MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "Post Validation failed on peers"); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent post valaidation req for %s " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + return ret; +} + +int32_t +gd_mgmt_v3_unlock_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_unlock_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_UNLOCK, + peerinfo, rsp.uuid); + free (rsp.dict.dict_val); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_unlock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_unlock_cbk_fn); +} + +int +gd_mgmt_v3_unlock (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + int32_t ret = -1; + gd1_mgmt_v3_unlock_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_UNLOCK, + gd_mgmt_v3_unlock_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_unlock_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_release_peer_locks (glusterd_conf_t *conf, glusterd_op_t op, + dict_t *dict, int32_t op_ret, + char **op_errstr, int npeers, + gf_boolean_t is_acquired) +{ + int32_t ret = -1; + int32_t peer_cnt = 0; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + peers = &conf->xaction_peers; + + /* If the lock has not been held during this + * transaction, do not send unlock requests */ + if (!is_acquired) + goto out; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending mgmt_v3 unlock req to other nodes in the cluster */ + gd_syncargs_init (&args, NULL); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_unlock (op, dict, peerinfo, &args, + MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unlock failed on peers"); + + if (!op_ret && args.errstr) + *op_errstr = gf_strdup (args.errstr); + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent unlock op req for %s " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); + +out: + return ret; +} + +int32_t +glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict) +{ + int32_t ret = -1; + int32_t op_ret = -1; + int32_t npeers = 0; + dict_t *req_dict = NULL; + dict_t *tmp_dict = NULL; + glusterd_conf_t *conf = NULL; + char *op_errstr = NULL; + xlator_t *this = NULL; + gf_boolean_t is_acquired = _gf_false; + uuid_t *originator_uuid = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (dict); + conf = this->private; + GF_ASSERT (conf); + + /* Save the MY_UUID as the originator_uuid. This originator_uuid + * will be used by is_origin_glusterd() to determine if a node + * is the originator node for a command. */ + originator_uuid = GF_CALLOC (1, sizeof(uuid_t), + gf_common_mt_uuid_t); + if (!originator_uuid) { + ret = -1; + goto out; + } + + uuid_copy (*originator_uuid, MY_UUID); + ret = dict_set_bin (dict, "originator_uuid", + originator_uuid, sizeof (uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set originator_uuid."); + goto out; + } + + /* Marking the operation as complete synctasked */ + ret = dict_set_int32 (dict, "is_synctasked", _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set synctasked flag."); + goto out; + } + + /* Use a copy at local unlock as cli response will be sent before + * the unlock and the volname in the dict might be removed */ + tmp_dict = dict_new(); + if (!tmp_dict) { + gf_log (this->name, GF_LOG_ERROR, "Unable to create dict"); + goto out; + } + dict_copy (dict, tmp_dict); + + /* BUILD PEERS LIST */ + INIT_LIST_HEAD (&conf->xaction_peers); + npeers = gd_build_peers_list (&conf->peers, &conf->xaction_peers, op); + + /* LOCKDOWN PHASE - Acquire mgmt_v3 locks */ + ret = glusterd_mgmt_v3_initiate_lockdown (conf, op, dict, &op_errstr, + npeers, &is_acquired); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "mgmt_v3 lockdown failed."); + goto out; + } + + /* BUILD PAYLOAD */ + ret = glusterd_mgmt_v3_build_payload (&req_dict, &op_errstr, dict, op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD, + gd_op_list[op]); + if (op_errstr == NULL) + gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD); + goto out; + } + + /* PRE-COMMIT VALIDATE PHASE */ + ret = glusterd_mgmt_v3_pre_validate (conf, op, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Pre Validation Failed"); + goto out; + } + + /* COMMIT OP PHASE */ + ret = glusterd_mgmt_v3_commit (conf, op, dict, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Commit Op Failed"); + goto out; + } + + /* POST-COMMIT VALIDATE PHASE */ + /* As of now, post_validate is not handling any other + commands other than snapshot. So as of now, I am + sending 0 (op_ret as 0). + */ + ret = glusterd_mgmt_v3_post_validate (conf, op, 0, dict, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Post Validation Failed"); + goto out; + } + + ret = 0; +out: + op_ret = ret; + /* UNLOCK PHASE FOR PEERS*/ + (void) glusterd_mgmt_v3_release_peer_locks (conf, op, dict, + op_ret, &op_errstr, + npeers, is_acquired); + + /* LOCAL VOLUME(S) UNLOCK */ + if (is_acquired) { + /* Trying to release multiple mgmt_v3 locks */ + ret = glusterd_multiple_mgmt_v3_unlock (tmp_dict, MY_UUID); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release mgmt_v3 locks on localhost"); + op_ret = ret; + } + } + + /* SEND CLI RESPONSE */ + glusterd_op_send_cli_response (op, op_ret, 0, req, dict, op_errstr); + + if (req_dict) + dict_unref (req_dict); + + if (tmp_dict) + dict_unref (tmp_dict); + + if (op_errstr) { + GF_FREE (op_errstr); + op_errstr = NULL; + } + + return 0; +} + +int32_t +glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict) +{ + int32_t ret = -1; + int32_t op_ret = -1; + int32_t npeers = 0; + dict_t *req_dict = NULL; + dict_t *tmp_dict = NULL; + glusterd_conf_t *conf = NULL; + char *op_errstr = NULL; + xlator_t *this = NULL; + gf_boolean_t is_acquired = _gf_false; + uuid_t *originator_uuid = NULL; + gf_boolean_t success = _gf_false; + char *tmp_errstr = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (dict); + conf = this->private; + GF_ASSERT (conf); + + /* Save the MY_UUID as the originator_uuid. This originator_uuid + * will be used by is_origin_glusterd() to determine if a node + * is the originator node for a command. */ + originator_uuid = GF_CALLOC (1, sizeof(uuid_t), + gf_common_mt_uuid_t); + if (!originator_uuid) { + ret = -1; + goto out; + } + + uuid_copy (*originator_uuid, MY_UUID); + ret = dict_set_bin (dict, "originator_uuid", + originator_uuid, sizeof (uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set originator_uuid."); + goto out; + } + + /* Marking the operation as complete synctasked */ + ret = dict_set_int32 (dict, "is_synctasked", _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set synctasked flag."); + goto out; + } + + /* Use a copy at local unlock as cli response will be sent before + * the unlock and the volname in the dict might be removed */ + tmp_dict = dict_new(); + if (!tmp_dict) { + gf_log (this->name, GF_LOG_ERROR, "Unable to create dict"); + goto out; + } + dict_copy (dict, tmp_dict); + + /* BUILD PEERS LIST */ + INIT_LIST_HEAD (&conf->xaction_peers); + npeers = gd_build_peers_list (&conf->peers, &conf->xaction_peers, op); + + /* LOCKDOWN PHASE - Acquire mgmt_v3 locks */ + ret = glusterd_mgmt_v3_initiate_lockdown (conf, op, dict, &op_errstr, + npeers, &is_acquired); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "mgmt_v3 lockdown failed."); + goto out; + } + + /* BUILD PAYLOAD */ + ret = glusterd_mgmt_v3_build_payload (&req_dict, &op_errstr, dict, op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD, + gd_op_list[op]); + if (op_errstr == NULL) + gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD); + goto out; + } + + /* PRE-COMMIT VALIDATE PHASE */ + ret = glusterd_mgmt_v3_pre_validate (conf, op, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Pre Validation Failed"); + goto out; + } + + /* BRICK OP PHASE for initiating barrier*/ + ret = dict_set_int32 (req_dict, "barrier", 1); + if (ret) + goto out; + ret = glusterd_mgmt_v3_brick_op (conf, op, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Brick Ops Failed"); + goto unbarrier; + } + + /* COMMIT OP PHASE */ + /* TODO: As of now, the plan is to do quorum check before sending the + commit fop and if the quorum succeeds, then commit is sent to all + the other glusterds. + snap create functionality now creates the in memory and on disk + objects for the snapshot (marking them as incomplete), takes the lvm + snapshot and then updates the status of the in memory and on disk + snap objects as complete. Suppose one of the glusterds goes down + after taking the lvm snapshot, but before updating the snap object, + then treat it as a snapshot create failure and trigger cleanup. + i.e the number of commit responses received by the originator + glusterd shold be the same as the number of peers it has sent the + request to (i.e npeers variable). If not, then originator glusterd + will initiate cleanup in post-validate fop. + Question: What if one of the other glusterds goes down as explained + above and along with it the originator glusterd also goes down? + Who will initiate the cleanup? + */ + ret = glusterd_mgmt_v3_commit (conf, op, dict, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Commit Op Failed"); + /* If the main op fails, we should save the error string. + Because, op_errstr will be used for unbarrier and + unlock ops also. We might lose the actual error that + caused the failure. + */ + tmp_errstr = op_errstr; + op_errstr = NULL; + goto unbarrier; + } + + success = _gf_true; +unbarrier: + /* BRICK OP PHASE for removing the barrier*/ + ret = dict_set_int32 (req_dict, "barrier", 0); + if (ret) + goto out; + ret = glusterd_mgmt_v3_brick_op (conf, op, req_dict, + &op_errstr, npeers); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Brick Ops Failed"); + goto out; + } + + ret = 0; + +out: + op_ret = ret; + + if (success == _gf_false) + op_ret = -1; + + /* POST-COMMIT VALIDATE PHASE */ + ret = glusterd_mgmt_v3_post_validate (conf, op, op_ret, dict, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Post Validation Failed"); + op_ret = -1; + } + + /* UNLOCK PHASE FOR PEERS*/ + (void) glusterd_mgmt_v3_release_peer_locks (conf, op, dict, + op_ret, &op_errstr, + npeers, is_acquired); + + /* If the commit op (snapshot taking) failed, then the error is stored + in tmp_errstr and unbarrier is called. Suppose, if unbarrier also + fails, then the error happened in unbarrier is logged and freed. + The error happened in commit op, which is stored in tmp_errstr + is sent to cli. + */ + if (tmp_errstr) { + if (op_errstr) { + gf_log (this->name, GF_LOG_ERROR, "unbarrier brick op" + "failed with the error %s", op_errstr); + GF_FREE (op_errstr); + op_errstr = NULL; + } + op_errstr = tmp_errstr; + } + + /* LOCAL VOLUME(S) UNLOCK */ + if (is_acquired) { + /* Trying to release multiple mgmt_v3 locks */ + ret = glusterd_multiple_mgmt_v3_unlock (tmp_dict, MY_UUID); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release mgmt_v3 locks on localhost"); + op_ret = ret; + } + } + + /* SEND CLI RESPONSE */ + glusterd_op_send_cli_response (op, op_ret, 0, req, dict, op_errstr); + + if (req_dict) + dict_unref (req_dict); + + if (tmp_dict) + dict_unref (tmp_dict); + + if (op_errstr) { + GF_FREE (op_errstr); + op_errstr = NULL; + } + + return 0; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt.h b/xlators/mgmt/glusterd/src/glusterd-mgmt.h new file mode 100644 index 000000000..b185a9bec --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-mgmt.h @@ -0,0 +1,45 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +#ifndef _GLUSTERD_MGMT_H_ +#define _GLUSTERD_MGMT_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +int32_t +gd_mgmt_v3_pre_validate_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict); + +int32_t +gd_mgmt_v3_brick_op_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict); + +int32_t +gd_mgmt_v3_commit_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict); + +int32_t +gd_mgmt_v3_post_validate_fn (glusterd_op_t op, int32_t op_ret, dict_t *dict, + char **op_errstr, dict_t *rsp_dict); + +int32_t +glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict); + +int32_t +glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict); + +int +glusterd_snap_pre_validate_use_rsp_dict (dict_t *dst, dict_t *src); + +#endif /* _GLUSTERD_MGMT_H_ */ diff --git a/xlators/mgmt/glusterd/src/glusterd-mountbroker.c b/xlators/mgmt/glusterd/src/glusterd-mountbroker.c index 32adf4ddd..0d67d1303 100644 --- a/xlators/mgmt/glusterd/src/glusterd-mountbroker.c +++ b/xlators/mgmt/glusterd/src/glusterd-mountbroker.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" @@ -244,12 +234,15 @@ const char *georep_mnt_desc_template = "xlator-option=\\*-dht.assert-no-child-down=true " "volfile-server=localhost " "client-pid=%d " - "volfile-id=%s " "user-map-root=%s " ")" "SUB+(" "log-file="DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"*/* " "log-level=* " + "volfile-id=* " + ")" + "MEET(" + "%s" ")"; const char *hadoop_mnt_desc_template = @@ -265,18 +258,69 @@ const char *hadoop_mnt_desc_template = ")"; int -make_georep_mountspec (gf_mount_spec_t *mspec, const char *volname, +make_georep_mountspec (gf_mount_spec_t *mspec, const char *volnames, char *user) { char *georep_mnt_desc = NULL; + char *meetspec = NULL; + char *vols = NULL; + char *vol = NULL; + char *p = NULL; + char *savetok = NULL; + char *fa[3] = {0,}; + size_t siz = 0; + int vc = 0; + int i = 0; int ret = 0; + vols = gf_strdup ((char *)volnames); + if (!vols) + goto out; + + for (vc = 1, p = vols; *p; p++) { + if (*p == ',') + vc++; + } + siz = strlen (volnames) + vc * strlen("volfile-id="); + meetspec = GF_CALLOC (1, siz + 1, gf_gld_mt_georep_meet_spec); + if (!meetspec) + goto out; + + for (p = vols;;) { + vol = strtok_r (p, ",", &savetok); + if (!vol) { + GF_ASSERT (vc == 0); + break; + } + p = NULL; + strcat (meetspec, "volfile-id="); + strcat (meetspec, vol); + if (--vc > 0) + strcat (meetspec, " "); + } + ret = gf_asprintf (&georep_mnt_desc, georep_mnt_desc_template, - GF_CLIENT_PID_GSYNCD, volname, user); - if (ret == -1) - return ret; + GF_CLIENT_PID_GSYNCD, user, meetspec); + if (ret == -1) { + georep_mnt_desc = NULL; + goto out; + } + + ret = parse_mount_pattern_desc (mspec, georep_mnt_desc); + + out: + fa[0] = meetspec; + fa[1] = vols; + fa[2] = georep_mnt_desc; - return parse_mount_pattern_desc (mspec, georep_mnt_desc); + for (i = 0; i < 3; i++) { + if (fa[i] == NULL) + ret = -1; + else + GF_FREE (fa[i]); + } + + return ret; } int @@ -425,6 +469,7 @@ evaluate_mount_request (gf_mount_spec_t *mspec, dict_t *argdict) break; case SET_INTERSECT: match = sd.common; + break; default: GF_ASSERT(!"unreached"); } @@ -642,8 +687,7 @@ glusterd_do_mount (char *label, dict_t *argdict, char **path, int *op_errno) *path = cookie; } - if (mtptemp) - GF_FREE (mtptemp); + GF_FREE (mtptemp); return ret; } diff --git a/xlators/mgmt/glusterd/src/glusterd-mountbroker.h b/xlators/mgmt/glusterd/src/glusterd-mountbroker.h index 729cf3fbc..426252ebe 100644 --- a/xlators/mgmt/glusterd/src/glusterd-mountbroker.h +++ b/xlators/mgmt/glusterd/src/glusterd-mountbroker.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #define MB_HIVE "mb_hive" typedef enum { diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c index f24cb9b13..1666f5e4d 100644 --- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c +++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H @@ -45,7 +35,9 @@ #include "glusterd-op-sm.h" #include "glusterd-utils.h" #include "glusterd-store.h" +#include "glusterd-hooks.h" #include "glusterd-volgen.h" +#include "glusterd-locks.h" #include "syscall.h" #include "cli1-xdr.h" #include "common-utils.h" @@ -55,9 +47,186 @@ #include <signal.h> #include <sys/wait.h> +#define ALL_VOLUME_OPTION_CHECK(volname, key, ret, op_errstr, label) \ + do { \ + gf_boolean_t _all = !strcmp ("all", volname); \ + gf_boolean_t _ratio = !strcmp (key, \ + GLUSTERD_QUORUM_RATIO_KEY); \ + if (_all && !_ratio) { \ + ret = -1; \ + *op_errstr = gf_strdup ("Not a valid option for all " \ + "volumes"); \ + goto label; \ + } else if (!_all && _ratio) { \ + ret = -1; \ + *op_errstr = gf_strdup ("Not a valid option for " \ + "single volume"); \ + goto label; \ + } \ + } while (0) + static struct list_head gd_op_sm_queue; pthread_mutex_t gd_op_sm_lock; glusterd_op_info_t opinfo = {{0},}; +uuid_t global_txn_id = {"\0"}; /* To be used in + * heterogeneous + * cluster with no + * transaction ids */ + +static dict_t *txn_opinfo; + +struct txn_opinfo_object_ { + glusterd_op_info_t opinfo; +}; +typedef struct txn_opinfo_object_ txn_opinfo_obj; + +int32_t +glusterd_txn_opinfo_dict_init () +{ + int32_t ret = -1; + + txn_opinfo = dict_new (); + if (!txn_opinfo) { + ret = -1; + goto out; + } + + ret = 0; +out: + return ret; +} + +void +glusterd_txn_opinfo_dict_fini () +{ + if (txn_opinfo) + dict_destroy (txn_opinfo); +} + +void +glusterd_txn_opinfo_init (glusterd_op_info_t *opinfo, + glusterd_op_sm_state_info_t *state, + glusterd_op_t *op, + dict_t *op_ctx, + rpcsvc_request_t *req) +{ + GF_ASSERT (opinfo); + + if (state) + opinfo->state = *state; + + if (op) + opinfo->op = *op; + + opinfo->op_ctx = dict_ref(op_ctx); + + if (req) + opinfo->req = req; + + return; +} + +int32_t +glusterd_get_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t *opinfo) +{ + int32_t ret = -1; + txn_opinfo_obj *opinfo_obj = NULL; + + if (!txn_id || !opinfo) { + gf_log ("", GF_LOG_ERROR, + "Empty transaction id or opinfo received."); + ret = -1; + goto out; + } + + ret = dict_get_bin(txn_opinfo, uuid_utoa (*txn_id), + (void **) &opinfo_obj); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get transaction opinfo"); + goto out; + } + + (*opinfo) = opinfo_obj->opinfo; + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_set_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t *opinfo) +{ + int32_t ret = -1; + txn_opinfo_obj *opinfo_obj = NULL; + + if (!txn_id) { + gf_log ("", GF_LOG_ERROR, "Empty transaction id received."); + ret = -1; + goto out; + } + + ret = dict_get_bin(txn_opinfo, uuid_utoa (*txn_id), + (void **) &opinfo_obj); + if (ret) { + opinfo_obj = GF_CALLOC (1, sizeof(txn_opinfo_obj), + gf_common_mt_txn_opinfo_obj_t); + if (!opinfo_obj) { + ret = -1; + goto out; + } + + ret = dict_set_bin(txn_opinfo, uuid_utoa (*txn_id), opinfo_obj, + sizeof(txn_opinfo_obj)); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set opinfo for transaction ID : %s", + uuid_utoa (*txn_id)); + goto out; + } + } + + opinfo_obj->opinfo = (*opinfo); + + ret = 0; +out: + if (ret) + if (opinfo_obj) + GF_FREE (opinfo_obj); + + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_clear_txn_opinfo (uuid_t *txn_id) +{ + int32_t ret = -1; + glusterd_op_info_t txn_op_info = {{0},}; + + if (!txn_id) { + gf_log ("", GF_LOG_ERROR, "Empty transaction id received."); + ret = -1; + goto out; + } + + ret = glusterd_get_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Transaction opinfo not found"); + goto out; + } + + dict_unref (txn_op_info.op_ctx); + + dict_del(txn_opinfo, uuid_utoa (*txn_id)); + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + static int glusterfs_port = GLUSTERD_DEFAULT_PORT; static char *glusterd_op_sm_state_names[] = { "Default", @@ -96,6 +265,8 @@ static char *glusterd_op_sm_event_names[] = { "GD_OP_EVENT_INVALID" }; +extern struct volopt_map_entry glusterd_volopt_map[]; + char* glusterd_op_sm_state_name_get (int state) { @@ -136,10 +307,10 @@ glusterd_is_volume_started (glusterd_volinfo_t *volinfo) } static int -glusterd_op_sm_inject_all_acc () +glusterd_op_sm_inject_all_acc (uuid_t *txn_id) { int32_t ret = -1; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, txn_id, NULL); gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -150,6 +321,13 @@ glusterd_brick_op_build_payload (glusterd_op_t op, glusterd_brickinfo_t *brickin { int ret = -1; gd1_mgmt_brick_op_req *brick_req = NULL; + char *volname = NULL; + char name[1024] = {0,}; + gf_xl_afr_op_t heal_op = GF_AFR_OP_INVALID; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (op < GD_OP_MAX); GF_ASSERT (op > GD_OP_NONE); @@ -161,10 +339,8 @@ glusterd_brick_op_build_payload (glusterd_op_t op, glusterd_brickinfo_t *brickin case GD_OP_STOP_VOLUME: brick_req = GF_CALLOC (1, sizeof (*brick_req), gf_gld_mt_mop_brick_req_t); - if (!brick_req) { - gf_log ("", GF_LOG_ERROR, "Out of Memory"); + if (!brick_req) goto out; - } brick_req->op = GLUSTERD_BRICK_TERMINATE; brick_req->name = ""; break; @@ -172,10 +348,8 @@ glusterd_brick_op_build_payload (glusterd_op_t op, glusterd_brickinfo_t *brickin brick_req = GF_CALLOC (1, sizeof (*brick_req), gf_gld_mt_mop_brick_req_t); - if (!brick_req) { - gf_log ("", GF_LOG_ERROR, "Out of Memory"); + if (!brick_req) goto out; - } brick_req->op = GLUSTERD_BRICK_XLATOR_INFO; brick_req->name = brickinfo->path; @@ -188,61 +362,209 @@ glusterd_brick_op_build_payload (glusterd_op_t op, glusterd_brickinfo_t *brickin if (!brick_req) goto out; - brick_req->op = GLUSTERD_BRICK_XLATOR_HEAL; + brick_req->op = GLUSTERD_BRICK_XLATOR_OP; brick_req->name = ""; + ret = dict_get_int32 (dict, "heal-op", (int32_t*)&heal_op); + if (ret) + goto out; + ret = dict_set_int32 (dict, "xl-op", heal_op); } break; case GD_OP_STATUS_VOLUME: { brick_req = GF_CALLOC (1, sizeof (*brick_req), gf_gld_mt_mop_brick_req_t); - if (!brick_req) { - gf_log (THIS->name, GF_LOG_ERROR, "Out of memory"); + if (!brick_req) goto out; - } brick_req->op = GLUSTERD_BRICK_STATUS; brick_req->name = ""; } break; + case GD_OP_REBALANCE: + case GD_OP_DEFRAG_BRICK_VOLUME: + brick_req = GF_CALLOC (1, sizeof (*brick_req), + gf_gld_mt_mop_brick_req_t); + if (!brick_req) + goto out; + + brick_req->op = GLUSTERD_BRICK_XLATOR_DEFRAG; + ret = dict_get_str (dict, "volname", &volname); + if (ret) + goto out; + snprintf (name, 1024, "%s-dht",volname); + brick_req->name = gf_strdup (name); + + break; + case GD_OP_SNAP: + brick_req = GF_CALLOC (1, sizeof (*brick_req), + gf_gld_mt_mop_brick_req_t); + if (!brick_req) + goto out; + + brick_req->op = GLUSTERD_VOLUME_BARRIER_OP; + ret = dict_get_str (dict, "volname", &volname); + if (ret) + goto out; + snprintf (name, 1024, "%s-server",volname); + brick_req->name = gf_strdup (name); + + break; default: goto out; break; } ret = dict_allocate_and_serialize (dict, &brick_req->input.input_val, - (size_t*)&brick_req->input.input_len); + &brick_req->input.input_len); + if (ret) + goto out; + *req = brick_req; + ret = 0; + +out: + if (ret && brick_req) + GF_FREE (brick_req); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterd_node_op_build_payload (glusterd_op_t op, gd1_mgmt_brick_op_req **req, + dict_t *dict) +{ + int ret = -1; + gd1_mgmt_brick_op_req *brick_req = NULL; + + GF_ASSERT (op < GD_OP_MAX); + GF_ASSERT (op > GD_OP_NONE); + GF_ASSERT (req); + + switch (op) { + case GD_OP_PROFILE_VOLUME: + brick_req = GF_CALLOC (1, sizeof (*brick_req), + gf_gld_mt_mop_brick_req_t); + if (!brick_req) + goto out; + + brick_req->op = GLUSTERD_NODE_PROFILE; + brick_req->name = ""; + + break; + + case GD_OP_STATUS_VOLUME: + brick_req = GF_CALLOC (1, sizeof (*brick_req), + gf_gld_mt_mop_brick_req_t); + if (!brick_req) + goto out; + + brick_req->op = GLUSTERD_NODE_STATUS; + brick_req->name = ""; + + break; + + default: + goto out; + } + + ret = dict_allocate_and_serialize (dict, &brick_req->input.input_val, + &brick_req->input.input_len); + if (ret) goto out; + *req = brick_req; ret = 0; out: if (ret && brick_req) GF_FREE (brick_req); - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +static int +glusterd_validate_quorum_options (xlator_t *this, char *fullkey, char *value, + char **op_errstr) +{ + int ret = 0; + char *key = NULL; + volume_option_t *opt = NULL; + + if (!glusterd_is_quorum_option (fullkey)) + goto out; + key = strchr (fullkey, '.'); + key++; + opt = xlator_volume_option_get (this, key); + ret = xlator_option_validate (this, key, value, opt, op_errstr); +out: return ret; } +static int +glusterd_check_client_op_version_support (char *volname, uint32_t op_version, + char **op_errstr) +{ + int ret = 0; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + rpc_transport_t *xprt = NULL; + + this = THIS; + GF_ASSERT(this); + priv = this->private; + GF_ASSERT(priv); + + pthread_mutex_lock (&priv->xprt_lock); + list_for_each_entry (xprt, &priv->xprt_list, list) { + if ((!strcmp(volname, xprt->peerinfo.volname)) && + ((op_version > xprt->peerinfo.max_op_version) || + (op_version < xprt->peerinfo.min_op_version))) { + ret = -1; + break; + } + } + pthread_mutex_unlock (&priv->xprt_lock); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "One or more clients " + "don't support the required op-version"); + ret = gf_asprintf (op_errstr, "One or more connected clients " + "cannot support the feature being set. " + "These clients need to be upgraded or " + "disconnected before running this command" + " again"); + return -1; + } + return 0; +} static int glusterd_op_stage_set_volume (dict_t *dict, char **op_errstr) { - int ret = 0; - char *volname = NULL; - int exists = 0; - char *key = NULL; - char *key_fixed = NULL; - char *value = NULL; - char str[100] = {0, }; - int count = 0; - int dict_count = 0; - char errstr[2048] = {0, }; - glusterd_volinfo_t *volinfo = NULL; - dict_t *val_dict = NULL; - gf_boolean_t global_opt = _gf_false; - glusterd_volinfo_t *voliter = NULL; - glusterd_conf_t *priv = NULL; - xlator_t *this = NULL; + int ret = -1; + char *volname = NULL; + int exists = 0; + char *key = NULL; + char *key_fixed = NULL; + char *value = NULL; + char str[100] = {0, }; + int count = 0; + int dict_count = 0; + char errstr[2048] = {0, }; + glusterd_volinfo_t *volinfo = NULL; + dict_t *val_dict = NULL; + gf_boolean_t global_opt = _gf_false; + glusterd_volinfo_t *voliter = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + uint32_t new_op_version = 0; + uint32_t local_new_op_version = 0; + uint32_t key_op_version = 0; + uint32_t local_key_op_version = 0; + gf_boolean_t origin_glusterd = _gf_true; + gf_boolean_t check_op_version = _gf_true; + gf_boolean_t all_vol = _gf_false; + struct volopt_map_entry *vme = NULL; GF_ASSERT (dict); this = THIS; @@ -254,35 +576,73 @@ glusterd_op_stage_set_volume (dict_t *dict, char **op_errstr) if (!val_dict) goto out; - ret = dict_get_int32 (dict, "count", &dict_count); + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); + goto out; + } + + /* Check if we can support the required op-version + * This check is not done on the originator glusterd. The originator + * glusterd sets this value. + */ + origin_glusterd = is_origin_glusterd (dict); + + if (!origin_glusterd) { + /* Check for v3.3.x origin glusterd */ + check_op_version = dict_get_str_boolean (dict, + "check-op-version", + _gf_false); + + if (check_op_version) { + ret = dict_get_uint32 (dict, "new-op-version", + &new_op_version); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get new_op_version"); + goto out; + } + if ((new_op_version > GD_OP_VERSION_MAX) || + (new_op_version < GD_OP_VERSION_MIN)) { + ret = -1; + snprintf (errstr, sizeof (errstr), + "Required op_version (%d) is not " + "supported", new_op_version); + gf_log (this->name, GF_LOG_ERROR, "%s", errstr); + goto out; + } + } + } + + ret = dict_get_int32 (dict, "count", &dict_count); if (ret) { - gf_log ("", GF_LOG_ERROR, "Count(dict),not set in Volume-Set"); + gf_log (this->name, GF_LOG_ERROR, + "Count(dict),not set in Volume-Set"); goto out; } - if ( dict_count == 0 ) { - /*No options would be specified of volume set help */ + if (dict_count == 0) { + /*No options would be specified of volume set help */ if (dict_get (dict, "help" )) { ret = 0; goto out; } if (dict_get (dict, "help-xml" )) { - #if (HAVE_LIB_XML) ret = 0; goto out; #else ret = -1; - gf_log ("", GF_LOG_ERROR, "libxml not present in the" - "system"); + gf_log (this->name, GF_LOG_ERROR, + "libxml not present in the system"); *op_errstr = gf_strdup ("Error: xml libraries not " "present to produce xml-output"); goto out; #endif } - gf_log ("", GF_LOG_ERROR, "No options received "); + gf_log (this->name, GF_LOG_ERROR, "No options received "); *op_errstr = gf_strdup ("Options not specified"); ret = -1; goto out; @@ -290,84 +650,171 @@ glusterd_op_stage_set_volume (dict_t *dict, char **op_errstr) ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } - exists = glusterd_check_volume_exists (volname); - if (!exists) { - snprintf (errstr, sizeof (errstr), "Volume %s does not exist", - volname); - gf_log ("", GF_LOG_ERROR, "%s", errstr); - *op_errstr = gf_strdup (errstr); - ret = -1; - goto out; - } + if (strcasecmp (volname, "all") != 0) { + exists = glusterd_check_volume_exists (volname); + if (!exists) { + snprintf (errstr, sizeof (errstr), + FMTSTR_CHECK_VOL_EXISTS, volname); + gf_log (this->name, GF_LOG_ERROR, "%s", errstr); + ret = -1; + goto out; + } - ret = glusterd_volinfo_find (volname, &volinfo); - if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to allocate memory"); - goto out; + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + FMTSTR_CHECK_VOL_EXISTS, volname); + goto out; + } + + ret = glusterd_validate_volume_id (dict, volinfo); + if (ret) + goto out; + } else { + all_vol = _gf_true; } + local_new_op_version = priv->op_version; + for ( count = 1; ret != 1 ; count++ ) { global_opt = _gf_false; sprintf (str, "key%d", count); ret = dict_get_str (dict, str, &key); - - if (ret) break; + sprintf (str, "value%d", count); + ret = dict_get_str (dict, str, &value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "invalid key,value pair in 'volume set'"); + ret = -1; + goto out; + } + + if (strcmp (key, "config.memory-accounting") == 0) { + gf_log (this->name, GF_LOG_DEBUG, + "enabling memory accounting for volume %s", + volname); + ret = 0; + } + + if (strcmp (key, "config.transport") == 0) { + gf_log (this->name, GF_LOG_DEBUG, + "changing transport-type for volume %s", + volname); + ret = 0; + /* if value is none of 'tcp/rdma/tcp,rdma' error out */ + if (!((strcasecmp (value, "rdma") == 0) || + (strcasecmp (value, "tcp") == 0) || + (strcasecmp (value, "tcp,rdma") == 0) || + (strcasecmp (value, "rdma,tcp") == 0))) { + ret = snprintf (errstr, sizeof (errstr), + "transport-type %s does " + "not exist", value); + /* lets not bother about above return value, + its a failure anyways */ + ret = -1; + goto out; + } + } + + if (is_key_glusterd_hooks_friendly (key)) + continue; + + for (vme = &glusterd_volopt_map[0]; vme->key; vme++) { + if ((vme->validate_fn) && + ((!strcmp (key, vme->key)) || + (!strcmp (key, strchr (vme->key, '.') + 1)))) { + ret = vme->validate_fn (dict, key, value, + op_errstr); + if (ret) + goto out; + break; + } + } + exists = glusterd_check_option_exists (key, &key_fixed); if (exists == -1) { ret = -1; goto out; } + if (!exists) { - gf_log ("", GF_LOG_ERROR, "Option with name: %s " - "does not exist", key); - ret = snprintf (errstr, 2048, + gf_log (this->name, GF_LOG_ERROR, + "Option with name: %s does not exist", key); + ret = snprintf (errstr, sizeof (errstr), "option : %s does not exist", key); if (key_fixed) - snprintf (errstr + ret, 2048 - ret, + snprintf (errstr + ret, sizeof (errstr) - ret, "\nDid you mean %s?", key_fixed); - *op_errstr = gf_strdup (errstr); - ret = -1; - goto out; - } - - sprintf (str, "value%d", count); - ret = dict_get_str (dict, str, &value); - - if (ret) { - gf_log ("", GF_LOG_ERROR, - "invalid key,value pair in 'volume set'"); ret = -1; goto out; } if (key_fixed) key = key_fixed; - - ret = glusterd_check_globaloption (key); + ALL_VOLUME_OPTION_CHECK (volname, key, ret, op_errstr, out); + ret = glusterd_validate_quorum_options (this, key, value, + op_errstr); if (ret) + goto out; + + local_key_op_version = glusterd_get_op_version_for_key (key); + if (local_key_op_version > local_new_op_version) + local_new_op_version = local_key_op_version; + + sprintf (str, "op-version%d", count); + if (origin_glusterd) { + ret = dict_set_uint32 (dict, str, local_key_op_version); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set key-op-version in dict"); + goto out; + } + } else if (check_op_version) { + ret = dict_get_uint32 (dict, str, &key_op_version); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get key-op-version from" + " dict"); + goto out; + } + if (local_key_op_version != key_op_version) { + ret = -1; + snprintf (errstr, sizeof (errstr), + "option: %s op-version mismatch", + key); + gf_log (this->name, GF_LOG_ERROR, + "%s, required op-version = %"PRIu32", " + "available op-version = %"PRIu32, + errstr, key_op_version, + local_key_op_version); + goto out; + } + } + + if (glusterd_check_globaloption (key)) global_opt = _gf_true; ret = dict_set_str (val_dict, key, value); if (ret) { - gf_log ("", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Unable to set the options in 'volume set'"); ret = -1; goto out; } *op_errstr = NULL; - if (!global_opt) + if (!global_opt && !all_vol) ret = glusterd_validate_reconfopts (volinfo, val_dict, op_errstr); - else { + else if (!all_vol) { voliter = NULL; list_for_each_entry (voliter, &priv->volumes, vol_list) { ret = glusterd_validate_globalopts (voliter, val_dict, op_errstr); @@ -377,8 +824,9 @@ glusterd_op_stage_set_volume (dict_t *dict, char **op_errstr) } if (ret) { - gf_log ("glusterd", GF_LOG_DEBUG, "Could not create temp " - "volfile, some option failed: %s", *op_errstr); + gf_log (this->name, GF_LOG_ERROR, "Could not create " + "temp volfile, some option failed: %s", + *op_errstr); goto out; } dict_del (val_dict, key); @@ -389,6 +837,35 @@ glusterd_op_stage_set_volume (dict_t *dict, char **op_errstr) } } + // Check if all the connected clients support the new op-version + ret = glusterd_check_client_op_version_support (volname, + local_new_op_version, + op_errstr); + if (ret) + goto out; + + if (origin_glusterd) { + ret = dict_set_uint32 (dict, "new-op-version", + local_new_op_version); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set new-op-version in dict"); + goto out; + } + /* Set this value in dict so other peers know to check for + * op-version. This is a hack for 3.3.x compatibility + * + * TODO: Remove this and the other places this is referred once + * 3.3.x compatibility is not required + */ + ret = dict_set_uint32 (dict, "check-op-version", + _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set check-op-version in dict"); + goto out; + } + } ret = 0; @@ -396,17 +873,18 @@ out: if (val_dict) dict_unref (val_dict); - if (key_fixed) - GF_FREE (key_fixed); + GF_FREE (key_fixed); + if (errstr[0] != '\0') + *op_errstr = gf_strdup (errstr); if (ret) { if (!(*op_errstr)) { *op_errstr = gf_strdup ("Error, Validation Failed"); - gf_log ("glsuterd", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "Error, Cannot Validate option :%s", *op_errstr); } else { - gf_log ("glsuterd", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "Error, Cannot Validate option"); } } @@ -422,28 +900,42 @@ glusterd_op_stage_reset_volume (dict_t *dict, char **op_errstr) char msg[2048] = {0}; char *key = NULL; char *key_fixed = NULL; + glusterd_volinfo_t *volinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } - exists = glusterd_check_volume_exists (volname); + if (strcasecmp (volname, "all") != 0) { + exists = glusterd_check_volume_exists (volname); + if (!exists) { + snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, + volname); + ret = -1; + goto out; + } + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, + volname); + goto out; + } - if (!exists) { - snprintf (msg, sizeof (msg), "Volume %s does not " - "exist", volname); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - ret = -1; - goto out; + ret = glusterd_validate_volume_id (dict, volinfo); + if (ret) + goto out; } ret = dict_get_str (dict, "key", &key); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "Unable to get option key"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get option key"); goto out; } if (strcmp(key, "all")) { @@ -453,24 +945,30 @@ glusterd_op_stage_reset_volume (dict_t *dict, char **op_errstr) goto out; } if (!exists) { - gf_log ("glusterd", GF_LOG_ERROR, - "Option %s does not exist", key); - ret = snprintf (msg, 2048, + ret = snprintf (msg, sizeof (msg), "Option %s does not exist", key); if (key_fixed) - snprintf (msg + ret, 2048 - ret, + snprintf (msg + ret, sizeof (msg) - ret, "\nDid you mean %s?", key_fixed); - *op_errstr = gf_strdup (msg); ret = -1; goto out; + } else if (exists > 0) { + if (key_fixed) + key = key_fixed; + ALL_VOLUME_OPTION_CHECK (volname, key, ret, + op_errstr, out); } } out: - if (key_fixed) - GF_FREE (key_fixed); + GF_FREE (key_fixed); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + if (msg[0] != '\0') { + gf_log (this->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + } + + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -486,6 +984,7 @@ glusterd_op_stage_sync_volume (dict_t *dict, char **op_errstr) gf_boolean_t exists = _gf_false; glusterd_peerinfo_t *peerinfo = NULL; char msg[2048] = {0,}; + glusterd_volinfo_t *volinfo = NULL; ret = dict_get_str (dict, "hostname", &hostname); if (ret) { @@ -495,8 +994,26 @@ glusterd_op_stage_sync_volume (dict_t *dict, char **op_errstr) goto out; } - ret = glusterd_is_local_addr (hostname); - if (ret) { + if (gf_is_local_addr (hostname)) { + //volname is not present in case of sync all + ret = dict_get_str (dict, "volname", &volname); + if (!ret) { + exists = glusterd_check_volume_exists (volname); + if (!exists) { + snprintf (msg, sizeof (msg), "Volume %s " + "does not exist", volname); + *op_errstr = gf_strdup (msg); + ret = -1; + goto out; + } + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) + goto out; + + } else { + ret = 0; + } + } else { ret = glusterd_friend_find (NULL, hostname, &peerinfo); if (ret) { snprintf (msg, sizeof (msg), "%s, is not a friend", @@ -512,22 +1029,7 @@ glusterd_op_stage_sync_volume (dict_t *dict, char **op_errstr) ret = -1; goto out; } - } else { - //volname is not present in case of sync all - ret = dict_get_str (dict, "volname", &volname); - if (!ret) { - exists = glusterd_check_volume_exists (volname); - if (!exists) { - snprintf (msg, sizeof (msg), "Volume %s " - "does not exist", volname); - *op_errstr = gf_strdup (msg); - ret = -1; - goto out; - } - } else { - ret = 0; - } } out: @@ -547,8 +1049,10 @@ glusterd_op_stage_status_volume (dict_t *dict, char **op_errstr) xlator_t *this = NULL; glusterd_conf_t *priv = NULL; glusterd_brickinfo_t *brickinfo = NULL; - glusterd_brickinfo_t *tmpbrickinfo = NULL; glusterd_volinfo_t *volinfo = NULL; + dict_t *vol_opts = NULL; + gf_boolean_t nfs_disabled = _gf_false; + gf_boolean_t shd_enabled = _gf_true; GF_ASSERT (dict); this = THIS; @@ -572,32 +1076,68 @@ glusterd_op_stage_status_volume (dict_t *dict, char **op_errstr) ret = glusterd_volinfo_find (volname, &volinfo); if (ret) { - snprintf (msg, sizeof(msg), "Volume %s does not exist", volname); + snprintf (msg, sizeof(msg), "Volume %s does not exist", + volname); gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); ret = -1; goto out; } - if ((cmd & GF_CLI_STATUS_BRICK) != 0) { - ret = dict_get_str (dict, "brick", &brick); - if (ret) + ret = glusterd_validate_volume_id (dict, volinfo); + if (ret) + goto out; + + ret = glusterd_is_volume_started (volinfo); + if (!ret) { + snprintf (msg, sizeof (msg), "Volume %s is not started", + volname); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + ret = -1; + goto out; + } + + vol_opts = volinfo->dict; + + if ((cmd & GF_CLI_STATUS_NFS) != 0) { + nfs_disabled = dict_get_str_boolean (vol_opts, "nfs.disable", + _gf_false); + if (nfs_disabled) { + ret = -1; + snprintf (msg, sizeof (msg), + "NFS server is disabled for volume %s", + volname); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + goto out; + } + } else if ((cmd & GF_CLI_STATUS_SHD) != 0) { + if (!glusterd_is_volume_replicate (volinfo)) { + ret = -1; + snprintf (msg, sizeof (msg), + "Volume %s is not of type replicate", + volname); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); goto out; + } - ret = glusterd_brickinfo_from_brick (brick, &brickinfo); - if (ret) { - snprintf (msg, sizeof (msg), "%s is not a brick", - brick); + shd_enabled = dict_get_str_boolean (vol_opts, + "cluster.self-heal-daemon", + _gf_true); + if (!shd_enabled) { + ret = -1; + snprintf (msg, sizeof (msg), + "Self-heal Daemon is disabled for volume %s", + volname); gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); goto out; } - ret = glusterd_volume_brickinfo_get (NULL, - brickinfo->hostname, - brickinfo->path, - volinfo, - &tmpbrickinfo, - GF_PATH_COMPLETE); + } else if ((cmd & GF_CLI_STATUS_BRICK) != 0) { + ret = dict_get_str (dict, "brick", &brick); + if (ret) + goto out; + ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, + &brickinfo); if (ret) { snprintf (msg, sizeof(msg), "No brick %s in" " volume %s", brick, volname); @@ -669,6 +1209,10 @@ glusterd_op_stage_stats_volume (dict_t *dict, char **op_errstr) goto out; } + ret = glusterd_validate_volume_id (dict, volinfo); + if (ret) + goto out; + ret = dict_get_int32 (dict, "op", &stats_op); if (ret) { snprintf (msg, sizeof (msg), "Volume profile op get failed"); @@ -715,60 +1259,82 @@ out: } -void +static int _delete_reconfig_opt (dict_t *this, char *key, data_t *value, void *data) { - int exists = 0; - int32_t is_force = 0; + int32_t *is_force = 0; GF_ASSERT (data); - is_force = *((int32_t*)data); - exists = glusterd_check_option_exists(key, NULL); - - if (exists != 1) - goto out; - - if ((!is_force) && - (_gf_true == glusterd_check_voloption_flags (key, - OPT_FLAG_FORCE))) - goto out; + is_force = (int32_t*)data; + + if (*is_force != 1) { + if (_gf_true == glusterd_check_voloption_flags (key, + OPT_FLAG_FORCE)) { + /* indicate to caller that we don't set the option + * due to being protected + */ + *is_force = *is_force | GD_OP_PROTECTED; + goto out; + } else { + *is_force = *is_force | GD_OP_UNPROTECTED; + } + } gf_log ("", GF_LOG_DEBUG, "deleting dict with key=%s,value=%s", key, value->data); dict_del (this, key); out: - return; + return 0; } -int +static int +_delete_reconfig_global_opt (dict_t *this, char *key, data_t *value, void *data) +{ + int32_t *is_force = 0; + + GF_ASSERT (data); + is_force = (int32_t*)data; + + if (strcmp (GLUSTERD_GLOBAL_OPT_VERSION, key) == 0) + goto out; + + _delete_reconfig_opt (this, key, value, data); +out: + return 0; +} + +static int glusterd_options_reset (glusterd_volinfo_t *volinfo, char *key, - int32_t is_force) + int32_t *is_force) { int ret = 0; data_t *value = NULL; + char *key_fixed = NULL; + xlator_t *this = NULL; - gf_log ("", GF_LOG_DEBUG, "Received volume set reset command"); - + this = THIS; + GF_ASSERT (this); GF_ASSERT (volinfo->dict); GF_ASSERT (key); if (!strncmp(key, "all", 3)) - dict_foreach (volinfo->dict, _delete_reconfig_opt, &is_force); + dict_foreach (volinfo->dict, _delete_reconfig_opt, is_force); else { value = dict_get (volinfo->dict, key); if (!value) { - gf_log ("glusterd", GF_LOG_ERROR, - "Could not get value"); + gf_log (this->name, GF_LOG_DEBUG, + "no value set for option %s", key); goto out; } - _delete_reconfig_opt (volinfo->dict, key, value, &is_force); + _delete_reconfig_opt (volinfo->dict, key, value, is_force); } - ret = glusterd_create_volfiles_and_notify_services (volinfo); + gd_update_volume_op_versions (volinfo); + ret = glusterd_create_volfiles_and_notify_services (volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to create volfile for" - " 'volume set'"); + gf_log (this->name, GF_LOG_ERROR, "Unable to create volfile for" + " 'volume reset'"); ret = -1; goto out; } @@ -786,23 +1352,120 @@ glusterd_options_reset (glusterd_volinfo_t *volinfo, char *key, ret = 0; out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + GF_FREE (key_fixed); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } +static int +glusterd_op_reset_all_volume_options (xlator_t *this, dict_t *dict) +{ + char *key = NULL; + char *key_fixed = NULL; + int ret = -1; + int32_t is_force = 0; + glusterd_conf_t *conf = NULL; + dict_t *dup_opt = NULL; + gf_boolean_t all = _gf_false; + char *next_version = NULL; + gf_boolean_t quorum_action = _gf_false; + + conf = this->private; + ret = dict_get_str (dict, "key", &key); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get key"); + goto out; + } + + ret = dict_get_int32 (dict, "force", &is_force); + if (ret) + is_force = 0; + + if (strcmp (key, "all")) { + ret = glusterd_check_option_exists (key, &key_fixed); + if (ret <= 0) { + gf_log (this->name, GF_LOG_ERROR, "Option %s does not " + "exist", key); + ret = -1; + goto out; + } + } else { + all = _gf_true; + } + + if (key_fixed) + key = key_fixed; + + ret = -1; + dup_opt = dict_new (); + if (!dup_opt) + goto out; + if (!all) { + dict_copy (conf->opts, dup_opt); + dict_del (dup_opt, key); + } + ret = glusterd_get_next_global_opt_version_str (conf->opts, + &next_version); + if (ret) + goto out; + + ret = dict_set_str (dup_opt, GLUSTERD_GLOBAL_OPT_VERSION, next_version); + if (ret) + goto out; + + ret = glusterd_store_options (this, dup_opt); + if (ret) + goto out; + + if (glusterd_is_quorum_changed (conf->opts, key, NULL)) + quorum_action = _gf_true; + + ret = dict_set_dynstr (conf->opts, GLUSTERD_GLOBAL_OPT_VERSION, + next_version); + if (ret) + goto out; + else + next_version = NULL; + + if (!all) { + dict_del (conf->opts, key); + } else { + dict_foreach (conf->opts, _delete_reconfig_global_opt, + &is_force); + } +out: + GF_FREE (key_fixed); + if (dup_opt) + dict_unref (dup_opt); + + gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret); + if (quorum_action) + glusterd_do_quorum_action (); + GF_FREE (next_version); + return ret; +} static int -glusterd_op_reset_volume (dict_t *dict) +glusterd_op_reset_volume (dict_t *dict, char **op_rspstr) { glusterd_volinfo_t *volinfo = NULL; int ret = -1; char *volname = NULL; char *key = NULL; + char *key_fixed = NULL; int32_t is_force = 0; + gf_boolean_t quorum_action = _gf_false; + xlator_t *this = NULL; + this = THIS; ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name " ); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name" ); + goto out; + } + + if (strcasecmp (volname, "all") == 0) { + ret = glusterd_op_reset_all_volume_options (this, dict); goto out; } @@ -812,32 +1475,64 @@ glusterd_op_reset_volume (dict_t *dict) ret = dict_get_str (dict, "key", &key); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "Unable to get option key"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get option key"); goto out; } ret = glusterd_volinfo_find (volname, &volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to allocate memory"); + gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS, + volname); goto out; } - ret = glusterd_options_reset (volinfo, key, is_force); + if (strcmp (key, "all") && + glusterd_check_option_exists (key, &key_fixed) != 1) { + gf_log (this->name, GF_LOG_ERROR, + "volinfo dict inconsistency: option %s not found", + key); + ret = -1; + goto out; + } + if (key_fixed) + key = key_fixed; + + if (glusterd_is_quorum_changed (volinfo->dict, key, NULL)) + quorum_action = _gf_true; + + ret = glusterd_options_reset (volinfo, key, &is_force); + if (ret == -1) { + gf_asprintf(op_rspstr, "Volume reset : failed"); + } else if (is_force & GD_OP_PROTECTED) { + if (is_force & GD_OP_UNPROTECTED) { + gf_asprintf (op_rspstr, "All unprotected fields were" + " reset. To reset the protected fields," + " use 'force'."); + } else { + ret = -1; + gf_asprintf (op_rspstr, "'%s' is protected. To reset" + " use 'force'.", key); + } + } out: - gf_log ("", GF_LOG_DEBUG, "'volume reset' returning %d", ret); - return ret; + GF_FREE (key_fixed); + if (quorum_action) + glusterd_do_quorum_action (); + gf_log (this->name, GF_LOG_DEBUG, "'volume reset' returning %d", ret); + return ret; } - int glusterd_stop_bricks (glusterd_volinfo_t *volinfo) { glusterd_brickinfo_t *brickinfo = NULL; list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { - if (glusterd_brick_stop (volinfo, brickinfo)) + /*TODO: Need to change @del_brick in brick_stop to _gf_true + * once we enable synctask in peer rpc prog */ + if (glusterd_brick_stop (volinfo, brickinfo, _gf_false)) return -1; } @@ -847,32 +1542,109 @@ glusterd_stop_bricks (glusterd_volinfo_t *volinfo) int glusterd_start_bricks (glusterd_volinfo_t *volinfo) { - glusterd_brickinfo_t *brickinfo = NULL; + int ret = -1; + glusterd_brickinfo_t *brickinfo = NULL; + + GF_ASSERT (volinfo); list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { - if (glusterd_brick_start (volinfo, brickinfo)) - return -1; + ret = glusterd_brick_start (volinfo, brickinfo, _gf_false); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to start %s:%s for %s", + brickinfo->hostname, brickinfo->path, + volinfo->volname); + goto out; + } } - return 0; + ret = 0; +out: + return ret; } static int -glusterd_volset_help (dict_t *dict) +glusterd_op_set_all_volume_options (xlator_t *this, dict_t *dict) { - int ret = -1; - gf_boolean_t xml_out = _gf_false; + char *key = NULL; + char *key_fixed = NULL; + char *value = NULL; + char *dup_value = NULL; + int ret = -1; + glusterd_conf_t *conf = NULL; + dict_t *dup_opt = NULL; + char *next_version = NULL; + gf_boolean_t quorum_action = _gf_false; + + conf = this->private; + ret = dict_get_str (dict, "key1", &key); + if (ret) + goto out; - if (dict_get (dict, "help" )) - xml_out = _gf_false; - else if (dict_get (dict, "help-xml" )) - xml_out = _gf_true; - else + ret = dict_get_str (dict, "value1", &value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "invalid key,value pair in 'volume set'"); + goto out; + } + ret = glusterd_check_option_exists (key, &key_fixed); + if (ret <= 0) { + gf_log (this->name, GF_LOG_ERROR, "Invalid key %s", key); + ret = -1; goto out; + } - ret = glusterd_get_volopt_content (xml_out); - out: - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + if (key_fixed) + key = key_fixed; + + ret = -1; + dup_opt = dict_new (); + if (!dup_opt) + goto out; + dict_copy (conf->opts, dup_opt); + ret = dict_set_str (dup_opt, key, value); + if (ret) + goto out; + + ret = glusterd_get_next_global_opt_version_str (conf->opts, + &next_version); + if (ret) + goto out; + + ret = dict_set_str (dup_opt, GLUSTERD_GLOBAL_OPT_VERSION, next_version); + if (ret) + goto out; + + dup_value = gf_strdup (value); + if (!dup_value) + goto out; + + ret = glusterd_store_options (this, dup_opt); + if (ret) + goto out; + + if (glusterd_is_quorum_changed (conf->opts, key, value)) + quorum_action = _gf_true; + + ret = dict_set_dynstr (conf->opts, GLUSTERD_GLOBAL_OPT_VERSION, + next_version); + if (ret) + goto out; + else + next_version = NULL; + + ret = dict_set_dynstr (conf->opts, key, dup_value); + if (ret) + goto out; +out: + GF_FREE (key_fixed); + if (dup_opt) + dict_unref (dup_opt); + + gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret); + if (quorum_action) + glusterd_do_quorum_action (); + GF_FREE (next_version); return ret; } @@ -889,9 +1661,14 @@ glusterd_op_set_volume (dict_t *dict) char *key_fixed = NULL; char *value = NULL; char str[50] = {0, }; + char *op_errstr = NULL; gf_boolean_t global_opt = _gf_false; + gf_boolean_t global_opts_set = _gf_false; glusterd_volinfo_t *voliter = NULL; int32_t dict_count = 0; + gf_boolean_t check_op_version = _gf_false; + uint32_t new_op_version = 0; + gf_boolean_t quorum_action = _gf_false; this = THIS; GF_ASSERT (this); @@ -901,68 +1678,111 @@ glusterd_op_set_volume (dict_t *dict) ret = dict_get_int32 (dict, "count", &dict_count); if (ret) { - gf_log ("", GF_LOG_ERROR, "Count(dict),not set in Volume-Set"); + gf_log (this->name, GF_LOG_ERROR, "Count(dict),not set in Volume-Set"); goto out; } - if ( dict_count == 0 ) { - ret = glusterd_volset_help (dict); - if (ret) - gf_log ("glusterd", GF_LOG_ERROR, "Volume set help" - "internal error"); + if (dict_count == 0) { + ret = glusterd_volset_help (NULL, &op_errstr); + if (ret) { + op_errstr = (op_errstr)? op_errstr: + "Volume set help internal error"; + gf_log (this->name, GF_LOG_ERROR, "%s", op_errstr); + } goto out; - } + } ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); + goto out; + } + + if (strcasecmp (volname, "all") == 0) { + ret = glusterd_op_set_all_volume_options (this, dict); goto out; } ret = glusterd_volinfo_find (volname, &volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to allocate memory"); + gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS, + volname); goto out; } - for ( count = 1; ret != -1 ; count++ ) { + // TODO: Remove this once v3.3 compatability is not required + check_op_version = dict_get_str_boolean (dict, "check-op-version", + _gf_false); + + if (check_op_version) { + ret = dict_get_uint32 (dict, "new-op-version", &new_op_version); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get new op-version from dict"); + goto out; + } + } + + for (count = 1; ret != -1 ; count++) { - global_opt = _gf_false; sprintf (str, "key%d", count); ret = dict_get_str (dict, str, &key); + if (ret) + break; + sprintf (str, "value%d", count); + ret = dict_get_str (dict, str, &value); if (ret) { - break; + gf_log (this->name, GF_LOG_ERROR, + "invalid key,value pair in 'volume set'"); + ret = -1; + goto out; } - if (!ret) { + if (strcmp (key, "config.memory-accounting") == 0) { + ret = gf_string2boolean (value, + &volinfo->memory_accounting); + } + + if (strcmp (key, "config.transport") == 0) { + gf_log (this->name, GF_LOG_INFO, + "changing transport-type for volume %s to %s", + volname, value); + ret = 0; + if (strcasecmp (value, "rdma") == 0) { + volinfo->transport_type = GF_TRANSPORT_RDMA; + } else if (strcasecmp (value, "tcp") == 0) { + volinfo->transport_type = GF_TRANSPORT_TCP; + } else if ((strcasecmp (value, "tcp,rdma") == 0) || + (strcasecmp (value, "rdma,tcp") == 0)) { + volinfo->transport_type = + GF_TRANSPORT_BOTH_TCP_RDMA; + } else { + ret = -1; + goto out; + } + } + + if (!is_key_glusterd_hooks_friendly (key)) { ret = glusterd_check_option_exists (key, &key_fixed); GF_ASSERT (ret); - if (ret == -1) { + if (ret <= 0) { key_fixed = NULL; goto out; } - ret = 0; } - ret = glusterd_check_globaloption (key); - if (ret) + global_opt = _gf_false; + if (glusterd_check_globaloption (key)) { global_opt = _gf_true; - - sprintf (str, "value%d", count); - ret = dict_get_str (dict, str, &value); - if (ret) { - gf_log ("", GF_LOG_ERROR, - "invalid key,value pair in 'volume set'"); - ret = -1; - goto out; + global_opts_set = _gf_true; } if (!global_opt) value = gf_strdup (value); if (!value) { - gf_log ("", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Unable to set the options in 'volume set'"); ret = -1; goto out; @@ -971,6 +1791,9 @@ glusterd_op_set_volume (dict_t *dict) if (key_fixed) key = key_fixed; + if (glusterd_is_quorum_changed (volinfo->dict, key, value)) + quorum_action = _gf_true; + if (global_opt) { list_for_each_entry (voliter, &priv->volumes, vol_list) { value = gf_strdup (value); @@ -978,8 +1801,7 @@ glusterd_op_set_volume (dict_t *dict) if (ret) goto out; } - } - else { + } else { ret = dict_set_dynstr (volinfo->dict, key, value); if (ret) goto out; @@ -987,21 +1809,35 @@ glusterd_op_set_volume (dict_t *dict) if (key_fixed) { GF_FREE (key_fixed); - key_fixed = NULL; } } - if ( count == 1 ) { - gf_log ("", GF_LOG_ERROR, "No options received "); + if (count == 1) { + gf_log (this->name, GF_LOG_ERROR, "No options received "); ret = -1; goto out; } - if (!global_opt) { + /* Update the cluster op-version before regenerating volfiles so that + * correct volfiles are generated + */ + if (new_op_version > priv->op_version) { + priv->op_version = new_op_version; + ret = glusterd_store_global_info (this); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to store op-version"); + goto out; + } + } + + if (!global_opts_set) { + gd_update_volume_op_versions (volinfo); ret = glusterd_create_volfiles_and_notify_services (volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to create volfile for" + gf_log (this->name, GF_LOG_ERROR, + "Unable to create volfile for" " 'volume set'"); ret = -1; goto out; @@ -1014,19 +1850,20 @@ glusterd_op_set_volume (dict_t *dict) if (GLUSTERD_STATUS_STARTED == volinfo->status) { ret = glusterd_nodesvcs_handle_reconfigure (volinfo); if (ret) { - gf_log ("", GF_LOG_WARNING, + gf_log (this->name, GF_LOG_WARNING, "Unable to restart NFS-Server"); goto out; } } - } - else { + } else { list_for_each_entry (voliter, &priv->volumes, vol_list) { volinfo = voliter; + gd_update_volume_op_versions (volinfo); ret = glusterd_create_volfiles_and_notify_services (volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to create volfile for" + gf_log (this->name, GF_LOG_ERROR, + "Unable to create volfile for" " 'volume set'"); ret = -1; goto out; @@ -1040,7 +1877,7 @@ glusterd_op_set_volume (dict_t *dict) if (GLUSTERD_STATUS_STARTED == volinfo->status) { ret = glusterd_nodesvcs_handle_reconfigure (volinfo); if (ret) { - gf_log ("", GF_LOG_WARNING, + gf_log (this->name, GF_LOG_WARNING, "Unable to restart NFS-Server"); goto out; } @@ -1048,11 +1885,11 @@ glusterd_op_set_volume (dict_t *dict) } } - ret = 0; out: - if (key_fixed) - GF_FREE (key_fixed); - gf_log ("", GF_LOG_DEBUG, "returning %d", ret); + GF_FREE (key_fixed); + gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret); + if (quorum_action) + glusterd_do_quorum_action (); return ret; } @@ -1084,7 +1921,7 @@ glusterd_op_sync_volume (dict_t *dict, char **op_errstr, goto out; } - if (glusterd_is_local_addr (hostname)) { + if (!gf_is_local_addr (hostname)) { ret = 0; goto out; } @@ -1255,12 +2092,249 @@ out: } static int +_add_brick_name_to_dict (dict_t *dict, char *key, glusterd_brickinfo_t *brick) +{ + int ret = -1; + char tmp[1024] = {0,}; + char *brickname = NULL; + xlator_t *this = NULL; + + GF_ASSERT (dict); + GF_ASSERT (key); + GF_ASSERT (brick); + + this = THIS; + GF_ASSERT (this); + + snprintf (tmp, sizeof (tmp), "%s:%s", brick->hostname, brick->path); + brickname = gf_strdup (tmp); + if (!brickname) { + gf_log (this->name, GF_LOG_ERROR, "Failed to dup brick name"); + goto out; + } + + ret = dict_set_dynstr (dict, key, brickname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add brick name to dict"); + goto out; + } + brickname = NULL; +out: + if (brickname) + GF_FREE (brickname); + return ret; +} + +static int +_add_remove_bricks_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, + char *prefix) +{ + int ret = -1; + int count = 0; + int i = 0; + char brick_key[1024] = {0,}; + char dict_key[1024] ={0,}; + char *brick = NULL; + xlator_t *this = NULL; + + GF_ASSERT (dict); + GF_ASSERT (volinfo); + GF_ASSERT (prefix); + + this = THIS; + GF_ASSERT (this); + + ret = dict_get_int32 (volinfo->rebal.dict, "count", &count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get brick count"); + goto out; + } + + snprintf (dict_key, sizeof (dict_key), "%s.count", prefix); + ret = dict_set_int32 (dict, dict_key, count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set brick count in dict"); + goto out; + } + + for (i = 1; i <= count; i++) { + memset (brick_key, 0, sizeof (brick_key)); + snprintf (brick_key, sizeof (brick_key), "brick%d", i); + + ret = dict_get_str (volinfo->rebal.dict, brick_key, &brick); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get %s", brick_key); + goto out; + } + + memset (dict_key, 0, sizeof (dict_key)); + snprintf (dict_key, sizeof (dict_key), "%s.%s", prefix, + brick_key); + ret = dict_set_str (dict, dict_key, brick); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add brick to dict"); + goto out; + } + brick = NULL; + } + +out: + return ret; +} + +/* This adds the respective task-id and all available parameters of a task into + * a dictionary + */ +static int +_add_task_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, int op, int index) +{ + + int ret = -1; + char key[128] = {0,}; + char *uuid_str = NULL; + int status = 0; + xlator_t *this = NULL; + + GF_ASSERT (dict); + GF_ASSERT (volinfo); + + this = THIS; + GF_ASSERT (this); + + switch (op) { + case GD_OP_REMOVE_BRICK: + snprintf (key, sizeof (key), "task%d", index); + ret = _add_remove_bricks_to_dict (dict, volinfo, key); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add remove bricks to dict"); + goto out; + } + case GD_OP_REBALANCE: + uuid_str = gf_strdup (uuid_utoa (volinfo->rebal.rebalance_id)); + status = volinfo->rebal.defrag_status; + break; + + case GD_OP_REPLACE_BRICK: + snprintf (key, sizeof (key), "task%d.src-brick", index); + ret = _add_brick_name_to_dict (dict, key, + volinfo->rep_brick.src_brick); + if (ret) + goto out; + memset (key, 0, sizeof (key)); + + snprintf (key, sizeof (key), "task%d.dst-brick", index); + ret = _add_brick_name_to_dict (dict, key, + volinfo->rep_brick.dst_brick); + if (ret) + goto out; + memset (key, 0, sizeof (key)); + + uuid_str = gf_strdup (uuid_utoa (volinfo->rep_brick.rb_id)); + status = volinfo->rep_brick.rb_status; + break; + + default: + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "%s operation doesn't have a" + " task_id", gd_op_list[op]); + goto out; + } + + snprintf (key, sizeof (key), "task%d.type", index); + ret = dict_set_str (dict, key, (char *)gd_op_list[op]); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting task type in dict"); + goto out; + } + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "task%d.id", index); + + if (!uuid_str) + goto out; + ret = dict_set_dynstr (dict, key, uuid_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting task id in dict"); + goto out; + } + uuid_str = NULL; + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "task%d.status", index); + ret = dict_set_int32 (dict, key, status); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting task status in dict"); + goto out; + } + +out: + if (uuid_str) + GF_FREE (uuid_str); + return ret; +} + +static int +glusterd_aggregate_task_status (dict_t *rsp_dict, glusterd_volinfo_t *volinfo) +{ + int ret = -1; + int tasks = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + if (!uuid_is_null (volinfo->rebal.rebalance_id)) { + ret = _add_task_to_dict (rsp_dict, volinfo, volinfo->rebal.op, + tasks); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add task details to dict"); + goto out; + } + tasks++; + } + + if (!uuid_is_null (volinfo->rep_brick.rb_id)) { + ret = _add_task_to_dict (rsp_dict, volinfo, GD_OP_REPLACE_BRICK, + tasks); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add task details to dict"); + goto out; + } + tasks++; + } + + ret = dict_set_int32 (rsp_dict, "tasks", tasks); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting tasks count in dict"); + goto out; + } + ret = 0; + +out: + return ret; +} + +static int glusterd_op_status_volume (dict_t *dict, char **op_errstr, dict_t *rsp_dict) { int ret = -1; - int brick_count = 0; + int node_count = 0; int brick_index = -1; + int other_count = 0; + int other_index = 0; uint32_t cmd = 0; char *volname = NULL; char *brick = NULL; @@ -1268,6 +2342,10 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, glusterd_volinfo_t *volinfo = NULL; glusterd_brickinfo_t *brickinfo = NULL; glusterd_conf_t *priv = NULL; + dict_t *vol_opts = NULL; + gf_boolean_t nfs_disabled = _gf_false; + gf_boolean_t shd_enabled = _gf_true; + gf_boolean_t origin_glusterd = _gf_false; this = THIS; GF_ASSERT (this); @@ -1277,23 +2355,21 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, GF_ASSERT (dict); + origin_glusterd = is_origin_glusterd (dict); + ret = dict_get_uint32 (dict, "cmd", &cmd); if (ret) goto out; - if (!rsp_dict) { - //this should happen only on source + if (origin_glusterd) { ret = 0; - rsp_dict = glusterd_op_get_ctx (); - if ((cmd & GF_CLI_STATUS_ALL)) { ret = glusterd_get_all_volnames (rsp_dict); if (ret) - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to get all volume " "names for status"); } - } ret = dict_set_uint32 (rsp_dict, "cmd", cmd); @@ -1309,37 +2385,57 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, ret = glusterd_volinfo_find (volname, &volinfo); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Volume with name: %s " + gf_log (this->name, GF_LOG_ERROR, "Volume with name: %s " "does not exist", volname); goto out; } + vol_opts = volinfo->dict; - if ((cmd & GF_CLI_STATUS_BRICK) != 0) { + if ((cmd & GF_CLI_STATUS_NFS) != 0) { + ret = glusterd_add_node_to_dict ("nfs", rsp_dict, 0, vol_opts); + if (ret) + goto out; + other_count++; + node_count++; + + } else if ((cmd & GF_CLI_STATUS_SHD) != 0) { + ret = glusterd_add_node_to_dict ("glustershd", rsp_dict, 0, + vol_opts); + if (ret) + goto out; + other_count++; + node_count++; + + } else if ((cmd & GF_CLI_STATUS_BRICK) != 0) { ret = dict_get_str (dict, "brick", &brick); if (ret) goto out; ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, - &brickinfo, - GF_PATH_COMPLETE); + &brickinfo); if (ret) goto out; - if (uuid_compare (brickinfo->uuid, priv->uuid)) + if (uuid_compare (brickinfo->uuid, MY_UUID)) goto out; - glusterd_add_brick_to_dict (volinfo, brickinfo, rsp_dict, 0); + glusterd_add_brick_to_dict (volinfo, brickinfo, rsp_dict, + ++brick_index); if (cmd & GF_CLI_STATUS_DETAIL) glusterd_add_brick_detail_to_dict (volinfo, brickinfo, - rsp_dict, 0); - ret = dict_set_int32 (rsp_dict, "count", 1); + rsp_dict, + brick_index); + node_count++; + } else if ((cmd & GF_CLI_STATUS_TASKS) != 0) { + ret = glusterd_aggregate_task_status (rsp_dict, volinfo); goto out; + } else { list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { brick_index++; - if (uuid_compare (brickinfo->uuid, priv->uuid)) + if (uuid_compare (brickinfo->uuid, MY_UUID)) continue; glusterd_add_brick_to_dict (volinfo, brickinfo, @@ -1351,14 +2447,79 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, rsp_dict, brick_index); } - brick_count++; + node_count++; } + + if ((cmd & GF_CLI_STATUS_MASK) == GF_CLI_STATUS_NONE) { + other_index = brick_index + 1; + + nfs_disabled = dict_get_str_boolean (vol_opts, + "nfs.disable", + _gf_false); + if (!nfs_disabled) { + ret = glusterd_add_node_to_dict ("nfs", + rsp_dict, + other_index, + vol_opts); + if (ret) + goto out; + other_index++; + other_count++; + node_count++; + } + + shd_enabled = dict_get_str_boolean + (vol_opts, "cluster.self-heal-daemon", + _gf_true); + if (glusterd_is_volume_replicate (volinfo) + && shd_enabled) { + ret = glusterd_add_node_to_dict ("glustershd", + rsp_dict, + other_index, + vol_opts); + if (ret) + goto out; + other_count++; + node_count++; + } + } + } + + ret = dict_set_int32 (rsp_dict, "brick-index-max", brick_index); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting brick-index-max to dict"); + goto out; + } + ret = dict_set_int32 (rsp_dict, "other-count", other_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting other-count to dict"); + goto out; + } + ret = dict_set_int32 (rsp_dict, "count", node_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting node count to dict"); + goto out; } - ret = dict_set_int32 (rsp_dict, "count", brick_count); + /* Active tasks */ + /* Tasks are added only for normal volume status request for either a + * single volume or all volumes, and only by the origin glusterd + */ + if (((cmd & GF_CLI_STATUS_MASK) != GF_CLI_STATUS_NONE) || + !(cmd & (GF_CLI_STATUS_VOL | GF_CLI_STATUS_ALL)) || + !origin_glusterd) + goto out; + + ret = glusterd_aggregate_task_status (rsp_dict, volinfo); + if (ret) + goto out; + ret = 0; out: - gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -1368,7 +2529,7 @@ glusterd_op_ac_none (glusterd_op_sm_event_t *event, void *ctx) { int ret = 0; - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } @@ -1382,6 +2543,7 @@ glusterd_op_ac_send_lock (glusterd_op_sm_event_t *event, void *ctx) xlator_t *this = NULL; glusterd_peerinfo_t *peerinfo = NULL; uint32_t pending_count = 0; + dict_t *dict = NULL; this = THIS; priv = this->private; @@ -1396,21 +2558,60 @@ glusterd_op_ac_send_lock (glusterd_op_sm_event_t *event, void *ctx) (glusterd_op_get_op() != GD_OP_SYNC_VOLUME)) continue; - proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_LOCK]; - if (proc->fn) { - ret = proc->fn (NULL, this, peerinfo); - if (ret) - continue; - pending_count++; + /* Based on the op_version, acquire a cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_LOCK]; + if (proc->fn) { + ret = proc->fn (NULL, this, peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send lock request " + "for operation 'Volume %s' to " + "peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + continue; + } + pending_count++; + } + } else { + dict = glusterd_op_get_ctx (); + dict_ref (dict); + + proc = &peerinfo->mgmt_v3->proctable + [GLUSTERD_MGMT_V3_LOCK]; + if (proc->fn) { + ret = dict_set_static_ptr (dict, "peerinfo", + peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to set peerinfo"); + dict_unref (dict); + goto out; + } + + ret = proc->fn (NULL, this, dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send mgmt_v3 lock " + "request for operation " + "'Volume %s' to peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + dict_unref (dict); + continue; + } + pending_count++; + } } } opinfo.pending_count = pending_count; if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_all_acc (); - - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } @@ -1423,17 +2624,12 @@ glusterd_op_ac_send_unlock (glusterd_op_sm_event_t *event, void *ctx) xlator_t *this = NULL; glusterd_peerinfo_t *peerinfo = NULL; uint32_t pending_count = 0; + dict_t *dict = NULL; this = THIS; priv = this->private; GF_ASSERT (priv); - /*ret = glusterd_unlock (priv->uuid); - - if (ret) - goto out; - */ - list_for_each_entry (peerinfo, &priv->peers, uuid_list) { GF_ASSERT (peerinfo); @@ -1443,23 +2639,62 @@ glusterd_op_ac_send_unlock (glusterd_op_sm_event_t *event, void *ctx) (glusterd_op_get_op() != GD_OP_SYNC_VOLUME)) continue; - proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_UNLOCK]; - if (proc->fn) { - ret = proc->fn (NULL, this, peerinfo); - if (ret) - continue; - pending_count++; + /* Based on the op_version, release the * + * cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_UNLOCK]; + if (proc->fn) { + ret = proc->fn (NULL, this, peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send unlock request " + "for operation 'Volume %s' to " + "peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + continue; + } + pending_count++; + } + } else { + dict = glusterd_op_get_ctx (); + dict_ref (dict); + + proc = &peerinfo->mgmt_v3->proctable + [GLUSTERD_MGMT_V3_UNLOCK]; + if (proc->fn) { + ret = dict_set_static_ptr (dict, "peerinfo", + peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to set peerinfo"); + dict_unref (dict); + goto out; + } + + ret = proc->fn (NULL, this, dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send mgmt_v3 unlock " + "request for operation " + "'Volume %s' to peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + dict_unref (dict); + continue; + } + pending_count++; + } } } opinfo.pending_count = pending_count; if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_all_acc (); - - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; - } static int @@ -1471,9 +2706,10 @@ glusterd_op_ac_ack_drain (glusterd_op_sm_event_t *event, void *ctx) opinfo.pending_count--; if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, NULL); - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } @@ -1487,41 +2723,97 @@ glusterd_op_ac_send_unlock_drain (glusterd_op_sm_event_t *event, void *ctx) static int glusterd_op_ac_lock (glusterd_op_sm_event_t *event, void *ctx) { - glusterd_op_lock_ctx_t *lock_ctx = NULL; - int32_t ret = 0; - + int32_t ret = 0; + char *volname = NULL; + glusterd_op_lock_ctx_t *lock_ctx = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; GF_ASSERT (event); GF_ASSERT (ctx); + this = THIS; + priv = this->private; + lock_ctx = (glusterd_op_lock_ctx_t *)ctx; - ret = glusterd_lock (lock_ctx->uuid); + /* If the req came from a node running on older op_version + * the dict won't be present. Based on it acquiring a cluster + * or mgmt_v3 lock */ + if (lock_ctx->dict == NULL) { + ret = glusterd_lock (lock_ctx->uuid); + glusterd_op_lock_send_resp (lock_ctx->req, ret); + } else { + ret = dict_get_str (lock_ctx->dict, "volname", &volname); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire volname"); + else { + ret = glusterd_mgmt_v3_lock (volname, lock_ctx->uuid, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock for %s", + volname); + } - gf_log ("", GF_LOG_DEBUG, "Lock Returned %d", ret); + glusterd_op_mgmt_v3_lock_send_resp (lock_ctx->req, + &event->txn_id, ret); - glusterd_op_lock_send_resp (lock_ctx->req, ret); + dict_unref (lock_ctx->dict); + } + gf_log (THIS->name, GF_LOG_DEBUG, "Lock Returned %d", ret); return ret; } static int glusterd_op_ac_unlock (glusterd_op_sm_event_t *event, void *ctx) { - int ret = 0; - glusterd_op_lock_ctx_t *lock_ctx = NULL; + int32_t ret = 0; + char *volname = NULL; + glusterd_op_lock_ctx_t *lock_ctx = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + GF_ASSERT (event); GF_ASSERT (ctx); + this = THIS; + priv = this->private; + lock_ctx = (glusterd_op_lock_ctx_t *)ctx; - ret = glusterd_unlock (lock_ctx->uuid); + /* If the req came from a node running on older op_version + * the dict won't be present. Based on it releasing the cluster + * or mgmt_v3 lock */ + if (lock_ctx->dict == NULL) { + ret = glusterd_unlock (lock_ctx->uuid); + glusterd_op_unlock_send_resp (lock_ctx->req, ret); + } else { + ret = dict_get_str (lock_ctx->dict, "volname", &volname); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire volname"); + else { + ret = glusterd_mgmt_v3_unlock (volname, lock_ctx->uuid, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to release lock for %s", volname); + } - gf_log ("", GF_LOG_DEBUG, "Unlock Returned %d", ret); + glusterd_op_mgmt_v3_unlock_send_resp (lock_ctx->req, + &event->txn_id, ret); - glusterd_op_unlock_send_resp (lock_ctx->req, ret); + dict_unref (lock_ctx->dict); + } + + gf_log (this->name, GF_LOG_DEBUG, "Unlock Returned %d", ret); + if (priv->pending_quorum_action) + glusterd_do_quorum_action (); return ret; } @@ -1538,7 +2830,7 @@ glusterd_op_ac_local_unlock (glusterd_op_sm_event_t *event, void *ctx) ret = glusterd_unlock (*originator); - gf_log ("", GF_LOG_DEBUG, "Unlock Returned %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Unlock Returned %d", ret); return ret; } @@ -1556,72 +2848,243 @@ glusterd_op_ac_rcvd_lock_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, + &event->txn_id, NULL); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); out: return ret; } +static int +glusterd_dict_set_volid (dict_t *dict, char *volname, char **op_errstr) +{ + int ret = -1; + glusterd_volinfo_t *volinfo = NULL; + char *volid = NULL; + char msg[1024] = {0,}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + if (!dict || !volname) + goto out; + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname); + goto out; + } + volid = gf_strdup (uuid_utoa (volinfo->volume_id)); + if (!volid) { + ret = -1; + goto out; + } + ret = dict_set_dynstr (dict, "vol-id", volid); + if (ret) { + snprintf (msg, sizeof (msg), "Failed to set volume id of volume" + " %s", volname); + goto out; + } +out: + if (msg[0] != '\0') { + gf_log (this->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + } + return ret; +} + int -glusterd_op_build_payload (dict_t **req) +glusterd_op_build_payload (dict_t **req, char **op_errstr, dict_t *op_ctx) { int ret = -1; void *ctx = NULL; + dict_t *dict = NULL; dict_t *req_dict = NULL; glusterd_op_t op = GD_OP_NONE; + char *volname = NULL; + uint32_t status_cmd = GF_CLI_STATUS_NONE; + char *errstr = NULL; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); + req_dict = dict_new (); if (!req_dict) goto out; - op = glusterd_op_get_op (); - ctx = (void*)glusterd_op_get_ctx (); - if (!ctx) { - gf_log ("", GF_LOG_ERROR, "Null Context for " - "op %d", op); - ret = -1; - goto out; + if (!op_ctx) { + op = glusterd_op_get_op (); + ctx = (void*)glusterd_op_get_ctx (); + if (!ctx) { + gf_log (this->name, GF_LOG_ERROR, "Null Context for " + "op %d", op); + ret = -1; + goto out; + } + + } else { +#define GD_SYNC_OPCODE_KEY "sync-mgmt-operation" + ret = dict_get_int32 (op_ctx, GD_SYNC_OPCODE_KEY, (int32_t*)&op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volume" + " operation"); + goto out; + } + ctx = op_ctx; +#undef GD_SYNC_OPCODE_KEY } + dict = ctx; switch (op) { case GD_OP_CREATE_VOLUME: { - dict_t *dict = ctx; ++glusterfs_port; - ret = dict_set_int32 (dict, "port", glusterfs_port); - if (ret) + ret = dict_set_int32 (dict, "port", + glusterfs_port); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set port in " + "dictionary"); + goto out; + } + dict_copy (dict, req_dict); + } + break; + + case GD_OP_GSYNC_CREATE: + case GD_OP_GSYNC_SET: + { + ret = glusterd_op_gsync_args_get (dict, + &errstr, + &volname, + NULL, NULL); + if (ret == 0) { + ret = glusterd_dict_set_volid + (dict, volname, op_errstr); + if (ret) + goto out; + } + dict_copy (dict, req_dict); + } + break; + + case GD_OP_SET_VOLUME: + { + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, + "volname is not present in " + "operation ctx"); goto out; + } + if (strcmp (volname, "help") && + strcmp (volname, "help-xml") && + strcasecmp (volname, "all")) { + ret = glusterd_dict_set_volid + (dict, volname, op_errstr); + if (ret) + goto out; + } + dict_destroy (req_dict); + req_dict = dict_ref (dict); + } + break; + + case GD_OP_SYNC_VOLUME: + { dict_copy (dict, req_dict); + break; + } + + case GD_OP_REMOVE_BRICK: + { + dict_t *dict = ctx; + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, + "volname is not present in " + "operation ctx"); + goto out; + } + + ret = glusterd_dict_set_volid (dict, volname, + op_errstr); + if (ret) + goto out; + + dict_destroy (req_dict); + req_dict = dict_ref (dict); } break; + case GD_OP_STATUS_VOLUME: + { + ret = dict_get_uint32 (dict, "cmd", + &status_cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Status command not present " + "in op ctx"); + goto out; + } + if (GF_CLI_STATUS_ALL & status_cmd) { + dict_copy (dict, req_dict); + break; + } + } + /*fall-through*/ case GD_OP_DELETE_VOLUME: case GD_OP_START_VOLUME: case GD_OP_STOP_VOLUME: case GD_OP_ADD_BRICK: case GD_OP_REPLACE_BRICK: - case GD_OP_SET_VOLUME: case GD_OP_RESET_VOLUME: - case GD_OP_REMOVE_BRICK: case GD_OP_LOG_ROTATE: - case GD_OP_SYNC_VOLUME: case GD_OP_QUOTA: - case GD_OP_GSYNC_SET: case GD_OP_PROFILE_VOLUME: - case GD_OP_STATUS_VOLUME: case GD_OP_REBALANCE: case GD_OP_HEAL_VOLUME: case GD_OP_STATEDUMP_VOLUME: + case GD_OP_CLEARLOCKS_VOLUME: + case GD_OP_DEFRAG_BRICK_VOLUME: { - dict_t *dict = ctx; + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, + "volname is not present in " + "operation ctx"); + goto out; + } + + if (strcasecmp (volname, "all")) { + ret = glusterd_dict_set_volid (dict, + volname, + op_errstr); + if (ret) + goto out; + } dict_copy (dict, req_dict); } break; + case GD_OP_COPY_FILE: + { + dict_copy (dict, req_dict); + break; + } + + case GD_OP_SYS_EXEC: + { + dict_copy (dict, req_dict); + break; + } + default: break; } @@ -1633,6 +3096,105 @@ out: return ret; } +gf_boolean_t +glusterd_is_get_op (xlator_t *this, glusterd_op_t op, dict_t *dict) +{ + char *key = NULL; + char *volname = NULL; + int ret = 0; + + if (op == GD_OP_STATUS_VOLUME) + return _gf_true; + + if ((op == GD_OP_SET_VOLUME)) { + //check for set volume help + ret = dict_get_str (dict, "volname", &volname); + if (volname && + ((strcmp (volname, "help") == 0) || + (strcmp (volname, "help-xml") == 0))) { + ret = dict_get_str (dict, "key1", &key); + if (ret < 0) + return _gf_true; + } + } + + return _gf_false; +} + +gf_boolean_t +glusterd_is_op_quorum_validation_required (xlator_t *this, glusterd_op_t op, + dict_t *dict) +{ + gf_boolean_t required = _gf_true; + char *key = NULL; + char *key_fixed = NULL; + int ret = -1; + + if (glusterd_is_get_op (this, op, dict)) { + required = _gf_false; + goto out; + } + if ((op != GD_OP_SET_VOLUME) && (op != GD_OP_RESET_VOLUME)) + goto out; + if (op == GD_OP_SET_VOLUME) + ret = dict_get_str (dict, "key1", &key); + else if (op == GD_OP_RESET_VOLUME) + ret = dict_get_str (dict, "key", &key); + if (ret) + goto out; + ret = glusterd_check_option_exists (key, &key_fixed); + if (ret <= 0) + goto out; + if (key_fixed) + key = key_fixed; + if (glusterd_is_quorum_option (key)) + required = _gf_false; +out: + GF_FREE (key_fixed); + return required; +} + +static int +glusterd_op_validate_quorum (xlator_t *this, glusterd_op_t op, + dict_t *dict, char **op_errstr) +{ + int ret = 0; + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + char *errstr = NULL; + + + errstr = "Quorum not met. Volume operation not allowed."; + if (!glusterd_is_op_quorum_validation_required (this, op, dict)) + goto out; + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + ret = 0; + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + ret = 0; + goto out; + } + + if (does_gd_meet_server_quorum (this)) { + ret = 0; + goto out; + } + + if (glusterd_is_volume_in_server_quorum (volinfo)) { + ret = -1; + *op_errstr = gf_strdup (errstr); + goto out; + } + ret = 0; +out: + return ret; +} + static int glusterd_op_ac_send_stage_op (glusterd_op_sm_event_t *event, void *ctx) { @@ -1653,14 +3215,32 @@ glusterd_op_ac_send_stage_op (glusterd_op_sm_event_t *event, void *ctx) op = glusterd_op_get_op (); - ret = glusterd_op_build_payload (&dict); - if (ret) + ret = glusterd_op_build_payload (&dict, &op_errstr, NULL); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD, + gd_op_list[op]); + if (op_errstr == NULL) + gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD); + opinfo.op_errstr = op_errstr; goto out; + } + + ret = glusterd_op_validate_quorum (this, op, dict, &op_errstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", op_errstr); + opinfo.op_errstr = op_errstr; + goto out; + } /* rsp_dict NULL from source */ ret = glusterd_op_stage_validate (op, dict, &op_errstr, NULL); if (ret) { - gf_log ("", GF_LOG_ERROR, "Staging failed"); + gf_log (this->name, GF_LOG_ERROR, LOGSTR_STAGE_FAIL, + gd_op_list[op], "localhost", + (op_errstr) ? ":" : " ", (op_errstr) ? op_errstr : " "); + if (op_errstr == NULL) + gf_asprintf (&op_errstr, OPERRSTR_STAGE_FAIL, + "localhost"); opinfo.op_errstr = op_errstr; goto out; } @@ -1679,13 +3259,19 @@ glusterd_op_ac_send_stage_op (glusterd_op_sm_event_t *event, void *ctx) if (proc->fn) { ret = dict_set_static_ptr (dict, "peerinfo", peerinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "failed to set peerinfo"); + gf_log (this->name, GF_LOG_ERROR, "failed to " + "set peerinfo"); goto out; } ret = proc->fn (NULL, this, dict); - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Failed to " + "send stage request for operation " + "'Volume %s' to peer %s", + gd_op_list[op], peerinfo->hostname); continue; + } pending_count++; } } @@ -1695,27 +3281,29 @@ out: if (dict) dict_unref (dict); if (ret) { - glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); + glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + &event->txn_id, NULL); opinfo.op_ret = ret; } - gf_log ("glusterd", GF_LOG_INFO, "Sent op req to %d peers", + gf_log (this->name, GF_LOG_DEBUG, "Sent stage op request for " + "'Volume %s' to %d peers", gd_op_list[op], opinfo.pending_count); if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } static int32_t -glusterd_op_start_rb_timer (dict_t *dict) +glusterd_op_start_rb_timer (dict_t *dict, uuid_t *txn_id) { int32_t op = 0; - struct timeval timeout = {0, }; + struct timespec timeout = {0, }; glusterd_conf_t *priv = NULL; int32_t ret = -1; dict_t *rb_ctx = NULL; @@ -1730,13 +3318,13 @@ glusterd_op_start_rb_timer (dict_t *dict) goto out; } - if (op == GF_REPLACE_OP_START || - op == GF_REPLACE_OP_ABORT) - timeout.tv_sec = 5; - else - timeout.tv_sec = 1; + if (op != GF_REPLACE_OP_START) { + ret = glusterd_op_sm_inject_all_acc (txn_id); + goto out; + } - timeout.tv_usec = 0; + timeout.tv_sec = 5; + timeout.tv_nsec = 0; rb_ctx = dict_copy (dict, rb_ctx); @@ -1746,6 +3334,17 @@ glusterd_op_start_rb_timer (dict_t *dict) ret = -1; goto out; } + + ret = dict_set_bin (rb_ctx, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Failed to set transaction id."); + goto out; + } else + gf_log ("", GF_LOG_DEBUG, + "transaction_id = %s", uuid_utoa (*txn_id)); + priv->timer = gf_timer_call_after (THIS->ctx, timeout, glusterd_do_replace_brick, (void *) rb_ctx); @@ -1756,6 +3355,360 @@ out: return ret; } +/* This function takes a dict and converts the uuid values of key specified + * into hostnames + */ +static int +glusterd_op_volume_dict_uuid_to_hostname (dict_t *dict, const char *key_fmt, + int idx_min, int idx_max) +{ + int ret = -1; + int i = 0; + char key[1024]; + char *uuid_str = NULL; + uuid_t uuid = {0,}; + char *hostname = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + GF_ASSERT (dict); + GF_ASSERT (key_fmt); + + for (i = idx_min; i < idx_max; i++) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), key_fmt, i); + ret = dict_get_str (dict, key, &uuid_str); + if (ret) + continue; + + gf_log (this->name, GF_LOG_DEBUG, "Got uuid %s", + uuid_str); + + ret = uuid_parse (uuid_str, uuid); + /* if parsing fails don't error out + * let the original value be retained + */ + if (ret) + continue; + + hostname = glusterd_uuid_to_hostname (uuid); + if (hostname) { + gf_log (this->name, GF_LOG_DEBUG, "%s -> %s", + uuid_str, hostname); + ret = dict_set_dynstr (dict, key, hostname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting hostname %s to dict", + hostname); + GF_FREE (hostname); + goto out; + } + } + } + +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +static int +reassign_defrag_status (dict_t *dict, char *key, gf_defrag_status_t *status) +{ + int ret = 0; + + if (!*status) + return ret; + + switch (*status) { + case GF_DEFRAG_STATUS_STARTED: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED; + break; + + case GF_DEFRAG_STATUS_STOPPED: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_STOPPED; + break; + + case GF_DEFRAG_STATUS_COMPLETE: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE; + break; + + case GF_DEFRAG_STATUS_FAILED: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_FAILED; + break; + default: + break; + } + + ret = dict_set_int32(dict, key, *status); + if (ret) + gf_log (THIS->name, GF_LOG_WARNING, + "failed to reset defrag %s in dict", key); + + return ret; +} + +/* Check and reassign the defrag_status enum got from the rebalance process + * of all peers so that the rebalance-status CLI command can display if a + * full-rebalance or just a fix-layout was carried out. + */ +static int +glusterd_op_check_peer_defrag_status (dict_t *dict, int count) +{ + glusterd_volinfo_t *volinfo = NULL; + gf_defrag_status_t status = GF_DEFRAG_STATUS_NOT_STARTED; + char key[256] = {0,}; + char *volname = NULL; + int ret = -1; + int i = 1; + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, "Unable to get volume name"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, FMTSTR_CHECK_VOL_EXISTS, + volname); + goto out; + } + + if (volinfo->rebal.defrag_cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) { + /* Fix layout was not issued; we don't need to reassign + the status */ + ret = 0; + goto out; + } + + do { + memset (key, 0, 256); + snprintf (key, 256, "status-%d", i); + ret = dict_get_int32 (dict, key, (int32_t *)&status); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, + "failed to get defrag %s", key); + goto out; + } + ret = reassign_defrag_status (dict, key, &status); + if (ret) + goto out; + i++; + } while (i <= count); + + ret = 0; +out: + return ret; + +} + +/* This function is used to modify the op_ctx dict before sending it back + * to cli. This is useful in situations like changing the peer uuids to + * hostnames etc. + */ +void +glusterd_op_modify_op_ctx (glusterd_op_t op, void *ctx) +{ + int ret = -1; + dict_t *op_ctx = NULL; + int brick_index_max = -1; + int other_count = 0; + int count = 0; + uint32_t cmd = GF_CLI_STATUS_NONE; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + if (ctx) + op_ctx = ctx; + else + op_ctx = glusterd_op_get_ctx(); + + if (!op_ctx) { + gf_log (this->name, GF_LOG_CRITICAL, + "Operation context is not present."); + goto out; + } + + switch (op) { + case GD_OP_STATUS_VOLUME: + ret = dict_get_uint32 (op_ctx, "cmd", &cmd); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Failed to get status cmd"); + goto out; + } + if (!(cmd & GF_CLI_STATUS_NFS || cmd & GF_CLI_STATUS_SHD || + (cmd & GF_CLI_STATUS_MASK) == GF_CLI_STATUS_NONE)) { + gf_log (this->name, GF_LOG_DEBUG, + "op_ctx modification not required for status " + "operation being performed"); + goto out; + } + + ret = dict_get_int32 (op_ctx, "brick-index-max", + &brick_index_max); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Failed to get brick-index-max"); + goto out; + } + + ret = dict_get_int32 (op_ctx, "other-count", &other_count); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Failed to get other-count"); + goto out; + } + + count = brick_index_max + other_count + 1; + + ret = glusterd_op_volume_dict_uuid_to_hostname (op_ctx, + "brick%d.path", + 0, count); + if (ret) + gf_log (this->name, GF_LOG_WARNING, + "Failed uuid to hostname conversion"); + + break; + + case GD_OP_PROFILE_VOLUME: + ret = dict_get_str_boolean (op_ctx, "nfs", _gf_false); + if (!ret) + goto out; + + ret = dict_get_int32 (op_ctx, "count", &count); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Failed to get brick count"); + goto out; + } + + ret = glusterd_op_volume_dict_uuid_to_hostname (op_ctx, + "%d-brick", + 1, (count + 1)); + if (ret) + gf_log (this->name, GF_LOG_WARNING, + "Failed uuid to hostname conversion"); + + break; + + /* For both rebalance and remove-brick status, the glusterd op is the + * same + */ + case GD_OP_DEFRAG_BRICK_VOLUME: + ret = dict_get_int32 (op_ctx, "count", &count); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Failed to get count"); + goto out; + } + + /* add 'node-name-%d' into op_ctx with value uuid_str. + this will be used to convert to hostname later */ + { + char key[1024]; + char *uuid_str = NULL; + int i; + + for (i = 1; i <= count; i++) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "node-uuid-%d", i); + ret = dict_get_str (op_ctx, key, &uuid_str); + if (!ret) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), + "node-name-%d", i); + ret = dict_set_str (op_ctx, key, + uuid_str); + } + } + } + + ret = glusterd_op_volume_dict_uuid_to_hostname (op_ctx, + "node-name-%d", + 1, (count + 1)); + if (ret) + gf_log (this->name, GF_LOG_WARNING, + "Failed uuid to hostname conversion"); + + ret = glusterd_op_check_peer_defrag_status (op_ctx, count); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to reset defrag status for fix-layout"); + break; + + default: + ret = 0; + gf_log (this->name, GF_LOG_DEBUG, + "op_ctx modification not required"); + break; + + } + +out: + if (ret) + gf_log (this->name, GF_LOG_WARNING, + "op_ctx modification failed"); + return; +} + +static int +glusterd_op_commit_hook (glusterd_op_t op, dict_t *op_ctx, + glusterd_commit_hook_type_t type) +{ + glusterd_conf_t *priv = NULL; + char hookdir[PATH_MAX] = {0, }; + char scriptdir[PATH_MAX] = {0, }; + char type_subdir[256] = {0, }; + char *cmd_subdir = NULL; + int ret = -1; + + priv = THIS->private; + switch (type) { + case GD_COMMIT_HOOK_NONE: + case GD_COMMIT_HOOK_MAX: + /*Won't be called*/ + break; + + case GD_COMMIT_HOOK_PRE: + strcpy (type_subdir, "pre"); + break; + case GD_COMMIT_HOOK_POST: + strcpy (type_subdir, "post"); + break; + } + + cmd_subdir = glusterd_hooks_get_hooks_cmd_subdir (op); + if (strlen (cmd_subdir) == 0) + return -1; + + GLUSTERD_GET_HOOKS_DIR (hookdir, GLUSTERD_HOOK_VER, priv); + snprintf (scriptdir, sizeof (scriptdir), "%s/%s/%s", + hookdir, cmd_subdir, type_subdir); + + switch (type) { + case GD_COMMIT_HOOK_NONE: + case GD_COMMIT_HOOK_MAX: + /*Won't be called*/ + break; + + case GD_COMMIT_HOOK_PRE: + ret = glusterd_hooks_run_hooks (scriptdir, op, op_ctx, + type); + break; + case GD_COMMIT_HOOK_POST: + ret = glusterd_hooks_post_stub_enqueue (scriptdir, op, + op_ctx); + break; + } + + return ret; +} + static int glusterd_op_ac_send_commit_op (glusterd_op_sm_event_t *event, void *ctx) { @@ -1775,19 +3728,32 @@ glusterd_op_ac_send_commit_op (glusterd_op_sm_event_t *event, void *ctx) priv = this->private; GF_ASSERT (priv); - op = glusterd_op_get_op (); - ret = glusterd_op_build_payload (&dict); + op = glusterd_op_get_op (); + op_dict = glusterd_op_get_ctx (); - if (ret) + ret = glusterd_op_build_payload (&dict, &op_errstr, NULL); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD, + gd_op_list[op]); + if (op_errstr == NULL) + gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD); + opinfo.op_errstr = op_errstr; goto out; + } ret = glusterd_op_commit_perform (op, dict, &op_errstr, NULL); //rsp_dict invalid for source if (ret) { - gf_log ("", GF_LOG_ERROR, "Commit failed"); + gf_log (this->name, GF_LOG_ERROR, LOGSTR_COMMIT_FAIL, + gd_op_list[op], "localhost", (op_errstr) ? ":" : " ", + (op_errstr) ? op_errstr : " "); + if (op_errstr == NULL) + gf_asprintf (&op_errstr, OPERRSTR_COMMIT_FAIL, + "localhost"); opinfo.op_errstr = op_errstr; goto out; } + list_for_each_entry (peerinfo, &priv->peers, uuid_list) { GF_ASSERT (peerinfo); @@ -1802,41 +3768,48 @@ glusterd_op_ac_send_commit_op (glusterd_op_sm_event_t *event, void *ctx) if (proc->fn) { ret = dict_set_static_ptr (dict, "peerinfo", peerinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "failed to set peerinfo"); + gf_log (this->name, GF_LOG_ERROR, + "failed to set peerinfo"); goto out; } ret = proc->fn (NULL, this, dict); - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Failed to " + "send commit request for operation " + "'Volume %s' to peer %s", + gd_op_list[op], peerinfo->hostname); continue; + } pending_count++; } } opinfo.pending_count = pending_count; - gf_log ("glusterd", GF_LOG_INFO, "Sent op req to %d peers", - opinfo.pending_count); + gf_log (this->name, GF_LOG_DEBUG, "Sent commit op req for 'Volume %s' " + "to %d peers", gd_op_list[op], opinfo.pending_count); out: if (dict) dict_unref (dict); if (ret) { - glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); + glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + &event->txn_id, NULL); opinfo.op_ret = ret; } if (!opinfo.pending_count) { if (op == GD_OP_REPLACE_BRICK) { - op_dict = glusterd_op_get_ctx (); - ret = glusterd_op_start_rb_timer (op_dict); + ret = glusterd_op_start_rb_timer (op_dict, + &event->txn_id); } else { - - ret = glusterd_op_sm_inject_all_acc (); + glusterd_op_modify_op_ctx (op, NULL); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); } goto err; } err: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; @@ -1855,10 +3828,11 @@ glusterd_op_ac_rcvd_stage_op_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_ACC, + &event->txn_id, NULL); out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -1876,10 +3850,11 @@ glusterd_op_ac_stage_op_failed (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, NULL); out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -1897,10 +3872,11 @@ glusterd_op_ac_commit_op_failed (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, NULL); out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -1911,6 +3887,10 @@ glusterd_op_ac_brick_op_failed (glusterd_op_sm_event_t *event, void *ctx) int ret = 0; glusterd_op_brick_rsp_ctx_t *ev_ctx = NULL; gf_boolean_t free_errstr = _gf_false; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (event); GF_ASSERT (ctx); @@ -1918,7 +3898,7 @@ glusterd_op_ac_brick_op_failed (glusterd_op_sm_event_t *event, void *ctx) ret = glusterd_remove_pending_entry (&opinfo.pending_bricks, ev_ctx->pending_node->node); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "unknown response received "); + gf_log (this->name, GF_LOG_ERROR, "unknown response received "); ret = -1; free_errstr = _gf_true; goto out; @@ -1936,7 +3916,8 @@ glusterd_op_ac_brick_op_failed (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.brick_pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, ev_ctx->commit_ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, ev_ctx->commit_ctx); out: if (ev_ctx->rsp_dict) @@ -1944,7 +3925,7 @@ out: if (free_errstr && ev_ctx->op_errstr) GF_FREE (ev_ctx->op_errstr); GF_FREE (ctx); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -1952,11 +3933,14 @@ out: static int glusterd_op_ac_rcvd_commit_op_acc (glusterd_op_sm_event_t *event, void *ctx) { - dict_t *dict = NULL; + dict_t *op_ctx = NULL; int ret = 0; gf_boolean_t commit_ack_inject = _gf_true; glusterd_op_t op = GD_OP_NONE; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); op = glusterd_op_get_op (); GF_ASSERT (event); @@ -1967,17 +3951,17 @@ glusterd_op_ac_rcvd_commit_op_acc (glusterd_op_sm_event_t *event, void *ctx) goto out; if (op == GD_OP_REPLACE_BRICK) { - dict = glusterd_op_get_ctx (); - if (!dict) { - gf_log (THIS->name, GF_LOG_CRITICAL, "Operation " + op_ctx = glusterd_op_get_ctx (); + if (!op_ctx) { + gf_log (this->name, GF_LOG_CRITICAL, "Operation " "context is not present."); ret = -1; goto out; } - ret = glusterd_op_start_rb_timer (dict); + ret = glusterd_op_start_rb_timer (op_ctx, &event->txn_id); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Couldn't start " + gf_log (this->name, GF_LOG_ERROR, "Couldn't start " "replace-brick operation."); goto out; } @@ -1986,13 +3970,17 @@ glusterd_op_ac_rcvd_commit_op_acc (glusterd_op_sm_event_t *event, void *ctx) goto out; } + out: if (commit_ack_inject) { if (ret) - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); - else if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, NULL); - + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + &event->txn_id, NULL); + else if (!opinfo.pending_count) { + glusterd_op_modify_op_ctx (op, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, + &event->txn_id, NULL); + } /*else do nothing*/ } @@ -2012,9 +4000,10 @@ glusterd_op_ac_rcvd_unlock_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, + &event->txn_id, NULL); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); out: return ret; @@ -2046,7 +4035,7 @@ glusterd_op_reset_ctx () } int32_t -glusterd_op_txn_complete () +glusterd_op_txn_complete (uuid_t *txn_id) { int32_t ret = -1; glusterd_conf_t *priv = NULL; @@ -2056,9 +4045,13 @@ glusterd_op_txn_complete () rpcsvc_request_t *req = NULL; void *ctx = NULL; char *op_errstr = NULL; + char *volname = NULL; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); - priv = THIS->private; + priv = this->private; GF_ASSERT (priv); op = glusterd_op_get_op (); @@ -2075,32 +4068,55 @@ glusterd_op_txn_complete () glusterd_op_reset_ctx (); glusterd_op_clear_errstr (); - ret = glusterd_unlock (priv->uuid); - - /* unlock cant/shouldnt fail here!! */ - if (ret) { - gf_log ("glusterd", GF_LOG_CRITICAL, - "Unable to clear local lock, ret: %d", ret); + /* Based on the op-version, we release the cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + ret = glusterd_unlock (MY_UUID); + /* unlock cant/shouldnt fail here!! */ + if (ret) + gf_log (this->name, GF_LOG_CRITICAL, + "Unable to clear local lock, ret: %d", ret); + else + gf_log (this->name, GF_LOG_DEBUG, "Cleared local lock"); } else { - gf_log ("glusterd", GF_LOG_INFO, "Cleared local lock"); + ret = dict_get_str (ctx, "volname", &volname); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire volname"); + + if (volname) { + ret = glusterd_mgmt_v3_unlock (volname, MY_UUID, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to release lock for %s", + volname); + } } ret = glusterd_op_send_cli_response (op, op_ret, op_errno, req, ctx, op_errstr); if (ret) { - gf_log ("", GF_LOG_ERROR, "Responding to cli failed, ret: %d", - ret); + gf_log (this->name, GF_LOG_ERROR, "Responding to cli failed, " + "ret: %d", ret); //Ignore this error, else state machine blocks ret = 0; } - glusterd_op_free_ctx (op, ctx); if (op_errstr && (strcmp (op_errstr, ""))) GF_FREE (op_errstr); - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + if (priv->pending_quorum_action) + glusterd_do_quorum_action (); + + /* Clearing the transaction opinfo */ + ret = glusterd_clear_txn_opinfo (txn_id); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to clear transaction's opinfo"); + + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -2111,9 +4127,9 @@ glusterd_op_ac_unlocked_all (glusterd_op_sm_event_t *event, void *ctx) GF_ASSERT (event); - ret = glusterd_op_txn_complete (); + ret = glusterd_op_txn_complete (&event->txn_id); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -2127,7 +4143,11 @@ glusterd_op_ac_stage_op (glusterd_op_sm_event_t *event, void *ctx) dict_t *rsp_dict = NULL; char *op_errstr = NULL; dict_t *dict = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (ctx); req_ctx = ctx; @@ -2136,8 +4156,8 @@ glusterd_op_ac_stage_op (glusterd_op_sm_event_t *event, void *ctx) rsp_dict = dict_new (); if (!rsp_dict) { - gf_log ("", GF_LOG_DEBUG, - "Out of memory"); + gf_log (this->name, GF_LOG_ERROR, + "Failed to get new dictionary"); return -1; } @@ -2145,16 +4165,31 @@ glusterd_op_ac_stage_op (glusterd_op_sm_event_t *event, void *ctx) rsp_dict); if (status) { - gf_log ("", GF_LOG_ERROR, "Validate failed: %d", status); + gf_log (this->name, GF_LOG_ERROR, "Stage failed on operation" + " 'Volume %s', Status : %d", gd_op_list[req_ctx->op], + status); } + txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + + if (txn_id) + uuid_copy (*txn_id, event->txn_id); + else + gf_log (this->name, GF_LOG_ERROR, "Out of Memory"); + + ret = dict_set_bin (rsp_dict, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to set transaction id."); + ret = glusterd_op_stage_send_resp (req_ctx->req, req_ctx->op, status, op_errstr, rsp_dict); if (op_errstr && (strcmp (op_errstr, ""))) GF_FREE (op_errstr); - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); if (rsp_dict) dict_unref (rsp_dict); @@ -2172,6 +4207,8 @@ glusterd_need_brick_op (glusterd_op_t op) switch (op) { case GD_OP_PROFILE_VOLUME: case GD_OP_STATUS_VOLUME: + case GD_OP_DEFRAG_BRICK_VOLUME: + case GD_OP_HEAL_VOLUME: ret = _gf_true; break; default: @@ -2181,7 +4218,7 @@ glusterd_need_brick_op (glusterd_op_t op) return ret; } -static dict_t* +dict_t* glusterd_op_init_commit_rsp_dict (glusterd_op_t op) { dict_t *rsp_dict = NULL; @@ -2204,12 +4241,16 @@ static int glusterd_op_ac_commit_op (glusterd_op_sm_event_t *event, void *ctx) { int ret = 0; - glusterd_req_ctx_t *req_ctx = NULL; + glusterd_req_ctx_t *req_ctx = NULL; int32_t status = 0; char *op_errstr = NULL; dict_t *dict = NULL; - dict_t *rsp_dict = NULL; + dict_t *rsp_dict = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (ctx); req_ctx = ctx; @@ -2219,24 +4260,46 @@ glusterd_op_ac_commit_op (glusterd_op_sm_event_t *event, void *ctx) rsp_dict = glusterd_op_init_commit_rsp_dict (req_ctx->op); if (NULL == rsp_dict) return -1; - status = glusterd_op_commit_perform (req_ctx->op, dict, &op_errstr, - rsp_dict); - if (status) { - gf_log (THIS->name, GF_LOG_ERROR, "Commit failed: %d", status); + + if (GD_OP_CLEARLOCKS_VOLUME == req_ctx->op) { + /*clear locks should be run only on + * originator glusterd*/ + status = 0; + + } else { + status = glusterd_op_commit_perform (req_ctx->op, dict, + &op_errstr, rsp_dict); } + if (status) + gf_log (this->name, GF_LOG_ERROR, "Commit of operation " + "'Volume %s' failed: %d", gd_op_list[req_ctx->op], + status); + + txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + + if (txn_id) + uuid_copy (*txn_id, event->txn_id); + else + gf_log (this->name, GF_LOG_ERROR, "Out of Memory"); + + ret = dict_set_bin (rsp_dict, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to set transaction id."); + ret = glusterd_op_commit_send_resp (req_ctx->req, req_ctx->op, status, op_errstr, rsp_dict); - glusterd_op_fini_ctx (); if (op_errstr && (strcmp (op_errstr, ""))) GF_FREE (op_errstr); if (rsp_dict) dict_unref (rsp_dict); - gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } @@ -2258,13 +4321,12 @@ glusterd_op_ac_send_commit_failed (glusterd_op_sm_event_t *event, void *ctx) opinfo.op_ret, opinfo.op_errstr, op_ctx); - glusterd_op_fini_ctx (); if (opinfo.op_errstr && (strcmp (opinfo.op_errstr, ""))) { GF_FREE (opinfo.op_errstr); opinfo.op_errstr = NULL; } - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } @@ -2295,6 +4357,7 @@ glusterd_op_stage_validate (glusterd_op_t op, dict_t *dict, char **op_errstr, dict_t *rsp_dict) { int ret = -1; + xlator_t *this = THIS; switch (op) { case GD_OP_CREATE_VOLUME: @@ -2342,6 +4405,10 @@ glusterd_op_stage_validate (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_op_stage_sync_volume (dict, op_errstr); break; + case GD_OP_GSYNC_CREATE: + ret = glusterd_op_stage_gsync_create (dict, op_errstr); + break; + case GD_OP_GSYNC_SET: ret = glusterd_op_stage_gsync_set (dict, op_errstr); break; @@ -2359,6 +4426,7 @@ glusterd_op_stage_validate (glusterd_op_t op, dict_t *dict, char **op_errstr, break; case GD_OP_REBALANCE: + case GD_OP_DEFRAG_BRICK_VOLUME: ret = glusterd_op_stage_rebalance (dict, op_errstr); break; @@ -2370,14 +4438,25 @@ glusterd_op_stage_validate (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_op_stage_statedump_volume (dict, op_errstr); break; + case GD_OP_CLEARLOCKS_VOLUME: + ret = glusterd_op_stage_clearlocks_volume (dict, + op_errstr); + break; + + case GD_OP_COPY_FILE: + ret = glusterd_op_stage_copy_file (dict, op_errstr); + break; + + case GD_OP_SYS_EXEC: + ret = glusterd_op_stage_sys_exec (dict, op_errstr); + break; default: - gf_log ("", GF_LOG_ERROR, "Unknown op %d", - op); + gf_log (this->name, GF_LOG_ERROR, "Unknown op %s", + gd_op_list[op]); } - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - + gf_log (this->name, GF_LOG_DEBUG, "OP = %d. Returning %d", op, ret); return ret; } @@ -2387,7 +4466,9 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, dict_t *rsp_dict) { int ret = -1; + xlator_t *this = THIS; + glusterd_op_commit_hook (op, dict, GD_COMMIT_HOOK_PRE); switch (op) { case GD_OP_CREATE_VOLUME: ret = glusterd_op_create_volume (dict, op_errstr); @@ -2418,7 +4499,7 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, break; case GD_OP_RESET_VOLUME: - ret = glusterd_op_reset_volume (dict); + ret = glusterd_op_reset_volume (dict, op_errstr); break; case GD_OP_REMOVE_BRICK: @@ -2433,6 +4514,11 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_op_sync_volume (dict, op_errstr, rsp_dict); break; + case GD_OP_GSYNC_CREATE: + ret = glusterd_op_gsync_create (dict, op_errstr, + rsp_dict); + break; + case GD_OP_GSYNC_SET: ret = glusterd_op_gsync_set (dict, op_errstr, rsp_dict); break; @@ -2443,7 +4529,7 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, break; case GD_OP_QUOTA: - ret = glusterd_op_quota (dict, op_errstr); + ret = glusterd_op_quota (dict, op_errstr, rsp_dict); break; case GD_OP_STATUS_VOLUME: @@ -2451,6 +4537,7 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, break; case GD_OP_REBALANCE: + case GD_OP_DEFRAG_BRICK_VOLUME: ret = glusterd_op_rebalance (dict, op_errstr, rsp_dict); break; @@ -2462,150 +4549,36 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_op_statedump_volume (dict, op_errstr); break; - default: - gf_log ("", GF_LOG_ERROR, "Unknown op %d", - op); + case GD_OP_CLEARLOCKS_VOLUME: + ret = glusterd_op_clearlocks_volume (dict, op_errstr, + rsp_dict); break; - } - - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - - return ret; -} - -void -_profile_volume_add_brick_rsp (dict_t *this, char *key, data_t *value, - void *data) -{ - char new_key[256] = {0}; - glusterd_pr_brick_rsp_conv_t *rsp_ctx = NULL; - data_t *new_value = NULL; - - rsp_ctx = data; - new_value = data_copy (value); - GF_ASSERT (new_value); - snprintf (new_key, sizeof (new_key), "%d-%s", rsp_ctx->count, key); - dict_set (rsp_ctx->dict, new_key, new_value); -} - -int -glusterd_profile_volume_brick_rsp (glusterd_brickinfo_t *brickinfo, - dict_t *rsp_dict, dict_t *op_ctx, - char **op_errstr) -{ - int ret = 0; - glusterd_pr_brick_rsp_conv_t rsp_ctx = {0}; - int32_t count = 0; - char brick[PATH_MAX+1024] = {0}; - char key[256] = {0}; - char *full_brick = NULL; - - GF_ASSERT (rsp_dict); - GF_ASSERT (op_ctx); - GF_ASSERT (op_errstr); - GF_ASSERT (brickinfo); - - ret = dict_get_int32 (op_ctx, "count", &count); - if (ret) { - count = 1; - } else { - count++; - } - snprintf (key, sizeof (key), "%d-brick", count); - snprintf (brick, sizeof (brick), "%s:%s", brickinfo->hostname, - brickinfo->path); - full_brick = gf_strdup (brick); - GF_ASSERT (full_brick); - ret = dict_set_dynstr (op_ctx, key, full_brick); - rsp_ctx.count = count; - rsp_ctx.dict = op_ctx; - dict_foreach (rsp_dict, _profile_volume_add_brick_rsp, &rsp_ctx); - dict_del (op_ctx, "count"); - ret = dict_set_int32 (op_ctx, "count", count); - return ret; -} - -void -_status_volume_add_brick_rsp (dict_t *this, char *key, data_t *value, - void *data) -{ - char new_key[256] = {0,}; - data_t *new_value = 0; - glusterd_pr_brick_rsp_conv_t *rsp_ctx = NULL; - - rsp_ctx = data; - new_value = data_copy (value); - snprintf (new_key, sizeof (new_key), "brick%d.%s", rsp_ctx->count, key); - dict_set (rsp_ctx->dict, new_key, new_value); - - return; -} - -int -glusterd_status_volume_brick_rsp (glusterd_brickinfo_t *brickinfo, - dict_t *rsp_dict, dict_t *op_ctx, - char **op_errstr) -{ - int ret = 0; - glusterd_pr_brick_rsp_conv_t rsp_ctx = {0}; - int32_t count = 0; - int index = 0; + case GD_OP_COPY_FILE: + ret = glusterd_op_copy_file (dict, op_errstr); + break; - GF_ASSERT (rsp_dict); - GF_ASSERT (op_ctx); - GF_ASSERT (op_errstr); - GF_ASSERT (brickinfo); + case GD_OP_SYS_EXEC: + ret = glusterd_op_sys_exec (dict, op_errstr, rsp_dict); + break; - ret = dict_get_int32 (op_ctx, "count", &count); - if (ret) { - count = 0; - } else { - count++; + default: + gf_log (this->name, GF_LOG_ERROR, "Unknown op %s", + gd_op_list[op]); + break; } - ret = dict_get_int32 (rsp_dict, "index", &index); - if (ret) - goto out; - dict_del (rsp_dict, "index"); - rsp_ctx.count = index; - rsp_ctx.dict = op_ctx; - dict_foreach (rsp_dict, _status_volume_add_brick_rsp, &rsp_ctx); - ret = dict_set_int32 (op_ctx, "count", count); + if (ret == 0) + glusterd_op_commit_hook (op, dict, GD_COMMIT_HOOK_POST); -out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } -int32_t -glusterd_handle_brick_rsp (glusterd_brickinfo_t *brickinfo, - glusterd_op_t op, dict_t *rsp_dict, dict_t *op_ctx, - char **op_errstr) -{ - int ret = 0; - - GF_ASSERT (op_errstr); - - switch (op) { - case GD_OP_PROFILE_VOLUME: - ret = glusterd_profile_volume_brick_rsp (brickinfo, rsp_dict, - op_ctx, op_errstr); - break; - case GD_OP_STATUS_VOLUME: - ret = glusterd_status_volume_brick_rsp (brickinfo, rsp_dict, - op_ctx, op_errstr); - break; - - default: - break; - } - - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - return ret; -} static int -glusterd_bricks_select_stop_volume (dict_t *dict, char **op_errstr) +glusterd_bricks_select_stop_volume (dict_t *dict, char **op_errstr, + struct list_head *selected) { int ret = 0; int flags = 0; @@ -2614,15 +4587,17 @@ glusterd_bricks_select_stop_volume (dict_t *dict, char **op_errstr) glusterd_brickinfo_t *brickinfo = NULL; glusterd_pending_node_t *pending_node = NULL; - ret = glusterd_op_stop_volume_args_get (dict, &volname, &flags); if (ret) goto out; ret = glusterd_volinfo_find (volname, &volinfo); - - if (ret) + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS, + volname); + gf_asprintf (op_errstr, FMTSTR_CHECK_VOL_EXISTS, volname); goto out; + } list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { if (glusterd_is_brick_started (brickinfo)) { @@ -2634,7 +4609,7 @@ glusterd_bricks_select_stop_volume (dict_t *dict, char **op_errstr) } else { pending_node->node = brickinfo; pending_node->type = GD_NODE_BRICK; - list_add_tail (&pending_node->list, &opinfo.pending_bricks); + list_add_tail (&pending_node->list, selected); pending_node = NULL; } } @@ -2645,7 +4620,8 @@ out: } static int -glusterd_bricks_select_remove_brick (dict_t *dict, char **op_errstr) +glusterd_bricks_select_remove_brick (dict_t *dict, char **op_errstr, + struct list_head *selected) { int ret = -1; char *volname = NULL; @@ -2658,6 +4634,8 @@ glusterd_bricks_select_remove_brick (dict_t *dict, char **op_errstr) glusterd_pending_node_t *pending_node = NULL; int32_t force = 0; + + ret = dict_get_str (dict, "volname", &volname); if (ret) { @@ -2694,8 +4672,7 @@ glusterd_bricks_select_remove_brick (dict_t *dict, char **op_errstr) } ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, - &brickinfo, - GF_PATH_COMPLETE); + &brickinfo); if (ret) goto out; if (glusterd_is_brick_started (brickinfo)) { @@ -2707,7 +4684,7 @@ glusterd_bricks_select_remove_brick (dict_t *dict, char **op_errstr) } else { pending_node->node = brickinfo; pending_node->type = GD_NODE_BRICK; - list_add_tail (&pending_node->list, &opinfo.pending_bricks); + list_add_tail (&pending_node->list, selected); pending_node = NULL; } } @@ -2719,7 +4696,8 @@ out: } static int -glusterd_bricks_select_profile_volume (dict_t *dict, char **op_errstr) +glusterd_bricks_select_profile_volume (dict_t *dict, char **op_errstr, + struct list_head *selected) { int ret = -1; char *volname = NULL; @@ -2732,11 +4710,14 @@ glusterd_bricks_select_profile_volume (dict_t *dict, char **op_errstr) glusterd_pending_node_t *pending_node = NULL; char *brick = NULL; + + this = THIS; GF_ASSERT (this); priv = this->private; GF_ASSERT (priv); + ret = dict_get_str (dict, "volname", &volname); if (ret) { gf_log ("glusterd", GF_LOG_ERROR, "volume name get failed"); @@ -2765,6 +4746,29 @@ glusterd_bricks_select_profile_volume (dict_t *dict, char **op_errstr) goto out; break; case GF_CLI_STATS_INFO: + ret = dict_get_str_boolean (dict, "nfs", _gf_false); + if (ret) { + if (!glusterd_is_nodesvc_online ("nfs")) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "NFS server" + " is not running"); + goto out; + } + pending_node = GF_CALLOC (1, sizeof (*pending_node), + gf_gld_mt_pending_node_t); + if (!pending_node) { + ret = -1; + goto out; + } + pending_node->node = priv->nfs; + pending_node->type = GD_NODE_NFS; + list_add_tail (&pending_node->list, selected); + pending_node = NULL; + + ret = 0; + goto out; + + } list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { if (glusterd_is_brick_started (brickinfo)) { pending_node = GF_CALLOC (1, sizeof (*pending_node), @@ -2776,7 +4780,7 @@ glusterd_bricks_select_profile_volume (dict_t *dict, char **op_errstr) pending_node->node = brickinfo; pending_node->type = GD_NODE_BRICK; list_add_tail (&pending_node->list, - &opinfo.pending_bricks); + selected); pending_node = NULL; } } @@ -2784,11 +4788,33 @@ glusterd_bricks_select_profile_volume (dict_t *dict, char **op_errstr) break; case GF_CLI_STATS_TOP: + ret = dict_get_str_boolean (dict, "nfs", _gf_false); + if (ret) { + if (!glusterd_is_nodesvc_online ("nfs")) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "NFS server" + " is not running"); + goto out; + } + pending_node = GF_CALLOC (1, sizeof (*pending_node), + gf_gld_mt_pending_node_t); + if (!pending_node) { + ret = -1; + goto out; + } + pending_node->node = priv->nfs; + pending_node->type = GD_NODE_NFS; + list_add_tail (&pending_node->list, selected); + pending_node = NULL; + + ret = 0; + goto out; + + } ret = dict_get_str (dict, "brick", &brick); if (!ret) { ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, - &brickinfo, - GF_PATH_COMPLETE); + &brickinfo); if (ret) goto out; @@ -2804,7 +4830,7 @@ glusterd_bricks_select_profile_volume (dict_t *dict, char **op_errstr) pending_node->node = brickinfo; pending_node->type = GD_NODE_BRICK; list_add_tail (&pending_node->list, - &opinfo.pending_bricks); + selected); pending_node = NULL; goto out; } @@ -2821,7 +4847,7 @@ glusterd_bricks_select_profile_volume (dict_t *dict, char **op_errstr) pending_node->node = brickinfo; pending_node->type = GD_NODE_BRICK; list_add_tail (&pending_node->list, - &opinfo.pending_bricks); + selected); pending_node = NULL; } } @@ -2851,31 +4877,305 @@ _add_rxlator_to_dict (dict_t *dict, char *volname, int index, int count) char key[128] = {0,}; char *xname = NULL; - snprintf (key, sizeof (key), "heal-%d", count); + snprintf (key, sizeof (key), "xl-%d", count); ret = gf_asprintf (&xname, "%s-replicate-%d", volname, index); if (ret == -1) goto out; ret = dict_set_dynstr (dict, key, xname); + if (ret) + goto out; + + ret = dict_set_int32 (dict, xname, index); +out: + return ret; +} + +int +get_replica_index_for_per_replica_cmd (glusterd_volinfo_t *volinfo, + dict_t *dict) { + int ret = 0; + char *hostname = NULL; + char *path = NULL; + int index = 0; + glusterd_brickinfo_t *brickinfo = NULL; + int cmd_replica_index = -1; + int replica_count = -1; + + + if (!dict) { + ret = -1; + goto out; + } + + ret = dict_get_str (dict, "per-replica-cmd-hostname", &hostname); + if (ret) + goto out; + ret = dict_get_str (dict, "per-replica-cmd-path", &path); + if (ret) + goto out; + + replica_count = volinfo->replica_count; + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (uuid_is_null (brickinfo->uuid)) + (void)glusterd_resolve_brick (brickinfo); + if (!strcmp (brickinfo->path, path) && + !strcmp (brickinfo->hostname, hostname)) { + cmd_replica_index = index/(replica_count); + goto out; + } + index++; + } + + +out: + if (ret) + cmd_replica_index = -1; + + return cmd_replica_index; +} + +int +_select_rxlators_with_local_bricks (xlator_t *this, glusterd_volinfo_t *volinfo, + dict_t *dict, cli_cmd_type type) +{ + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_conf_t *priv = NULL; + int index = 0; + int rxlator_count = 0; + int replica_count = 0; + gf_boolean_t add = _gf_false; + int ret = 0; + int cmd_replica_index = -1; + + priv = this->private; + replica_count = volinfo->replica_count; + + if (type == PER_REPLICA) { + + cmd_replica_index = get_replica_index_for_per_replica_cmd + (volinfo, dict); + if (cmd_replica_index == -1) { + ret = -1; + goto err; + } + } + + index = 1; + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (uuid_is_null (brickinfo->uuid)) + (void)glusterd_resolve_brick (brickinfo); + + switch (type) { + case ALL_REPLICA: + if (!uuid_compare (MY_UUID, brickinfo->uuid)) + add = _gf_true; + break; + case PER_REPLICA: + if (!uuid_compare (MY_UUID, brickinfo->uuid) && + ((index-1)/replica_count == cmd_replica_index)) + + add = _gf_true; + break; + } + + if (index % replica_count == 0) { + if (add) { + _add_rxlator_to_dict (dict, volinfo->volname, + (index-1)/replica_count, + rxlator_count); + rxlator_count++; + } + add = _gf_false; + } + + index++; + } +err: + if (ret) + rxlator_count = -1; + + return rxlator_count; +} + +int +_select_rxlators_for_full_self_heal (xlator_t *this, + glusterd_volinfo_t *volinfo, + dict_t *dict) +{ + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_conf_t *priv = NULL; + int index = 1; + int rxlator_count = 0; + int replica_count = 0; + uuid_t candidate = {0}; + + priv = this->private; + replica_count = volinfo->replica_count; + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (uuid_is_null (brickinfo->uuid)) + (void)glusterd_resolve_brick (brickinfo); + + if (uuid_compare (brickinfo->uuid, candidate) > 0) + uuid_copy (candidate, brickinfo->uuid); + + if (index % replica_count == 0) { + if (!uuid_compare (MY_UUID, candidate)) { + _add_rxlator_to_dict (dict, volinfo->volname, + (index-1)/replica_count, + rxlator_count); + rxlator_count++; + } + uuid_clear (candidate); + } + + index++; + } + return rxlator_count; +} + + +static int +glusterd_bricks_select_snap (dict_t *dict, char **op_errstr, + struct list_head *selected) +{ + int ret = -1; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + glusterd_pending_node_t *pending_node = NULL; + glusterd_volinfo_t *volinfo = NULL; + char *volname = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + int brick_index = -1; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to get" + " volname"); + goto out; + } + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) + goto out; + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + brick_index++; + if (uuid_compare (brickinfo->uuid, MY_UUID) || + !glusterd_is_brick_started (brickinfo)) { + continue; + } + pending_node = GF_CALLOC (1, sizeof (*pending_node), + gf_gld_mt_pending_node_t); + if (!pending_node) { + ret = -1; + goto out; + } + pending_node->node = brickinfo; + pending_node->type = GD_NODE_BRICK; + pending_node->index = brick_index; + list_add_tail (&pending_node->list, + selected); + pending_node = NULL; + } + + ret = 0; + out: + gf_log (THIS->name, GF_LOG_DEBUG, "Returning ret %d", ret); return ret; } static int -glusterd_bricks_select_heal_volume (dict_t *dict, char **op_errstr) +fill_shd_status_for_local_bricks (dict_t *dict, glusterd_volinfo_t *volinfo, + cli_cmd_type type, dict_t *req_dict) +{ + glusterd_brickinfo_t *brickinfo = NULL; + char msg[1024] = {0,}; + char key[1024] = {0,}; + char value[1024] = {0,}; + int index = 0; + int ret = 0; + xlator_t *this = NULL; + int cmd_replica_index = -1; + + this = THIS; + snprintf (msg, sizeof (msg), "self-heal-daemon is not running on"); + + if (type == PER_REPLICA) { + cmd_replica_index = get_replica_index_for_per_replica_cmd + (volinfo, req_dict); + if (cmd_replica_index == -1) { + gf_log (THIS->name, GF_LOG_ERROR, "Could not find the " + "replica index for per replica type command"); + ret = -1; + goto out; + } + } + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (uuid_is_null (brickinfo->uuid)) + (void)glusterd_resolve_brick (brickinfo); + + if (uuid_compare (MY_UUID, brickinfo->uuid)) { + index++; + continue; + } + + if (type == PER_REPLICA) { + if (cmd_replica_index != (index/volinfo->replica_count)) { + index++; + continue; + } + + } + snprintf (key, sizeof (key), "%d-status",index); + snprintf (value, sizeof (value), "%s %s",msg, + uuid_utoa(MY_UUID)); + ret = dict_set_dynstr (dict, key, gf_strdup(value)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to" + "set the dictionary for shd status msg"); + goto out; + } + snprintf (key, sizeof (key), "%d-shd-status",index); + ret = dict_set_str (dict, key, "off"); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to" + " set dictionary for shd status msg"); + goto out; + } + + index++; + } + +out: + return ret; + +} + + +static int +glusterd_bricks_select_heal_volume (dict_t *dict, char **op_errstr, + struct list_head *selected, + dict_t *rsp_dict) { int ret = -1; char *volname = NULL; glusterd_conf_t *priv = NULL; glusterd_volinfo_t *volinfo = NULL; - glusterd_brickinfo_t *brickinfo = NULL; xlator_t *this = NULL; char msg[2048] = {0,}; - int replica_count = 0; - int index = 1; - int rxlator_count = 0; - uuid_t candidate = {0}; glusterd_pending_node_t *pending_node = NULL; + gf_xl_afr_op_t heal_op = GF_AFR_OP_INVALID; + int rxlator_count = 0; this = THIS; GF_ASSERT (this); @@ -2898,48 +5198,102 @@ glusterd_bricks_select_heal_volume (dict_t *dict, char **op_errstr) goto out; } - replica_count = volinfo->replica_count; + ret = dict_get_int32 (dict, "heal-op", (int32_t*)&heal_op); + if (ret || (heal_op == GF_AFR_OP_INVALID)) { + gf_log ("glusterd", GF_LOG_ERROR, "heal op invalid"); + goto out; + } - index = 1; - list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { - if (uuid_is_null (brickinfo->uuid)) - (void)glusterd_resolve_brick (brickinfo); + switch (heal_op) { + case GF_AFR_OP_INDEX_SUMMARY: + case GF_AFR_OP_STATISTICS_HEAL_COUNT: + if (!glusterd_is_nodesvc_online ("glustershd")) { + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, "Received " + "empty ctx."); + goto out; + } - if (uuid_compare (brickinfo->uuid, candidate) > 0) - uuid_copy (candidate, brickinfo->uuid); + ret = fill_shd_status_for_local_bricks (rsp_dict, + volinfo, + ALL_REPLICA, + dict); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "fill the shd status for the local " + "bricks"); + goto out; - if (index % replica_count == 0) { - if (!uuid_compare (priv->uuid, candidate)) { - _add_rxlator_to_dict (dict, volname, - (index-1)/replica_count, - rxlator_count); - rxlator_count++; + } + break; + case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA: + if (!glusterd_is_nodesvc_online ("glustershd")) { + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, "Received " + "empty ctx."); + goto out; } - uuid_clear (candidate); + ret = fill_shd_status_for_local_bricks (rsp_dict, + volinfo, + PER_REPLICA, + dict); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "fill the shd status for the local" + " bricks."); + goto out; + } + break; + default: + break; + } - index++; + + switch (heal_op) { + case GF_AFR_OP_HEAL_FULL: + rxlator_count = _select_rxlators_for_full_self_heal (this, + volinfo, + dict); + break; + case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA: + rxlator_count = _select_rxlators_with_local_bricks (this, + volinfo, + dict, + PER_REPLICA); + break; + default: + rxlator_count = _select_rxlators_with_local_bricks (this, + volinfo, + dict, + ALL_REPLICA); + break; } + if (!rxlator_count) + goto out; + if (rxlator_count == -1){ + gf_log (this->name, GF_LOG_ERROR, "Could not determine the" + "translator count"); + ret = -1; + goto out; + } + ret = dict_set_int32 (dict, "count", rxlator_count); if (ret) goto out; - if (rxlator_count) { - pending_node = GF_CALLOC (1, sizeof (*pending_node), - gf_gld_mt_pending_node_t); - if (!pending_node) { - ret = -1; - goto out; - } else { - pending_node->node = priv->shd; - pending_node->type = GD_NODE_SHD; - list_add_tail (&pending_node->list, - &opinfo.pending_bricks); - pending_node = NULL; - } + pending_node = GF_CALLOC (1, sizeof (*pending_node), + gf_gld_mt_pending_node_t); + if (!pending_node) { + ret = -1; + goto out; + } else { + pending_node->node = priv->shd; + pending_node->type = GD_NODE_SHD; + list_add_tail (&pending_node->list, selected); + pending_node = NULL; } - out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning ret %d", ret); return ret; @@ -2947,7 +5301,57 @@ out: } static int -glusterd_bricks_select_status_volume (dict_t *dict, char **op_errstr) +glusterd_bricks_select_rebalance_volume (dict_t *dict, char **op_errstr, + struct list_head *selected) +{ + int ret = -1; + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + xlator_t *this = NULL; + char msg[2048] = {0,}; + glusterd_pending_node_t *pending_node = NULL; + + this = THIS; + GF_ASSERT (this); + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, "volume name get failed"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (msg, sizeof (msg), "Volume %s does not exist", + volname); + + *op_errstr = gf_strdup (msg); + gf_log ("", GF_LOG_ERROR, "%s", msg); + goto out; + } + pending_node = GF_CALLOC (1, sizeof (*pending_node), + gf_gld_mt_pending_node_t); + if (!pending_node) { + ret = -1; + goto out; + } else { + pending_node->node = volinfo; + pending_node->type = GD_NODE_REBALANCE; + list_add_tail (&pending_node->list, + &opinfo.pending_bricks); + pending_node = NULL; + } + +out: + return ret; +} + + + + +static int +glusterd_bricks_select_status_volume (dict_t *dict, char **op_errstr, + struct list_head *selected) { int ret = -1; int cmd = 0; @@ -2969,7 +5373,7 @@ glusterd_bricks_select_status_volume (dict_t *dict, char **op_errstr) ret = dict_get_int32 (dict, "cmd", &cmd); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Unable to get status type"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get status type"); goto out; } @@ -2982,13 +5386,15 @@ glusterd_bricks_select_status_volume (dict_t *dict, char **op_errstr) case GF_CLI_STATUS_INODE: case GF_CLI_STATUS_FD: case GF_CLI_STATUS_CALLPOOL: + case GF_CLI_STATUS_NFS: + case GF_CLI_STATUS_SHD: break; default: goto out; } ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Unable to get volname"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volname"); goto out; } ret = glusterd_volinfo_find (volname, &volinfo); @@ -2999,18 +5405,17 @@ glusterd_bricks_select_status_volume (dict_t *dict, char **op_errstr) if ( (cmd & GF_CLI_STATUS_BRICK) != 0) { ret = dict_get_str (dict, "brick", &brickname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Unable to get brick"); goto out; } ret = glusterd_volume_brickinfo_get_by_brick (brickname, volinfo, - &brickinfo, - GF_PATH_COMPLETE); + &brickinfo); if (ret) goto out; - if (uuid_compare (brickinfo->uuid, priv->uuid)|| + if (uuid_compare (brickinfo->uuid, MY_UUID)|| !glusterd_is_brick_started (brickinfo)) goto out; @@ -3023,13 +5428,51 @@ glusterd_bricks_select_status_volume (dict_t *dict, char **op_errstr) pending_node->node = brickinfo; pending_node->type = GD_NODE_BRICK; pending_node->index = 0; - list_add_tail (&pending_node->list, &opinfo.pending_bricks); + list_add_tail (&pending_node->list, selected); + + ret = 0; + } else if ((cmd & GF_CLI_STATUS_NFS) != 0) { + if (!glusterd_is_nodesvc_online ("nfs")) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "NFS server is not running"); + goto out; + } + pending_node = GF_CALLOC (1, sizeof (*pending_node), + gf_gld_mt_pending_node_t); + if (!pending_node) { + ret = -1; + goto out; + } + pending_node->node = priv->nfs; + pending_node->type = GD_NODE_NFS; + pending_node->index = 0; + list_add_tail (&pending_node->list, selected); + + ret = 0; + } else if ((cmd & GF_CLI_STATUS_SHD) != 0) { + if (!glusterd_is_nodesvc_online ("glustershd")) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "Self-heal daemon is not running"); + goto out; + } + pending_node = GF_CALLOC (1, sizeof (*pending_node), + gf_gld_mt_pending_node_t); + if (!pending_node) { + ret = -1; + goto out; + } + pending_node->node = priv->shd; + pending_node->type = GD_NODE_SHD; + pending_node->index = 0; + list_add_tail (&pending_node->list, selected); ret = 0; } else { list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { brick_index++; - if (uuid_compare (brickinfo->uuid, priv->uuid) || + if (uuid_compare (brickinfo->uuid, MY_UUID) || !glusterd_is_brick_started (brickinfo)) { continue; } @@ -3044,8 +5487,7 @@ glusterd_bricks_select_status_volume (dict_t *dict, char **op_errstr) pending_node->node = brickinfo; pending_node->type = GD_NODE_BRICK; pending_node->index = brick_index; - list_add_tail (&pending_node->list, - &opinfo.pending_bricks); + list_add_tail (&pending_node->list, selected); pending_node = NULL; } } @@ -3062,6 +5504,7 @@ glusterd_op_ac_send_brick_op (glusterd_op_sm_event_t *event, void *ctx) xlator_t *this = NULL; glusterd_op_t op = GD_OP_NONE; glusterd_req_ctx_t *req_ctx = NULL; + char *op_errstr = NULL; this = THIS; priv = this->private; @@ -3073,10 +5516,18 @@ glusterd_op_ac_send_brick_op (glusterd_op_sm_event_t *event, void *ctx) gf_gld_mt_op_allack_ctx_t); op = glusterd_op_get_op (); req_ctx->op = op; - uuid_copy (req_ctx->uuid, priv->uuid); - ret = glusterd_op_build_payload (&req_ctx->dict); - if (ret)//TODO:what to do?? + uuid_copy (req_ctx->uuid, MY_UUID); + ret = glusterd_op_build_payload (&req_ctx->dict, &op_errstr, + NULL); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD, + gd_op_list[op]); + if (op_errstr == NULL) + gf_asprintf (&op_errstr, + OPERRSTR_BUILD_PAYLOAD); + opinfo.op_errstr = op_errstr; goto out; + } } proc = &priv->gfs_mgmt->proctable[GLUSTERD_BRICK_OP]; @@ -3088,11 +5539,12 @@ glusterd_op_ac_send_brick_op (glusterd_op_sm_event_t *event, void *ctx) if (!opinfo.pending_count && !opinfo.brick_pending_count) { glusterd_clear_pending_nodes (&opinfo.pending_bricks); - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, req_ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, req_ctx); } out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } @@ -3105,10 +5557,14 @@ glusterd_op_ac_rcvd_brick_op_acc (glusterd_op_sm_event_t *event, void *ctx) glusterd_op_brick_rsp_ctx_t *ev_ctx = NULL; char *op_errstr = NULL; glusterd_op_t op = GD_OP_NONE; + gd_node_type type = GD_NODE_NONE; dict_t *op_ctx = NULL; glusterd_req_ctx_t *req_ctx = NULL; void *pending_entry = NULL; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (event); GF_ASSERT (ctx); ev_ctx = ctx; @@ -3119,11 +5575,12 @@ glusterd_op_ac_rcvd_brick_op_acc (glusterd_op_sm_event_t *event, void *ctx) op = req_ctx->op; op_ctx = glusterd_op_get_ctx (); pending_entry = ev_ctx->pending_node->node; + type = ev_ctx->pending_node->type; ret = glusterd_remove_pending_entry (&opinfo.pending_bricks, pending_entry); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "unknown response received "); + gf_log (this->name, GF_LOG_ERROR, "unknown response received "); ret = -1; goto out; } @@ -3131,25 +5588,27 @@ glusterd_op_ac_rcvd_brick_op_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.brick_pending_count > 0) opinfo.brick_pending_count--; - glusterd_handle_brick_rsp (pending_entry, op, ev_ctx->rsp_dict, - op_ctx, &op_errstr); + glusterd_handle_node_rsp (req_ctx->dict, pending_entry, op, ev_ctx->rsp_dict, + op_ctx, &op_errstr, type); if (opinfo.brick_pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, ev_ctx->commit_ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, &event->txn_id, + ev_ctx->commit_ctx); out: if (ev_ctx->rsp_dict) dict_unref (ev_ctx->rsp_dict); GF_FREE (ev_ctx); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } int32_t -glusterd_op_bricks_select (glusterd_op_t op, dict_t *dict, char **op_errstr) +glusterd_op_bricks_select (glusterd_op_t op, dict_t *dict, char **op_errstr, + struct list_head *selected, dict_t *rsp_dict) { int ret = 0; @@ -3160,30 +5619,42 @@ glusterd_op_bricks_select (glusterd_op_t op, dict_t *dict, char **op_errstr) switch (op) { case GD_OP_STOP_VOLUME: - ret = glusterd_bricks_select_stop_volume (dict, op_errstr); + ret = glusterd_bricks_select_stop_volume (dict, op_errstr, + selected); break; case GD_OP_REMOVE_BRICK: - ret = glusterd_bricks_select_remove_brick (dict, op_errstr); + ret = glusterd_bricks_select_remove_brick (dict, op_errstr, + selected); break; case GD_OP_PROFILE_VOLUME: - ret = glusterd_bricks_select_profile_volume (dict, op_errstr); + ret = glusterd_bricks_select_profile_volume (dict, op_errstr, + selected); break; case GD_OP_HEAL_VOLUME: - ret = glusterd_bricks_select_heal_volume (dict, op_errstr); + ret = glusterd_bricks_select_heal_volume (dict, op_errstr, + selected, rsp_dict); break; case GD_OP_STATUS_VOLUME: - ret = glusterd_bricks_select_status_volume (dict, op_errstr); + ret = glusterd_bricks_select_status_volume (dict, op_errstr, + selected); break; + case GD_OP_DEFRAG_BRICK_VOLUME: + ret = glusterd_bricks_select_rebalance_volume (dict, op_errstr, + selected); + break; + case GD_OP_SNAP: + ret = glusterd_bricks_select_snap (dict, op_errstr, selected); + break; default: break; } - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -3500,7 +5971,7 @@ glusterd_op_sm_new_event (glusterd_op_sm_event_type_t event_type, int glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type, - void *ctx) + uuid_t *txn_id, void *ctx) { int32_t ret = -1; glusterd_op_sm_event_t *event = NULL; @@ -3515,7 +5986,10 @@ glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type, event->ctx = ctx; - gf_log ("glusterd", GF_LOG_DEBUG, "Enqueue event: '%s'", + if (txn_id) + uuid_copy (event->txn_id, *txn_id); + + gf_log (THIS->name, GF_LOG_DEBUG, "Enqueue event: '%s'", glusterd_op_sm_event_name_get (event->event)); list_add_tail (&event->list, &gd_op_sm_queue); @@ -3570,11 +6044,21 @@ glusterd_op_sm () glusterd_op_sm_event_t *event = NULL; glusterd_op_sm_event_t *tmp = NULL; int ret = -1; + int lock_err = 0; glusterd_op_sm_ac_fn handler = NULL; glusterd_op_sm_t *state = NULL; glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; + xlator_t *this = NULL; + glusterd_op_info_t txn_op_info; - (void ) pthread_mutex_lock (&gd_op_sm_lock); + this = THIS; + GF_ASSERT (this); + + if ((lock_err = pthread_mutex_trylock (&gd_op_sm_lock))) { + gf_log (this->name, GF_LOG_ERROR, "lock failed due to %s", + strerror (lock_err)); + goto lock_failed; + } while (!list_empty (&gd_op_sm_queue)) { @@ -3582,9 +6066,24 @@ glusterd_op_sm () list_del_init (&event->list); event_type = event->event; - gf_log ("", GF_LOG_DEBUG, "Dequeued event of type: '%s'", + gf_log (this->name, GF_LOG_DEBUG, "Dequeued event of " + "type: '%s'", glusterd_op_sm_event_name_get(event_type)); + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", + uuid_utoa (event->txn_id)); + + ret = glusterd_get_txn_opinfo (&event->txn_id, + &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get transaction's opinfo"); + glusterd_destroy_op_event_ctx (event); + GF_FREE (event); + continue; + } else + opinfo = txn_op_info; + state = glusterd_op_state_table[opinfo.state.state]; GF_ASSERT (state); @@ -3595,7 +6094,7 @@ glusterd_op_sm () ret = handler (event, event->ctx); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "handler returned: %d", ret); glusterd_destroy_op_event_ctx (event); GF_FREE (event); @@ -3606,7 +6105,7 @@ glusterd_op_sm () event_type); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Unable to transition" "state from '%s' to '%s'", glusterd_op_sm_state_name_get(opinfo.state.state), @@ -3615,8 +6114,27 @@ glusterd_op_sm () return ret; } + if ((state[event_type].next_state == + GD_OP_STATE_DEFAULT) && + (event_type == GD_OP_EVENT_UNLOCK)) { + /* Clearing the transaction opinfo */ + ret = glusterd_clear_txn_opinfo(&event->txn_id); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to clear " + "transaction's opinfo"); + } else { + ret = glusterd_set_txn_opinfo (&event->txn_id, + &opinfo); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to set " + "transaction's opinfo"); + } + glusterd_destroy_op_event_ctx (event); GF_FREE (event); + } } @@ -3624,6 +6142,8 @@ glusterd_op_sm () (void ) pthread_mutex_unlock (&gd_op_sm_lock); ret = 0; +lock_failed: + return ret; } @@ -3668,48 +6188,6 @@ glusterd_op_clear_op (glusterd_op_t op) } int32_t -glusterd_op_init_ctx (glusterd_op_t op) -{ - int ret = 0; - dict_t *dict = NULL; - - GF_ASSERT (GD_OP_NONE < op && op < GD_OP_MAX); - - if (_gf_false == glusterd_need_brick_op (op)) { - gf_log ("", GF_LOG_DEBUG, "Received op: %d, returning", op); - goto out; - } - dict = dict_new (); - if (dict == NULL) { - ret = -1; - goto out; - } - ret = glusterd_op_set_ctx (dict); - if (ret) - goto out; -out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - return ret; -} - - - -int32_t -glusterd_op_fini_ctx () -{ - dict_t *dict = NULL; - - dict = glusterd_op_get_ctx (); - if (dict) - dict_unref (dict); - - glusterd_op_reset_ctx (); - return 0; -} - - - -int32_t glusterd_op_free_ctx (glusterd_op_t op, void *ctx) { @@ -3733,6 +6211,8 @@ glusterd_op_free_ctx (glusterd_op_t op, void *ctx) case GD_OP_REBALANCE: case GD_OP_HEAL_VOLUME: case GD_OP_STATEDUMP_VOLUME: + case GD_OP_CLEARLOCKS_VOLUME: + case GD_OP_DEFRAG_BRICK_VOLUME: dict_unref (ctx); break; default: @@ -3761,4 +6241,3 @@ glusterd_op_sm_init () pthread_mutex_init (&gd_op_sm_lock, NULL); return 0; } - diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.h b/xlators/mgmt/glusterd/src/glusterd-op-sm.h index e658e6ece..cf57b78e0 100644 --- a/xlators/mgmt/glusterd/src/glusterd-op-sm.h +++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _GLUSTERD_OP_SM_H_ #define _GLUSTERD_OP_SM_H_ @@ -25,8 +15,8 @@ #include "config.h" #endif -#ifndef GSYNC_CONF -#define GSYNC_CONF GEOREP"/gsyncd.conf" +#ifndef GSYNC_CONF_TEMPLATE +#define GSYNC_CONF_TEMPLATE GEOREP"/gsyncd_template.conf" #endif #include <pthread.h> @@ -42,6 +32,8 @@ #include "protocol-common.h" #define GD_VOLUME_NAME_MAX 256 +#define GD_OP_PROTECTED (0x02) +#define GD_OP_UNPROTECTED (0x04) typedef enum glusterd_op_sm_state_ { GD_OP_STATE_DEFAULT = 0, @@ -85,6 +77,7 @@ struct glusterd_op_sm_event_ { struct list_head list; void *ctx; glusterd_op_sm_event_type_t event; + uuid_t txn_id; }; typedef struct glusterd_op_sm_event_ glusterd_op_sm_event_t; @@ -127,6 +120,7 @@ typedef struct glusterd_op_log_filename_ctx_ glusterd_op_log_filename_ctx_t; struct glusterd_op_lock_ctx_ { uuid_t uuid; + dict_t *dict; rpcsvc_request_t *req; }; @@ -154,21 +148,36 @@ typedef struct glusterd_pr_brick_rsp_conv_t { dict_t *dict; } glusterd_pr_brick_rsp_conv_t; +typedef struct glusterd_heal_rsp_conv_ { + dict_t *dict; + glusterd_volinfo_t *volinfo; + xlator_t *this; +} glusterd_heal_rsp_conv_t; + typedef struct glusterd_status_rsp_conv_ { int count; + int brick_index_max; + int other_count; dict_t *dict; } glusterd_status_rsp_conv_t; typedef struct glusterd_gsync_status_temp { dict_t *rsp_dict; glusterd_volinfo_t *volinfo; + char *node; }glusterd_gsync_status_temp_t; + +typedef enum cli_cmd_type_ { + PER_REPLICA, + ALL_REPLICA, + } cli_cmd_type; + int glusterd_op_sm_new_event (glusterd_op_sm_event_type_t event_type, glusterd_op_sm_event_t **new_event); int glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type, - void *ctx); + uuid_t *txn_id, void *ctx); int glusterd_op_sm_init (); @@ -183,7 +192,7 @@ int32_t glusterd_op_set_op (glusterd_op_t op); int -glusterd_op_build_payload (dict_t **req); +glusterd_op_build_payload (dict_t **req, char **op_errstr, dict_t *op_ctx); int32_t glusterd_op_stage_validate (glusterd_op_t op, dict_t *req, char **op_errstr, @@ -193,6 +202,13 @@ int32_t glusterd_op_commit_perform (glusterd_op_t op, dict_t *req, char **op_errstr, dict_t* dict); +int32_t +glusterd_op_txn_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx, + char *err_str, size_t err_len); + +int32_t +glusterd_op_txn_complete (); + void * glusterd_op_get_ctx (); @@ -218,13 +234,8 @@ glusterd_check_option_exists(char *optstring, char **completion); int set_xlator_option (dict_t *dict, char *key, char *value); -char * -glusterd_check_brick_rb_part (char *bricks, int count, glusterd_volinfo_t *volinfo); - void glusterd_do_replace_brick (void *data); -int -glusterd_options_reset (glusterd_volinfo_t *volinfo, char *key, int32_t is_force); char* glusterd_op_sm_state_name_get (int state); @@ -232,18 +243,25 @@ glusterd_op_sm_state_name_get (int state); char* glusterd_op_sm_event_name_get (int event); int32_t -glusterd_op_bricks_select (glusterd_op_t op, dict_t *dict, char **op_errstr); +glusterd_op_bricks_select (glusterd_op_t op, dict_t *dict, char **op_errstr, + struct list_head *selected, dict_t *rsp_dict); int glusterd_brick_op_build_payload (glusterd_op_t op, glusterd_brickinfo_t *brickinfo, gd1_mgmt_brick_op_req **req, dict_t *dict); +int +glusterd_node_op_build_payload (glusterd_op_t op, gd1_mgmt_brick_op_req **req, + dict_t *dict); int32_t -glusterd_handle_brick_rsp (glusterd_brickinfo_t *brickinfo, - glusterd_op_t op, dict_t *rsp_dict, dict_t *ctx_dict, - char **op_errstr); -int32_t -glusterd_op_init_ctx (glusterd_op_t op); -int32_t -glusterd_op_fini_ctx (); +glusterd_handle_brick_rsp (void *pending_entry, glusterd_op_t op, + dict_t *rsp_dict, dict_t *ctx_dict, char **op_errstr, + gd_node_type type); + +dict_t* +glusterd_op_init_commit_rsp_dict (glusterd_op_t op); + +void +glusterd_op_modify_op_ctx (glusterd_op_t op, void *op_ctx); + int32_t glusterd_volume_stats_read_perf (char *brick_path, int32_t blk_size, int32_t blk_count, double *throughput, double *time); @@ -259,14 +277,23 @@ glusterd_are_all_volumes_stopped (); int glusterd_stop_bricks (glusterd_volinfo_t *volinfo); int -glusterd_get_gsync_status_mst_slv( glusterd_volinfo_t *volinfo, - char *slave, dict_t *rsp_dict); +gsync_status (char *master, char *slave, char *conf_path, int *status); + int -gsync_status (char *master, char *slave, int *status); +glusterd_check_gsync_running (glusterd_volinfo_t *volinfo, gf_boolean_t *flag); int -glusterd_gsync_get_param_file (char *prmfile, const char *ext, char *master, - char *slave, char *gl_workdir); +glusterd_defrag_volume_node_rsp (dict_t *req_dict, dict_t *rsp_dict, + dict_t *op_ctx); int -glusterd_check_gsync_running (glusterd_volinfo_t *volinfo, gf_boolean_t *flag); +glusterd_is_valid_vg (glusterd_brickinfo_t *brick, int check_tag, char *msg); + +int32_t +glusterd_get_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t *opinfo); + +int32_t +glusterd_set_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t *opinfo); + +int32_t +glusterd_clear_txn_opinfo (uuid_t *txn_id); #endif diff --git a/xlators/mgmt/glusterd/src/glusterd-pmap.c b/xlators/mgmt/glusterd/src/glusterd-pmap.c index 08e7aedca..a153ca1a9 100644 --- a/xlators/mgmt/glusterd/src/glusterd-pmap.c +++ b/xlators/mgmt/glusterd/src/glusterd-pmap.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2010-2013 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H @@ -62,8 +52,8 @@ pmap_port_isfree (int port) } -struct pmap_registry * -pmap_registry_new (void) +static struct pmap_registry * +pmap_registry_new (xlator_t *this) { struct pmap_registry *pmap = NULL; int i = 0; @@ -79,8 +69,8 @@ pmap_registry_new (void) pmap->ports[i].type = GF_PMAP_PORT_FOREIGN; } - pmap->base_port = GF_DEFAULT_BASE_PORT + 2; - pmap->last_alloc = GF_DEFAULT_BASE_PORT + 2; + pmap->base_port = pmap->last_alloc = + ((glusterd_conf_t *)(this->private))->base_port; return pmap; } @@ -96,7 +86,7 @@ pmap_registry_get (xlator_t *this) pmap = priv->pmap; if (!pmap) { - pmap = pmap_registry_new (); + pmap = pmap_registry_new (this); if (!pmap) return NULL; priv->pmap = pmap; @@ -228,8 +218,7 @@ pmap_registry_bind (xlator_t *this, int port, const char *brickname, p = port; pmap->ports[p].type = type; - if (pmap->ports[p].brickname) - free (pmap->ports[p].brickname); + free (pmap->ports[p].brickname); pmap->ports[p].brickname = strdup (brickname); pmap->ports[p].type = type; pmap->ports[p].xprt = xprt; @@ -281,8 +270,7 @@ remove: gf_log ("pmap", GF_LOG_INFO, "removing brick %s on port %d", pmap->ports[p].brickname, p); - if (pmap->ports[p].brickname) - free (pmap->ports[p].brickname); + free (pmap->ports[p].brickname); pmap->ports[p].brickname = NULL; pmap->ports[p].xprt = NULL; @@ -292,15 +280,17 @@ out: } int -gluster_pmap_portbybrick (rpcsvc_request_t *req) +__gluster_pmap_portbybrick (rpcsvc_request_t *req) { pmap_port_by_brick_req args = {0,}; pmap_port_by_brick_rsp rsp = {0,}; char *brick = NULL; int port = 0; + int ret = -1; - if (!xdr_to_generic (req->msg[0], &args, - (xdrproc_t)xdr_pmap_port_by_brick_req)) { + ret = xdr_to_generic (req->msg[0], &args, + (xdrproc_t)xdr_pmap_port_by_brick_req); + if (ret < 0) { req->rpc_err = GARBAGE_ARGS; goto fail; } @@ -317,21 +307,29 @@ gluster_pmap_portbybrick (rpcsvc_request_t *req) fail: glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_pmap_port_by_brick_rsp); - if (args.brick) - free (args.brick);//malloced by xdr + free (args.brick);//malloced by xdr return 0; } int -gluster_pmap_brickbyport (rpcsvc_request_t *req) +gluster_pmap_portbybrick (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __gluster_pmap_portbybrick); +} + + +int +__gluster_pmap_brickbyport (rpcsvc_request_t *req) { pmap_brick_by_port_req args = {0,}; pmap_brick_by_port_rsp rsp = {0,}; + int ret = -1; - if (!xdr_to_generic (req->msg[0], &args, - (xdrproc_t)xdr_pmap_brick_by_port_req)) { + ret = xdr_to_generic (req->msg[0], &args, + (xdrproc_t)xdr_pmap_brick_by_port_req); + if (ret < 0) { req->rpc_err = GARBAGE_ARGS; goto fail; } @@ -349,6 +347,14 @@ fail: return 0; } + +int +gluster_pmap_brickbyport (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __gluster_pmap_brickbyport); +} + + static int glusterd_brick_update_signin (glusterd_brickinfo_t *brickinfo, gf_boolean_t value) @@ -359,14 +365,16 @@ glusterd_brick_update_signin (glusterd_brickinfo_t *brickinfo, } int -gluster_pmap_signup (rpcsvc_request_t *req) +__gluster_pmap_signup (rpcsvc_request_t *req) { pmap_signup_req args = {0,}; pmap_signup_rsp rsp = {0,}; + int ret = -1; - if (!xdr_to_generic (req->msg[0], &args, - (xdrproc_t)xdr_pmap_signup_req)) { + ret = xdr_to_generic (req->msg[0], &args, + (xdrproc_t)xdr_pmap_signup_req); + if (ret < 0) { req->rpc_err = GARBAGE_ARGS; goto fail; } @@ -377,22 +385,28 @@ gluster_pmap_signup (rpcsvc_request_t *req) fail: glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_pmap_signup_rsp); - if (args.brick) - free (args.brick);//malloced by xdr + free (args.brick);//malloced by xdr return 0; } int -gluster_pmap_signin (rpcsvc_request_t *req) +gluster_pmap_signup (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __gluster_pmap_signup); +} + +int +__gluster_pmap_signin (rpcsvc_request_t *req) { pmap_signin_req args = {0,}; pmap_signin_rsp rsp = {0,}; glusterd_brickinfo_t *brickinfo = NULL; int ret = -1; - if (!xdr_to_generic (req->msg[0], &args, - (xdrproc_t)xdr_pmap_signin_req)) { + ret = xdr_to_generic (req->msg[0], &args, + (xdrproc_t)xdr_pmap_signin_req); + if (ret < 0) { req->rpc_err = GARBAGE_ARGS; goto fail; } @@ -405,8 +419,7 @@ gluster_pmap_signin (rpcsvc_request_t *req) fail: glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_pmap_signin_rsp); - if (args.brick) - free (args.brick);//malloced by xdr + free (args.brick);//malloced by xdr if (!ret) glusterd_brick_update_signin (brickinfo, _gf_true); @@ -415,17 +428,24 @@ fail: } +int +gluster_pmap_signin (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __gluster_pmap_signin); +} + int -gluster_pmap_signout (rpcsvc_request_t *req) +__gluster_pmap_signout (rpcsvc_request_t *req) { pmap_signout_req args = {0,}; pmap_signout_rsp rsp = {0,}; int ret = -1; glusterd_brickinfo_t *brickinfo = NULL; - if (!xdr_to_generic (req->msg[0], &args, - (xdrproc_t)xdr_pmap_signout_req)) { + ret = xdr_to_generic (req->msg[0], &args, + (xdrproc_t)xdr_pmap_signout_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto fail; @@ -439,8 +459,7 @@ gluster_pmap_signout (rpcsvc_request_t *req) fail: glusterd_submit_reply (req, &rsp, NULL, 0, NULL, (xdrproc_t)xdr_pmap_signout_rsp); - if (args.brick) - free (args.brick);//malloced by xdr + free (args.brick);//malloced by xdr if (!ret) glusterd_brick_update_signin (brickinfo, _gf_false); @@ -448,18 +467,19 @@ fail: return 0; } +int +gluster_pmap_signout (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __gluster_pmap_signout); +} + rpcsvc_actor_t gluster_pmap_actors[] = { - [GF_PMAP_NULL] = {"NULL", GF_PMAP_NULL, NULL, NULL, NULL, 0}, - [GF_PMAP_PORTBYBRICK] = {"PORTBYBRICK", GF_PMAP_PORTBYBRICK, - gluster_pmap_portbybrick, NULL, NULL, 0}, - [GF_PMAP_BRICKBYPORT] = {"BRICKBYPORT", GF_PMAP_BRICKBYPORT, - gluster_pmap_brickbyport, NULL, NULL, 0}, - [GF_PMAP_SIGNIN] = {"SIGNIN", GF_PMAP_SIGNIN, - gluster_pmap_signin, NULL, NULL, 0}, - [GF_PMAP_SIGNOUT] = {"SIGNOUT", GF_PMAP_SIGNOUT, - gluster_pmap_signout, NULL, NULL, 0}, - [GF_PMAP_SIGNUP] = {"SIGNUP", GF_PMAP_SIGNUP, - gluster_pmap_signup, NULL, NULL, 0}, + [GF_PMAP_NULL] = {"NULL", GF_PMAP_NULL, NULL, NULL, 0, DRC_NA}, + [GF_PMAP_PORTBYBRICK] = {"PORTBYBRICK", GF_PMAP_PORTBYBRICK, gluster_pmap_portbybrick, NULL, 0, DRC_NA}, + [GF_PMAP_BRICKBYPORT] = {"BRICKBYPORT", GF_PMAP_BRICKBYPORT, gluster_pmap_brickbyport, NULL, 0, DRC_NA}, + [GF_PMAP_SIGNIN] = {"SIGNIN", GF_PMAP_SIGNIN, gluster_pmap_signin, NULL, 0, DRC_NA}, + [GF_PMAP_SIGNOUT] = {"SIGNOUT", GF_PMAP_SIGNOUT, gluster_pmap_signout, NULL, 0, DRC_NA}, + [GF_PMAP_SIGNUP] = {"SIGNUP", GF_PMAP_SIGNUP, gluster_pmap_signup, NULL, 0, DRC_NA}, }; diff --git a/xlators/mgmt/glusterd/src/glusterd-pmap.h b/xlators/mgmt/glusterd/src/glusterd-pmap.h index 8f1bb4500..6336ee998 100644 --- a/xlators/mgmt/glusterd/src/glusterd-pmap.h +++ b/xlators/mgmt/glusterd/src/glusterd-pmap.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _GLUSTERD_PMAP_H_ #define _GLUSTERD_PMAP_H_ @@ -38,6 +28,8 @@ #include "rpcsvc.h" +#define GF_IANA_PRIV_PORTS_START 49152 /* RFC 6335 */ + struct pmap_port_status { gf_pmap_port_type_t type; char *brickname; diff --git a/xlators/mgmt/glusterd/src/glusterd-quota.c b/xlators/mgmt/glusterd/src/glusterd-quota.c index 881cf66b4..318267199 100644 --- a/xlators/mgmt/glusterd/src/glusterd-quota.c +++ b/xlators/mgmt/glusterd/src/glusterd-quota.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" @@ -35,7 +25,7 @@ #include <sys/wait.h> int -glusterd_handle_quota (rpcsvc_request_t *req) +__glusterd_handle_quota (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; @@ -44,11 +34,15 @@ glusterd_handle_quota (rpcsvc_request_t *req) char operation[256] = {0, }; char *volname = NULL; int32_t type = 0; + char msg[2048] = {0,}; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; @@ -62,8 +56,10 @@ glusterd_handle_quota (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, "failed to " + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (msg, sizeof (msg), "Unable to decode the " + "command"); goto out; } else { dict->extra_stdfree = cli_req.dict.dict_val; @@ -72,16 +68,18 @@ glusterd_handle_quota (rpcsvc_request_t *req) ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_WARNING, "Unable to get volume name, while" - "handling quota command"); + snprintf (msg, sizeof (msg), "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name, " + "while handling quota command"); goto out; } ret = dict_get_int32 (dict, "type", &type); if (ret) { - gf_log ("", GF_LOG_WARNING, "Unable to get type of cmd. , while" - "handling quota command"); - goto out; + snprintf (msg, sizeof (msg), "Unable to get type of command"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get type of cmd, " + "while handling quota command"); + goto out; } switch (type) { @@ -101,25 +99,25 @@ glusterd_handle_quota (rpcsvc_request_t *req) strncpy (operation, "remove", sizeof (operation)); break; } - gf_cmd_log ("volume quota", " %s command on %s", operation, volname); - ret = glusterd_op_begin (req, GD_OP_QUOTA, dict); - gf_cmd_log ("volume quota", " %s command on %s %s", operation,volname, - (ret != 0)? "FAILED" : "SUCCEEDED"); + ret = glusterd_op_begin_synctask (req, GD_OP_QUOTA, dict); out: - glusterd_friend_sm (); - glusterd_op_sm (); - if (ret) { - if (dict) - dict_unref (dict); + if (msg[0] == '\0') + snprintf (msg, sizeof (msg), "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, msg); } return ret; } +int +glusterd_handle_quota (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_quota); +} + int32_t glusterd_check_if_quota_trans_enabled (glusterd_volinfo_t *volinfo) { @@ -237,8 +235,7 @@ _glusterd_quota_remove_limits (char **quota_limits, char *path, } out: - if (limits) - GF_FREE (limits); + GF_FREE (limits); return ret; } @@ -246,11 +243,11 @@ out: int32_t glusterd_quota_initiate_fs_crawl (glusterd_conf_t *priv, char *volname) { - int32_t ret = 0; - pid_t pid; - char mountdir [] = "/tmp/mntXXXXXX"; - runner_t runner = {0,}; - int status = 0; + pid_t pid; + int32_t ret = 0; + int status = 0; + char mountdir[] = "/tmp/mntXXXXXX"; + runner_t runner = {0}; if (mkdtemp (mountdir) == NULL) { gf_log ("glusterd", GF_LOG_DEBUG, @@ -260,12 +257,15 @@ glusterd_quota_initiate_fs_crawl (glusterd_conf_t *priv, char *volname) } runinit (&runner); - runner_add_args (&runner, SBIN_DIR"/glusterfs", "-s", - "localhost", "--volfile-id", volname, "-l", - DEFAULT_LOG_FILE_DIRECTORY"/quota-crawl.log", + runner_add_args (&runner, SBIN_DIR"/glusterfs", + "-s", "localhost", + "--volfile-id", volname, + "-l", DEFAULT_LOG_FILE_DIRECTORY"/quota-crawl.log", mountdir, NULL); + synclock_unlock (&priv->big_lock); ret = runner_run_reuse (&runner); + synclock_lock (&priv->big_lock); if (ret == -1) { runner_log (&runner, "glusterd", GF_LOG_DEBUG, "command failed"); runner_end (&runner); @@ -384,18 +384,17 @@ glusterd_quota_get_limit_usages (glusterd_conf_t *priv, glusterd_volinfo_t *volinfo, char *volname, dict_t *dict, - char **op_errstr) + char **op_errstr, + dict_t *rsp_dict) { int32_t i = 0; int32_t ret = 0; int32_t count = 0; char *path = NULL; - dict_t *ctx = NULL; char cmd_str [1024] = {0, }; char *ret_str = NULL; - ctx = glusterd_op_get_ctx (); - if (ctx == NULL) + if (rsp_dict == NULL) return 0; ret = dict_get_int32 (dict, "count", &count); @@ -419,7 +418,7 @@ glusterd_quota_get_limit_usages (glusterd_conf_t *priv, } if (ret_str) { - ret = dict_set_dynstr (ctx, "limit_list", ret_str); + ret = dict_set_dynstr (rsp_dict, "limit_list", ret_str); } out: return ret; @@ -530,6 +529,13 @@ glusterd_quota_limit_usage (glusterd_volinfo_t *volinfo, dict_t *dict, char **op GF_VALIDATE_OR_GOTO ("glusterd", volinfo, out); GF_VALIDATE_OR_GOTO ("glusterd", op_errstr, out); + ret = glusterd_check_if_quota_trans_enabled (volinfo); + if (ret == -1) { + *op_errstr = gf_strdup ("Quota is disabled, please enable " + "quota"); + goto out; + } + ret = glusterd_volinfo_get (volinfo, VKEY_FEATURES_LIMIT_USAGE, "a_limits); if (ret) { @@ -667,12 +673,11 @@ out: int -glusterd_op_quota (dict_t *dict, char **op_errstr) +glusterd_op_quota (dict_t *dict, char **op_errstr, dict_t *rsp_dict) { glusterd_volinfo_t *volinfo = NULL; int32_t ret = -1; char *volname = NULL; - dict_t *ctx = NULL; int type = -1; gf_boolean_t start_crawl = _gf_false; glusterd_conf_t *priv = NULL; @@ -737,7 +742,7 @@ glusterd_op_quota (dict_t *dict, char **op_errstr) } ret = glusterd_quota_get_limit_usages (priv, volinfo, volname, - dict, op_errstr); + dict, op_errstr, rsp_dict); goto out; } @@ -760,12 +765,11 @@ create_vol: ret = 0; out: - ctx = glusterd_op_get_ctx (); - if (ctx && start_crawl == _gf_true) + if (rsp_dict && start_crawl == _gf_true) glusterd_quota_initiate_fs_crawl (priv, volname); - if (ctx && *op_errstr) { - ret = dict_set_dynstr (ctx, "errstr", *op_errstr); + if (rsp_dict && *op_errstr) { + ret = dict_set_dynstr (rsp_dict, "errstr", *op_errstr); if (ret) { GF_FREE (*op_errstr); gf_log ("", GF_LOG_DEBUG, diff --git a/xlators/mgmt/glusterd/src/glusterd-rebalance.c b/xlators/mgmt/glusterd/src/glusterd-rebalance.c index e5a907eaa..b7b974c68 100644 --- a/xlators/mgmt/glusterd/src/glusterd-rebalance.c +++ b/xlators/mgmt/glusterd/src/glusterd-rebalance.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" @@ -41,581 +31,336 @@ #include "glusterd-utils.h" #include "glusterd-store.h" #include "run.h" +#include "glusterd-volgen.h" #include "syscall.h" #include "cli1-xdr.h" #include "xdr-generic.h" -/* return values - 0: success, +ve: stopped, -ve: failure */ +int32_t +glusterd_brick_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe); int -gf_glusterd_rebalance_move_data (glusterd_volinfo_t *volinfo, const char *dir) +glusterd_defrag_start_validate (glusterd_volinfo_t *volinfo, char *op_errstr, + size_t len, glusterd_op_t op) { - int ret = -1; - DIR *fd = NULL; - glusterd_defrag_info_t *defrag = NULL; - struct dirent *entry = NULL; - struct stat stbuf = {0,}; - char full_path[PATH_MAX] = {0,}; - char linkinfo[PATH_MAX] = {0,}; - char force_string[64] = {0,}; - - if (!volinfo->defrag) - goto out; - - defrag = volinfo->defrag; - - fd = opendir (dir); - if (!fd) + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + /* Check only if operation is not remove-brick */ + if ((GD_OP_REMOVE_BRICK != op) && + !gd_is_remove_brick_committed (volinfo)) { + gf_log (this->name, GF_LOG_DEBUG, "A remove-brick task on " + "volume %s is not yet committed", volinfo->volname); + snprintf (op_errstr, len, "A remove-brick task on volume %s is" + " not yet committed. Either commit or stop the " + "remove-brick task.", volinfo->volname); goto out; - - if ((defrag->cmd == GF_DEFRAG_CMD_START_MIGRATE_DATA_FORCE) || - (defrag->cmd == GF_DEFRAG_CMD_START_FORCE)) { - strcpy (force_string, "force"); - } else { - strcpy (force_string, "not-force"); } - while ((entry = readdir (fd))) { - if (!entry) - break; - - /* We have to honor 'stop' (or 'pause'|'commit') as early - as possible */ - if (volinfo->defrag_status != - GF_DEFRAG_STATUS_MIGRATE_DATA_STARTED) { - /* It can be one of 'stopped|paused|commit' etc */ - closedir (fd); - ret = 1; - goto out; - } - - if (!strcmp (entry->d_name, ".") || !strcmp (entry->d_name, "..")) - continue; - - snprintf (full_path, PATH_MAX, "%s/%s", dir, entry->d_name); - - ret = lstat (full_path, &stbuf); - if (ret == -1) - continue; - - if (S_ISDIR (stbuf.st_mode)) - continue; - - defrag->num_files_lookedup += 1; - - /* TODO: bring in feature to support hardlink rebalance */ - if (stbuf.st_nlink > 1) - continue; - - /* if distribute is present, it will honor this key. - -1 is returned if distribute is not present or file doesn't - have a link-file. If file has link-file, the path of - link-file will be the value, and also that guarantees - that file has to be mostly migrated */ - ret = sys_lgetxattr (full_path, GF_XATTR_LINKINFO_KEY, - &linkinfo, PATH_MAX); - if (ret <= 0) - continue; - - ret = sys_lsetxattr (full_path, "distribute.migrate-data", - force_string, strlen (force_string), 0); - - /* if errno is not ENOSPC or ENOTCONN, we can still continue - with rebalance process */ - if ((ret == -1) && ((errno != ENOSPC) || - (errno != ENOTCONN))) - continue; - - if ((ret == -1) && (errno == ENOTCONN)) { - /* Most probably mount point went missing (mostly due - to a brick down), say rebalance failure to user, - let him restart it if everything is fine */ - volinfo->defrag_status = GF_DEFRAG_STATUS_FAILED; - break; - } - - if ((ret == -1) && (errno == ENOSPC)) { - /* rebalance process itself failed, may be - remote brick went down, or write failed due to - disk full etc etc.. */ - volinfo->defrag_status = GF_DEFRAG_STATUS_FAILED; - break; - } - - LOCK (&defrag->lock); - { - defrag->total_files += 1; - defrag->total_data += stbuf.st_size; - } - UNLOCK (&defrag->lock); + if (glusterd_is_defrag_on (volinfo)) { + gf_log (this->name, GF_LOG_DEBUG, + "rebalance on volume %s already started", + volinfo->volname); + snprintf (op_errstr, len, "Rebalance on %s is already started", + volinfo->volname); + goto out; } - closedir (fd); - fd = opendir (dir); - if (!fd) + if (glusterd_is_rb_started (volinfo) || + glusterd_is_rb_paused (volinfo)) { + gf_log (this->name, GF_LOG_DEBUG, + "Rebalance failed as replace brick is in progress on volume %s", + volinfo->volname); + snprintf (op_errstr, len, "Rebalance failed as replace brick is in progress on " + "volume %s", volinfo->volname); goto out; - while ((entry = readdir (fd))) { - if (!entry) - break; - - /* We have to honor 'stop' (or 'pause'|'commit') as early - as possible */ - if (volinfo->defrag_status != - GF_DEFRAG_STATUS_MIGRATE_DATA_STARTED) { - /* It can be one of 'stopped|paused|commit' etc */ - closedir (fd); - ret = 1; - goto out; - } - - if (!strcmp (entry->d_name, ".") || !strcmp (entry->d_name, "..")) - continue; - - snprintf (full_path, 1024, "%s/%s", dir, entry->d_name); - - ret = lstat (full_path, &stbuf); - if (ret == -1) - continue; - - if (!S_ISDIR (stbuf.st_mode)) - continue; - - ret = gf_glusterd_rebalance_move_data (volinfo, full_path); - if (ret) - break; } - closedir (fd); - - if (!entry) - ret = 0; + ret = 0; out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } -/* return values - 0: success, +ve: stopped, -ve: failure */ -int -gf_glusterd_rebalance_fix_layout (glusterd_volinfo_t *volinfo, const char *dir) -{ - int ret = -1; - char full_path[1024] = {0,}; - struct stat stbuf = {0,}; - DIR *fd = NULL; - struct dirent *entry = NULL; - - if (!volinfo->defrag) - goto out; - - fd = opendir (dir); - if (!fd) - goto out; - - while ((entry = readdir (fd))) { - if (!entry) - break; - - /* We have to honor 'stop' (or 'pause'|'commit') as early - as possible */ - if (volinfo->defrag_status != - GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED) { - /* It can be one of 'stopped|paused|commit' etc */ - closedir (fd); - ret = 1; - goto out; - } - - if (!strcmp (entry->d_name, ".") || !strcmp (entry->d_name, "..")) - continue; - - snprintf (full_path, 1024, "%s/%s", dir, entry->d_name); - - ret = lstat (full_path, &stbuf); - if (ret == -1) - continue; - if (S_ISDIR (stbuf.st_mode)) { - /* Fix the layout of the directory */ - /* TODO: isn't error code not important ? */ - sys_lsetxattr (full_path, "distribute.fix.layout", - "yes", 3, 0); - - volinfo->defrag->total_files += 1; - - /* Traverse into subdirectory */ - ret = gf_glusterd_rebalance_fix_layout (volinfo, - full_path); - if (ret) - break; - } - } - closedir (fd); - - if (!entry) - ret = 0; +int32_t +__glusterd_defrag_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, void *data) +{ + glusterd_volinfo_t *volinfo = NULL; + glusterd_defrag_info_t *defrag = NULL; + int ret = 0; + char pidfile[PATH_MAX]; + glusterd_conf_t *priv = NULL; -out: - return ret; -} + priv = THIS->private; + if (!priv) + return 0; -void * -glusterd_defrag_start (void *data) -{ - glusterd_volinfo_t *volinfo = data; - glusterd_defrag_info_t *defrag = NULL; - int ret = -1; - struct stat stbuf = {0,}; + volinfo = mydata; + if (!volinfo) + return 0; - THIS = volinfo->xl; - defrag = volinfo->defrag; + defrag = volinfo->rebal.defrag; if (!defrag) - goto out; + return 0; - sleep (1); - ret = lstat (defrag->mount, &stbuf); - if ((ret == -1) && (errno == ENOTCONN)) { - /* Wait for some more time before starting rebalance */ - sleep (2); - ret = lstat (defrag->mount, &stbuf); - if (ret == -1) { - volinfo->defrag_status = GF_DEFRAG_STATUS_FAILED; - volinfo->rebalance_files = 0; - volinfo->rebalance_data = 0; - volinfo->lookedup_files = 0; - goto out; - } - } + if ((event == RPC_CLNT_DISCONNECT) && defrag->connected) + volinfo->rebal.defrag = NULL; - /* Fix the root ('/') first */ - sys_lsetxattr (defrag->mount, "distribute.fix.layout", - "yes", 3, 0); + GLUSTERD_GET_DEFRAG_PID_FILE(pidfile, volinfo, priv); - if ((defrag->cmd == GF_DEFRAG_CMD_START) || - (defrag->cmd == GF_DEFRAG_CMD_START_LAYOUT_FIX)) { - /* root's layout got fixed */ - defrag->total_files = 1; + switch (event) { + case RPC_CLNT_CONNECT: + { + if (defrag->connected) + return 0; - /* Step 1: Fix layout of all the directories */ - ret = gf_glusterd_rebalance_fix_layout (volinfo, defrag->mount); - if (ret < 0) - volinfo->defrag_status = GF_DEFRAG_STATUS_FAILED; - /* in both 'stopped' or 'failure' cases goto out */ - if (ret) { - goto out; + LOCK (&defrag->lock); + { + defrag->connected = 1; } + UNLOCK (&defrag->lock); - /* Completed first step */ - volinfo->defrag_status = GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE; + gf_log ("", GF_LOG_DEBUG, "%s got RPC_CLNT_CONNECT", + rpc->conn.trans->name); + break; } - if (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) { - /* It was used by number of layout fixes on directories */ - defrag->total_files = 0; - - volinfo->defrag_status = GF_DEFRAG_STATUS_MIGRATE_DATA_STARTED; + case RPC_CLNT_DISCONNECT: + { + if (!defrag->connected) + return 0; - /* Step 2: Iterate over directories to move data */ - ret = gf_glusterd_rebalance_move_data (volinfo, defrag->mount); - if (ret < 0) - volinfo->defrag_status = GF_DEFRAG_STATUS_FAILED; - /* in both 'stopped' or 'failure' cases goto out */ - if (ret) { - goto out; + LOCK (&defrag->lock); + { + defrag->connected = 0; } + UNLOCK (&defrag->lock); - /* Completed second step */ - volinfo->defrag_status = GF_DEFRAG_STATUS_MIGRATE_DATA_COMPLETE; - } - - /* Completed whole process */ - if ((defrag->cmd == GF_DEFRAG_CMD_START) || - (defrag->cmd == GF_DEFRAG_CMD_START_FORCE)) - volinfo->defrag_status = GF_DEFRAG_STATUS_COMPLETE; - - volinfo->rebalance_files = defrag->total_files; - volinfo->rebalance_data = defrag->total_data; - volinfo->lookedup_files = defrag->num_files_lookedup; -out: - volinfo->defrag = NULL; - if (defrag) { - gf_log ("rebalance", GF_LOG_INFO, "rebalance on %s complete", - defrag->mount); + if (!glusterd_is_service_running (pidfile, NULL)) { + if (volinfo->rebal.defrag_status == + GF_DEFRAG_STATUS_STARTED) { + volinfo->rebal.defrag_status = + GF_DEFRAG_STATUS_FAILED; + } + } - ret = runcmd ("umount", "-l", defrag->mount, NULL); - LOCK_DESTROY (&defrag->lock); + glusterd_store_perform_node_state_store (volinfo); - if (defrag->cbk_fn) { - defrag->cbk_fn (volinfo, volinfo->defrag_status); + if (defrag->rpc) { + rpc_clnt_unref (defrag->rpc); + defrag->rpc = NULL; } + if (defrag->cbk_fn) + defrag->cbk_fn (volinfo, + volinfo->rebal.defrag_status); GF_FREE (defrag); + gf_log ("", GF_LOG_DEBUG, "%s got RPC_CLNT_DISCONNECT", + rpc->conn.trans->name); + break; } - return NULL; + default: + gf_log ("", GF_LOG_TRACE, + "got some other RPC event %d", event); + ret = 0; + break; + } + + return ret; } -int -glusterd_defrag_stop_validate (glusterd_volinfo_t *volinfo, - char *op_errstr, size_t len) +int32_t +glusterd_defrag_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, void *data) { - int ret = -1; - if (glusterd_is_defrag_on (volinfo) == 0) { - snprintf (op_errstr, len, "Rebalance on %s is either Completed " - "or not yet started", volinfo->volname); - goto out; - } - ret = 0; -out: - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); - return ret; + return glusterd_big_locked_notify (rpc, mydata, event, + data, __glusterd_defrag_notify); } int -glusterd_defrag_stop (glusterd_volinfo_t *volinfo, u_quad_t *files, - u_quad_t *size, char *op_errstr, size_t len) +glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr, + size_t len, int cmd, defrag_cbk_fn_t cbk, + glusterd_op_t op) { - /* TODO: set a variable 'stop_defrag' here, it should be checked - in defrag loop */ - int ret = -1; + int ret = -1; + glusterd_defrag_info_t *defrag = NULL; + runner_t runner = {0,}; + glusterd_conf_t *priv = NULL; + char defrag_path[PATH_MAX]; + char sockfile[PATH_MAX] = {0,}; + char pidfile[PATH_MAX] = {0,}; + char logfile[PATH_MAX] = {0,}; + dict_t *options = NULL; + char valgrind_logfile[PATH_MAX] = {0,}; + + priv = THIS->private; + GF_ASSERT (volinfo); - GF_ASSERT (files); - GF_ASSERT (size); GF_ASSERT (op_errstr); - if (!volinfo) { - ret = -1; + ret = glusterd_defrag_start_validate (volinfo, op_errstr, len, op); + if (ret) goto out; - } - - ret = glusterd_defrag_stop_validate (volinfo, op_errstr, len); - if (ret) { - /* rebalance may be happening on other nodes */ - ret = 0; + if (!volinfo->rebal.defrag) + volinfo->rebal.defrag = + GF_CALLOC (1, sizeof (*volinfo->rebal.defrag), + gf_gld_mt_defrag_info); + if (!volinfo->rebal.defrag) goto out; - } - ret = 0; - if (volinfo->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED) { - goto out; - } + defrag = volinfo->rebal.defrag; - LOCK (&volinfo->defrag->lock); - { - volinfo->defrag_status = GF_DEFRAG_STATUS_STOPPED; - *files = volinfo->defrag->total_files; - *size = volinfo->defrag->total_data; - } - UNLOCK (&volinfo->defrag->lock); + defrag->cmd = cmd; - ret = 0; -out: - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); - return ret; -} + volinfo->rebal.defrag_cmd = cmd; + volinfo->rebal.op = op; -int -glusterd_defrag_status_get (glusterd_volinfo_t *volinfo, - dict_t *dict) -{ - int ret = 0; - uint64_t files = 0; - uint64_t size = 0; - uint64_t lookup = 0; + LOCK_INIT (&defrag->lock); - if (!volinfo || !dict) - goto out; + volinfo->rebal.defrag_status = GF_DEFRAG_STATUS_STARTED; - ret = 0; - if (volinfo->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED) - goto out; + glusterd_volinfo_reset_defrag_stats (volinfo); + glusterd_store_perform_node_state_store (volinfo); - if (volinfo->defrag) { - LOCK (&volinfo->defrag->lock); - { - files = volinfo->defrag->total_files; - size = volinfo->defrag->total_data; - lookup = volinfo->defrag->num_files_lookedup; - } - UNLOCK (&volinfo->defrag->lock); - } else { - files = volinfo->rebalance_files; - size = volinfo->rebalance_data; - lookup = volinfo->lookedup_files; + GLUSTERD_GET_DEFRAG_DIR (defrag_path, volinfo, priv); + ret = mkdir_p (defrag_path, 0777, _gf_true); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to create " + "directory %s", defrag_path); + goto out; } - ret = dict_set_uint64 (dict, "files", files); - if (ret) - gf_log (THIS->name, GF_LOG_WARNING, - "failed to set file count"); - - ret = dict_set_uint64 (dict, "size", size); - if (ret) - gf_log (THIS->name, GF_LOG_WARNING, - "failed to set size of xfer"); - - ret = dict_set_uint64 (dict, "lookups", lookup); - if (ret) - gf_log (THIS->name, GF_LOG_WARNING, - "failed to set lookedup file count"); - - ret = dict_set_int32 (dict, "status", volinfo->defrag_status); - if (ret) - gf_log (THIS->name, GF_LOG_WARNING, - "failed to set status"); + GLUSTERD_GET_DEFRAG_SOCK_FILE (sockfile, volinfo, priv); + GLUSTERD_GET_DEFRAG_PID_FILE (pidfile, volinfo, priv); + snprintf (logfile, PATH_MAX, "%s/%s-rebalance.log", + DEFAULT_LOG_FILE_DIRECTORY, volinfo->volname); + runinit (&runner); -out: - return 0; -} + if (priv->valgrind) { + snprintf (valgrind_logfile, PATH_MAX, + "%s/valgrind-%s-rebalance.log", + DEFAULT_LOG_FILE_DIRECTORY, + volinfo->volname); -void -glusterd_rebalance_cmd_attempted_log (int cmd, char *volname) -{ - switch (cmd) { - case GF_DEFRAG_CMD_START_LAYOUT_FIX: - gf_cmd_log ("Volume rebalance"," on volname: %s " - "cmd: start fix layout , attempted", - volname); - gf_log ("glusterd", GF_LOG_INFO, "Received rebalance " - "volume start layout fix on %s", volname); - break; - case GF_DEFRAG_CMD_START_MIGRATE_DATA: - case GF_DEFRAG_CMD_START_MIGRATE_DATA_FORCE: - gf_cmd_log ("Volume rebalance"," on volname: %s " - "cmd: start data migrate attempted", - volname); - gf_log ("glusterd", GF_LOG_INFO, "Received rebalance " - "volume start migrate data on %s", volname); - break; - case GF_DEFRAG_CMD_START: - gf_cmd_log ("Volume rebalance"," on volname: %s " - "cmd: start, attempted", volname); - gf_log ("glusterd", GF_LOG_INFO, "Received rebalance " - "volume start on %s", volname); - break; - case GF_DEFRAG_CMD_STOP: - gf_cmd_log ("Volume rebalance"," on volname: %s " - "cmd: stop, attempted", volname); - gf_log ("glusterd", GF_LOG_INFO, "Received rebalance " - "volume stop on %s", volname); - break; - default: - break; + runner_add_args (&runner, "valgrind", "--leak-check=full", + "--trace-children=yes", "--track-origins=yes", + NULL); + runner_argprintf (&runner, "--log-file=%s", valgrind_logfile); } -} -void -glusterd_rebalance_cmd_log (int cmd, char *volname, int status) -{ - if (cmd != GF_DEFRAG_CMD_STATUS) { - gf_cmd_log ("volume rebalance"," on volname: %s %d %s", - volname, cmd, ((status)?"FAILED":"SUCCESS")); + runner_add_args (&runner, SBIN_DIR"/glusterfs", + "-s", "localhost", "--volfile-id", volinfo->volname, + "--xlator-option", "*dht.use-readdirp=yes", + "--xlator-option", "*dht.lookup-unhashed=yes", + "--xlator-option", "*dht.assert-no-child-down=yes", + "--xlator-option", "*replicate*.data-self-heal=off", + "--xlator-option", + "*replicate*.metadata-self-heal=off", + "--xlator-option", "*replicate*.entry-self-heal=off", + "--xlator-option", "*replicate*.readdir-failover=off", + "--xlator-option", "*dht.readdir-optimize=on", + NULL); + runner_add_arg (&runner, "--xlator-option"); + runner_argprintf ( &runner, "*dht.rebalance-cmd=%d",cmd); + runner_add_arg (&runner, "--xlator-option"); + runner_argprintf (&runner, "*dht.node-uuid=%s", uuid_utoa(MY_UUID)); + runner_add_arg (&runner, "--socket-file"); + runner_argprintf (&runner, "%s",sockfile); + runner_add_arg (&runner, "--pid-file"); + runner_argprintf (&runner, "%s",pidfile); + runner_add_arg (&runner, "-l"); + runner_argprintf (&runner, logfile); + if (volinfo->memory_accounting) + runner_add_arg (&runner, "--mem-accounting"); + + ret = runner_run_nowait (&runner); + if (ret) { + gf_log ("glusterd", GF_LOG_DEBUG, "rebalance command failed"); + goto out; } -} -int -glusterd_defrag_start_validate (glusterd_volinfo_t *volinfo, char *op_errstr, - size_t len) -{ - int ret = -1; + sleep (5); - if (glusterd_is_defrag_on (volinfo)) { - gf_log ("glusterd", GF_LOG_DEBUG, - "rebalance on volume %s already started", - volinfo->volname); - snprintf (op_errstr, len, "Rebalance on %s is already started", - volinfo->volname); + /* Setting frame-timeout to 10mins (600seconds). + * Unix domain sockets ensures that the connection is reliable. The + * default timeout of 30mins used for unreliable network connections is + * too long for unix domain socket connections. + */ + ret = rpc_transport_unix_options_build (&options, sockfile, 600); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Unix options build failed"); goto out; } - if (glusterd_is_rb_started (volinfo) || - glusterd_is_rb_paused (volinfo)) { - gf_log ("glusterd", GF_LOG_DEBUG, - "Rebalance failed as replace brick is in progress on volume %s", - volinfo->volname); - snprintf (op_errstr, len, "Rebalance failed as replace brick is in progress on " - "volume %s", volinfo->volname); + synclock_unlock (&priv->big_lock); + ret = glusterd_rpc_create (&defrag->rpc, options, + glusterd_defrag_notify, volinfo); + synclock_lock (&priv->big_lock); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "RPC create failed"); goto out; } - ret = 0; + + if (cbk) + defrag->cbk_fn = cbk; + out: - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } + int -glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr, - size_t len, int cmd, defrag_cbk_fn_t cbk) +glusterd_rebalance_rpc_create (glusterd_volinfo_t *volinfo, + glusterd_conf_t *priv, int cmd) { - int ret = -1; - glusterd_defrag_info_t *defrag = NULL; - runner_t runner = {0,}; - glusterd_conf_t *priv = NULL; + dict_t *options = NULL; + char sockfile[PATH_MAX] = {0,}; + int ret = -1; + glusterd_defrag_info_t *defrag = NULL; - priv = THIS->private; + if (!volinfo->rebal.defrag) + volinfo->rebal.defrag = + GF_CALLOC (1, sizeof (*volinfo->rebal.defrag), + gf_gld_mt_defrag_info); - GF_ASSERT (volinfo); - GF_ASSERT (op_errstr); - - ret = glusterd_defrag_start_validate (volinfo, op_errstr, len); - if (ret) - goto out; - if (!volinfo->defrag) - volinfo->defrag = GF_CALLOC (1, sizeof (glusterd_defrag_info_t), - gf_gld_mt_defrag_info); - if (!volinfo->defrag) + if (!volinfo->rebal.defrag) goto out; - defrag = volinfo->defrag; + defrag = volinfo->rebal.defrag; defrag->cmd = cmd; LOCK_INIT (&defrag->lock); - snprintf (defrag->mount, 1024, "%s/mount/%s", - priv->workdir, volinfo->volname); - /* Create a directory, mount glusterfs over it, start glusterfs-defrag */ - runinit (&runner); - runner_add_args (&runner, "mkdir", "-p", defrag->mount, NULL); - ret = runner_run_reuse (&runner); - if (ret) { - runner_log (&runner, "glusterd", GF_LOG_DEBUG, "command failed"); - runner_end (&runner); - goto out; - } - runner_end (&runner); - runinit (&runner); - runner_add_args (&runner, SBIN_DIR"/glusterfs", - "-s", "localhost", "--volfile-id", volinfo->volname, - "--xlator-option", "*dht.use-readdirp=yes", - "--xlator-option", "*dht.lookup-unhashed=yes", - "--xlator-option", "*dht.assert-no-child-down=yes", - defrag->mount, NULL); - ret = runner_run_reuse (&runner); + GLUSTERD_GET_DEFRAG_SOCK_FILE (sockfile, volinfo, priv); + + /* Setting frame-timeout to 10mins (600seconds). + * Unix domain sockets ensures that the connection is reliable. The + * default timeout of 30mins used for unreliable network connections is + * too long for unix domain socket connections. + */ + ret = rpc_transport_unix_options_build (&options, sockfile, 600); if (ret) { - runner_log (&runner, "glusterd", GF_LOG_DEBUG, "command failed"); - runner_end (&runner); + gf_log (THIS->name, GF_LOG_ERROR, "Unix options build failed"); goto out; } - runner_end (&runner); - - volinfo->defrag_status = GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED; - if ((cmd == GF_DEFRAG_CMD_START_MIGRATE_DATA) || - (cmd == GF_DEFRAG_CMD_START_MIGRATE_DATA_FORCE)) { - volinfo->defrag_status = GF_DEFRAG_STATUS_MIGRATE_DATA_STARTED; - } - if (cbk) - defrag->cbk_fn = cbk; - - ret = pthread_create (&defrag->th, NULL, glusterd_defrag_start, - volinfo); + synclock_unlock (&priv->big_lock); + ret = glusterd_rpc_create (&defrag->rpc, options, + glusterd_defrag_notify, volinfo); + synclock_lock (&priv->big_lock); if (ret) { - runinit (&runner); - runner_add_args (&runner, "umount", "-l", defrag->mount, NULL); - ret = runner_run_reuse (&runner); - if (ret) - runner_log (&runner, "glusterd", GF_LOG_DEBUG, "command failed"); - runner_end (&runner); + gf_log (THIS->name, GF_LOG_ERROR, "RPC create failed"); + goto out; } + ret = 0; out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -658,26 +403,32 @@ out: } int -glusterd_handle_defrag_volume (rpcsvc_request_t *req) +__glusterd_handle_defrag_volume (rpcsvc_request_t *req) { - int32_t ret = -1; - gf_cli_req cli_req = {{0,}}; - glusterd_conf_t *priv = NULL; - dict_t *dict = NULL; - char *volname = NULL; - gf_cli_defrag_type cmd = 0; + int32_t ret = -1; + gf_cli_req cli_req = {{0,}}; + glusterd_conf_t *priv = NULL; + dict_t *dict = NULL; + char *volname = NULL; + gf_cli_defrag_type cmd = 0; + char msg[2048] = {0,}; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); - priv = THIS->private; + priv = this->private; + GF_ASSERT (priv); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; } - if (cli_req.dict.dict_len) { + + if (cli_req.dict.dict_len) { /* Unserialize the dictionary */ dict = dict_new (); @@ -685,34 +436,39 @@ glusterd_handle_defrag_volume (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, - "failed to " + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (msg, sizeof (msg), "Unable to decode the " + "command"); goto out; } } ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, - "Failed to get volname"); + snprintf (msg, sizeof (msg), "Failed to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); goto out; } ret = dict_get_int32 (dict, "rebalance-command", (int32_t*)&cmd); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, - "Failed to get command"); + snprintf (msg, sizeof (msg), "Failed to get command"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); goto out; } - glusterd_rebalance_cmd_attempted_log (cmd, volname); - - ret = dict_set_static_bin (dict, "node-uuid", priv->uuid, 16); + ret = dict_set_static_bin (dict, "node-uuid", MY_UUID, 16); if (ret) goto out; - ret = glusterd_op_begin (req, GD_OP_REBALANCE, dict); + if ((cmd == GF_DEFRAG_CMD_STATUS) || + (cmd == GF_DEFRAG_CMD_STOP)) { + ret = glusterd_op_begin (req, GD_OP_DEFRAG_BRICK_VOLUME, + dict, msg, sizeof (msg)); + } else + ret = glusterd_op_begin (req, GD_OP_REBALANCE, dict, + msg, sizeof (msg)); out: @@ -720,57 +476,100 @@ out: glusterd_op_sm (); if (ret) { - if (dict) - dict_unref (dict); - ret = glusterd_op_send_cli_response (GD_OP_REBALANCE, ret, 0, req, - NULL, "operation failed"); + if (msg[0] == '\0') + snprintf (msg, sizeof (msg), "Operation failed"); + ret = glusterd_op_send_cli_response (GD_OP_REBALANCE, ret, 0, + req, dict, msg); + } - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val);//malloced by xdr + free (cli_req.dict.dict_val);//malloced by xdr return 0; } +int +glusterd_handle_defrag_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, __glusterd_handle_defrag_volume); +} + int glusterd_op_stage_rebalance (dict_t *dict, char **op_errstr) { - char *volname = NULL; - int ret = 0; - int32_t cmd = 0; - char msg[2048] = {0}; - glusterd_volinfo_t *volinfo = NULL; + char *volname = NULL; + int ret = 0; + int32_t cmd = 0; + char msg[2048] = {0}; + glusterd_volinfo_t *volinfo = NULL; + char *task_id_str = NULL; + dict_t *op_ctx = NULL; + xlator_t *this = 0; + + this = THIS; + GF_ASSERT (this); ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, "volname not found"); + gf_log (this->name, GF_LOG_DEBUG, "volname not found"); goto out; } + ret = dict_get_int32 (dict, "rebalance-command", &cmd); if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, "cmd not found"); + gf_log (this->name, GF_LOG_DEBUG, "cmd not found"); goto out; } ret = glusterd_rebalance_cmd_validate (cmd, volname, &volinfo, msg, sizeof (msg)); if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, "failed to validate"); + gf_log (this->name, GF_LOG_DEBUG, "failed to validate"); goto out; } switch (cmd) { case GF_DEFRAG_CMD_START: case GF_DEFRAG_CMD_START_LAYOUT_FIX: - case GF_DEFRAG_CMD_START_MIGRATE_DATA: - case GF_DEFRAG_CMD_START_MIGRATE_DATA_FORCE: - ret = glusterd_defrag_start_validate (volinfo, - msg, sizeof (msg)); + case GF_DEFRAG_CMD_START_FORCE: + if (is_origin_glusterd (dict)) { + op_ctx = glusterd_op_get_ctx (); + if (!op_ctx) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "Failed to get op_ctx"); + goto out; + } + + ret = glusterd_generate_and_set_task_id + (op_ctx, GF_REBALANCE_TID_KEY); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to generate task-id"); + goto out; + } + } else { + ret = dict_get_str (dict, GF_REBALANCE_TID_KEY, + &task_id_str); + if (ret) { + snprintf (msg, sizeof (msg), + "Missing rebalance-id"); + gf_log (this->name, GF_LOG_WARNING, "%s", msg); + ret = 0; + } + } + ret = glusterd_defrag_start_validate (volinfo, msg, + sizeof (msg), + GD_OP_REBALANCE); if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, - "start validate failed"); + gf_log (this->name, GF_LOG_DEBUG, + "start validate failed"); goto out; } + break; + case GF_DEFRAG_CMD_STATUS: + case GF_DEFRAG_CMD_STOP: + break; default: break; } @@ -787,101 +586,164 @@ out: int glusterd_op_rebalance (dict_t *dict, char **op_errstr, dict_t *rsp_dict) { - char *volname = NULL; - int ret = 0; - int32_t cmd = 0; - char msg[2048] = {0}; - glusterd_volinfo_t *volinfo = NULL; - uint64_t files = 0; - uint64_t size = 0; - void *node_uuid = NULL; - glusterd_conf_t *priv = NULL; - dict_t *tmp_dict = NULL; - - priv = THIS->private; + char *volname = NULL; + int ret = 0; + int32_t cmd = 0; + char msg[2048] = {0}; + glusterd_volinfo_t *volinfo = NULL; + glusterd_conf_t *priv = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_brickinfo_t *tmp = NULL; + gf_boolean_t volfile_update = _gf_false; + char *task_id_str = NULL; + dict_t *ctx = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, "volname not given"); + gf_log (this->name, GF_LOG_DEBUG, "volname not given"); goto out; } ret = dict_get_int32 (dict, "rebalance-command", &cmd); if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, "command not given"); + gf_log (this->name, GF_LOG_DEBUG, "command not given"); goto out; } + ret = glusterd_rebalance_cmd_validate (cmd, volname, &volinfo, msg, sizeof (msg)); if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, "cmd validate failed"); + gf_log (this->name, GF_LOG_DEBUG, "cmd validate failed"); goto out; } - if ((cmd != GF_DEFRAG_CMD_STATUS) && - (cmd != GF_DEFRAG_CMD_STOP)) { - ret = dict_get_ptr (dict, "node-uuid", &node_uuid); - if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, "node-uuid not found"); - goto out; - } - - /* perform this on only the node which has - issued the command */ - if (uuid_compare (node_uuid, priv->uuid)) { - gf_log (THIS->name, GF_LOG_DEBUG, - "not the source node %s", uuid_utoa (priv->uuid)); - goto out; + /* Set task-id, if available, in op_ctx dict for operations other than + * start + */ + if (cmd == GF_DEFRAG_CMD_STATUS || cmd == GF_DEFRAG_CMD_STOP) { + if (!uuid_is_null (volinfo->rebal.rebalance_id)) { + ctx = glusterd_op_get_ctx (); + if (!ctx) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get op_ctx"); + ret = -1; + goto out; + } + + if (GD_OP_REMOVE_BRICK == volinfo->rebal.op) + ret = glusterd_copy_uuid_to_dict + (volinfo->rebal.rebalance_id, ctx, + GF_REMOVE_BRICK_TID_KEY); + else + ret = glusterd_copy_uuid_to_dict + (volinfo->rebal.rebalance_id, ctx, + GF_REBALANCE_TID_KEY); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set task-id"); + goto out; + } } } switch (cmd) { case GF_DEFRAG_CMD_START: case GF_DEFRAG_CMD_START_LAYOUT_FIX: - case GF_DEFRAG_CMD_START_MIGRATE_DATA: - case GF_DEFRAG_CMD_START_MIGRATE_DATA_FORCE: + case GF_DEFRAG_CMD_START_FORCE: + ret = dict_get_str (dict, GF_REBALANCE_TID_KEY, &task_id_str); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Missing rebalance " + "id"); + ret = 0; + } else { + uuid_parse (task_id_str, volinfo->rebal.rebalance_id) ; + volinfo->rebal.op = GD_OP_REBALANCE; + } ret = glusterd_handle_defrag_start (volinfo, msg, sizeof (msg), - cmd, NULL); + cmd, NULL, GD_OP_REBALANCE); break; - case GF_DEFRAG_CMD_STOP: - ret = glusterd_defrag_stop (volinfo, &files, &size, - msg, sizeof (msg)); - if (!ret && rsp_dict) { - ret = dict_set_uint64 (rsp_dict, "files", files); - if (ret) - gf_log (THIS->name, GF_LOG_WARNING, - "failed to set file count"); - - ret = dict_set_uint64 (rsp_dict, "size", size); - if (ret) - gf_log (THIS->name, GF_LOG_WARNING, - "failed to set xfer size"); - - /* Don't want to propagate errors from dict_set() */ + case GF_DEFRAG_CMD_STOP: + /* Clear task-id only on explicitly stopping rebalance. + * Also clear the stored operation, so it doesn't cause trouble + * with future rebalance/remove-brick starts + */ + uuid_clear (volinfo->rebal.rebalance_id); + volinfo->rebal.op = GD_OP_NONE; + + /* Fall back to the old volume file in case of decommission*/ + list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks, + brick_list) { + if (!brickinfo->decommissioned) + continue; + brickinfo->decommissioned = 0; + volfile_update = _gf_true; + } + + if (volfile_update == _gf_false) { ret = 0; + break; } - break; - case GF_DEFRAG_CMD_STATUS: - if (rsp_dict) - tmp_dict = rsp_dict; + ret = glusterd_create_volfiles_and_notify_services (volinfo); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to create volfiles"); + goto out; + } + + ret = glusterd_store_volinfo (volinfo, + GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to store volinfo"); + goto out; + } - /* On source node, there will be no 'rsp_dict' */ - if (!tmp_dict) - tmp_dict = glusterd_op_get_ctx (GD_OP_REBALANCE); + ret = 0; + break; - ret = glusterd_defrag_status_get (volinfo, tmp_dict); + case GF_DEFRAG_CMD_STATUS: break; default: break; } - glusterd_rebalance_cmd_log (cmd, volname, ret); - out: if (ret && op_errstr && msg[0]) *op_errstr = gf_strdup (msg); return ret; } + +int32_t +glusterd_defrag_event_notify_handle (dict_t *dict) +{ + glusterd_volinfo_t *volinfo = NULL; + char *volname = NULL; + int32_t ret = -1; + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Failed to get volname"); + return ret; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Failed to get volinfo for %s" + , volname); + return ret; + } + + ret = glusterd_defrag_volume_status_update (volinfo, dict); + + if (ret) + gf_log ("", GF_LOG_ERROR, "Failed to update status"); + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-replace-brick.c b/xlators/mgmt/glusterd/src/glusterd-replace-brick.c index 92ae963a1..54b830870 100644 --- a/xlators/mgmt/glusterd/src/glusterd-replace-brick.c +++ b/xlators/mgmt/glusterd/src/glusterd-replace-brick.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" @@ -36,6 +26,13 @@ #include <signal.h> +#define GLUSTERD_GET_RB_MNTPT(path, len, volinfo) \ + snprintf (path, len, \ + DEFAULT_VAR_RUN_DIRECTORY"/%s-"RB_CLIENT_MOUNTPOINT, \ + volinfo->volname); + +extern uuid_t global_txn_id; + int glusterd_get_replace_op_str (gf1_cli_replace_op op, char *op_str) { @@ -74,7 +71,7 @@ out: } int -glusterd_handle_replace_brick (rpcsvc_request_t *req) +__glusterd_handle_replace_brick (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; @@ -85,17 +82,21 @@ glusterd_handle_replace_brick (rpcsvc_request_t *req) char operation[256]; glusterd_op_t cli_op = GD_OP_REPLACE_BRICK; char *volname = NULL; + char msg[2048] = {0,}; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; + GF_ASSERT (this); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; } - gf_log ("glusterd", GF_LOG_INFO, "Received replace brick req"); + gf_log (this->name, GF_LOG_INFO, "Received replace brick req"); if (cli_req.dict.dict_len) { /* Unserialize the dictionary */ @@ -105,127 +106,77 @@ glusterd_handle_replace_brick (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (msg, sizeof (msg), "Unable to decode the " + "command"); goto out; } } ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "could not get volname"); + snprintf (msg, sizeof (msg), "Could not get volume name"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); goto out; } ret = dict_get_int32 (dict, "operation", &op); if (ret) { - gf_log ("", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "dict_get on operation failed"); + snprintf (msg, sizeof (msg), "Could not get operation"); goto out; } ret = dict_get_str (dict, "src-brick", &src_brick); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get src brick"); + snprintf (msg, sizeof (msg), "Failed to get src brick"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); goto out; } - gf_log ("", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "src brick=%s", src_brick); ret = dict_get_str (dict, "dst-brick", &dst_brick); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get dest brick"); + snprintf (msg, sizeof (msg), "Failed to get dest brick"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); goto out; } (void) glusterd_get_replace_op_str (op, operation); - gf_log ("", GF_LOG_DEBUG, "dst brick=%s", dst_brick); - gf_log ("glusterd", GF_LOG_INFO, "Received replace brick %s request", + gf_log (this->name, GF_LOG_DEBUG, "dst brick=%s", dst_brick); + gf_log (this->name, GF_LOG_INFO, "Received replace brick %s request", operation); - gf_cmd_log ("Volume replace-brick","volname: %s src_brick:%s" - " dst_brick:%s op:%s", volname, src_brick, dst_brick, - operation); - ret = glusterd_op_begin (req, GD_OP_REPLACE_BRICK, dict); - gf_cmd_log ("Volume replace-brick","on volname: %s %s", volname, - (ret) ? "FAILED" : "SUCCESS"); + ret = glusterd_op_begin (req, GD_OP_REPLACE_BRICK, dict, + msg, sizeof (msg)); out: - if (ret && dict) - dict_unref (dict); - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val);//malloced by xdr + free (cli_req.dict.dict_val);//malloced by xdr glusterd_friend_sm (); glusterd_op_sm (); - if (ret) + if (ret) { + if (msg[0] == '\0') + snprintf (msg, sizeof (msg), "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, msg); + } return ret; } - -char * -glusterd_check_brick_rb_part (char *bricks, int count, glusterd_volinfo_t *volinfo) +int +glusterd_handle_replace_brick (rpcsvc_request_t *req) { - char *saveptr = NULL; - char *brick = NULL; - char *brick_list = NULL; - int ret = 0; - glusterd_brickinfo_t *brickinfo = NULL; - uint32_t i = 0; - char *str = NULL; - char msg[2048] = {0,}; - - brick_list = gf_strdup (bricks); - if (!brick_list) { - gf_log ("glusterd", GF_LOG_ERROR, - "Out of memory"); - ret = -1; - goto out; - } - - if (count) - brick = strtok_r (brick_list+1, " \n", &saveptr); - - - while ( i < count) { - ret = glusterd_brickinfo_from_brick (brick, &brickinfo); - if (ret) { - snprintf (msg, sizeof(msg), "Unable to" - " get brickinfo"); - gf_log ("", GF_LOG_ERROR, "%s", msg); - ret = -1; - goto out; - } - - if (glusterd_is_replace_running (volinfo, brickinfo)) { - snprintf (msg, sizeof(msg), "Volume %s: replace brick is running" - " and the brick %s:%s you are trying to add is the destination brick" - " for replace brick", volinfo->volname, brickinfo->hostname, brickinfo->path); - ret = -1; - goto out; - } - - glusterd_brickinfo_delete (brickinfo); - brickinfo = NULL; - brick = strtok_r (NULL, " \n", &saveptr); - i++; - } - -out: - if (brick_list) - GF_FREE(brick_list); - if (brickinfo) - glusterd_brickinfo_delete (brickinfo); - if (ret) - str = gf_strdup (msg); - return str; + return glusterd_big_locked_handler (req, + __glusterd_handle_replace_brick); } static int @@ -237,7 +188,7 @@ glusterd_get_rb_dst_brickinfo (glusterd_volinfo_t *volinfo, if (!volinfo || !brickinfo) goto out; - *brickinfo = volinfo->dst_brick; + *brickinfo = volinfo->rep_brick.dst_brick; ret = 0; @@ -250,6 +201,7 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, dict_t *rsp_dict) { int ret = 0; + int32_t port = 0; char *src_brick = NULL; char *dst_brick = NULL; char *volname = NULL; @@ -258,47 +210,53 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, glusterd_brickinfo_t *src_brickinfo = NULL; char *host = NULL; char *path = NULL; - char msg[2048] = {0}; + char msg[2048] = {0}; char *dup_dstbrick = NULL; glusterd_peerinfo_t *peerinfo = NULL; glusterd_brickinfo_t *dst_brickinfo = NULL; - gf_boolean_t is_run = _gf_false; + gf_boolean_t is_run = _gf_false; dict_t *ctx = NULL; glusterd_conf_t *priv = NULL; - char voldir[PATH_MAX] = {0}; - char pidfile[PATH_MAX] = {0}; + char *savetok = NULL; + char pidfile[PATH_MAX] = {0}; + char *task_id_str = NULL; + xlator_t *this = NULL; + gf_boolean_t is_force = _gf_false; - priv = THIS->private; + this = THIS; + GF_ASSERT (this); + + priv = this->private; GF_ASSERT (priv); ret = dict_get_str (dict, "src-brick", &src_brick); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get src brick"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get src brick"); goto out; } - gf_log ("", GF_LOG_DEBUG, "src brick=%s", src_brick); + gf_log (this->name, GF_LOG_DEBUG, "src brick=%s", src_brick); ret = dict_get_str (dict, "dst-brick", &dst_brick); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get dest brick"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get dest brick"); goto out; } - gf_log ("", GF_LOG_DEBUG, "dst brick=%s", dst_brick); + gf_log (this->name, GF_LOG_DEBUG, "dst brick=%s", dst_brick); ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } ret = dict_get_int32 (dict, "operation", (int32_t *)&replace_op); if (ret) { - gf_log ("", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "dict get on replace-brick operation failed"); goto out; } @@ -323,7 +281,7 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, !glusterd_is_valid_volfpath (volname, dst_brick)) { snprintf (msg, sizeof (msg), "brick path %s is too " "long.", dst_brick); - gf_log ("", GF_LOG_ERROR, "%s", msg); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); *op_errstr = gf_strdup (msg); ret = -1; @@ -332,10 +290,10 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, ret = glusterd_check_gsync_running (volinfo, &is_run); if (ret && (is_run == _gf_false)) - gf_log ("", GF_LOG_WARNING, "Unable to get the status" + gf_log (this->name, GF_LOG_WARNING, "Unable to get the status" " of active "GEOREP" session"); if (is_run) { - gf_log ("", GF_LOG_WARNING, GEOREP" sessions active" + gf_log (this->name, GF_LOG_WARNING, GEOREP" sessions active" "for the volume %s ", volname); snprintf (msg, sizeof(msg), GEOREP" sessions are active " "for the volume %s.\nStop "GEOREP " sessions " @@ -350,29 +308,61 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, if (glusterd_is_defrag_on(volinfo)) { snprintf (msg, sizeof(msg), "Volume name %s rebalance is in " "progress. Please retry after completion", volname); - gf_log ("glusterd", GF_LOG_ERROR, "%s", msg); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); *op_errstr = gf_strdup (msg); ret = -1; goto out; } + ctx = glusterd_op_get_ctx(); + switch (replace_op) { case GF_REPLACE_OP_START: if (glusterd_is_rb_started (volinfo)) { - gf_log ("", GF_LOG_ERROR, "Replace brick is already " - "started for volume "); + snprintf (msg, sizeof (msg), "Replace brick is already " + "started for volume"); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); ret = -1; goto out; } + if (is_origin_glusterd (dict)) { + if (!ctx) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "Failed to get op_ctx"); + goto out; + } + + ret = glusterd_generate_and_set_task_id + (ctx, GF_REPLACE_BRICK_TID_KEY); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to generate task-id"); + goto out; + } + + } else { + ret = dict_get_str (dict, GF_REPLACE_BRICK_TID_KEY, + &task_id_str); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Missing replace-brick-id"); + ret = 0; + } + } + is_force = dict_get_str_boolean (dict, "force", _gf_false); + break; + case GF_REPLACE_OP_PAUSE: if (glusterd_is_rb_paused (volinfo)) { - gf_log ("", GF_LOG_ERROR, "Replace brick is already " - "paused for volume "); + gf_log (this->name, GF_LOG_ERROR, "Replace brick is " + "already paused for volume "); ret = -1; goto out; } else if (!glusterd_is_rb_started(volinfo)) { - gf_log ("", GF_LOG_ERROR, "Replace brick is not" + gf_log (this->name, GF_LOG_ERROR, "Replace brick is not" " started for volume "); ret = -1; goto out; @@ -380,9 +370,8 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, break; case GF_REPLACE_OP_ABORT: - if ((!glusterd_is_rb_paused (volinfo)) && - (!glusterd_is_rb_started (volinfo))) { - gf_log ("", GF_LOG_ERROR, "Replace brick is not" + if (!glusterd_is_rb_ongoing (volinfo)) { + gf_log (this->name, GF_LOG_ERROR, "Replace brick is not" " started or paused for volume "); ret = -1; goto out; @@ -390,15 +379,17 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, break; case GF_REPLACE_OP_COMMIT: - if (!glusterd_is_rb_started (volinfo)) { - gf_log ("", GF_LOG_ERROR, "Replace brick is not " + if (!glusterd_is_rb_ongoing (volinfo)) { + gf_log (this->name, GF_LOG_ERROR, "Replace brick is not " "started for volume "); ret = -1; goto out; } break; - case GF_REPLACE_OP_COMMIT_FORCE: break; + case GF_REPLACE_OP_COMMIT_FORCE: + is_force = _gf_true; + break; case GF_REPLACE_OP_STATUS: @@ -411,7 +402,7 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, goto out; } - gf_log (THIS->name, GF_LOG_ERROR, "%s", *op_errstr); + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); ret = -1; goto out; } @@ -423,8 +414,7 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, } ret = glusterd_volume_brickinfo_get_by_brick (src_brick, volinfo, - &src_brickinfo, - GF_PATH_COMPLETE); + &src_brickinfo); if (ret) { snprintf (msg, sizeof (msg), "brick: %s does not exist in " "volume: %s", src_brick, volname); @@ -432,10 +422,9 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, goto out; } - ctx = glusterd_op_get_ctx(); if (ctx) { if (!glusterd_is_fuse_available ()) { - gf_log ("glusterd", GF_LOG_ERROR, "Unable to open /dev/" + gf_log (this->name, GF_LOG_ERROR, "Unable to open /dev/" "fuse (%s), replace-brick command failed", strerror (errno)); snprintf (msg, sizeof(msg), "Fuse unavailable\n " @@ -446,8 +435,8 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, } } - if (!glusterd_is_local_addr (src_brickinfo->hostname)) { - gf_log ("", GF_LOG_DEBUG, + if (gf_is_local_addr (src_brickinfo->hostname)) { + gf_log (this->name, GF_LOG_DEBUG, "I AM THE SOURCE HOST"); if (src_brickinfo->port && rsp_dict) { ret = dict_set_int32 (rsp_dict, "src-brick-port", @@ -459,11 +448,10 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, } } - GLUSTERD_GET_VOLUME_DIR (voldir, volinfo, priv); - GLUSTERD_GET_BRICK_PIDFILE (pidfile, voldir, - src_brickinfo->hostname, - src_brickinfo->path); - if (!glusterd_is_service_running (pidfile, NULL)) { + GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, src_brickinfo, + priv); + if ((replace_op != GF_REPLACE_OP_COMMIT_FORCE) && + !glusterd_is_service_running (pidfile, NULL)) { snprintf(msg, sizeof(msg), "Source brick %s:%s " "is not online.", src_brickinfo->hostname, src_brickinfo->path); @@ -478,53 +466,61 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, dup_dstbrick = gf_strdup (dst_brick); if (!dup_dstbrick) { ret = -1; - gf_log ("", GF_LOG_ERROR, "Memory allocation failed"); + gf_log (this->name, GF_LOG_ERROR, "Memory allocation failed"); goto out; } - host = strtok (dup_dstbrick, ":"); - path = strtok (NULL, ":"); + host = strtok_r (dup_dstbrick, ":", &savetok); + path = strtok_r (NULL, ":", &savetok); if (!host || !path) { - gf_log ("", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "dst brick %s is not of form <HOSTNAME>:<export-dir>", dst_brick); ret = -1; goto out; } - if (!glusterd_brickinfo_get (NULL, host, path, NULL)) { - snprintf(msg, sizeof(msg), "Brick: %s:%s already in use", - host, path); + + ret = glusterd_brickinfo_new_from_brick (dst_brick, &dst_brickinfo); + if (ret) + goto out; + + ret = glusterd_new_brick_validate (dst_brick, dst_brickinfo, + msg, sizeof (msg)); + if (ret) { *op_errstr = gf_strdup (msg); ret = -1; + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); goto out; } - if ((volinfo->rb_status ==GF_RB_STATUS_NONE) && - (replace_op == GF_REPLACE_OP_START)) { - ret = glusterd_brickinfo_from_brick (dst_brick, &dst_brickinfo); - volinfo->src_brick = src_brickinfo; - volinfo->dst_brick = dst_brickinfo; - } else { - ret = glusterd_get_rb_dst_brickinfo (volinfo, &dst_brickinfo); + if (!glusterd_is_rb_ongoing (volinfo) && + (replace_op == GF_REPLACE_OP_START || + replace_op == GF_REPLACE_OP_COMMIT_FORCE)) { + + volinfo->rep_brick.src_brick = src_brickinfo; + volinfo->rep_brick.dst_brick = dst_brickinfo; } if (glusterd_rb_check_bricks (volinfo, src_brickinfo, dst_brickinfo)) { ret = -1; - *op_errstr = gf_strdup ("incorrect source or " + *op_errstr = gf_strdup ("Incorrect source or " "destination brick"); if (*op_errstr) - gf_log (THIS->name, GF_LOG_ERROR, "%s", *op_errstr); + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); goto out; } - if (!glusterd_is_local_addr (host)) { - ret = glusterd_brick_create_path (host, path, - volinfo->volume_id, 0777, - op_errstr); + if (!glusterd_is_rb_ongoing (volinfo) && + gf_is_local_addr (host)) { + ret = glusterd_validate_and_create_brickpath (dst_brickinfo, + volinfo->volume_id, + op_errstr, is_force); if (ret) goto out; - } else { + } + + if (!gf_is_local_addr (host)) { ret = glusterd_friend_find (NULL, host, &peerinfo); if (ret) { snprintf (msg, sizeof (msg), "%s, is not a friend", @@ -549,13 +545,72 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, goto out; } } + + if (replace_op == GF_REPLACE_OP_START && + gf_is_local_addr (volinfo->rep_brick.dst_brick->hostname)) { + port = pmap_registry_alloc (THIS); + if (!port) { + gf_log (THIS->name, GF_LOG_CRITICAL, + "No free ports available"); + ret = -1; + goto out; + } + + ctx = glusterd_op_get_ctx(); + ret = dict_set_int32 ((ctx)?ctx:rsp_dict, "dst-brick-port", + port); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to set dst " + "brick port"); + goto out; + } + volinfo->rep_brick.dst_brick->port = port; + } + ret = 0; out: - if (dup_dstbrick) - GF_FREE (dup_dstbrick); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + GF_FREE (dup_dstbrick); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + + return ret; +} + +static int +rb_set_mntfd (int mntfd) +{ + int ret = -1; + dict_t *ctx = NULL; + + ctx = glusterd_op_get_ctx (); + if (!ctx) { + gf_log (THIS->name, GF_LOG_CRITICAL, "Failed to get op ctx"); + goto out; + } + ret = dict_set_int32 (ctx, "mntfd", mntfd); + if (ret) + gf_log (THIS->name, GF_LOG_DEBUG, "Failed to set mnt fd " + "in op ctx"); +out: + return ret; +} +static int +rb_get_mntfd (int *mntfd) +{ + int ret = -1; + dict_t *ctx = NULL; + + ctx = glusterd_op_get_ctx (); + if (!ctx) { + gf_log (THIS->name, GF_LOG_CRITICAL, "Failed to get op ctx"); + goto out; + } + ret = dict_get_int32 (ctx, "mntfd", mntfd); + if (ret) + gf_log (THIS->name, GF_LOG_DEBUG, "Failed to get mnt fd " + "from op ctx"); +out: return ret; } @@ -581,6 +636,8 @@ rb_regenerate_volfiles (glusterd_volinfo_t *volinfo, ret = glusterd_create_rb_volfiles (volinfo, brickinfo); + dict_del (dict, "enable-pump"); + out: return ret; } @@ -602,7 +659,8 @@ rb_src_brick_restart (glusterd_volinfo_t *volinfo, ret); } - ret = glusterd_volume_stop_glusterfs (volinfo, src_brickinfo); + ret = glusterd_volume_stop_glusterfs (volinfo, src_brickinfo, + _gf_false); if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to stop " "glusterfs, ret: %d", ret); @@ -629,7 +687,8 @@ rb_src_brick_restart (glusterd_volinfo_t *volinfo, } sleep (2); - ret = glusterd_volume_start_glusterfs (volinfo, src_brickinfo); + ret = glusterd_volume_start_glusterfs (volinfo, src_brickinfo, + _gf_false); if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to start " "glusterfs, ret: %d", ret); @@ -651,22 +710,19 @@ rb_send_xattr_command (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *dst_brickinfo, const char *xattr_key, const char *value) { - glusterd_conf_t *priv = NULL; - xlator_t *this = NULL; - char mount_point_path[PATH_MAX] = {0,}; - int ret = -1; - - this = THIS; - priv = this->private; + int ret = -1; + int mntfd = -1; - snprintf (mount_point_path, PATH_MAX, "%s/vols/%s/%s", - priv->workdir, volinfo->volname, RB_CLIENT_MOUNTPOINT); + ret = rb_get_mntfd (&mntfd); + if (ret) + goto out; - ret = sys_lsetxattr (mount_point_path, xattr_key, value, - strlen (value) + 1, 0); + ret = sys_fsetxattr (mntfd, xattr_key, value, strlen (value) + 1, 0); if (ret) - gf_log (this->name, GF_LOG_DEBUG, "setxattr on key: " - "%s failed", xattr_key); + gf_log (THIS->name, GF_LOG_DEBUG, "setxattr on key: " + "%s, reason: %s", xattr_key, strerror (errno)); + +out: return ret; } @@ -681,9 +737,7 @@ rb_spawn_dst_brick (glusterd_volinfo_t *volinfo, priv = THIS->private; - port = pmap_registry_alloc (THIS); - brickinfo->port = port; - + port = brickinfo->port; GF_ASSERT (port); runinit (&runner); @@ -694,9 +748,13 @@ rb_spawn_dst_brick (glusterd_volinfo_t *volinfo, priv->workdir, volinfo->volname); runner_add_arg (&runner, "--xlator-option"); runner_argprintf (&runner, "src-server.listen-port=%d", port); + if (volinfo->memory_accounting) + runner_add_arg (&runner, "--mem-accounting"); - ret = runner_run (&runner); + ret = runner_run_nowait (&runner); if (ret) { + pmap_registry_remove (THIS, 0, brickinfo->path, + GF_PMAP_PORT_BRICKSERVER, NULL); gf_log ("", GF_LOG_DEBUG, "Could not start glusterfs"); goto out; @@ -716,51 +774,70 @@ static int rb_spawn_glusterfs_client (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo) { - glusterd_conf_t *priv = NULL; - char cmd_str[8192] = {0,}; - runner_t runner = {0,}; - struct stat buf; - int ret = -1; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + runner_t runner = {0,}; + struct stat buf = {0,}; + char mntpt[PATH_MAX] = {0,}; + int mntfd = -1; + int ret = -1; - priv = THIS->private; + this = THIS; + priv = this->private; + GLUSTERD_GET_RB_MNTPT (mntpt, sizeof (mntpt), volinfo); runinit (&runner); runner_add_arg (&runner, SBIN_DIR"/glusterfs"); runner_argprintf (&runner, "-f" "%s/vols/%s/"RB_CLIENTVOL_FILENAME, priv->workdir, volinfo->volname); - runner_argprintf (&runner, "%s/vols/%s/"RB_CLIENT_MOUNTPOINT, - priv->workdir, volinfo->volname); + runner_add_arg (&runner, mntpt); + if (volinfo->memory_accounting) + runner_add_arg (&runner, "--mem-accounting"); - ret = runner_run (&runner); + ret = runner_run_reuse (&runner); if (ret) { - gf_log ("", GF_LOG_DEBUG, - "Could not start glusterfs"); + runner_log (&runner, this->name, GF_LOG_DEBUG, + "Could not start glusterfs"); + runner_end (&runner); goto out; + } else { + runner_log (&runner, this->name, GF_LOG_DEBUG, + "Successfully started glusterfs"); + runner_end (&runner); } - gf_log ("", GF_LOG_DEBUG, - "Successfully started glusterfs: brick=%s:%s", - brickinfo->hostname, brickinfo->path); + ret = stat (mntpt, &buf); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "stat on mount point %s " + "failed", mntpt); + goto out; + } - memset (cmd_str, 0, sizeof (cmd_str)); + mntfd = open (mntpt, O_DIRECTORY); + if (mntfd == -1) + goto out; - snprintf (cmd_str, 4096, "%s/vols/%s/%s", - priv->workdir, volinfo->volname, - RB_CLIENT_MOUNTPOINT); + ret = rb_set_mntfd (mntfd); + if (ret) + goto out; - ret = stat (cmd_str, &buf); + runinit (&runner); + runner_add_args (&runner, "/bin/umount", "-l", mntpt, NULL); + ret = runner_run_reuse (&runner); if (ret) { - gf_log ("", GF_LOG_DEBUG, - "stat on mountpoint failed"); - goto out; - } - - gf_log ("", GF_LOG_DEBUG, - "stat on mountpoint succeeded"); + runner_log (&runner, this->name, GF_LOG_DEBUG, + "Lazy unmount failed on maintenance client"); + runner_end (&runner); + goto out; + } else { + runner_log (&runner, this->name, GF_LOG_DEBUG, + "Successfully unmounted maintenance client"); + runner_end (&runner); + } - ret = 0; out: + return ret; } @@ -770,6 +847,8 @@ static const char *client_volfile_str = "volume mnt-client\n" " option remote-subvolume %s\n" " option remote-port %d\n" " option transport-type %s\n" + " option username %s\n" + " option password %s\n" "end-volume\n" "volume mnt-wb\n" " type performance/write-behind\n" @@ -781,23 +860,33 @@ rb_generate_client_volfile (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *src_brickinfo) { glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; FILE *file = NULL; char filename[PATH_MAX] = {0, }; int ret = -1; + int fd = -1; char *ttype = NULL; - priv = THIS->private; + this = THIS; + priv = this->private; - gf_log ("", GF_LOG_DEBUG, - "Creating volfile"); + gf_log (this->name, GF_LOG_DEBUG, "Creating volfile"); snprintf (filename, PATH_MAX, "%s/vols/%s/%s", priv->workdir, volinfo->volname, RB_CLIENTVOL_FILENAME); + fd = open (filename, O_CREAT | O_RDONLY, S_IRUSR | S_IWUSR); + if (fd < 0) { + gf_log (this->name, GF_LOG_ERROR, + "%s", strerror (errno)); + goto out; + } + close (fd); + file = fopen (filename, "w+"); if (!file) { - gf_log ("", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "Open of volfile failed"); ret = -1; goto out; @@ -812,7 +901,10 @@ rb_generate_client_volfile (glusterd_volinfo_t *volinfo, } fprintf (file, client_volfile_str, src_brickinfo->hostname, - src_brickinfo->path, src_brickinfo->port, ttype); + src_brickinfo->path, + src_brickinfo->port, ttype, + glusterd_auth_get_username (volinfo), + glusterd_auth_get_password (volinfo)); fclose (file); GF_FREE (ttype); @@ -826,6 +918,7 @@ out: static const char *dst_brick_volfile_str = "volume src-posix\n" " type storage/posix\n" " option directory %s\n" + " option volume-id %s\n" "end-volume\n" "volume %s\n" " type features/locks\n" @@ -833,6 +926,8 @@ static const char *dst_brick_volfile_str = "volume src-posix\n" "end-volume\n" "volume src-server\n" " type protocol/server\n" + " option auth.login.%s.allow %s\n" + " option auth.login.%s.password %s\n" " option auth.addr.%s.allow *\n" " option transport-type %s\n" " subvolumes %s\n" @@ -843,23 +938,34 @@ rb_generate_dst_brick_volfile (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *dst_brickinfo) { glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; FILE *file = NULL; char filename[PATH_MAX] = {0, }; int ret = -1; + int fd = -1; char *trans_type = NULL; - priv = THIS->private; + this = THIS; + priv = this->private; - gf_log ("", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "Creating volfile"); snprintf (filename, PATH_MAX, "%s/vols/%s/%s", priv->workdir, volinfo->volname, RB_DSTBRICKVOL_FILENAME); + fd = creat (filename, S_IRUSR | S_IWUSR); + if (fd < 0) { + gf_log (this->name, GF_LOG_ERROR, + "%s", strerror (errno)); + goto out; + } + close (fd); + file = fopen (filename, "w+"); if (!file) { - gf_log ("", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "Open of volfile failed"); ret = -1; goto out; @@ -871,9 +977,17 @@ rb_generate_dst_brick_volfile (glusterd_volinfo_t *volinfo, goto out; } - fprintf (file, dst_brick_volfile_str, dst_brickinfo->path, - dst_brickinfo->path, dst_brickinfo->path, - trans_type, dst_brickinfo->path); + fprintf (file, dst_brick_volfile_str, + dst_brickinfo->path, + uuid_utoa (volinfo->volume_id), + dst_brickinfo->path, + dst_brickinfo->path, + glusterd_auth_get_username (volinfo), + glusterd_auth_get_username (volinfo), + glusterd_auth_get_password (volinfo), + dst_brickinfo->path, + trans_type, + dst_brickinfo->path); GF_FREE (trans_type); @@ -890,20 +1004,14 @@ static int rb_mountpoint_mkdir (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *src_brickinfo) { - glusterd_conf_t *priv = NULL; - char mount_point_path[PATH_MAX] = {0,}; - int ret = -1; - - priv = THIS->private; - - snprintf (mount_point_path, PATH_MAX, "%s/vols/%s/%s", - priv->workdir, volinfo->volname, - RB_CLIENT_MOUNTPOINT); + char mntpt[PATH_MAX] = {0,}; + int ret = -1; - ret = mkdir (mount_point_path, 0777); + GLUSTERD_GET_RB_MNTPT (mntpt, sizeof (mntpt), volinfo); + ret = mkdir (mntpt, 0777); if (ret && (errno != EEXIST)) { - gf_log ("", GF_LOG_DEBUG, "mkdir failed, errno: %d", - errno); + gf_log ("", GF_LOG_DEBUG, "mkdir failed, due to %s", + strerror (errno)); goto out; } @@ -917,19 +1025,14 @@ static int rb_mountpoint_rmdir (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *src_brickinfo) { - glusterd_conf_t *priv = NULL; - char mount_point_path[PATH_MAX] = {0,}; + char mntpt[PATH_MAX] = {0,}; int ret = -1; - priv = THIS->private; - - snprintf (mount_point_path, PATH_MAX, "%s/vols/%s/%s", - priv->workdir, volinfo->volname, - RB_CLIENT_MOUNTPOINT); - - ret = rmdir (mount_point_path); + GLUSTERD_GET_RB_MNTPT (mntpt, sizeof (mntpt), volinfo); + ret = rmdir (mntpt); if (ret) { - gf_log ("", GF_LOG_DEBUG, "rmdir failed"); + gf_log ("", GF_LOG_DEBUG, "rmdir failed, reason: %s", + strerror (errno)); goto out; } @@ -941,42 +1044,42 @@ out: static int rb_destroy_maintenance_client (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *src_brickinfo) + glusterd_brickinfo_t *src_brickinfo) { + xlator_t *this = NULL; glusterd_conf_t *priv = NULL; - runner_t runner = {0,}; - char filename[PATH_MAX] = {0,}; + char volfile[PATH_MAX] = {0,}; int ret = -1; + int mntfd = -1; - priv = THIS->private; + this = THIS; + priv = this->private; - runinit (&runner); - runner_add_args (&runner, "/bin/umount", "-f", NULL); - runner_argprintf (&runner, "%s/vols/%s/"RB_CLIENT_MOUNTPOINT, - priv->workdir, volinfo->volname); + ret = rb_get_mntfd (&mntfd); + if (ret) + goto out; - ret = runner_run (&runner); + ret = close (mntfd); if (ret) { - gf_log ("", GF_LOG_DEBUG, - "umount failed on maintenance client"); + gf_log (this->name, GF_LOG_DEBUG, "Failed to close mount " + "point directory"); goto out; } ret = rb_mountpoint_rmdir (volinfo, src_brickinfo); if (ret) { - gf_log ("", GF_LOG_DEBUG, - "rmdir of mountpoint failed"); + gf_log (this->name, GF_LOG_DEBUG, "rmdir of mountpoint " + "failed"); goto out; } - snprintf (filename, PATH_MAX, "%s/vols/%s/%s", - priv->workdir, volinfo->volname, - RB_CLIENTVOL_FILENAME); + snprintf (volfile, PATH_MAX, "%s/vols/%s/%s", priv->workdir, + volinfo->volname, RB_CLIENTVOL_FILENAME); - ret = unlink (filename); + ret = unlink (volfile); if (ret) { - gf_log ("", GF_LOG_DEBUG, - "unlink failed"); + gf_log ("", GF_LOG_DEBUG, "unlink of %s failed, reason: %s", + volfile, strerror (errno)); goto out; } @@ -1065,23 +1168,18 @@ rb_get_xattr_command (glusterd_volinfo_t *volinfo, const char *xattr_key, char *value) { - glusterd_conf_t *priv = NULL; - xlator_t *this = NULL; - char mount_point_path[PATH_MAX] = {0,}; - int ret = -1; + int ret = -1; + int mntfd = -1; - this = THIS; - priv = THIS->private; - - snprintf (mount_point_path, PATH_MAX, "%s/vols/%s/%s", - priv->workdir, volinfo->volname, - RB_CLIENT_MOUNTPOINT); + ret = rb_get_mntfd (&mntfd); + if (ret) + goto out; - ret = sys_lgetxattr (mount_point_path, xattr_key, value, 8192); + ret = sys_fgetxattr (mntfd, xattr_key, value, 8192); if (ret < 0) { - gf_log (this->name, GF_LOG_DEBUG, "getxattr on key: %s " - "failed", xattr_key); + gf_log (THIS->name, GF_LOG_DEBUG, "getxattr on key: %s " + "failed, reason: %s", xattr_key, strerror (errno)); goto out; } @@ -1096,11 +1194,17 @@ rb_send_cmd (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *dst, gf1_cli_replace_op op) { - char start_value[8192] = {0,}; - char status[8192] = {0,}; - char *status_reply = NULL; - dict_t *ctx = NULL; - int ret = 0; + char start_value[8192] = {0,}; + char status_str[8192] = {0,}; + char *status_reply = NULL; + char *tmp = NULL; + char *save_ptr = NULL; + char filename[PATH_MAX] = {0,}; + char *current_file = NULL; + uint64_t files = 0; + int status = 0; + dict_t *ctx = NULL; + int ret = 0; GF_ASSERT (volinfo); GF_ASSERT (src); @@ -1144,7 +1248,7 @@ rb_send_cmd (glusterd_volinfo_t *volinfo, { ret = rb_get_xattr_command (volinfo, src, dst, RB_PUMP_CMD_STATUS, - status); + status_str); if (ret) goto out; @@ -1157,7 +1261,76 @@ rb_send_cmd (glusterd_volinfo_t *volinfo, goto out; } - status_reply = gf_strdup (status); + /* Split status reply into different parts */ + tmp = strtok_r (status_str, ":", &save_ptr); + if (!tmp) { + ret = -1; + gf_log (THIS->name, GF_LOG_ERROR, + "Couldn't tokenize status string"); + goto out; + } + sscanf (tmp, "status=%d", &status); + ret = dict_set_int32 (ctx, "status", status); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Couldn't " + "set rb status in context"); + goto out; + } + + tmp = NULL; + tmp = strtok_r (NULL, ":", &save_ptr); + if (!tmp) { + ret = -1; + gf_log (THIS->name, GF_LOG_ERROR, + "Couldn't tokenize status string"); + goto out; + } + sscanf (tmp, "no_of_files=%"SCNu64, &files); + ret = dict_set_uint64 (ctx, "files", files); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Couldn't " + "set rb files in context"); + goto out; + } + + if (status == 0) { + tmp = NULL; + tmp = strtok_r (NULL, ":", &save_ptr); + if (!tmp) { + ret = -1; + gf_log (THIS->name, GF_LOG_ERROR, + "Couldn't tokenize status " + "string"); + goto out; + } + sscanf (tmp, "current_file=%s", filename); + current_file = gf_strdup (filename); + ret = dict_set_dynstr (ctx, "current_file", + current_file); + if (ret) { + GF_FREE (current_file); + gf_log (THIS->name, GF_LOG_ERROR, + "Couldn't set rb current file " + "in context"); + goto out; + } + } + if (status) { + ret = gf_asprintf (&status_reply, + "Number of files migrated = %" + PRIu64"\tMigration complete", + files); + } else { + ret = gf_asprintf (&status_reply, + "Number of files migrated = %" + PRIu64"\tCurrent file = %s", + files, filename); + } + if (ret == -1) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to create status_reply string"); + goto out; + } ret = dict_set_dynstr (ctx, "status-reply", status_reply); if (ret) { @@ -1231,14 +1404,11 @@ rb_update_srcbrick_port (glusterd_brickinfo_t *src_brickinfo, dict_t *rsp_dict, this = THIS; - ctx = glusterd_op_get_ctx (); - if (ctx) { - dict_ret = dict_get_int32 (req_dict, "src-brick-port", &src_port); - if (src_port) - src_brickinfo->port = src_port; - } + dict_ret = dict_get_int32 (req_dict, "src-brick-port", &src_port); + if (src_port) + src_brickinfo->port = src_port; - if (!glusterd_is_local_addr (src_brickinfo->hostname)) { + if (gf_is_local_addr (src_brickinfo->hostname)) { gf_log ("", GF_LOG_INFO, "adding src-brick port no"); @@ -1287,15 +1457,12 @@ rb_update_dstbrick_port (glusterd_brickinfo_t *dst_brickinfo, dict_t *rsp_dict, int dict_ret = 0; int dst_port = 0; - ctx = glusterd_op_get_ctx (); - if (ctx) { - dict_ret = dict_get_int32 (req_dict, "dst-brick-port", &dst_port); - if (dst_port) - dst_brickinfo->port = dst_port; + dict_ret = dict_get_int32 (req_dict, "dst-brick-port", &dst_port); + if (!dict_ret) + dst_brickinfo->port = dst_port; - } - if (!glusterd_is_local_addr (dst_brickinfo->hostname)) { + if (gf_is_local_addr (dst_brickinfo->hostname)) { gf_log ("", GF_LOG_INFO, "adding dst-brick port no"); @@ -1334,18 +1501,18 @@ glusterd_op_perform_replace_brick (glusterd_volinfo_t *volinfo, GF_ASSERT (volinfo); - ret = glusterd_brickinfo_from_brick (new_brick, - &new_brickinfo); + ret = glusterd_brickinfo_new_from_brick (new_brick, + &new_brickinfo); if (ret) goto out; - ret = glusterd_volume_brickinfo_get_by_brick (old_brick, volinfo, - &old_brickinfo, - GF_PATH_COMPLETE); + ret = glusterd_resolve_brick (new_brickinfo); + if (ret) goto out; - ret = glusterd_resolve_brick (new_brickinfo); + ret = glusterd_volume_brickinfo_get_by_brick (old_brick, + volinfo, &old_brickinfo); if (ret) goto out; @@ -1363,13 +1530,13 @@ glusterd_op_perform_replace_brick (glusterd_volinfo_t *volinfo, goto out; if (GLUSTERD_STATUS_STARTED == volinfo->status) { - ret = glusterd_brick_start (volinfo, new_brickinfo); + ret = glusterd_brick_start (volinfo, new_brickinfo, _gf_false); if (ret) goto out; } - out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -1388,6 +1555,7 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) char *dst_brick = NULL; glusterd_brickinfo_t *src_brickinfo = NULL; glusterd_brickinfo_t *dst_brickinfo = NULL; + char *task_id_str = NULL; this = THIS; GF_ASSERT (this); @@ -1397,26 +1565,23 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) ret = dict_get_str (dict, "src-brick", &src_brick); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get src brick"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get src brick"); goto out; } - gf_log (this->name, GF_LOG_DEBUG, - "src brick=%s", src_brick); + gf_log (this->name, GF_LOG_DEBUG, "src brick=%s", src_brick); ret = dict_get_str (dict, "dst-brick", &dst_brick); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get dst brick"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get dst brick"); goto out; } - gf_log (this->name, GF_LOG_DEBUG, - "dst brick=%s", dst_brick); + gf_log (this->name, GF_LOG_DEBUG, "dst brick=%s", dst_brick); ret = dict_get_str (dict, "volname", &volname); - if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } @@ -1429,29 +1594,30 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) ret = glusterd_volinfo_find (volname, &volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to allocate memory"); + gf_log (this->name, GF_LOG_ERROR, "Unable to allocate memory"); goto out; } ret = glusterd_volume_brickinfo_get_by_brick (src_brick, volinfo, - &src_brickinfo, - GF_PATH_COMPLETE); + &src_brickinfo); if (ret) { - gf_log ("", GF_LOG_DEBUG, "Unable to get src-brickinfo"); + gf_log (this->name, GF_LOG_DEBUG, + "Unable to get src-brickinfo"); goto out; } ret = glusterd_get_rb_dst_brickinfo (volinfo, &dst_brickinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get " + gf_log (this->name, GF_LOG_ERROR, "Unable to get " "replace brick destination brickinfo"); goto out; } ret = glusterd_resolve_brick (dst_brickinfo); if (ret) { - gf_log ("", GF_LOG_DEBUG, "Unable to resolve dst-brickinfo"); + gf_log (this->name, GF_LOG_DEBUG, + "Unable to resolve dst-brickinfo"); goto out; } @@ -1460,45 +1626,82 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) if (ret) goto out; + if ((GF_REPLACE_OP_START != replace_op)) { - ret = rb_update_dstbrick_port (dst_brickinfo, rsp_dict, - dict, replace_op); - if (ret) - goto out; + + /* Set task-id, if available, in op_ctx dict for operations + * other than start + */ + if (is_origin_glusterd (dict)) { + ctx = glusterd_op_get_ctx(); + if (!ctx) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "get op_ctx"); + ret = -1; + goto out; + } + if (!uuid_is_null (volinfo->rep_brick.rb_id)) { + ret = glusterd_copy_uuid_to_dict + (volinfo->rep_brick.rb_id, ctx, + GF_REPLACE_BRICK_TID_KEY); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set " + "replace-brick-id"); + goto out; + } + } + } } + ret = rb_update_dstbrick_port (dst_brickinfo, rsp_dict, + dict, replace_op); + if (ret) + goto out; switch (replace_op) { case GF_REPLACE_OP_START: { - if (!glusterd_is_local_addr (dst_brickinfo->hostname)) { - gf_log ("", GF_LOG_INFO, + ret = dict_get_str (dict, GF_REPLACE_BRICK_TID_KEY, &task_id_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Missing replace-brick-id"); + ret = 0; + } else { + uuid_parse (task_id_str, volinfo->rep_brick.rb_id); + } + + if (gf_is_local_addr (dst_brickinfo->hostname)) { + gf_log (this->name, GF_LOG_INFO, "I AM THE DESTINATION HOST"); if (!glusterd_is_rb_paused (volinfo)) { - ret = rb_spawn_destination_brick (volinfo, dst_brickinfo); + ret = rb_spawn_destination_brick + (volinfo, dst_brickinfo); if (ret) { - gf_log ("", GF_LOG_DEBUG, - "Failed to spawn destination brick"); + gf_log (this->name, GF_LOG_DEBUG, + "Failed to spawn destination " + "brick"); goto out; } } else { - gf_log ("", GF_LOG_ERROR, "Replace brick is already " - "started=> no need to restart dst brick "); + gf_log (this->name, GF_LOG_ERROR, + "Replace brick is already started=> no " + "need to restart dst brick "); } } - if (!glusterd_is_local_addr (src_brickinfo->hostname)) { + if (gf_is_local_addr (src_brickinfo->hostname)) { ret = rb_src_brick_restart (volinfo, src_brickinfo, 1); if (ret) { - gf_log ("", GF_LOG_DEBUG, - "Could not restart src-brick"); + gf_log (this->name, GF_LOG_DEBUG, + "Could not restart src-brick"); goto out; } } - if (!glusterd_is_local_addr (dst_brickinfo->hostname)) { - gf_log ("", GF_LOG_INFO, + if (gf_is_local_addr (dst_brickinfo->hostname)) { + gf_log (this->name, GF_LOG_INFO, "adding dst-brick port no"); ret = rb_update_dstbrick_port (dst_brickinfo, rsp_dict, @@ -1512,7 +1715,6 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) } case GF_REPLACE_OP_COMMIT: - case GF_REPLACE_OP_COMMIT_FORCE: { ctx = glusterd_op_get_ctx (); if (ctx) { @@ -1520,71 +1722,64 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) dst_brickinfo, GF_REPLACE_OP_COMMIT); if (ret) { - gf_log ("", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Commit operation failed"); goto out; } } - - ret = dict_set_int32 (volinfo->dict, "enable-pump", 0); - gf_log ("", GF_LOG_DEBUG, - "Received commit - will be adding dst brick and " - "removing src brick"); - - if (!glusterd_is_local_addr (dst_brickinfo->hostname) && - replace_op != GF_REPLACE_OP_COMMIT_FORCE) { - gf_log ("", GF_LOG_INFO, + } + /* fall through */ + case GF_REPLACE_OP_COMMIT_FORCE: + { + if (gf_is_local_addr (dst_brickinfo->hostname)) { + gf_log (this->name, GF_LOG_DEBUG, "I AM THE DESTINATION HOST"); - ret = rb_kill_destination_brick (volinfo, dst_brickinfo); + ret = rb_kill_destination_brick (volinfo, + dst_brickinfo); if (ret) { - gf_log ("", GF_LOG_DEBUG, - "Failed to kill destination brick"); + gf_log (this->name, GF_LOG_CRITICAL, + "Unable to cleanup dst brick"); goto out; } } - if (ret) { - gf_log ("", GF_LOG_CRITICAL, - "Unable to cleanup dst brick"); - goto out; - } - - ret = glusterd_nodesvcs_stop (volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Unable to stop nfs server, ret: %d", ret); } ret = glusterd_op_perform_replace_brick (volinfo, src_brick, dst_brick); if (ret) { - gf_log ("", GF_LOG_CRITICAL, "Unable to add " - "dst-brick: %s to volume: %s", - dst_brick, volinfo->volname); + gf_log (this->name, GF_LOG_CRITICAL, "Unable to add " + "dst-brick: %s to volume: %s", dst_brick, + volinfo->volname); (void) glusterd_nodesvcs_handle_graph_change (volinfo); goto out; } - volinfo->defrag_status = 0; + volinfo->rebal.defrag_status = 0; ret = glusterd_nodesvcs_handle_graph_change (volinfo); if (ret) { - gf_log ("", GF_LOG_CRITICAL, + gf_log (this->name, GF_LOG_CRITICAL, "Failed to generate nfs volume file"); } ret = glusterd_fetchspec_notify (THIS); glusterd_set_rb_status (volinfo, GF_RB_STATUS_NONE); - glusterd_brickinfo_delete (volinfo->dst_brick); - volinfo->src_brick = volinfo->dst_brick = NULL; + glusterd_brickinfo_delete (volinfo->rep_brick.dst_brick); + volinfo->rep_brick.src_brick = NULL; + volinfo->rep_brick.dst_brick = NULL; + uuid_clear (volinfo->rep_brick.rb_id); } break; case GF_REPLACE_OP_PAUSE: { - gf_log ("", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "Received pause - doing nothing"); ctx = glusterd_op_get_ctx (); if (ctx) { @@ -1592,7 +1787,7 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) dst_brickinfo, GF_REPLACE_OP_PAUSE); if (ret) { - gf_log ("", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Pause operation failed"); goto out; } @@ -1611,18 +1806,13 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) dst_brickinfo, GF_REPLACE_OP_ABORT); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Abort operation failed"); goto out; } } - ret = dict_set_int32 (volinfo->dict, "enable-pump", 0); - if (ret) { - gf_log (THIS->name, GF_LOG_CRITICAL, "Unable to disable pump"); - } - - if (!glusterd_is_local_addr (src_brickinfo->hostname)) { + if (gf_is_local_addr (src_brickinfo->hostname)) { ret = rb_src_brick_restart (volinfo, src_brickinfo, 0); if (ret) { @@ -1633,8 +1823,8 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) } } - if (!glusterd_is_local_addr (dst_brickinfo->hostname)) { - gf_log (THIS->name, GF_LOG_INFO, + if (gf_is_local_addr (dst_brickinfo->hostname)) { + gf_log (this->name, GF_LOG_INFO, "I AM THE DESTINATION HOST"); ret = rb_kill_destination_brick (volinfo, dst_brickinfo); if (ret) { @@ -1644,14 +1834,15 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) } } glusterd_set_rb_status (volinfo, GF_RB_STATUS_NONE); - glusterd_brickinfo_delete (volinfo->dst_brick); - volinfo->src_brick = volinfo->dst_brick = NULL; + glusterd_brickinfo_delete (volinfo->rep_brick.dst_brick); + volinfo->rep_brick.src_brick = NULL; + volinfo->rep_brick.dst_brick = NULL; } break; case GF_REPLACE_OP_STATUS: { - gf_log ("", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "received status - doing nothing"); ctx = glusterd_op_get_ctx (); if (ctx) { @@ -1659,7 +1850,7 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) ret = dict_set_str (ctx, "status-reply", "replace brick has been paused"); if (ret) - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to set pump status" " in ctx"); goto out; @@ -1683,7 +1874,7 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); if (ret) - gf_log (THIS->name, GF_LOG_ERROR, "Couldn't store" + gf_log (this->name, GF_LOG_ERROR, "Couldn't store" " replace brick operation's state"); out: @@ -1704,6 +1895,7 @@ glusterd_do_replace_brick (void *data) glusterd_brickinfo_t *src_brickinfo = NULL; glusterd_brickinfo_t *dst_brickinfo = NULL; glusterd_conf_t *priv = NULL; + uuid_t *txn_id = &global_txn_id; int ret = 0; @@ -1723,6 +1915,10 @@ glusterd_do_replace_brick (void *data) gf_log ("", GF_LOG_DEBUG, "Replace brick operation detected"); + ret = dict_get_bin (dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + ret = dict_get_int32 (dict, "operation", &op); if (ret) { gf_log ("", GF_LOG_DEBUG, @@ -1761,8 +1957,7 @@ glusterd_do_replace_brick (void *data) } ret = glusterd_volume_brickinfo_get_by_brick (src_brick, volinfo, - &src_brickinfo, - GF_PATH_COMPLETE); + &src_brickinfo); if (ret) { gf_log ("", GF_LOG_DEBUG, "Unable to get src-brickinfo"); goto out; @@ -1819,12 +2014,11 @@ glusterd_do_replace_brick (void *data) out: if (ret) - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + txn_id, NULL); else - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, NULL); - - if (dict) - dict_unref (dict); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, + txn_id, NULL); glusterd_op_sm (); } diff --git a/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c b/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c index 21d8ab0ac..d5200a4ae 100644 --- a/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c +++ b/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H @@ -43,6 +33,7 @@ extern glusterd_op_info_t opinfo; +extern uuid_t global_txn_id; int32_t glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret, @@ -59,10 +50,11 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret, int32_t status = 0; int32_t count = 0; gf_cli_rsp rsp = {0,}; + xlator_t *this = NULL; - GF_ASSERT (THIS); - - conf = THIS->private; + this = THIS; + GF_ASSERT (this); + conf = this->private; GF_ASSERT (conf); @@ -82,16 +74,18 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret, break; } case GD_OP_REBALANCE: + case GD_OP_DEFRAG_BRICK_VOLUME: { if (ctx) { ret = dict_get_int32 (ctx, "status", &status); if (ret) { - gf_log (THIS->name, GF_LOG_TRACE, + gf_log (this->name, GF_LOG_TRACE, "failed to get status"); } } break; } + case GD_OP_GSYNC_CREATE: case GD_OP_GSYNC_SET: { if (ctx) { @@ -115,7 +109,7 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret, if (ctx && dict_get_int32 (ctx, "count", &count)) { ret = dict_set_int32 (ctx, "count", 0); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to set count in dictionary"); } } @@ -124,13 +118,14 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret, case GD_OP_START_BRICK: case GD_OP_STOP_BRICK: { - gf_log ("", GF_LOG_DEBUG, "not supported op %d", op); + gf_log (this->name, GF_LOG_DEBUG, "op '%s' not supported", + gd_op_list[op]); break; } case GD_OP_NONE: case GD_OP_MAX: { - gf_log ("", GF_LOG_ERROR, "invalid operation %d", op); + gf_log (this->name, GF_LOG_ERROR, "invalid operation"); break; } case GD_OP_CREATE_VOLUME: @@ -141,15 +136,33 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret, case GD_OP_ADD_BRICK: case GD_OP_LOG_ROTATE: case GD_OP_SYNC_VOLUME: - case GD_OP_HEAL_VOLUME: case GD_OP_STATEDUMP_VOLUME: case GD_OP_REPLACE_BRICK: case GD_OP_STATUS_VOLUME: case GD_OP_SET_VOLUME: + case GD_OP_LIST_VOLUME: + case GD_OP_CLEARLOCKS_VOLUME: + case GD_OP_HEAL_VOLUME: + case GD_OP_SNAP: { /*nothing specific to be done*/ break; } + case GD_OP_COPY_FILE: + { + if (ctx) + ret = dict_get_str (ctx, "errstr", &errstr); + break; + } + case GD_OP_SYS_EXEC: + { + if (ctx) { + ret = dict_get_str (ctx, "errstr", &errstr); + ret = dict_set_str (ctx, "glusterd_workdir", + conf->workdir); + } + break; + } } rsp.op_ret = op_ret; @@ -164,9 +177,9 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret, if (ctx) { ret = dict_allocate_and_serialize (ctx, &rsp.dict.dict_val, - (size_t*)&rsp.dict.dict_len); + &rsp.dict.dict_len); if (ret < 0 ) - gf_log (THIS->name, GF_LOG_ERROR, "failed to " + gf_log (this->name, GF_LOG_ERROR, "failed to " "serialize buffer"); else free_ptr = rsp.dict.dict_val; @@ -179,17 +192,31 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret, cli_rsp = &rsp; xdrproc = (xdrproc_t) xdr_gf_cli_rsp; - ret = glusterd_submit_reply (req, cli_rsp, NULL, 0, NULL, - xdrproc); + glusterd_to_cli (req, cli_rsp, NULL, 0, NULL, + xdrproc, ctx); + ret = 0; - if (free_ptr) - GF_FREE (free_ptr); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + GF_FREE (free_ptr); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } int -glusterd3_1_probe_cbk (struct rpc_req *req, struct iovec *iov, +glusterd_big_locked_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe, fop_cbk_fn_t fn) +{ + glusterd_conf_t *priv = THIS->private; + int ret = -1; + + synclock_lock (&priv->big_lock); + ret = fn (req, iov, count, myframe); + synclock_unlock (&priv->big_lock); + + return ret; +} + +int +__glusterd_probe_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) { gd1_mgmt_probe_rsp rsp = {{0},}; @@ -222,7 +249,9 @@ glusterd3_1_probe_cbk (struct rpc_req *req, struct iovec *iov, if (ctx->req) { glusterd_xfer_cli_probe_resp (ctx->req, rsp.op_ret, rsp.op_errno, - ctx->hostname, ctx->port); + rsp.op_errstr, + ctx->hostname, ctx->port, + ctx->dict); } glusterd_destroy_probe_ctx (ctx); @@ -249,7 +278,9 @@ glusterd3_1_probe_cbk (struct rpc_req *req, struct iovec *iov, if (ctx->req) { glusterd_xfer_cli_probe_resp (ctx->req, rsp.op_ret, rsp.op_errno, - ctx->hostname, ctx->port); + rsp.op_errstr, + ctx->hostname, ctx->port, + ctx->dict); } glusterd_destroy_probe_ctx (ctx); @@ -283,14 +314,22 @@ glusterd3_1_probe_cbk (struct rpc_req *req, struct iovec *iov, gf_log ("glusterd", GF_LOG_INFO, "Received resp to probe req"); out: - if (rsp.hostname) - free (rsp.hostname);//malloced by xdr + free (rsp.hostname);//malloced by xdr GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe)); return ret; } int -glusterd3_1_friend_add_cbk (struct rpc_req * req, struct iovec *iov, +glusterd_probe_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_probe_cbk); +} + + +int +__glusterd_friend_add_cbk (struct rpc_req * req, struct iovec *iov, int count, void *myframe) { gd1_mgmt_friend_rsp rsp = {{0},}; @@ -369,21 +408,29 @@ out: if (ctx->req)//reverse probe doesn't have req ret = glusterd_xfer_cli_probe_resp (ctx->req, op_ret, op_errno, - ctx->hostname, ctx->port); + NULL, ctx->hostname, + ctx->port, ctx->dict); if (!ret) { glusterd_friend_sm (); glusterd_op_sm (); } if (ctx) glusterd_destroy_probe_ctx (ctx); - if (rsp.hostname) - free (rsp.hostname);//malloced by xdr + free (rsp.hostname);//malloced by xdr GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe)); return ret; } int -glusterd3_1_friend_remove_cbk (struct rpc_req * req, struct iovec *iov, +glusterd_friend_add_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_friend_add_cbk); +} + +int +__glusterd_friend_remove_cbk (struct rpc_req * req, struct iovec *iov, int count, void *myframe) { gd1_mgmt_friend_rsp rsp = {{0},}; @@ -459,8 +506,8 @@ inject: respond: - ret = glusterd_xfer_cli_deprobe_resp (ctx->req, op_ret, op_errno, - ctx->hostname); + ret = glusterd_xfer_cli_deprobe_resp (ctx->req, op_ret, op_errno, NULL, + ctx->hostname, ctx->dict); if (!ret && move_sm_now) { glusterd_friend_sm (); glusterd_op_sm (); @@ -471,37 +518,62 @@ respond: glusterd_destroy_probe_ctx (ctx); } - if (rsp.hostname) - free (rsp.hostname);//malloced by xdr + free (rsp.hostname);//malloced by xdr GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe)); return ret; } +int +glusterd_friend_remove_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_friend_remove_cbk); +} + int32_t -glusterd3_1_friend_update_cbk (struct rpc_req *req, struct iovec *iov, +__glusterd_friend_update_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) { - int ret = -1; - int32_t op_ret = 0; - char str[50] = {0,}; + int ret = -1; + gd1_mgmt_friend_update_rsp rsp = {{0}, }; + xlator_t *this = NULL; GF_ASSERT (req); + this = THIS; if (-1 == req->rpc_status) { + gf_log (this->name, GF_LOG_ERROR, "RPC Error"); goto out; } - gf_log ("glusterd", GF_LOG_INFO, - "Received %s from uuid: %s", - (op_ret)?"RJT":"ACC", str); + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_friend_update_rsp); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to serialize friend" + " update repsonse"); + goto out; + } + ret = 0; out: + gf_log (this->name, GF_LOG_INFO, "Received %s from uuid: %s", + (ret)?"RJT":"ACC", uuid_utoa (rsp.uuid)); + GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe)); return ret; } +int +glusterd_friend_update_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_friend_update_cbk); +} + int32_t -glusterd3_1_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov, +__glusterd_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) { gd1_mgmt_cluster_lock_rsp rsp = {{0},}; @@ -509,7 +581,11 @@ glusterd3_1_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov, int32_t op_ret = -1; glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); if (-1 == req->rpc_status) { @@ -520,7 +596,8 @@ glusterd3_1_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov, ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_cluster_lock_rsp); if (ret < 0) { - gf_log ("", GF_LOG_ERROR, "error"); + gf_log (this->name, GF_LOG_ERROR, "Failed to decode lock " + "response received from peer"); rsp.op_ret = -1; rsp.op_errno = EINVAL; goto out; @@ -529,25 +606,28 @@ glusterd3_1_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov, out: op_ret = rsp.op_ret; - gf_log ("glusterd", GF_LOG_INFO, - "Received %s from uuid: %s", - (op_ret)?"RJT":"ACC", uuid_utoa (rsp.uuid)); + gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG, + "Received lock %s from uuid: %s", (op_ret) ? "RJT" : "ACC", + uuid_utoa (rsp.uuid)); ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); if (ret) { - gf_log ("", GF_LOG_CRITICAL, "Lock response received from " - "unknown peer: %s", uuid_utoa (rsp.uuid)); + gf_log (this->name, GF_LOG_CRITICAL, "Lock response received " + "from unknown peer: %s", uuid_utoa (rsp.uuid)); } if (op_ret) { event_type = GD_OP_EVENT_RCVD_RJT; opinfo.op_ret = op_ret; + opinfo.op_errstr = gf_strdup ("Another transaction could be in " + "progress. Please try again after" + " sometime."); } else { event_type = GD_OP_EVENT_RCVD_ACC; } - ret = glusterd_op_sm_inject_event (event_type, NULL); + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); if (!ret) { glusterd_friend_sm (); @@ -559,16 +639,27 @@ out: } int32_t -glusterd3_1_cluster_unlock_cbk (struct rpc_req *req, struct iovec *iov, - int count, void *myframe) +glusterd_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) { - gd1_mgmt_cluster_lock_rsp rsp = {{0},}; + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_cluster_lock_cbk); +} + +static int32_t +glusterd_mgmt_v3_lock_peers_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + gd1_mgmt_v3_lock_rsp rsp = {{0},}; int ret = -1; int32_t op_ret = -1; glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = NULL; - + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); if (-1 == req->rpc_status) { @@ -577,9 +668,12 @@ glusterd3_1_cluster_unlock_cbk (struct rpc_req *req, struct iovec *iov, goto out; } - ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_rsp); + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp); if (ret < 0) { - gf_log ("", GF_LOG_ERROR, "error"); + gf_log (this->name, GF_LOG_ERROR, + "Failed to decode mgmt_v3 lock " + "response received from peer"); rsp.op_ret = -1; rsp.op_errno = EINVAL; goto out; @@ -588,25 +682,30 @@ glusterd3_1_cluster_unlock_cbk (struct rpc_req *req, struct iovec *iov, out: op_ret = rsp.op_ret; - gf_log ("glusterd", GF_LOG_INFO, - "Received %s from uuid: %s", - (op_ret)?"RJT":"ACC", uuid_utoa (rsp.uuid)); + txn_id = &rsp.txn_id; - ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); + gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG, + "Received mgmt_v3 lock %s from uuid: %s", + (op_ret) ? "RJT" : "ACC", uuid_utoa (rsp.uuid)); + ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); if (ret) { - gf_log ("", GF_LOG_CRITICAL, "Unlock response received from " - "unknown peer %s", uuid_utoa (rsp.uuid)); + gf_log (this->name, GF_LOG_CRITICAL, + "mgmt_v3 lock response received " + "from unknown peer: %s", uuid_utoa (rsp.uuid)); } if (op_ret) { event_type = GD_OP_EVENT_RCVD_RJT; opinfo.op_ret = op_ret; + opinfo.op_errstr = gf_strdup ("Another transaction could be in " + "progress. Please try again after" + " sometime."); } else { event_type = GD_OP_EVENT_RCVD_ACC; } - ret = glusterd_op_sm_inject_event (event_type, NULL); + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); if (!ret) { glusterd_friend_sm (); @@ -617,197 +716,167 @@ out: return ret; } -static int32_t -glusterd_append_gsync_status (dict_t *dst, dict_t *src) +int32_t +glusterd_mgmt_v3_lock_peers_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) { - int ret = 0; - char *stop_msg = NULL; - - ret = dict_get_str (src, "gsync-status", &stop_msg); - if (ret) { - ret = 0; - goto out; - } - - ret = dict_set_dynstr (dst, "gsync-status", gf_strdup (stop_msg)); - if (ret) { - gf_log ("glusterd", GF_LOG_WARNING, "Unable to set the stop" - "message in the ctx dictionary"); - goto out; - } - - ret = 0; - out: - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); - return ret; - + return glusterd_big_locked_cbk (req, iov, count, myframe, + glusterd_mgmt_v3_lock_peers_cbk_fn); } static int32_t -glusterd_append_status_dicts (dict_t *dst, dict_t *src) +glusterd_mgmt_v3_unlock_peers_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) { - int dst_count = 0; - int src_count = 0; - int i = 0; - int ret = 0; - char mst[PATH_MAX] = {0,}; - char slv[PATH_MAX] = {0, }; - char sts[PATH_MAX] = {0, }; - char *mst_val = NULL; - char *slv_val = NULL; - char *sts_val = NULL; - - GF_ASSERT (dst); - - if (src == NULL) - goto out; + gd1_mgmt_v3_unlock_rsp rsp = {{0},}; + int ret = -1; + int32_t op_ret = -1; + glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = NULL; - ret = dict_get_int32 (dst, "gsync-count", &dst_count); - if (ret) - dst_count = 0; + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); - ret = dict_get_int32 (src, "gsync-count", &src_count); - if (ret || !src_count) { - gf_log ("", GF_LOG_DEBUG, "Source brick empty"); - ret = 0; + if (-1 == req->rpc_status) { + rsp.op_ret = -1; + rsp.op_errno = EINVAL; goto out; } - for (i = 1; i <= src_count; i++) { - snprintf (mst, sizeof(mst), "master%d", i); - snprintf (slv, sizeof(slv), "slave%d", i); - snprintf (sts, sizeof(sts), "status%d", i); + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to decode mgmt_v3 unlock " + "response received from peer"); + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } - ret = dict_get_str (src, mst, &mst_val); - if (ret) - goto out; +out: + op_ret = rsp.op_ret; - ret = dict_get_str (src, slv, &slv_val); - if (ret) - goto out; + txn_id = &rsp.txn_id; - ret = dict_get_str (src, sts, &sts_val); - if (ret) - goto out; + gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG, + "Received mgmt_v3 unlock %s from uuid: %s", + (op_ret) ? "RJT" : "ACC", + uuid_utoa (rsp.uuid)); - snprintf (mst, sizeof(mst), "master%d", i+dst_count); - snprintf (slv, sizeof(slv), "slave%d", i+dst_count); - snprintf (sts, sizeof(sts), "status%d", i+dst_count); + ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); - ret = dict_set_dynstr (dst, mst, gf_strdup (mst_val)); - if (ret) - goto out; + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, + "mgmt_v3 unlock response received " + "from unknown peer: %s", uuid_utoa (rsp.uuid)); + } - ret = dict_set_dynstr (dst, slv, gf_strdup (slv_val)); - if (ret) - goto out; + if (op_ret) { + event_type = GD_OP_EVENT_RCVD_RJT; + opinfo.op_ret = op_ret; + opinfo.op_errstr = gf_strdup ("Another transaction could be in " + "progress. Please try again after" + " sometime."); + } else { + event_type = GD_OP_EVENT_RCVD_ACC; + } - ret = dict_set_dynstr (dst, sts, gf_strdup (sts_val)); - if (ret) - goto out; + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); + if (!ret) { + glusterd_friend_sm (); + glusterd_op_sm (); } - ret = dict_set_int32 (dst, "gsync-count", dst_count+src_count); - - out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe)); return ret; - } -static int32_t -glusterd_gsync_use_rsp_dict (dict_t *rsp_dict, char *op_errstr) +int32_t +glusterd_mgmt_v3_unlock_peers_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) { - dict_t *ctx = NULL; - int ret = 0; + return glusterd_big_locked_cbk (req, iov, count, myframe, + glusterd_mgmt_v3_unlock_peers_cbk_fn); +} - ctx = glusterd_op_get_ctx (); - if (!ctx) { - gf_log ("", GF_LOG_ERROR, - "Operation Context is not present"); - GF_ASSERT (0); - } +int32_t +__glusterd_cluster_unlock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + gd1_mgmt_cluster_lock_rsp rsp = {{0},}; + int ret = -1; + int32_t op_ret = -1; + glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; - if (rsp_dict) { - ret = glusterd_append_status_dicts (ctx, rsp_dict); - if (ret) - goto out; + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); - ret = glusterd_append_gsync_status (ctx, rsp_dict); - if (ret) - goto out; - } - if (strcmp ("", op_errstr)) { - ret = dict_set_dynstr (ctx, "errstr", gf_strdup(op_errstr)); - if (ret) - goto out; + if (-1 == req->rpc_status) { + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; } - ret = 0; - out: - gf_log ("", GF_LOG_DEBUG, "Returning %d ", ret); - return ret; -} -static int32_t -glusterd_rb_use_rsp_dict (dict_t *rsp_dict) -{ - int32_t src_port = 0; - int32_t dst_port = 0; - int ret = 0; - dict_t *ctx = NULL; - - - ctx = glusterd_op_get_ctx (); - if (!ctx) { - gf_log ("", GF_LOG_ERROR, - "Operation Context is not present"); - GF_ASSERT (0); + ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_rsp); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode unlock " + "response received from peer"); + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; } - if (rsp_dict) { - ret = dict_get_int32 (rsp_dict, "src-brick-port", &src_port); - if (ret == 0) { - gf_log ("", GF_LOG_DEBUG, - "src-brick-port=%d found", src_port); - } +out: + op_ret = rsp.op_ret; - ret = dict_get_int32 (rsp_dict, "dst-brick-port", &dst_port); - if (ret == 0) { - gf_log ("", GF_LOG_DEBUG, - "dst-brick-port=%d found", dst_port); - } + gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG, + "Received unlock %s from uuid: %s", + (op_ret)?"RJT":"ACC", uuid_utoa (rsp.uuid)); + + ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, "Unlock response received " + "from unknown peer %s", uuid_utoa (rsp.uuid)); } - if (src_port) { - ret = dict_set_int32 (ctx, "src-brick-port", - src_port); - if (ret) { - gf_log ("", GF_LOG_DEBUG, - "Could not set src-brick"); - goto out; - } + if (op_ret) { + event_type = GD_OP_EVENT_RCVD_RJT; + opinfo.op_ret = op_ret; + } else { + event_type = GD_OP_EVENT_RCVD_ACC; } - if (dst_port) { - ret = dict_set_int32 (ctx, "dst-brick-port", - dst_port); - if (ret) { - gf_log ("", GF_LOG_DEBUG, - "Could not set dst-brick"); - goto out; - } + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); + if (!ret) { + glusterd_friend_sm (); + glusterd_op_sm (); } -out: + GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe)); return ret; +} +int32_t +glusterd_cluster_unlock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_cluster_unlock_cbk); } int32_t -glusterd3_1_stage_op_cbk (struct rpc_req *req, struct iovec *iov, +__glusterd_stage_op_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) { gd1_mgmt_stage_op_rsp rsp = {{0},}; @@ -818,22 +887,32 @@ glusterd3_1_stage_op_cbk (struct rpc_req *req, struct iovec *iov, dict_t *dict = NULL; char err_str[2048] = {0}; char *peer_str = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); if (-1 == req->rpc_status) { rsp.op_ret = -1; rsp.op_errno = EINVAL; - rsp.op_errstr = "error"; + /* use standard allocation because to keep uniformity + in freeing it */ + rsp.op_errstr = strdup ("error"); goto out; } ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_stage_op_rsp); if (ret < 0) { - gf_log ("", GF_LOG_ERROR, "error"); + gf_log (this->name, GF_LOG_ERROR, "Failed to decode stage " + "response received from peer"); rsp.op_ret = -1; rsp.op_errno = EINVAL; - rsp.op_errstr = "error"; + /* use standard allocation because to keep uniformity + in freeing it */ + rsp.op_errstr = strdup ("Failed to decode stage response " + "received from peer."); goto out; } @@ -845,7 +924,7 @@ glusterd3_1_stage_op_cbk (struct rpc_req *req, struct iovec *iov, rsp.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize rsp-buffer to dictionary"); event_type = GD_OP_EVENT_RCVD_RJT; @@ -858,15 +937,19 @@ glusterd3_1_stage_op_cbk (struct rpc_req *req, struct iovec *iov, out: op_ret = rsp.op_ret; - gf_log ("glusterd", GF_LOG_INFO, - "Received %s from uuid: %s", - (op_ret)?"RJT":"ACC", uuid_utoa (rsp.uuid)); + gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG, + "Received stage %s from uuid: %s", + (op_ret) ? "RJT" : "ACC", uuid_utoa (rsp.uuid)); + + ret = dict_get_bin (dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); if (ret) { - gf_log ("", GF_LOG_CRITICAL, "Stage response received from " - "unknown peer: %s", uuid_utoa (rsp.uuid)); + gf_log (this->name, GF_LOG_CRITICAL, "Stage response received " + "from unknown peer: %s", uuid_utoa (rsp.uuid)); } if (op_ret) { @@ -879,12 +962,11 @@ out: peer_str = peerinfo->hostname; else peer_str = uuid_utoa (rsp.uuid); - snprintf (err_str, sizeof (err_str), "Operation failed " - "on %s", peer_str); + snprintf (err_str, sizeof (err_str), + OPERRSTR_STAGE_FAIL, peer_str); opinfo.op_errstr = gf_strdup (err_str); } if (!opinfo.op_errstr) { - gf_log ("", GF_LOG_ERROR, "memory allocation failed"); ret = -1; goto out; } @@ -894,218 +976,39 @@ out: switch (rsp.op) { case GD_OP_REPLACE_BRICK: - glusterd_rb_use_rsp_dict (dict); + glusterd_rb_use_rsp_dict (NULL, dict); break; } - ret = glusterd_op_sm_inject_event (event_type, NULL); + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); if (!ret) { glusterd_friend_sm (); glusterd_op_sm (); } - if (rsp.op_errstr && strcmp (rsp.op_errstr, "error")) - free (rsp.op_errstr); //malloced by xdr + free (rsp.op_errstr); //malloced by xdr if (dict) { if (!dict->extra_stdfree && rsp.dict.dict_val) free (rsp.dict.dict_val); //malloced by xdr dict_unref (dict); } else { - if (rsp.dict.dict_val) - free (rsp.dict.dict_val); //malloced by xdr + free (rsp.dict.dict_val); //malloced by xdr } GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe)); return ret; } -static int32_t -glusterd_sync_use_rsp_dict (dict_t *rsp_dict) -{ - int ret = 0; - - GF_ASSERT (rsp_dict); - - if (!rsp_dict) { - goto out; - } - - ret = glusterd_import_friend_volumes (rsp_dict); -out: - return ret; - -} - -void -_profile_volume_add_friend_rsp (dict_t *this, char *key, data_t *value, - void *data) -{ - char new_key[256] = {0}; - glusterd_pr_brick_rsp_conv_t *rsp_ctx = NULL; - data_t *new_value = NULL; - int brick_count = 0; - char brick_key[256]; - - if (strcmp (key, "count") == 0) - return; - sscanf (key, "%d%s", &brick_count, brick_key); - rsp_ctx = data; - new_value = data_copy (value); - GF_ASSERT (new_value); - snprintf (new_key, sizeof (new_key), "%d%s", - rsp_ctx->count + brick_count, brick_key); - dict_set (rsp_ctx->dict, new_key, new_value); -} - -int -glusterd_profile_volume_use_rsp_dict (dict_t *rsp_dict) -{ - int ret = 0; - glusterd_pr_brick_rsp_conv_t rsp_ctx = {0}; - int32_t brick_count = 0; - int32_t count = 0; - dict_t *ctx_dict = NULL; - glusterd_op_t op = GD_OP_NONE; - - GF_ASSERT (rsp_dict); - - ret = dict_get_int32 (rsp_dict, "count", &brick_count); - if (ret) { - ret = 0; //no bricks in the rsp - goto out; - } - - op = glusterd_op_get_op (); - GF_ASSERT (GD_OP_PROFILE_VOLUME == op); - ctx_dict = glusterd_op_get_ctx (); - - ret = dict_get_int32 (ctx_dict, "count", &count); - rsp_ctx.count = count; - rsp_ctx.dict = ctx_dict; - dict_foreach (rsp_dict, _profile_volume_add_friend_rsp, &rsp_ctx); - dict_del (ctx_dict, "count"); - ret = dict_set_int32 (ctx_dict, "count", count + brick_count); -out: - return ret; -} - -void -glusterd_volume_status_add_peer_rsp (dict_t *this, char *key, data_t *value, - void *data) +int32_t +glusterd_stage_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) { - glusterd_status_rsp_conv_t *rsp_ctx = NULL; - data_t *new_value = NULL; - int32_t ret = 0; - - if (strcmp (key, "count") == 0) - return; - - rsp_ctx = data; - new_value = data_copy (value); - GF_ASSERT (new_value); - - ret = dict_set (rsp_ctx->dict, key, new_value); - if (ret) - gf_log ("", GF_LOG_ERROR, "Unable to set key: %s in dict", - key); - - return; + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_stage_op_cbk); } -int -glusterd_volume_status_use_rsp_dict (dict_t *rsp_dict) -{ - int ret = 0; - glusterd_status_rsp_conv_t rsp_ctx = {0}; - int32_t brick_count = 0; - int32_t count = 0; - dict_t *ctx_dict = NULL; - glusterd_op_t op = GD_OP_NONE; - - GF_ASSERT (rsp_dict); - - ret = dict_get_int32 (rsp_dict, "count", &brick_count); - if (ret) { - ret = 0; //no bricks in the rsp - goto out; - } - - op = glusterd_op_get_op (); - GF_ASSERT (GD_OP_STATUS_VOLUME == op); - ctx_dict = glusterd_op_get_ctx (op); - - ret = dict_get_int32 (ctx_dict, "count", &count); - rsp_ctx.count = count; - rsp_ctx.dict = ctx_dict; - dict_foreach (rsp_dict, glusterd_volume_status_add_peer_rsp, &rsp_ctx); - - ret = dict_set_int32 (ctx_dict, "count", count + brick_count); -out: - return ret; -} - -int -glusterd_volume_rebalance_use_rsp_dict (dict_t *rsp_dict) -{ - int ret = 0; - dict_t *ctx_dict = NULL; - glusterd_op_t op = GD_OP_NONE; - uint64_t value = 0; - int32_t value32 = 0; - - GF_ASSERT (rsp_dict); - - op = glusterd_op_get_op (); - GF_ASSERT (GD_OP_REBALANCE == op); - - ctx_dict = glusterd_op_get_ctx (op); - - if (!ctx_dict) - goto out; - - ret = dict_get_uint64 (rsp_dict, "files", &value); - if (!ret) { - ret = dict_set_uint64 (ctx_dict, "files", value); - if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, - "failed to set the file count"); - } - } - - ret = dict_get_uint64 (rsp_dict, "size", &value); - if (!ret) { - ret = dict_set_uint64 (ctx_dict, "size", value); - if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, - "failed to set the size of migration"); - } - } - - ret = dict_get_uint64 (rsp_dict, "lookups", &value); - if (!ret) { - ret = dict_set_uint64 (ctx_dict, "lookups", value); - if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, - "failed to set lookuped file count"); - } - } - - ret = dict_get_int32 (rsp_dict, "status", &value32); - if (!ret) { - ret = dict_set_int32 (ctx_dict, "status", value32); - if (ret) { - gf_log (THIS->name, GF_LOG_DEBUG, - "failed to set status"); - } - } - -out: - return ret; -} - - int32_t -glusterd3_1_commit_op_cbk (struct rpc_req *req, struct iovec *iov, +__glusterd_commit_op_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) { gd1_mgmt_commit_op_rsp rsp = {{0},}; @@ -1116,24 +1019,34 @@ glusterd3_1_commit_op_cbk (struct rpc_req *req, struct iovec *iov, dict_t *dict = NULL; char err_str[2048] = {0}; char *peer_str = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); if (-1 == req->rpc_status) { rsp.op_ret = -1; rsp.op_errno = EINVAL; - rsp.op_errstr = "error"; + /* use standard allocation because to keep uniformity + in freeing it */ + rsp.op_errstr = strdup ("error"); event_type = GD_OP_EVENT_RCVD_RJT; goto out; } ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_commit_op_rsp); if (ret < 0) { - gf_log ("", GF_LOG_ERROR, "error"); + gf_log (this->name, GF_LOG_ERROR, "Failed to decode commit " + "response received from peer"); rsp.op_ret = -1; rsp.op_errno = EINVAL; - rsp.op_errstr = "error"; + /* use standard allocation because to keep uniformity + in freeing it */ + rsp.op_errstr = strdup ("Failed to decode commit response " + "received from peer."); event_type = GD_OP_EVENT_RCVD_RJT; goto out; } @@ -1146,7 +1059,7 @@ glusterd3_1_commit_op_cbk (struct rpc_req *req, struct iovec *iov, rsp.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize rsp-buffer to dictionary"); event_type = GD_OP_EVENT_RCVD_RJT; @@ -1158,15 +1071,20 @@ glusterd3_1_commit_op_cbk (struct rpc_req *req, struct iovec *iov, op_ret = rsp.op_ret; - gf_log ("glusterd", GF_LOG_INFO, - "Received %s from uuid: %s", + gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG, + "Received commit %s from uuid: %s", (op_ret)?"RJT":"ACC", uuid_utoa (rsp.uuid)); + ret = dict_get_bin (dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); if (ret) { - gf_log ("", GF_LOG_CRITICAL, "Commit response received from " - "unknown peer: %s", uuid_utoa (rsp.uuid)); + gf_log (this->name, GF_LOG_CRITICAL, "Commit response for " + "'Volume %s' received from unknown peer: %s", + gd_op_list[opinfo.op], uuid_utoa (rsp.uuid)); } if (op_ret) { @@ -1179,12 +1097,11 @@ glusterd3_1_commit_op_cbk (struct rpc_req *req, struct iovec *iov, peer_str = peerinfo->hostname; else peer_str = uuid_utoa (rsp.uuid); - snprintf (err_str, sizeof (err_str), "Operation failed " - "on %s", peer_str); + snprintf (err_str, sizeof (err_str), + OPERRSTR_COMMIT_FAIL, peer_str); opinfo.op_errstr = gf_strdup (err_str); } if (!opinfo.op_errstr) { - gf_log ("", GF_LOG_ERROR, "memory allocation failed"); ret = -1; goto out; } @@ -1192,37 +1109,44 @@ glusterd3_1_commit_op_cbk (struct rpc_req *req, struct iovec *iov, event_type = GD_OP_EVENT_RCVD_ACC; switch (rsp.op) { case GD_OP_REPLACE_BRICK: - ret = glusterd_rb_use_rsp_dict (dict); + ret = glusterd_rb_use_rsp_dict (NULL, dict); if (ret) goto out; break; case GD_OP_SYNC_VOLUME: - ret = glusterd_sync_use_rsp_dict (dict); + ret = glusterd_sync_use_rsp_dict (NULL, dict); if (ret) goto out; break; case GD_OP_PROFILE_VOLUME: - ret = glusterd_profile_volume_use_rsp_dict (dict); + ret = glusterd_profile_volume_use_rsp_dict (NULL, dict); if (ret) goto out; break; case GD_OP_GSYNC_SET: - ret = glusterd_gsync_use_rsp_dict (dict, rsp.op_errstr); + ret = glusterd_gsync_use_rsp_dict (NULL, dict, rsp.op_errstr); if (ret) goto out; break; case GD_OP_STATUS_VOLUME: - ret = glusterd_volume_status_use_rsp_dict (dict); + ret = glusterd_volume_status_copy_to_op_ctx_dict (NULL, dict); if (ret) goto out; break; case GD_OP_REBALANCE: - ret = glusterd_volume_rebalance_use_rsp_dict (dict); + case GD_OP_DEFRAG_BRICK_VOLUME: + ret = glusterd_volume_rebalance_use_rsp_dict (NULL, dict); + if (ret) + goto out; + break; + + case GD_OP_HEAL_VOLUME: + ret = glusterd_volume_heal_use_rsp_dict (NULL, dict); if (ret) goto out; @@ -1234,7 +1158,7 @@ glusterd3_1_commit_op_cbk (struct rpc_req *req, struct iovec *iov, } out: - ret = glusterd_op_sm_inject_event (event_type, NULL); + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); if (!ret) { glusterd_friend_sm (); @@ -1243,17 +1167,22 @@ out: if (dict) dict_unref (dict); - if (rsp.op_errstr && strcmp (rsp.op_errstr, "error")) - free (rsp.op_errstr); //malloced by xdr + free (rsp.op_errstr); //malloced by xdr GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe)); return ret; } - +int32_t +glusterd_commit_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_commit_op_cbk); +} int32_t -glusterd3_1_probe (call_frame_t *frame, xlator_t *this, - void *data) +glusterd_rpc_probe (call_frame_t *frame, xlator_t *this, + void *data) { gd1_mgmt_probe_req req = {{0},}; int ret = 0; @@ -1283,26 +1212,25 @@ glusterd3_1_probe (call_frame_t *frame, xlator_t *this, if (ret) goto out; - uuid_copy (req.uuid, priv->uuid); + uuid_copy (req.uuid, MY_UUID); req.hostname = gf_strdup (hostname); req.port = port; ret = glusterd_submit_request (peerinfo->rpc, &req, frame, peerinfo->peer, GLUSTERD_PROBE_QUERY, - NULL, this, glusterd3_1_probe_cbk, + NULL, this, glusterd_probe_cbk, (xdrproc_t)xdr_gd1_mgmt_probe_req); out: - if (req.hostname) - GF_FREE (req.hostname); + GF_FREE (req.hostname); gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); return ret; } int32_t -glusterd3_1_friend_add (call_frame_t *frame, xlator_t *this, - void *data) +glusterd_rpc_friend_add (call_frame_t *frame, xlator_t *this, + void *data) { gd1_mgmt_friend_req req = {{0},}; int ret = 0; @@ -1328,24 +1256,23 @@ glusterd3_1_friend_add (call_frame_t *frame, xlator_t *this, if (ret) goto out; - uuid_copy (req.uuid, priv->uuid); + uuid_copy (req.uuid, MY_UUID); req.hostname = peerinfo->hostname; req.port = peerinfo->port; ret = dict_allocate_and_serialize (vols, &req.vols.vols_val, - (size_t *)&req.vols.vols_len); + &req.vols.vols_len); if (ret) goto out; ret = glusterd_submit_request (peerinfo->rpc, &req, frame, peerinfo->peer, GLUSTERD_FRIEND_ADD, - NULL, this, glusterd3_1_friend_add_cbk, + NULL, this, glusterd_friend_add_cbk, (xdrproc_t)xdr_gd1_mgmt_friend_req); out: - if (req.vols.vols_val) - GF_FREE (req.vols.vols_val); + GF_FREE (req.vols.vols_val); if (vols) dict_unref (vols); @@ -1355,8 +1282,8 @@ out: } int32_t -glusterd3_1_friend_remove (call_frame_t *frame, xlator_t *this, - void *data) +glusterd_rpc_friend_remove (call_frame_t *frame, xlator_t *this, + void *data) { gd1_mgmt_friend_req req = {{0},}; int ret = 0; @@ -1376,12 +1303,12 @@ glusterd3_1_friend_remove (call_frame_t *frame, xlator_t *this, peerinfo = event->peerinfo; - uuid_copy (req.uuid, priv->uuid); + uuid_copy (req.uuid, MY_UUID); req.hostname = peerinfo->hostname; req.port = peerinfo->port; ret = glusterd_submit_request (peerinfo->rpc, &req, frame, peerinfo->peer, GLUSTERD_FRIEND_REMOVE, NULL, - this, glusterd3_1_friend_remove_cbk, + this, glusterd_friend_remove_cbk, (xdrproc_t)xdr_gd1_mgmt_friend_req); out: @@ -1391,15 +1318,13 @@ out: int32_t -glusterd3_1_friend_update (call_frame_t *frame, xlator_t *this, - void *data) +glusterd_rpc_friend_update (call_frame_t *frame, xlator_t *this, + void *data) { gd1_mgmt_friend_update req = {{0},}; int ret = 0; glusterd_conf_t *priv = NULL; dict_t *friends = NULL; - char *dict_buf = NULL; - size_t len = -1; call_frame_t *dummy_frame = NULL; glusterd_peerinfo_t *peerinfo = NULL; @@ -1414,32 +1339,29 @@ glusterd3_1_friend_update (call_frame_t *frame, xlator_t *this, if (ret) goto out; - ret = dict_allocate_and_serialize (friends, &dict_buf, (size_t *)&len); + ret = dict_allocate_and_serialize (friends, &req.friends.friends_val, + &req.friends.friends_len); if (ret) goto out; - req.friends.friends_val = dict_buf; - req.friends.friends_len = len; - - uuid_copy (req.uuid, priv->uuid); + uuid_copy (req.uuid, MY_UUID); dummy_frame = create_frame (this, this->ctx->pool); ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame, peerinfo->peer, GLUSTERD_FRIEND_UPDATE, NULL, - this, glusterd3_1_friend_update_cbk, + this, glusterd_friend_update_cbk, (xdrproc_t)xdr_gd1_mgmt_friend_update); out: - if (req.friends.friends_val) - GF_FREE (req.friends.friends_val); + GF_FREE (req.friends.friends_val); gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); return ret; } int32_t -glusterd3_1_cluster_lock (call_frame_t *frame, xlator_t *this, +glusterd_cluster_lock (call_frame_t *frame, xlator_t *this, void *data) { gd1_mgmt_cluster_lock_req req = {{0},}; @@ -1465,15 +1387,143 @@ glusterd3_1_cluster_lock (call_frame_t *frame, xlator_t *this, ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame, peerinfo->mgmt, GLUSTERD_MGMT_CLUSTER_LOCK, NULL, - this, glusterd3_1_cluster_lock_cbk, + this, glusterd_cluster_lock_cbk, (xdrproc_t)xdr_gd1_mgmt_cluster_lock_req); out: - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_mgmt_v3_lock_peers (call_frame_t *frame, xlator_t *this, + void *data) +{ + gd1_mgmt_v3_lock_req req = {{0},}; + int ret = -1; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_conf_t *priv = NULL; + call_frame_t *dummy_frame = NULL; + dict_t *dict = NULL; + uuid_t *txn_id = NULL; + + if (!this) + goto out; + + dict = data; + + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_ptr (dict, "peerinfo", VOID (&peerinfo)); + if (ret) + goto out; + + //peerinfo should not be in payload + dict_del (dict, "peerinfo"); + + glusterd_get_uuid (&req.uuid); + + ret = dict_allocate_and_serialize (dict, &req.dict.dict_val, + &req.dict.dict_len); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to serialize dict " + "to request buffer"); + goto out; + } + + /* Sending valid transaction ID to peers */ + ret = dict_get_bin (dict, "transaction_id", + (void **)&txn_id); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get transaction id."); + goto out; + } else { + gf_log (this->name, GF_LOG_DEBUG, + "Transaction_id = %s", uuid_utoa (*txn_id)); + uuid_copy (req.txn_id, *txn_id); + } + + dummy_frame = create_frame (this, this->ctx->pool); + if (!dummy_frame) + goto out; + + ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame, + peerinfo->mgmt_v3, + GLUSTERD_MGMT_V3_LOCK, NULL, + this, glusterd_mgmt_v3_lock_peers_cbk, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_req); +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } int32_t -glusterd3_1_cluster_unlock (call_frame_t *frame, xlator_t *this, +glusterd_mgmt_v3_unlock_peers (call_frame_t *frame, xlator_t *this, + void *data) +{ + gd1_mgmt_v3_unlock_req req = {{0},}; + int ret = -1; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_conf_t *priv = NULL; + call_frame_t *dummy_frame = NULL; + dict_t *dict = NULL; + uuid_t *txn_id = NULL; + + if (!this) + goto out; + + dict = data; + + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_ptr (dict, "peerinfo", VOID (&peerinfo)); + if (ret) + goto out; + + //peerinfo should not be in payload + dict_del (dict, "peerinfo"); + + glusterd_get_uuid (&req.uuid); + + ret = dict_allocate_and_serialize (dict, &req.dict.dict_val, + &req.dict.dict_len); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to serialize dict " + "to request buffer"); + goto out; + } + + /* Sending valid transaction ID to peers */ + ret = dict_get_bin (dict, "transaction_id", + (void **)&txn_id); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get transaction id."); + goto out; + } else { + gf_log (this->name, GF_LOG_DEBUG, + "Transaction_id = %s", uuid_utoa (*txn_id)); + uuid_copy (req.txn_id, *txn_id); + } + + dummy_frame = create_frame (this, this->ctx->pool); + if (!dummy_frame) + goto out; + + ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame, + peerinfo->mgmt_v3, + GLUSTERD_MGMT_V3_UNLOCK, NULL, + this, glusterd_mgmt_v3_unlock_peers_cbk, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_req); +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_cluster_unlock (call_frame_t *frame, xlator_t *this, void *data) { gd1_mgmt_cluster_lock_req req = {{0},}; @@ -1499,15 +1549,15 @@ glusterd3_1_cluster_unlock (call_frame_t *frame, xlator_t *this, ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame, peerinfo->mgmt, GLUSTERD_MGMT_CLUSTER_UNLOCK, NULL, - this, glusterd3_1_cluster_unlock_cbk, + this, glusterd_cluster_unlock_cbk, (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_req); out: - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } int32_t -glusterd3_1_stage_op (call_frame_t *frame, xlator_t *this, +glusterd_stage_op (call_frame_t *frame, xlator_t *this, void *data) { gd1_mgmt_stage_op_req req = {{0,},}; @@ -1538,9 +1588,12 @@ glusterd3_1_stage_op (call_frame_t *frame, xlator_t *this, req.op = glusterd_op_get_op (); ret = dict_allocate_and_serialize (dict, &req.buf.buf_val, - (size_t *)&req.buf.buf_len); - if (ret) + &req.buf.buf_len); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to serialize dict " + "to request buffer"); goto out; + } dummy_frame = create_frame (this, this->ctx->pool); @@ -1550,19 +1603,19 @@ glusterd3_1_stage_op (call_frame_t *frame, xlator_t *this, ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame, peerinfo->mgmt, GLUSTERD_MGMT_STAGE_OP, NULL, - this, glusterd3_1_stage_op_cbk, + this, glusterd_stage_op_cbk, (xdrproc_t)xdr_gd1_mgmt_stage_op_req); out: if ((_gf_true == is_alloc) && req.buf.buf_val) GF_FREE (req.buf.buf_val); - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } int32_t -glusterd3_1_commit_op (call_frame_t *frame, xlator_t *this, +glusterd_commit_op (call_frame_t *frame, xlator_t *this, void *data) { gd1_mgmt_commit_op_req req = {{0,},}; @@ -1592,9 +1645,12 @@ glusterd3_1_commit_op (call_frame_t *frame, xlator_t *this, req.op = glusterd_op_get_op (); ret = dict_allocate_and_serialize (dict, &req.buf.buf_val, - (size_t *)&req.buf.buf_len); - if (ret) + &req.buf.buf_len); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to serialize dict to " + "request buffer"); goto out; + } dummy_frame = create_frame (this, this->ctx->pool); if (!dummy_frame) @@ -1603,19 +1659,19 @@ glusterd3_1_commit_op (call_frame_t *frame, xlator_t *this, ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame, peerinfo->mgmt, GLUSTERD_MGMT_COMMIT_OP, NULL, - this, glusterd3_1_commit_op_cbk, + this, glusterd_commit_op_cbk, (xdrproc_t)xdr_gd1_mgmt_commit_op_req); out: if ((_gf_true == is_alloc) && req.buf.buf_val) GF_FREE (req.buf.buf_val); - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } int32_t -glusterd3_1_brick_op_cbk (struct rpc_req *req, struct iovec *iov, +__glusterd_brick_op_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) { gd1_mgmt_brick_op_rsp rsp = {0}; @@ -1628,6 +1684,11 @@ glusterd3_1_brick_op_cbk (struct rpc_req *req, struct iovec *iov, int index = 0; glusterd_req_ctx_t *req_ctx = NULL; glusterd_pending_node_t *node = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); frame = myframe; @@ -1636,14 +1697,17 @@ glusterd3_1_brick_op_cbk (struct rpc_req *req, struct iovec *iov, if (-1 == req->rpc_status) { rsp.op_ret = -1; rsp.op_errno = EINVAL; - rsp.op_errstr = "error"; + /* use standard allocation because to keep uniformity + in freeing it */ + rsp.op_errstr = strdup ("error"); event_type = GD_OP_EVENT_RCVD_RJT; goto out; } ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_brick_op_rsp); if (ret < 0) { - gf_log ("", GF_LOG_ERROR, "error"); + gf_log (this->name, GF_LOG_ERROR, "Failed to decode brick op " + "response received"); rsp.op_ret = -1; rsp.op_errno = EINVAL; rsp.op_errstr = strdup ("Unable to decode brick op response"); @@ -1659,8 +1723,7 @@ glusterd3_1_brick_op_cbk (struct rpc_req *req, struct iovec *iov, rsp.output.output_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, - "failed to " + gf_log (this->name, GF_LOG_ERROR, "Failed to " "unserialize rsp-buffer to dictionary"); event_type = GD_OP_EVENT_RCVD_RJT; goto out; @@ -1677,7 +1740,7 @@ glusterd3_1_brick_op_cbk (struct rpc_req *req, struct iovec *iov, index = node->index; ret = dict_set_int32 (dict, "index", index); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Error setting index on brick status rsp dict"); rsp.op_ret = -1; event_type = GD_OP_EVENT_RCVD_RJT; @@ -1685,6 +1748,11 @@ glusterd3_1_brick_op_cbk (struct rpc_req *req, struct iovec *iov, } } out: + + ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + ev_ctx = GF_CALLOC (1, sizeof (*ev_ctx), gf_gld_mt_brick_rsp_ctx_t); GF_ASSERT (ev_ctx); if (op_ret) { @@ -1697,7 +1765,7 @@ out: ev_ctx->pending_node = frame->cookie; ev_ctx->rsp_dict = dict; ev_ctx->commit_ctx = frame->local; - ret = glusterd_op_sm_inject_event (event_type, ev_ctx); + ret = glusterd_op_sm_inject_event (event_type, txn_id, ev_ctx); if (!ret) { glusterd_friend_sm (); glusterd_op_sm (); @@ -1705,16 +1773,24 @@ out: if (ret && dict) dict_unref (dict); - if (rsp.op_errstr && strcmp (rsp.op_errstr, "error")) - free (rsp.op_errstr); //malloced by xdr + free (rsp.op_errstr); //malloced by xdr GLUSTERD_STACK_DESTROY (frame); return ret; } int32_t -glusterd3_1_brick_op (call_frame_t *frame, xlator_t *this, +glusterd_brick_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + __glusterd_brick_op_cbk); +} + +int32_t +glusterd_brick_op (call_frame_t *frame, xlator_t *this, void *data) { + gd1_mgmt_brick_op_req *req = NULL; int ret = 0; glusterd_conf_t *priv = NULL; @@ -1722,8 +1798,10 @@ glusterd3_1_brick_op (call_frame_t *frame, xlator_t *this, char *op_errstr = NULL; int pending_bricks = 0; glusterd_pending_node_t *pending_node; - glusterd_req_ctx_t *req_ctx = NULL; + glusterd_req_ctx_t *req_ctx = NULL; struct rpc_clnt *rpc = NULL; + dict_t *op_ctx = NULL; + uuid_t *txn_id = &global_txn_id; if (!this) { ret = -1; @@ -1735,31 +1813,72 @@ glusterd3_1_brick_op (call_frame_t *frame, xlator_t *this, req_ctx = data; GF_ASSERT (req_ctx); INIT_LIST_HEAD (&opinfo.pending_bricks); - ret = glusterd_op_bricks_select (req_ctx->op, req_ctx->dict, &op_errstr); + ret = glusterd_op_bricks_select (req_ctx->op, req_ctx->dict, &op_errstr, + &opinfo.pending_bricks, NULL); if (ret) { - gf_log ("", GF_LOG_ERROR, "Brick Op failed"); + gf_log (this->name, GF_LOG_ERROR, "Failed to select bricks " + "while performing brick op during 'Volume %s'", + gd_op_list[opinfo.op]); opinfo.op_errstr = op_errstr; goto out; } + + ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + list_for_each_entry (pending_node, &opinfo.pending_bricks, list) { dummy_frame = create_frame (this, this->ctx->pool); if (!dummy_frame) continue; - ret = glusterd_brick_op_build_payload (req_ctx->op, - pending_node->node, - (gd1_mgmt_brick_op_req **)&req, - req_ctx->dict); - if (ret) - goto out; + if ((pending_node->type == GD_NODE_NFS) || + ((pending_node->type == GD_NODE_SHD) && + (req_ctx->op == GD_OP_STATUS_VOLUME))) + ret = glusterd_node_op_build_payload + (req_ctx->op, + (gd1_mgmt_brick_op_req **)&req, + req_ctx->dict); + else { + ret = glusterd_brick_op_build_payload + (req_ctx->op, pending_node->node, + (gd1_mgmt_brick_op_req **)&req, + req_ctx->dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "build brick op payload during " + "'Volume %s'", gd_op_list[req_ctx->op]); + goto out; + } + } dummy_frame->local = data; dummy_frame->cookie = pending_node; rpc = glusterd_pending_node_get_rpc (pending_node); if (!rpc) { + if (pending_node->type == GD_NODE_REBALANCE) { + opinfo.brick_pending_count = 0; + ret = 0; + if (req) { + GF_FREE (req->input.input_val); + GF_FREE (req); + req = NULL; + } + GLUSTERD_STACK_DESTROY (dummy_frame); + + op_ctx = glusterd_op_get_ctx (); + if (!op_ctx) + goto out; + glusterd_defrag_volume_node_rsp (req_ctx->dict, + NULL, op_ctx); + + goto out; + } + ret = -1; gf_log (this->name, GF_LOG_ERROR, "Brick Op failed " "due to rpc failure."); @@ -1769,11 +1888,10 @@ glusterd3_1_brick_op (call_frame_t *frame, xlator_t *this, ret = glusterd_submit_request (rpc, req, dummy_frame, priv->gfs_mgmt, req->op, NULL, - this, glusterd3_1_brick_op_cbk, + this, glusterd_brick_op_cbk, (xdrproc_t)xdr_gd1_mgmt_brick_op_req); if (req) { - if (req->input.input_val) - GF_FREE (req->input.input_val); + GF_FREE (req->input.input_val); GF_FREE (req); req = NULL; } @@ -1781,38 +1899,46 @@ glusterd3_1_brick_op (call_frame_t *frame, xlator_t *this, pending_bricks++; } - gf_log ("glusterd", GF_LOG_DEBUG, "Sent op req to %d bricks", + gf_log (this->name, GF_LOG_DEBUG, "Sent brick op req for operation " + "'Volume %s' to %d bricks", gd_op_list[req_ctx->op], pending_bricks); opinfo.brick_pending_count = pending_bricks; out: if (ret) { - glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, data); + glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + txn_id, data); opinfo.op_ret = ret; } - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } struct rpc_clnt_procedure gd_brick_actors[GLUSTERD_BRICK_MAXVALUE] = { [GLUSTERD_BRICK_NULL] = {"NULL", NULL }, - [GLUSTERD_BRICK_OP] = {"BRICK_OP", glusterd3_1_brick_op}, + [GLUSTERD_BRICK_OP] = {"BRICK_OP", glusterd_brick_op}, }; struct rpc_clnt_procedure gd_peer_actors[GLUSTERD_FRIEND_MAXVALUE] = { [GLUSTERD_FRIEND_NULL] = {"NULL", NULL }, - [GLUSTERD_PROBE_QUERY] = {"PROBE_QUERY", glusterd3_1_probe}, - [GLUSTERD_FRIEND_ADD] = {"FRIEND_ADD", glusterd3_1_friend_add}, - [GLUSTERD_FRIEND_REMOVE] = {"FRIEND_REMOVE", glusterd3_1_friend_remove}, - [GLUSTERD_FRIEND_UPDATE] = {"FRIEND_UPDATE", glusterd3_1_friend_update}, + [GLUSTERD_PROBE_QUERY] = {"PROBE_QUERY", glusterd_rpc_probe}, + [GLUSTERD_FRIEND_ADD] = {"FRIEND_ADD", glusterd_rpc_friend_add}, + [GLUSTERD_FRIEND_REMOVE] = {"FRIEND_REMOVE", glusterd_rpc_friend_remove}, + [GLUSTERD_FRIEND_UPDATE] = {"FRIEND_UPDATE", glusterd_rpc_friend_update}, }; struct rpc_clnt_procedure gd_mgmt_actors[GLUSTERD_MGMT_MAXVALUE] = { [GLUSTERD_MGMT_NULL] = {"NULL", NULL }, - [GLUSTERD_MGMT_CLUSTER_LOCK] = {"CLUSTER_LOCK", glusterd3_1_cluster_lock}, - [GLUSTERD_MGMT_CLUSTER_UNLOCK] = {"CLUSTER_UNLOCK", glusterd3_1_cluster_unlock}, - [GLUSTERD_MGMT_STAGE_OP] = {"STAGE_OP", glusterd3_1_stage_op}, - [GLUSTERD_MGMT_COMMIT_OP] = {"COMMIT_OP", glusterd3_1_commit_op}, + [GLUSTERD_MGMT_CLUSTER_LOCK] = {"CLUSTER_LOCK", glusterd_cluster_lock}, + [GLUSTERD_MGMT_CLUSTER_UNLOCK] = {"CLUSTER_UNLOCK", glusterd_cluster_unlock}, + [GLUSTERD_MGMT_STAGE_OP] = {"STAGE_OP", glusterd_stage_op}, + [GLUSTERD_MGMT_COMMIT_OP] = {"COMMIT_OP", glusterd_commit_op}, +}; + +struct rpc_clnt_procedure gd_mgmt_v3_actors[GLUSTERD_MGMT_V3_MAXVALUE] = { + [GLUSTERD_MGMT_V3_NULL] = {"NULL", NULL }, + [GLUSTERD_MGMT_V3_LOCK] = {"MGMT_V3_LOCK", glusterd_mgmt_v3_lock_peers}, + [GLUSTERD_MGMT_V3_UNLOCK] = {"MGMT_V3_UNLOCK", glusterd_mgmt_v3_unlock_peers}, }; struct rpc_clnt_program gd_mgmt_prog = { @@ -1839,3 +1965,10 @@ struct rpc_clnt_program gd_peer_prog = { .numproc = GLUSTERD_FRIEND_MAXVALUE, }; +struct rpc_clnt_program gd_mgmt_v3_prog = { + .progname = "glusterd mgmt v3", + .prognum = GD_MGMT_V3_PROGRAM, + .progver = GD_MGMT_V3_VERSION, + .proctable = gd_mgmt_v3_actors, + .numproc = GLUSTERD_MGMT_V3_MAXVALUE, +}; diff --git a/xlators/mgmt/glusterd/src/glusterd-sm.c b/xlators/mgmt/glusterd/src/glusterd-sm.c index 259c247f4..c671edf68 100644 --- a/xlators/mgmt/glusterd/src/glusterd-sm.c +++ b/xlators/mgmt/glusterd/src/glusterd-sm.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H @@ -100,8 +90,7 @@ glusterd_destroy_probe_ctx (glusterd_probe_ctx_t *ctx) if (!ctx) return; - if (ctx->hostname) - GF_FREE (ctx->hostname); + GF_FREE (ctx->hostname); GF_FREE (ctx); } @@ -113,8 +102,7 @@ glusterd_destroy_friend_req_ctx (glusterd_friend_req_ctx_t *ctx) if (ctx->vols) dict_unref (ctx->vols); - if (ctx->hostname) - GF_FREE (ctx->hostname); + GF_FREE (ctx->hostname); GF_FREE (ctx); } @@ -123,8 +111,7 @@ glusterd_destroy_friend_update_ctx (glusterd_friend_update_ctx_t *ctx) { if (!ctx) return; - if (ctx->hostname) - GF_FREE (ctx->hostname); + GF_FREE (ctx->hostname); GF_FREE (ctx); } @@ -257,12 +244,9 @@ glusterd_ac_reverse_probe_begin (glusterd_friend_sm_event_t *event, void *ctx) out: if (ret) { - if (new_event) - GF_FREE (new_event); - if (new_ev_ctx->hostname) - GF_FREE (new_ev_ctx->hostname); - if (new_ev_ctx) - GF_FREE (new_ev_ctx); + GF_FREE (new_event); + GF_FREE (new_ev_ctx->hostname); + GF_FREE (new_ev_ctx); } gf_log ("", GF_LOG_DEBUG, "returning with %d", ret); return ret; @@ -415,7 +399,9 @@ glusterd_ac_send_friend_remove_req (glusterd_friend_sm_event_t *event, if (ctx) ret = glusterd_xfer_cli_deprobe_resp (ctx->req, ret, 0, - ctx->hostname); + NULL, + ctx->hostname, + ctx->dict); glusterd_friend_sm (); glusterd_op_sm (); @@ -444,22 +430,36 @@ out: return ret; } +static gf_boolean_t +glusterd_should_update_peer (glusterd_peerinfo_t *peerinfo, + glusterd_peerinfo_t *cur_peerinfo) +{ + gf_boolean_t is_valid = _gf_false; + + if ((peerinfo == cur_peerinfo) || + (peerinfo->state.state == GD_FRIEND_STATE_BEFRIENDED)) + is_valid = _gf_true; + + return is_valid; +} + static int glusterd_ac_send_friend_update (glusterd_friend_sm_event_t *event, void *ctx) { - int ret = 0; - glusterd_peerinfo_t *peerinfo = NULL; - rpc_clnt_procedure_t *proc = NULL; - xlator_t *this = NULL; - glusterd_friend_update_ctx_t ev_ctx = {{0}}; - glusterd_conf_t *priv = NULL; - dict_t *friends = NULL; - char key[100] = {0,}; - char *dup_buf = NULL; - int32_t count = 0; + int ret = 0; + glusterd_peerinfo_t *cur_peerinfo = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + rpc_clnt_procedure_t *proc = NULL; + xlator_t *this = NULL; + glusterd_friend_update_ctx_t ev_ctx = {{0}}; + glusterd_conf_t *priv = NULL; + dict_t *friends = NULL; + char key[100] = {0,}; + char *dup_buf = NULL; + int32_t count = 0; GF_ASSERT (event); - peerinfo = event->peerinfo; + cur_peerinfo = event->peerinfo; this = THIS; priv = this->private; @@ -478,6 +478,9 @@ glusterd_ac_send_friend_update (glusterd_friend_sm_event_t *event, void *ctx) goto out; list_for_each_entry (peerinfo, &priv->peers, uuid_list) { + if (!glusterd_should_update_peer (peerinfo, cur_peerinfo)) + continue; + count++; snprintf (key, sizeof (key), "friend%d.uuid", count); dup_buf = gf_strdup (uuid_utoa (peerinfo->uuid)); @@ -500,6 +503,9 @@ glusterd_ac_send_friend_update (glusterd_friend_sm_event_t *event, void *ctx) if (!peerinfo->connected || !peerinfo->peer) continue; + if (!glusterd_should_update_peer (peerinfo, cur_peerinfo)) + continue; + ret = dict_set_static_ptr (friends, "peerinfo", peerinfo); if (ret) { gf_log ("", GF_LOG_ERROR, "failed to set peerinfo"); @@ -533,7 +539,7 @@ glusterd_peer_detach_cleanup (glusterd_conf_t *priv) list_for_each_entry_safe (volinfo,tmp_volinfo, &priv->volumes, vol_list) { if (!glusterd_friend_contains_vol_bricks (volinfo, - priv->uuid)) { + MY_UUID)) { gf_log (THIS->name, GF_LOG_INFO, "Deleting stale volume %s", volinfo->volname); ret = glusterd_delete_volume (volinfo); @@ -645,7 +651,8 @@ glusterd_ac_handle_friend_add_req (glusterd_friend_sm_event_t *event, void *ctx) uuid_copy (peerinfo->uuid, ev_ctx->uuid); //Build comparison logic here. - ret = glusterd_compare_friend_data (ev_ctx->vols, &status); + ret = glusterd_compare_friend_data (ev_ctx->vols, &status, + peerinfo->hostname); if (ret) goto out; @@ -682,7 +689,8 @@ glusterd_ac_handle_friend_add_req (glusterd_friend_sm_event_t *event, void *ctx) glusterd_friend_sm_inject_event (new_event); ret = glusterd_xfer_friend_add_resp (ev_ctx->req, ev_ctx->hostname, - ev_ctx->port, op_ret, op_errno); + peerinfo->hostname, ev_ctx->port, + op_ret, op_errno); out: gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); @@ -954,6 +962,27 @@ glusterd_destroy_friend_event_context (glusterd_friend_sm_event_t *event) } } +gf_boolean_t +gd_does_peer_affect_quorum (glusterd_friend_sm_state_t old_state, + glusterd_friend_sm_event_type_t event_type, + glusterd_peerinfo_t *peerinfo) +{ + gf_boolean_t affects = _gf_false; + + //When glusterd comes up with friends in BEFRIENDED state in store, + //wait until compare-data happens. + if ((old_state == GD_FRIEND_STATE_BEFRIENDED) && + (event_type != GD_FRIEND_EVENT_RCVD_ACC) && + (event_type != GD_FRIEND_EVENT_LOCAL_ACC)) + goto out; + if ((peerinfo->state.state == GD_FRIEND_STATE_BEFRIENDED) + && peerinfo->connected) { + affects = _gf_true; + } +out: + return affects; +} + int glusterd_friend_sm () { @@ -965,6 +994,8 @@ glusterd_friend_sm () glusterd_peerinfo_t *peerinfo = NULL; glusterd_friend_sm_event_type_t event_type = 0; gf_boolean_t is_await_conn = _gf_false; + gf_boolean_t quorum_action = _gf_false; + glusterd_friend_sm_state_t old_state = GD_FRIEND_STATE_DEFAULT; while (!list_empty (&gd_friend_sm_queue)) { list_for_each_entry_safe (event, tmp, &gd_friend_sm_queue, list) { @@ -984,6 +1015,7 @@ glusterd_friend_sm () glusterd_friend_sm_event_name_get (event_type)); + old_state = peerinfo->state.state; state = glusterd_friend_state_table[peerinfo->state.state]; GF_ASSERT (state); @@ -1024,6 +1056,15 @@ glusterd_friend_sm () goto out; } + if (gd_does_peer_affect_quorum (old_state, event_type, + peerinfo)) { + peerinfo->quorum_contrib = QUORUM_UP; + if (peerinfo->quorum_action) { + peerinfo->quorum_action = _gf_false; + quorum_action = _gf_true; + } + } + ret = glusterd_store_peerinfo (peerinfo); glusterd_destroy_friend_event_context (event); @@ -1037,6 +1078,25 @@ glusterd_friend_sm () ret = 0; out: + if (quorum_action) { + /* When glusterd is restarted, it needs to wait until the 'friends' view + * of the volumes settle, before it starts any of the internal daemons. + * + * Every friend that was part of the cluster, would send its + * cluster-view, 'our' way. For every friend, who belongs to + * a partition which has a different cluster-view from our + * partition, we may update our cluster-view. For subsequent + * friends from that partition would agree with us, if the first + * friend wasn't rejected. For every first friend, whom we agreed with, + * we would need to start internal daemons/bricks belonging to the + * new volumes. + * glusterd_spawn_daemons calls functions that are idempotent. ie, + * the functions spawn process(es) only if they are not started yet. + * + * */ + glusterd_spawn_daemons (NULL); + glusterd_do_quorum_action (); + } return ret; } diff --git a/xlators/mgmt/glusterd/src/glusterd-sm.h b/xlators/mgmt/glusterd/src/glusterd-sm.h index 7baaf0236..b9bedbe69 100644 --- a/xlators/mgmt/glusterd/src/glusterd-sm.h +++ b/xlators/mgmt/glusterd/src/glusterd-sm.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _GLUSTERD_SM_H_ #define _GLUSTERD_SM_H_ @@ -37,14 +27,21 @@ #include "byte-order.h" //#include "glusterd.h" #include "rpcsvc.h" - -struct glusterd_store_handle_ { - char *path; - int fd; - FILE *read; -}; - -typedef struct glusterd_store_handle_ glusterd_store_handle_t; +#include "store.h" + +typedef enum gd_quorum_contribution_ { + QUORUM_NONE, + QUORUM_WAITING, + QUORUM_DOWN, + QUORUM_UP +} gd_quorum_contrib_t; + +typedef enum gd_quorum_status_ { + QUORUM_UNKNOWN, + QUORUM_NOT_APPLICABLE, + QUORUM_MEETS, + QUORUM_DOES_NOT_MEET +} gd_quorum_status_t; typedef enum glusterd_friend_sm_state_ { GD_FRIEND_STATE_DEFAULT = 0, @@ -89,7 +86,9 @@ typedef struct glusterd_sm_tr_log_ { struct glusterd_peerinfo_ { uuid_t uuid; - char uuid_str[50]; + char uuid_str[50]; /* Retrieve this using + * gd_peer_uuid_str () + */ glusterd_peer_state_info_t state; char *hostname; int port; @@ -98,9 +97,13 @@ struct glusterd_peerinfo_ { struct rpc_clnt *rpc; rpc_clnt_prog_t *mgmt; rpc_clnt_prog_t *peer; + rpc_clnt_prog_t *mgmt_v3; int connected; - glusterd_store_handle_t *shandle; + gf_store_handle_t *shandle; glusterd_sm_tr_log_t sm_log; + gf_boolean_t quorum_action; + gd_quorum_contrib_t quorum_contrib; + gf_boolean_t locked; }; typedef struct glusterd_peerinfo_ glusterd_peerinfo_t; @@ -114,11 +117,13 @@ typedef enum glusterd_ev_gen_mode_ { typedef struct glusterd_peer_ctx_args_ { rpcsvc_request_t *req; glusterd_ev_gen_mode_t mode; + dict_t *dict; } glusterd_peerctx_args_t; typedef struct glusterd_peer_ctx_ { glusterd_peerctx_args_t args; glusterd_peerinfo_t *peerinfo; + char *errstr; } glusterd_peerctx_t; typedef enum glusterd_friend_sm_event_type_ { @@ -179,6 +184,7 @@ typedef struct glusterd_probe_ctx_ { char *hostname; rpcsvc_request_t *req; int port; + dict_t *dict; } glusterd_probe_ctx_t; int glusterd_friend_sm_new_event (glusterd_friend_sm_event_type_t event_type, diff --git a/xlators/mgmt/glusterd/src/glusterd-snapshot.c b/xlators/mgmt/glusterd/src/glusterd-snapshot.c new file mode 100644 index 000000000..9b811cd05 --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-snapshot.c @@ -0,0 +1,5590 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include <inttypes.h> +#include <sys/types.h> +#include <unistd.h> +#include <sys/resource.h> +#include <sys/statvfs.h> +#include <sys/mount.h> + +#include "globals.h" +#include "compat.h" +#include "protocol-common.h" +#include "xlator.h" +#include "logging.h" +#include "timer.h" +#include "glusterd-mem-types.h" +#include "glusterd.h" +#include "glusterd-sm.h" +#include "glusterd-op-sm.h" +#include "glusterd-utils.h" +#include "glusterd-store.h" +#include "run.h" +#include "glusterd-volgen.h" +#include "glusterd-mgmt.h" +#include "glusterd-syncop.h" + +#include "syscall.h" +#include "cli1-xdr.h" +#include "xdr-generic.h" + +#ifdef GF_LINUX_HOST_OS +#include <mntent.h> +#endif + +char snap_mount_folder[PATH_MAX]; + +static int32_t +glusterd_find_missed_snap (dict_t *rsp_dict, glusterd_volinfo_t *vol, + char *snap_uuid, struct list_head *peers, + int32_t op); + +/* This function will restore a snapshot volumes + * + * @param dict dictionary containing snapshot restore request + * @param op_errstr In case of any failure error message will be returned + * in this variable + * @return Negative value on Failure and 0 in success + */ +int +glusterd_snapshot_restore (dict_t *dict, char **op_errstr, dict_t *rsp_dict) +{ + int ret = -1; + char *volname = NULL; + char *snapname = NULL; + xlator_t *this = NULL; + glusterd_volinfo_t *snap_volinfo = NULL; + glusterd_volinfo_t *volinfo = NULL; + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "snap name"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (NULL == snap) { + ret = gf_asprintf (op_errstr, "Snap (%s) not found", + snapname); + if (ret < 0) { + goto out; + } + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); + ret = -1; + goto out; + } + + /* TODO : As of now there is only volume in snapshot. + * Change this when multiple volume snapshot is introduced + */ + snap_volinfo = list_entry (snap->volumes.next, glusterd_volinfo_t, + vol_list); + + ret = glusterd_volinfo_find (snap_volinfo->parent_volname, &volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not get volinfo of " + "%s", snap_volinfo->parent_volname); + goto out; + } + + if (is_origin_glusterd (dict) == _gf_true) { + /* From origin glusterd check if * + * any peers with snap bricks is down */ + ret = glusterd_find_missed_snap (rsp_dict, snap_volinfo, + snap_volinfo->volname, + &priv->peers, + GF_SNAP_OPTION_TYPE_RESTORE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to find missed snap restores"); + goto out; + } + } + + ret = gd_restore_snap_volume (rsp_dict, volinfo, snap_volinfo); + if (ret) { + /* No need to update op_errstr because it is assumed + * that the called function will do that in case of + * failure. + */ + gf_log (this->name, GF_LOG_ERROR, "Failed to restore " + "snap for %s volume", volname); + goto out; + } + + ret = 0; + + /* TODO: Need to check if we need to delete the snap after the + * operation is successful or not. Also need to persist the state + * of restore operation in the store. + */ +out: + return ret; +} + +/* This function is called before actual restore is taken place. This function + * will validate whether the snapshot volumes are ready to be restored or not. + * + * @param dict dictionary containing snapshot restore request + * @param op_errstr In case of any failure error message will be returned + * in this variable + * @param rsp_dict response dictionary + * @return Negative value on Failure and 0 in success + */ +int +glusterd_snapshot_restore_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int ret = -1; + int32_t i = 0; + int32_t volcount = 0; + gf_boolean_t snap_restored = _gf_false; + char key[PATH_MAX] = {0, }; + char *volname = NULL; + char *snapname = NULL; + glusterd_volinfo_t *volinfo = NULL; + glusterd_snap_t *snap = NULL; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "snap name"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (NULL == snap) { + ret = gf_asprintf (op_errstr, "Snap (%s) not found", + snapname); + if (ret < 0) { + goto out; + } + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); + ret = -1; + goto out; + } + + snap_restored = snap->snap_restored; + + if (snap_restored) { + ret = gf_asprintf (op_errstr, "Snap (%s) is already " + "restored", snapname); + if (ret < 0) { + goto out; + } + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); + ret = -1; + goto out; + } + + ret = dict_set_str (rsp_dict, "snapname", snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "snap name"); + goto out; + } + + ret = dict_get_int32 (dict, "volcount", &volcount); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volume count"); + goto out; + } + + /* Snapshot restore will only work if all the volumes, + that are part of the snapshot, are stopped. */ + for (i = 1; i <= volcount; ++i) { + snprintf (key, sizeof (key), "volname%d", i); + ret = dict_get_str (dict, key, &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "get volume name"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + ret = gf_asprintf (op_errstr, "Volume (%s) not found", + volname); + if (ret < 0) { + goto out; + } + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); + ret = -1; + goto out; + } + + if (glusterd_is_volume_started (volinfo)) { + ret = gf_asprintf (op_errstr, "Volume (%s) has been " + "started. Volume needs to be stopped before restoring " + "a snapshot.", volname); + if (ret < 0) { + goto out; + } + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); + ret = -1; + goto out; + } + } + + ret = 0; +out: + return ret; +} + +int +snap_max_hard_limits_validate (dict_t *dict, char *volname, + uint64_t value, char **op_errstr) +{ + char err_str[PATH_MAX] = ""; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = -1; + uint64_t max_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + + if (volname) { + ret = glusterd_volinfo_find (volname, &volinfo); + if (!ret) { + if (volinfo->is_snap_volume) { + ret = -1; + snprintf (err_str, PATH_MAX, + "%s is a snap volume. Configuring " + "snap-max-hard-limit for a snap " + "volume is prohibited.", volname); + goto out; + } + } + } + + if (value) { + /* Max limit for the system is GLUSTERD_SNAPS_MAX_HARD_LIMIT + * but max limit for a volume is conf->snap_max_hard_limit. + */ + if (volname) { + max_limit = conf->snap_max_hard_limit; + } else { + max_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT; + } + } + + if ((value < 0) || (value > max_limit)) { + ret = -1; + snprintf (err_str, PATH_MAX, "Invalid snap-max-hard-limit" + "%"PRIu64 ". Expected range 0 - %"PRIu64, + value, max_limit); + goto out; + } + + ret = 0; +out: + if (ret) { + *op_errstr = gf_strdup (err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + } + return ret; +} + +int +glusterd_snapshot_config_prevalidate (dict_t *dict, char **op_errstr) +{ + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + xlator_t *this = NULL; + int ret = -1; + int config_command = 0; + char err_str[PATH_MAX] = {0,}; + glusterd_conf_t *conf = NULL; + uint64_t value = 0; + uint64_t hard_limit = 0; + uint64_t soft_limit = 0; + gf_loglevel_t loglevel = GF_LOG_ERROR; + uint64_t max_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + + ret = dict_get_int32 (dict, "config-command", &config_command); + if (ret) { + snprintf (err_str, sizeof (err_str), + "failed to get config-command type"); + goto out; + } + + ret = dict_get_uint64 (dict, "snap-max-hard-limit", &hard_limit); + + ret = dict_get_uint64 (dict, "snap-max-soft-limit", &soft_limit); + + ret = dict_get_str (dict, "volname", &volname); + + if (volname) { + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, sizeof (err_str), + "Volume %s does not exist.", volname); + goto out; + } + } + + switch (config_command) { + case GF_SNAP_CONFIG_TYPE_SET: + if (hard_limit) { + /* Validations for snap-max-hard-limits */ + ret = snap_max_hard_limits_validate (dict, volname, + hard_limit, op_errstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "snap-max-hard-limit validation " + "failed."); + goto out; + } + } + + if (soft_limit) { + max_limit = GLUSTERD_SNAPS_MAX_SOFT_LIMIT_PERCENT; + if ((soft_limit < 0) || (soft_limit > max_limit)) { + ret = -1; + snprintf (err_str, PATH_MAX, "Invalid " + "snap-max-soft-limit ""%" + PRIu64 ". Expected range 0 - %"PRIu64, + value, max_limit); + goto out; + } + break; + } + default: + break; + } + + ret = 0; +out: + + if (ret && err_str[0] != '\0') { + gf_log (this->name, loglevel, "%s", err_str); + *op_errstr = gf_strdup (err_str); + } + + return ret; +} + +int +glusterd_snap_create_pre_val_use_rsp_dict (dict_t *dst, dict_t *src) +{ + char *snap_brick_dir = NULL; + char *snap_device = NULL; + char *tmpstr = NULL; + char key[PATH_MAX] = ""; + char snapbrckcnt[PATH_MAX] = ""; + char snapbrckord[PATH_MAX] = ""; + int ret = -1; + int64_t i = -1; + int64_t j = -1; + int64_t volume_count = 0; + int64_t brick_count = 0; + int64_t brick_order = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dst); + GF_ASSERT (src); + + ret = dict_get_int64 (src, "volcount", &volume_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "get the volume count"); + goto out; + } + + for (i = 0; i < volume_count; i++) { + memset (snapbrckcnt, '\0', sizeof(snapbrckcnt)); + ret = snprintf (snapbrckcnt, sizeof(snapbrckcnt) - 1, + "vol%ld_brickcount", i+1); + ret = dict_get_int64 (src, snapbrckcnt, &brick_count); + if (ret) { + gf_log (this->name, GF_LOG_TRACE, + "No bricks for this volume in this dict"); + continue; + } + + for (j = 0; j < brick_count; j++) { + /* Fetching data from source dict */ + snprintf (key, sizeof(key) - 1, + "vol%ld.brickdir%ld", i+1, j); + + ret = dict_get_ptr (src, key, + (void **)&snap_brick_dir); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Unable to fetch %s", key); + continue; + } + + snprintf (key, sizeof(key) - 1, + "vol%ld.brick_snapdevice%ld", i+1, j); + + ret = dict_get_ptr (src, key, + (void **)&snap_device); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch snap_device"); + goto out; + } + + snprintf (snapbrckord, sizeof(snapbrckord) - 1, + "vol%ld.brick%ld.order", i+1, j); + + ret = dict_get_int64 (src, snapbrckord, &brick_order); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get brick order"); + goto out; + } + + /* Adding the data in the dst dict */ + snprintf (key, sizeof(key) - 1, + "vol%ld.brickdir%ld", i+1, brick_order); + + tmpstr = gf_strdup (snap_brick_dir); + if (!tmpstr) { + gf_log (this->name, GF_LOG_ERROR, + "Out Of Memory"); + ret = -1; + goto out; + } + ret = dict_set_dynstr (dst, key, tmpstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set %s", key); + GF_FREE (tmpstr); + goto out; + } + + snprintf (key, sizeof(key) - 1, + "vol%ld.brick_snapdevice%ld", + i+1, brick_order); + + tmpstr = gf_strdup (snap_device); + if (!tmpstr) { + ret = -1; + goto out; + } + ret = dict_set_dynstr (dst, key, tmpstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set %s", key); + GF_FREE (tmpstr); + goto out; + } + + } + } + + ret = 0; +out: + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_snap_pre_validate_use_rsp_dict (dict_t *dst, dict_t *src) +{ + int ret = -1; + int32_t snap_command = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + if (!dst || !src) { + gf_log (this->name, GF_LOG_ERROR, "Source or Destination " + "dict is empty."); + goto out; + } + + ret = dict_get_int32 (dst, "type", &snap_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case GF_SNAP_OPTION_TYPE_CREATE: + ret = glusterd_snap_create_pre_val_use_rsp_dict (dst, src); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to use " + "rsp dict"); + goto out; + } + break; + default: + break; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterd_snapshot_create_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + char *volname = NULL; + char *snapname = NULL; + char *device = NULL; + char *tmpstr = NULL; + char *brick_dir = NULL; + char snap_brick_dir[PATH_MAX] = ""; + char *mnt_pt = NULL; + char key[PATH_MAX] = ""; + char snap_mount[PATH_MAX] = ""; + char snap_volname[64] = ""; + char err_str[PATH_MAX] = ""; + int ret = -1; + int64_t i = 0; + int64_t volcount = 0; + int64_t brick_count = 0; + int64_t brick_order = 0; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_volinfo_t *volinfo = NULL; + xlator_t *this = NULL; + uuid_t *snap_volid = NULL; + gf_loglevel_t loglevel = GF_LOG_ERROR; + glusterd_conf_t *conf = NULL; + int64_t effective_max_limit = 0; + + this = THIS; + GF_ASSERT (op_errstr); + conf = this->private; + GF_ASSERT (conf); + + ret = dict_get_int64 (dict, "volcount", &volcount); + if (ret) { + snprintf (err_str, sizeof (err_str), "Failed to " + "get the volume count"); + goto out; + } + if (volcount <= 0) { + snprintf (err_str, sizeof (err_str), "Invalid volume count %ld " + "supplied", volcount); + ret = -1; + goto out; + } + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + snprintf (err_str, sizeof (err_str), "Failed to get snapname"); + goto out; + } + + if (glusterd_find_snap_by_name (snapname)) { + ret = -1; + snprintf (err_str, sizeof (err_str), "Snap %s already exists", + snapname); + goto out; + } + + for (i = 1; i <= volcount; i++) { + snprintf (key, sizeof (key), "volname%ld", i); + ret = dict_get_str (dict, key, &volname); + if (ret) { + snprintf (err_str, sizeof (err_str), + "failed to get volume name"); + goto out; + } + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, sizeof (err_str), + "Volume (%s) does not exist ", volname); + goto out; + } + + ret = -1; + if (!glusterd_is_volume_started (volinfo)) { + snprintf (err_str, sizeof (err_str), "volume %s is " + "not started", volinfo->volname); + loglevel = GF_LOG_WARNING; + goto out; + } + if (glusterd_is_defrag_on (volinfo)) { + snprintf (err_str, sizeof (err_str), + "rebalance process is running for the " + "volume %s", volname); + loglevel = GF_LOG_WARNING; + goto out; + } + /* TODO: Also check whether geo replication is running */ + + if (volinfo->is_snap_volume == _gf_true) { + snprintf (err_str, sizeof (err_str), + "Volume %s is a snap volume", volname); + loglevel = GF_LOG_WARNING; + goto out; + } + + if (volinfo->snap_max_hard_limit < conf->snap_max_hard_limit) + effective_max_limit = volinfo->snap_max_hard_limit; + else + effective_max_limit = conf->snap_max_hard_limit; + + if (volinfo->snap_count >= effective_max_limit) { + snprintf (err_str, sizeof (err_str), + "The number of existing snaps has reached " + "the effective maximum limit of %"PRIu64" ," + "for the volume %s", effective_max_limit, + volname); + loglevel = GF_LOG_WARNING; + goto out; + } + + snprintf (key, sizeof(key) - 1, "vol%ld_volid", i); + ret = dict_get_bin (dict, key, (void **)&snap_volid); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch snap_volid"); + goto out; + } + + /* snap volume uuid is used as lvm snapshot name. + This will avoid restrictions on snapshot names + provided by user */ + GLUSTERD_GET_UUID_NOHYPHEN (snap_volname, *snap_volid); + + brick_count = 0; + brick_order = 0; + /* Adding snap bricks mount paths to the dict */ + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (uuid_compare (brickinfo->uuid, MY_UUID)) { + brick_order++; + continue; + } + + if (!glusterd_is_brick_started (brickinfo)) { + gf_log (this->name, GF_LOG_WARNING, + "brick %s:%s is not started", + brickinfo->hostname, + brickinfo->path); + brick_order++; + brick_count++; + continue; + } + + device = glusterd_get_brick_mount_details (brickinfo); + if (!device) { + snprintf (err_str, sizeof (err_str), + "getting device name for the brick " + "%s:%s failed", brickinfo->hostname, + brickinfo->path); + ret = -1; + goto out; + } + + device = glusterd_build_snap_device_path (device, + snap_volname); + if (!device) { + snprintf (err_str, sizeof (err_str), + "cannot copy the snapshot device " + "name (volname: %s, snapname: %s)", + volinfo->volname, snapname); + loglevel = GF_LOG_WARNING; + ret = -1; + goto out; + } + + snprintf (key, sizeof(key), + "vol%ld.brick_snapdevice%ld", i, + brick_count); + ret = dict_set_dynstr (rsp_dict, key, device); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set %s", key); + GF_FREE (device); + goto out; + } + + ret = glusterd_get_brick_root (brickinfo->path, + &mnt_pt); + if (ret) { + snprintf (err_str, sizeof (err_str), + "could not get the root of the brick path %s", + brickinfo->path); + loglevel = GF_LOG_WARNING; + goto out; + } + if (strncmp (brickinfo->path, mnt_pt, strlen(mnt_pt))) { + snprintf (err_str, sizeof (err_str), + "brick: %s brick mount: %s", + brickinfo->path, mnt_pt); + loglevel = GF_LOG_WARNING; + goto out; + } + + brick_dir = &brickinfo->path[strlen (mnt_pt)]; + brick_dir++; + + snprintf (snap_brick_dir, sizeof (snap_brick_dir), + "/%s", brick_dir); + + tmpstr = gf_strdup (snap_brick_dir); + if (!tmpstr) { + ret = -1; + goto out; + } + snprintf (key, sizeof(key), "vol%ld.brickdir%ld", i, + brick_count); + ret = dict_set_dynstr (rsp_dict, key, tmpstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set %s", snap_mount); + goto out; + } + tmpstr = NULL; + + snprintf (key, sizeof(key) - 1, "vol%ld.brick%ld.order", + i, brick_count); + ret = dict_set_int64 (rsp_dict, key, brick_order); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set %s", key); + goto out; + } + + brick_count++; + brick_order++; + } + snprintf (key, sizeof(key) - 1, "vol%ld_brickcount", i); + ret = dict_set_int64 (rsp_dict, key, brick_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set %s", + key); + goto out; + } + } + + ret = dict_set_int64 (rsp_dict, "volcount", volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set volcount"); + goto out; + } + + ret = 0; +out: + if (ret) + GF_FREE (tmpstr); + + if (ret && err_str[0] != '\0') { + gf_log (this->name, loglevel, "%s", err_str); + *op_errstr = gf_strdup (err_str); + } + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +glusterd_snap_t* +glusterd_new_snap_object() +{ + glusterd_snap_t *snap = NULL; + + snap = GF_CALLOC (1, sizeof (*snap), gf_gld_mt_snap_t); + + if (snap) { + if (LOCK_INIT (&snap->lock)) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed initiating" + " snap lock"); + GF_FREE (snap); + return NULL; + } + + INIT_LIST_HEAD (&snap->snap_list); + INIT_LIST_HEAD (&snap->volumes); + snap->snapname[0] = 0; + snap->snap_status = GD_SNAP_STATUS_INIT; + } + + return snap; + +}; + +/* Function glusterd_list_add_snapvol adds the volinfo object (snapshot volume) + to the snapshot object list and to the parent volume list */ +int32_t +glusterd_list_add_snapvol (glusterd_volinfo_t *origin_vol, + glusterd_volinfo_t *snap_vol) +{ + int ret = -1; + glusterd_snap_t *snap = NULL; + + GF_VALIDATE_OR_GOTO ("glusterd", origin_vol, out); + GF_VALIDATE_OR_GOTO ("glusterd", snap_vol, out); + + snap = snap_vol->snapshot; + GF_ASSERT (snap); + + list_add_tail (&snap_vol->vol_list, &snap->volumes); + LOCK (&origin_vol->lock); + { + list_add_order (&snap_vol->snapvol_list, + &origin_vol->snap_volumes, + glusterd_compare_snap_vol_time); + origin_vol->snap_count++; + } + UNLOCK (&origin_vol->lock); + + gf_log (THIS->name, GF_LOG_DEBUG, "Snap %s added to the list", + snap->snapname); + ret = 0; +out: + return ret; +} + +glusterd_snap_t* +glusterd_find_snap_by_name (char *snapname) +{ + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + GF_ASSERT (snapname); + + + list_for_each_entry (snap, &priv->snapshots, snap_list) { + if (!strcmp (snap->snapname, snapname)) { + gf_log (THIS->name, GF_LOG_DEBUG, "Found " + "snap %s (%s)", snap->snapname, + uuid_utoa (snap->snap_id)); + goto out; + } + } + snap = NULL; +out: + return snap; +} + +glusterd_snap_t* +glusterd_find_snap_by_id (uuid_t snap_id) +{ + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + + if (uuid_is_null(snap_id)) + goto out; + + list_for_each_entry (snap, &priv->snapshots, snap_list) { + if (!uuid_compare (snap->snap_id, snap_id)) { + gf_log (THIS->name, GF_LOG_DEBUG, "Found " + "snap %s (%s)", snap->snapname, + uuid_utoa (snap->snap_id)); + goto out; + } + } + snap = NULL; +out: + return snap; +} + +int +glusterd_do_lvm_snapshot_remove (glusterd_volinfo_t *snap_vol, + glusterd_brickinfo_t *brickinfo, + const char *mount_pt, const char *snap_device) +{ + int ret = -1; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + runner_t runner = {0,}; + char msg[1024] = {0, }; + char pidfile[PATH_MAX] = {0, }; + pid_t pid = -1; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + if (!brickinfo) { + gf_log (this->name, GF_LOG_ERROR, "brickinfo NULL"); + goto out; + } + + GF_ASSERT (snap_vol); + GF_ASSERT (mount_pt); + GF_ASSERT (snap_device); + + GLUSTERD_GET_BRICK_PIDFILE (pidfile, snap_vol, brickinfo, priv); + if (glusterd_is_service_running (pidfile, &pid)) { + ret = kill (pid, SIGKILL); + if (ret && errno != ESRCH) { + gf_log (this->name, GF_LOG_ERROR, "Unable to kill pid " + "%d reason : %s", pid, strerror(errno)); + goto out; + } + } + + runinit (&runner); + snprintf (msg, sizeof (msg), "umount the snapshot mounted path %s", + mount_pt); + runner_add_args (&runner, "umount", mount_pt, NULL); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + + /* We need not do synclock_unlock => runner_run => synclock_lock here. + Because it is needed if we are running a glusterfs process in + runner_run, so that when the glusterfs process started wants to + communicate to glusterd, glusterd wont be able to respond if it + has held the big lock. So we do unlock, run glusterfs process + (thus communicate to glusterd), lock. But since this is not a + glusterfs command that is being run, unlocking and then relocking + is not needed. + */ + ret = runner_run (&runner); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "unmounting the " + "path %s (brick: %s) failed (%s)", mount_pt, + brickinfo->path, strerror (errno)); + goto out; + } + + runinit (&runner); + snprintf (msg, sizeof(msg), "remove snapshot of the brick %s:%s, " + "device: %s", brickinfo->hostname, brickinfo->path, + snap_device); + runner_add_args (&runner, "/sbin/lvremove", "-f", snap_device, NULL); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + + ret = runner_run (&runner); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "removing snapshot of the " + "brick (%s:%s) of device %s failed", + brickinfo->hostname, brickinfo->path, snap_device); + goto out; + } + +out: + return ret; +} + +int32_t +glusterd_lvm_snapshot_remove (dict_t *rsp_dict, glusterd_volinfo_t *snap_vol) +{ + char *mnt_pt = NULL; + struct mntent *entry = NULL; + int32_t brick_count = -1; + int32_t ret = -1; + glusterd_brickinfo_t *brickinfo = NULL; + xlator_t *this = NULL; + FILE *mtab = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap_vol); + + if (!snap_vol) { + gf_log (this->name, GF_LOG_ERROR, "snap volinfo is NULL"); + goto out; + } + + brick_count = -1; + list_for_each_entry (brickinfo, &snap_vol->bricks, brick_list) { + brick_count++; + if (uuid_compare (brickinfo->uuid, MY_UUID)) + continue; + + if (brickinfo->snap_status == -1) { + gf_log (this->name, GF_LOG_INFO, + "snapshot was pending. lvm not present " + "for brick %s:%s of the snap %s.", + brickinfo->hostname, brickinfo->path, + snap_vol->snapshot->snapname); + + /* Adding missed delete to the dict */ + ret = glusterd_add_missed_snaps_to_dict + (rsp_dict, + snap_vol->volname, + brickinfo, + brick_count + 1, + GF_SNAP_OPTION_TYPE_DELETE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add missed snapshot info " + "for %s:%s in the rsp_dict", + brickinfo->hostname, + brickinfo->path); + goto out; + } + + continue; + } + + ret = glusterd_get_brick_root (brickinfo->path, &mnt_pt); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "getting the root " + "of the brick for volume %s (snap %s) failed ", + snap_vol->volname, snap_vol->snapshot->snapname); + goto out; + } + + entry = glusterd_get_mnt_entry_info (mnt_pt, mtab); + if (!entry) { + gf_log (this->name, GF_LOG_WARNING, "getting the mount" + " entry for the brick %s:%s of the snap %s " + "(volume: %s) failed", brickinfo->hostname, + brickinfo->path, snap_vol->snapshot->snapname, + snap_vol->volname); + ret = -1; + goto out; + } + ret = glusterd_do_lvm_snapshot_remove (snap_vol, brickinfo, + mnt_pt, + entry->mnt_fsname); + if (mtab) + endmntent (mtab); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "remove the snapshot %s (%s)", + brickinfo->path, entry->mnt_fsname); + goto out; + } + + } + + ret = 0; +out: + return ret; +} + +int32_t +glusterd_snap_volume_remove (dict_t *rsp_dict, + glusterd_volinfo_t *snap_vol, + gf_boolean_t remove_lvm, + gf_boolean_t force) +{ + int ret = -1; + int save_ret = 0; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_volinfo_t *origin_vol = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap_vol); + + if (!snap_vol) { + gf_log(this->name, GF_LOG_WARNING, "snap_vol in NULL"); + ret = -1; + goto out; + } + + list_for_each_entry (brickinfo, &snap_vol->bricks, brick_list) { + if (uuid_compare (brickinfo->uuid, MY_UUID)) + continue; + + ret = glusterd_brick_stop (snap_vol, brickinfo, _gf_false); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to stop " + "brick for volume %s", snap_vol->volname); + save_ret = ret; + + /* Continue to cleaning up the snap in case of error + if force flag is enabled */ + if (!force) + goto out; + } + } + + /* Only remove the backend lvm when required */ + if (remove_lvm) { + ret = glusterd_lvm_snapshot_remove (rsp_dict, snap_vol); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to remove " + "lvm snapshot volume %s", snap_vol->volname); + save_ret = ret; + if (!force) + goto out; + } + } + + ret = glusterd_store_delete_volume (snap_vol); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to remove volume %s " + "from store", snap_vol->volname); + save_ret = ret; + if (!force) + goto out; + } + + if (!list_empty(&snap_vol->snapvol_list)) { + ret = glusterd_volinfo_find (snap_vol->parent_volname, + &origin_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "parent volinfo %s for volume %s", + snap_vol->parent_volname, snap_vol->volname); + save_ret = ret; + if (!force) + goto out; + } + origin_vol->snap_count--; + } + + ret = glusterd_volinfo_delete (snap_vol); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to remove volinfo " + "%s ", snap_vol->volname); + save_ret = ret; + if (!force) + goto out; + } + + if (save_ret) + ret = save_ret; +out: + gf_log (this->name, GF_LOG_TRACE, "returning %d", ret); + return ret; +} + +int32_t +glusterd_snapobject_delete (glusterd_snap_t *snap) +{ + if (snap == NULL) { + gf_log(THIS->name, GF_LOG_WARNING, "snap is NULL"); + return -1; + } + + list_del_init (&snap->snap_list); + list_del_init (&snap->volumes); + if (LOCK_DESTROY(&snap->lock)) + gf_log (THIS->name, GF_LOG_WARNING, "Failed destroying lock" + "of snap %s", snap->snapname); + + GF_FREE (snap->description); + GF_FREE (snap); + + return 0; +} + +int32_t +glusterd_snap_remove (dict_t *rsp_dict, + glusterd_snap_t *snap, + gf_boolean_t remove_lvm, + gf_boolean_t force) +{ + int ret = -1; + int save_ret = 0; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *tmp = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap); + + if (!snap) { + gf_log(this->name, GF_LOG_WARNING, "snap is NULL"); + ret = -1; + goto out; + } + + list_for_each_entry_safe (snap_vol, tmp, &snap->volumes, vol_list) { + ret = glusterd_snap_volume_remove (rsp_dict, snap_vol, + remove_lvm, force); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to remove " + "volinfo %s for snap %s", snap_vol->volname, + snap->snapname); + save_ret = ret; + + /* Continue to cleaning up the snap in case of error + if force flag is enabled */ + if (!force) + goto out; + } + } + + ret = glusterd_store_delete_snap (snap); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to remove snap %s " + "from store", snap->snapname); + save_ret = ret; + if (!force) + goto out; + } + + ret = glusterd_snapobject_delete (snap); + if (ret) + gf_log (this->name, GF_LOG_WARNING, "Failed to delete " + "snap object %s", snap->snapname); + + if (save_ret) + ret = save_ret; +out: + gf_log (THIS->name, GF_LOG_TRACE, "returning %d", ret); + return ret; +} + +static int +glusterd_snapshot_get_snapvol_detail (dict_t *dict, + glusterd_volinfo_t *snap_vol, + char *keyprefix, int detail) +{ + int ret = -1; + int snap_limit = 0; + char key[PATH_MAX] = {0,}; + char *value = NULL; + glusterd_volinfo_t *origin_vol = NULL; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + + this = THIS; + conf = this->private; + GF_ASSERT (conf); + + GF_ASSERT (dict); + GF_ASSERT (snap_vol); + GF_ASSERT (keyprefix); + + /* Volume Name */ + value = gf_strdup (snap_vol->volname); + if (!value) + goto out; + + snprintf (key, sizeof (key), "%s.volname", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "volume name in dictionary: %s", key); + goto out; + } + + /* Volume ID */ + value = gf_strdup (uuid_utoa (snap_vol->volume_id)); + if (NULL == value) { + ret = -1; + goto out; + } + + snprintf (key, sizeof (key), "%s.vol-id", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "volume id in dictionary: %s", key); + goto out; + } + value = NULL; + + /* volume status */ + snprintf (key, sizeof (key), "%s.vol-status", keyprefix); + switch (snap_vol->status) { + case GLUSTERD_STATUS_STARTED: + ret = dict_set_str (dict, key, "Started"); + break; + case GLUSTERD_STATUS_STOPPED: + ret = dict_set_str (dict, key, "Stopped"); + break; + case GD_SNAP_STATUS_NONE: + ret = dict_set_str (dict, key, "None"); + break; + default: + gf_log (this->name, GF_LOG_ERROR, "Invalid volume status"); + ret = -1; + goto out; + } + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set volume status" + " in dictionary: %s", key); + goto out; + } + + + ret = glusterd_volinfo_find (snap_vol->parent_volname, &origin_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to get the parent " + "volinfo for the volume %s", snap_vol->volname); + goto out; + } + + /* Snaps available */ + if (conf->snap_max_hard_limit < origin_vol->snap_max_hard_limit) { + snap_limit = conf->snap_max_hard_limit; + gf_log(this->name, GF_LOG_DEBUG, "system snap-max-hard-limit is" + " lesser than volume snap-max-hard-limit, " + "snap-max-hard-limit value is set to %d", snap_limit); + } else { + snap_limit = origin_vol->snap_max_hard_limit; + gf_log(this->name, GF_LOG_DEBUG, "volume snap-max-hard-limit is" + " lesser than system snap-max-hard-limit, " + "snap-max-hard-limit value is set to %d", snap_limit); + } + + snprintf (key, sizeof (key), "%s.snaps-available", keyprefix); + if (snap_limit > origin_vol->snap_count) + ret = dict_set_int32 (dict, key, + snap_limit - origin_vol->snap_count); + else + ret = dict_set_int32 (dict, key, 0); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set available snaps"); + goto out; + } + + snprintf (key, sizeof (key), "%s.snapcount", keyprefix); + ret = dict_set_int32 (dict, key, origin_vol->snap_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not save snapcount"); + goto out; + } + + if (!detail) + goto out; + + /* Parent volume name */ + value = gf_strdup (snap_vol->parent_volname); + if (!value) + goto out; + + snprintf (key, sizeof (key), "%s.origin-volname", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set parent " + "volume name in dictionary: %s", key); + goto out; + } + value = NULL; + + ret = 0; +out: + if (value) + GF_FREE (value); + + return ret; +} + +static int +glusterd_snapshot_get_snap_detail (dict_t *dict, glusterd_snap_t *snap, + char *keyprefix, glusterd_volinfo_t *volinfo) +{ + int ret = -1; + int volcount = 0; + char key[PATH_MAX] = {0,}; + char *value = NULL; + char *timestr = NULL; + struct tm *tmptr = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *tmp_vol = NULL; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (dict); + GF_ASSERT (snap); + GF_ASSERT (keyprefix); + + /* Snap Name */ + value = gf_strdup (snap->snapname); + if (!value) + goto out; + + snprintf (key, sizeof (key), "%s.snapname", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "snap name in dictionary"); + goto out; + } + + /* Snap ID */ + value = gf_strdup (uuid_utoa (snap->snap_id)); + if (NULL == value) { + ret = -1; + goto out; + } + + snprintf (key, sizeof (key), "%s.snap-id", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "snap id in dictionary"); + goto out; + } + value = NULL; + + tmptr = localtime (&(snap->time_stamp)); + if (NULL == tmptr) { + gf_log (this->name, GF_LOG_ERROR, "Failed to convert " + "time_t to *tm"); + ret = -1; + goto out; + } + + timestr = GF_CALLOC (1, PATH_MAX, gf_gld_mt_char); + if (NULL == timestr) { + ret = -1; + goto out; + } + + ret = strftime (timestr, PATH_MAX, "%Y-%m-%d %H:%M:%S", tmptr); + if (0 == ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to convert time_t " + "to string"); + ret = -1; + goto out; + } + + snprintf (key, sizeof (key), "%s.snap-time", keyprefix); + ret = dict_set_dynstr (dict, key, timestr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "snap time stamp in dictionary"); + goto out; + } + timestr = NULL; + + /* If snap description is provided then add that into dictionary */ + if (NULL != snap->description) { + value = gf_strdup (snap->description); + if (NULL == value) { + ret = -1; + goto out; + } + + snprintf (key, sizeof (key), "%s.snap-desc", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "snap description in dictionary"); + goto out; + } + value = NULL; + } + + snprintf (key, sizeof (key), "%s.snap-status", keyprefix); + switch (snap->snap_status) { + case GD_SNAP_STATUS_INIT: + ret = dict_set_str (dict, key, "Init"); + break; + case GD_SNAP_STATUS_IN_USE: + ret = dict_set_str (dict, key, "In-use"); + break; + case GD_SNAP_STATUS_DECOMMISSION: + ret = dict_set_str (dict, key, "Decommisioned"); + break; + case GD_SNAP_STATUS_RESTORED: + ret = dict_set_str (dict, key, "Restored"); + break; + case GD_SNAP_STATUS_NONE: + ret = dict_set_str (dict, key, "None"); + break; + default: + gf_log (this->name, GF_LOG_ERROR, "Invalid snap status"); + ret = -1; + goto out; + } + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snap status " + "in dictionary"); + goto out; + } + + if (volinfo) { + volcount = 1; + snprintf (key, sizeof (key), "%s.vol%d", keyprefix, volcount); + ret = glusterd_snapshot_get_snapvol_detail (dict, + volinfo, key, 0); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "get volume detail %s for snap %s", + snap_vol->volname, snap->snapname); + goto out; + } + goto done; + } + + list_for_each_entry_safe (snap_vol, tmp_vol, &snap->volumes, vol_list) { + volcount++; + snprintf (key, sizeof (key), "%s.vol%d", keyprefix, volcount); + ret = glusterd_snapshot_get_snapvol_detail (dict, + snap_vol, key, 1); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "get volume detail %s for snap %s", + snap_vol->volname, snap->snapname); + goto out; + } + } + +done: + snprintf (key, sizeof (key), "%s.vol-count", keyprefix); + ret = dict_set_int32 (dict, key, volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set %s", + key); + goto out; + } + + ret = 0; +out: + if (value) + GF_FREE (value); + + if (timestr) + GF_FREE(timestr); + + return ret; +} + +static int +glusterd_snapshot_get_all_snap_info (dict_t *dict) +{ + int ret = -1; + int snapcount = 0; + char key[PATH_MAX] = {0,}; + glusterd_snap_t *snap = NULL; + glusterd_snap_t *tmp_snap = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + + /* General parameter validation */ + GF_ASSERT (dict); + + list_for_each_entry_safe (snap, tmp_snap, &priv->snapshots, snap_list) { + snapcount++; + snprintf (key, sizeof (key), "snap%d", snapcount); + ret = glusterd_snapshot_get_snap_detail (dict, snap, key, NULL); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "snapdetail for snap %s", snap->snapname); + goto out; + } + } + + ret = dict_set_int32 (dict, "snap-count", snapcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount"); + goto out; + } + + ret = 0; +out: + return ret; +} + +int +glusterd_snapshot_get_info_by_volume (dict_t *dict, char *volname, + char *err_str, size_t len) +{ + int ret = -1; + int snapcount = 0; + int snap_limit = 0; + char *value = NULL; + char key[PATH_MAX] = ""; + glusterd_volinfo_t *volinfo = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *tmp_vol = NULL; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + + this = THIS; + conf = this->private; + GF_ASSERT (conf); + + GF_ASSERT (dict); + GF_ASSERT (volname); + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, len, "Volume (%s) does not exist", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + goto out; + } + + /* Snaps available */ + if (conf->snap_max_hard_limit < volinfo->snap_max_hard_limit) { + snap_limit = conf->snap_max_hard_limit; + gf_log(this->name, GF_LOG_DEBUG, "system snap-max-hard-limit is" + " lesser than volume snap-max-hard-limit, " + "snap-max-hard-limit value is set to %d", snap_limit); + } else { + snap_limit = volinfo->snap_max_hard_limit; + gf_log(this->name, GF_LOG_DEBUG, "volume snap-max-hard-limit is" + " lesser than system snap-max-hard-limit, " + "snap-max-hard-limit value is set to %d", snap_limit); + } + + if (snap_limit > volinfo->snap_count) + ret = dict_set_int32 (dict, "snaps-available", + snap_limit - volinfo->snap_count); + else + ret = dict_set_int32 (dict, "snaps-available", 0); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set available snaps"); + goto out; + } + + /* Origin volume name */ + value = gf_strdup (volinfo->volname); + if (!value) + goto out; + + ret = dict_set_dynstr (dict, "origin-volname", value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set parent " + "volume name in dictionary: %s", key); + goto out; + } + value = NULL; + + list_for_each_entry_safe (snap_vol, tmp_vol, &volinfo->snap_volumes, + snapvol_list) { + snapcount++; + snprintf (key, sizeof (key), "snap%d", snapcount); + ret = glusterd_snapshot_get_snap_detail (dict, + snap_vol->snapshot, + key, snap_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "snapdetail for snap %s", + snap_vol->snapshot->snapname); + goto out; + } + } + ret = dict_set_int32 (dict, "snap-count", snapcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount"); + goto out; + } + + ret = 0; +out: + if (value) + GF_FREE (value); + + return ret; +} + +/* This function will be called from RPC handler routine. + * This function is responsible for getting the requested + * snapshot info into the dictionary. + * + * @param req RPC request object. Required for sending a response back. + * @param op glusterd operation. Required for sending a response back. + * @param dict pointer to dictionary which will contain both + * request and response key-pair values. + * @return -1 on error and 0 on success + */ +int +glusterd_handle_snapshot_info (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + int8_t snap_driven = 1; + char *volname = NULL; + char *snapname = NULL; + glusterd_snap_t *snap = NULL; + xlator_t *this = NULL; + int32_t cmd = GF_SNAP_INFO_TYPE_ALL; + + this = THIS; + GF_ASSERT (this); + + GF_VALIDATE_OR_GOTO (this->name, req, out); + GF_VALIDATE_OR_GOTO (this->name, dict, out); + + + ret = dict_get_int32 (dict, "cmd", &cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get type " + "of snapshot info"); + goto out; + } + + switch (cmd) { + case GF_SNAP_INFO_TYPE_ALL: + { + ret = glusterd_snapshot_get_all_snap_info (dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get info of all snaps"); + goto out; + } + break; + } + + case GF_SNAP_INFO_TYPE_SNAP: + { + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get snap name"); + goto out; + } + + ret = dict_set_int32 (dict, "snap-count", 1); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set snapcount"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + snprintf (err_str, len, + "Snap (%s) does not exist", snapname); + gf_log (this->name, GF_LOG_ERROR, + "%s", err_str); + ret = -1; + goto out; + } + ret = glusterd_snapshot_get_snap_detail (dict, snap, + "snap1", NULL); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get snap detail of snap " + "%s", snap->snapname); + goto out; + } + break; + } + + case GF_SNAP_INFO_TYPE_VOL: + { + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get volname"); + goto out; + } + ret = glusterd_snapshot_get_info_by_volume (dict, + volname, err_str, len); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get volume info of volume " + "%s", volname); + goto out; + } + snap_driven = 0; + break; + } + } + + ret = dict_set_int8 (dict, "snap-driven", snap_driven); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snap-driven"); + goto out; + } + + /* If everything is successful then send the response back to cli. + * In case of failure the caller of this function will take care + of the response */ + ret = glusterd_op_send_cli_response (op, 0, 0, req, dict, err_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to send cli " + "response"); + goto out; + } + + ret = 0; + +out: + return ret; +} + +/* This function sets all the snapshot names in the dictionary */ +int +glusterd_snapshot_get_all_snapnames (dict_t *dict) +{ + int ret = -1; + int snapcount = 0; + char *snapname = NULL; + char key[PATH_MAX] = {0,}; + glusterd_snap_t *snap = NULL; + glusterd_snap_t *tmp_snap = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (dict); + + list_for_each_entry_safe (snap, tmp_snap, &priv->snapshots, snap_list) { + snapcount++; + snapname = gf_strdup (snap->snapname); + if (!snapname) { + gf_log (this->name, GF_LOG_ERROR, "strdup failed"); + ret = -1; + goto out; + } + snprintf (key, sizeof (key), "snapname%d", snapcount); + ret = dict_set_dynstr (dict, key, snapname); + if (ret) { + GF_FREE (snapname); + gf_log (this->name, GF_LOG_ERROR, "Failed to set %s", + key); + goto out; + } + } + + ret = dict_set_int32 (dict, "snap-count", snapcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount"); + goto out; + } + + ret = 0; +out: + + return ret; +} + +/* This function sets all the snapshot names + under a given volume in the dictionary */ +int +glusterd_snapshot_get_vol_snapnames (dict_t *dict, glusterd_volinfo_t *volinfo) +{ + int ret = -1; + int snapcount = 0; + char *snapname = NULL; + char key[PATH_MAX] = {0,}; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *tmp_vol = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (dict); + GF_ASSERT (volinfo); + + list_for_each_entry_safe (snap_vol, tmp_vol, + &volinfo->snap_volumes, snapvol_list) { + snapcount++; + snapname = gf_strdup (snap_vol->snapshot->snapname); + if (!snapname) { + gf_log (this->name, GF_LOG_ERROR, + "strdup failed"); + ret = -1; + goto out; + } + snprintf (key, sizeof (key), "snapname%d", snapcount); + ret = dict_set_dynstr (dict, key, snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "set %s", key); + GF_FREE (snapname); + goto out; + } + } + + ret = dict_set_int32 (dict, "snap-count", snapcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount"); + goto out; + } + + ret = 0; +out: + + return ret; +} + +int +glusterd_handle_snapshot_list (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + + GF_VALIDATE_OR_GOTO (this->name, req, out); + GF_VALIDATE_OR_GOTO (this->name, dict, out); + + /* Ignore error for getting volname as it is optional */ + ret = dict_get_str (dict, "volname", &volname); + + if (NULL == volname) { + ret = glusterd_snapshot_get_all_snapnames (dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get snapshot list"); + goto out; + } + } else { + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, len, + "Volume (%s) does not exist", volname); + gf_log (this->name, GF_LOG_ERROR, + "%s", err_str); + goto out; + } + + ret = glusterd_snapshot_get_vol_snapnames (dict, volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get snapshot list for volume %s", + volname); + goto out; + } + } + + /* If everything is successful then send the response back to cli. + In case of failure the caller of this function will take of response.*/ + ret = glusterd_op_send_cli_response (op, 0, 0, req, dict, err_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to send cli " + "response"); + goto out; + } + + ret = 0; + +out: + return ret; +} + +/* This is a snapshot create handler function. This function will be + * executed in the originator node. This function is responsible for + * calling mgmt_v3 framework to do the actual snap creation on all the bricks + * + * @param req RPC request object + * @param op gluster operation + * @param dict dictionary containing snapshot restore request + * @param err_str In case of an err this string should be populated + * @param len length of err_str buffer + * + * @return Negative value on Failure and 0 in success + */ +int +glusterd_handle_snapshot_create (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + char *volname = NULL; + char *snapname = NULL; + int64_t volcount = 0; + xlator_t *this = NULL; + char key[PATH_MAX] = ""; + char *username = NULL; + char *password = NULL; + uuid_t *uuid_ptr = NULL; + uuid_t tmp_uuid = {0}; + int i = 0; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (dict); + GF_ASSERT (err_str); + + ret = dict_get_int64 (dict, "volcount", &volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "get the volume count"); + goto out; + } + if (volcount <= 0) { + gf_log (this->name, GF_LOG_ERROR, "Invalid volume count %ld " + "supplied", volcount); + ret = -1; + goto out; + } + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to get the snapname"); + goto out; + } + + if (strlen(snapname) >= GLUSTERD_MAX_SNAP_NAME) { + snprintf (err_str, len, "snapname cannot exceed 255 " + "characters"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + ret = -1; + goto out; + } + + uuid_ptr = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + if (!uuid_ptr) { + gf_log (this->name, GF_LOG_ERROR, "Out Of Memory"); + ret = -1; + goto out; + } + + uuid_generate (*uuid_ptr); + ret = dict_set_bin (dict, "snap-id", uuid_ptr, sizeof(uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to set snap-id"); + GF_FREE (uuid_ptr); + goto out; + } + uuid_ptr = NULL; + + ret = dict_set_int64 (dict, "snap-time", (int64_t)time(NULL)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to set snap-time"); + goto out; + } + + for (i = 1; i <= volcount; i++) { + snprintf (key, sizeof (key), "volname%d", i); + ret = dict_get_str (dict, key, &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get volume name"); + goto out; + } + + /* generate internal username and password for the snap*/ + uuid_generate (tmp_uuid); + username = gf_strdup (uuid_utoa (tmp_uuid)); + snprintf (key, sizeof(key), "volume%d_username", i); + ret = dict_set_dynstr (dict, key, username); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snap " + "username for volume %s", volname); + GF_FREE (username); + goto out; + } + + uuid_generate (tmp_uuid); + password = gf_strdup (uuid_utoa (tmp_uuid)); + snprintf (key, sizeof(key), "volume%d_password", i); + ret = dict_set_dynstr (dict, key, password); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snap " + "password for volume %s", volname); + GF_FREE (password); + goto out; + } + + uuid_ptr = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + if (!uuid_ptr) { + gf_log (this->name, GF_LOG_ERROR, "Out Of Memory"); + ret = -1; + goto out; + } + + snprintf (key, sizeof(key) - 1, "vol%d_volid", i); + uuid_generate (*uuid_ptr); + ret = dict_set_bin (dict, key, uuid_ptr, sizeof(uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set snap_volid"); + GF_FREE (uuid_ptr); + goto out; + } + } + + ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to initiate snap " + "phases"); + } + +out: + return ret; +} + +/* This is a snapshot status handler function. This function will be + * executed in a originator node. This function is responsible for + * calling mgmt v3 framework to get the actual snapshot status from + * all the bricks + * + * @param req RPC request object + * @param op gluster operation + * @param dict dictionary containing snapshot status request + * @param err_str In case of an err this string should be populated + * @param len length of err_str buffer + * + * return : 0 in case of success. + * -1 in case of failure. + * + */ +int +glusterd_handle_snapshot_status (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + char *volname = NULL; + char *snapname = NULL; + char *buf = NULL; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + int32_t cmd = -1; + int i = 0; + dict_t *voldict = NULL; + char key[PATH_MAX] = ""; + glusterd_volinfo_t *volinfo = NULL; + glusterd_snap_t *snap = NULL; + glusterd_volinfo_t *snap_volinfo = NULL; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + + GF_ASSERT (conf); + GF_ASSERT (req); + GF_ASSERT (dict); + GF_ASSERT (err_str); + + ret = dict_get_int32 (dict, "cmd", &cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not get status type"); + goto out; + } + switch (cmd) { + case GF_SNAP_STATUS_TYPE_ALL: + { + /* IF we give "gluster snapshot status" + * then lock is held on all snaps. + * This is the place where necessary information + * (snapname and snapcount)is populated in dictionary + * for locking. + */ + ++i; + list_for_each_entry (snap, &conf->snapshots, snap_list) + { + snprintf (key, sizeof (key), "snapname%d", i); + buf = gf_strdup (snap->snapname); + if (!buf) { + ret = -1; + goto out; + } + ret = dict_set_dynstr (dict, key, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save snapname (%s) " + "in the dictionary", + snap->snapname); + GF_FREE (buf); + goto out; + } + + buf = NULL; + i++; + } + + ret = dict_set_int32 (dict, "snapcount", i - 1); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not " + "save snapcount in the dictionary"); + goto out; + } + break; + } + + case GF_SNAP_STATUS_TYPE_SNAP: + { + /* IF we give "gluster snapshot status <snapname>" + * then lock is held on single snap. + * This is the place where necessary information + * (snapname)is populated in dictionary + * for locking. + */ + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to fetch snap name"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + snprintf (err_str, len, "Snap (%s)" + "does not exist", snapname); + gf_log(this->name, GF_LOG_ERROR, + "%s", err_str); + ret = -1; + goto out; + } + break; + } + case GF_SNAP_STATUS_TYPE_VOL: + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to fetch volname"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, len, "Volume (%s) " + "does not exist", volname); + gf_log (this->name, GF_LOG_ERROR, + "%s", err_str); + goto out; + } + + i = 1; + list_for_each_entry (snap_volinfo, + &volinfo->snap_volumes, snapvol_list) { + snprintf (key, sizeof (key), "snapname%d", i); + + buf = gf_strdup + (snap_volinfo->snapshot->snapname); + if (!buf) { + ret = -1; + goto out; + } + + ret = dict_set_dynstr (dict, key, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save snapname"); + GF_FREE (buf); + goto out; + } + + buf = NULL; + i++; + } + + ret = dict_set_int32 (dict, "snapcount", i-1); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save snapcount"); + goto out; + } + break; + default: + { + gf_log (this->name, GF_LOG_ERROR, "Unknown type"); + ret = -1; + goto out; + } + } + + /* Volume lock is not necessary for snapshot status, hence + * turning it off + */ + ret = dict_set_int8 (dict, "hold_vol_locks", 0); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Setting volume lock " + "flag failed"); + goto out; + } + + ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to initiate " + "snap phases"); + goto out; + } + + ret = 0; + +out: + if (voldict) { + dict_unref (voldict); + } + return ret; +} + + +/* This is a snapshot restore handler function. This function will be + * executed in the originator node. This function is responsible for + * calling mgmt_v3 framework to do the actual restore on all the bricks + * + * @param req RPC request object + * @param op gluster operation + * @param dict dictionary containing snapshot restore request + * @param err_str In case of an err this string should be populated + * @param len length of err_str buffer + * + * @return Negative value on Failure and 0 in success + */ +int +glusterd_handle_snapshot_restore (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + char *snapname = NULL; + char *buf = NULL; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + glusterd_snap_t *snap = NULL; + glusterd_volinfo_t *snap_volinfo = NULL; + int32_t i = 0; + char key[PATH_MAX] = ""; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + + GF_ASSERT (conf); + GF_ASSERT (req); + GF_ASSERT (dict); + GF_ASSERT (err_str); + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "get snapname"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + snprintf (err_str, len, "Snap (%s) does not exist", snapname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + ret = -1; + goto out; + } + + list_for_each_entry (snap_volinfo, &snap->volumes, vol_list) { + i++; + snprintf (key, sizeof (key), "volname%d", i); + buf = gf_strdup (snap_volinfo->parent_volname); + if (!buf) { + ret = -1; + goto out; + } + ret = dict_set_dynstr (dict, key, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not set " + "parent volume name %s in the dict", + snap_volinfo->parent_volname); + GF_FREE (buf); + goto out; + } + buf = NULL; + } + + ret = dict_set_int32 (dict, "volcount", i); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save volume count"); + goto out; + } + + ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to initiate snap " + "phases"); + goto out; + } + + ret = 0; + +out: + return ret; +} + +glusterd_snap_t* +glusterd_create_snap_object (dict_t *dict, dict_t *rsp_dict) +{ + char *snapname = NULL; + uuid_t *snap_id = NULL; + char *description = NULL; + glusterd_snap_t *snap = NULL; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int ret = -1; + int64_t time_stamp = 0; + + this = THIS; + priv = this->private; + + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + /* Fetch snapname, description, id and time from dict */ + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snapname"); + goto out; + } + + /* Ignore ret value for description*/ + ret = dict_get_str (dict, "description", &description); + + ret = dict_get_bin (dict, "snap-id", (void **)&snap_id); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snap_id"); + goto out; + } + + ret = dict_get_int64 (dict, "snap-time", &time_stamp); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snap-time"); + goto out; + } + if (time_stamp <= 0) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Invalid time-stamp: %ld", + time_stamp); + goto out; + } + + list_for_each_entry (snap, &priv->snapshots, snap_list) { + if (!strcmp (snap->snapname, snapname) || + !uuid_compare (snap->snap_id, *snap_id)) { + gf_log (THIS->name, GF_LOG_ERROR, + "Found duplicate snap %s (%s)", + snap->snapname, uuid_utoa (snap->snap_id)); + ret = -1; + break; + } + } + if (ret) { + snap = NULL; + goto out; + } + + snap = glusterd_new_snap_object (); + if (!snap) { + gf_log (this->name, GF_LOG_ERROR, "Could not create " + "the snap object for snap %s", snapname); + goto out; + } + + strcpy (snap->snapname, snapname); + uuid_copy (snap->snap_id, *snap_id); + snap->time_stamp = (time_t)time_stamp; + /* Set the status as GD_SNAP_STATUS_INIT and once the backend snapshot + is taken and snap is really ready to use, set the status to + GD_SNAP_STATUS_IN_USE. This helps in identifying the incomplete + snapshots and cleaning them up. + */ + snap->snap_status = GD_SNAP_STATUS_INIT; + if (description) { + snap->description = gf_strdup (description); + if (snap->description == NULL) { + gf_log (this->name, GF_LOG_ERROR, + "Saving the Snap Description Failed"); + ret = -1; + goto out; + } + } + + ret = glusterd_store_snap (snap); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Could not store snap" + "object %s", snap->snapname); + goto out; + } + + list_add_order (&snap->snap_list, &priv->snapshots, + glusterd_compare_snap_time); + + gf_log (this->name, GF_LOG_TRACE, "Snap %s added to the list", + snap->snapname); + + ret = 0; + +out: + if (ret) { + if (snap) + glusterd_snap_remove (rsp_dict, snap, + _gf_true, _gf_true); + snap = NULL; + } + + return snap; +} + +/* This function is called to get the device path of the snap lvm. Usually + if /dev/mapper/<group-name>-<lvm-name> is the device for the lvm, + then the snap device will be /dev/<group-name>/<snapname>. + This function takes care of building the path for the snap device. +*/ +char * +glusterd_build_snap_device_path (char *device, char *snapname) +{ + char snap[PATH_MAX] = ""; + char msg[1024] = ""; + char volgroup[PATH_MAX] = ""; + char *snap_device = NULL; + xlator_t *this = NULL; + runner_t runner = {0,}; + char *ptr = NULL; + int ret = -1; + + this = THIS; + GF_ASSERT (this); + if (!device) { + gf_log (this->name, GF_LOG_ERROR, "device is NULL"); + goto out; + } + if (!snapname) { + gf_log (this->name, GF_LOG_ERROR, "snapname is NULL"); + goto out; + } + + runinit (&runner); + runner_add_args (&runner, "/sbin/lvs", "--noheadings", "-o", "vg_name", + device, NULL); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + snprintf (msg, sizeof (msg), "Get volume group for device %s", device); + runner_log (&runner, this->name, GF_LOG_DEBUG, msg); + ret = runner_start (&runner); + if (ret == -1) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volume group " + "for device %s", device); + runner_end (&runner); + goto out; + } + ptr = fgets(volgroup, sizeof(volgroup), + runner_chio (&runner, STDOUT_FILENO)); + if (!ptr || !strlen(volgroup)) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volume group " + "for snap %s", snapname); + runner_end (&runner); + ret = -1; + goto out; + } + runner_end (&runner); + + snprintf (snap, sizeof(snap), "/dev/%s/%s", gf_trim(volgroup), + snapname); + snap_device = gf_strdup (snap); + if (!snap_device) { + gf_log (this->name, GF_LOG_WARNING, "Cannot copy the " + "snapshot device name for snapname: %s)", snapname); + } + +out: + return snap_device; +} + +/* This function actually calls the command (or the API) for taking the + snapshot of the backend brick filesystem. If this is successful, + then call the glusterd_snap_create function to create the snap object + for glusterd +*/ +char * +glusterd_take_lvm_snapshot (glusterd_volinfo_t *snap_vol, + glusterd_brickinfo_t *brickinfo) +{ + char msg[NAME_MAX] = ""; + char buf[PATH_MAX] = ""; + char *snap_device = NULL; + char *ptr = NULL; + char *device = NULL; + int ret = -1; + gf_boolean_t match = _gf_false; + runner_t runner = {0,}; + xlator_t *this = NULL; + + this = THIS; + + if (!brickinfo) { + gf_log (this->name, GF_LOG_ERROR, "brickinfo NULL"); + goto out; + } + + device = glusterd_get_brick_mount_details (brickinfo); + if (!device) { + gf_log (this->name, GF_LOG_ERROR, "getting device name for " + "the brick %s:%s failed", brickinfo->hostname, + brickinfo->path); + goto out; + } + + /* Figuring out if setactivationskip flag is supported or not */ + runinit (&runner); + snprintf (msg, sizeof (msg), "running lvcreate help"); + runner_add_args (&runner, "/sbin/lvcreate", "--help", NULL); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + ret = runner_start (&runner); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to run lvcreate help"); + runner_end (&runner); + goto out; + } + + /* Looking for setactivationskip in lvcreate --help */ + do { + ptr = fgets(buf, sizeof(buf), + runner_chio (&runner, STDOUT_FILENO)); + if (ptr) { + if (strstr(buf, "setactivationskip")) { + match = _gf_true; + break; + } + } + } while (ptr != NULL); + runner_end (&runner); + + /* Takng the actual snapshot */ + runinit (&runner); + snprintf (msg, sizeof (msg), "taking snapshot of the brick %s:%s", + brickinfo->hostname, brickinfo->path); + if (match == _gf_true) + runner_add_args (&runner, "/sbin/lvcreate", "-s", device, + "--setactivationskip", "n", "--name", + snap_vol->volname, NULL); + else + runner_add_args (&runner, "/sbin/lvcreate", "-s", device, + "--name", snap_vol->volname, NULL); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + ret = runner_start (&runner); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "taking snapshot of the " + "brick (%s:%s) of device %s failed", + brickinfo->hostname, brickinfo->path, device); + runner_end (&runner); + goto out; + } + runner_end (&runner); + + snap_device = glusterd_build_snap_device_path (device, + snap_vol->volname); + if (!snap_device) { + gf_log (this->name, GF_LOG_WARNING, "Cannot copy the snapshot " + "device name for snap %s (volume id: %s)", + snap_vol->snapshot->snapname, snap_vol->volname); + ret = -1; + goto out; + } + +out: + return snap_device; +} + +int32_t +glusterd_snap_brick_create (char *device, glusterd_volinfo_t *snap_volinfo, + glusterd_brickinfo_t *original_brickinfo, + int32_t brick_count, char *snap_brick_dir) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + char snap_brick_mount_path[PATH_MAX] = ""; + char snap_brick_path[PATH_MAX] = ""; + char msg[1024] = ""; + struct stat statbuf = {0, }; + runner_t runner = {0, }; + + this = THIS; + priv = this->private; + + GF_ASSERT (device); + GF_ASSERT (snap_volinfo); + GF_ASSERT (original_brickinfo); + GF_ASSERT (snap_brick_dir); + + snprintf (snap_brick_mount_path, sizeof (snap_brick_mount_path), + "%s/%s/brick%d", snap_mount_folder, snap_volinfo->volname, + brick_count+1); + + snprintf (snap_brick_path, sizeof (snap_brick_path), "%s%s", + snap_brick_mount_path, snap_brick_dir); + + ret = mkdir_p (snap_brick_mount_path, 0777, _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "creating the brick directory" + " %s for the snapshot %s(device: %s) failed", + snap_brick_mount_path, snap_volinfo->volname, device); + goto out; + } + /* mount the snap logical device on the directory inside + /run/gluster/snaps/<snapname>/@snap_brick_mount_path + Way to mount the snap brick via mount api is this. + ret = mount (device, snap_brick_mount_path, entry->mnt_type, + MS_MGC_VAL, "nouuid"); + But for now, mounting using runner apis. + */ + runinit (&runner); + snprintf (msg, sizeof (msg), "mounting snapshot of the brick %s:%s", + original_brickinfo->hostname, original_brickinfo->path); + runner_add_args (&runner, "mount", "-o", "nouuid", device, + snap_brick_mount_path, NULL); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + + /* let glusterd get blocked till snapshot is over */ + synclock_unlock (&priv->big_lock); + ret = runner_run (&runner); + synclock_lock (&priv->big_lock); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "mounting the snapshot " + "logical device %s failed (error: %s)", device, + strerror (errno)); + goto out; + } else + gf_log (this->name, GF_LOG_DEBUG, "mounting the snapshot " + "logical device %s successful", device); + + ret = stat (snap_brick_path, &statbuf); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "stat of the brick %s" + "(brick mount: %s) failed (%s)", snap_brick_path, + snap_brick_mount_path, strerror (errno)); + goto out; + } + ret = sys_lsetxattr (snap_brick_path, + GF_XATTR_VOL_ID_KEY, + snap_volinfo->volume_id, 16, + XATTR_REPLACE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "extended attribute %s on %s. Reason: " + "%s, snap: %s", GF_XATTR_VOL_ID_KEY, + snap_brick_path, strerror (errno), + snap_volinfo->volname); + goto out; + } + +out: + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "unmounting the snap brick" + " mount %s", snap_brick_mount_path); + umount (snap_brick_mount_path); + } + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Added missed_snap_entry to rsp_dict */ +int32_t +glusterd_add_missed_snaps_to_dict (dict_t *rsp_dict, char *snap_uuid, + glusterd_brickinfo_t *brickinfo, + int32_t brick_number, int32_t op) +{ + char *buf = NULL; + char missed_snap_entry[PATH_MAX] = ""; + char name_buf[PATH_MAX] = ""; + int32_t missed_snap_count = -1; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap_uuid); + GF_ASSERT (brickinfo); + + snprintf (missed_snap_entry, sizeof(missed_snap_entry), + "%s:%s=%d:%s:%d:%d", uuid_utoa(brickinfo->uuid), + snap_uuid, brick_number, brickinfo->path, op, + GD_MISSED_SNAP_PENDING); + + buf = gf_strdup (missed_snap_entry); + if (!buf) { + ret = -1; + goto out; + } + + /* Fetch the missed_snap_count from the dict */ + ret = dict_get_int32 (rsp_dict, "missed_snap_count", + &missed_snap_count); + if (ret) { + /* Initialize the missed_snap_count for the first time */ + missed_snap_count = 0; + } + + /* Setting the missed_snap_entry in the rsp_dict */ + snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d", + missed_snap_count); + ret = dict_set_dynstr (rsp_dict, name_buf, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set missed_snap_entry (%s) " + "in the rsp_dict.", buf); + GF_FREE (buf); + goto out; + } + missed_snap_count++; + + /* Setting the new missed_snap_count in the dict */ + ret = dict_set_int32 (rsp_dict, "missed_snap_count", + missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set missed_snap_count for %s " + "in the rsp_dict.", missed_snap_entry); + goto out; + } + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int32_t +glusterd_add_bricks_to_snap_volume (dict_t *dict, dict_t *rsp_dict, + glusterd_volinfo_t *snap_vol, + glusterd_brickinfo_t *original_brickinfo, + glusterd_brickinfo_t *snap_brickinfo, + char **snap_brick_dir, int64_t volcount, + int32_t brick_count) +{ + char key[PATH_MAX] = ""; + char snap_brick_path[PATH_MAX] = ""; + char *snap_device = NULL; + gf_boolean_t add_missed_snap = _gf_false; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap_vol); + GF_ASSERT (original_brickinfo); + GF_ASSERT (snap_brickinfo); + GF_ASSERT (snap_brick_dir); + + snprintf (key, sizeof(key) - 1, "vol%ld.brickdir%d", volcount, + brick_count); + ret = dict_get_ptr (dict, key, (void **)snap_brick_dir); + if (ret) { + /* Using original brickinfo here because it will be a + * pending snapshot and storing the original brickinfo + * will help in mapping while recreating the missed snapshot + */ + gf_log (this->name, GF_LOG_WARNING, "Unable to fetch " + "snap mount path (%s). Using original brickinfo", key); + snap_brickinfo->snap_status = -1; + strcpy (snap_brick_path, original_brickinfo->path); + + /* In origiator node add snaps missed + * from different nodes to the dict + */ + if (is_origin_glusterd (dict) == _gf_true) + add_missed_snap = _gf_true; + } else { + /* Create brick-path in the format /var/run/gluster/snaps/ * + * <snap-uuid>/<original-brick#>/snap-brick-dir * + */ + snprintf (snap_brick_path, sizeof(snap_brick_path), + "%s/%s/brick%d%s", snap_mount_folder, + snap_vol->volname, brick_count+1, + *snap_brick_dir); + } + + if ((snap_brickinfo->snap_status != -1) && + (!uuid_compare (original_brickinfo->uuid, MY_UUID)) && + (!glusterd_is_brick_started (original_brickinfo))) { + /* In case if the brick goes down after prevalidate. */ + gf_log (this->name, GF_LOG_WARNING, "brick %s:%s is not" + " started (snap: %s)", + original_brickinfo->hostname, + original_brickinfo->path, + snap_vol->snapshot->snapname); + + snap_brickinfo->snap_status = -1; + strcpy (snap_brick_path, original_brickinfo->path); + add_missed_snap = _gf_true; + } + + if (add_missed_snap) { + ret = glusterd_add_missed_snaps_to_dict (rsp_dict, + snap_vol->volname, + original_brickinfo, + brick_count + 1, + GF_SNAP_OPTION_TYPE_CREATE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to add missed" + " snapshot info for %s:%s in the rsp_dict", + original_brickinfo->hostname, + original_brickinfo->path); + goto out; + } + } + + snprintf (key, sizeof(key), "vol%ld.brick_snapdevice%d", + volcount, brick_count); + ret = dict_get_ptr (dict, key, (void **)&snap_device); + if (ret) { + /* If the device name is empty, so will be the brick path + * Hence the missed snap has already been added above + */ + gf_log (this->name, GF_LOG_ERROR, "Unable to fetch " + "snap device (%s). Leaving empty", key); + } else + strcpy (snap_brickinfo->device_path, snap_device); + + ret = gf_canonicalize_path (snap_brick_path); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to canonicalize path"); + goto out; + } + + strcpy (snap_brickinfo->hostname, original_brickinfo->hostname); + strcpy (snap_brickinfo->path, snap_brick_path); + uuid_copy (snap_brickinfo->uuid, original_brickinfo->uuid); + list_add_tail (&snap_brickinfo->brick_list, &snap_vol->bricks); + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int32_t +glusterd_take_brick_snapshot (glusterd_volinfo_t *origin_vol, + glusterd_volinfo_t *snap_vol, dict_t *rsp_dict, + glusterd_brickinfo_t *original_brickinfo, + glusterd_brickinfo_t *snap_brickinfo, + char *snap_brick_dir, int32_t brick_count) +{ + char *device = NULL; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (origin_vol); + GF_ASSERT (snap_vol); + GF_ASSERT (rsp_dict); + GF_ASSERT (original_brickinfo); + GF_ASSERT (snap_brickinfo); + GF_ASSERT (snap_brick_dir); + + device = glusterd_take_lvm_snapshot (snap_vol, original_brickinfo); + /* Fail the snapshot even though snapshot on one of + the bricks fails. At the end when we check whether + the snapshot volume meets quorum or not, then the + the snapshot can either be treated as success, or + in case of failure we can undo the changes and return + failure to cli. */ + if (!device) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to take snapshot of %s:%s", + original_brickinfo->hostname, + original_brickinfo->path); + goto out; + } + + /* create the complete brick here */ + ret = glusterd_snap_brick_create (device, snap_vol, + original_brickinfo, + brick_count, snap_brick_dir); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "not able to" + " create the brickinfo for the snap %s" + ", volume %s", snap_vol->snapshot->snapname, + origin_vol->volname); + goto out; + } + +out: + if (device) + GF_FREE (device); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Look for disconnected peers, for missed snap creates or deletes */ +static int32_t +glusterd_find_missed_snap (dict_t *rsp_dict, glusterd_volinfo_t *vol, + char *snap_uuid, struct list_head *peers, + int32_t op) +{ + int32_t brick_count = -1; + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (peers); + GF_ASSERT (vol); + GF_ASSERT (snap_uuid); + + brick_count = 0; + list_for_each_entry (brickinfo, &vol->bricks, brick_list) { + if (!uuid_compare (brickinfo->uuid, MY_UUID)) { + /* If the brick belongs to the same node */ + brick_count++; + continue; + } + + list_for_each_entry (peerinfo, peers, uuid_list) { + if (uuid_compare (peerinfo->uuid, brickinfo->uuid)) { + /* If the brick doesnt belong to this peer */ + continue; + } + + /* Found peer who owns the brick, * + * if peer is not connected or not * + * friend add it to missed snap list */ + if (!(peerinfo->connected) || + (peerinfo->state.state != + GD_FRIEND_STATE_BEFRIENDED)) { + ret = glusterd_add_missed_snaps_to_dict + (rsp_dict, + snap_uuid, + brickinfo, + brick_count + 1, + op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add missed snapshot " + "info for %s:%s in the " + "rsp_dict", brickinfo->hostname, + brickinfo->path); + goto out; + } + } + } + brick_count++; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +glusterd_volinfo_t * +glusterd_do_snap_vol (glusterd_volinfo_t *origin_vol, glusterd_snap_t *snap, + dict_t *dict, dict_t *rsp_dict, int64_t volcount) +{ + char key[PATH_MAX] = ""; + char *snap_brick_dir = NULL; + char *username = NULL; + char *password = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_conf_t *priv = NULL; + glusterd_volinfo_t *snap_vol = NULL; + uuid_t *snap_volid = NULL; + int32_t ret = -1; + int32_t brick_count = 0; + glusterd_brickinfo_t *snap_brickinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (origin_vol); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + /* fetch username, password and vol_id from dict*/ + snprintf (key, sizeof(key), "volume%ld_username", volcount); + ret = dict_get_str (dict, key, &username); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get %s for " + "snap %s", key, snap->snapname); + goto out; + } + + snprintf (key, sizeof(key), "volume%ld_password", volcount); + ret = dict_get_str (dict, key, &password); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get %s for " + "snap %s", key, snap->snapname); + goto out; + } + snprintf (key, sizeof(key) - 1, "vol%ld_volid", volcount); + ret = dict_get_bin (dict, key, (void **)&snap_volid); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch snap_volid"); + goto out; + } + + /* We are not setting the username and password here as + * we need to set the user name and password passed in + * the dictionary + */ + ret = glusterd_volinfo_dup (origin_vol, &snap_vol, _gf_false); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to duplicate volinfo " + "for the snapshot %s", snap->snapname); + goto out; + } + + /* uuid is used as lvm snapshot name. + This will avoid restrictions on snapshot names provided by user */ + GLUSTERD_GET_UUID_NOHYPHEN (snap_vol->volname, *snap_volid); + uuid_copy (snap_vol->volume_id, *snap_volid); + snap_vol->is_snap_volume = _gf_true; + strcpy (snap_vol->parent_volname, origin_vol->volname); + snap_vol->snapshot = snap; + + glusterd_auth_set_username (snap_vol, username); + glusterd_auth_set_password (snap_vol, password); + + /* Adding snap brickinfos to the snap volinfo */ + brick_count = 0; + list_for_each_entry (brickinfo, &origin_vol->bricks, brick_list) { + snap_brickinfo = NULL; + + ret = glusterd_brickinfo_new (&snap_brickinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "initializing the brick for the snap " + "volume failed (snapname: %s)", snap->snapname); + goto out; + } + + ret = glusterd_add_bricks_to_snap_volume (dict, rsp_dict, + snap_vol, + brickinfo, + snap_brickinfo, + &snap_brick_dir, + volcount, + brick_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add the snap brick for " + "%s:%s to the snap volume", + brickinfo->hostname, brickinfo->path); + GF_FREE (snap_brickinfo); + goto out; + } + + /* Take snapshot of the brick */ + if ((uuid_compare (brickinfo->uuid, MY_UUID)) || + (snap_brickinfo->snap_status == -1)) { + brick_count++; + continue; + } + + ret = glusterd_take_brick_snapshot (origin_vol, snap_vol, + rsp_dict, brickinfo, + snap_brickinfo, + snap_brick_dir, + brick_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to take snapshot for %s:%s", + brickinfo->hostname, brickinfo->path); + goto out; + } + + brick_count++; + } + + /*TODO: the quorum check of the snap volume here */ + + ret = glusterd_store_volinfo (snap_vol, + GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to store snapshot " + "volinfo (%s) for snap %s", snap_vol->volname, + snap->snapname); + goto out; + } + + ret = generate_brick_volfiles (snap_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "generating the brick " + "volfiles for the snap %s (volume: %s) failed", + snap->snapname, origin_vol->volname); + goto out; + } + + ret = generate_client_volfiles (snap_vol, GF_CLIENT_TRUSTED); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "generating the trusted " + "client volfiles for the snap %s (volume: %s) failed", + snap->snapname, origin_vol->volname); + goto out; + } + ret = generate_client_volfiles (snap_vol, GF_CLIENT_OTHER); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "generating the client " + "volfiles for the snap %s (volume: %s) failed", + snap->snapname, origin_vol->volname); + goto out; + } + + ret = glusterd_list_add_snapvol (origin_vol, snap_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "could not add the snap " + "volume %s to the list", snap_vol->volname); + goto out; + } + + list_for_each_entry (brickinfo, &snap_vol->bricks, brick_list) { + if (uuid_compare (brickinfo->uuid, MY_UUID)) + continue; + + if (brickinfo->snap_status == -1) { + gf_log (this->name, GF_LOG_INFO, + "not starting snap brick %s:%s for " + "for the snap %s (volume: %s)", + brickinfo->hostname, brickinfo->path, + snap->snapname, origin_vol->volname); + continue; + } + + ret = glusterd_brick_start (snap_vol, brickinfo, _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "starting the " + "brick %s:%s for the snap %s (volume: %s) " + "failed", brickinfo->hostname, brickinfo->path, + snap->snapname, origin_vol->volname); + goto out; + } + } + + snap_vol->status = GLUSTERD_STATUS_STARTED; + ret = glusterd_store_volinfo (snap_vol, + GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to store snap volinfo"); + goto out; + } + +out: + if (ret) { + if (snap_vol) + glusterd_snap_volume_remove (rsp_dict, snap_vol, + _gf_true, _gf_true); + snap_vol = NULL; + } + + return snap_vol; +} + +/* This is a snapshot remove handler function. This function will be + * executed in the originator node. This function is responsible for + * calling mgmt v3 framework to do the actual remove on all the bricks + * + * @param req RPC request object + * @param op gluster operation + * @param dict dictionary containing snapshot remove request + * @param err_str In case of an err this string should be populated + * @param len length of err_str buffer + * + * @return Negative value on Failure and 0 in success + */ +int +glusterd_handle_snapshot_remove (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + int64_t volcount = 0; + char *snapname = NULL; + char *volname = NULL; + char key[PATH_MAX] = ""; + glusterd_snap_t *snap = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *tmp = NULL; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (req); + GF_ASSERT (dict); + GF_ASSERT (err_str); + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get snapname"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + snprintf (err_str, len, "Snap (%s) does not exist", snapname); + gf_log (this->name, GF_LOG_ERROR, + "%s", err_str); + ret = -1; + goto out; + } + + /* Set volnames in the dict to get mgmt_v3 lock */ + list_for_each_entry_safe (snap_vol, tmp, &snap->volumes, vol_list) { + volcount++; + volname = gf_strdup (snap_vol->parent_volname); + if (!volname) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "strdup failed"); + goto out; + } + + snprintf (key, sizeof (key), "volname%ld", volcount); + ret = dict_set_dynstr (dict, key, volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "volume name in dictionary"); + GF_FREE (volname); + goto out; + } + volname = NULL; + } + ret = dict_set_int64 (dict, "volcount", volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set volcount"); + goto out; + } + + ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to initiate snap " + "phases"); + goto out; + } + + ret = 0; +out: + return ret; +} + +int +glusterd_snapshot_remove_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int32_t ret = -1; + char *snapname = NULL; + xlator_t *this = NULL; + glusterd_snap_t *snap = NULL; + + this = THIS; + + if (!dict || !op_errstr) { + gf_log (this->name, GF_LOG_ERROR, "input parameters NULL"); + goto out; + } + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Getting the snap name " + "failed"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + gf_log (this->name, GF_LOG_ERROR, "Snap %s does not exist", + snapname); + ret = -1; + goto out; + } + ret = 0; +out: + return ret; +} + +int +glusterd_snapshot_status_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int ret = -1; + char *snapname = NULL; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + int32_t cmd = -1; + glusterd_volinfo_t *volinfo = NULL; + char *volname = NULL; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + + GF_ASSERT (conf); + GF_ASSERT (op_errstr); + if (!dict) { + gf_log (this->name, GF_LOG_ERROR, "Input dict is NULL"); + goto out; + } + + ret = dict_get_int32 (dict, "cmd", &cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not fetch status cmd"); + goto out; + } + + switch (cmd) { + case GF_SNAP_STATUS_TYPE_ALL: + { + break; + } + case GF_SNAP_STATUS_TYPE_SNAP: + { + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not fetch snapname"); + goto out; + } + + if (!glusterd_find_snap_by_name (snapname)) { + ret = gf_asprintf (op_errstr, "Snap (%s) " + "not found", snapname); + if (ret < 0) { + goto out; + } + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Snap (%s) " + "not found", snapname); + goto out; + } + break; + } + case GF_SNAP_STATUS_TYPE_VOL: + { + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not fetch volname"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + ret = gf_asprintf (op_errstr, "Volume (%s)" + "not found", volname); + if (ret < 0) { + goto out; + } + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Volume " + "%s not present", volname); + goto out; + } + break; + + } + default: + { + gf_log (this->name, GF_LOG_ERROR, "Invalid command"); + break; + } + } + ret = 0; + +out: + return ret; +} + +int32_t +glusterd_snapshot_remove_commit (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int32_t ret = -1; + char *snapname = NULL; + char *dup_snapname = NULL; + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + glusterd_volinfo_t *snap_volinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + GF_ASSERT (op_errstr); + + priv = this->private; + GF_ASSERT (priv); + + if (!dict || !op_errstr) { + gf_log (this->name, GF_LOG_ERROR, "input parameters NULL"); + goto out; + } + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Getting the snap name " + "failed"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + gf_log (this->name, GF_LOG_ERROR, "Snap %s does not exist", + snapname); + ret = -1; + goto out; + } + + if (is_origin_glusterd (dict) == _gf_true) { + /* TODO : As of now there is only volume in snapshot. + * Change this when multiple volume snapshot is introduced + */ + snap_volinfo = list_entry (snap->volumes.next, + glusterd_volinfo_t, + vol_list); + if (!snap_volinfo) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch snap_volinfo"); + ret = -1; + goto out; + } + + /* From origin glusterd check if * + * any peers with snap bricks is down */ + ret = glusterd_find_missed_snap (rsp_dict, snap_volinfo, + snap_volinfo->volname, + &priv->peers, + GF_SNAP_OPTION_TYPE_DELETE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to find missed snap deletes"); + goto out; + } + } + + ret = glusterd_snap_remove (rsp_dict, snap, _gf_true, _gf_false); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to remove snap %s", + snapname); + goto out; + } + + dup_snapname = gf_strdup (snapname); + if (!dup_snapname) { + gf_log (this->name, GF_LOG_ERROR, "Strdup failed"); + ret = -1; + goto out; + } + + ret = dict_set_dynstr (rsp_dict, "snapname", dup_snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set the snapname"); + GF_FREE (dup_snapname); + goto out; + } + + ret = 0; +out: + return ret; +} + +int32_t +glusterd_do_snap_cleanup (dict_t *dict, char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + char *name = NULL; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + glusterd_snap_t *snap = NULL; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + GF_ASSERT (conf); + + if (!dict || !op_errstr) { + gf_log (this->name, GF_LOG_ERROR, "input parameters NULL"); + goto out; + } + + ret = dict_get_str (dict, "snapname", &name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "getting the snap " + "name failed (volume: %s)", volinfo->volname); + goto out; + } + + /* + If the snapname is not found that means the failure happened at + staging, or in commit, before the snap object is created, in which + case there is nothing to cleanup. So set ret to 0. + */ + snap = glusterd_find_snap_by_name (name); + if (!snap) { + gf_log (this->name, GF_LOG_INFO, "snap %s is not found", name); + ret = 0; + goto out; + } + + ret = glusterd_snap_remove (rsp_dict, snap, _gf_true, _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "removing the snap %s failed", + name); + goto out; + } + + name = NULL; + + ret = 0; + +out: + + return ret; +} + +/* In case of a successful, delete or create operation, during post_validate * + * look for missed snap operations and update the missed snap lists */ +int32_t +glusterd_snapshot_update_snaps_post_validate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int32_t ret = -1; + int32_t missed_snap_count = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + GF_ASSERT (op_errstr); + + ret = dict_get_int32 (dict, "missed_snap_count", + &missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "No missed snaps"); + ret = 0; + goto out; + } + + ret = glusterd_store_update_missed_snaps (dict, missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to update missed_snaps_list"); + goto out; + } + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_snapshot_create_commit (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int ret = -1; + int64_t i = 0; + int64_t volcount = 0; + char *snapname = NULL; + char *volname = NULL; + char *tmp_name = NULL; + char key[PATH_MAX] = ""; + xlator_t *this = NULL; + glusterd_snap_t *snap = NULL; + glusterd_volinfo_t *origin_vol = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_conf_t *priv = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(dict); + GF_ASSERT(op_errstr); + GF_ASSERT(rsp_dict); + priv = this->private; + GF_ASSERT(priv); + + ret = dict_get_int64 (dict, "volcount", &volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "get the volume count"); + goto out; + } + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snapname"); + goto out; + } + tmp_name = gf_strdup (snapname); + if (!tmp_name) { + gf_log (this->name, GF_LOG_ERROR, "Out of memory"); + ret = -1; + goto out; + } + + ret = dict_set_dynstr (rsp_dict, "snapname", tmp_name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set snapname in rsp_dict"); + GF_FREE (tmp_name); + goto out; + } + tmp_name = NULL; + + snap = glusterd_create_snap_object (dict, rsp_dict); + if (!snap) { + gf_log (this->name, GF_LOG_ERROR, "creating the" + "snap object %s failed", snapname); + ret = -1; + goto out; + } + + for (i = 1; i <= volcount; i++) { + snprintf (key, sizeof (key), "volname%ld", i); + ret = dict_get_str (dict, key, &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get volume name"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &origin_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get the volinfo for " + "the volume %s", volname); + goto out; + } + + /* TODO: Create a stub where the bricks are + added parallely by worker threads so that + the snap creating happens parallely. */ + snap_vol = glusterd_do_snap_vol (origin_vol, snap, dict, + rsp_dict, i); + if (!snap_vol) { + ret = -1; + gf_log (this->name, GF_LOG_WARNING, "taking the " + "snapshot of the volume %s failed", volname); + goto out; + } + } + + snap->snap_status = GD_SNAP_STATUS_IN_USE; + ret = glusterd_store_snap (snap); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Could not store snap" + "object %s", snap->snapname); + goto out; + } + + ret = 0; + +out: + if (ret) { + if (snap) + glusterd_snap_remove (rsp_dict, snap, + _gf_true, _gf_true); + snap = NULL; + } + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +snap_max_hard_limit_set_commit (dict_t *dict, uint64_t value, + char *volname, char **op_errstr) +{ + char err_str[PATH_MAX] = ""; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (volname); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + + /* TODO: Initiate auto deletion when there is a limit change */ + if (!volname) { + /* For system limit */ + conf->snap_max_hard_limit = value; + + ret = glusterd_store_global_info (this); + if (ret) { + snprintf (err_str, PATH_MAX, "Failed to store " + "snap-max-hard-limit for system"); + goto out; + } + } else { + /* For one volume */ + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, PATH_MAX, "Failed to get the" + " volinfo for volume %s", volname); + goto out; + } + + volinfo->snap_max_hard_limit = value; + + ret = glusterd_store_volinfo (volinfo, + GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + snprintf (err_str, PATH_MAX, "Failed to store " + "snap-max-hard-limit for volume %s", volname); + goto out; + } + } + + ret = 0; +out: + if (ret) { + *op_errstr = gf_strdup (err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + } + return ret; +} + +int +snap_max_limits_display_commit (dict_t *rsp_dict, char *volname, + char **op_errstr) +{ + char err_str[PATH_MAX] = ""; + char buf[PATH_MAX] = ""; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = -1; + uint64_t active_hard_limit = 0; + uint64_t snap_max_limit = 0; + uint64_t soft_limit_value = -1; + uint64_t count = 0; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (volname); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + + if (!volname) { + /* For system limit */ + list_for_each_entry (volinfo, &conf->volumes, vol_list) { + if (volinfo->is_snap_volume == _gf_true) + continue; + snap_max_limit = volinfo->snap_max_hard_limit; + if (snap_max_limit > conf->snap_max_hard_limit) + active_hard_limit = conf->snap_max_hard_limit; + else + active_hard_limit = snap_max_limit; + soft_limit_value = (active_hard_limit * + conf->snap_max_soft_limit) / 100; + + snprintf (buf, sizeof(buf), "volume%ld-volname", count); + ret = dict_set_str (rsp_dict, buf, volinfo->volname); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-hard-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, snap_max_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-active-hard-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, + active_hard_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-soft-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, soft_limit_value); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + count++; + } + + ret = dict_set_uint64 (rsp_dict, "voldisplaycount", count); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set voldisplaycount"); + goto out; + } + } else { + /* For one volume */ + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, PATH_MAX, "Failed to get the" + " volinfo for volume %s", volname); + goto out; + } + + snap_max_limit = volinfo->snap_max_hard_limit; + if (snap_max_limit > conf->snap_max_hard_limit) + active_hard_limit = conf->snap_max_hard_limit; + else + active_hard_limit = snap_max_limit; + + soft_limit_value = (active_hard_limit * + conf->snap_max_soft_limit) / 100; + + snprintf (buf, sizeof(buf), "volume%ld-volname", count); + ret = dict_set_str (rsp_dict, buf, volinfo->volname); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-hard-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, snap_max_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-active-hard-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, active_hard_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-soft-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, soft_limit_value); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + count++; + + ret = dict_set_uint64 (rsp_dict, "voldisplaycount", count); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set voldisplaycount"); + goto out; + } + + } + + ret = dict_set_uint64 (rsp_dict, "snap-max-hard-limit", + conf->snap_max_hard_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set sys-snap-max-hard-limit "); + goto out; + } + + ret = dict_set_uint64 (rsp_dict, "snap-max-soft-limit", + conf->snap_max_soft_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set sys-snap-max-hard-limit "); + goto out; + } + + ret = 0; +out: + if (ret) { + *op_errstr = gf_strdup (err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + } + return ret; +} + +int +glusterd_snapshot_config_commit (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + char *volname = NULL; + xlator_t *this = NULL; + int ret = -1; + char err_str[PATH_MAX] = {0,}; + glusterd_conf_t *conf = NULL; + int config_command = 0; + uint64_t hard_limit = 0; + uint64_t soft_limit = 0; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + + ret = dict_get_int32 (dict, "config-command", &config_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get config-command type"); + goto out; + } + + /* Ignore the return value of the following dict_get, + * as they are optional + */ + ret = dict_get_str (dict, "volname", &volname); + + ret = dict_get_uint64 (dict, "snap-max-hard-limit", &hard_limit); + + ret = dict_get_uint64 (dict, "snap-max-soft-limit", &soft_limit); + + switch (config_command) { + case GF_SNAP_CONFIG_TYPE_SET: + if (hard_limit) { + /* Commit ops for snap-max-hard-limit */ + ret = snap_max_hard_limit_set_commit (dict, hard_limit, + volname, + op_errstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "snap-max-hard-limit set " + "commit failed."); + goto out; + } + } + + if (soft_limit) { + /* For system limit */ + conf->snap_max_soft_limit = soft_limit; + + ret = glusterd_store_global_info (this); + if (ret) { + snprintf (err_str, PATH_MAX, "Failed to store " + "snap-max-soft-limit for system"); + *op_errstr = gf_strdup (err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", + err_str); + goto out; + } + } + break; + + case GF_SNAP_CONFIG_DISPLAY: + /* Reading data from local node only */ + if (!is_origin_glusterd (dict)) { + ret = 0; + break; + } + + ret = snap_max_limits_display_commit (rsp_dict, volname, + op_errstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "snap-max-limit " + "display commit failed."); + goto out; + } + break; + default: + break; + } + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_get_brick_lvm_details (dict_t *rsp_dict, + glusterd_brickinfo_t *brickinfo, char *volname, + char *device, char *key_prefix) +{ + + int ret = -1; + glusterd_conf_t *priv = NULL; + runner_t runner = {0,}; + xlator_t *this = NULL; + char msg[PATH_MAX] = ""; + char buf[PATH_MAX] = ""; + char *ptr = NULL; + char *token = NULL; + char key[PATH_MAX] = ""; + char *value = NULL; + + GF_ASSERT (rsp_dict); + GF_ASSERT (brickinfo); + GF_ASSERT (volname); + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + device = glusterd_get_brick_mount_details (brickinfo); + if (!device) { + gf_log (this->name, GF_LOG_ERROR, "Getting device name for " + "the brick %s:%s failed", brickinfo->hostname, + brickinfo->path); + goto out; + } + runinit (&runner); + snprintf (msg, sizeof (msg), "running lvs command, " + "for getting snap status"); + /* Using lvs command fetch the Volume Group name, + * Percentage of data filled and Logical Volume size + * + * "-o" argument is used to get the desired information, + * example : "lvs /dev/VolGroup/thin_vol -o vgname,lv_size", + * will get us Volume Group name and Logical Volume size. + * + * Here separator used is ":", + * for the above given command with separator ":", + * The output will be "vgname:lvsize" + */ + runner_add_args (&runner, "lvs", device, "--noheading", "-o", + "vg_name,data_percent,lv_size", + "--separator", ":", NULL); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + ret = runner_start (&runner); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not perform lvs action"); + goto end; + } + do { + ptr = fgets (buf, sizeof (buf), + runner_chio (&runner, STDOUT_FILENO)); + + if (ptr == NULL) + break; + token = strtok (buf, ":"); + if (token != NULL) { + while (token && token[0] == ' ') + token++; + if (!token) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "Invalid vg entry"); + goto end; + } + value = gf_strdup (token); + if (!value) { + ret = -1; + goto end; + } + ret = snprintf (key, sizeof (key), "%s.vgname", + key_prefix); + if (ret < 0) { + goto end; + } + + ret = dict_set_dynstr (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save vgname "); + goto end; + } + } + + token = strtok (NULL, ":"); + if (token != NULL) { + value = gf_strdup (token); + if (!value) { + ret = -1; + goto end; + } + ret = snprintf (key, sizeof (key), "%s.data", + key_prefix); + if (ret < 0) { + goto end; + } + + ret = dict_set_dynstr (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save data percent "); + goto end; + } + } + token = strtok (NULL, ":"); + if (token != NULL) { + value = gf_strdup (token); + if (!value) { + ret = -1; + goto end; + } + ret = snprintf (key, sizeof (key), "%s.lvsize", + key_prefix); + if (ret < 0) { + goto end; + } + + ret = dict_set_dynstr (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save meta data percent "); + goto end; + } + } + + } while (ptr != NULL); + + ret = 0; + +end: + runner_end (&runner); + +out: + if (ret && value) { + GF_FREE (value); + } + + return ret; +} + +int +glusterd_get_single_brick_status (char **op_errstr, dict_t *rsp_dict, + char *keyprefix, int index, + glusterd_volinfo_t *snap_volinfo, + glusterd_brickinfo_t *brickinfo) +{ + int ret = -1; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + char key[PATH_MAX] = ""; + char *device = NULL; + char *value = NULL; + char brick_path[PATH_MAX] = ""; + char pidfile[PATH_MAX] = ""; + pid_t pid = -1; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + GF_ASSERT (keyprefix); + GF_ASSERT (snap_volinfo); + GF_ASSERT (brickinfo); + + ret = snprintf (key, sizeof (key), "%s.brick%d.path", keyprefix, + index); + if (ret < 0) { + goto out; + } + + ret = snprintf (brick_path, sizeof (brick_path), + "%s:%s", brickinfo->hostname, brickinfo->path); + if (ret < 0) { + goto out; + } + + value = gf_strdup (brick_path); + if (!value) { + ret = -1; + goto out; + } + + ret = dict_set_dynstr (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to store " + "brick_path %s", brickinfo->path); + goto out; + } + + if (brickinfo->snap_status == -1) { + /* Setting vgname as "Pending Snapshot" */ + value = gf_strdup ("Pending Snapshot"); + if (!value) { + ret = -1; + goto out; + } + + snprintf (key, sizeof (key), "%s.brick%d.vgname", + keyprefix, index); + ret = dict_set_dynstr (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save vgname "); + goto out; + } + + ret = 0; + goto out; + } + value = NULL; + + ret = snprintf (key, sizeof (key), "%s.brick%d.status", + keyprefix, index); + if (ret < 0) { + goto out; + } + + if (brickinfo->status == GF_BRICK_STOPPED) { + value = gf_strdup ("No"); + if (!value) { + ret = -1; + goto out; + } + ret = dict_set_str (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save brick status"); + goto out; + } + value = NULL; + } else { + value = gf_strdup ("Yes"); + if (!value) { + ret = -1; + goto out; + } + ret = dict_set_str (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save brick status"); + goto out; + } + value = NULL; + + GLUSTERD_GET_BRICK_PIDFILE (pidfile, snap_volinfo, + brickinfo, priv); + ret = glusterd_is_service_running (pidfile, &pid); + + ret = snprintf (key, sizeof (key), "%s.brick%d.pid", + keyprefix, index); + if (ret < 0) { + goto out; + } + + ret = dict_set_int32 (rsp_dict, key, pid); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save pid %d", pid); + goto out; + } + } + + ret = snprintf (key, sizeof (key), "%s.brick%d", + keyprefix, index); + if (ret < 0) { + goto out; + } + + ret = glusterd_get_brick_lvm_details (rsp_dict, brickinfo, + snap_volinfo->volname, + device, key); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "brick LVM details"); + goto out; + } +out: + if (ret && value) { + GF_FREE (value); + } + + return ret; +} + +int +glusterd_get_single_snap_status (char **op_errstr, dict_t *rsp_dict, + char *keyprefix, glusterd_snap_t *snap) +{ + int ret = -1; + xlator_t *this = NULL; + char key[PATH_MAX] = ""; + char brickkey[PATH_MAX] = ""; + glusterd_volinfo_t *snap_volinfo = NULL; + glusterd_volinfo_t *tmp_volinfo = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + int volcount = 0; + int brickcount = 0; + + this = THIS; + GF_ASSERT (this); + + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + GF_ASSERT (keyprefix); + GF_ASSERT (snap); + + list_for_each_entry_safe (snap_volinfo, tmp_volinfo, &snap->volumes, + vol_list) { + ret = snprintf (key, sizeof (key), "%s.vol%d", keyprefix, + volcount); + if (ret < 0) { + goto out; + } + list_for_each_entry (brickinfo, &snap_volinfo->bricks, + brick_list) { + if (!glusterd_is_local_brick (this, snap_volinfo, + brickinfo)) { + brickcount++; + continue; + } + + ret = glusterd_get_single_brick_status (op_errstr, + rsp_dict, key, brickcount, + snap_volinfo, brickinfo); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Getting " + "single snap status failed"); + goto out; + } + brickcount++; + } + ret = snprintf (brickkey, sizeof (brickkey), "%s.brickcount", + key); + if (ret < 0) { + goto out; + } + + ret = dict_set_int32 (rsp_dict, brickkey, brickcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save brick count"); + goto out; + } + volcount++; + } + + ret = snprintf (key, sizeof (key), "%s.volcount", keyprefix); + if (ret < 0) { + goto out; + } + + ret = dict_set_int32 (rsp_dict, key, volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save volcount"); + goto out; + } + +out: + + return ret; +} + +int +glusterd_get_each_snap_object_status (char **op_errstr, dict_t *rsp_dict, + glusterd_snap_t *snap, char *keyprefix) +{ + int ret = -1; + char key[PATH_MAX] = ""; + char *temp = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap); + GF_ASSERT (keyprefix); + + /* TODO : Get all the snap volume info present in snap object, + * as of now, There will be only one snapvolinfo per snap object + */ + ret = snprintf (key, sizeof (key), "%s.snapname", keyprefix); + if (ret < 0) { + goto out; + } + + temp = gf_strdup (snap->snapname); + if (temp == NULL) { + ret = -1; + goto out; + } + ret = dict_set_dynstr (rsp_dict, key, temp); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not save " + "snap name"); + goto out; + } + + temp = NULL; + + ret = snprintf (key, sizeof (key), "%s.uuid", keyprefix); + if (ret < 0) { + goto out; + } + + temp = gf_strdup (uuid_utoa (snap->snap_id)); + if (temp == NULL) { + ret = -1; + goto out; + } + + ret = dict_set_dynstr (rsp_dict, key, temp); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not save " + "snap UUID"); + goto out; + } + + temp = NULL; + + ret = glusterd_get_single_snap_status (op_errstr, rsp_dict, keyprefix, + snap); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not get single snap status"); + goto out; + } + + ret = snprintf (key, sizeof (key), "%s.volcount", keyprefix); + if (ret < 0) { + goto out; + } + + ret = dict_set_int32 (rsp_dict, key, 1); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not save volcount"); + goto out; + } +out: + if (ret && temp) + GF_FREE (temp); + + return ret; +} + +int +glusterd_get_snap_status_of_volume (char **op_errstr, dict_t *rsp_dict, + char *volname, char *keyprefix) { + int ret = -1; + glusterd_volinfo_t *snap_volinfo = NULL; + glusterd_volinfo_t *temp_volinfo = NULL; + glusterd_volinfo_t *volinfo = NULL; + char key[PATH_MAX] = ""; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int i = 0; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + GF_ASSERT (volname); + GF_ASSERT (keyprefix); + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volinfo of " + "volume %s", volname); + goto out; + } + + list_for_each_entry_safe (snap_volinfo, temp_volinfo, + &volinfo->snap_volumes, snapvol_list) { + ret = snprintf (key, sizeof (key), "status.snap%d", i); + if (ret < 0) { + goto out; + } + + ret = glusterd_get_each_snap_object_status (op_errstr, + rsp_dict, snap_volinfo->snapshot, key); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Function : " + "glusterd_get_single_snap_status failed"); + goto out; + } + i++; + } + + ret = dict_set_int32 (rsp_dict, "status.snapcount", i); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to save snapcount"); + ret = -1; + goto out; + } +out: + return ret; +} + +int +glusterd_get_all_snapshot_status (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int32_t i = 0; + int ret = -1; + char key[PATH_MAX] = ""; + glusterd_conf_t *priv = NULL; + glusterd_snap_t *snap = NULL; + glusterd_snap_t *tmp_snap = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + list_for_each_entry_safe (snap, tmp_snap, + &priv->snapshots, snap_list) { + ret = snprintf (key, sizeof (key), "status.snap%d", i); + if (ret < 0) { + goto out; + } + + ret = glusterd_get_each_snap_object_status (op_errstr, + rsp_dict, snap, key); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not get " + "the details of a snap object: %s", + snap->snapname); + goto out; + } + i++; + } + + ret = dict_set_int32 (rsp_dict, "status.snapcount", i); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not save snapcount"); + goto out; + } + + ret = 0; +out: + return ret; +} + + +int +glusterd_snapshot_status_commit (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + xlator_t *this = NULL; + int ret = -1; + glusterd_conf_t *conf = NULL; + char *get_buffer = NULL; + int32_t cmd = -1; + char *snapname = NULL; + glusterd_snap_t *snap = NULL; + char *volname = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + ret = dict_get_int32 (dict, "cmd", &cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get status cmd type"); + goto out; + } + + ret = dict_set_int32 (rsp_dict, "cmd", cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save status cmd in rsp dictionary"); + goto out; + } + switch (cmd) { + case GF_SNAP_STATUS_TYPE_ALL: + { + ret = glusterd_get_all_snapshot_status (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "get snapshot status"); + goto out; + } + break; + } + case GF_SNAP_STATUS_TYPE_SNAP: + { + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "get snap name"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + ret = gf_asprintf (op_errstr, "Snap (%s) " + "not found", snapname); + if (ret < 0) { + goto out; + } + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "get snap volinfo"); + goto out; + } + ret = glusterd_get_each_snap_object_status (op_errstr, + rsp_dict, snap, "status.snap0"); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "get status of snap %s", get_buffer); + goto out; + } + break; + } + case GF_SNAP_STATUS_TYPE_VOL: + { + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to" + " get volume name"); + goto out; + } + + ret = glusterd_get_snap_status_of_volume (op_errstr, + rsp_dict, volname, "status.vol0"); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Function :" + " glusterd_get_snap_status_of_volume " + "failed"); + goto out; + } + } + } + ret = 0; +out: + return ret; +} + +int32_t +glusterd_snapshot_create_postvalidate (dict_t *dict, int32_t op_ret, + char **op_errstr, dict_t *rsp_dict) +{ + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int ret = -1; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + priv = this->private; + GF_ASSERT (priv); + + if (op_ret) { + ret = glusterd_do_snap_cleanup (dict, op_errstr, rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "cleanup operation " + "failed"); + goto out; + } + } else { + ret = glusterd_snapshot_update_snaps_post_validate (dict, + op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "create snapshot"); + goto out; + } + } + + ret = 0; +out: + return ret; +} + +int32_t +glusterd_snapshot (dict_t *dict, char **op_errstr, dict_t *rsp_dict) +{ + + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int32_t snap_command = 0; + int ret = -1; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_int32 (dict, "type", &snap_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case (GF_SNAP_OPTION_TYPE_CREATE): + ret = glusterd_snapshot_create_commit (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "create snapshot"); + goto out; + } + break; + + case GF_SNAP_OPTION_TYPE_CONFIG: + ret = glusterd_snapshot_config_commit (dict, op_errstr, + rsp_dict); + break; + + case GF_SNAP_OPTION_TYPE_DELETE: + ret = glusterd_snapshot_remove_commit (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "delete snapshot"); + goto out; + } + break; + + case GF_SNAP_OPTION_TYPE_RESTORE: + ret = glusterd_snapshot_restore (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Failed to " + "restore snapshot"); + goto out; + } + + break; + + case GF_SNAP_OPTION_TYPE_STATUS: + ret = glusterd_snapshot_status_commit (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "show snapshot status"); + goto out; + } + break; + + + default: + gf_log (this->name, GF_LOG_WARNING, "invalid snap command"); + goto out; + break; + } + + ret = 0; + +out: + return ret; +} + +int +glusterd_snapshot_brickop (dict_t *dict, char **op_errstr, dict_t *rsp_dict) +{ + int ret = -1; + int64_t vol_count = 0; + int64_t count = 1; + char key[1024] = {0,}; + char *volname = NULL; + int32_t snap_command = 0; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + ret = dict_get_int32 (dict, "type", &snap_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case GF_SNAP_OPTION_TYPE_CREATE: + ret = dict_get_int64 (dict, "volcount", &vol_count); + if (ret) + goto out; + while (count <= vol_count) { + snprintf (key, 1024, "volname%"PRId64, count); + ret = dict_get_str (dict, key, &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get volname"); + goto out; + } + ret = dict_set_str (dict, "volname", volname); + if (ret) + goto out; + + ret = gd_brick_op_phase (GD_OP_SNAP, NULL, dict, + op_errstr); + if (ret) + goto out; + volname = NULL; + count++; + } + + dict_del (dict, "volname"); + ret = 0; + break; + case GF_SNAP_OPTION_TYPE_DELETE: + break; + default: + break; + } + +out: + return ret; +} + +int +glusterd_snapshot_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int snap_command = 0; + xlator_t *this = NULL; + int ret = -1; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + ret = dict_get_int32 (dict, "type", &snap_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case (GF_SNAP_OPTION_TYPE_CREATE): + ret = glusterd_snapshot_create_prevalidate (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot create " + "pre-validation failed"); + goto out; + } + break; + + case (GF_SNAP_OPTION_TYPE_CONFIG): + ret = glusterd_snapshot_config_prevalidate (dict, op_errstr); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot config " + "pre-validation failed"); + goto out; + } + break; + + case GF_SNAP_OPTION_TYPE_RESTORE: + ret = glusterd_snapshot_restore_prevalidate (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot restore " + "validation failed"); + goto out; + } + break; + case GF_SNAP_OPTION_TYPE_DELETE: + ret = glusterd_snapshot_remove_prevalidate (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot remove " + "validation failed"); + goto out; + } + break; + + case GF_SNAP_OPTION_TYPE_STATUS: + ret = glusterd_snapshot_status_prevalidate (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot status " + "validation failed"); + goto out; + } + break; + + default: + gf_log (this->name, GF_LOG_WARNING, "invalid snap command"); + goto out; + } + + ret = 0; +out: + return ret; +} + +int +glusterd_snapshot_postvalidate (dict_t *dict, int32_t op_ret, char **op_errstr, + dict_t *rsp_dict) +{ + int snap_command = 0; + xlator_t *this = NULL; + int ret = -1; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + ret = dict_get_int32 (dict, "type", &snap_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case GF_SNAP_OPTION_TYPE_CREATE: + ret = glusterd_snapshot_create_postvalidate (dict, op_ret, + op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot create " + "post-validation failed"); + goto out; + } + break; + + case GF_SNAP_OPTION_TYPE_DELETE: + case GF_SNAP_OPTION_TYPE_RESTORE: + ret = glusterd_snapshot_update_snaps_post_validate (dict, + op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "update missed snaps list"); + goto out; + } + break; + + default: + gf_log (this->name, GF_LOG_WARNING, "invalid snap command"); + goto out; + } + + ret = 0; +out: + return ret; +} + +int +glusterd_handle_snapshot_fn (rpcsvc_request_t *req) +{ + int32_t ret = 0; + dict_t *dict = NULL; + gf_cli_req cli_req = {{0},}; + glusterd_op_t cli_op = GD_OP_SNAP; + int type = 0; + glusterd_conf_t *conf = NULL; + char *host_uuid = NULL; + char err_str[2048] = {0,}; + xlator_t *this = NULL; + char *volname = NULL; + + GF_ASSERT (req); + + this = THIS; + GF_ASSERT (this); + conf = this->private; + GF_ASSERT (conf); + + ret = xdr_to_generic (req->msg[0], &cli_req, + (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (cli_req.dict.dict_len > 0) { + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (cli_req.dict.dict_val, + cli_req.dict.dict_len, + &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); + goto out; + } + + dict->extra_stdfree = cli_req.dict.dict_val; + + host_uuid = gf_strdup (uuid_utoa(MY_UUID)); + if (host_uuid == NULL) { + snprintf (err_str, sizeof (err_str), "Failed to get " + "the uuid of local glusterd"); + ret = -1; + goto out; + } + ret = dict_set_dynstr (dict, "host-uuid", host_uuid); + if (ret) { + GF_FREE (host_uuid); + goto out; + } + + + } else { + gf_log (this->name, GF_LOG_ERROR, "request dict length is %d", + cli_req.dict.dict_len); + goto out; + } + + ret = dict_get_int32 (dict, "type", &type); + if (ret < 0) { + snprintf (err_str, sizeof (err_str), "Command type not found"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + goto out; + } + + switch (type) { + case GF_SNAP_OPTION_TYPE_CREATE: + ret = glusterd_handle_snapshot_create (req, cli_op, dict, + err_str, sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot create " + "failed: %s", err_str); + } + break; + case GF_SNAP_OPTION_TYPE_RESTORE: + ret = glusterd_handle_snapshot_restore (req, cli_op, dict, + err_str, sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot restore " + "failed: %s", err_str); + } + + break; + case GF_SNAP_OPTION_TYPE_INFO: + ret = glusterd_handle_snapshot_info (req, cli_op, dict, + err_str, sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot info " + "failed"); + } + break; + case GF_SNAP_OPTION_TYPE_LIST: + ret = glusterd_handle_snapshot_list (req, cli_op, dict, + err_str, sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot list " + "failed"); + } + break; + case GF_SNAP_OPTION_TYPE_CONFIG: + /* TODO : Type of lock to be taken when we are setting + * limits system wide + */ + ret = dict_get_str (dict, "volname", &volname); + if (!volname) { + ret = dict_set_int32 (dict, "hold_vol_locks", + _gf_false); + if (ret) { + gf_log ("cli", GF_LOG_ERROR, + "Unable to set hold_vol_locks value " + "as _gf_false"); + goto out; + } + + } + ret = glusterd_mgmt_v3_initiate_all_phases (req, cli_op, dict); + break; + case GF_SNAP_OPTION_TYPE_DELETE: + ret = glusterd_handle_snapshot_remove (req, cli_op, dict, + err_str, + sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot delete " + "failed: %s", err_str); + } + break; + case GF_SNAP_OPTION_TYPE_START: + case GF_SNAP_OPTION_TYPE_STOP: + case GF_SNAP_OPTION_TYPE_STATUS: + ret = glusterd_handle_snapshot_status (req, cli_op, dict, + err_str, + sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot status " + "failed: %s", err_str); + } + break; + default: + gf_log (this->name, GF_LOG_ERROR, "Unkown snapshot request " + "type (%d)", type); + ret = -1; /* Failure */ + } + +out: + if (ret) { + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); + ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, + dict, err_str); + } + + return ret; +} + +int +glusterd_handle_snapshot (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, glusterd_handle_snapshot_fn); +} + +static inline void +glusterd_free_snap_op (glusterd_snap_op_t *snap_op) +{ + if (snap_op) { + if (snap_op->brick_path) + GF_FREE (snap_op->brick_path); + + GF_FREE (snap_op); + } +} + +/* Look for duplicates and accordingly update the list */ +int32_t +glusterd_update_missed_snap_entry (glusterd_missed_snap_info *missed_snapinfo, + glusterd_snap_op_t *missed_snap_op) +{ + int32_t ret = -1; + glusterd_snap_op_t *snap_opinfo = NULL; + gf_boolean_t match = _gf_false; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(missed_snapinfo); + GF_ASSERT(missed_snap_op); + + list_for_each_entry (snap_opinfo, &missed_snapinfo->snap_ops, + snap_ops_list) { + if ((!strcmp (snap_opinfo->brick_path, + missed_snap_op->brick_path)) && + (snap_opinfo->op == missed_snap_op->op)) { + /* If two entries have conflicting status + * GD_MISSED_SNAP_DONE takes precedence + */ + if ((snap_opinfo->status == GD_MISSED_SNAP_PENDING) && + (missed_snap_op->status == GD_MISSED_SNAP_DONE)) { + snap_opinfo->status = GD_MISSED_SNAP_DONE; + gf_log (this->name, GF_LOG_INFO, + "Updating missed snap status " + "for %s:%d:%s:%d as DONE", + missed_snapinfo->node_snap_info, + snap_opinfo->brick_num, + snap_opinfo->brick_path, + snap_opinfo->op); + ret = 0; + glusterd_free_snap_op (missed_snap_op); + goto out; + } + match = _gf_true; + break; + } else if ((snap_opinfo->brick_num == + missed_snap_op->brick_num) && + (snap_opinfo->op == GF_SNAP_OPTION_TYPE_CREATE) && + (missed_snap_op->op == + GF_SNAP_OPTION_TYPE_DELETE)) { + /* Optimizing create and delete entries for the same + * brick and same node + */ + gf_log (this->name, GF_LOG_INFO, + "Updating missed snap status " + "for %s:%d:%s:%d as DONE", + missed_snapinfo->node_snap_info, + snap_opinfo->brick_num, + snap_opinfo->brick_path, + snap_opinfo->op); + snap_opinfo->status = GD_MISSED_SNAP_DONE; + ret = 0; + glusterd_free_snap_op (missed_snap_op); + goto out; + } + } + + if (match == _gf_true) { + gf_log (this->name, GF_LOG_INFO, + "Duplicate entry. Not updating"); + glusterd_free_snap_op (missed_snap_op); + } else { + list_add_tail (&missed_snap_op->snap_ops_list, + &missed_snapinfo->snap_ops); + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Add new missed snap entry to the missed_snaps list. */ +int32_t +glusterd_store_missed_snaps_list (char *missed_info, int32_t brick_num, + char *brick_path, int32_t snap_op, + int32_t snap_status) +{ + int32_t ret = -1; + glusterd_missed_snap_info *missed_snapinfo = NULL; + glusterd_snap_op_t *missed_snap_op = NULL; + glusterd_conf_t *priv = NULL; + gf_boolean_t match = _gf_false; + gf_boolean_t free_missed_snap_info = _gf_false; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(missed_info); + GF_ASSERT(brick_path); + + priv = this->private; + GF_ASSERT (priv); + + /* Create the snap_op object consisting of the * + * snap id and the op */ + ret = glusterd_missed_snap_op_new (&missed_snap_op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create new missed snap object."); + ret = -1; + goto out; + } + + missed_snap_op->brick_path = gf_strdup(brick_path); + if (!missed_snap_op->brick_path) { + ret = -1; + goto out; + } + missed_snap_op->brick_num = brick_num; + missed_snap_op->op = snap_op; + missed_snap_op->status = snap_status; + + /* Look for other entries for the same node and same snap */ + list_for_each_entry (missed_snapinfo, &priv->missed_snaps_list, + missed_snaps) { + if (!strcmp (missed_snapinfo->node_snap_info, + missed_info)) { + /* Found missed snapshot info for * + * the same node and same snap */ + match = _gf_true; + break; + } + } + + if (match == _gf_false) { + /* First snap op missed for the brick */ + ret = glusterd_missed_snapinfo_new (&missed_snapinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create missed snapinfo"); + goto out; + } + free_missed_snap_info = _gf_true; + missed_snapinfo->node_snap_info = gf_strdup(missed_info); + if (!missed_snapinfo->node_snap_info) { + ret = -1; + goto out; + } + + list_add_tail (&missed_snap_op->snap_ops_list, + &missed_snapinfo->snap_ops); + list_add_tail (&missed_snapinfo->missed_snaps, + &priv->missed_snaps_list); + + ret = 0; + goto out; + } else { + ret = glusterd_update_missed_snap_entry (missed_snapinfo, + missed_snap_op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to update existing missed snap entry."); + goto out; + } + } + +out: + if (ret) { + glusterd_free_snap_op (missed_snap_op); + + if (missed_snapinfo && + (free_missed_snap_info == _gf_true)) { + if (missed_snapinfo->node_snap_info) + GF_FREE (missed_snapinfo->node_snap_info); + + GF_FREE (missed_snapinfo); + } + } + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Add missing snap entries to the in-memory conf->missed_snap_list */ +int32_t +glusterd_add_missed_snaps_to_list (dict_t *dict, int32_t missed_snap_count) +{ + char *buf = NULL; + char *tmp = NULL; + char *save_ptr = NULL; + char *nodeid = NULL; + char *snap_uuid = NULL; + char *brick_path = NULL; + char missed_info[PATH_MAX] = ""; + char name_buf[PATH_MAX] = ""; + int32_t i = -1; + int32_t ret = -1; + int32_t brick_num = -1; + int32_t snap_op = -1; + int32_t snap_status = -1; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(dict); + + priv = this->private; + GF_ASSERT (priv); + + /* We can update the missed_snaps_list without acquiring * + * any additional locks as big lock will be held. */ + for (i = 0; i < missed_snap_count; i++) { + snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d", + i); + ret = dict_get_str (dict, name_buf, &buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch %s", name_buf); + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, "missed_snap_entry = %s", + buf); + + /* Need to make a duplicate string coz the same dictionary * + * is resent to the non-originator nodes */ + tmp = gf_strdup (buf); + if (!tmp) { + ret = -1; + goto out; + } + + /* Fetch the node-id, snap-id, brick_num, + * brick_path, snap_op and snap status + */ + nodeid = strtok_r (tmp, ":", &save_ptr); + snap_uuid = strtok_r (NULL, "=", &save_ptr); + brick_num = atoi(strtok_r (NULL, ":", &save_ptr)); + brick_path = strtok_r (NULL, ":", &save_ptr); + snap_op = atoi(strtok_r (NULL, ":", &save_ptr)); + snap_status = atoi(strtok_r (NULL, ":", &save_ptr)); + + if (!nodeid || !snap_uuid || !brick_path || + brick_num < 1 || snap_op < 1 || + snap_status < 1) { + gf_log (this->name, GF_LOG_ERROR, + "Invalid missed_snap_entry"); + ret = -1; + goto out; + } + + snprintf (missed_info, sizeof(missed_info), "%s:%s", + nodeid, snap_uuid); + + ret = glusterd_store_missed_snaps_list (missed_info, + brick_num, + brick_path, + snap_op, + snap_status); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to store missed snaps_list"); + goto out; + } + + GF_FREE (tmp); + tmp = NULL; + } + +out: + if (tmp) + GF_FREE (tmp); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-store.c b/xlators/mgmt/glusterd/src/glusterd-store.c index 823fb37cd..1c2ec58e8 100644 --- a/xlators/mgmt/glusterd/src/glusterd-store.c +++ b/xlators/mgmt/glusterd/src/glusterd-store.c @@ -1,21 +1,13 @@ /* - Copyright (c) 2007-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. + Copyright (c) 2007-2013 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. */ + #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" @@ -42,6 +34,8 @@ #include "glusterd-sm.h" #include "glusterd-op-sm.h" #include "glusterd-utils.h" +#include "glusterd-hooks.h" +#include "store.h" #include "glusterd-store.h" #include "rpc-clnt.h" @@ -51,102 +45,8 @@ #include <inttypes.h> #include <dirent.h> -static int32_t -glusterd_store_mkdir (char *path) -{ - int32_t ret = -1; - - ret = mkdir (path, 0777); - - if ((-1 == ret) && (EEXIST != errno)) { - gf_log ("", GF_LOG_ERROR, "mkdir() failed on path %s," - "errno: %s", path, strerror (errno)); - } else { - ret = 0; - } - - return ret; -} - -int32_t -glusterd_store_handle_create_on_absence (glusterd_store_handle_t **shandle, - char *path) -{ - GF_ASSERT (shandle); - int32_t ret = 0; - - if (*shandle == NULL) { - ret = glusterd_store_handle_new (path, shandle); - - if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to create store" - " handle for path: %s", path); - } - } - return ret; -} - -int32_t -glusterd_store_mkstemp (glusterd_store_handle_t *shandle) -{ - int fd = -1; - char tmppath[PATH_MAX] = {0,}; - - GF_ASSERT (shandle); - GF_ASSERT (shandle->path); - - snprintf (tmppath, sizeof (tmppath), "%s.tmp", shandle->path); - fd = open (tmppath, O_RDWR | O_CREAT | O_TRUNC, 0644); - if (fd <= 0) { - gf_log ("glusterd", GF_LOG_ERROR, "Failed to open %s, " - "error: %s", tmppath, strerror (errno)); - } - - return fd; -} - -int32_t -glusterd_store_rename_tmppath (glusterd_store_handle_t *shandle) -{ - int32_t ret = -1; - char tmppath[PATH_MAX] = {0,}; - - GF_ASSERT (shandle); - GF_ASSERT (shandle->path); - - snprintf (tmppath, sizeof (tmppath), "%s.tmp", shandle->path); - ret = rename (tmppath, shandle->path); - if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "Failed to mv %s to %s, " - "error: %s", tmppath, shandle->path, strerror (errno)); - } - - return ret; -} - -int32_t -glusterd_store_unlink_tmppath (glusterd_store_handle_t *shandle) -{ - int32_t ret = -1; - char tmppath[PATH_MAX] = {0,}; - - GF_ASSERT (shandle); - GF_ASSERT (shandle->path); - - snprintf (tmppath, sizeof (tmppath), "%s.tmp", shandle->path); - ret = unlink (tmppath); - if (ret && (errno != ENOENT)) { - gf_log ("glusterd", GF_LOG_ERROR, "Failed to mv %s to %s, " - "error: %s", tmppath, shandle->path, strerror (errno)); - } else { - ret = 0; - } - - return ret; -} - -static void -glusterd_replace_slash_with_hipen (char *str) +void +glusterd_replace_slash_with_hyphen (char *str) { char *ptr = NULL; @@ -171,7 +71,7 @@ glusterd_store_create_brick_dir (glusterd_volinfo_t *volinfo) GF_ASSERT (priv); GLUSTERD_GET_BRICK_DIR (brickdirpath, volinfo, priv); - ret = glusterd_store_mkdir (brickdirpath); + ret = gf_store_mkdir (brickdirpath); return ret; } @@ -185,7 +85,7 @@ glusterd_store_key_vol_brick_set (glusterd_brickinfo_t *brickinfo, GF_ASSERT (len >= PATH_MAX); snprintf (key_vol_brick, len, "%s", brickinfo->path); - glusterd_replace_slash_with_hipen (key_vol_brick); + glusterd_replace_slash_with_hyphen (key_vol_brick); } static void @@ -232,20 +132,31 @@ glusterd_store_is_valid_brickpath (char *volname, char *brick) glusterd_brickinfo_t *brickinfo = NULL; glusterd_volinfo_t *volinfo = NULL; int32_t ret = 0; + size_t volname_len = strlen (volname); + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); - ret = glusterd_brickinfo_from_brick (brick, &brickinfo); + ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo); if (ret) { - gf_log ("", GF_LOG_WARNING, "brick path validation failed"); + gf_log (this->name, GF_LOG_WARNING, "Failed to create brick " + "info for brick %s", brick); ret = 0; goto out; } ret = glusterd_volinfo_new (&volinfo); if (ret) { - gf_log ("", GF_LOG_WARNING, "brick path validation failed"); + gf_log (this->name, GF_LOG_WARNING, "Failed to create volinfo"); + ret = 0; + goto out; + } + if (volname_len >= sizeof (volinfo->volname)) { + gf_log (this->name, GF_LOG_WARNING, "volume name too long"); ret = 0; goto out; } - strncpy (volinfo->volname, volname, sizeof (volinfo->volname)); + memcpy (volinfo->volname, volname, volname_len+1); glusterd_store_brickinfopath_set (volinfo, brickinfo, brickpath, sizeof (brickpath)); @@ -273,7 +184,11 @@ glusterd_store_volinfo_brick_fname_write (int vol_fd, brick_count); glusterd_store_brickinfofname_set (brickinfo, brickfname, sizeof (brickfname)); - ret = glusterd_store_save_value (vol_fd, key, brickfname); + ret = gf_store_save_value (vol_fd, key, brickfname); + if (ret) + goto out; + +out: return ret; } @@ -288,9 +203,9 @@ glusterd_store_create_brick_shandle_on_absence (glusterd_volinfo_t *volinfo, GF_ASSERT (brickinfo); glusterd_store_brickinfopath_set (volinfo, brickinfo, brickpath, - sizeof (brickpath)); - ret = glusterd_store_handle_create_on_absence (&brickinfo->shandle, - brickpath); + sizeof (brickpath)); + ret = gf_store_handle_create_on_absence (&brickinfo->shandle, + brickpath); return ret; } @@ -303,30 +218,50 @@ glusterd_store_brickinfo_write (int fd, glusterd_brickinfo_t *brickinfo) GF_ASSERT (brickinfo); GF_ASSERT (fd > 0); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_HOSTNAME, - brickinfo->hostname); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_HOSTNAME, + brickinfo->hostname); if (ret) goto out; - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_PATH, - brickinfo->path); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_PATH, + brickinfo->path); if (ret) goto out; snprintf (value, sizeof(value), "%d", brickinfo->port); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_PORT, - value); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_PORT, value); snprintf (value, sizeof(value), "%d", brickinfo->rdma_port); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_RDMA_PORT, - value); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_RDMA_PORT, + value); snprintf (value, sizeof(value), "%d", brickinfo->decommissioned); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED, - value); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED, + value); + if (ret) + goto out; + + if (strlen(brickinfo->device_path) > 0) { + snprintf (value, sizeof(value), "%s", brickinfo->device_path); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH, + value); + if (ret) + goto out; + } + snprintf (value, sizeof(value), "%d", brickinfo->snap_status); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS, + value); + if (ret) + goto out; + + if (!brickinfo->vg[0]) + goto out; + + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_VGNAME, + brickinfo->vg); out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -337,7 +272,7 @@ glusterd_store_perform_brick_store (glusterd_brickinfo_t *brickinfo) int32_t ret = -1; GF_ASSERT (brickinfo); - fd = glusterd_store_mkstemp (brickinfo->shandle); + fd = gf_store_mkstemp (brickinfo->shandle); if (fd <= 0) { ret = -1; goto out; @@ -347,13 +282,12 @@ glusterd_store_perform_brick_store (glusterd_brickinfo_t *brickinfo) if (ret) goto out; - ret = glusterd_store_rename_tmppath (brickinfo->shandle); out: if (ret && (fd > 0)) - glusterd_store_unlink_tmppath (brickinfo->shandle); + gf_store_unlink_tmppath (brickinfo->shandle); if (fd > 0) close (fd); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -382,31 +316,29 @@ glusterd_store_brickinfo (glusterd_volinfo_t *volinfo, goto out; ret = glusterd_store_perform_brick_store (brickinfo); + out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } int32_t -glusterd_store_delete_brick (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo) +glusterd_store_delete_brick (glusterd_brickinfo_t *brickinfo, char *delete_path) { int32_t ret = -1; glusterd_conf_t *priv = NULL; - char path[PATH_MAX] = {0,}; char brickpath[PATH_MAX] = {0,}; char *ptr = NULL; char *tmppath = NULL; + xlator_t *this = NULL; - GF_ASSERT (volinfo); + this = THIS; + GF_ASSERT (this); GF_ASSERT (brickinfo); - priv = THIS->private; - + priv = this->private; GF_ASSERT (priv); - GLUSTERD_GET_BRICK_DIR (path, volinfo, priv); - tmppath = gf_strdup (brickinfo->path); ptr = strchr (tmppath, '/'); @@ -416,16 +348,17 @@ glusterd_store_delete_brick (glusterd_volinfo_t *volinfo, ptr = strchr (tmppath, '/'); } - snprintf (brickpath, sizeof (brickpath), "%s/%s:%s", - path, brickinfo->hostname, tmppath); + snprintf (brickpath, sizeof (brickpath), + "%s/"GLUSTERD_BRICK_INFO_DIR"/%s:%s", delete_path, + brickinfo->hostname, tmppath); GF_FREE (tmppath); ret = unlink (brickpath); if ((ret < 0) && (errno != ENOENT)) { - gf_log ("", GF_LOG_ERROR, "Unlink failed on %s, reason: %s", - brickpath, strerror(errno)); + gf_log (this->name, GF_LOG_DEBUG, "Unlink failed on %s, " + "reason: %s", brickpath, strerror(errno)); ret = -1; goto out; } else { @@ -434,15 +367,15 @@ glusterd_store_delete_brick (glusterd_volinfo_t *volinfo, out: if (brickinfo->shandle) { - glusterd_store_handle_destroy (brickinfo->shandle); + gf_store_handle_destroy (brickinfo->shandle); brickinfo->shandle = NULL; } - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } int32_t -glusterd_store_remove_bricks (glusterd_volinfo_t *volinfo) +glusterd_store_remove_bricks (glusterd_volinfo_t *volinfo, char *delete_path) { int32_t ret = 0; glusterd_brickinfo_t *tmp = NULL; @@ -451,19 +384,24 @@ glusterd_store_remove_bricks (glusterd_volinfo_t *volinfo) DIR *dir = NULL; struct dirent *entry = NULL; char path[PATH_MAX] = {0,}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (volinfo); list_for_each_entry (tmp, &volinfo->bricks, brick_list) { - ret = glusterd_store_delete_brick (volinfo, tmp); + ret = glusterd_store_delete_brick (tmp, delete_path); if (ret) goto out; } - priv = THIS->private; + priv = this->private; GF_ASSERT (priv); - GLUSTERD_GET_BRICK_DIR (brickdir, volinfo, priv); + snprintf (brickdir, sizeof (brickdir), "%s/%s", delete_path, + GLUSTERD_BRICK_INFO_DIR); dir = opendir (brickdir); @@ -474,7 +412,7 @@ glusterd_store_remove_bricks (glusterd_volinfo_t *volinfo) brickdir, entry->d_name); ret = unlink (path); if (ret && errno != ENOENT) { - gf_log ("", GF_LOG_ERROR, "Unable to unlink %s, " + gf_log (this->name, GF_LOG_DEBUG, "Unable to unlink %s, " "reason: %s", path, strerror(errno)); } glusterd_for_each_entry (entry, dir); @@ -485,17 +423,21 @@ glusterd_store_remove_bricks (glusterd_volinfo_t *volinfo) ret = rmdir (brickdir); out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } -static void +static int _storeslaves (dict_t *this, char *key, data_t *value, void *data) { int32_t ret = 0; - glusterd_store_handle_t *shandle = NULL; + gf_store_handle_t *shandle = NULL; + xlator_t *xl = NULL; + + xl = THIS; + GF_ASSERT (xl); - shandle = (glusterd_store_handle_t*)data; + shandle = (gf_store_handle_t*)data; GF_ASSERT (shandle); GF_ASSERT (shandle->fd > 0); @@ -504,32 +446,37 @@ _storeslaves (dict_t *this, char *key, data_t *value, void *data) GF_ASSERT (value && value->data); if ((!shandle) || (shandle->fd <= 0) || (!shandle->path)) - return; + return -1; if (!key) - return; + return -1; if (!value || !value->data) - return; + return -1; - gf_log ("", GF_LOG_DEBUG, "Storing in volinfo:key= %s, val=%s", + gf_log (xl->name, GF_LOG_DEBUG, "Storing in volinfo:key= %s, val=%s", key, value->data); - ret = glusterd_store_save_value (shandle->fd, key, (char*)value->data); + ret = gf_store_save_value (shandle->fd, key, (char*)value->data); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to write into store" + gf_log (xl->name, GF_LOG_ERROR, "Unable to write into store" " handle for path: %s", shandle->path); - return; + return -1; } + return 0; } -void _storeopts (dict_t *this, char *key, data_t *value, void *data) +int _storeopts (dict_t *this, char *key, data_t *value, void *data) { int32_t ret = 0; int32_t exists = 0; - glusterd_store_handle_t *shandle = NULL; + gf_store_handle_t *shandle = NULL; + xlator_t *xl = NULL; - shandle = (glusterd_store_handle_t*)data; + xl = THIS; + GF_ASSERT (xl); + + shandle = (gf_store_handle_t*)data; GF_ASSERT (shandle); GF_ASSERT (shandle->fd > 0); @@ -538,94 +485,158 @@ void _storeopts (dict_t *this, char *key, data_t *value, void *data) GF_ASSERT (value && value->data); if ((!shandle) || (shandle->fd <= 0) || (!shandle->path)) - return; + return -1; if (!key) - return; + return -1; if (!value || !value->data) - return; + return -1; + + if (is_key_glusterd_hooks_friendly (key)) { + exists = 1; + + } else { + exists = glusterd_check_option_exists (key, NULL); + } - exists = glusterd_check_option_exists (key, NULL); if (1 == exists) { - gf_log ("", GF_LOG_DEBUG, "Storing in volinfo:key= %s, val=%s", - key, value->data); + gf_log (xl->name, GF_LOG_DEBUG, "Storing in volinfo:key= %s, " + "val=%s", key, value->data); + } else { - gf_log ("", GF_LOG_DEBUG, "Discarding:key= %s, val=%s", + gf_log (xl->name, GF_LOG_DEBUG, "Discarding:key= %s, val=%s", key, value->data); - return; + return 0; } - ret = glusterd_store_save_value (shandle->fd, key, (char*)value->data); + ret = gf_store_save_value (shandle->fd, key, (char*)value->data); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to write into store" - " handle for path: %s", shandle->path); - return; + gf_log (xl->name, GF_LOG_ERROR, "Unable to write into store" + " handle for path: %s", shandle->path); + return -1; } + return 0; } int32_t glusterd_volume_exclude_options_write (int fd, glusterd_volinfo_t *volinfo) { + char *str = NULL; + char buf[PATH_MAX] = {0,}; + int32_t ret = -1; + GF_ASSERT (fd > 0); GF_ASSERT (volinfo); - char buf[PATH_MAX] = {0,}; - int32_t ret = -1; - snprintf (buf, sizeof (buf), "%d", volinfo->type); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_TYPE, buf); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_TYPE, buf); if (ret) goto out; snprintf (buf, sizeof (buf), "%d", volinfo->brick_count); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_COUNT, buf); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_COUNT, buf); if (ret) goto out; snprintf (buf, sizeof (buf), "%d", volinfo->status); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_STATUS, buf); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_STATUS, buf); if (ret) goto out; snprintf (buf, sizeof (buf), "%d", volinfo->sub_count); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_SUB_COUNT, - buf); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_SUB_COUNT, buf); if (ret) goto out; snprintf (buf, sizeof (buf), "%d", volinfo->stripe_count); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_STRIPE_CNT, - buf); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_STRIPE_CNT, buf); if (ret) goto out; snprintf (buf, sizeof (buf), "%d", volinfo->replica_count); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_REPLICA_CNT, - buf); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_REPLICA_CNT, + buf); if (ret) goto out; snprintf (buf, sizeof (buf), "%d", volinfo->version); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_VERSION, - buf); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_VERSION, buf); if (ret) goto out; snprintf (buf, sizeof (buf), "%d", volinfo->transport_type); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_TRANSPORT, - buf); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_TRANSPORT, buf); if (ret) goto out; - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_ID, - uuid_utoa (volinfo->volume_id)); + snprintf (buf, sizeof (buf), "%s", volinfo->parent_volname); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PARENT_VOLNAME, buf); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to store " + GLUSTERD_STORE_KEY_PARENT_VOLNAME); + goto out; + } + + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_ID, + uuid_utoa (volinfo->volume_id)); if (ret) goto out; + str = glusterd_auth_get_username (volinfo); + if (str) { + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_USERNAME, + str); + if (ret) + goto out; + } + + str = glusterd_auth_get_password (volinfo); + if (str) { + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PASSWORD, + str); + if (ret) + goto out; + } + + snprintf (buf, sizeof (buf), "%d", volinfo->op_version); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_OP_VERSION, buf); + if (ret) + goto out; + + snprintf (buf, sizeof (buf), "%d", volinfo->client_op_version); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION, + buf); + if (ret) + goto out; + if (volinfo->caps) { + snprintf (buf, sizeof (buf), "%d", volinfo->caps); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_CAPS, + buf); + if (ret) + goto out; + } + + snprintf (buf, sizeof (buf), "%d", volinfo->is_volume_restored); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_IS_RESTORED, buf); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Unable to write is_volume_restored"); + goto out; + } + + snprintf (buf, sizeof (buf), "%"PRIu64, volinfo->snap_max_hard_limit); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT, + buf); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Unable to write snap-max-hard-limit"); + goto out; + } + out: if (ret) - gf_log ("", GF_LOG_ERROR, "Unable to write volume values" - " for %s", volinfo->volname); + gf_log (THIS->name, GF_LOG_ERROR, "Unable to write volume " + "values for %s", volinfo->volname); return ret; } @@ -639,8 +650,7 @@ glusterd_store_voldirpath_set (glusterd_volinfo_t *volinfo, char *voldirpath, priv = THIS->private; GF_ASSERT (priv); - snprintf (voldirpath, len, "%s/%s/%s", priv->workdir, - GLUSTERD_VOLUME_DIR_PREFIX, volinfo->volname); + GLUSTERD_GET_VOLUME_DIR (voldirpath, volinfo, priv); } static int32_t @@ -653,8 +663,30 @@ glusterd_store_create_volume_dir (glusterd_volinfo_t *volinfo) glusterd_store_voldirpath_set (volinfo, voldirpath, sizeof (voldirpath)); - ret = glusterd_store_mkdir (voldirpath); - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + ret = gf_store_mkdir (voldirpath); + + gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); + return ret; +} + +static int32_t +glusterd_store_create_snap_dir (glusterd_snap_t *snap) +{ + int32_t ret = -1; + char snapdirpath[PATH_MAX] = {0,}; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + GF_ASSERT (snap); + + GLUSTERD_GET_SNAP_DIR (snapdirpath, snap, priv); + + ret = mkdir_p (snapdirpath, 0755, _gf_true); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to create snaps dir " + "%s", snapdirpath); + } return ret; } @@ -662,7 +694,7 @@ int32_t glusterd_store_volinfo_write (int fd, glusterd_volinfo_t *volinfo) { int32_t ret = -1; - glusterd_store_handle_t *shandle = NULL; + gf_store_handle_t *shandle = NULL; GF_ASSERT (fd > 0); GF_ASSERT (volinfo); GF_ASSERT (volinfo->shandle); @@ -678,7 +710,50 @@ glusterd_store_volinfo_write (int fd, glusterd_volinfo_t *volinfo) dict_foreach (volinfo->gsync_slaves, _storeslaves, shandle); shandle->fd = 0; out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_store_snapinfo_write (glusterd_snap_t *snap) +{ + int32_t ret = -1; + int fd = 0; + char buf[PATH_MAX] = ""; + + GF_ASSERT (snap); + + fd = gf_store_mkstemp (snap->shandle); + if (fd <= 0) + goto out; + + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_ID, + uuid_utoa (snap->snap_id)); + if (ret) + goto out; + + snprintf (buf, sizeof (buf), "%d", snap->snap_status); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_STATUS, buf); + if (ret) + goto out; + + snprintf (buf, sizeof (buf), "%d", snap->snap_restored); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_RESTORED, buf); + if (ret) + goto out; + + if (snap->description) { + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_DESC, + snap->description); + if (ret) + goto out; + } + + snprintf (buf, sizeof (buf), "%ld", snap->time_stamp); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_TIMESTAMP, buf); + +out: + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -689,7 +764,7 @@ glusterd_store_rbstatepath_set (glusterd_volinfo_t *volinfo, char *rbstatepath, char voldirpath[PATH_MAX] = {0,}; GF_ASSERT (volinfo); GF_ASSERT (rbstatepath); - GF_ASSERT (len >= PATH_MAX); + GF_ASSERT (len <= PATH_MAX); glusterd_store_voldirpath_set (volinfo, voldirpath, sizeof (voldirpath)); @@ -704,13 +779,59 @@ glusterd_store_volfpath_set (glusterd_volinfo_t *volinfo, char *volfpath, char voldirpath[PATH_MAX] = {0,}; GF_ASSERT (volinfo); GF_ASSERT (volfpath); - GF_ASSERT (len >= PATH_MAX); + GF_ASSERT (len <= PATH_MAX); glusterd_store_voldirpath_set (volinfo, voldirpath, sizeof (voldirpath)); snprintf (volfpath, len, "%s/%s", voldirpath, GLUSTERD_VOLUME_INFO_FILE); } +static void +glusterd_store_node_state_path_set (glusterd_volinfo_t *volinfo, + char *node_statepath, size_t len) +{ + char voldirpath[PATH_MAX] = {0,}; + GF_ASSERT (volinfo); + GF_ASSERT (node_statepath); + GF_ASSERT (len <= PATH_MAX); + + glusterd_store_voldirpath_set (volinfo, voldirpath, + sizeof (voldirpath)); + snprintf (node_statepath, len, "%s/%s", voldirpath, + GLUSTERD_NODE_STATE_FILE); +} + +static void +glusterd_store_missed_snaps_list_path_set (char *missed_snaps_list, + size_t len) +{ + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + GF_ASSERT (missed_snaps_list); + GF_ASSERT (len <= PATH_MAX); + + snprintf (missed_snaps_list, len, "%s/snaps/" + GLUSTERD_MISSED_SNAPS_LIST_FILE, priv->workdir); +} + +static void +glusterd_store_snapfpath_set (glusterd_snap_t *snap, char *snap_fpath, + size_t len) +{ + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + GF_ASSERT (snap); + GF_ASSERT (snap_fpath); + GF_ASSERT (len <= PATH_MAX); + + snprintf (snap_fpath, len, "%s/snaps/%s/%s", priv->workdir, + snap->snapname, GLUSTERD_SNAP_INFO_FILE); +} + int32_t glusterd_store_create_rbstate_shandle_on_absence (glusterd_volinfo_t *volinfo) { @@ -720,8 +841,8 @@ glusterd_store_create_rbstate_shandle_on_absence (glusterd_volinfo_t *volinfo) GF_ASSERT (volinfo); glusterd_store_rbstatepath_set (volinfo, rbstatepath, sizeof (rbstatepath)); - ret = glusterd_store_handle_create_on_absence (&volinfo->rb_shandle, - rbstatepath); + ret = gf_store_handle_create_on_absence (&volinfo->rb_shandle, + rbstatepath); return ret; } @@ -734,8 +855,60 @@ glusterd_store_create_vol_shandle_on_absence (glusterd_volinfo_t *volinfo) GF_ASSERT (volinfo); glusterd_store_volfpath_set (volinfo, volfpath, sizeof (volfpath)); - ret = glusterd_store_handle_create_on_absence (&volinfo->shandle, - volfpath); + ret = gf_store_handle_create_on_absence (&volinfo->shandle, volfpath); + return ret; +} + +int32_t +glusterd_store_create_nodestate_sh_on_absence (glusterd_volinfo_t *volinfo) +{ + char node_state_path[PATH_MAX] = {0}; + int32_t ret = 0; + + GF_ASSERT (volinfo); + + glusterd_store_node_state_path_set (volinfo, node_state_path, + sizeof (node_state_path)); + ret = + gf_store_handle_create_on_absence (&volinfo->node_state_shandle, + node_state_path); + + return ret; +} + +static int32_t +glusterd_store_create_missed_snaps_list_shandle_on_absence () +{ + char missed_snaps_list[PATH_MAX] = ""; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + priv = this->private; + GF_ASSERT (priv); + + glusterd_store_missed_snaps_list_path_set (missed_snaps_list, + sizeof(missed_snaps_list)); + + ret = gf_store_handle_create_on_absence + (&priv->missed_snaps_list_shandle, + missed_snaps_list); + return ret; +} + +int32_t +glusterd_store_create_snap_shandle_on_absence (glusterd_snap_t *snap) +{ + char snapfpath[PATH_MAX] = {0}; + int32_t ret = 0; + + GF_ASSERT (snap); + + glusterd_store_snapfpath_set (snap, snapfpath, sizeof (snapfpath)); + ret = gf_store_handle_create_on_absence (&snap->shandle, snapfpath); return ret; } @@ -756,7 +929,7 @@ glusterd_store_brickinfos (glusterd_volinfo_t *volinfo, int vol_fd) brick_count++; } out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -764,36 +937,56 @@ int32_t glusterd_store_rbstate_write (int fd, glusterd_volinfo_t *volinfo) { int ret = -1; + int port = 0; char buf[PATH_MAX] = {0, }; GF_ASSERT (fd > 0); GF_ASSERT (volinfo); - snprintf (buf, sizeof (buf), "%d", volinfo->rb_status); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_RB_STATUS, - buf); + snprintf (buf, sizeof (buf), "%d", volinfo->rep_brick.rb_status); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_RB_STATUS, buf); if (ret) goto out; - if (volinfo->rb_status > GF_RB_STATUS_NONE) { + if (volinfo->rep_brick.rb_status > GF_RB_STATUS_NONE) { snprintf (buf, sizeof (buf), "%s:%s", - volinfo->src_brick->hostname, - volinfo->src_brick->path); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_RB_SRC_BRICK, - buf); + volinfo->rep_brick.src_brick->hostname, + volinfo->rep_brick.src_brick->path); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_RB_SRC_BRICK, + buf); if (ret) goto out; snprintf (buf, sizeof (buf), "%s:%s", - volinfo->dst_brick->hostname, - volinfo->dst_brick->path); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_RB_DST_BRICK, - buf); + volinfo->rep_brick.dst_brick->hostname, + volinfo->rep_brick.dst_brick->path); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_RB_DST_BRICK, + buf); if (ret) goto out; + + switch (volinfo->transport_type) { + case GF_TRANSPORT_RDMA: + port = volinfo->rep_brick.dst_brick->rdma_port; + break; + + case GF_TRANSPORT_TCP: + case GF_TRANSPORT_BOTH_TCP_RDMA: + port = volinfo->rep_brick.dst_brick->port; + break; + } + + snprintf (buf, sizeof (buf), "%d", port); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_RB_DST_PORT, + buf); + if (ret) + goto out; + uuid_unparse (volinfo->rep_brick.rb_id, buf); + ret = gf_store_save_value (fd, GF_REPLACE_BRICK_TID_KEY, buf); } + ret = 0; out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; @@ -806,7 +999,7 @@ glusterd_store_perform_rbstate_store (glusterd_volinfo_t *volinfo) int32_t ret = -1; GF_ASSERT (volinfo); - fd = glusterd_store_mkstemp (volinfo->rb_shandle); + fd = gf_store_mkstemp (volinfo->rb_shandle); if (fd <= 0) { ret = -1; goto out; @@ -816,14 +1009,79 @@ glusterd_store_perform_rbstate_store (glusterd_volinfo_t *volinfo) if (ret) goto out; - ret = glusterd_store_rename_tmppath (volinfo->rb_shandle); + ret = gf_store_rename_tmppath (volinfo->rb_shandle); + if (ret) + goto out; out: if (ret && (fd > 0)) - glusterd_store_unlink_tmppath (volinfo->rb_shandle); + gf_store_unlink_tmppath (volinfo->rb_shandle); if (fd > 0) close (fd); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_store_node_state_write (int fd, glusterd_volinfo_t *volinfo) +{ + int ret = -1; + char buf[PATH_MAX] = {0, }; + + GF_ASSERT (fd > 0); + GF_ASSERT (volinfo); + + if (volinfo->rebal.defrag_cmd == GF_DEFRAG_CMD_STATUS) { + ret = 0; + goto out; + } + + snprintf (buf, sizeof (buf), "%d", volinfo->rebal.defrag_cmd); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_DEFRAG, buf); + if (ret) + goto out; + + snprintf (buf, sizeof (buf), "%d", volinfo->rebal.op); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_DEFRAG_OP, buf); + if (ret) + goto out; + + if (volinfo->rebal.defrag_cmd) { + uuid_unparse (volinfo->rebal.rebalance_id, buf); + ret = gf_store_save_value (fd, GF_REBALANCE_TID_KEY, buf); + } +out: + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_store_perform_node_state_store (glusterd_volinfo_t *volinfo) +{ + int fd = -1; + int32_t ret = -1; + GF_ASSERT (volinfo); + + fd = gf_store_mkstemp (volinfo->node_state_shandle); + if (fd <= 0) { + ret = -1; + goto out; + } + + ret = glusterd_store_node_state_write (fd, volinfo); + if (ret) + goto out; + + ret = gf_store_rename_tmppath (volinfo->node_state_shandle); + if (ret) + goto out; + +out: + if (ret && (fd > 0)) + gf_store_unlink_tmppath (volinfo->node_state_shandle); + if (fd > 0) + close (fd); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -834,7 +1092,7 @@ glusterd_store_perform_volume_store (glusterd_volinfo_t *volinfo) int32_t ret = -1; GF_ASSERT (volinfo); - fd = glusterd_store_mkstemp (volinfo->shandle); + fd = gf_store_mkstemp (volinfo->shandle); if (fd <= 0) { ret = -1; goto out; @@ -848,13 +1106,12 @@ glusterd_store_perform_volume_store (glusterd_volinfo_t *volinfo) if (ret) goto out; - ret = glusterd_store_rename_tmppath (volinfo->shandle); out: if (ret && (fd > 0)) - glusterd_store_unlink_tmppath (volinfo->shandle); + gf_store_unlink_tmppath (volinfo->shandle); if (fd > 0) close (fd); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -870,7 +1127,148 @@ glusterd_perform_volinfo_version_action (glusterd_volinfo_t *volinfo, case GLUSTERD_VOLINFO_VER_AC_INCREMENT: volinfo->version++; break; + case GLUSTERD_VOLINFO_VER_AC_DECREMENT: + volinfo->version--; + break; + } +} + +void +glusterd_store_bricks_cleanup_tmp (glusterd_volinfo_t *volinfo) +{ + glusterd_brickinfo_t *brickinfo = NULL; + + GF_ASSERT (volinfo); + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + gf_store_unlink_tmppath (brickinfo->shandle); + } +} + +void +glusterd_store_volume_cleanup_tmp (glusterd_volinfo_t *volinfo) +{ + GF_ASSERT (volinfo); + + glusterd_store_bricks_cleanup_tmp (volinfo); + + gf_store_unlink_tmppath (volinfo->shandle); + + gf_store_unlink_tmppath (volinfo->rb_shandle); + + gf_store_unlink_tmppath (volinfo->node_state_shandle); +} + +void +glusterd_store_snap_cleanup_tmp (glusterd_snap_t *snap) +{ + GF_ASSERT (snap); + + gf_store_unlink_tmppath (snap->shandle); +} + +int32_t +glusterd_store_brickinfos_atomic_update (glusterd_volinfo_t *volinfo) +{ + int ret = -1; + glusterd_brickinfo_t *brickinfo = NULL; + + GF_ASSERT (volinfo); + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + ret = gf_store_rename_tmppath (brickinfo->shandle); + if (ret) + goto out; + } +out: + return ret; +} + +int32_t +glusterd_store_volinfo_atomic_update (glusterd_volinfo_t *volinfo) +{ + int ret = -1; + GF_ASSERT (volinfo); + + ret = gf_store_rename_tmppath (volinfo->shandle); + if (ret) + goto out; + +out: + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, "Couldn't rename " + "temporary file(s): Reason %s", strerror (errno)); + return ret; +} + +int32_t +glusterd_store_volume_atomic_update (glusterd_volinfo_t *volinfo) +{ + int ret = -1; + GF_ASSERT (volinfo); + + ret = glusterd_store_brickinfos_atomic_update (volinfo); + if (ret) + goto out; + + ret = glusterd_store_volinfo_atomic_update (volinfo); + +out: + return ret; +} + +int32_t +glusterd_store_snap_atomic_update (glusterd_snap_t *snap) +{ + int ret = -1; + GF_ASSERT (snap); + + ret = gf_store_rename_tmppath (snap->shandle); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, "Couldn't rename " + "temporary file(s): Reason %s", strerror (errno)); + + return ret; +} + +int32_t +glusterd_store_snap (glusterd_snap_t *snap) +{ + int32_t ret = -1; + + GF_ASSERT (snap); + + ret = glusterd_store_create_snap_dir (snap); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to create snap dir"); + goto out; + } + + ret = glusterd_store_create_snap_shandle_on_absence (snap); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to create snap info " + "file"); + goto out; + } + + ret = glusterd_store_snapinfo_write (snap); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to write snap info"); + goto out; } + + ret = glusterd_store_snap_atomic_update (snap); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR,"Failed to do automic update"); + goto out; + } + +out: + if (ret) + glusterd_store_snap_cleanup_tmp (snap); + + gf_log (THIS->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; } int32_t @@ -893,61 +1291,113 @@ glusterd_store_volinfo (glusterd_volinfo_t *volinfo, glusterd_volinfo_ver_ac_t a if (ret) goto out; + ret = glusterd_store_create_nodestate_sh_on_absence (volinfo); + if (ret) + goto out; + ret = glusterd_store_perform_volume_store (volinfo); if (ret) goto out; + ret = glusterd_store_volume_atomic_update (volinfo); + if (ret) { + glusterd_perform_volinfo_version_action (volinfo, + GLUSTERD_VOLINFO_VER_AC_DECREMENT); + goto out; + } + ret = glusterd_store_perform_rbstate_store (volinfo); if (ret) goto out; + ret = glusterd_store_perform_node_state_store (volinfo); + if (ret) + goto out; + //checksum should be computed at the end ret = glusterd_volume_compute_cksum (volinfo); if (ret) goto out; + out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + if (ret) + glusterd_store_volume_cleanup_tmp (volinfo); + + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } - int32_t glusterd_store_delete_volume (glusterd_volinfo_t *volinfo) { - char pathname[PATH_MAX] = {0,}; - int32_t ret = 0; - glusterd_conf_t *priv = NULL; - DIR *dir = NULL; - struct dirent *entry = NULL; - char path[PATH_MAX] = {0,}; - struct stat st = {0, }; + char pathname[PATH_MAX] = {0,}; + int32_t ret = 0; + glusterd_conf_t *priv = NULL; + DIR *dir = NULL; + struct dirent *entry = NULL; + char path[PATH_MAX] = {0,}; + char delete_path[PATH_MAX] = {0,}; + char trashdir[PATH_MAX] = {0,}; + struct stat st = {0, }; + xlator_t *this = NULL; + gf_boolean_t rename_fail = _gf_false; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (volinfo); - priv = THIS->private; + priv = this->private; GF_ASSERT (priv); - snprintf (pathname, sizeof (pathname), "%s/vols/%s", priv->workdir, - volinfo->volname); - dir = opendir (pathname); - if (!dir) + GLUSTERD_GET_VOLUME_DIR (pathname, volinfo, priv); + + snprintf (delete_path, sizeof (delete_path), + "%s/"GLUSTERD_TRASH"/%s.deleted", priv->workdir, + uuid_utoa (volinfo->volume_id)); + + snprintf (trashdir, sizeof (trashdir), "%s/"GLUSTERD_TRASH, + priv->workdir); + + ret = mkdir (trashdir, 0777); + if (ret && errno != EEXIST) { + gf_log (this->name, GF_LOG_ERROR, "Failed to create trash " + "directory, reason : %s", strerror (errno)); + ret = -1; goto out; - ret = glusterd_store_remove_bricks (volinfo); + } + ret = rename (pathname, delete_path); if (ret) { - gf_log ("", GF_LOG_ERROR, "Remove bricks failed for %s", + gf_log (this->name, GF_LOG_ERROR, "Failed to rename volume " + "directory for volume %s", volinfo->volname); + rename_fail = _gf_true; + goto out; + } + + dir = opendir (delete_path); + if (!dir) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to open directory %s." + " Reason : %s", delete_path, strerror (errno)); + ret = 0; + goto out; + } + ret = glusterd_store_remove_bricks (volinfo, delete_path); + + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Remove bricks failed for %s", volinfo->volname); } glusterd_for_each_entry (entry, dir); while (entry) { - snprintf (path, PATH_MAX, "%s/%s", pathname, entry->d_name); + snprintf (path, PATH_MAX, "%s/%s", delete_path, entry->d_name); ret = stat (path, &st); if (ret == -1) { - gf_log ("", GF_LOG_ERROR, "Failed to stat entry: %s:%s", - path, strerror (errno)); + gf_log (this->name, GF_LOG_DEBUG, "Failed to stat " + "entry %s : %s", path, strerror (errno)); goto stat_failed; } @@ -956,12 +1406,13 @@ glusterd_store_delete_volume (glusterd_volinfo_t *volinfo) else ret = unlink (path); - if (ret) - gf_log ("", GF_LOG_INFO, "errno:%d (%s)", errno, - strerror (errno)); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, " Failed to remove " + "%s. Reason : %s", path, strerror (errno)); + } - gf_log ("", GF_LOG_INFO, "%s %s", - ret?"Failed to remove":"Removed", + gf_log (this->name, GF_LOG_DEBUG, "%s %s", + ret ? "Failed to remove":"Removed", entry->d_name); stat_failed: memset (path, 0, sizeof(path)); @@ -970,339 +1421,320 @@ stat_failed: ret = closedir (dir); if (ret) { - gf_log ("", GF_LOG_ERROR, "Failed to close dir, errno:%d", - errno); + gf_log (this->name, GF_LOG_DEBUG, "Failed to close dir %s. " + "Reason : %s",delete_path, strerror (errno)); } - ret = rmdir (pathname); + ret = rmdir (delete_path); if (ret) { - gf_log ("", GF_LOG_ERROR, "Failed to rmdir: %s, err: %s", - pathname, strerror (errno)); + gf_log (this->name, GF_LOG_DEBUG, "Failed to rmdir: %s,err: %s", + delete_path, strerror (errno)); + } + ret = rmdir (trashdir); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to rmdir: %s, Reason:" + " %s", trashdir, strerror (errno)); } - out: if (volinfo->shandle) { - glusterd_store_handle_destroy (volinfo->shandle); + gf_store_handle_destroy (volinfo->shandle); volinfo->shandle = NULL; } - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + ret = (rename_fail == _gf_true) ? -1: 0; + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } +/*TODO: cleanup the duplicate code and implement a generic function for + * deleting snap/volume depending on the parameter flag */ +int32_t +glusterd_store_delete_snap (glusterd_snap_t *snap) +{ + char pathname[PATH_MAX] = {0,}; + int32_t ret = 0; + glusterd_conf_t *priv = NULL; + DIR *dir = NULL; + struct dirent *entry = NULL; + char path[PATH_MAX] = {0,}; + char delete_path[PATH_MAX] = {0,}; + char trashdir[PATH_MAX] = {0,}; + struct stat st = {0, }; + xlator_t *this = NULL; + gf_boolean_t rename_fail = _gf_false; -int -glusterd_store_read_and_tokenize (FILE *file, char *str, - char **iter_key, char **iter_val, - glusterd_store_op_errno_t *store_errno) -{ - int32_t ret = -1; + this = THIS; + priv = this->private; + GF_ASSERT (priv); + + GF_ASSERT (snap); + GLUSTERD_GET_SNAP_DIR (pathname, snap, priv); - GF_ASSERT (file); - GF_ASSERT (str); - GF_ASSERT (iter_key); - GF_ASSERT (iter_val); - GF_ASSERT (store_errno); + snprintf (delete_path, sizeof (delete_path), + "%s/"GLUSTERD_TRASH"/snap-%s.deleted", priv->workdir, + uuid_utoa (snap->snap_id)); - ret = fscanf (file, "%s", str); - if (ret <= 0 || feof (file)) { + snprintf (trashdir, sizeof (trashdir), "%s/"GLUSTERD_TRASH, + priv->workdir); + + ret = mkdir (trashdir, 0777); + if (ret && errno != EEXIST) { + gf_log (this->name, GF_LOG_ERROR, "Failed to create trash " + "directory, reason : %s", strerror (errno)); ret = -1; - *store_errno = GD_STORE_EOF; goto out; } - *iter_key = strtok (str, "="); - if (*iter_key == NULL) { - ret = -1; - *store_errno = GD_STORE_KEY_NULL; + ret = rename (pathname, delete_path); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to rename snap " + "directory %s to %s", snap->snapname, delete_path); + rename_fail = _gf_true; goto out; } - *iter_val = strtok (NULL, "="); - if (*iter_key == NULL) { - ret = -1; - *store_errno = GD_STORE_VALUE_NULL; + dir = opendir (delete_path); + if (!dir) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to open directory %s." + " Reason : %s", delete_path, strerror (errno)); + ret = 0; goto out; } - *store_errno = GD_STORE_SUCCESS; - ret = 0; -out: - return ret; -} - -int32_t -glusterd_store_retrieve_value (glusterd_store_handle_t *handle, - char *key, char **value) -{ - int32_t ret = -1; - char *scan_str = NULL; - char *iter_key = NULL; - char *iter_val = NULL; - char *free_str = NULL; - struct stat st = {0,}; - glusterd_store_op_errno_t store_errno = GD_STORE_SUCCESS; - - GF_ASSERT (handle); + glusterd_for_each_entry (entry, dir); + while (entry) { + snprintf (path, PATH_MAX, "%s/%s", delete_path, entry->d_name); + ret = stat (path, &st); + if (ret == -1) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to stat " + "entry %s : %s", path, strerror (errno)); + goto stat_failed; + } - handle->fd = open (handle->path, O_RDWR); + if (S_ISDIR (st.st_mode)) + ret = rmdir (path); + else + ret = unlink (path); - if (!handle->read) - handle->read = fdopen (handle->fd, "r"); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, " Failed to remove " + "%s. Reason : %s", path, strerror (errno)); + } - if (!handle->read) { - gf_log ("", GF_LOG_ERROR, "Unable to open file %s errno: %d", - handle->path, errno); - goto out; + gf_log (this->name, GF_LOG_DEBUG, "%s %s", + ret ? "Failed to remove":"Removed", + entry->d_name); +stat_failed: + memset (path, 0, sizeof(path)); + glusterd_for_each_entry (entry, dir); } - ret = fstat (handle->fd, &st); - if (ret < 0) { - gf_log ("glusterd", GF_LOG_WARNING, - "stat on file failed"); - ret = -1; - store_errno = GD_STORE_STAT_FAILED; - goto out; + ret = closedir (dir); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to close dir %s. " + "Reason : %s",delete_path, strerror (errno)); } - scan_str = GF_CALLOC (1, st.st_size, - gf_gld_mt_char); - if (scan_str == NULL) { - ret = -1; - store_errno = GD_STORE_ENOMEM; - goto out; + ret = rmdir (delete_path); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to rmdir: %s,err: %s", + delete_path, strerror (errno)); + } + ret = rmdir (trashdir); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to rmdir: %s, Reason:" + " %s", trashdir, strerror (errno)); } - free_str = scan_str; - - do { - ret = glusterd_store_read_and_tokenize (handle->read, scan_str, - &iter_key, &iter_val, - &store_errno); - if (ret < 0) { - goto out; - } - - gf_log ("", GF_LOG_DEBUG, "key %s read", iter_key); - - if (!strcmp (key, iter_key)) { - gf_log ("", GF_LOG_DEBUG, "key %s found", key); - ret = 0; - if (iter_val) - *value = gf_strdup (iter_val); - goto out; - } - } while (1); out: - if (handle->fd > 0) { - close (handle->fd); - handle->read = NULL; + if (snap->shandle) { + gf_store_handle_destroy (snap->shandle); + snap->shandle = NULL; } + ret = (rename_fail == _gf_true) ? -1: 0; - if (free_str) - GF_FREE (free_str); - + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } -int32_t -glusterd_store_save_value (int fd, char *key, char *value) +int +glusterd_store_global_info (xlator_t *this) { - int32_t ret = -1; - FILE *fp = NULL; + int ret = -1; + glusterd_conf_t *conf = NULL; + char op_version_str[15] = {0,}; + char path[PATH_MAX] = {0,}; + gf_store_handle_t *handle = NULL; + char *uuid_str = NULL; + char buf[256] = {0, }; - GF_ASSERT (fd > 0); - GF_ASSERT (key); - GF_ASSERT (value); + conf = this->private; - fp = fdopen (fd, "a+"); - if (fp == NULL) { - gf_log ("", GF_LOG_WARNING, "fdopen failed."); - ret = -1; + uuid_str = gf_strdup (uuid_utoa (MY_UUID)); + if (!uuid_str) goto out; - } - ret = fprintf (fp, "%s=%s\n", key, value); - if (ret < 0) { - gf_log ("", GF_LOG_WARNING, "Unable to store key: %s," - "value: %s, error: %s", key, value, - strerror (errno)); - ret = -1; + if (!conf->handle) { + snprintf (path, PATH_MAX, "%s/%s", conf->workdir, + GLUSTERD_INFO_FILE); + ret = gf_store_handle_new (path, &handle); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get store handle"); + goto out; + } + + conf->handle = handle; + } else + handle = conf->handle; + + /* These options need to be available for all users */ + ret = chmod (handle->path, 0644); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "chmod error for %s: %s", + GLUSTERD_INFO_FILE, strerror (errno)); goto out; } - ret = fflush (fp); - if (feof (fp)) { - gf_log ("", GF_LOG_WARNING, - "fflush failed, error: %s", - strerror (errno)); + handle->fd = gf_store_mkstemp (handle); + if (handle->fd <= 0) { ret = -1; goto out; } - ret = 0; -out: - - gf_log ("", GF_LOG_DEBUG, "returning: %d", ret); - return ret; -} - -int32_t -glusterd_store_handle_new (char *path, glusterd_store_handle_t **handle) -{ - int32_t ret = -1; - glusterd_store_handle_t *shandle = NULL; - int fd = -1; - char *spath = NULL; - - shandle = GF_CALLOC (1, sizeof (*shandle), gf_gld_mt_store_handle_t); - if (!shandle) - goto out; - - spath = gf_strdup (path); - - if (!spath) - goto out; - - fd = open (path, O_RDWR | O_CREAT | O_APPEND, 0644); - if (fd <= 0) { - gf_log ("glusterd", GF_LOG_ERROR, "Failed to open file: %s, " - "error: %s", path, strerror (errno)); + ret = gf_store_save_value (handle->fd, GLUSTERD_STORE_UUID_KEY, + uuid_str); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, + "Storing uuid failed ret = %d", ret); goto out; } - shandle->path = spath; - *handle = shandle; - - ret = 0; - -out: - if (fd > 0) - close (fd); - - if (ret == -1) { - if (spath) - GF_FREE (spath); - if (shandle) { - GF_FREE (shandle); - } + snprintf (op_version_str, 15, "%d", conf->op_version); + ret = gf_store_save_value (handle->fd, GD_OP_VERSION_KEY, + op_version_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Storing op-version failed ret = %d", ret); + goto out; } - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - return ret; -} - -int -glusterd_store_handle_retrieve (char *path, glusterd_store_handle_t **handle) -{ - int32_t ret = -1; - struct stat statbuf = {0}; - - ret = stat (path, &statbuf); + snprintf (buf, sizeof (buf), "%"PRIu64, conf->snap_max_hard_limit); + ret = gf_store_save_value (handle->fd, + GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT, buf); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "Unable to retrieve store " - "handle for %s, error: %s", path, strerror (errno)); + gf_log (this->name, GF_LOG_ERROR, + "Storing snap-max-hard-limit failed ret = %d", ret); goto out; } - ret = glusterd_store_handle_new (path, handle); -out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - return ret; -} -int32_t -glusterd_store_handle_destroy (glusterd_store_handle_t *handle) -{ - int32_t ret = -1; - - if (!handle) { - ret = 0; + snprintf (buf, sizeof (buf), "%"PRIu64, conf->snap_max_soft_limit); + ret = gf_store_save_value (handle->fd, + GLUSTERD_STORE_KEY_SNAP_MAX_SOFT_LIMIT, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Storing snap-max-soft-limit failed ret = %d", ret); goto out; } - GF_FREE (handle->path); + ret = gf_store_rename_tmppath (handle); +out: + if (ret && (handle->fd > 0)) + gf_store_unlink_tmppath (handle); - GF_FREE (handle); + if (handle->fd > 0) { + close (handle->fd); + handle->fd = 0; + } - ret = 0; + if (uuid_str) + GF_FREE (uuid_str); -out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to store glusterd global-info"); return ret; } -int32_t -glusterd_store_uuid () +int +glusterd_retrieve_op_version (xlator_t *this, int *op_version) { - glusterd_conf_t *priv = NULL; - char path[PATH_MAX] = {0,}; - int32_t ret = -1; - glusterd_store_handle_t *handle = NULL; - - priv = THIS->private; + char *op_version_str = NULL; + glusterd_conf_t *priv = NULL; + int ret = -1; + int tmp_version = 0; + char *tmp = NULL; + char path[PATH_MAX] = {0,}; + gf_store_handle_t *handle = NULL; - snprintf (path, PATH_MAX, "%s/%s", priv->workdir, - GLUSTERD_INFO_FILE); + priv = this->private; if (!priv->handle) { - ret = glusterd_store_handle_new (path, &handle); + snprintf (path, PATH_MAX, "%s/%s", priv->workdir, + GLUSTERD_INFO_FILE); + ret = gf_store_handle_retrieve (path, &handle); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get store" - " handle!"); + gf_log ("", GF_LOG_DEBUG, "Unable to get store " + "handle!"); goto out; } priv->handle = handle; - } else { - handle = priv->handle; } - handle->fd = open (handle->path, O_RDWR | O_CREAT | O_TRUNC, 0644); - if (handle->fd <= 0) { - ret = -1; + ret = gf_store_retrieve_value (priv->handle, GD_OP_VERSION_KEY, + &op_version_str); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "No previous op_version present"); goto out; } - ret = glusterd_store_save_value (handle->fd, GLUSTERD_STORE_UUID_KEY, - uuid_utoa (priv->uuid)); - if (ret) { - gf_log ("", GF_LOG_CRITICAL, "Storing uuid failed" - "ret = %d", ret); + tmp_version = strtol (op_version_str, &tmp, 10); + if ((tmp_version <= 0) || (tmp && strlen (tmp) > 1)) { + gf_log (this->name, GF_LOG_WARNING, "invalid version number"); goto out; } + *op_version = tmp_version; + ret = 0; out: - if (handle->fd > 0) { - close (handle->fd); - handle->fd = 0; - } - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + if (op_version_str) + GF_FREE (op_version_str); + return ret; } -int32_t -glusterd_retrieve_uuid () +int +glusterd_retrieve_sys_snap_max_limit (xlator_t *this, uint64_t *limit, + char *key) { - char *uuid_str = NULL; - int32_t ret = -1; - glusterd_store_handle_t *handle = NULL; - glusterd_conf_t *priv = NULL; - char path[PATH_MAX] = {0,}; + char *limit_str = NULL; + glusterd_conf_t *priv = NULL; + int ret = -1; + uint64_t tmp_limit = 0; + char *tmp = NULL; + char path[PATH_MAX] = {0,}; + gf_store_handle_t *handle = NULL; - priv = THIS->private; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (limit); + GF_ASSERT (key); if (!priv->handle) { snprintf (path, PATH_MAX, "%s/%s", priv->workdir, GLUSTERD_INFO_FILE); - ret = glusterd_store_handle_retrieve (path, &handle); + ret = gf_store_handle_retrieve (path, &handle); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get store " + gf_log ("", GF_LOG_DEBUG, "Unable to get store " "handle!"); goto out; } @@ -1310,262 +1742,152 @@ glusterd_retrieve_uuid () priv->handle = handle; } - ret = glusterd_store_retrieve_value (priv->handle, - GLUSTERD_STORE_UUID_KEY, - &uuid_str); - + ret = gf_store_retrieve_value (priv->handle, + key, + &limit_str); if (ret) { - gf_log ("", GF_LOG_INFO, "No previous uuid is present"); - goto out; - } - - uuid_parse (uuid_str, priv->uuid); - -out: - if (uuid_str) - GF_FREE (uuid_str); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - return ret; -} - -int32_t -glusterd_store_iter_new (glusterd_store_handle_t *shandle, - glusterd_store_iter_t **iter) -{ - int32_t ret = -1; - glusterd_store_iter_t *tmp_iter = NULL; - int fd = -1; - - GF_ASSERT (shandle); - GF_ASSERT (iter); - - tmp_iter = GF_CALLOC (1, sizeof (*tmp_iter), - gf_gld_mt_store_iter_t); - - if (!tmp_iter) { - gf_log ("", GF_LOG_ERROR, "Out of Memory"); + gf_log (this->name, GF_LOG_DEBUG, + "No previous %s present", key); goto out; } - fd = open (shandle->path, O_RDWR); - - if (fd < 0) { - gf_log ("", GF_LOG_ERROR, "Unable to open %s, errno: %d", - shandle->path, errno); + tmp_limit = strtoul (limit_str, &tmp, 10); + if ((tmp_limit <= 0) || (tmp && strlen (tmp) > 1)) { + gf_log (this->name, GF_LOG_WARNING, "invalid version number"); goto out; } - tmp_iter->fd = fd; + *limit = tmp_limit; - tmp_iter->file = fdopen (tmp_iter->fd, "r"); - - if (!tmp_iter->file) { - gf_log ("", GF_LOG_ERROR, "Unable to open file %s errno: %d", - shandle->path, errno); - goto out; - } - - strncpy (tmp_iter->filepath, shandle->path, sizeof (tmp_iter->filepath)); - *iter = tmp_iter; ret = 0; - out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); - return ret; -} - -int32_t -glusterd_store_validate_key_value (char *storepath, char *key, char*val, - glusterd_store_op_errno_t *op_errno) -{ - int ret = 0; - - GF_ASSERT (op_errno); - GF_ASSERT (storepath); - - if ((key == NULL) && (val == NULL)) { - ret = -1; - gf_log ("glusterd", GF_LOG_ERROR, "Glusterd store may be " - "corrupted, Invalid key and value (null) in %s", - storepath); - *op_errno = GD_STORE_KEY_VALUE_NULL; - } else if (key == NULL) { - ret = -1; - gf_log ("glusterd", GF_LOG_ERROR, "Glusterd store may be " - "corrupted, Invalid key (null) in %s", storepath); - *op_errno = GD_STORE_KEY_NULL; - } else if (val == NULL) { - ret = -1; - gf_log ("glusterd", GF_LOG_ERROR, "Glusterd store may be " - "corrupted, Invalid value (null) for key %s in %s", - key, storepath); - *op_errno = GD_STORE_VALUE_NULL; - } else { - ret = 0; - *op_errno = GD_STORE_SUCCESS; - } + if (limit_str) + GF_FREE (limit_str); return ret; } - -int32_t -glusterd_store_iter_get_next (glusterd_store_iter_t *iter, - char **key, char **value, - glusterd_store_op_errno_t *op_errno) +static int +glusterd_restore_op_version (xlator_t *this) { - int32_t ret = -1; - char *scan_str = NULL; - char *free_str = NULL; - char *iter_key = NULL; - char *iter_val = NULL; - struct stat st = {0,}; - glusterd_store_op_errno_t store_errno = GD_STORE_SUCCESS; - - GF_ASSERT (iter); - GF_ASSERT (iter->file); - GF_ASSERT (key); - GF_ASSERT (value); - - ret = fstat (iter->fd, &st); - if (ret < 0) { - gf_log ("glusterd", GF_LOG_WARNING, - "stat on file failed"); - ret = -1; - store_errno = GD_STORE_STAT_FAILED; - goto out; - } - - scan_str = GF_CALLOC (1, st.st_size, - gf_gld_mt_char); - if (scan_str == NULL) { - ret = -1; - store_errno = GD_STORE_ENOMEM; - goto out; - } - - *key = NULL; - *value = NULL; + glusterd_conf_t *conf = NULL; + int ret = 0; + int op_version = 0; - free_str = scan_str; + conf = this->private; - ret = glusterd_store_read_and_tokenize (iter->file, scan_str, - &iter_key, &iter_val, - &store_errno); - if (ret < 0) { - goto out; - } - - - ret = glusterd_store_validate_key_value (iter->filepath, iter_key, - iter_val, &store_errno); - if (ret) - goto out; - - *value = gf_strdup (iter_val); - - *key = gf_strdup (iter_key); - if (!iter_key || !iter_val) { - ret = -1; - store_errno = GD_STORE_ENOMEM; - goto out; + ret = glusterd_retrieve_sys_snap_max_limit (this, + &conf->snap_max_hard_limit, + GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Unable to retrieve system snap-max-hard-limit, " + "setting it to default value(%d)", + GLUSTERD_SNAPS_MAX_HARD_LIMIT); + conf->snap_max_hard_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT; } - ret = 0; - -out: + ret = glusterd_retrieve_sys_snap_max_limit (this, + &conf->snap_max_soft_limit, + GLUSTERD_STORE_KEY_SNAP_MAX_SOFT_LIMIT); if (ret) { - if (*key) { - GF_FREE (*key); - *key = NULL; - } - if (*value) { - GF_FREE (*value); - *value = NULL; - } + gf_log (this->name, GF_LOG_WARNING, + "Unable to retrieve system snap-max-soft-limit, " + "setting it to default value(%d)", + GLUSTERD_SNAPS_DEF_SOFT_LIMIT_PERCENT); + conf->snap_max_soft_limit = GLUSTERD_SNAPS_DEF_SOFT_LIMIT_PERCENT; } - if (free_str) - GF_FREE (free_str); - if (op_errno) - *op_errno = store_errno; - - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); - return ret; -} -int32_t -glusterd_store_iter_get_matching (glusterd_store_iter_t *iter, - char *key, char **value) -{ - int32_t ret = -1; - char *tmp_key = NULL; - char *tmp_value = NULL; - - ret = glusterd_store_iter_get_next (iter, &tmp_key, &tmp_value, - NULL); - while (!ret) { - if (!strncmp (key, tmp_key, strlen (key))){ - *value = tmp_value; - GF_FREE (tmp_key); + ret = glusterd_retrieve_op_version (this, &op_version); + if (!ret) { + if ((op_version < GD_OP_VERSION_MIN) || + (op_version > GD_OP_VERSION_MAX)) { + gf_log (this->name, GF_LOG_ERROR, + "wrong op-version (%d) retrieved", op_version); + ret = -1; goto out; } - GF_FREE (tmp_key); - GF_FREE (tmp_value); - ret = glusterd_store_iter_get_next (iter, &tmp_key, - &tmp_value, NULL); + conf->op_version = op_version; + gf_log ("glusterd", GF_LOG_INFO, + "retrieved op-version: %d", conf->op_version); + goto out; + } + + /* op-version can be missing from the store file in 2 cases, + * 1. This is a new install of glusterfs + * 2. This is an upgrade of glusterfs from a version without op-version + * to a version with op-version (eg. 3.3 -> 3.4) + * + * Detection of a new install or an upgrade from an older install can be + * done by checking for the presence of the its peer-id in the store + * file. If peer-id is present, the installation is an upgrade else, it + * is a new install. + * + * For case 1, set op-version to GD_OP_VERSION_MAX. + * For case 2, set op-version to GD_OP_VERSION_MIN. + */ + ret = glusterd_retrieve_uuid(); + if (ret) { + gf_log (this->name, GF_LOG_INFO, "Detected new install. Setting" + " op-version to maximum : %d", GD_OP_VERSION_MAX); + conf->op_version = GD_OP_VERSION_MAX; + } else { + gf_log (this->name, GF_LOG_INFO, "Upgrade detected. Setting" + " op-version to minimum : %d", GD_OP_VERSION_MIN); + conf->op_version = GD_OP_VERSION_MIN; } + ret = 0; out: return ret; } int32_t -glusterd_store_iter_destroy (glusterd_store_iter_t *iter) +glusterd_retrieve_uuid () { + char *uuid_str = NULL; int32_t ret = -1; + gf_store_handle_t *handle = NULL; + glusterd_conf_t *priv = NULL; + char path[PATH_MAX] = {0,}; - GF_ASSERT (iter); - GF_ASSERT (iter->fd > 0); + priv = THIS->private; - ret = fclose (iter->file); + if (!priv->handle) { + snprintf (path, PATH_MAX, "%s/%s", priv->workdir, + GLUSTERD_INFO_FILE); + ret = gf_store_handle_retrieve (path, &handle); + + if (ret) { + gf_log ("", GF_LOG_DEBUG, "Unable to get store" + "handle!"); + goto out; + } + + priv->handle = handle; + } + + ret = gf_store_retrieve_value (priv->handle, GLUSTERD_STORE_UUID_KEY, + &uuid_str); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to close fd: %d, ret: %d, " - "errno: %d" ,iter->fd, ret, errno); + gf_log ("", GF_LOG_DEBUG, "No previous uuid is present"); + goto out; } - GF_FREE (iter); + uuid_parse (uuid_str, priv->uuid); +out: + GF_FREE (uuid_str); + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } -char* -glusterd_store_strerror (glusterd_store_op_errno_t op_errno) -{ - switch (op_errno) { - case GD_STORE_SUCCESS: - return "Success"; - case GD_STORE_KEY_NULL: - return "Invalid Key"; - case GD_STORE_VALUE_NULL: - return "Invalid Value"; - case GD_STORE_KEY_VALUE_NULL: - return "Invalid Key and Value"; - case GD_STORE_EOF: - return "No data"; - case GD_STORE_ENOMEM: - return "No memory"; - default: - return "Invalid errno"; - } - return "Invalid errno"; -} int32_t glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) { - int32_t ret = 0; glusterd_brickinfo_t *brickinfo = NULL; - glusterd_store_iter_t *iter = NULL; + gf_store_iter_t *iter = NULL; char *key = NULL; char *value = NULL; char brickdir[PATH_MAX] = {0,}; @@ -1573,20 +1895,19 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) glusterd_conf_t *priv = NULL; int32_t brick_count = 0; char tmpkey[4096] = {0,}; - glusterd_store_iter_t *tmpiter = NULL; + gf_store_iter_t *tmpiter = NULL; char *tmpvalue = NULL; - struct pmap_registry *pmap = NULL; - glusterd_store_op_errno_t op_errno = GD_STORE_SUCCESS; + struct pmap_registry *pmap = NULL; + gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; GF_ASSERT (volinfo); GF_ASSERT (volinfo->volname); priv = THIS->private; - GLUSTERD_GET_BRICK_DIR (brickdir, volinfo, priv) - - ret = glusterd_store_iter_new (volinfo->shandle, &tmpiter); + GLUSTERD_GET_BRICK_DIR (brickdir, volinfo, priv); + ret = gf_store_iter_new (volinfo->shandle, &tmpiter); if (ret) goto out; @@ -1597,30 +1918,28 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) goto out; snprintf (tmpkey, sizeof (tmpkey), "%s-%d", GLUSTERD_STORE_KEY_VOL_BRICK,brick_count); - ret = glusterd_store_iter_get_matching (tmpiter, tmpkey, - &tmpvalue); + ret = gf_store_iter_get_matching (tmpiter, tmpkey, &tmpvalue); snprintf (path, sizeof (path), "%s/%s", brickdir, tmpvalue); GF_FREE (tmpvalue); tmpvalue = NULL; - ret = glusterd_store_handle_retrieve (path, &brickinfo->shandle); + ret = gf_store_handle_retrieve (path, &brickinfo->shandle); if (ret) goto out; - ret = glusterd_store_iter_new (brickinfo->shandle, &iter); + ret = gf_store_iter_new (brickinfo->shandle, &iter); if (ret) goto out; - ret = glusterd_store_iter_get_next (iter, &key, &value, - &op_errno); + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); if (ret) { gf_log ("glusterd", GF_LOG_ERROR, "Unable to iterate " "the store for brick: %s, reason: %s", path, - glusterd_store_strerror (op_errno)); + gf_store_strerror (op_errno)); goto out; } while (!ret) { @@ -1634,22 +1953,52 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_PORT, strlen (GLUSTERD_STORE_KEY_BRICK_PORT))) { gf_string2int (value, &brickinfo->port); - /* This is required to have proper ports - assigned to bricks after restart */ - pmap = pmap_registry_get (THIS); - if (pmap->last_alloc <= brickinfo->port) - pmap->last_alloc = brickinfo->port + 1; + + if (brickinfo->port < priv->base_port) { + /* This is required to adhere to the + IANA standards */ + brickinfo->port = 0; + } else { + /* This is required to have proper ports + assigned to bricks after restart */ + pmap = pmap_registry_get (THIS); + if (pmap->last_alloc <= brickinfo->port) + pmap->last_alloc = + brickinfo->port + 1; + } } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_RDMA_PORT, strlen (GLUSTERD_STORE_KEY_BRICK_RDMA_PORT))) { gf_string2int (value, &brickinfo->rdma_port); - /* This is required to have proper ports - assigned to bricks after restart */ - pmap = pmap_registry_get (THIS); - if (pmap->last_alloc <= brickinfo->rdma_port) - pmap->last_alloc = brickinfo->rdma_port + 1; + + if (brickinfo->rdma_port < priv->base_port) { + /* This is required to adhere to the + IANA standards */ + brickinfo->rdma_port = 0; + } else { + /* This is required to have proper ports + assigned to bricks after restart */ + pmap = pmap_registry_get (THIS); + if (pmap->last_alloc <= + brickinfo->rdma_port) + pmap->last_alloc = + brickinfo->rdma_port +1; + } + } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED, strlen (GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED))) { gf_string2int (value, &brickinfo->decommissioned); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH, + strlen (GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH))) { + strncpy (brickinfo->device_path, value, + sizeof (brickinfo->device_path)); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS, + strlen (GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS))) { + gf_string2int (value, &brickinfo->snap_status); + } else if (!strncmp (key, + GLUSTERD_STORE_KEY_BRICK_VGNAME, + strlen (GLUSTERD_STORE_KEY_BRICK_VGNAME))) { + strncpy (brickinfo->vg, value, + sizeof (brickinfo->vg)); } else { gf_log ("", GF_LOG_ERROR, "Unknown key: %s", key); @@ -1660,13 +2009,13 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) key = NULL; value = NULL; - ret = glusterd_store_iter_get_next (iter, &key, &value, - &op_errno); + ret = gf_store_iter_get_next (iter, &key, &value, + &op_errno); } if (op_errno != GD_STORE_EOF) goto out; - ret = glusterd_store_iter_destroy (iter); + ret = gf_store_iter_destroy (iter); if (ret) goto out; @@ -1675,7 +2024,7 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) brick_count++; } - ret = glusterd_store_iter_destroy (tmpiter); + ret = gf_store_iter_destroy (tmpiter); if (ret) goto out; out: @@ -1686,64 +2035,79 @@ out: int32_t -glusterd_store_retrieve_rbstate (char *volname) +glusterd_store_retrieve_rbstate (glusterd_volinfo_t *volinfo) { int32_t ret = -1; - glusterd_volinfo_t *volinfo = NULL; - glusterd_store_iter_t *iter = NULL; + gf_store_iter_t *iter = NULL; char *key = NULL; char *value = NULL; char volpath[PATH_MAX] = {0,}; glusterd_conf_t *priv = NULL; char path[PATH_MAX] = {0,}; - glusterd_store_op_errno_t op_errno = GD_STORE_SUCCESS; + gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; + xlator_t *this = NULL; - priv = THIS->private; - - ret = glusterd_volinfo_find (volname, &volinfo); - if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Couldn't get" - "volinfo for %s.", volname); - goto out; - } + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (volinfo); GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, priv); snprintf (path, sizeof (path), "%s/%s", volpath, GLUSTERD_VOLUME_RBSTATE_FILE); - ret = glusterd_store_handle_retrieve (path, &volinfo->rb_shandle); + ret = gf_store_handle_retrieve (path, &volinfo->rb_shandle); if (ret) goto out; - ret = glusterd_store_iter_new (volinfo->rb_shandle, &iter); + ret = gf_store_iter_new (volinfo->rb_shandle, &iter); if (ret) goto out; - ret = glusterd_store_iter_get_next (iter, &key, &value, &op_errno); + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); if (ret) goto out; while (!ret) { if (!strncmp (key, GLUSTERD_STORE_KEY_RB_STATUS, strlen (GLUSTERD_STORE_KEY_RB_STATUS))) { - volinfo->rb_status = atoi (value); + volinfo->rep_brick.rb_status = atoi (value); } - if (volinfo->rb_status > GF_RB_STATUS_NONE) { + if (volinfo->rep_brick.rb_status > GF_RB_STATUS_NONE) { if (!strncmp (key, GLUSTERD_STORE_KEY_RB_SRC_BRICK, strlen (GLUSTERD_STORE_KEY_RB_SRC_BRICK))) { - ret = glusterd_brickinfo_from_brick (value, - &volinfo->src_brick); + ret = glusterd_brickinfo_new_from_brick (value, + &volinfo->rep_brick.src_brick); if (ret) goto out; } else if (!strncmp (key, GLUSTERD_STORE_KEY_RB_DST_BRICK, strlen (GLUSTERD_STORE_KEY_RB_DST_BRICK))) { - ret = glusterd_brickinfo_from_brick (value, - &volinfo->dst_brick); + ret = glusterd_brickinfo_new_from_brick (value, + &volinfo->rep_brick.dst_brick); if (ret) goto out; + } else if (!strncmp (key, GLUSTERD_STORE_KEY_RB_DST_PORT, + strlen (GLUSTERD_STORE_KEY_RB_DST_PORT))) { + switch (volinfo->transport_type) { + case GF_TRANSPORT_RDMA: + volinfo->rep_brick.dst_brick->rdma_port = + atoi (value); + break; + + case GF_TRANSPORT_TCP: + case GF_TRANSPORT_BOTH_TCP_RDMA: + volinfo->rep_brick.dst_brick->port = + atoi (value); + break; + } + } else if (!strncmp (key, GF_REPLACE_BRICK_TID_KEY, + strlen (GF_REPLACE_BRICK_TID_KEY))) { + uuid_parse (value, + volinfo->rep_brick.rb_id); } } @@ -1752,66 +2116,144 @@ glusterd_store_retrieve_rbstate (char *volname) key = NULL; value = NULL; - ret = glusterd_store_iter_get_next (iter, &key, &value, - &op_errno); + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); } if (op_errno != GD_STORE_EOF) goto out; - ret = glusterd_store_iter_destroy (iter); + ret = gf_store_iter_destroy (iter); if (ret) goto out; out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); return ret; } int32_t -glusterd_store_retrieve_volume (char *volname) +glusterd_store_retrieve_node_state (glusterd_volinfo_t *volinfo) { int32_t ret = -1; - glusterd_volinfo_t *volinfo = NULL; - glusterd_store_iter_t *iter = NULL; + gf_store_iter_t *iter = NULL; char *key = NULL; char *value = NULL; char volpath[PATH_MAX] = {0,}; glusterd_conf_t *priv = NULL; char path[PATH_MAX] = {0,}; - int exists = 0; - glusterd_store_op_errno_t op_errno = GD_STORE_SUCCESS; + gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; + xlator_t *this = NULL; - ret = glusterd_volinfo_new (&volinfo); + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (volinfo); + GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, priv); + snprintf (path, sizeof (path), "%s/%s", volpath, + GLUSTERD_NODE_STATE_FILE); + + ret = gf_store_handle_retrieve (path, &volinfo->node_state_shandle); if (ret) goto out; - strncpy (volinfo->volname, volname, GLUSTERD_MAX_VOLUME_NAME); + ret = gf_store_iter_new (volinfo->node_state_shandle, &iter); - priv = THIS->private; + if (ret) + goto out; - GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, priv); - snprintf (path, sizeof (path), "%s/%s", volpath, - GLUSTERD_VOLUME_INFO_FILE); + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); + if (ret) + goto out; - ret = glusterd_store_handle_retrieve (path, &volinfo->shandle); + while (ret == 0) { + if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_DEFRAG, + strlen (GLUSTERD_STORE_KEY_VOL_DEFRAG))) { + volinfo->rebal.defrag_cmd = atoi (value); + } - if (ret) + if (volinfo->rebal.defrag_cmd) { + if (!strncmp (key, GF_REBALANCE_TID_KEY, + strlen (GF_REBALANCE_TID_KEY))) + uuid_parse (value, volinfo->rebal.rebalance_id); + + if (!strncmp (key, GLUSTERD_STORE_KEY_DEFRAG_OP, + strlen (GLUSTERD_STORE_KEY_DEFRAG_OP))) + volinfo->rebal.op = atoi (value); + } + + GF_FREE (key); + GF_FREE (value); + key = NULL; + value = NULL; + + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); + } + + if (op_errno != GD_STORE_EOF) goto out; - ret = glusterd_store_iter_new (volinfo->shandle, &iter); + ret = gf_store_iter_destroy (iter); if (ret) goto out; - ret = glusterd_store_iter_get_next (iter, &key, &value, &op_errno); - if (ret) +out: + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); + + return ret; +} + + +int +glusterd_store_update_volinfo (glusterd_volinfo_t *volinfo) +{ + int ret = -1; + int exists = 0; + char *key = NULL; + char *value = NULL; + char volpath[PATH_MAX] = {0,}; + char path[PATH_MAX] = {0,}; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + gf_store_iter_t *iter = NULL; + gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; + + this = THIS; + GF_ASSERT (this); + conf = THIS->private; + GF_ASSERT (volinfo); + + GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, conf); + + snprintf (path, sizeof (path), "%s/%s", volpath, + GLUSTERD_VOLUME_INFO_FILE); + + ret = gf_store_handle_retrieve (path, &volinfo->shandle); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "volinfo handle is NULL"); + goto out; + } + + ret = gf_store_iter_new (volinfo->shandle, &iter); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get new store " + "iter"); + goto out; + } + + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get next store " + "iter"); goto out; + } while (!ret) { + gf_log ("", GF_LOG_DEBUG, "key = %s value = %s", key, value); if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_TYPE, strlen (GLUSTERD_STORE_KEY_VOL_TYPE))) { volinfo->type = atoi (value); @@ -1850,6 +2292,16 @@ glusterd_store_retrieve_volume (char *volname) gf_log ("", GF_LOG_WARNING, "failed to parse uuid"); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_USERNAME, + strlen (GLUSTERD_STORE_KEY_USERNAME))) { + + glusterd_auth_set_username (volinfo, value); + + } else if (!strncmp (key, GLUSTERD_STORE_KEY_PASSWORD, + strlen (GLUSTERD_STORE_KEY_PASSWORD))) { + + glusterd_auth_set_password (volinfo, value); + } else if (strstr (key, "slave")) { ret = dict_set_dynstr (volinfo->gsync_slaves, key, gf_strdup (value)); @@ -1860,28 +2312,57 @@ glusterd_store_retrieve_volume (char *volname) } gf_log ("", GF_LOG_DEBUG, "Parsed as "GEOREP" " " slave:key=%s,value:%s", key, value); - } - else { - exists = glusterd_check_option_exists (key, NULL); - if (exists == -1) { + + } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_OP_VERSION, + strlen (GLUSTERD_STORE_KEY_VOL_OP_VERSION))) { + volinfo->op_version = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION, + strlen (GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION))) { + volinfo->client_op_version = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_CAPS, + strlen (GLUSTERD_STORE_KEY_VOL_CAPS))) { + volinfo->caps = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT, + strlen (GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT))) { + volinfo->snap_max_hard_limit = (uint64_t) atoll (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_IS_RESTORED, + strlen (GLUSTERD_STORE_KEY_VOL_IS_RESTORED))) { + volinfo->is_volume_restored = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_PARENT_VOLNAME, + strlen (GLUSTERD_STORE_KEY_PARENT_VOLNAME))) { + strncpy (volinfo->parent_volname, value, sizeof(volinfo->parent_volname) - 1); + } else { + + if (is_key_glusterd_hooks_friendly (key)) { + exists = 1; + + } else { + exists = glusterd_check_option_exists (key, + NULL); + } + + switch (exists) { + case -1: ret = -1; goto out; - } - if (exists) { + + case 0: + gf_log ("", GF_LOG_ERROR, "Unknown key: %s", + key); + break; + + case 1: ret = dict_set_str(volinfo->dict, key, - gf_strdup (value)); + gf_strdup (value)); if (ret) { gf_log ("",GF_LOG_ERROR, "Error in " - "dict_set_str"); + "dict_set_str"); goto out; } gf_log ("", GF_LOG_DEBUG, "Parsed as Volume-" - "set:key=%s,value:%s", - key, value); + "set:key=%s,value:%s", key, value); + break; } - else - gf_log ("", GF_LOG_ERROR, "Unknown key: %s", - key); } GF_FREE (key); @@ -1889,8 +2370,7 @@ glusterd_store_retrieve_volume (char *volname) key = NULL; value = NULL; - ret = glusterd_store_iter_get_next (iter, &key, &value, - &op_errno); + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); } /* backward compatibility */ @@ -1924,18 +2404,64 @@ glusterd_store_retrieve_volume (char *volname) break; } - volinfo->dist_leaf_count = (volinfo->stripe_count * - volinfo->replica_count); + volinfo->dist_leaf_count = glusterd_get_dist_leaf_count (volinfo); + + volinfo->subvol_count = (volinfo->brick_count / + volinfo->dist_leaf_count); + + /* Only calculate volume op-versions if they are not found */ + if (!volinfo->op_version && !volinfo->client_op_version) + gd_update_volume_op_versions (volinfo); } if (op_errno != GD_STORE_EOF) goto out; - ret = glusterd_store_iter_destroy (iter); + ret = gf_store_iter_destroy (iter); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to destroy store " + "iter"); + goto out; + } + ret = 0; +out: + return ret; +} + +glusterd_volinfo_t* +glusterd_store_retrieve_volume (char *volname, glusterd_snap_t *snap) +{ + int32_t ret = -1; + glusterd_volinfo_t *volinfo = NULL; + glusterd_volinfo_t *origin_volinfo = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (volname); + + ret = glusterd_volinfo_new (&volinfo); if (ret) goto out; + priv = THIS->private; + + strncpy (volinfo->volname, volname, GLUSTERD_MAX_VOLUME_NAME); + volinfo->snapshot = snap; + if (snap) + volinfo->is_snap_volume = _gf_true; + + ret = glusterd_store_update_volinfo (volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to update volinfo " + "for %s volume", volname); + goto out; + } + ret = glusterd_store_retrieve_bricks (volinfo); if (ret) goto out; @@ -1944,19 +2470,130 @@ glusterd_store_retrieve_volume (char *volname) if (ret) goto out; - list_add_tail (&volinfo->vol_list, &priv->volumes); - + if (!snap) { + list_add_tail (&volinfo->vol_list, &priv->volumes); + } else { + ret = glusterd_volinfo_find (volinfo->parent_volname, + &origin_volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Parent volinfo " + "not found for %s volume", volname); + goto out; + } + glusterd_list_add_snapvol (origin_volinfo, volinfo); + } out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + if (ret) { + if (volinfo) + glusterd_volinfo_delete (volinfo); + volinfo = NULL; + } + + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); + + return volinfo; +} + +inline void +glusterd_store_set_options_path (glusterd_conf_t *conf, char *path, size_t len) +{ + snprintf (path, len, "%s/options", conf->workdir); +} + +int +_store_global_opts (dict_t *this, char *key, data_t *value, void *data) +{ + gf_store_handle_t *shandle = data; + + gf_store_save_value (shandle->fd, key, (char*)value->data); + return 0; +} + +int32_t +glusterd_store_options (xlator_t *this, dict_t *opts) +{ + gf_store_handle_t *shandle = NULL; + glusterd_conf_t *conf = NULL; + char path[PATH_MAX] = {0}; + int fd = -1; + int32_t ret = -1; + + conf = this->private; + glusterd_store_set_options_path (conf, path, sizeof (path)); + ret = gf_store_handle_new (path, &shandle); + if (ret) + goto out; + + fd = gf_store_mkstemp (shandle); + if (fd <= 0) { + ret = -1; + goto out; + } + + shandle->fd = fd; + dict_foreach (opts, _store_global_opts, shandle); + shandle->fd = 0; + ret = gf_store_rename_tmppath (shandle); + if (ret) + goto out; +out: + gf_store_handle_destroy (shandle); + if (fd >=0 ) + close (fd); return ret; } +int32_t +glusterd_store_retrieve_options (xlator_t *this) +{ + char path[PATH_MAX] = {0}; + glusterd_conf_t *conf = NULL; + gf_store_handle_t *shandle = NULL; + gf_store_iter_t *iter = NULL; + char *key = NULL; + char *value = NULL; + gf_store_op_errno_t op_errno = 0; + int ret = -1; + + conf = this->private; + glusterd_store_set_options_path (conf, path, sizeof (path)); + + ret = gf_store_handle_retrieve (path, &shandle); + if (ret) + goto out; + + ret = gf_store_iter_new (shandle, &iter); + if (ret) + goto out; + + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); + while (!ret) { + ret = dict_set_dynstr (conf->opts, key, value); + if (ret) { + GF_FREE (key); + GF_FREE (value); + goto out; + } + GF_FREE (key); + key = NULL; + value = NULL; + + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); + } + if (op_errno != GD_STORE_EOF) + goto out; + ret = 0; +out: + gf_store_iter_destroy (iter); + gf_store_handle_destroy (shandle); + return ret; +} int32_t -glusterd_store_retrieve_volumes (xlator_t *this) +glusterd_store_retrieve_volumes (xlator_t *this, glusterd_snap_t *snap) { - int32_t ret = 0; + int32_t ret = -1; char path[PATH_MAX] = {0,}; glusterd_conf_t *priv = NULL; DIR *dir = NULL; @@ -1968,39 +2605,58 @@ glusterd_store_retrieve_volumes (xlator_t *this) GF_ASSERT (priv); - snprintf (path, PATH_MAX, "%s/%s", priv->workdir, - GLUSTERD_VOLUME_DIR_PREFIX); + if (snap) + snprintf (path, PATH_MAX, "%s/snaps/%s", priv->workdir, + snap->snapname); + else + snprintf (path, PATH_MAX, "%s/%s", priv->workdir, + GLUSTERD_VOLUME_DIR_PREFIX); dir = opendir (path); if (!dir) { gf_log ("", GF_LOG_ERROR, "Unable to open dir %s", path); - ret = -1; goto out; } glusterd_for_each_entry (entry, dir); while (entry) { - ret = glusterd_store_retrieve_volume (entry->d_name); - if (ret) { + if ( entry->d_type != DT_DIR ) + goto next; + + volinfo = glusterd_store_retrieve_volume (entry->d_name, snap); + if (!volinfo) { gf_log ("", GF_LOG_ERROR, "Unable to restore " "volume: %s", entry->d_name); + ret = -1; goto out; } - ret = glusterd_store_retrieve_rbstate (entry->d_name); + ret = glusterd_store_retrieve_rbstate (volinfo); if (ret) { /* Backward compatibility */ gf_log ("", GF_LOG_INFO, "Creating a new rbstate " "for volume: %s.", entry->d_name); - ret = glusterd_volinfo_find (entry->d_name, &volinfo); ret = glusterd_store_create_rbstate_shandle_on_absence (volinfo); ret = glusterd_store_perform_rbstate_store (volinfo); } + + ret = glusterd_store_retrieve_node_state (volinfo); + if (ret) { + /* Backward compatibility */ + gf_log ("", GF_LOG_INFO, "Creating a new node_state " + "for volume: %s.", entry->d_name); + glusterd_store_create_nodestate_sh_on_absence (volinfo); + ret = glusterd_store_perform_node_state_store (volinfo); + + } + +next: glusterd_for_each_entry (entry, dir); } + ret = 0; out: if (dir) closedir (dir); @@ -2010,6 +2666,482 @@ out: } int32_t +glusterd_resolve_snap_bricks (xlator_t *this, glusterd_snap_t *snap) +{ + int32_t ret = -1; + glusterd_volinfo_t *volinfo = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + + GF_ASSERT (this); + GF_VALIDATE_OR_GOTO (this->name, snap, out); + + list_for_each_entry (volinfo, &snap->volumes, vol_list) { + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + ret = glusterd_resolve_brick (brickinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "resolve brick failed in restore"); + goto out; + } + } + } + + ret = 0; + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); + + return ret; +} + +int +glusterd_store_update_snap (glusterd_snap_t *snap) +{ + int ret = -1; + char *key = NULL; + char *value = NULL; + char snappath[PATH_MAX] = {0,}; + char path[PATH_MAX] = {0,}; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + gf_store_iter_t *iter = NULL; + gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; + + this = THIS; + conf = this->private; + GF_ASSERT (snap); + + GLUSTERD_GET_SNAP_DIR (snappath, snap, conf); + + snprintf (path, sizeof (path), "%s/%s", snappath, + GLUSTERD_SNAP_INFO_FILE); + + ret = gf_store_handle_retrieve (path, &snap->shandle); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "snap handle is NULL"); + goto out; + } + + ret = gf_store_iter_new (snap->shandle, &iter); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get new store " + "iter"); + goto out; + } + + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get next store " + "iter"); + goto out; + } + + while (!ret) { + gf_log (this->name, GF_LOG_DEBUG, "key = %s value = %s", + key, value); + + if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_ID, + strlen (GLUSTERD_STORE_KEY_SNAP_ID))) { + ret = uuid_parse (value, snap->snap_id); + if (ret) + gf_log (this->name, GF_LOG_WARNING, + "Failed to parse uuid"); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_RESTORED, + strlen (GLUSTERD_STORE_KEY_SNAP_RESTORED))) { + snap->snap_restored = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_STATUS, + strlen (GLUSTERD_STORE_KEY_SNAP_STATUS))) { + snap->snap_status = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_DESC, + strlen (GLUSTERD_STORE_KEY_SNAP_DESC))) { + snap->description = gf_strdup (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_TIMESTAMP, + strlen (GLUSTERD_STORE_KEY_SNAP_TIMESTAMP))) { + snap->time_stamp = atoi (value); + } + + GF_FREE (key); + GF_FREE (value); + key = NULL; + value = NULL; + + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); + } + + if (op_errno != GD_STORE_EOF) + goto out; + + ret = gf_store_iter_destroy (iter); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to destroy store " + "iter"); + } + +out: + return ret; +} + +int32_t +glusterd_store_retrieve_snap (char *snapname) +{ + int32_t ret = -1; + dict_t *dict = NULL; + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (snapname); + + dict = dict_new(); + if (!dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create dict"); + ret = -1; + goto out; + } + + snap = glusterd_new_snap_object (); + if (!snap) { + gf_log (this->name, GF_LOG_ERROR, "Failed to create " + " snap object"); + goto out; + } + + strncpy (snap->snapname, snapname, strlen(snapname)); + ret = glusterd_store_update_snap (snap); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to update snapshot " + "for %s snap", snapname); + goto out; + } + + ret = glusterd_store_retrieve_volumes (this, snap); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to retrieve " + "snap volumes for snap %s", snapname); + goto out; + } + + /* Unlike bricks of normal volumes which are resolved at the end of + the glusterd restore, the bricks belonging to the snap volumes of + each snap should be resolved as part of snapshot restore itself. + Because if the snapshot has to be removed, then resolving bricks + helps glusterd in understanding what all bricks have its own uuid + and killing those bricks. + */ + ret = glusterd_resolve_snap_bricks (this, snap); + if (ret) + gf_log (this->name, GF_LOG_WARNING, "resolving the snap bricks" + " failed (snap: %s)", snap?snap->snapname:""); + + /* When the snapshot command from cli is received, the on disk and + in memory structures for the snapshot are created (with the status) + being marked as GD_SNAP_STATUS_INIT. Once the backend snapshot is + taken, the status is changed to GD_SNAP_STATUS_IN_USE. If glusterd + dies after taking the backend snapshot, but before updating the + status, then when glusterd comes up, it should treat that snapshot + as a failed snapshot and clean it up. + */ + if (snap->snap_status != GD_SNAP_STATUS_IN_USE) { + ret = glusterd_snap_remove (dict, snap, _gf_true, _gf_true); + if (ret) + gf_log (this->name, GF_LOG_WARNING, "failed to remove" + " the snapshot %s", snap->snapname); + goto out; + } + + /* TODO: list_add_order can do 'N-square' comparisions and + is not efficient. Find a better solution to store the snap + in order */ + list_add_order (&snap->snap_list, &priv->snapshots, + glusterd_compare_snap_time); + +out: + if (dict) + dict_unref (dict); + + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); + return ret; +} + +/* Read the missed_snap_list and update the in-memory structs */ +int32_t +glusterd_store_retrieve_missed_snaps_list (xlator_t *this) +{ + char buf[PATH_MAX] = ""; + char path[PATH_MAX] = ""; + char *missed_node_info = NULL; + char *brick_path = NULL; + char *value = NULL; + char *save_ptr = NULL; + FILE *fp = NULL; + int32_t brick_num = -1; + int32_t snap_op = -1; + int32_t snap_status = -1; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; + gf_store_op_errno_t store_errno = GD_STORE_SUCCESS; + + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + /* Get the path of the missed_snap_list */ + glusterd_store_missed_snaps_list_path_set (path, sizeof(path)); + + fp = fopen (path, "r"); + if (!fp) { + /* If errno is ENOENT then there are no missed snaps yet */ + if (errno != ENOENT) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "Failed to open %s. " + "Error: %s", path, strerror(errno)); + } else { + gf_log (this->name, GF_LOG_INFO, + "No missed snaps list."); + ret = 0; + } + goto out; + } + + do { + ret = gf_store_read_and_tokenize (fp, buf, + &missed_node_info, &value, + &store_errno); + if (ret) { + if (store_errno == GD_STORE_EOF) { + gf_log (this->name, + GF_LOG_DEBUG, + "EOF for missed_snap_list"); + ret = 0; + break; + } + gf_log (this->name, GF_LOG_ERROR, + "Failed to fetch data from " + "missed_snaps_list. Error: %s", + gf_store_strerror (store_errno)); + goto out; + } + + /* Fetch the brick_num, brick_path, snap_op and snap status */ + brick_num = atoi(strtok_r (value, ":", &save_ptr)); + brick_path = strtok_r (NULL, ":", &save_ptr); + snap_op = atoi(strtok_r (NULL, ":", &save_ptr)); + snap_status = atoi(strtok_r (NULL, ":", &save_ptr)); + + if (!missed_node_info || !brick_path || + brick_num < 1 || snap_op < 1 || + snap_status < 1) { + gf_log (this->name, GF_LOG_ERROR, + "Invalid missed_snap_entry"); + ret = -1; + goto out; + } + + ret = glusterd_store_missed_snaps_list (missed_node_info, + brick_num, + brick_path, + snap_op, + snap_status); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to store missed snaps_list"); + goto out; + } + + } while (store_errno == GD_STORE_SUCCESS); + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); + return ret; +} + +int32_t +glusterd_store_retrieve_snaps (xlator_t *this) +{ + int32_t ret = 0; + char path[PATH_MAX] = {0,}; + glusterd_conf_t *priv = NULL; + DIR *dir = NULL; + struct dirent *entry = NULL; + + GF_ASSERT (this); + priv = this->private; + + GF_ASSERT (priv); + + snprintf (path, PATH_MAX, "%s/snaps", priv->workdir); + + dir = opendir (path); + + if (!dir) { + /* If snaps dir doesn't exists ignore the error for + backward compatibility */ + if (errno != ENOENT) { + ret = -1; + gf_log ("", GF_LOG_ERROR, "Unable to open dir %s", path); + } + goto out; + } + + glusterd_for_each_entry (entry, dir); + + while (entry) { + if (entry->d_type == DT_DIR) { + ret = glusterd_store_retrieve_snap (entry->d_name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to restore snapshot: %s", + entry->d_name); + goto out; + } + } + + glusterd_for_each_entry (entry, dir); + } + + /* Retrieve missed_snaps_list */ + ret = glusterd_store_retrieve_missed_snaps_list (this); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Failed to retrieve missed_snaps_list"); + goto out; + } + +out: + if (dir) + closedir (dir); + gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + + return ret; +} + +/* Writes all the contents of conf->missed_snap_list */ +int32_t +glusterd_store_write_missed_snapinfo (int32_t fd) +{ + char value[PATH_MAX] = ""; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; + glusterd_missed_snap_info *missed_snapinfo = NULL; + glusterd_snap_op_t *snap_opinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + + priv = this->private; + GF_ASSERT (priv); + + /* Write the missed_snap_entry */ + list_for_each_entry (missed_snapinfo, &priv->missed_snaps_list, + missed_snaps) { + list_for_each_entry (snap_opinfo, + &missed_snapinfo->snap_ops, + snap_ops_list) { + snprintf (value, sizeof(value), "%d:%s:%d:%d", + snap_opinfo->brick_num, + snap_opinfo->brick_path, + snap_opinfo->op, snap_opinfo->status); + ret = gf_store_save_value + (fd, + missed_snapinfo->node_snap_info, + value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to write missed snapinfo"); + goto out; + } + } + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Adds the missed snap entries to the in-memory conf->missed_snap_list * + * and writes them to disk */ +int32_t +glusterd_store_update_missed_snaps (dict_t *dict, int32_t missed_snap_count) +{ + int32_t fd = -1; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(dict); + + priv = this->private; + GF_ASSERT (priv); + + if (missed_snap_count < 1) { + gf_log (this->name, GF_LOG_DEBUG, "No missed snaps"); + ret = 0; + goto out; + } + + ret = glusterd_store_create_missed_snaps_list_shandle_on_absence (); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to obtain " + "missed_snaps_list store handle."); + goto out; + } + + fd = gf_store_mkstemp (priv->missed_snaps_list_shandle); + if (fd <= 0) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create tmp file"); + ret = -1; + goto out; + } + + ret = glusterd_add_missed_snaps_to_list (dict, missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add missed snaps to list"); + goto out; + } + + ret = glusterd_store_write_missed_snapinfo (fd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to write missed snaps to disk"); + goto out; + } + + ret = gf_store_rename_tmppath (priv->missed_snaps_list_shandle); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to rename the tmp file"); + goto out; + } +out: + if (ret && (fd > 0)) { + ret = gf_store_unlink_tmppath (priv->missed_snaps_list_shandle); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to unlink the tmp file"); + } + ret = -1; + } + + if (fd > 0) + close (fd); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t glusterd_store_delete_peerinfo (glusterd_peerinfo_t *peerinfo) { int32_t ret = -1; @@ -2057,7 +3189,7 @@ glusterd_store_delete_peerinfo (glusterd_peerinfo_t *peerinfo) out: if (peerinfo->shandle) { - glusterd_store_handle_destroy (peerinfo->shandle); + gf_store_handle_destroy (peerinfo->shandle); peerinfo->shandle = NULL; } gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); @@ -2083,7 +3215,7 @@ glusterd_store_create_peer_dir () char path[PATH_MAX]; glusterd_store_peerinfo_dirpath_set (path, sizeof (path)); - ret = glusterd_store_mkdir (path); + ret = gf_store_mkdir (path); gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); return ret; @@ -2127,8 +3259,8 @@ glusterd_store_peerinfo_hostname_shandle_create (glusterd_peerinfo_t *peerinfo) glusterd_store_hostname_peerpath_set (peerinfo, peerfpath, sizeof (peerfpath)); - ret = glusterd_store_handle_create_on_absence (&peerinfo->shandle, - peerfpath); + ret = gf_store_handle_create_on_absence (&peerinfo->shandle, + peerfpath); return ret; } @@ -2140,8 +3272,8 @@ glusterd_store_peerinfo_uuid_shandle_create (glusterd_peerinfo_t *peerinfo) glusterd_store_uuid_peerpath_set (peerinfo, peerfpath, sizeof (peerfpath)); - ret = glusterd_store_handle_create_on_absence (&peerinfo->shandle, - peerfpath); + ret = gf_store_handle_create_on_absence (&peerinfo->shandle, + peerfpath); return ret; } @@ -2157,7 +3289,7 @@ glusterd_peerinfo_hostname_shandle_check_destroy (glusterd_peerinfo_t *peerinfo) ret = stat (peerfpath, &stbuf); if (!ret) { if (peerinfo->shandle) - glusterd_store_handle_destroy (peerinfo->shandle); + gf_store_handle_destroy (peerinfo->shandle); peerinfo->shandle = NULL; ret = unlink (peerfpath); } @@ -2186,18 +3318,21 @@ glusterd_store_peer_write (int fd, glusterd_peerinfo_t *peerinfo) char buf[50] = {0}; int32_t ret = 0; - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_PEER_UUID, - uuid_utoa (peerinfo->uuid)); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PEER_UUID, + uuid_utoa (peerinfo->uuid)); if (ret) goto out; snprintf (buf, sizeof (buf), "%d", peerinfo->state.state); - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_PEER_STATE, buf); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PEER_STATE, buf); + if (ret) + goto out; + + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PEER_HOSTNAME "1", + peerinfo->hostname); if (ret) goto out; - ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_PEER_HOSTNAME "1", - peerinfo->hostname); out: gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); return ret; @@ -2211,7 +3346,7 @@ glusterd_store_perform_peer_store (glusterd_peerinfo_t *peerinfo) GF_ASSERT (peerinfo); - fd = glusterd_store_mkstemp (peerinfo->shandle); + fd = gf_store_mkstemp (peerinfo->shandle); if (fd <= 0) { ret = -1; goto out; @@ -2221,10 +3356,10 @@ glusterd_store_perform_peer_store (glusterd_peerinfo_t *peerinfo) if (ret) goto out; - ret = glusterd_store_rename_tmppath (peerinfo->shandle); + ret = gf_store_rename_tmppath (peerinfo->shandle); out: if (ret && (fd > 0)) - glusterd_store_unlink_tmppath (peerinfo->shandle); + gf_store_unlink_tmppath (peerinfo->shandle); if (fd > 0) close (fd); gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); @@ -2264,13 +3399,13 @@ glusterd_store_retrieve_peers (xlator_t *this) uuid_t uuid = {0,}; char *hostname = NULL; int32_t state = 0; - glusterd_store_handle_t *shandle = NULL; + gf_store_handle_t *shandle = NULL; char filepath[PATH_MAX] = {0,}; - glusterd_store_iter_t *iter = NULL; + gf_store_iter_t *iter = NULL; char *key = NULL; char *value = NULL; glusterd_peerctx_args_t args = {0}; - glusterd_store_op_errno_t op_errno = GD_STORE_SUCCESS; + gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; GF_ASSERT (this); priv = this->private; @@ -2292,16 +3427,15 @@ glusterd_store_retrieve_peers (xlator_t *this) while (entry) { snprintf (filepath, PATH_MAX, "%s/%s", path, entry->d_name); - ret = glusterd_store_handle_retrieve (filepath, &shandle); + ret = gf_store_handle_retrieve (filepath, &shandle); if (ret) goto out; - ret = glusterd_store_iter_new (shandle, &iter); + ret = gf_store_iter_new (shandle, &iter); if (ret) goto out; - ret = glusterd_store_iter_get_next (iter, &key, &value, - &op_errno); + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); if (ret) goto out; @@ -2329,17 +3463,16 @@ glusterd_store_retrieve_peers (xlator_t *this) key = NULL; value = NULL; - ret = glusterd_store_iter_get_next (iter, &key, &value, - &op_errno); + ret = gf_store_iter_get_next (iter, &key, &value, + &op_errno); } if (op_errno != GD_STORE_EOF) goto out; - (void) glusterd_store_iter_destroy (iter); + (void) gf_store_iter_destroy (iter); - args.mode = GD_MODE_SWITCH_ON; ret = glusterd_friend_add (hostname, 0, state, &uuid, - NULL, &peerinfo, 1, &args); + &peerinfo, 1, NULL); GF_FREE (hostname); if (ret) @@ -2349,6 +3482,13 @@ glusterd_store_retrieve_peers (xlator_t *this) glusterd_for_each_entry (entry, dir); } + args.mode = GD_MODE_ON; + list_for_each_entry (peerinfo, &priv->peers, uuid_list) { + ret = glusterd_friend_rpc_create (this, peerinfo, &args); + if (ret) + goto out; + } + out: if (dir) closedir (dir); @@ -2395,8 +3535,18 @@ glusterd_restore () this = THIS; - ret = glusterd_store_retrieve_volumes (this); + ret = glusterd_restore_op_version (this); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to restore op_version"); + goto out; + } + + ret = glusterd_store_retrieve_volumes (this, NULL); + if (ret) + goto out; + ret = glusterd_store_retrieve_snaps (this); if (ret) goto out; diff --git a/xlators/mgmt/glusterd/src/glusterd-store.h b/xlators/mgmt/glusterd/src/glusterd-store.h index f1413955b..1b5cebc0c 100644 --- a/xlators/mgmt/glusterd/src/glusterd-store.h +++ b/xlators/mgmt/glusterd/src/glusterd-store.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _GLUSTERD_HA_H_ #define _GLUSTERD_HA_H_ @@ -30,6 +20,7 @@ #include "glusterfs.h" #include "xlator.h" +#include "run.h" #include "logging.h" #include "call-stub.h" #include "fd.h" @@ -40,35 +31,59 @@ typedef enum glusterd_store_ver_ac_{ GLUSTERD_VOLINFO_VER_AC_NONE = 0, GLUSTERD_VOLINFO_VER_AC_INCREMENT = 1, + GLUSTERD_VOLINFO_VER_AC_DECREMENT = 2, } glusterd_volinfo_ver_ac_t; -#define GLUSTERD_STORE_UUID_KEY "UUID" - -#define GLUSTERD_STORE_KEY_VOL_TYPE "type" -#define GLUSTERD_STORE_KEY_VOL_COUNT "count" -#define GLUSTERD_STORE_KEY_VOL_STATUS "status" -#define GLUSTERD_STORE_KEY_VOL_PORT "port" -#define GLUSTERD_STORE_KEY_VOL_SUB_COUNT "sub_count" -#define GLUSTERD_STORE_KEY_VOL_STRIPE_CNT "stripe_count" -#define GLUSTERD_STORE_KEY_VOL_REPLICA_CNT "replica_count" -#define GLUSTERD_STORE_KEY_VOL_BRICK "brick" -#define GLUSTERD_STORE_KEY_VOL_VERSION "version" -#define GLUSTERD_STORE_KEY_VOL_TRANSPORT "transport-type" -#define GLUSTERD_STORE_KEY_VOL_ID "volume-id" -#define GLUSTERD_STORE_KEY_RB_STATUS "rb_status" -#define GLUSTERD_STORE_KEY_RB_SRC_BRICK "rb_src" -#define GLUSTERD_STORE_KEY_RB_DST_BRICK "rb_dst" - -#define GLUSTERD_STORE_KEY_BRICK_HOSTNAME "hostname" -#define GLUSTERD_STORE_KEY_BRICK_PATH "path" -#define GLUSTERD_STORE_KEY_BRICK_PORT "listen-port" -#define GLUSTERD_STORE_KEY_BRICK_RDMA_PORT "rdma.listen-port" +#define GLUSTERD_STORE_UUID_KEY "UUID" + +#define GLUSTERD_STORE_KEY_VOL_TYPE "type" +#define GLUSTERD_STORE_KEY_VOL_COUNT "count" +#define GLUSTERD_STORE_KEY_VOL_STATUS "status" +#define GLUSTERD_STORE_KEY_VOL_PORT "port" +#define GLUSTERD_STORE_KEY_VOL_SUB_COUNT "sub_count" +#define GLUSTERD_STORE_KEY_VOL_STRIPE_CNT "stripe_count" +#define GLUSTERD_STORE_KEY_VOL_REPLICA_CNT "replica_count" +#define GLUSTERD_STORE_KEY_VOL_BRICK "brick" +#define GLUSTERD_STORE_KEY_VOL_VERSION "version" +#define GLUSTERD_STORE_KEY_VOL_TRANSPORT "transport-type" +#define GLUSTERD_STORE_KEY_VOL_ID "volume-id" +#define GLUSTERD_STORE_KEY_VOL_IS_RESTORED "is-volume-restored" +#define GLUSTERD_STORE_KEY_RB_STATUS "rb_status" +#define GLUSTERD_STORE_KEY_RB_SRC_BRICK "rb_src" +#define GLUSTERD_STORE_KEY_RB_DST_BRICK "rb_dst" +#define GLUSTERD_STORE_KEY_RB_DST_PORT "rb_port" +#define GLUSTERD_STORE_KEY_VOL_DEFRAG "rebalance_status" +#define GLUSTERD_STORE_KEY_DEFRAG_OP "rebalance_op" +#define GLUSTERD_STORE_KEY_USERNAME "username" +#define GLUSTERD_STORE_KEY_PASSWORD "password" +#define GLUSTERD_STORE_KEY_PARENT_VOLNAME "parent_volname" +#define GLUSTERD_STORE_KEY_VOL_OP_VERSION "op-version" +#define GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION "client-op-version" + +#define GLUSTERD_STORE_KEY_SNAP_NAME "name" +#define GLUSTERD_STORE_KEY_SNAP_ID "snap-id" +#define GLUSTERD_STORE_KEY_SNAP_DESC "desc" +#define GLUSTERD_STORE_KEY_SNAP_TIMESTAMP "time-stamp" +#define GLUSTERD_STORE_KEY_SNAP_STATUS "status" +#define GLUSTERD_STORE_KEY_SNAP_RESTORED "snap-restored" +#define GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT "snap-max-hard-limit" +#define GLUSTERD_STORE_KEY_SNAP_MAX_SOFT_LIMIT "snap-max-soft-limit" + +#define GLUSTERD_STORE_KEY_BRICK_HOSTNAME "hostname" +#define GLUSTERD_STORE_KEY_BRICK_PATH "path" +#define GLUSTERD_STORE_KEY_BRICK_PORT "listen-port" +#define GLUSTERD_STORE_KEY_BRICK_RDMA_PORT "rdma.listen-port" #define GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED "decommissioned" +#define GLUSTERD_STORE_KEY_BRICK_VGNAME "vg" +#define GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH "device_path" +#define GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS "snap-status" -#define GLUSTERD_STORE_KEY_PEER_UUID "uuid" -#define GLUSTERD_STORE_KEY_PEER_HOSTNAME "hostname" -#define GLUSTERD_STORE_KEY_PEER_STATE "state" +#define GLUSTERD_STORE_KEY_PEER_UUID "uuid" +#define GLUSTERD_STORE_KEY_PEER_HOSTNAME "hostname" +#define GLUSTERD_STORE_KEY_PEER_STATE "state" + +#define GLUSTERD_STORE_KEY_VOL_CAPS "caps" #define glusterd_for_each_entry(entry, dir) \ do {\ @@ -76,21 +91,13 @@ typedef enum glusterd_store_ver_ac_{ if (dir) {\ entry = readdir (dir);\ while (entry && (!strcmp (entry->d_name, ".") ||\ + !fnmatch ("*.tmp", entry->d_name, 0) ||\ !strcmp (entry->d_name, ".."))) {\ entry = readdir (dir);\ }\ }\ } while (0); \ -typedef enum { - GD_STORE_SUCCESS, - GD_STORE_KEY_NULL, - GD_STORE_VALUE_NULL, - GD_STORE_KEY_VALUE_NULL, - GD_STORE_EOF, - GD_STORE_ENOMEM, - GD_STORE_STAT_FAILED -} glusterd_store_op_errno_t; int32_t glusterd_store_volinfo (glusterd_volinfo_t *volinfo, glusterd_volinfo_ver_ac_t ac); @@ -99,17 +106,7 @@ int32_t glusterd_store_delete_volume (glusterd_volinfo_t *volinfo); int32_t -glusterd_store_uuid (); - -int32_t -glusterd_store_handle_new (char *path, glusterd_store_handle_t **handle); - -int32_t -glusterd_store_save_value (int fd, char *key, char *value); - -int32_t -glusterd_store_retrieve_value (glusterd_store_handle_t *handle, - char *key, char **value); +glusterd_store_delete_snap (glusterd_snap_t *snap); int32_t glusterd_retrieve_uuid (); @@ -121,11 +118,8 @@ int32_t glusterd_store_delete_peerinfo (glusterd_peerinfo_t *peerinfo); int32_t -glusterd_store_delete_brick (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo); - -int32_t -glusterd_store_handle_destroy (glusterd_store_handle_t *handle); +glusterd_store_delete_brick (glusterd_brickinfo_t *brickinfo, + char *delete_path); int32_t glusterd_restore (); @@ -135,4 +129,36 @@ glusterd_perform_volinfo_version_action (glusterd_volinfo_t *volinfo, glusterd_volinfo_ver_ac_t ac); gf_boolean_t glusterd_store_is_valid_brickpath (char *volname, char *brick); + +int32_t +glusterd_store_perform_node_state_store (glusterd_volinfo_t *volinfo); + +int +glusterd_retrieve_op_version (xlator_t *this, int *op_version); + +int +glusterd_store_global_info (xlator_t *this); + +int32_t +glusterd_store_retrieve_options (xlator_t *this); + +int32_t +glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo); + +int32_t +glusterd_store_options (xlator_t *this, dict_t *opts); + +void +glusterd_replace_slash_with_hyphen (char *str); + +int32_t +glusterd_store_perform_volume_store (glusterd_volinfo_t *volinfo); + +int32_t +glusterd_store_snap (glusterd_snap_t *snap); + +int32_t +glusterd_store_update_missed_snaps (dict_t *dict, + int32_t missed_snap_count); + #endif diff --git a/xlators/mgmt/glusterd/src/glusterd-syncop.c b/xlators/mgmt/glusterd/src/glusterd-syncop.c new file mode 100644 index 000000000..438df8266 --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-syncop.c @@ -0,0 +1,1639 @@ +/* + Copyright (c) 2012-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +/* rpc related syncops */ +#include "rpc-clnt.h" +#include "protocol-common.h" +#include "xdr-generic.h" +#include "glusterd1-xdr.h" +#include "glusterd-syncop.h" + +#include "glusterd.h" +#include "glusterd-op-sm.h" +#include "glusterd-utils.h" +#include "glusterd-locks.h" + +extern glusterd_op_info_t opinfo; + +void +gd_synctask_barrier_wait (struct syncargs *args, int count) +{ + glusterd_conf_t *conf = THIS->private; + + synclock_unlock (&conf->big_lock); + synctask_barrier_wait (args, count); + synclock_lock (&conf->big_lock); + + syncbarrier_destroy (&args->barrier); +} + +static void +gd_mgmt_v3_collate_errors (struct syncargs *args, int op_ret, int op_errno, + char *op_errstr, int op_code, + glusterd_peerinfo_t *peerinfo, u_char *uuid) +{ + char err_str[PATH_MAX] = "Please check log file for details."; + char op_err[PATH_MAX] = ""; + int len = -1; + char *peer_str = NULL; + + if (op_ret) { + args->op_ret = op_ret; + args->op_errno = op_errno; + + if (peerinfo) + peer_str = peerinfo->hostname; + else + peer_str = uuid_utoa (uuid); + + if (op_errstr && strcmp (op_errstr, "")) { + len = snprintf (err_str, sizeof(err_str) - 1, + "Error: %s", op_errstr); + err_str[len] = '\0'; + } + + switch (op_code){ + case GLUSTERD_MGMT_V3_LOCK: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Locking failed " + "on %s. %s", peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_UNLOCK: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Unlocking failed " + "on %s. %s", peer_str, err_str); + break; + } + } + op_err[len] = '\0'; + + if (args->errstr) { + len = snprintf (err_str, sizeof(err_str) - 1, + "%s\n%s", args->errstr, + op_err); + GF_FREE (args->errstr); + args->errstr = NULL; + } else + len = snprintf (err_str, sizeof(err_str) - 1, + "%s", op_err); + err_str[len] = '\0'; + + gf_log ("", GF_LOG_ERROR, "%s", op_err); + args->errstr = gf_strdup (err_str); + } + + return; +} + +static void +gd_collate_errors (struct syncargs *args, int op_ret, int op_errno, + char *op_errstr, int op_code, + glusterd_peerinfo_t *peerinfo, u_char *uuid) +{ + char err_str[PATH_MAX] = "Please check log file for details."; + char op_err[PATH_MAX] = ""; + int len = -1; + char *peer_str = NULL; + + if (op_ret) { + args->op_ret = op_ret; + args->op_errno = op_errno; + + if (peerinfo) + peer_str = peerinfo->hostname; + else + peer_str = uuid_utoa (uuid); + + if (op_errstr && strcmp (op_errstr, "")) { + len = snprintf (err_str, sizeof(err_str) - 1, + "Error: %s", op_errstr); + err_str[len] = '\0'; + } + + switch (op_code){ + case GLUSTERD_MGMT_CLUSTER_UNLOCK : + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Unlocking failed on %s. %s", + peer_str, err_str); + break; + } + case GLUSTERD_MGMT_STAGE_OP : + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Staging failed on %s. %s", + peer_str, err_str); + break; + } + case GLUSTERD_MGMT_COMMIT_OP : + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Commit failed on %s. %s", + peer_str, err_str); + break; + } + } + op_err[len] = '\0'; + + if (args->errstr) { + len = snprintf (err_str, sizeof(err_str) - 1, + "%s\n%s", args->errstr, + op_err); + GF_FREE (args->errstr); + args->errstr = NULL; + } else + len = snprintf (err_str, sizeof(err_str) - 1, + "%s", op_err); + err_str[len] = '\0'; + + gf_log ("", GF_LOG_ERROR, "%s", op_err); + args->errstr = gf_strdup (err_str); + } + + return; +} + +void +gd_syncargs_init (struct syncargs *args, dict_t *op_ctx) +{ + args->dict = op_ctx; + pthread_mutex_init (&args->lock_dict, NULL); +} + +static void +gd_stage_op_req_free (gd1_mgmt_stage_op_req *req) +{ + if (!req) + return; + + GF_FREE (req->buf.buf_val); + GF_FREE (req); +} + +static void +gd_commit_op_req_free (gd1_mgmt_commit_op_req *req) +{ + if (!req) + return; + + GF_FREE (req->buf.buf_val); + GF_FREE (req); +} + +static void +gd_brick_op_req_free (gd1_mgmt_brick_op_req *req) +{ + if (!req) + return; + + if (strcmp (req->name, "") != 0) + GF_FREE (req->name); + GF_FREE (req->input.input_val); + GF_FREE (req); +} + +int +gd_syncop_submit_request (struct rpc_clnt *rpc, void *req, void *local, + void *cookie, rpc_clnt_prog_t *prog, int procnum, + fop_cbk_fn_t cbkfn, xdrproc_t xdrproc) +{ + int ret = -1; + struct iobuf *iobuf = NULL; + struct iobref *iobref = NULL; + int count = 0; + struct iovec iov = {0, }; + ssize_t req_size = 0; + call_frame_t *frame = NULL; + + GF_ASSERT (rpc); + if (!req) + goto out; + + req_size = xdr_sizeof (xdrproc, req); + iobuf = iobuf_get2 (rpc->ctx->iobuf_pool, req_size); + if (!iobuf) + goto out; + + iobref = iobref_new (); + if (!iobref) + goto out; + + frame = create_frame (THIS, THIS->ctx->pool); + if (!frame) + goto out; + + iobref_add (iobref, iobuf); + + iov.iov_base = iobuf->ptr; + iov.iov_len = iobuf_pagesize (iobuf); + + /* Create the xdr payload */ + ret = xdr_serialize_generic (iov, req, xdrproc); + if (ret == -1) + goto out; + + iov.iov_len = ret; + count = 1; + + frame->local = local; + frame->cookie = cookie; + + /* Send the msg */ + ret = rpc_clnt_submit (rpc, prog, procnum, cbkfn, + &iov, count, NULL, 0, iobref, + frame, NULL, 0, NULL, 0, NULL); + + /* TODO: do we need to start ping also? */ + +out: + iobref_unref (iobref); + iobuf_unref (iobuf); + + return ret; +} + +/* Defined in glusterd-rpc-ops.c */ +extern struct rpc_clnt_program gd_mgmt_prog; +extern struct rpc_clnt_program gd_brick_prog; +extern struct rpc_clnt_program gd_mgmt_v3_prog; + +int +glusterd_syncop_aggr_rsp_dict (glusterd_op_t op, dict_t *aggr, dict_t *rsp) +{ + int ret = 0; + + switch (op) { + case GD_OP_REPLACE_BRICK: + ret = glusterd_rb_use_rsp_dict (aggr, rsp); + if (ret) + goto out; + break; + + case GD_OP_SYNC_VOLUME: + ret = glusterd_sync_use_rsp_dict (aggr, rsp); + if (ret) + goto out; + break; + + case GD_OP_PROFILE_VOLUME: + ret = glusterd_profile_volume_use_rsp_dict (aggr, rsp); + if (ret) + goto out; + break; + + case GD_OP_GSYNC_CREATE: + break; + + case GD_OP_GSYNC_SET: + ret = glusterd_gsync_use_rsp_dict (aggr, rsp, NULL); + if (ret) + goto out; + break; + + case GD_OP_STATUS_VOLUME: + ret = glusterd_volume_status_copy_to_op_ctx_dict (aggr, rsp); + if (ret) + goto out; + break; + + case GD_OP_REBALANCE: + case GD_OP_DEFRAG_BRICK_VOLUME: + ret = glusterd_volume_rebalance_use_rsp_dict (aggr, rsp); + if (ret) + goto out; + break; + + case GD_OP_HEAL_VOLUME: + ret = glusterd_volume_heal_use_rsp_dict (aggr, rsp); + if (ret) + goto out; + + break; + + case GD_OP_QUOTA: + case GD_OP_CLEARLOCKS_VOLUME: + ret = glusterd_use_rsp_dict (aggr, rsp); + if (ret) + goto out; + + break; + + case GD_OP_SYS_EXEC: + ret = glusterd_sys_exec_output_rsp_dict (aggr, rsp); + if (ret) + goto out; + break; + + case GD_OP_SNAP: + ret = glusterd_snap_use_rsp_dict (aggr, rsp); + if (ret) + goto out; + break; + + default: + break; + } +out: + return ret; +} + +int32_t +_gd_syncop_mgmt_lock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_cluster_lock_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int op_ret = -1; + int op_errno = -1; + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_cluster_lock_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + /* Set peer as locked, so we unlock only the locked peers */ + if (rsp.op_ret == 0) + peerinfo->locked = _gf_true; + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; +out: + gd_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_CLUSTER_LOCK, peerinfo, rsp.uuid); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_syncop_mgmt_lock_cbk (struct rpc_req *req, struct iovec *iov, int count, + void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + _gd_syncop_mgmt_lock_cbk); +} + +int +gd_syncop_mgmt_lock (glusterd_peerinfo_t *peerinfo, struct syncargs *args, + uuid_t my_uuid, uuid_t recv_uuid) +{ + int ret = -1; + gd1_mgmt_cluster_lock_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + + uuid_copy (req.uuid, my_uuid); + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_prog, + GLUSTERD_MGMT_CLUSTER_LOCK, + gd_syncop_mgmt_lock_cbk, + (xdrproc_t) xdr_gd1_mgmt_cluster_lock_req); + synclock_lock (&conf->big_lock); + return ret; +} + +int32_t +gd_syncop_mgmt_v3_lock_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_lock_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int op_ret = -1; + int op_errno = -1; + + GF_ASSERT(req); + GF_ASSERT(iov); + GF_ASSERT(myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_LOCK, + peerinfo, rsp.uuid); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_syncop_mgmt_v3_lock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_syncop_mgmt_v3_lock_cbk_fn); +} + +int +gd_syncop_mgmt_v3_lock (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid, uuid_t txn_id) +{ + int ret = -1; + gd1_mgmt_v3_lock_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + + GF_ASSERT(op_ctx); + GF_ASSERT(peerinfo); + GF_ASSERT(args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + uuid_copy (req.txn_id, txn_id); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_LOCK, + gd_syncop_mgmt_v3_lock_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_lock_req); + synclock_lock (&conf->big_lock); +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +gd_syncop_mgmt_v3_unlock_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_unlock_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int op_ret = -1; + int op_errno = -1; + + GF_ASSERT(req); + GF_ASSERT(iov); + GF_ASSERT(myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + /* Set peer as locked, so we unlock only the locked peers */ + if (rsp.op_ret == 0) + peerinfo->locked = _gf_true; + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_UNLOCK, + peerinfo, rsp.uuid); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_syncop_mgmt_v3_unlock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_syncop_mgmt_v3_unlock_cbk_fn); +} + +int +gd_syncop_mgmt_v3_unlock (dict_t *op_ctx, glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid, uuid_t txn_id) +{ + int ret = -1; + gd1_mgmt_v3_unlock_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + + GF_ASSERT(op_ctx); + GF_ASSERT(peerinfo); + GF_ASSERT(args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + uuid_copy (req.txn_id, txn_id); + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_UNLOCK, + gd_syncop_mgmt_v3_unlock_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_unlock_req); + synclock_lock (&conf->big_lock); +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +_gd_syncop_mgmt_unlock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_cluster_unlock_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int op_ret = -1; + int op_errno = -1; + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + peerinfo->locked = _gf_false; + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; +out: + gd_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_CLUSTER_UNLOCK, peerinfo, rsp.uuid); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_syncop_mgmt_unlock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + _gd_syncop_mgmt_unlock_cbk); +} + + +int +gd_syncop_mgmt_unlock (glusterd_peerinfo_t *peerinfo, struct syncargs *args, + uuid_t my_uuid, uuid_t recv_uuid) +{ + int ret = -1; + gd1_mgmt_cluster_unlock_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + + uuid_copy (req.uuid, my_uuid); + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_prog, + GLUSTERD_MGMT_CLUSTER_UNLOCK, + gd_syncop_mgmt_unlock_cbk, + (xdrproc_t) xdr_gd1_mgmt_cluster_lock_req); + synclock_lock (&conf->big_lock); + return ret; +} + +int32_t +_gd_syncop_stage_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int ret = -1; + gd1_mgmt_stage_op_rsp rsp = {{0},}; + struct syncargs *args = NULL; + xlator_t *this = NULL; + dict_t *rsp_dict = NULL; + call_frame_t *frame = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + int op_ret = -1; + int op_errno = -1; + + this = THIS; + frame = myframe; + args = frame->local; + frame->local = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_stage_op_rsp); + if (ret < 0) + goto out; + + if (rsp.dict.dict_len) { + /* Unserialize the dictionary */ + rsp_dict = dict_new (); + + ret = dict_unserialize (rsp.dict.dict_val, + rsp.dict.dict_len, + &rsp_dict); + if (ret < 0) { + GF_FREE (rsp.dict.dict_val); + goto out; + } else { + rsp_dict->extra_stdfree = rsp.dict.dict_val; + } + } + + ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, "Staging response " + "for 'Volume %s' received from unknown " + "peer: %s", gd_op_list[rsp.op], + uuid_utoa (rsp.uuid)); + goto out; + } + + uuid_copy (args->uuid, rsp.uuid); + if (rsp.op == GD_OP_REPLACE_BRICK) { + pthread_mutex_lock (&args->lock_dict); + { + ret = glusterd_syncop_aggr_rsp_dict (rsp.op, args->dict, + rsp_dict); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from " + " node/brick"); + } + pthread_mutex_unlock (&args->lock_dict); + } + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + +out: + gd_collate_errors (args, op_ret, op_errno, rsp.op_errstr, + GLUSTERD_MGMT_STAGE_OP, peerinfo, rsp.uuid); + + if (rsp_dict) + dict_unref (rsp_dict); + + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_syncop_stage_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + _gd_syncop_stage_op_cbk); +} + + +int +gd_syncop_mgmt_stage_op (struct rpc_clnt *rpc, struct syncargs *args, + uuid_t my_uuid, uuid_t recv_uuid, int op, + dict_t *dict_out, dict_t *op_ctx) +{ + gd1_mgmt_stage_op_req *req = NULL; + glusterd_conf_t *conf = THIS->private; + int ret = -1; + + req = GF_CALLOC (1, sizeof (*req), gf_gld_mt_mop_stage_req_t); + if (!req) + goto out; + + uuid_copy (req->uuid, my_uuid); + req->op = op; + + ret = dict_allocate_and_serialize (dict_out, + &req->buf.buf_val, &req->buf.buf_len); + if (ret) + goto out; + + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (rpc, req, args, NULL, &gd_mgmt_prog, + GLUSTERD_MGMT_STAGE_OP, + gd_syncop_stage_op_cbk, + (xdrproc_t) xdr_gd1_mgmt_stage_op_req); + synclock_lock (&conf->big_lock); +out: + gd_stage_op_req_free (req); + return ret; + +} + +int32_t +_gd_syncop_brick_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + struct syncargs *args = NULL; + gd1_mgmt_brick_op_rsp rsp = {0,}; + int ret = -1; + call_frame_t *frame = NULL; + + frame = myframe; + args = frame->local; + frame->local = NULL; + + /* initialize */ + args->op_ret = -1; + args->op_errno = EINVAL; + + if (-1 == req->rpc_status) { + args->op_errno = ENOTCONN; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_brick_op_rsp); + if (ret < 0) + goto out; + + if (rsp.output.output_len) { + args->dict = dict_new (); + if (!args->dict) { + ret = -1; + args->op_errno = ENOMEM; + goto out; + } + + ret = dict_unserialize (rsp.output.output_val, + rsp.output.output_len, + &args->dict); + if (ret < 0) + goto out; + } + + args->op_ret = rsp.op_ret; + args->op_errno = rsp.op_errno; + args->errstr = gf_strdup (rsp.op_errstr); + +out: + if ((rsp.op_errstr) && (strcmp (rsp.op_errstr, "") != 0)) + free (rsp.op_errstr); + free (rsp.output.output_val); + + STACK_DESTROY (frame->root); + __wake (args); + + return 0; +} + +int32_t +gd_syncop_brick_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + _gd_syncop_brick_op_cbk); +} + +int +gd_syncop_mgmt_brick_op (struct rpc_clnt *rpc, glusterd_pending_node_t *pnode, + int op, dict_t *dict_out, dict_t *op_ctx, + char **errstr) +{ + struct syncargs args = {0, }; + gd1_mgmt_brick_op_req *req = NULL; + int ret = 0; + xlator_t *this = NULL; + + this = THIS; + args.op_ret = -1; + args.op_errno = ENOTCONN; + + if ((pnode->type == GD_NODE_NFS) || + ((pnode->type == GD_NODE_SHD) && + (op == GD_OP_STATUS_VOLUME))) { + ret = glusterd_node_op_build_payload + (op, &req, dict_out); + + } else { + ret = glusterd_brick_op_build_payload + (op, pnode->node, &req, dict_out); + + } + + if (ret) + goto out; + + GD_SYNCOP (rpc, (&args), NULL, gd_syncop_brick_op_cbk, req, + &gd_brick_prog, req->op, xdr_gd1_mgmt_brick_op_req); + + if (args.errstr) { + if ((strlen(args.errstr) > 0) && errstr) + *errstr = args.errstr; + else + GF_FREE (args.errstr); + } + + if (GD_OP_STATUS_VOLUME == op) { + ret = dict_set_int32 (args.dict, "index", pnode->index); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting index on brick status" + " rsp dict"); + args.op_ret = -1; + goto out; + } + } + if (args.op_ret == 0) + glusterd_handle_node_rsp (dict_out, pnode->node, op, + args.dict, op_ctx, errstr, + pnode->type); + +out: + errno = args.op_errno; + if (args.dict) + dict_unref (args.dict); + gd_brick_op_req_free (req); + return args.op_ret; + +} + +int32_t +_gd_syncop_commit_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int ret = -1; + gd1_mgmt_commit_op_rsp rsp = {{0},}; + struct syncargs *args = NULL; + xlator_t *this = NULL; + dict_t *rsp_dict = NULL; + call_frame_t *frame = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + int op_ret = -1; + int op_errno = -1; + + this = THIS; + frame = myframe; + args = frame->local; + frame->local = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_commit_op_rsp); + if (ret < 0) { + goto out; + } + + if (rsp.dict.dict_len) { + /* Unserialize the dictionary */ + rsp_dict = dict_new (); + + ret = dict_unserialize (rsp.dict.dict_val, + rsp.dict.dict_len, + &rsp_dict); + if (ret < 0) { + GF_FREE (rsp.dict.dict_val); + goto out; + } else { + rsp_dict->extra_stdfree = rsp.dict.dict_val; + } + } + + ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, "Commit response " + "for 'Volume %s' received from unknown " + "peer: %s", gd_op_list[rsp.op], + uuid_utoa (rsp.uuid)); + goto out; + } + + uuid_copy (args->uuid, rsp.uuid); + pthread_mutex_lock (&args->lock_dict); + { + ret = glusterd_syncop_aggr_rsp_dict (rsp.op, args->dict, + rsp_dict); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from " + " node/brick"); + } + pthread_mutex_unlock (&args->lock_dict); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + +out: + gd_collate_errors (args, op_ret, op_errno, rsp.op_errstr, + GLUSTERD_MGMT_COMMIT_OP, peerinfo, rsp.uuid); + if (rsp_dict) + dict_unref (rsp_dict); + + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + + return 0; +} + +int32_t +gd_syncop_commit_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + _gd_syncop_commit_op_cbk); +} + + +int +gd_syncop_mgmt_commit_op (struct rpc_clnt *rpc, struct syncargs *args, + uuid_t my_uuid, uuid_t recv_uuid, + int op, dict_t *dict_out, dict_t *op_ctx) +{ + glusterd_conf_t *conf = THIS->private; + gd1_mgmt_commit_op_req *req = NULL; + int ret = -1; + + req = GF_CALLOC (1, sizeof (*req), gf_gld_mt_mop_commit_req_t); + if (!req) + goto out; + + uuid_copy (req->uuid, my_uuid); + req->op = op; + + ret = dict_allocate_and_serialize (dict_out, + &req->buf.buf_val, &req->buf.buf_len); + if (ret) + goto out; + + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (rpc, req, args, NULL, &gd_mgmt_prog, + GLUSTERD_MGMT_COMMIT_OP , + gd_syncop_commit_op_cbk, + (xdrproc_t) xdr_gd1_mgmt_commit_op_req); + synclock_lock (&conf->big_lock); +out: + gd_commit_op_req_free (req); + return ret; +} + + +int +gd_build_peers_list (struct list_head *peers, struct list_head *xact_peers, + glusterd_op_t op) +{ + glusterd_peerinfo_t *peerinfo = NULL; + int npeers = 0; + + list_for_each_entry (peerinfo, peers, uuid_list) { + if (!peerinfo->connected) + continue; + if (op != GD_OP_SYNC_VOLUME && + peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) + continue; + + list_add_tail (&peerinfo->op_peers_list, xact_peers); + npeers++; + } + return npeers; +} + +int +gd_lock_op_phase (glusterd_conf_t *conf, glusterd_op_t op, dict_t *op_ctx, + char **op_errstr, int npeers, uuid_t txn_id) +{ + int ret = -1; + int peer_cnt = 0; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + + peers = &conf->xaction_peers; + + if (!npeers) { + ret = 0; + goto out; + } + + this = THIS; + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + if (conf->op_version < 3) { + /* Reset lock status */ + peerinfo->locked = _gf_false; + gd_syncop_mgmt_lock (peerinfo, &args, + MY_UUID, peer_uuid); + } else + gd_syncop_mgmt_v3_lock (op, op_ctx, peerinfo, &args, + MY_UUID, peer_uuid, txn_id); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + else { + ret = gf_asprintf (op_errstr, "Another transaction could be " + "in progress. Please try again after " + "sometime."); + if (ret == -1) + *op_errstr = NULL; + + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire lock"); + + } + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent lock op req for 'Volume %s' " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + return ret; +} + +int +gd_stage_op_phase (struct list_head *peers, glusterd_op_t op, dict_t *op_ctx, + dict_t *req_dict, char **op_errstr, int npeers) +{ + int ret = -1; + int peer_cnt = 0; + dict_t *rsp_dict = NULL; + char *hostname = NULL; + xlator_t *this = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + uuid_t tmp_uuid = {0}; + char *errstr = NULL; + struct syncargs args = {0}; + + this = THIS; + rsp_dict = dict_new (); + if (!rsp_dict) + goto out; + + ret = glusterd_op_stage_validate (op, req_dict, op_errstr, rsp_dict); + if (ret) { + hostname = "localhost"; + goto stage_done; + } + + if ((op == GD_OP_REPLACE_BRICK)) { + ret = glusterd_syncop_aggr_rsp_dict (op, op_ctx, rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from node/brick"); + goto out; + } + } + dict_unref (rsp_dict); + rsp_dict = NULL; + +stage_done: + if (ret) { + gf_log (this->name, GF_LOG_ERROR, LOGSTR_STAGE_FAIL, + gd_op_list[op], hostname, (*op_errstr) ? ":" : " ", + (*op_errstr) ? *op_errstr : " "); + if (*op_errstr == NULL) + gf_asprintf (op_errstr, OPERRSTR_STAGE_FAIL, hostname); + goto out; + } + + if (!npeers) { + ret = 0; + goto out; + } + + gd_syncargs_init (&args, op_ctx); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + ret = gd_syncop_mgmt_stage_op (peerinfo->rpc, &args, + MY_UUID, tmp_uuid, + op, req_dict, op_ctx); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + else if (dict_get_str (op_ctx, "errstr", &errstr) == 0) + *op_errstr = gf_strdup (errstr); + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent stage op req for 'Volume %s' " + "to %d peers", gd_op_list[op], peer_cnt); +out: + if (rsp_dict) + dict_unref (rsp_dict); + return ret; +} + +int +gd_commit_op_phase (struct list_head *peers, glusterd_op_t op, dict_t *op_ctx, + dict_t *req_dict, char **op_errstr, int npeers) +{ + dict_t *rsp_dict = NULL; + int peer_cnt = -1; + int ret = -1; + char *hostname = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + uuid_t tmp_uuid = {0}; + char *errstr = NULL; + struct syncargs args = {0}; + + this = THIS; + rsp_dict = dict_new (); + if (!rsp_dict) { + ret = -1; + goto out; + } + + ret = glusterd_op_commit_perform (op, req_dict, op_errstr, rsp_dict); + if (ret) { + hostname = "localhost"; + goto commit_done; + } + if (op != GD_OP_SYNC_VOLUME) { + ret = glusterd_syncop_aggr_rsp_dict (op, op_ctx, rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response " + "from node/brick"); + goto out; + } + } + dict_unref (rsp_dict); + rsp_dict = NULL; + +commit_done: + if (ret) { + gf_log (this->name, GF_LOG_ERROR, LOGSTR_COMMIT_FAIL, + gd_op_list[op], hostname, (*op_errstr) ? ":" : " ", + (*op_errstr) ? *op_errstr : " "); + if (*op_errstr == NULL) + gf_asprintf (op_errstr, OPERRSTR_COMMIT_FAIL, + hostname); + goto out; + } + + if (!npeers) { + ret = 0; + goto out; + } + + gd_syncargs_init (&args, op_ctx); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + ret = gd_syncop_mgmt_commit_op (peerinfo->rpc, &args, + MY_UUID, tmp_uuid, + op, req_dict, op_ctx); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + ret = args.op_ret; + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + else if (dict_get_str (op_ctx, "errstr", &errstr) == 0) + *op_errstr = gf_strdup (errstr); + + gf_log (this->name, GF_LOG_DEBUG, "Sent commit op req for 'Volume %s' " + "to %d peers", gd_op_list[op], peer_cnt); +out: + if (!ret) + glusterd_op_modify_op_ctx (op, op_ctx); + + if (rsp_dict) + dict_unref (rsp_dict); + + GF_FREE (args.errstr); + args.errstr = NULL; + + return ret; +} + +int +gd_unlock_op_phase (glusterd_conf_t *conf, glusterd_op_t op, int op_ret, + rpcsvc_request_t *req, dict_t *op_ctx, char *op_errstr, + int npeers, char *volname, gf_boolean_t is_acquired, + uuid_t txn_id) +{ + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_peerinfo_t *tmp = NULL; + uuid_t tmp_uuid = {0}; + int peer_cnt = 0; + int ret = -1; + xlator_t *this = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + + peers = &conf->xaction_peers; + + if (!npeers) { + ret = 0; + goto out; + } + + /* If the lock has not been held during this + * transaction, do not send unlock requests */ + if (!is_acquired) + goto out; + + this = THIS; + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry_safe (peerinfo, tmp, peers, op_peers_list) { + if (conf->op_version < 3) { + /* Only unlock peers that were locked */ + if (peerinfo->locked) + gd_syncop_mgmt_unlock (peerinfo, &args, + MY_UUID, tmp_uuid); + } else + gd_syncop_mgmt_v3_unlock (op_ctx, peerinfo, + &args, MY_UUID, + tmp_uuid, txn_id); + peer_cnt++; + list_del_init (&peerinfo->op_peers_list); + } + gd_synctask_barrier_wait((&args), peer_cnt); + ret = args.op_ret; + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to unlock " + "on some peer(s)"); + } + +out: + glusterd_op_send_cli_response (op, op_ret, 0, req, op_ctx, op_errstr); + glusterd_op_clear_op (op); + if (is_acquired) { + /* Based on the op-version, we release * + * the cluster or mgmt_v3 lock */ + if (conf->op_version < 3) + glusterd_unlock (MY_UUID); + else { + ret = glusterd_mgmt_v3_unlock (volname, MY_UUID, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to release lock for %s", + volname); + } + } + + return 0; +} + +int +gd_get_brick_count (struct list_head *bricks) +{ + glusterd_pending_node_t *pending_node = NULL; + int npeers = 0; + list_for_each_entry (pending_node, bricks, list) { + npeers++; + } + return npeers; +} + +int +gd_brick_op_phase (glusterd_op_t op, dict_t *op_ctx, dict_t *req_dict, + char **op_errstr) +{ + glusterd_pending_node_t *pending_node = NULL; + struct list_head selected = {0,}; + xlator_t *this = NULL; + int brick_count = 0; + int ret = -1; + rpc_clnt_t *rpc = NULL; + dict_t *rsp_dict = NULL; + glusterd_conf_t *conf = NULL; + + this = THIS; + conf = this->private; + rsp_dict = dict_new (); + if (!rsp_dict) { + ret = -1; + goto out; + } + + INIT_LIST_HEAD (&selected); + ret = glusterd_op_bricks_select (op, req_dict, op_errstr, &selected, rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + (*op_errstr)? *op_errstr: "Brick op failed. Check " + "glusterd log file for more details."); + goto out; + } + + if (op == GD_OP_HEAL_VOLUME) { + ret = glusterd_syncop_aggr_rsp_dict (op, op_ctx, rsp_dict); + if (ret) + goto out; + } + dict_unref (rsp_dict); + rsp_dict = NULL; + + brick_count = 0; + list_for_each_entry (pending_node, &selected, list) { + rpc = glusterd_pending_node_get_rpc (pending_node); + if (!rpc) { + if (pending_node->type == GD_NODE_REBALANCE) { + ret = 0; + glusterd_defrag_volume_node_rsp (req_dict, + NULL, op_ctx); + goto out; + } + + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Brick Op failed " + "due to rpc failure."); + goto out; + } + ret = gd_syncop_mgmt_brick_op (rpc, pending_node, op, req_dict, + op_ctx, op_errstr); + if (ret) + goto out; + + brick_count++; + } + + ret = 0; +out: + if (rsp_dict) + dict_unref (rsp_dict); + gf_log (this->name, GF_LOG_DEBUG, "Sent op req to %d bricks", + brick_count); + return ret; +} + +void +gd_sync_task_begin (dict_t *op_ctx, rpcsvc_request_t * req) +{ + int ret = -1; + int npeers = 0; + dict_t *req_dict = NULL; + glusterd_conf_t *conf = NULL; + glusterd_op_t op = 0; + int32_t tmp_op = 0; + char *op_errstr = NULL; + char *tmp = NULL; + char *volname = NULL; + xlator_t *this = NULL; + gf_boolean_t is_acquired = _gf_false; + uuid_t *txn_id = NULL; + uuid_t *originator_uuid = NULL; + glusterd_op_info_t txn_opinfo; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + GF_ASSERT (conf); + + /* Generate a transaction-id for this operation and + * save it in the dict */ + txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + if (!txn_id) { + ret = -1; + goto out; + } + + uuid_generate (*txn_id); + + ret = dict_set_bin (op_ctx, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set transaction id."); + goto out; + } else + gf_log (this->name, GF_LOG_DEBUG, + "Transaction_id = %s", uuid_utoa (*txn_id)); + + /* Save the MY_UUID as the originator_uuid */ + originator_uuid = GF_CALLOC (1, sizeof(uuid_t), + gf_common_mt_uuid_t); + if (!originator_uuid) { + ret = -1; + goto out; + } + + uuid_copy (*originator_uuid, MY_UUID); + ret = dict_set_bin (op_ctx, "originator_uuid", + originator_uuid, sizeof (uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set originator_uuid."); + goto out; + } + + ret = dict_get_int32 (op_ctx, GD_SYNC_OPCODE_KEY, &tmp_op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volume " + "operation"); + goto out; + } + + op = tmp_op; + + /* Based on the op_version, acquire a cluster or mgmt_v3 lock */ + if (conf->op_version < 3) { + ret = glusterd_lock (MY_UUID); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock"); + gf_asprintf (&op_errstr, + "Another transaction is in progress. " + "Please try again after sometime."); + goto out; + } + } else { + + /* If no volname is given as a part of the command, locks will + * not be held */ + ret = dict_get_str (op_ctx, "volname", &tmp); + if (ret) { + gf_log ("", GF_LOG_DEBUG, "Failed to get volume " + "name"); + goto local_locking_done; + } else { + /* Use a copy of volname, as cli response will be + * sent before the unlock, and the volname in the + * dict, might be removed */ + volname = gf_strdup (tmp); + if (!volname) + goto out; + } + + ret = glusterd_mgmt_v3_lock (volname, MY_UUID, "vol"); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock for %s", volname); + gf_asprintf (&op_errstr, + "Another transaction is in progress " + "for %s. Please try again after sometime.", + volname); + goto out; + } + } + + is_acquired = _gf_true; + +local_locking_done: + + /* Save opinfo for this transaction with the transaction id */ + glusterd_txn_opinfo_init (&txn_opinfo, NULL, &op, NULL, NULL); + ret = glusterd_set_txn_opinfo (txn_id, &txn_opinfo); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to set transaction's opinfo"); + + opinfo = txn_opinfo; + + INIT_LIST_HEAD (&conf->xaction_peers); + + /* Make 'volume status tasks' command a local operation. + * This is accomplished by setting npeers to 0. + */ + if (!glusterd_is_status_tasks_op (op, op_ctx)) + npeers = gd_build_peers_list (&conf->peers, + &conf->xaction_peers, op); + + /* If no volname is given as a part of the command, locks will + * not be held */ + if (volname) { + ret = gd_lock_op_phase (conf, op, op_ctx, &op_errstr, npeers, *txn_id); + if (ret) + goto out; + } + + ret = glusterd_op_build_payload (&req_dict, &op_errstr, op_ctx); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD, + gd_op_list[op]); + if (op_errstr == NULL) + gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD); + goto out; + } + + ret = gd_stage_op_phase (&conf->xaction_peers, op, op_ctx, req_dict, + &op_errstr, npeers); + if (ret) + goto out; + + ret = gd_brick_op_phase (op, op_ctx, req_dict, &op_errstr); + if (ret) + goto out; + + ret = gd_commit_op_phase (&conf->xaction_peers, op, op_ctx, req_dict, + &op_errstr, npeers); + if (ret) + goto out; + + ret = 0; +out: + (void) gd_unlock_op_phase (conf, op, ret, req, op_ctx, op_errstr, + npeers, volname, is_acquired, *txn_id); + + /* Clearing the transaction opinfo */ + ret = glusterd_clear_txn_opinfo (txn_id); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to clear transaction's opinfo"); + + if (volname) + GF_FREE (volname); + + if (req_dict) + dict_unref (req_dict); + + if (op_errstr) { + GF_FREE (op_errstr); + op_errstr = NULL; + } + + return; +} + +int32_t +glusterd_op_begin_synctask (rpcsvc_request_t *req, glusterd_op_t op, + void *dict) +{ + int ret = 0; + + ret = dict_set_int32 (dict, GD_SYNC_OPCODE_KEY, op); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "dict set failed for setting operations"); + goto out; + } + + gd_sync_task_begin (dict, req); + ret = 0; +out: + + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-syncop.h b/xlators/mgmt/glusterd/src/glusterd-syncop.h new file mode 100644 index 000000000..e83ea2f4c --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-syncop.h @@ -0,0 +1,71 @@ +/* + Copyright (c) 2012-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +#ifndef __RPC_SYNCOP_H +#define __RPC_SYNCOP_H + +#include "syncop.h" +#include "glusterd-sm.h" +#include "glusterd.h" + +#define GD_SYNC_OPCODE_KEY "sync-mgmt-operation" + +/* gd_syncop_* */ +#define GD_SYNCOP(rpc, stb, cookie, cbk, req, prog, procnum, xdrproc) do { \ + int ret = 0; \ + struct synctask *task = NULL; \ + glusterd_conf_t *conf= THIS->private; \ + \ + task = synctask_get (); \ + stb->task = task; \ + \ + /*This is to ensure that the brick_op_cbk is able to \ + * take the big lock*/ \ + synclock_unlock (&conf->big_lock); \ + ret = gd_syncop_submit_request (rpc, req, stb, cookie, \ + prog, procnum, cbk, \ + (xdrproc_t)xdrproc); \ + if (!ret) \ + synctask_yield (stb->task); \ + synclock_lock (&conf->big_lock); \ + } while (0) + + +int gd_syncop_submit_request (struct rpc_clnt *rpc, void *req, void *local, + void *cookie, rpc_clnt_prog_t *prog, int procnum, + fop_cbk_fn_t cbkfn, xdrproc_t xdrproc); + + +int gd_syncop_mgmt_lock (glusterd_peerinfo_t *peerinfo, struct syncargs *arg, + uuid_t my_uuid, uuid_t recv_uuid); +int gd_syncop_mgmt_unlock (glusterd_peerinfo_t *peerinfo, struct syncargs *arg, + uuid_t my_uuid, uuid_t recv_uuid); +int gd_syncop_mgmt_stage_op (struct rpc_clnt *rpc, struct syncargs *arg, + uuid_t my_uuid, uuid_t recv_uuid, int op, + dict_t *dict_out, dict_t *op_ctx); +int gd_syncop_mgmt_commit_op (struct rpc_clnt *rpc, struct syncargs *arg, + uuid_t my_uuid, uuid_t recv_uuid, int op, + dict_t *dict_out, dict_t *op_ctx); + +void +gd_synctask_barrier_wait (struct syncargs *args, int count); + +int +gd_build_peers_list (struct list_head *peers, struct list_head *xact_peers, + glusterd_op_t op); +int +gd_brick_op_phase (glusterd_op_t op, dict_t *op_ctx, dict_t *req_dict, + char **op_errstr); + +int +glusterd_syncop_aggr_rsp_dict (glusterd_op_t op, dict_t *aggr, dict_t *rsp); + +void +gd_syncargs_init (struct syncargs *args, dict_t *op_ctx); +#endif /* __RPC_SYNCOP_H */ diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c index c0462d9c7..e8ae05851 100644 --- a/xlators/mgmt/glusterd/src/glusterd-utils.c +++ b/xlators/mgmt/glusterd/src/glusterd-utils.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" @@ -32,7 +22,7 @@ #include "timer.h" #include "defaults.h" #include "compat.h" -#include "md5.h" +#include "syncop.h" #include "run.h" #include "compat-errno.h" #include "statedump.h" @@ -45,9 +35,10 @@ #include "glusterd-store.h" #include "glusterd-volgen.h" #include "glusterd-pmap.h" +#include "glusterfs-acl.h" +#include "glusterd-locks.h" #include "xdr-generic.h" - #include <sys/resource.h> #include <inttypes.h> #include <signal.h> @@ -59,21 +50,48 @@ #include <unistd.h> #include <fnmatch.h> #include <sys/statvfs.h> +#include <ifaddrs.h> +#ifdef HAVE_BD_XLATOR +#include <lvm2app.h> +#endif + +#ifdef GF_LINUX_HOST_OS +#include <mntent.h> +#endif #ifdef GF_SOLARIS_HOST_OS #include <sys/sockio.h> #endif -#define MOUNT_PROGRAM 100005 -#define NFS_PROGRAM 100003 -#define NFSV3_VERSION 3 -#define MOUNTV3_VERSION 3 -#define MOUNTV1_VERSION 1 +#define NFS_PROGRAM 100003 +#define NFSV3_VERSION 3 + +#define MOUNT_PROGRAM 100005 +#define MOUNTV3_VERSION 3 +#define MOUNTV1_VERSION 1 + +#define NLM_PROGRAM 100021 +#define NLMV4_VERSION 4 +#define NLMV1_VERSION 1 + +#define CEILING_POS(X) (((X)-(int)(X)) > 0 ? (int)((X)+1) : (int)(X)) -char *glusterd_sock_dir = "/tmp"; static glusterd_lock_t lock; +char* +gd_peer_uuid_str (glusterd_peerinfo_t *peerinfo) +{ + if ((peerinfo == NULL) || uuid_is_null (peerinfo->uuid)) + return NULL; + + if (peerinfo->uuid_str[0] == '\0') + uuid_utoa_r (peerinfo->uuid, peerinfo->uuid_str); + + return peerinfo->uuid_str; +} + + int32_t glusterd_get_lock_owner (uuid_t *uuid) { @@ -111,129 +129,6 @@ glusterd_is_fuse_available () return _gf_false; } -gf_boolean_t -glusterd_is_loopback_localhost (const struct sockaddr *sa, char *hostname) -{ - GF_ASSERT (sa); - - gf_boolean_t is_local = _gf_false; - const struct in_addr *addr4 = NULL; - const struct in6_addr *addr6 = NULL; - uint8_t *ap = NULL; - struct in6_addr loopbackaddr6 = IN6ADDR_LOOPBACK_INIT; - - switch (sa->sa_family) { - case AF_INET: - addr4 = &(((struct sockaddr_in *)sa)->sin_addr); - ap = (uint8_t*)&addr4->s_addr; - if (ap[0] == 127) - is_local = _gf_true; - break; - - case AF_INET6: - addr6 = &(((struct sockaddr_in6 *)sa)->sin6_addr); - if (memcmp (addr6, &loopbackaddr6, - sizeof (loopbackaddr6)) == 0) - is_local = _gf_true; - break; - - default: - if (hostname) - gf_log ("glusterd", GF_LOG_ERROR, - "unknown address family %d for %s", - sa->sa_family, hostname); - break; - } - - return is_local; -} - -char * -get_ip_from_addrinfo (struct addrinfo *addr, char **ip) -{ - char buf[64]; - void *in_addr = NULL; - struct sockaddr_in *s4 = NULL; - struct sockaddr_in6 *s6 = NULL; - - switch (addr->ai_family) - { - case AF_INET: - s4 = (struct sockaddr_in *)addr->ai_addr; - in_addr = &s4->sin_addr; - break; - - case AF_INET6: - s6 = (struct sockaddr_in6 *)addr->ai_addr; - in_addr = &s6->sin6_addr; - break; - - default: - gf_log ("glusterd", GF_LOG_ERROR, "Invalid family"); - return NULL; - } - - if (!inet_ntop(addr->ai_family, in_addr, buf, sizeof(buf))) { - gf_log ("glusterd", GF_LOG_ERROR, "String conversion failed"); - return NULL; - } - - *ip = strdup (buf); - return *ip; -} - -int32_t -glusterd_is_local_addr (char *hostname) -{ - int32_t ret = -1; - struct addrinfo *result = NULL; - struct addrinfo *res = NULL; - int32_t found = 0; - int sd = -1; - char *ip = NULL; - - ret = getaddrinfo (hostname, NULL, NULL, &result); - - if (ret != 0) { - gf_log ("", GF_LOG_ERROR, "error in getaddrinfo: %s\n", - gai_strerror(ret)); - goto out; - } - - for (res = result; res != NULL; res = res->ai_next) { - found = glusterd_is_loopback_localhost (res->ai_addr, hostname); - if (found) - goto out; - } - - for (res = result; res != NULL; res = res->ai_next) { - gf_log ("glusterd", GF_LOG_DEBUG, "%s ", get_ip_from_addrinfo (res, &ip)); - sd = socket (res->ai_family, SOCK_DGRAM, 0); - if (sd == -1) - goto out; - /*If bind succeeds then its a local address*/ - ret = bind (sd, res->ai_addr, res->ai_addrlen); - if (ret == 0) { - found = _gf_true; - gf_log ("glusterd", GF_LOG_INFO, "%s is local", get_ip_from_addrinfo (res, &ip)); - close (sd); - break; - } - close (sd); - } - -out: - if (result) - freeaddrinfo (result); - - if (found) - gf_log ("glusterd", GF_LOG_DEBUG, "%s is local", hostname); - else - gf_log ("glusterd", GF_LOG_DEBUG, "%s is not local", hostname); - - return !found; -} - int32_t glusterd_lock (uuid_t uuid) { @@ -242,13 +137,17 @@ glusterd_lock (uuid_t uuid) char new_owner_str[50]; char owner_str[50]; int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (uuid); glusterd_get_lock_owner (&owner); if (!uuid_is_null (owner)) { - gf_log ("glusterd", GF_LOG_ERROR, "Unable to get lock" + gf_log (this->name, GF_LOG_ERROR, "Unable to get lock" " for uuid: %s, lock held by: %s", uuid_utoa_r (uuid, new_owner_str), uuid_utoa_r (owner, owner_str)); @@ -258,7 +157,7 @@ glusterd_lock (uuid_t uuid) ret = glusterd_set_lock_owner (uuid); if (!ret) { - gf_log ("glusterd", GF_LOG_INFO, "Cluster lock held by" + gf_log (this->name, GF_LOG_DEBUG, "Cluster lock held by" " %s", uuid_utoa (uuid)); } @@ -274,21 +173,25 @@ glusterd_unlock (uuid_t uuid) char new_owner_str[50]; char owner_str[50]; int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (uuid); glusterd_get_lock_owner (&owner); - if (NULL == owner) { - gf_log ("glusterd", GF_LOG_ERROR, "Cluster lock not held!"); + if (uuid_is_null (owner)) { + gf_log (this->name, GF_LOG_ERROR, "Cluster lock not held!"); goto out; } ret = uuid_compare (uuid, owner); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "Cluster lock held by %s" - " ,unlock req from %s!", uuid_utoa_r (owner ,owner_str) + gf_log (this->name, GF_LOG_ERROR, "Cluster lock held by %s ," + "unlock req from %s!", uuid_utoa_r (owner ,owner_str) , uuid_utoa_r (uuid, new_owner_str)); goto out; } @@ -296,7 +199,7 @@ glusterd_unlock (uuid_t uuid) ret = glusterd_unset_lock_owner (uuid); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "Unable to clear cluster " + gf_log (this->name, GF_LOG_ERROR, "Unable to clear cluster " "lock"); goto out; } @@ -317,7 +220,7 @@ glusterd_get_uuid (uuid_t *uuid) GF_ASSERT (priv); - uuid_copy (*uuid, priv->uuid); + uuid_copy (*uuid, MY_UUID); return 0; } @@ -510,7 +413,7 @@ glusterd_check_volume_exists (char *volname) ret = stat (pathname, &stbuf); if (ret) { - gf_log ("", GF_LOG_DEBUG, "Volume %s does not exist." + gf_log (THIS->name, GF_LOG_DEBUG, "Volume %s does not exist." "stat failed with errno : %d on path: %s", volname, errno, pathname); return _gf_false; @@ -533,25 +436,30 @@ glusterd_volinfo_new (glusterd_volinfo_t **volinfo) if (!new_volinfo) goto out; + LOCK_INIT (&new_volinfo->lock); INIT_LIST_HEAD (&new_volinfo->vol_list); + INIT_LIST_HEAD (&new_volinfo->snapvol_list); INIT_LIST_HEAD (&new_volinfo->bricks); + INIT_LIST_HEAD (&new_volinfo->snap_volumes); new_volinfo->dict = dict_new (); if (!new_volinfo->dict) { - if (new_volinfo) - GF_FREE (new_volinfo); + GF_FREE (new_volinfo); goto out; } new_volinfo->gsync_slaves = dict_new (); if (!new_volinfo->gsync_slaves) { - if (new_volinfo) - GF_FREE (new_volinfo); + GF_FREE (new_volinfo); goto out; } + snprintf (new_volinfo->parent_volname, GLUSTERD_MAX_VOLUME_NAME, "N/A"); + + new_volinfo->snap_max_hard_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT; + new_volinfo->xl = THIS; *volinfo = new_volinfo; @@ -559,10 +467,269 @@ glusterd_volinfo_new (glusterd_volinfo_t **volinfo) ret = 0; out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +/* This function will create a new volinfo and then + * dup the entries from volinfo to the new_volinfo. + * + * @param volinfo volinfo which will be duplicated + * @param dup_volinfo new volinfo which will be created + * @param set_userauth if this true then auth info is also set + * + * @return 0 on success else -1 + */ +int32_t +glusterd_volinfo_dup (glusterd_volinfo_t *volinfo, + glusterd_volinfo_t **dup_volinfo, + gf_boolean_t set_userauth) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_volinfo_t *new_volinfo = NULL; + + this = THIS; + GF_ASSERT (this); + GF_VALIDATE_OR_GOTO (this->name, volinfo, out); + GF_VALIDATE_OR_GOTO (this->name, dup_volinfo, out); + + ret = glusterd_volinfo_new (&new_volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "not able to create the " + "duplicate volinfo for the volume %s", + volinfo->volname); + goto out; + } + + new_volinfo->type = volinfo->type; + new_volinfo->replica_count = volinfo->replica_count; + new_volinfo->stripe_count = volinfo->stripe_count; + new_volinfo->dist_leaf_count = volinfo->dist_leaf_count; + new_volinfo->sub_count = volinfo->sub_count; + new_volinfo->transport_type = volinfo->transport_type; + new_volinfo->nfs_transport_type = volinfo->nfs_transport_type; + new_volinfo->brick_count = volinfo->brick_count; + + dict_copy (volinfo->dict, new_volinfo->dict); + gd_update_volume_op_versions (new_volinfo); + + if (set_userauth) { + glusterd_auth_set_username (new_volinfo, + volinfo->auth.username); + glusterd_auth_set_password (new_volinfo, + volinfo->auth.password); + } + + *dup_volinfo = new_volinfo; + ret = 0; +out: + if (ret && (NULL != new_volinfo)) { + (void) glusterd_volinfo_delete (new_volinfo); + } return ret; } +/* This function will duplicate brickinfo + * + * @param brickinfo Source brickinfo + * @param dup_brickinfo Destination brickinfo + * + * @return 0 on success else -1 + */ +int32_t +glusterd_brickinfo_dup (glusterd_brickinfo_t *brickinfo, + glusterd_brickinfo_t *dup_brickinfo) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + GF_VALIDATE_OR_GOTO (this->name, brickinfo, out); + GF_VALIDATE_OR_GOTO (this->name, dup_brickinfo, out); + + strcpy (dup_brickinfo->hostname, brickinfo->hostname); + strcpy (dup_brickinfo->path, brickinfo->path); + strcpy (dup_brickinfo->device_path, brickinfo->device_path); + ret = gf_canonicalize_path (dup_brickinfo->path); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to canonicalize " + "brick path"); + goto out; + } + uuid_copy (dup_brickinfo->uuid, brickinfo->uuid); + + dup_brickinfo->port = brickinfo->port; + dup_brickinfo->rdma_port = brickinfo->rdma_port; + if (NULL != brickinfo->logfile) { + dup_brickinfo->logfile = gf_strdup (brickinfo->logfile); + if (NULL == dup_brickinfo->logfile) { + ret = -1; + goto out; + } + } + dup_brickinfo->status = brickinfo->status; + dup_brickinfo->snap_status = brickinfo->snap_status; +out: + return ret; +} + +/* This function will copy snap volinfo to the new + * passed volinfo and regenerate backend store files + * for the restored snap. + * + * @param new_volinfo new volinfo + * @param snap_volinfo volinfo of snap volume + * + * @return 0 on success and -1 on failure + * + * TODO: Duplicate all members of volinfo, e.g. geo-rep sync slaves + */ +int32_t +glusterd_snap_volinfo_restore (dict_t *rsp_dict, + glusterd_volinfo_t *new_volinfo, + glusterd_volinfo_t *snap_volinfo) +{ + int32_t brick_count = -1; + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_brickinfo_t *new_brickinfo = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + + GF_VALIDATE_OR_GOTO (this->name, new_volinfo, out); + GF_VALIDATE_OR_GOTO (this->name, snap_volinfo, out); + + brick_count = 0; + list_for_each_entry (brickinfo, &snap_volinfo->bricks, brick_list) { + ret = glusterd_brickinfo_new (&new_brickinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to create " + "new brickinfo"); + goto out; + } + + /* Duplicate brickinfo */ + ret = glusterd_brickinfo_dup (brickinfo, new_brickinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to dup " + "brickinfo"); + goto out; + } + + /* If the brick is not of this peer, or snapshot is missed * + * for the brick do not replace the xattr for it */ + if ((!uuid_compare (brickinfo->uuid, MY_UUID)) && + (brickinfo->snap_status != -1)) { + /* We need to replace the volume id of all the bricks + * to the volume id of the origin volume. new_volinfo + * has the origin volume's volume id*/ + ret = sys_lsetxattr (new_brickinfo->path, + GF_XATTR_VOL_ID_KEY, + new_volinfo->volume_id, + sizeof (new_volinfo->volume_id), + XATTR_REPLACE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "set extended attribute %s on %s. " + "Reason: %s, snap: %s", + GF_XATTR_VOL_ID_KEY, + new_brickinfo->path, strerror (errno), + new_volinfo->volname); + goto out; + } + } + + /* If a snapshot is pending for this brick then + * restore should also be pending + */ + if (brickinfo->snap_status == -1) { + /* Adding missed delete to the dict */ + ret = glusterd_add_missed_snaps_to_dict + (rsp_dict, + snap_volinfo->volname, + brickinfo, + brick_count + 1, + GF_SNAP_OPTION_TYPE_RESTORE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add missed snapshot info " + "for %s:%s in the rsp_dict", + brickinfo->hostname, + brickinfo->path); + goto out; + } + } + + list_add_tail (&new_brickinfo->brick_list, + &new_volinfo->bricks); + /* ownership of new_brickinfo is passed to new_volinfo */ + new_brickinfo = NULL; + brick_count++; + } + + /* Regenerate all volfiles */ + ret = glusterd_create_volfiles_and_notify_services (new_volinfo); + +out: + if (ret && (NULL != new_brickinfo)) { + (void) glusterd_brickinfo_delete (new_brickinfo); + } + + return ret; +} + +void +glusterd_auth_cleanup (glusterd_volinfo_t *volinfo) { + + GF_ASSERT (volinfo); + + GF_FREE (volinfo->auth.username); + + GF_FREE (volinfo->auth.password); +} + +char * +glusterd_auth_get_username (glusterd_volinfo_t *volinfo) { + + GF_ASSERT (volinfo); + + return volinfo->auth.username; +} + +char * +glusterd_auth_get_password (glusterd_volinfo_t *volinfo) { + + GF_ASSERT (volinfo); + + return volinfo->auth.password; +} + +int32_t +glusterd_auth_set_username (glusterd_volinfo_t *volinfo, char *username) { + + GF_ASSERT (volinfo); + GF_ASSERT (username); + + volinfo->auth.username = gf_strdup (username); + return 0; +} + +int32_t +glusterd_auth_set_password (glusterd_volinfo_t *volinfo, char *password) { + + GF_ASSERT (volinfo); + GF_ASSERT (password); + + volinfo->auth.password = gf_strdup (password); + return 0; +} + int32_t glusterd_brickinfo_delete (glusterd_brickinfo_t *brickinfo) { @@ -572,8 +739,7 @@ glusterd_brickinfo_delete (glusterd_brickinfo_t *brickinfo) list_del_init (&brickinfo->brick_list); - if (brickinfo->logfile) - GF_FREE (brickinfo->logfile); + GF_FREE (brickinfo->logfile); GF_FREE (brickinfo); ret = 0; @@ -598,7 +764,7 @@ glusterd_volume_brickinfos_delete (glusterd_volinfo_t *volinfo) } out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -610,6 +776,7 @@ glusterd_volinfo_delete (glusterd_volinfo_t *volinfo) GF_ASSERT (volinfo); list_del_init (&volinfo->vol_list); + list_del_init (&volinfo->snapvol_list); ret = glusterd_volume_brickinfos_delete (volinfo); if (ret) @@ -618,18 +785,18 @@ glusterd_volinfo_delete (glusterd_volinfo_t *volinfo) dict_unref (volinfo->dict); if (volinfo->gsync_slaves) dict_unref (volinfo->gsync_slaves); - if (volinfo->logdir) - GF_FREE (volinfo->logdir); + GF_FREE (volinfo->logdir); + + glusterd_auth_cleanup (volinfo); GF_FREE (volinfo); ret = 0; out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } - int32_t glusterd_brickinfo_new (glusterd_brickinfo_t **brickinfo) { @@ -651,7 +818,7 @@ glusterd_brickinfo_new (glusterd_brickinfo_t **brickinfo) ret = 0; out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -659,17 +826,21 @@ int32_t glusterd_resolve_brick (glusterd_brickinfo_t *brickinfo) { int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (brickinfo); ret = glusterd_hostname_to_uuid (brickinfo->hostname, brickinfo->uuid); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } int32_t -glusterd_brickinfo_from_brick (char *brick, - glusterd_brickinfo_t **brickinfo) +glusterd_brickinfo_new_from_brick (char *brick, + glusterd_brickinfo_t **brickinfo) { int32_t ret = -1; glusterd_brickinfo_t *new_brickinfo = NULL; @@ -677,6 +848,7 @@ glusterd_brickinfo_from_brick (char *brick, char *path = NULL; char *tmp_host = NULL; char *tmp_path = NULL; + char *vg = NULL; GF_ASSERT (brick); GF_ASSERT (brickinfo); @@ -692,7 +864,21 @@ glusterd_brickinfo_from_brick (char *brick, GF_ASSERT (path); ret = glusterd_brickinfo_new (&new_brickinfo); + if (ret) + goto out; +#ifdef HAVE_BD_XLATOR + vg = strchr (path, '?'); + /* ? is used as a delimiter for vg */ + if (vg) { + strncpy (new_brickinfo->vg, vg + 1, PATH_MAX - 1); + *vg = '\0'; + } + new_brickinfo->caps = CAPS_BD; +#else + vg = NULL; /* Avoid compiler warnings when BD not enabled */ +#endif + ret = gf_canonicalize_path (path); if (ret) goto out; @@ -703,23 +889,272 @@ glusterd_brickinfo_from_brick (char *brick, ret = 0; out: - if (tmp_host) - GF_FREE (tmp_host); + GF_FREE (tmp_host); if (tmp_host) GF_FREE (tmp_path); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +static gf_boolean_t +_is_prefix (char *str1, char *str2) +{ + GF_ASSERT (str1); + GF_ASSERT (str2); + + int i = 0; + int len1 = 0; + int len2 = 0; + int small_len = 0; + char *bigger = NULL; + gf_boolean_t prefix = _gf_true; + + len1 = strlen (str1); + len2 = strlen (str2); + small_len = min (len1, len2); + for (i = 0; i < small_len; i++) { + if (str1[i] != str2[i]) { + prefix = _gf_false; + break; + } + } + + if (len1 < len2) + bigger = str2; + + else if (len1 > len2) + bigger = str1; + + else + return prefix; + + if (bigger[small_len] != '/') + prefix = _gf_false; + + return prefix; +} + +/* Checks if @path is available in the peer identified by @uuid + * 'availability' is determined by querying current state of volumes + * in the cluster. */ +gf_boolean_t +glusterd_is_brickpath_available (uuid_t uuid, char *path) +{ + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_volinfo_t *volinfo = NULL; + glusterd_conf_t *priv = NULL; + gf_boolean_t available = _gf_false; + char tmp_path[PATH_MAX+1] = {0}; + char tmp_brickpath[PATH_MAX+1] = {0}; + + priv = THIS->private; + + strncpy (tmp_path, path, PATH_MAX); + /* path may not yet exist */ + if (!realpath (path, tmp_path)) { + if (errno != ENOENT) { + goto out; + } + /* When realpath(3) fails, tmp_path is undefined. */ + strncpy(tmp_path,path,PATH_MAX); + } + + list_for_each_entry (volinfo, &priv->volumes, vol_list) { + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (uuid_compare (uuid, brickinfo->uuid)) + continue; + + if (!realpath (brickinfo->path, tmp_brickpath)) { + if (errno == ENOENT) + strncpy (tmp_brickpath, brickinfo->path, + PATH_MAX); + else + goto out; + } + + if (_is_prefix (tmp_brickpath, tmp_path)) + goto out; + } + } + available = _gf_true; +out: + return available; +} + +#ifdef HAVE_BD_XLATOR +/* + * Sets the tag of the format "trusted.glusterfs.volume-id:<uuid>" in + * the brick VG. It is used to avoid using same VG for another brick. + * @volume-id - gfid, @brick - brick info, @msg - Error message returned + * to the caller + */ +int +glusterd_bd_set_vg_tag (unsigned char *volume_id, glusterd_brickinfo_t *brick, + char *msg, int msg_size) +{ + lvm_t handle = NULL; + vg_t vg = NULL; + char *uuid = NULL; + int ret = -1; + + gf_asprintf (&uuid, "%s:%s", GF_XATTR_VOL_ID_KEY, + uuid_utoa (volume_id)); + if (!uuid) { + snprintf (msg, sizeof(*msg), "Could not allocate memory " + "for tag"); + return -1; + } + + handle = lvm_init (NULL); + if (!handle) { + snprintf (msg, sizeof(*msg), "lvm_init failed"); + goto out; + } + + vg = lvm_vg_open (handle, brick->vg, "w", 0); + if (!vg) { + snprintf (msg, sizeof(*msg), "Could not open VG %s", + brick->vg); + goto out; + } + + if (lvm_vg_add_tag (vg, uuid) < 0) { + snprintf (msg, sizeof(*msg), "Could not set tag %s for " + "VG %s", uuid, brick->vg); + goto out; + } + lvm_vg_write (vg); + ret = 0; +out: + GF_FREE (uuid); + + if (vg) + lvm_vg_close (vg); + if (handle) + lvm_quit (handle); + + return ret; +} +#endif + +int +glusterd_validate_and_create_brickpath (glusterd_brickinfo_t *brickinfo, + uuid_t volume_id, char **op_errstr, + gf_boolean_t is_force) +{ + int ret = -1; + char parentdir[PATH_MAX] = {0,}; + struct stat parent_st = {0,}; + struct stat brick_st = {0,}; + struct stat root_st = {0,}; + char msg[2048] = {0,}; + gf_boolean_t is_created = _gf_false; + + ret = mkdir (brickinfo->path, 0777); + if (ret) { + if (errno != EEXIST) { + snprintf (msg, sizeof (msg), "Failed to create brick " + "directory for brick %s:%s. Reason : %s ", + brickinfo->hostname, brickinfo->path, + strerror (errno)); + goto out; + } + } else { + is_created = _gf_true; + } + + ret = lstat (brickinfo->path, &brick_st); + if (ret) { + snprintf (msg, sizeof (msg), "lstat failed on %s. Reason : %s", + brickinfo->path, strerror (errno)); + goto out; + } + + if ((!is_created) && (!S_ISDIR (brick_st.st_mode))) { + snprintf (msg, sizeof (msg), "The provided path %s which is " + "already present, is not a directory", + brickinfo->path); + ret = -1; + goto out; + } + + snprintf (parentdir, sizeof (parentdir), "%s/..", brickinfo->path); + + ret = lstat ("/", &root_st); + if (ret) { + snprintf (msg, sizeof (msg), "lstat failed on /. Reason : %s", + strerror (errno)); + goto out; + } + + ret = lstat (parentdir, &parent_st); + if (ret) { + snprintf (msg, sizeof (msg), "lstat failed on %s. Reason : %s", + parentdir, strerror (errno)); + goto out; + } + + if (!is_force) { + if (brick_st.st_dev != parent_st.st_dev) { + snprintf (msg, sizeof (msg), "The brick %s:%s is a " + "mount point. Please create a sub-directory " + "under the mount point and use that as the " + "brick directory. Or use 'force' at the end " + "of the command if you want to override this " + "behavior.", brickinfo->hostname, + brickinfo->path); + ret = -1; + goto out; + } + else if (parent_st.st_dev == root_st.st_dev) { + snprintf (msg, sizeof (msg), "The brick %s:%s is " + "is being created in the root partition. It " + "is recommended that you don't use the " + "system's root partition for storage backend." + " Or use 'force' at the end of the command if" + " you want to override this behavior.", + brickinfo->hostname, brickinfo->path); + ret = -1; + goto out; + } + } + +#ifdef HAVE_BD_XLATOR + if (brickinfo->vg[0]) { + ret = glusterd_bd_set_vg_tag (volume_id, brickinfo, msg, + sizeof(msg)); + if (ret) + goto out; + } +#endif + ret = glusterd_check_and_set_brick_xattr (brickinfo->hostname, + brickinfo->path, volume_id, + op_errstr, is_force); + if (ret) + goto out; + + ret = 0; + +out: + if (ret && is_created) + rmdir (brickinfo->path); + if (ret && !*op_errstr && msg[0] != '\0') + *op_errstr = gf_strdup (msg); + return ret; } int32_t glusterd_volume_brickinfo_get (uuid_t uuid, char *hostname, char *path, glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t **brickinfo, - gf_path_match_t path_match) + glusterd_brickinfo_t **brickinfo) { glusterd_brickinfo_t *brickiter = NULL; uuid_t peer_uuid = {0}; int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; if (uuid) { uuid_copy (peer_uuid, uuid); @@ -731,77 +1166,48 @@ glusterd_volume_brickinfo_get (uuid_t uuid, char *hostname, char *path, ret = -1; list_for_each_entry (brickiter, &volinfo->bricks, brick_list) { - if (uuid_is_null (brickiter->uuid) && - glusterd_resolve_brick (brickiter)) + if ((uuid_is_null (brickiter->uuid)) && + (glusterd_resolve_brick (brickiter) != 0)) goto out; if (uuid_compare (peer_uuid, brickiter->uuid)) continue; - if (!strcmp (brickiter->path, path)) { - gf_log (THIS->name, GF_LOG_INFO, "Found brick"); + if (strcmp (brickiter->path, path) == 0) { + gf_log (this->name, GF_LOG_DEBUG, LOGSTR_FOUND_BRICK, + brickiter->hostname, brickiter->path, + volinfo->volname); ret = 0; if (brickinfo) *brickinfo = brickiter; break; } - if (path_match != GF_PATH_PARTIAL) - continue; - - if (!fnmatch (path, brickiter->path, FNM_LEADING_DIR) || - !fnmatch (brickiter->path, path, FNM_LEADING_DIR)) { - gf_log (THIS->name, GF_LOG_ERROR, - "paths %s and %s are recursive", - path, brickiter->path); - *brickinfo = brickiter; - ret = 0; - break; - } } out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } int32_t glusterd_volume_brickinfo_get_by_brick (char *brick, glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t **brickinfo, - gf_path_match_t path_match) + glusterd_brickinfo_t **brickinfo) { int32_t ret = -1; - char *hostname = NULL; - char *path = NULL; - char *tmp_host = NULL; - char *tmp_path = NULL; + glusterd_brickinfo_t *tmp_brickinfo = NULL; GF_ASSERT (brick); GF_ASSERT (volinfo); - gf_log ("", GF_LOG_INFO, "brick: %s", brick); - - tmp_host = gf_strdup (brick); - if (tmp_host) - get_host_name (tmp_host, &hostname); - tmp_path = gf_strdup (brick); - if (tmp_path) - get_path_name (tmp_path, &path); - - if (!hostname || !path) { - gf_log ("", GF_LOG_ERROR, - "brick %s is not of form <HOSTNAME>:<export-dir>", - brick); - ret = -1; + ret = glusterd_brickinfo_new_from_brick (brick, &tmp_brickinfo); + if (ret) goto out; - } - ret = glusterd_volume_brickinfo_get (NULL, hostname, path, volinfo, - brickinfo, path_match); + ret = glusterd_volume_brickinfo_get (NULL, tmp_brickinfo->hostname, + tmp_brickinfo->path, volinfo, + brickinfo); + (void) glusterd_brickinfo_delete (tmp_brickinfo); out: - if (tmp_host) - GF_FREE (tmp_host); - if (tmp_path) - GF_FREE (tmp_path); gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -815,7 +1221,7 @@ glusterd_is_brick_decommissioned (glusterd_volinfo_t *volinfo, char *hostname, int ret = -1; ret = glusterd_volume_brickinfo_get (NULL, hostname, path, volinfo, - &brickinfo, GF_PATH_COMPLETE); + &brickinfo); if (ret) goto out; decommissioned = brickinfo->decommissioned; @@ -828,20 +1234,95 @@ glusterd_friend_cleanup (glusterd_peerinfo_t *peerinfo) { GF_ASSERT (peerinfo); glusterd_peerctx_t *peerctx = NULL; + gf_boolean_t quorum_action = _gf_false; + glusterd_conf_t *priv = THIS->private; + if (peerinfo->quorum_contrib != QUORUM_NONE) + quorum_action = _gf_true; if (peerinfo->rpc) { + /* cleanup the saved-frames before last unref */ + synclock_unlock (&priv->big_lock); + rpc_clnt_connection_cleanup (&peerinfo->rpc->conn); + synclock_lock (&priv->big_lock); + peerctx = peerinfo->rpc->mydata; peerinfo->rpc->mydata = NULL; peerinfo->rpc = rpc_clnt_unref (peerinfo->rpc); peerinfo->rpc = NULL; - if (peerctx) + if (peerctx) { + GF_FREE (peerctx->errstr); GF_FREE (peerctx); + } } glusterd_peer_destroy (peerinfo); + if (quorum_action) + glusterd_do_quorum_action (); return 0; } +int +glusterd_volinfo_find_by_volume_id (uuid_t volume_id, glusterd_volinfo_t **volinfo) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_volinfo_t *voliter = NULL; + glusterd_conf_t *priv = NULL; + + if (!volume_id) + return -1; + + this = THIS; + priv = this->private; + + list_for_each_entry (voliter, &priv->volumes, vol_list) { + if (uuid_compare (volume_id, voliter->volume_id)) + continue; + *volinfo = voliter; + ret = 0; + gf_log (this->name, GF_LOG_DEBUG, "Volume %s found", + voliter->volname); + break; + } + return ret; +} + +int +glusterd_snap_volinfo_find_by_volume_id (uuid_t volume_id, + glusterd_volinfo_t **volinfo) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_volinfo_t *voliter = NULL; + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (volinfo); + + if (uuid_is_null(volume_id)) { + gf_log (this->name, GF_LOG_WARNING, "Volume UUID is NULL"); + goto out; + } + + list_for_each_entry (snap, &priv->snapshots, snap_list) { + list_for_each_entry (voliter, &snap->volumes, vol_list) { + if (uuid_compare (volume_id, voliter->volume_id)) + continue; + *volinfo = voliter; + ret = 0; + goto out; + } + } + + gf_log (this->name, GF_LOG_WARNING, "Snap volume not found"); +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + int32_t glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo) { @@ -851,134 +1332,149 @@ glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo) glusterd_conf_t *priv = NULL; GF_ASSERT (volname); - this = THIS; GF_ASSERT (this); priv = this->private; + GF_ASSERT (priv); list_for_each_entry (tmp_volinfo, &priv->volumes, vol_list) { if (!strcmp (tmp_volinfo->volname, volname)) { - gf_log ("", GF_LOG_DEBUG, "Volume %s found", volname); + gf_log (this->name, GF_LOG_DEBUG, "Volume %s found", + volname); ret = 0; *volinfo = tmp_volinfo; break; } } - - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } - int32_t -glusterd_service_stop (const char *service, char *pidfile, int sig, - gf_boolean_t force_kill) +glusterd_snap_volinfo_find (char *snap_volname, glusterd_snap_t *snap, + glusterd_volinfo_t **volinfo) { - int32_t ret = -1; - pid_t pid = -1; - FILE *file = NULL; - gf_boolean_t is_locked = _gf_false; + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_conf_t *priv = NULL; - file = fopen (pidfile, "r+"); + this = THIS; + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (snap); + GF_ASSERT (snap_volname); - if (!file) { - gf_log ("", GF_LOG_ERROR, "Unable to open pidfile: %s", - pidfile); - if (errno == ENOENT) { - gf_log ("",GF_LOG_TRACE, "%s may not be running", - service); + list_for_each_entry (snap_vol, &snap->volumes, vol_list) { + if (!strcmp (snap_vol->volname, snap_volname)) { ret = 0; + *volinfo = snap_vol; goto out; } - ret = -1; - goto out; } - ret = lockf (fileno (file), F_TLOCK, 0); - if (!ret) { - is_locked = _gf_true; - ret = unlink (pidfile); - if (ret && (ENOENT != errno)) { - gf_log ("", GF_LOG_ERROR, "Unable to " - "unlink stale pidfile: %s", pidfile); - } else if (ret && (ENOENT == errno)){ + + gf_log (this->name, GF_LOG_WARNING, "Snap volume %s not found", + snap_volname); +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_snap_volinfo_find_from_parent_volname (char *origin_volname, + glusterd_snap_t *snap, + glusterd_volinfo_t **volinfo) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_conf_t *priv = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (snap); + GF_ASSERT (origin_volname); + + list_for_each_entry (snap_vol, &snap->volumes, vol_list) { + if (!strcmp (snap_vol->parent_volname, origin_volname)) { ret = 0; - gf_log ("", GF_LOG_INFO, "Brick already stopped"); + *volinfo = snap_vol; + goto out; } - goto out; } + gf_log (this->name, GF_LOG_DEBUG, "Snap volume not found(snap: %s, " + "origin-volume: %s", snap->snapname, origin_volname); - ret = fscanf (file, "%d", &pid); - if (ret <= 0) { - gf_log ("", GF_LOG_ERROR, "Unable to read pidfile: %s", - pidfile); - ret = -1; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_service_stop (const char *service, char *pidfile, int sig, + gf_boolean_t force_kill) +{ + int32_t ret = -1; + pid_t pid = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + if (!glusterd_is_service_running (pidfile, &pid)) { + ret = 0; + gf_log (this->name, GF_LOG_INFO, "%s already stopped", service); goto out; } - fclose (file); - file = NULL; - - gf_log ("", GF_LOG_INFO, "Stopping gluster %s running in pid: %d", - service, pid); + gf_log (this->name, GF_LOG_DEBUG, "Stopping gluster %s running in pid: " + "%d", service, pid); ret = kill (pid, sig); + if (!force_kill) + goto out; - if (force_kill) { - sleep (1); - file = fopen (pidfile, "r+"); - if (!file) { - ret = 0; - goto out; - } - ret = lockf (fileno (file), F_TLOCK, 0); - if (ret && ((EAGAIN == errno) || (EACCES == errno))) { - ret = kill (pid, SIGKILL); - if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to " - "kill pid %d reason: %s", pid, - strerror(errno)); - goto out; - } - - } else if (0 == ret){ - is_locked = _gf_true; - } - ret = unlink (pidfile); - if (ret && (ENOENT != errno)) { - gf_log ("", GF_LOG_ERROR, "Unable to " - "unlink pidfile: %s", pidfile); + sleep (1); + if (glusterd_is_service_running (pidfile, NULL)) { + ret = kill (pid, SIGKILL); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "kill pid %d reason: %s", pid, + strerror(errno)); goto out; } } ret = 0; out: - if (is_locked && file) - if (lockf (fileno (file), F_ULOCK, 0) < 0) - gf_log ("", GF_LOG_WARNING, "Cannot unlock pidfile: %s" - " reason: %s", pidfile, strerror(errno)); - if (file) - fclose (file); return ret; } void +glusterd_set_socket_filepath (char *sock_filepath, char *sockpath, size_t len) +{ + char md5_sum[MD5_DIGEST_LENGTH*2+1] = {0,}; + + md5_wrapper ((unsigned char *) sock_filepath, strlen(sock_filepath), md5_sum); + snprintf (sockpath, len, "%s/%s.socket", GLUSTERD_SOCK_DIR, md5_sum); +} + +void glusterd_set_brick_socket_filepath (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo, char *sockpath, size_t len) { char export_path[PATH_MAX] = {0,}; char sock_filepath[PATH_MAX] = {0,}; - char md5_sum[MD5_DIGEST_LEN*2+1] = {0,}; char volume_dir[PATH_MAX] = {0,}; xlator_t *this = NULL; glusterd_conf_t *priv = NULL; int expected_file_len = 0; - expected_file_len = strlen (glusterd_sock_dir) + strlen ("/") + - MD5_DIGEST_LEN*2 + strlen (".socket") + 1; + expected_file_len = strlen (GLUSTERD_SOCK_DIR) + strlen ("/") + + MD5_DIGEST_LENGTH*2 + strlen (".socket") + 1; GF_ASSERT (len >= expected_file_len); this = THIS; GF_ASSERT (this); @@ -989,10 +1485,8 @@ glusterd_set_brick_socket_filepath (glusterd_volinfo_t *volinfo, GLUSTERD_REMOVE_SLASH_FROM_PATH (brickinfo->path, export_path); snprintf (sock_filepath, PATH_MAX, "%s/run/%s-%s", volume_dir, brickinfo->hostname, export_path); - _get_md5_str (md5_sum, sizeof (md5_sum), - (uint8_t*)sock_filepath, strlen (sock_filepath)); - snprintf (sockpath, len, "%s/%s.socket", glusterd_sock_dir, md5_sum); + glusterd_set_socket_filepath (sock_filepath, sockpath, len); } /* connection happens only if it is not aleady connected, @@ -1000,26 +1494,41 @@ glusterd_set_brick_socket_filepath (glusterd_volinfo_t *volinfo, */ int32_t glusterd_brick_connect (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo) + glusterd_brickinfo_t *brickinfo, char *socketpath) { int ret = 0; - char socketpath[PATH_MAX] = {0}; + char volume_id_str[64]; + char *brickid = NULL; dict_t *options = NULL; struct rpc_clnt *rpc = NULL; + glusterd_conf_t *priv = THIS->private; GF_ASSERT (volinfo); GF_ASSERT (brickinfo); + GF_ASSERT (socketpath); if (brickinfo->rpc == NULL) { - glusterd_set_brick_socket_filepath (volinfo, brickinfo, - socketpath, - sizeof (socketpath)); - ret = rpc_clnt_transport_unix_options_build (&options, socketpath); + /* Setting frame-timeout to 10mins (600seconds). + * Unix domain sockets ensures that the connection is reliable. + * The default timeout of 30mins used for unreliable network + * connections is too long for unix domain socket connections. + */ + ret = rpc_transport_unix_options_build (&options, socketpath, + 600); if (ret) goto out; + + uuid_utoa_r (volinfo->volume_id, volume_id_str); + ret = gf_asprintf (&brickid, "%s:%s:%s", volume_id_str, + brickinfo->hostname, brickinfo->path); + if (ret < 0) + goto out; + + synclock_unlock (&priv->big_lock); ret = glusterd_rpc_create (&rpc, options, glusterd_brick_rpc_notify, - brickinfo); + brickid); + synclock_lock (&priv->big_lock); if (ret) goto out; brickinfo->rpc = rpc; @@ -1029,28 +1538,54 @@ out: return ret; } +/* Caller should ensure that brick process is not running*/ +static void +_reap_brick_process (char *pidfile, char *brickpath) +{ + unlink (pidfile); + /* Brick process is not running and pmap may have an entry for it.*/ + pmap_registry_remove (THIS, 0, brickpath, + GF_PMAP_PORT_BRICKSERVER, NULL); +} + +static int +_mk_rundir_p (glusterd_volinfo_t *volinfo) +{ + char voldir[PATH_MAX] = {0,}; + char rundir[PATH_MAX] = {0,}; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + int ret = -1; + + this = THIS; + priv = this->private; + GLUSTERD_GET_VOLUME_DIR (voldir, volinfo, priv); + snprintf (rundir, sizeof (rundir)-1, "%s/run", voldir); + ret = mkdir_p (rundir, 0777, _gf_true); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "Failed to create rundir"); + return ret; +} + int32_t glusterd_volume_start_glusterfs (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo) + glusterd_brickinfo_t *brickinfo, + gf_boolean_t wait) { int32_t ret = -1; xlator_t *this = NULL; glusterd_conf_t *priv = NULL; - char pidfile[PATH_MAX] = {0,}; + char pidfile[PATH_MAX+1] = {0,}; char volfile[PATH_MAX] = {0,}; - char path[PATH_MAX] = {0,}; runner_t runner = {0,}; - char rundir[PATH_MAX] = {0,}; char exp_path[PATH_MAX] = {0,}; char logfile[PATH_MAX] = {0,}; int port = 0; int rdma_port = 0; - FILE *file = NULL; - gf_boolean_t is_locked = _gf_false; char socketpath[PATH_MAX] = {0}; -#ifdef DEBUG + char glusterd_uuid[1024] = {0,}; char valgrind_logfile[PATH_MAX] = {0}; -#endif + GF_ASSERT (volinfo); GF_ASSERT (brickinfo); @@ -1058,105 +1593,93 @@ glusterd_volume_start_glusterfs (glusterd_volinfo_t *volinfo, GF_ASSERT (this); priv = this->private; + GF_ASSERT (priv); - GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv); - snprintf (rundir, PATH_MAX, "%s/run", path); - ret = mkdir (rundir, 0777); - - if ((ret == -1) && (EEXIST != errno)) { - gf_log ("", GF_LOG_ERROR, "Unable to create rundir %s", - rundir); + if (brickinfo->snap_status == -1) { + gf_log (this->name, GF_LOG_INFO, + "Snapshot is pending on %s:%s. " + "Hence not starting the brick", + brickinfo->hostname, + brickinfo->path); + ret = 0; goto out; } + ret = _mk_rundir_p (volinfo); + if (ret) + goto out; + glusterd_set_brick_socket_filepath (volinfo, brickinfo, socketpath, sizeof (socketpath)); - GLUSTERD_GET_BRICK_PIDFILE (pidfile, path, brickinfo->hostname, - brickinfo->path); - - file = fopen (pidfile, "r+"); - if (file) { - ret = lockf (fileno (file), F_TLOCK, 0); - if (ret && ((EAGAIN == errno) || (EACCES == errno))) { - ret = 0; - gf_log ("", GF_LOG_INFO, "brick %s:%s " - "already started", brickinfo->hostname, - brickinfo->path); - goto connect; - } - } - ret = pmap_registry_search (this, brickinfo->path, - GF_PMAP_PORT_BRICKSERVER); - if (ret) { - ret = 0; - file = fopen (pidfile, "r+"); - if (file) { - ret = lockf (fileno (file), F_TLOCK, 0); - if (ret && ((EAGAIN == errno) || (EACCES == errno))) { - ret = 0; - gf_log ("", GF_LOG_INFO, "brick %s:%s " - "already started", brickinfo->hostname, - brickinfo->path); - goto connect; - } else if (0 == ret) { - is_locked = _gf_true; - } - } - /* This means, pmap has the entry, remove it */ - ret = pmap_registry_remove (this, 0, brickinfo->path, - GF_PMAP_PORT_BRICKSERVER, NULL); - } - unlink (pidfile); + GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, brickinfo, priv); + if (glusterd_is_service_running (pidfile, NULL)) + goto connect; - gf_log ("", GF_LOG_INFO, "About to start glusterfs" - " for brick %s:%s", brickinfo->hostname, - brickinfo->path); - GLUSTERD_REMOVE_SLASH_FROM_PATH (brickinfo->path, exp_path); - snprintf (volfile, PATH_MAX, "%s.%s.%s", volinfo->volname, - brickinfo->hostname, exp_path); - - if (!brickinfo->logfile && volinfo->logdir) { - snprintf (logfile, PATH_MAX, "%s/%s.log", volinfo->logdir, - exp_path); - brickinfo->logfile = gf_strdup (logfile); - } else if (!brickinfo->logfile) { - snprintf (logfile, PATH_MAX, "%s/bricks/%s.log", - DEFAULT_LOG_FILE_DIRECTORY, exp_path); - brickinfo->logfile = gf_strdup (logfile); - } + _reap_brick_process (pidfile, brickinfo->path); port = brickinfo->port; if (!port) port = pmap_registry_alloc (THIS); + /* Build the exp_path, before starting the glusterfsd even in + valgrind mode. Otherwise all the glusterfsd processes start + writing the valgrind log to the same file. + */ + GLUSTERD_REMOVE_SLASH_FROM_PATH (brickinfo->path, exp_path); runinit (&runner); -#ifdef DEBUG if (priv->valgrind) { + /* Run bricks with valgrind */ if (volinfo->logdir) { snprintf (valgrind_logfile, PATH_MAX, - "%s/valgrind-%s-%s.log", volinfo->logdir, + "%s/valgrind-%s-%s.log", + volinfo->logdir, volinfo->volname, exp_path); } else { - snprintf (valgrind_logfile, PATH_MAX, - "%s/bricks/valgrnd-%s-%s.log", - DEFAULT_LOG_FILE_DIRECTORY, - volinfo->volname, exp_path); + snprintf (valgrind_logfile, PATH_MAX, + "%s/bricks/valgrind-%s-%s.log", + DEFAULT_LOG_FILE_DIRECTORY, + volinfo->volname, exp_path); } - /* Run bricks with valgrind */ + runner_add_args (&runner, "valgrind", "--leak-check=full", - "--trace-children=yes", NULL); + "--trace-children=yes", "--track-origins=yes", + NULL); runner_argprintf (&runner, "--log-file=%s", valgrind_logfile); - } -#endif - runner_add_args (&runner, SBIN_DIR"/glusterfsd", - "-s", "localhost", "--volfile-id", volfile, + } + + if (volinfo->is_snap_volume) { + snprintf (volfile, PATH_MAX,"/%s/%s/%s.%s.%s", + GLUSTERD_VOL_SNAP_DIR_PREFIX, + volinfo->snapshot->snapname, volinfo->volname, + brickinfo->hostname, exp_path); + } else { + snprintf (volfile, PATH_MAX, "%s.%s.%s", volinfo->volname, + brickinfo->hostname, exp_path); + } + + if (volinfo->logdir) { + snprintf (logfile, PATH_MAX, "%s/%s.log", + volinfo->logdir, exp_path); + } else { + snprintf (logfile, PATH_MAX, "%s/bricks/%s.log", + DEFAULT_LOG_FILE_DIRECTORY, exp_path); + } + if (!brickinfo->logfile) + brickinfo->logfile = gf_strdup (logfile); + + (void) snprintf (glusterd_uuid, 1024, "*-posix.glusterd-uuid=%s", + uuid_utoa (MY_UUID)); + runner_add_args (&runner, SBIN_DIR"/glusterfsd", + "-s", brickinfo->hostname, "--volfile-id", volfile, "-p", pidfile, "-S", socketpath, "--brick-name", brickinfo->path, - "-l", brickinfo->logfile, NULL); + "-l", brickinfo->logfile, + "--xlator-option", glusterd_uuid, + NULL); - runner_add_arg (&runner, "--brick-port"); + runner_add_arg (&runner, "--brick-port"); if (volinfo->transport_type != GF_TRANSPORT_BOTH_TCP_RDMA) { runner_argprintf (&runner, "%d", port); } else { @@ -1173,26 +1696,34 @@ glusterd_volume_start_glusterfs (glusterd_volinfo_t *volinfo, runner_argprintf (&runner, "%s-server.listen-port=%d", volinfo->volname, port); + if (volinfo->memory_accounting) + runner_add_arg (&runner, "--mem-accounting"); + runner_log (&runner, "", GF_LOG_DEBUG, "Starting GlusterFS"); - ret = runner_run (&runner); + if (wait) { + synclock_unlock (&priv->big_lock); + ret = runner_run (&runner); + synclock_lock (&priv->big_lock); - if (ret == 0) { - //pmap_registry_bind (THIS, port, brickinfo->path); - brickinfo->port = port; - brickinfo->rdma_port = rdma_port; + } else { + ret = runner_run_nowait (&runner); } -connect: - ret = glusterd_brick_connect (volinfo, brickinfo); if (ret) goto out; + + brickinfo->port = port; + brickinfo->rdma_port = rdma_port; + +connect: + ret = glusterd_brick_connect (volinfo, brickinfo, socketpath); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to connect to brick %s:%s on %s", + brickinfo->hostname, brickinfo->path, socketpath); + goto out; + } out: - if (is_locked && file) - if (lockf (fileno (file), F_ULOCK, 0) < 0) - gf_log ("", GF_LOG_WARNING, "Cannot unlock pidfile: %s" - " reason: %s", pidfile, strerror(errno)); - if (file) - fclose (file); return ret; } @@ -1220,7 +1751,7 @@ glusterd_brick_unlink_socket_file (glusterd_volinfo_t *volinfo, if (ret && (ENOENT == errno)) { ret = 0; } else { - gf_log ("glusterd", GF_LOG_ERROR, "Failed to remove %s" + gf_log (this->name, GF_LOG_ERROR, "Failed to remove %s" " error: %s", socketpath, strerror (errno)); } @@ -1230,24 +1761,37 @@ glusterd_brick_unlink_socket_file (glusterd_volinfo_t *volinfo, int32_t glusterd_brick_disconnect (glusterd_brickinfo_t *brickinfo) { + rpc_clnt_t *rpc = NULL; + glusterd_conf_t *priv = THIS->private; + GF_ASSERT (brickinfo); - if (brickinfo->rpc) { - rpc_clnt_unref (brickinfo->rpc); - brickinfo->rpc = NULL; + if (!brickinfo) { + gf_log_callingfn ("glusterd", GF_LOG_WARNING, "!brickinfo"); + return -1; } + + rpc = brickinfo->rpc; + brickinfo->rpc = NULL; + + if (rpc) { + synclock_unlock (&priv->big_lock); + rpc_clnt_unref (rpc); + synclock_lock (&priv->big_lock); + } + return 0; } int32_t glusterd_volume_stop_glusterfs (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo) + glusterd_brickinfo_t *brickinfo, + gf_boolean_t del_brick) { - xlator_t *this = NULL; - glusterd_conf_t *priv = NULL; - char pidfile[PATH_MAX] = {0,}; - char path[PATH_MAX] = {0,}; - int ret = 0; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + char pidfile[PATH_MAX] = {0,}; + int ret = 0; GF_ASSERT (volinfo); GF_ASSERT (brickinfo); @@ -1256,17 +1800,22 @@ glusterd_volume_stop_glusterfs (glusterd_volinfo_t *volinfo, GF_ASSERT (this); priv = this->private; - (void) glusterd_brick_disconnect (brickinfo); - - GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv); - GLUSTERD_GET_BRICK_PIDFILE (pidfile, path, brickinfo->hostname, - brickinfo->path); + if (del_brick) + list_del_init (&brickinfo->brick_list); - ret = glusterd_service_stop ("brick", pidfile, SIGTERM, _gf_false); - if (ret == 0) { - glusterd_set_brick_status (brickinfo, GF_BRICK_STOPPED); - (void) glusterd_brick_unlink_socket_file (volinfo, brickinfo); + if (GLUSTERD_STATUS_STARTED == volinfo->status) { + (void) glusterd_brick_disconnect (brickinfo); + GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, brickinfo, priv); + ret = glusterd_service_stop ("brick", pidfile, SIGTERM, _gf_false); + if (ret == 0) { + glusterd_set_brick_status (brickinfo, GF_BRICK_STOPPED); + (void) glusterd_brick_unlink_socket_file (volinfo, brickinfo); + } } + + if (del_brick) + glusterd_delete_brick (volinfo, brickinfo); + return ret; } @@ -1296,6 +1845,114 @@ out: return ret; } +/* Free LINE[0..N-1] and then the LINE buffer. */ +static void +free_lines (char **line, size_t n) +{ + size_t i; + for (i = 0; i < n; i++) + GF_FREE (line[i]); + GF_FREE (line); +} + +char ** +glusterd_readin_file (const char *filepath, int *line_count) +{ + int ret = -1; + int n = 8; + int counter = 0; + char buffer[PATH_MAX + 256] = {0}; + char **lines = NULL; + FILE *fp = NULL; + void *p; + + fp = fopen (filepath, "r"); + if (!fp) + goto out; + + lines = GF_CALLOC (1, n * sizeof (*lines), gf_gld_mt_charptr); + if (!lines) + goto out; + + for (counter = 0; fgets (buffer, sizeof (buffer), fp); counter++) { + + if (counter == n-1) { + n *= 2; + p = GF_REALLOC (lines, n * sizeof (char *)); + if (!p) { + free_lines (lines, n/2); + lines = NULL; + goto out; + } + lines = p; + } + + lines[counter] = gf_strdup (buffer); + } + + lines[counter] = NULL; + /* Reduce allocation to minimal size. */ + p = GF_REALLOC (lines, (counter + 1) * sizeof (char *)); + if (!p) { + free_lines (lines, counter); + lines = NULL; + goto out; + } + lines = p; + + *line_count = counter; + ret = 0; + + out: + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, "%s", strerror (errno)); + if (fp) + fclose (fp); + + return lines; +} + +int +glusterd_compare_lines (const void *a, const void *b) { + + return strcmp(* (char * const *) a, * (char * const *) b); +} + +int +glusterd_sort_and_redirect (const char *src_filepath, int dest_fd) +{ + int ret = -1; + int line_count = 0; + int counter = 0; + char **lines = NULL; + + + if (!src_filepath || dest_fd < 0) + goto out; + + lines = glusterd_readin_file (src_filepath, &line_count); + if (!lines) + goto out; + + qsort (lines, line_count, sizeof (*lines), glusterd_compare_lines); + + for (counter = 0; lines[counter]; counter++) { + + ret = write (dest_fd, lines[counter], + strlen (lines[counter])); + if (ret < 0) + goto out; + + GF_FREE (lines[counter]); + } + + ret = 0; + out: + GF_FREE (lines); + + return ret; +} + int glusterd_volume_compute_cksum (glusterd_volinfo_t *volinfo) { @@ -1310,10 +1967,10 @@ glusterd_volume_compute_cksum (glusterd_volinfo_t *volinfo) char sort_filepath[PATH_MAX] = {0}; gf_boolean_t unlink_sortfile = _gf_false; int sort_fd = 0; - runner_t runner; + xlator_t *this = NULL; GF_ASSERT (volinfo); - + this = THIS; priv = THIS->private; GF_ASSERT (priv); @@ -1322,10 +1979,10 @@ glusterd_volume_compute_cksum (glusterd_volinfo_t *volinfo) snprintf (cksum_path, sizeof (cksum_path), "%s/%s", path, GLUSTERD_CKSUM_FILE); - fd = open (cksum_path, O_RDWR | O_APPEND | O_CREAT| O_TRUNC, 0644); + fd = open (cksum_path, O_RDWR | O_APPEND | O_CREAT| O_TRUNC, 0600); if (-1 == fd) { - gf_log (THIS->name, GF_LOG_ERROR, "Unable to open %s, errno: %d", + gf_log (this->name, GF_LOG_ERROR, "Unable to open %s, errno: %d", cksum_path, errno); ret = -1; goto out; @@ -1333,12 +1990,14 @@ glusterd_volume_compute_cksum (glusterd_volinfo_t *volinfo) snprintf (filepath, sizeof (filepath), "%s/%s", path, GLUSTERD_VOLUME_INFO_FILE); - snprintf (sort_filepath, sizeof (sort_filepath), "/tmp/%s.XXXXXX", - volinfo->volname); + snprintf (sort_filepath, sizeof (sort_filepath), + "/tmp/%s.XXXXXX", volinfo->volname); + sort_fd = mkstemp (sort_filepath); if (sort_fd < 0) { - gf_log (THIS->name, GF_LOG_ERROR, "Could not generate temp file, " - "reason: %s for volume: %s", strerror (errno), + gf_log (this->name, GF_LOG_ERROR, "Could not generate temp " + "file, reason: %s for %s: %s", strerror (errno), + (volinfo->is_snap_volume)?"snap":"volume", volinfo->volname); goto out; } else { @@ -1346,21 +2005,21 @@ glusterd_volume_compute_cksum (glusterd_volinfo_t *volinfo) } /* sort the info file, result in sort_filepath */ - runinit (&runner); - runner_add_args (&runner, "sort", filepath, NULL); - runner_redir (&runner, STDOUT_FILENO, sort_fd); - ret = runner_run (&runner); - close (sort_fd); + ret = glusterd_sort_and_redirect (filepath, sort_fd); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to sort file %s to %s", - filepath, sort_filepath); + gf_log (this->name, GF_LOG_ERROR, "sorting info file failed"); goto out; } + + ret = close (sort_fd); + if (ret) + goto out; + ret = get_checksum_for_path (sort_filepath, &cksum); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Unable to get checksum" + gf_log (this->name, GF_LOG_ERROR, "Unable to get checksum" " for path: %s", sort_filepath); goto out; } @@ -1379,55 +2038,88 @@ glusterd_volume_compute_cksum (glusterd_volinfo_t *volinfo) goto out; volinfo->cksum = cksum; - out: if (fd > 0) close (fd); if (unlink_sortfile) unlink (sort_filepath); - gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } -void -_add_volinfo_dict_to_prdict (dict_t *this, char *key, data_t *value, void *data) +int +_add_dict_to_prdict (dict_t *this, char *key, data_t *value, void *data) { - glusterd_voldict_ctx_t *ctx = NULL; + glusterd_dict_ctx_t *ctx = NULL; char optkey[512] = {0,}; int ret = -1; ctx = data; - snprintf (optkey, sizeof (optkey), "volume%d.%s%d", ctx->count, + snprintf (optkey, sizeof (optkey), "%s.%s%d", ctx->prefix, ctx->key_name, ctx->opt_count); ret = dict_set_str (ctx->dict, optkey, key); if (ret) gf_log ("", GF_LOG_ERROR, "option add for %s%d %s", - ctx->key_name, ctx->count, key); - snprintf (optkey, sizeof (optkey), "volume%d.%s%d", ctx->count, + ctx->key_name, ctx->opt_count, key); + snprintf (optkey, sizeof (optkey), "%s.%s%d", ctx->prefix, ctx->val_name, ctx->opt_count); ret = dict_set_str (ctx->dict, optkey, value->data); if (ret) gf_log ("", GF_LOG_ERROR, "option add for %s%d %s", - ctx->val_name, ctx->count, value->data); + ctx->val_name, ctx->opt_count, value->data); ctx->opt_count++; - return; + return ret; +} + +int32_t +glusterd_add_bricks_hname_path_to_dict (dict_t *dict, + glusterd_volinfo_t *volinfo) +{ + glusterd_brickinfo_t *brickinfo = NULL; + int ret = 0; + char key[256] = {0}; + int index = 0; + + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + snprintf (key, sizeof (key), "%d-hostname", index); + ret = dict_set_str (dict, key, brickinfo->hostname); + if (ret) + goto out; + + snprintf (key, sizeof (key), "%d-path", index); + ret = dict_set_str (dict, key, brickinfo->path); + if (ret) + goto out; + + index++; + } +out: + return ret; } int32_t glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, dict_t *dict, int32_t count) { - int32_t ret = -1; - char key[512] = {0,}; - glusterd_brickinfo_t *brickinfo = NULL; - int32_t i = 1; - char *volume_id_str = NULL; - char *src_brick = NULL; - char *dst_brick = NULL; - glusterd_voldict_ctx_t ctx = {0}; + int32_t ret = -1; + char prefix[512] = {0,}; + char key[512] = {0,}; + glusterd_brickinfo_t *brickinfo = NULL; + int32_t i = 1; + char *volume_id_str = NULL; + char *src_brick = NULL; + char *dst_brick = NULL; + char *str = NULL; + glusterd_dict_ctx_t ctx = {0}; + char *rebalance_id_str = NULL; + char *rb_id_str = NULL; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (dict); GF_ASSERT (volinfo); @@ -1442,6 +2134,15 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; + snprintf (key, sizeof (key), "volume%d.is_volume_restored", count); + ret = dict_set_int32 (dict, key, volinfo->is_volume_restored); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to set " + "is_volume_restored option for %s volume", + volinfo->volname); + goto out; + } + memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "volume%d.brick_count", count); ret = dict_set_int32 (dict, key, volinfo->brick_count); @@ -1496,30 +2197,91 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; - volume_id_str = gf_strdup (uuid_utoa (volinfo->volume_id)); - if (!volume_id_str) + snprintf (key, sizeof (key), "volume%d.is_snap_volume", count); + ret = dict_set_uint32 (dict, key, volinfo->is_snap_volume); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Unable to set %s", key); + goto out; + } + + snprintf (key, sizeof (key), "volume%d.snap-max-hard-limit", count); + ret = dict_set_uint64 (dict, key, volinfo->snap_max_hard_limit); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Unable to set %s", key); goto out; + } + volume_id_str = gf_strdup (uuid_utoa (volinfo->volume_id)); + if (!volume_id_str) { + ret = -1; + goto out; + } memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "volume%d.volume_id", count); ret = dict_set_dynstr (dict, key, volume_id_str); if (ret) goto out; + volume_id_str = NULL; + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.username", count); + str = glusterd_auth_get_username (volinfo); + if (str) { + ret = dict_set_dynstr (dict, key, gf_strdup (str)); + if (ret) + goto out; + } + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.password", count); + str = glusterd_auth_get_password (volinfo); + if (str) { + ret = dict_set_dynstr (dict, key, gf_strdup (str)); + if (ret) + goto out; + } + + memset (key, 0, sizeof (key)); + snprintf (key, 256, "volume%d.rebalance", count); + ret = dict_set_int32 (dict, key, volinfo->rebal.defrag_cmd); + if (ret) + goto out; + + if (volinfo->rebal.defrag_cmd) { + rebalance_id_str = gf_strdup (uuid_utoa + (volinfo->rebal.rebalance_id)); + if (!rebalance_id_str) { + ret = -1; + goto out; + } + memset (key, 0, sizeof (key)); + snprintf (key, 256, "volume%d.rebalance-id", count); + ret = dict_set_dynstr (dict, key, rebalance_id_str); + if (ret) + goto out; + rebalance_id_str = NULL; + } + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.rebalance-op", count); + ret = dict_set_uint32 (dict, key, volinfo->rebal.op); + if (ret) + goto out; memset (key, 0, sizeof (key)); snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_STATUS, count); - ret = dict_set_int32 (dict, key, volinfo->rb_status); + ret = dict_set_int32 (dict, key, volinfo->rep_brick.rb_status); if (ret) goto out; - if (volinfo->rb_status > GF_RB_STATUS_NONE) { + if (volinfo->rep_brick.rb_status > GF_RB_STATUS_NONE) { memset (key, 0, sizeof (key)); snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_SRC_BRICK, count); gf_asprintf (&src_brick, "%s:%s", - volinfo->src_brick->hostname, - volinfo->src_brick->path); + volinfo->rep_brick.src_brick->hostname, + volinfo->rep_brick.src_brick->path); ret = dict_set_dynstr (dict, key, src_brick); if (ret) goto out; @@ -1528,21 +2290,35 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_DST_BRICK, count); gf_asprintf (&dst_brick, "%s:%s", - volinfo->dst_brick->hostname, - volinfo->dst_brick->path); + volinfo->rep_brick.dst_brick->hostname, + volinfo->rep_brick.dst_brick->path); ret = dict_set_dynstr (dict, key, dst_brick); if (ret) goto out; + + rb_id_str = gf_strdup (uuid_utoa (volinfo->rep_brick.rb_id)); + if (!rb_id_str) { + ret = -1; + goto out; + } + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.rb_id", count); + ret = dict_set_dynstr (dict, key, rb_id_str); + if (ret) + goto out; + rb_id_str = NULL; } + snprintf (prefix, sizeof (prefix), "volume%d", count); ctx.dict = dict; - ctx.count = count; + ctx.prefix = prefix; ctx.opt_count = 1; ctx.key_name = "key"; ctx.val_name = "value"; GF_ASSERT (volinfo->dict); - dict_foreach (volinfo->dict, _add_volinfo_dict_to_prdict, &ctx); + dict_foreach (volinfo->dict, _add_dict_to_prdict, &ctx); ctx.opt_count--; memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "volume%d.opt-count", count); @@ -1551,13 +2327,13 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, goto out; ctx.dict = dict; - ctx.count = count; + ctx.prefix = prefix; ctx.opt_count = 1; ctx.key_name = "slave-num"; ctx.val_name = "slave-val"; GF_ASSERT (volinfo->gsync_slaves); - dict_foreach (volinfo->gsync_slaves, _add_volinfo_dict_to_prdict, &ctx); + dict_foreach (volinfo->gsync_slaves, _add_dict_to_prdict, &ctx); ctx.opt_count--; memset (key, 0, sizeof (key)); @@ -1581,11 +2357,48 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; + snprintf (key, sizeof (key), "volume%d.brick%d.snap_status", + count, i); + ret = dict_set_int32 (dict, key, brickinfo->snap_status); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set snap_status for %s:%s", + brickinfo->hostname, + brickinfo->path); + goto out; + } + + snprintf (key, sizeof (key), "volume%d.brick%d.device_path", + count, i); + ret = dict_set_str (dict, key, brickinfo->device_path); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set snap_device for %s:%s", + brickinfo->hostname, + brickinfo->path); + goto out; + } + i++; } + /* Add volume op-versions to dict. This prevents volume inconsistencies + * in the cluster + */ + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.op-version", count); + ret = dict_set_int32 (dict, key, volinfo->op_version); + if (ret) + goto out; + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.client-op-version", count); + ret = dict_set_int32 (dict, key, volinfo->client_op_version); out: + GF_FREE (volume_id_str); + GF_FREE (rebalance_id_str); + GF_FREE (rb_id_str); + gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); return ret; @@ -1599,6 +2412,7 @@ glusterd_build_volume_dict (dict_t **vols) glusterd_conf_t *priv = NULL; glusterd_volinfo_t *volinfo = NULL; int32_t count = 0; + glusterd_dict_ctx_t ctx = {0}; priv = THIS->private; @@ -1619,6 +2433,17 @@ glusterd_build_volume_dict (dict_t **vols) if (ret) goto out; + ctx.dict = dict; + ctx.prefix = "global"; + ctx.opt_count = 1; + ctx.key_name = "key"; + ctx.val_name = "val"; + dict_foreach (priv->opts, _add_dict_to_prdict, &ctx); + ctx.opt_count--; + ret = dict_set_int32 (dict, "global-opt-count", ctx.opt_count); + if (ret) + goto out; + *vols = dict; out: gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); @@ -1629,7 +2454,8 @@ out: } int32_t -glusterd_compare_friend_volume (dict_t *vols, int32_t count, int32_t *status) +glusterd_compare_friend_volume (dict_t *vols, int32_t count, int32_t *status, + char *hostname) { int32_t ret = -1; @@ -1665,8 +2491,8 @@ glusterd_compare_friend_volume (dict_t *vols, int32_t count, int32_t *status) //Mismatch detected ret = 0; gf_log ("", GF_LOG_ERROR, "Version of volume %s differ." - "local version = %d, remote version = %d", - volinfo->volname, volinfo->version, version); + "local version = %d, remote version = %d on peer %s", + volinfo->volname, volinfo->version, version, hostname); *status = GLUSTERD_VOL_COMP_UPDATE_REQ; goto out; } else if (version < volinfo->version) { @@ -1685,8 +2511,8 @@ glusterd_compare_friend_volume (dict_t *vols, int32_t count, int32_t *status) if (cksum != volinfo->cksum) { ret = 0; gf_log ("", GF_LOG_ERROR, "Cksums of volume %s differ." - " local cksum = %d, remote cksum = %d", - volinfo->volname, volinfo->cksum, cksum); + " local cksum = %u, remote cksum = %u on peer %s", + volinfo->volname, volinfo->cksum, cksum, hostname); *status = GLUSTERD_VOL_COMP_RJT; goto out; } @@ -1700,8 +2526,8 @@ out: } static int32_t -import_prdict_volinfo_dict (dict_t *vols, dict_t *dst_dict, char *key_prefix, - char *value_prefix, int opt_count, int count) +import_prdict_dict (dict_t *vols, dict_t *dst_dict, char *key_prefix, + char *value_prefix, int opt_count, char *prefix) { char key[512] = {0,}; int32_t ret = 0; @@ -1713,8 +2539,8 @@ import_prdict_volinfo_dict (dict_t *vols, dict_t *dst_dict, char *key_prefix, while (i <= opt_count) { memset (key, 0, sizeof (key)); - snprintf (key, sizeof (key), "volume%d.%s%d", - count, key_prefix, i); + snprintf (key, sizeof (key), "%s.%s%d", + prefix, key_prefix, i); ret = dict_get_str (vols, key, &opt_key); if (ret) { snprintf (msg, sizeof (msg), "Volume dict key not " @@ -1723,8 +2549,8 @@ import_prdict_volinfo_dict (dict_t *vols, dict_t *dst_dict, char *key_prefix, } memset (key, 0, sizeof (key)); - snprintf (key, sizeof (key), "volume%d.%s%d", - count, value_prefix, i); + snprintf (key, sizeof (key), "%s.%s%d", + prefix, value_prefix, i); ret = dict_get_str (vols, key, &opt_val); if (ret) { snprintf (msg, sizeof (msg), "Volume dict value not " @@ -1753,6 +2579,265 @@ out: } +gf_boolean_t +glusterd_is_quorum_option (char *option) +{ + gf_boolean_t res = _gf_false; + int i = 0; + char *keys[] = {GLUSTERD_QUORUM_TYPE_KEY, + GLUSTERD_QUORUM_RATIO_KEY, NULL}; + + for (i = 0; keys[i]; i++) { + if (strcmp (option, keys[i]) == 0) { + res = _gf_true; + break; + } + } + return res; +} + +gf_boolean_t +glusterd_is_quorum_changed (dict_t *options, char *option, char *value) +{ + int ret = 0; + gf_boolean_t reconfigured = _gf_false; + gf_boolean_t all = _gf_false; + char *oldquorum = NULL; + char *newquorum = NULL; + char *oldratio = NULL; + char *newratio = NULL; + + if ((strcmp ("all", option) != 0) && + !glusterd_is_quorum_option (option)) + goto out; + + if (strcmp ("all", option) == 0) + all = _gf_true; + + if (all || (strcmp (GLUSTERD_QUORUM_TYPE_KEY, option) == 0)) { + newquorum = value; + ret = dict_get_str (options, GLUSTERD_QUORUM_TYPE_KEY, + &oldquorum); + } + + if (all || (strcmp (GLUSTERD_QUORUM_RATIO_KEY, option) == 0)) { + newratio = value; + ret = dict_get_str (options, GLUSTERD_QUORUM_RATIO_KEY, + &oldratio); + } + + reconfigured = _gf_true; + + if (oldquorum && newquorum && (strcmp (oldquorum, newquorum) == 0)) + reconfigured = _gf_false; + if (oldratio && newratio && (strcmp (oldratio, newratio) == 0)) + reconfigured = _gf_false; + + if ((oldratio == NULL) && (newratio == NULL) && (oldquorum == NULL) && + (newquorum == NULL)) + reconfigured = _gf_false; +out: + return reconfigured; +} + +static inline gf_boolean_t +_is_contributing_to_quorum (gd_quorum_contrib_t contrib) +{ + if ((contrib == QUORUM_UP) || (contrib == QUORUM_DOWN)) + return _gf_true; + return _gf_false; +} + +static inline gf_boolean_t +_does_quorum_meet (int active_count, int quorum_count) +{ + return (active_count >= quorum_count); +} + +int +glusterd_get_quorum_cluster_counts (xlator_t *this, int *active_count, + int *quorum_count) +{ + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_conf_t *conf = NULL; + int ret = -1; + int inquorum_count = 0; + char *val = NULL; + double quorum_percentage = 0.0; + gf_boolean_t ratio = _gf_false; + int count = 0; + + conf = this->private; + //Start with counting self + inquorum_count = 1; + if (active_count) + *active_count = 1; + list_for_each_entry (peerinfo, &conf->peers, uuid_list) { + if (peerinfo->quorum_contrib == QUORUM_WAITING) + goto out; + + if (_is_contributing_to_quorum (peerinfo->quorum_contrib)) + inquorum_count = inquorum_count + 1; + + if (active_count && (peerinfo->quorum_contrib == QUORUM_UP)) + *active_count = *active_count + 1; + } + + ret = dict_get_str (conf->opts, GLUSTERD_QUORUM_RATIO_KEY, &val); + if (ret == 0) { + ratio = _gf_true; + ret = gf_string2percent (val, &quorum_percentage); + if (!ret) + ratio = _gf_true; + } + if (ratio) + count = CEILING_POS (inquorum_count * + quorum_percentage / 100.0); + else + count = (inquorum_count * 50 / 100) + 1; + + *quorum_count = count; + ret = 0; +out: + return ret; +} + +gf_boolean_t +glusterd_is_volume_in_server_quorum (glusterd_volinfo_t *volinfo) +{ + gf_boolean_t res = _gf_false; + char *quorum_type = NULL; + int ret = 0; + + ret = dict_get_str (volinfo->dict, GLUSTERD_QUORUM_TYPE_KEY, + &quorum_type); + if (ret) + goto out; + + if (strcmp (quorum_type, GLUSTERD_SERVER_QUORUM) == 0) + res = _gf_true; +out: + return res; +} + +gf_boolean_t +glusterd_is_any_volume_in_server_quorum (xlator_t *this) +{ + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + + conf = this->private; + list_for_each_entry (volinfo, &conf->volumes, vol_list) { + if (glusterd_is_volume_in_server_quorum (volinfo)) { + return _gf_true; + } + } + return _gf_false; +} + +gf_boolean_t +does_gd_meet_server_quorum (xlator_t *this) +{ + int quorum_count = 0; + int active_count = 0; + gf_boolean_t in = _gf_false; + glusterd_conf_t *conf = NULL; + int ret = -1; + + conf = this->private; + ret = glusterd_get_quorum_cluster_counts (this, &active_count, + &quorum_count); + if (ret) + goto out; + + if (!_does_quorum_meet (active_count, quorum_count)) { + goto out; + } + + in = _gf_true; +out: + return in; +} + +int +glusterd_spawn_daemons (void *opaque) +{ + glusterd_conf_t *conf = THIS->private; + gf_boolean_t start_bricks = !conf->restart_done; + + if (start_bricks) { + glusterd_restart_bricks (conf); + conf->restart_done = _gf_true; + } + glusterd_restart_gsyncds (conf); + glusterd_restart_rebalance (conf); + return 0; +} + +void +glusterd_do_volume_quorum_action (xlator_t *this, glusterd_volinfo_t *volinfo, + gf_boolean_t meets_quorum) +{ + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_conf_t *conf = NULL; + + conf = this->private; + if (volinfo->status != GLUSTERD_STATUS_STARTED) + goto out; + + if (!glusterd_is_volume_in_server_quorum (volinfo)) + meets_quorum = _gf_true; + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (!glusterd_is_local_brick (this, volinfo, brickinfo)) + continue; + if (meets_quorum) + glusterd_brick_start (volinfo, brickinfo, _gf_false); + else + glusterd_brick_stop (volinfo, brickinfo, _gf_false); + } +out: + return; +} + +int +glusterd_do_quorum_action () +{ + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = 0; + int active_count = 0; + int quorum_count = 0; + gf_boolean_t meets = _gf_false; + + this = THIS; + conf = this->private; + + conf->pending_quorum_action = _gf_true; + ret = glusterd_lock (conf->uuid); + if (ret) + goto out; + + { + ret = glusterd_get_quorum_cluster_counts (this, &active_count, + &quorum_count); + if (ret) + goto unlock; + + if (_does_quorum_meet (active_count, quorum_count)) + meets = _gf_true; + list_for_each_entry (volinfo, &conf->volumes, vol_list) { + glusterd_do_volume_quorum_action (this, volinfo, meets); + } + } +unlock: + (void)glusterd_unlock (conf->uuid); + conf->pending_quorum_action = _gf_false; +out: + return ret; +} + int32_t glusterd_import_friend_volume_opts (dict_t *vols, int count, glusterd_volinfo_t *volinfo) @@ -1761,6 +2846,7 @@ glusterd_import_friend_volume_opts (dict_t *vols, int count, int32_t ret = -1; int opt_count = 0; char msg[2048] = {0}; + char volume_prefix[1024] = {0}; memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "volume%d.opt-count", count); @@ -1771,8 +2857,9 @@ glusterd_import_friend_volume_opts (dict_t *vols, int count, goto out; } - ret = import_prdict_volinfo_dict (vols, volinfo->dict, "key", - "value", opt_count, count); + snprintf (volume_prefix, sizeof (volume_prefix), "volume%d", count); + ret = import_prdict_dict (vols, volinfo->dict, "key", "value", + opt_count, volume_prefix); if (ret) { snprintf (msg, sizeof (msg), "Unable to import options dict " "specified for %s", volinfo->volname); @@ -1788,9 +2875,8 @@ glusterd_import_friend_volume_opts (dict_t *vols, int count, goto out; } - ret = import_prdict_volinfo_dict (vols, volinfo->gsync_slaves, - "slave-num", "slave-val", opt_count, - count); + ret = import_prdict_dict (vols, volinfo->gsync_slaves, "slave-num", + "slave-val", opt_count, volume_prefix); if (ret) { snprintf (msg, sizeof (msg), "Unable to import gsync sessions " "specified for %s", volinfo->volname); @@ -1811,6 +2897,8 @@ glusterd_import_new_brick (dict_t *vols, int32_t vol_count, { char key[512] = {0,}; int ret = -1; + int32_t snap_status = 0; + char *snap_device = NULL; char *hostname = NULL; char *path = NULL; glusterd_brickinfo_t *new_brickinfo = NULL; @@ -1838,12 +2926,31 @@ glusterd_import_new_brick (dict_t *vols, int32_t vol_count, goto out; } + snprintf (key, sizeof (key), "volume%d.brick%d.snap_status", + vol_count, brick_count); + ret = dict_get_int32 (vols, key, &snap_status); + if (ret) { + snprintf (msg, sizeof (msg), "%s missing in payload", key); + goto out; + } + + snprintf (key, sizeof (key), "volume%d.brick%d.device_path", + vol_count, brick_count); + ret = dict_get_str (vols, key, &snap_device); + if (ret) { + snprintf (msg, sizeof (msg), "%s missing in payload", key); + goto out; + } + ret = glusterd_brickinfo_new (&new_brickinfo); if (ret) goto out; strcpy (new_brickinfo->path, path); strcpy (new_brickinfo->hostname, hostname); + strcpy (new_brickinfo->device_path, snap_device); + new_brickinfo->snap_status = snap_status; + //peerinfo might not be added yet (void) glusterd_resolve_brick (new_brickinfo); ret = 0; @@ -1893,7 +3000,13 @@ glusterd_import_volinfo (dict_t *vols, int count, char msg[2048] = {0}; char *src_brick = NULL; char *dst_brick = NULL; + char *str = NULL; int rb_status = 0; + char *rebalance_id_str = NULL; + char *rb_id_str = NULL; + int op_version = 0; + int client_op_version = 0; + uint32_t is_snap_volume = 0; GF_ASSERT (vols); GF_ASSERT (volinfo); @@ -1905,6 +3018,22 @@ glusterd_import_volinfo (dict_t *vols, int count, goto out; } + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.is_snap_volume", count); + ret = dict_get_uint32 (vols, key, &is_snap_volume); + if (ret) { + snprintf (msg, sizeof (msg), "%s missing in payload for %s", + key, volname); + goto out; + } + + if (is_snap_volume == _gf_true) { + gf_log (THIS->name, GF_LOG_DEBUG, + "Not syncing snap volume %s", volname); + ret = 0; + goto out; + } + ret = glusterd_volinfo_new (&new_volinfo); if (ret) goto out; @@ -2001,6 +3130,26 @@ glusterd_import_volinfo (dict_t *vols, int count, goto out; } + uuid_parse (volume_id_str, new_volinfo->volume_id); + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.username", count); + ret = dict_get_str (vols, key, &str); + if (!ret) { + ret = glusterd_auth_set_username (new_volinfo, str); + if (ret) + goto out; + } + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.password", count); + ret = dict_get_str (vols, key, &str); + if (!ret) { + ret = glusterd_auth_set_password (new_volinfo, str); + if (ret) + goto out; + } + memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "volume%d.transport_type", count); ret = dict_get_uint32 (vols, key, &new_volinfo->transport_type); @@ -2010,16 +3159,67 @@ glusterd_import_volinfo (dict_t *vols, int count, goto out; } - uuid_parse (volume_id_str, new_volinfo->volume_id); + new_volinfo->is_snap_volume = is_snap_volume; + + snprintf (key, sizeof (key), "volume%d.is_volume_restored", count); + ret = dict_get_uint32 (vols, key, &new_volinfo->is_volume_restored); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to get " + "is_volume_restored option for %s", + volname); + goto out; + } + + snprintf (key, sizeof (key), "volume%d.snap-max-hard-limit", count); + ret = dict_get_uint64 (vols, key, &new_volinfo->snap_max_hard_limit); + if (ret) { + snprintf (msg, sizeof (msg), "%s missing in payload for %s", + key, volname); + goto out; + } + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.rebalance", count); + ret = dict_get_uint32 (vols, key, &new_volinfo->rebal.defrag_cmd); + if (ret) { + snprintf (msg, sizeof (msg), "%s missing in payload for %s", + key, volname); + goto out; + } + + if (new_volinfo->rebal.defrag_cmd) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.rebalance-id", count); + ret = dict_get_str (vols, key, &rebalance_id_str); + if (ret) { + /* This is not present in older glusterfs versions, + * so don't error out + */ + ret = 0; + } else { + uuid_parse (rebalance_id_str, + new_volinfo->rebal.rebalance_id); + } + } + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.rebalance-op", count); + ret = dict_get_uint32 (vols, key,(uint32_t *) &new_volinfo->rebal.op); + if (ret) { + /* This is not present in older glusterfs versions, + * so don't error out + */ + ret = 0; + } memset (key, 0, sizeof (key)); snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_STATUS, count); ret = dict_get_int32 (vols, key, &rb_status); if (ret) goto out; - new_volinfo->rb_status = rb_status; + new_volinfo->rep_brick.rb_status = rb_status; - if (new_volinfo->rb_status > GF_RB_STATUS_NONE) { + if (new_volinfo->rep_brick.rb_status > GF_RB_STATUS_NONE) { memset (key, 0, sizeof (key)); snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_SRC_BRICK, @@ -2028,8 +3228,8 @@ glusterd_import_volinfo (dict_t *vols, int count, if (ret) goto out; - ret = glusterd_brickinfo_from_brick (src_brick, - &new_volinfo->src_brick); + ret = glusterd_brickinfo_new_from_brick (src_brick, + &new_volinfo->rep_brick.src_brick); if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to create" " src brickinfo"); @@ -2043,19 +3243,65 @@ glusterd_import_volinfo (dict_t *vols, int count, if (ret) goto out; - ret = glusterd_brickinfo_from_brick (dst_brick, - &new_volinfo->dst_brick); + ret = glusterd_brickinfo_new_from_brick (dst_brick, + &new_volinfo->rep_brick.dst_brick); if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to create" " dst brickinfo"); goto out; } + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.rb_id", count); + ret = dict_get_str (vols, key, &rb_id_str); + if (ret) { + /* This is not present in older glusterfs versions, + * so don't error out + */ + ret = 0; + } else { + uuid_parse (rb_id_str, new_volinfo->rep_brick.rb_id); + } } ret = glusterd_import_friend_volume_opts (vols, count, new_volinfo); if (ret) goto out; + + /* Import the volume's op-versions if available else set it to 1. + * Not having op-versions implies this informtation was obtained from a + * op-version 1 friend (gluster-3.3), ergo the cluster is at op-version + * 1 and all volumes are at op-versions 1. + * + * Either both the volume op-versions should be absent or both should be + * present. Only one being present is a failure + */ + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.op-version", count); + ret = dict_get_int32 (vols, key, &op_version); + if (ret) + ret = 0; + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.client-op-version", count); + ret = dict_get_int32 (vols, key, &client_op_version); + if (ret) + ret = 0; + + if (op_version && client_op_version) { + new_volinfo->op_version = op_version; + new_volinfo->client_op_version = client_op_version; + } else if (((op_version == 0) && (client_op_version != 0)) || + ((op_version != 0) && (client_op_version == 0))) { + ret = -1; + gf_log ("glusterd", GF_LOG_ERROR, + "Only one volume op-version found"); + goto out; + } else { + new_volinfo->op_version = 1; + new_volinfo->client_op_version = 1; + } + ret = glusterd_import_bricks (vols, count, new_volinfo); if (ret) goto out; @@ -2105,8 +3351,7 @@ glusterd_volinfo_copy_brick_portinfo (glusterd_volinfo_t *new_volinfo, ret = glusterd_volume_brickinfo_get (new_brickinfo->uuid, new_brickinfo->hostname, new_brickinfo->path, - old_volinfo, &old_brickinfo, - GF_PATH_COMPLETE); + old_volinfo, &old_brickinfo); if ((0 == ret) && glusterd_is_brick_started (old_brickinfo)) { new_brickinfo->port = old_brickinfo->port; } @@ -2132,10 +3377,13 @@ glusterd_volinfo_stop_stale_bricks (glusterd_volinfo_t *new_volinfo, ret = glusterd_volume_brickinfo_get (old_brickinfo->uuid, old_brickinfo->hostname, old_brickinfo->path, - new_volinfo, &new_brickinfo, - GF_PATH_COMPLETE); + new_volinfo, &new_brickinfo); if (ret) { - ret = glusterd_brick_stop (old_volinfo, old_brickinfo); + /*TODO: may need to switch to 'atomic' flavour of + * brick_stop, once we make peer rpc program also + * synctask enabled*/ + ret = glusterd_brick_stop (old_volinfo, old_brickinfo, + _gf_false); if (ret) gf_log ("glusterd", GF_LOG_ERROR, "Failed to " "stop brick %s:%s", old_brickinfo->hostname, @@ -2160,7 +3408,8 @@ glusterd_delete_stale_volume (glusterd_volinfo_t *stale_volinfo, * stop stale bricks. Stale volume information is going to be deleted. * Which deletes the valid brick information inside stale volinfo. * We dont want brick_rpc_notify to access already deleted brickinfo. - * Disconnect valid bricks. + * Disconnect all bricks from stale_volinfo (unconditionally), since + * they are being deleted subsequently. */ if (glusterd_is_volume_started (stale_volinfo)) { if (glusterd_is_volume_started (valid_volinfo)) { @@ -2169,10 +3418,12 @@ glusterd_delete_stale_volume (glusterd_volinfo_t *stale_volinfo, //Only valid bricks will be running now. (void) glusterd_volinfo_copy_brick_portinfo (valid_volinfo, stale_volinfo); - (void) glusterd_volume_disconnect_all_bricks (stale_volinfo); + } else { (void) glusterd_stop_bricks (stale_volinfo); } + + (void) glusterd_volume_disconnect_all_bricks (stale_volinfo); } /* Delete all the bricks and stores and vol files. They will be created * again by the valid_volinfo. Volume store delete should not be @@ -2182,7 +3433,7 @@ glusterd_delete_stale_volume (glusterd_volinfo_t *stale_volinfo, (void) glusterd_delete_all_bricks (stale_volinfo); if (stale_volinfo->shandle) { unlink (stale_volinfo->shandle->path); - (void) glusterd_store_handle_destroy (stale_volinfo->shandle); + (void) gf_store_handle_destroy (stale_volinfo->shandle); stale_volinfo->shandle = NULL; } (void) glusterd_volinfo_delete (stale_volinfo); @@ -2209,6 +3460,12 @@ glusterd_import_friend_volume (dict_t *vols, size_t count) if (ret) goto out; + if (!new_volinfo) { + gf_log (this->name, GF_LOG_DEBUG, + "Not importing snap volume"); + goto out; + } + ret = glusterd_volinfo_find (new_volinfo->volname, &old_volinfo); if (0 == ret) { (void) glusterd_delete_stale_volume (old_volinfo, new_volinfo); @@ -2254,8 +3511,97 @@ out: return ret; } +int +glusterd_get_global_opt_version (dict_t *opts, uint32_t *version) +{ + int ret = -1; + char *version_str = NULL; + + ret = dict_get_str (opts, GLUSTERD_GLOBAL_OPT_VERSION, &version_str); + if (ret) + goto out; + + ret = gf_string2uint (version_str, version); + if (ret) + goto out; + ret = 0; +out: + return ret; +} + +int +glusterd_get_next_global_opt_version_str (dict_t *opts, char **version_str) +{ + int ret = -1; + char version_string[64] = {0}; + uint32_t version = 0; + + ret = glusterd_get_global_opt_version (opts, &version); + if (ret) + goto out; + version++; + snprintf (version_string, sizeof (version_string), "%"PRIu32, version); + *version_str = gf_strdup (version_string); + if (*version_str) + ret = 0; +out: + return ret; +} + int32_t -glusterd_compare_friend_data (dict_t *vols, int32_t *status) +glusterd_import_global_opts (dict_t *friend_data) +{ + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + int ret = -1; + dict_t *import_options = NULL; + int count = 0; + uint32_t local_version = 0; + uint32_t remote_version = 0; + + this = THIS; + conf = this->private; + + ret = dict_get_int32 (friend_data, "global-opt-count", &count); + if (ret) { + //old version peer + ret = 0; + goto out; + } + + import_options = dict_new (); + if (!import_options) + goto out; + ret = import_prdict_dict (friend_data, import_options, "key", "val", + count, "global"); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to import" + " global options"); + goto out; + } + + ret = glusterd_get_global_opt_version (conf->opts, &local_version); + if (ret) + goto out; + ret = glusterd_get_global_opt_version (import_options, &remote_version); + if (ret) + goto out; + if (remote_version > local_version) { + ret = glusterd_store_options (this, import_options); + if (ret) + goto out; + dict_unref (conf->opts); + conf->opts = dict_ref (import_options); + } + ret = 0; +out: + if (import_options) + dict_unref (import_options); + return ret; +} + +int32_t +glusterd_compare_friend_data (dict_t *vols, int32_t *status, char *hostname) { int32_t ret = -1; int32_t count = 0; @@ -2272,7 +3618,8 @@ glusterd_compare_friend_data (dict_t *vols, int32_t *status) goto out; while (i <= count) { - ret = glusterd_compare_friend_volume (vols, i, status); + ret = glusterd_compare_friend_volume (vols, i, status, + hostname); if (ret) goto out; @@ -2291,6 +3638,9 @@ glusterd_compare_friend_data (dict_t *vols, int32_t *status) stale_nfs = _gf_true; if (glusterd_is_nodesvc_running ("glustershd")) stale_shd = _gf_true; + ret = glusterd_import_global_opts (vols); + if (ret) + goto out; ret = glusterd_import_friend_volumes (vols); if (ret) goto out; @@ -2387,42 +3737,52 @@ glusterd_get_nodesvc_volfile (char *server, char *workdir, } void -glusterd_shd_set_running (gf_boolean_t status) +glusterd_nodesvc_set_online_status (char *server, gf_boolean_t status) { glusterd_conf_t *priv = NULL; + GF_ASSERT (server); priv = THIS->private; GF_ASSERT (priv); GF_ASSERT (priv->shd); + GF_ASSERT (priv->nfs); - priv->shd->running = status; + if (!strcmp("glustershd", server)) + priv->shd->online = status; + else if (!strcmp ("nfs", server)) + priv->nfs->online = status; } gf_boolean_t -glusterd_shd_is_running () +glusterd_is_nodesvc_online (char *server) { glusterd_conf_t *conf = NULL; + gf_boolean_t online = _gf_false; + GF_ASSERT (server); conf = THIS->private; GF_ASSERT (conf); GF_ASSERT (conf->shd); + GF_ASSERT (conf->nfs); + + if (!strcmp (server, "glustershd")) + online = conf->shd->online; + else if (!strcmp (server, "nfs")) + online = conf->nfs->online; - return conf->shd->running; + return online; } int32_t -glusterd_shd_set_socket_filepath (char *rundir, uuid_t uuid, - char *socketpath, int len) +glusterd_nodesvc_set_socket_filepath (char *rundir, uuid_t uuid, + char *socketpath, int len) { char sockfilepath[PATH_MAX] = {0,}; - char md5_str[PATH_MAX] = {0,}; snprintf (sockfilepath, sizeof (sockfilepath), "%s/run-%s", rundir, uuid_utoa (uuid)); - _get_md5_str (md5_str, sizeof (md5_str), - (uint8_t *)sockfilepath, sizeof (sockfilepath)); - snprintf (socketpath, len, "%s/%s.socket", glusterd_sock_dir, - md5_str); + + glusterd_set_socket_filepath (sockfilepath, socketpath, len); return 0; } @@ -2432,6 +3792,9 @@ glusterd_pending_node_get_rpc (glusterd_pending_node_t *pending_node) struct rpc_clnt *rpc = NULL; glusterd_brickinfo_t *brickinfo = NULL; nodesrv_t *shd = NULL; + glusterd_volinfo_t *volinfo = NULL; + nodesrv_t *nfs = NULL; + GF_VALIDATE_OR_GOTO (THIS->name, pending_node, out); GF_VALIDATE_OR_GOTO (THIS->name, pending_node->node, out); @@ -2443,6 +3806,15 @@ glusterd_pending_node_get_rpc (glusterd_pending_node_t *pending_node) shd = pending_node->node; rpc = shd->rpc; + } else if (pending_node->type == GD_NODE_REBALANCE) { + volinfo = pending_node->node; + if (volinfo->rebal.defrag) + rpc = volinfo->rebal.defrag->rpc; + + } else if (pending_node->type == GD_NODE_NFS) { + nfs = pending_node->node; + rpc = nfs->rpc; + } else { GF_ASSERT (0); } @@ -2452,19 +3824,27 @@ out: } struct rpc_clnt* -glusterd_shd_get_rpc (void) +glusterd_nodesvc_get_rpc (char *server) { glusterd_conf_t *priv = NULL; + struct rpc_clnt *rpc = NULL; + GF_ASSERT (server); priv = THIS->private; GF_ASSERT (priv); GF_ASSERT (priv->shd); + GF_ASSERT (priv->nfs); + + if (!strcmp (server, "glustershd")) + rpc = priv->shd->rpc; + else if (!strcmp (server, "nfs")) + rpc = priv->nfs->rpc; - return priv->shd->rpc; + return rpc; } int32_t -glusterd_shd_set_rpc (struct rpc_clnt *rpc) +glusterd_nodesvc_set_rpc (char *server, struct rpc_clnt *rpc) { int ret = 0; xlator_t *this = NULL; @@ -2475,43 +3855,77 @@ glusterd_shd_set_rpc (struct rpc_clnt *rpc) priv = this->private; GF_ASSERT (priv); GF_ASSERT (priv->shd); + GF_ASSERT (priv->nfs); - priv->shd->rpc = rpc; + if (!strcmp ("glustershd", server)) + priv->shd->rpc = rpc; + else if (!strcmp ("nfs", server)) + priv->nfs->rpc = rpc; return ret; } int32_t -glusterd_shd_connect (char *socketpath) { +glusterd_nodesvc_connect (char *server, char *socketpath) { int ret = 0; dict_t *options = NULL; struct rpc_clnt *rpc = NULL; + glusterd_conf_t *priv = THIS->private; - ret = rpc_clnt_transport_unix_options_build (&options, socketpath); - if (ret) - goto out; - ret = glusterd_rpc_create (&rpc, options, - glusterd_shd_rpc_notify, - NULL); - if (ret) - goto out; - (void) glusterd_shd_set_rpc (rpc); + rpc = glusterd_nodesvc_get_rpc (server); + + if (rpc == NULL) { + /* Setting frame-timeout to 10mins (600seconds). + * Unix domain sockets ensures that the connection is reliable. + * The default timeout of 30mins used for unreliable network + * connections is too long for unix domain socket connections. + */ + ret = rpc_transport_unix_options_build (&options, socketpath, + 600); + if (ret) + goto out; + synclock_unlock (&priv->big_lock); + ret = glusterd_rpc_create (&rpc, options, + glusterd_nodesvc_rpc_notify, + server); + synclock_lock (&priv->big_lock); + if (ret) + goto out; + (void) glusterd_nodesvc_set_rpc (server, rpc); + } out: return ret; } int32_t -glusterd_nodesvc_start (char *server, gf_boolean_t pmap_signin) +glusterd_nodesvc_disconnect (char *server) { - int32_t ret = -1; - xlator_t *this = NULL; - glusterd_conf_t *priv = NULL; - char pidfile[PATH_MAX] = {0,}; - char logfile[PATH_MAX] = {0,}; - char volfile[PATH_MAX] = {0,}; - char rundir[PATH_MAX] = {0,}; - char shd_sockfpath[PATH_MAX] = {0,}; - char volfileid[256] = {0}; + struct rpc_clnt *rpc = NULL; + + rpc = glusterd_nodesvc_get_rpc (server); + (void)glusterd_nodesvc_set_rpc (server, NULL); + + if (rpc) + rpc_clnt_unref (rpc); + + return 0; +} + +int32_t +glusterd_nodesvc_start (char *server) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + runner_t runner = {0,}; + char pidfile[PATH_MAX] = {0,}; + char logfile[PATH_MAX] = {0,}; + char volfile[PATH_MAX] = {0,}; + char rundir[PATH_MAX] = {0,}; + char sockfpath[PATH_MAX] = {0,}; + char volfileid[256] = {0}; + char glusterd_uuid_option[1024] = {0}; + char valgrind_logfile[PATH_MAX] = {0}; this = THIS; GF_ASSERT(this); @@ -2519,7 +3933,7 @@ glusterd_nodesvc_start (char *server, gf_boolean_t pmap_signin) priv = this->private; glusterd_get_nodesvc_rundir (server, priv->workdir, - rundir, sizeof (rundir)); + rundir, sizeof (rundir)); ret = mkdir (rundir, 0777); if ((ret == -1) && (EEXIST != errno)) { @@ -2529,9 +3943,9 @@ glusterd_nodesvc_start (char *server, gf_boolean_t pmap_signin) } glusterd_get_nodesvc_pidfile (server, priv->workdir, - pidfile, sizeof (pidfile)); + pidfile, sizeof (pidfile)); glusterd_get_nodesvc_volfile (server, priv->workdir, - volfile, sizeof (volfile)); + volfile, sizeof (volfile)); ret = access (volfile, F_OK); if (ret) { gf_log ("", GF_LOG_ERROR, "%s Volfile %s is not present", @@ -2543,29 +3957,43 @@ glusterd_nodesvc_start (char *server, gf_boolean_t pmap_signin) server); snprintf (volfileid, sizeof (volfileid), "gluster/%s", server); - if (!strcmp (server, "glustershd")) { - glusterd_shd_set_socket_filepath (rundir, - priv->uuid, - shd_sockfpath, - sizeof (shd_sockfpath)); - } - - //TODO: kp:change the assumption that shd is the one which signs in - // use runner_add_args? - if (pmap_signin) { - ret = runcmd (SBIN_DIR"/glusterfs", "-s", "localhost", - "--volfile-id", volfileid, - "-p", pidfile, "-l", logfile, - "-S", shd_sockfpath, NULL); - if (!ret) - glusterd_shd_connect (shd_sockfpath); + glusterd_nodesvc_set_socket_filepath (rundir, MY_UUID, + sockfpath, sizeof (sockfpath)); + runinit (&runner); + + if (priv->valgrind) { + snprintf (valgrind_logfile, PATH_MAX, + "%s/valgrind-%s.log", + DEFAULT_LOG_FILE_DIRECTORY, + server); + + runner_add_args (&runner, "valgrind", "--leak-check=full", + "--trace-children=yes", "--track-origins=yes", + NULL); + runner_argprintf (&runner, "--log-file=%s", valgrind_logfile); } - else { - ret = runcmd (SBIN_DIR"/glusterfs", "-f", volfile, - "-p", pidfile, "-l", logfile, NULL); + + runner_add_args (&runner, SBIN_DIR"/glusterfs", + "-s", "localhost", + "--volfile-id", volfileid, + "-p", pidfile, + "-l", logfile, + "-S", sockfpath, NULL); + + if (!strcmp (server, "glustershd")) { + snprintf (glusterd_uuid_option, sizeof (glusterd_uuid_option), + "*replicate*.node-uuid=%s", uuid_utoa (MY_UUID)); + runner_add_args (&runner, "--xlator-option", + glusterd_uuid_option, NULL); } + runner_log (&runner, "", GF_LOG_DEBUG, + "Starting the nfs/glustershd services"); + ret = runner_run_nowait (&runner); + if (ret == 0) { + glusterd_nodesvc_connect (server, sockfpath); + } out: return ret; } @@ -2573,13 +4001,13 @@ out: int glusterd_nfs_server_start () { - return glusterd_nodesvc_start ("nfs", _gf_false); + return glusterd_nodesvc_start ("nfs"); } int glusterd_shd_start () { - return glusterd_nodesvc_start ("glustershd", _gf_true); + return glusterd_nodesvc_start ("glustershd"); } gf_boolean_t @@ -2594,6 +4022,31 @@ glusterd_is_nodesvc_running (char *server) } int32_t +glusterd_nodesvc_unlink_socket_file (char *server) +{ + int ret = 0; + char sockfpath[PATH_MAX] = {0,}; + char rundir[PATH_MAX] = {0,}; + glusterd_conf_t *priv = THIS->private; + + glusterd_get_nodesvc_rundir (server, priv->workdir, + rundir, sizeof (rundir)); + + glusterd_nodesvc_set_socket_filepath (rundir, MY_UUID, + sockfpath, sizeof (sockfpath)); + + ret = unlink (sockfpath); + if (ret && (ENOENT == errno)) { + ret = 0; + } else { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to remove %s" + " error: %s", sockfpath, strerror (errno)); + } + + return ret; +} + +int32_t glusterd_nodesvc_stop (char *server, int sig) { char pidfile[PATH_MAX] = {0,}; @@ -2602,9 +4055,17 @@ glusterd_nodesvc_stop (char *server, int sig) if (!glusterd_is_nodesvc_running (server)) goto out; + + (void)glusterd_nodesvc_disconnect (server); + glusterd_get_nodesvc_pidfile (server, priv->workdir, pidfile, sizeof (pidfile)); ret = glusterd_service_stop (server, pidfile, sig, _gf_true); + + if (ret == 0) { + glusterd_nodesvc_set_online_status (server, _gf_false); + (void)glusterd_nodesvc_unlink_socket_file (server); + } out: return ret; } @@ -2627,6 +4088,20 @@ glusterd_nfs_pmap_deregister () else gf_log ("", GF_LOG_ERROR, "De-register NFSV3 is unsuccessful"); + if (pmap_unset (NLM_PROGRAM, NLMV4_VERSION)) + gf_log ("", GF_LOG_INFO, "De-registered NLM v4 successfully"); + else + gf_log ("", GF_LOG_ERROR, "De-registration of NLM v4 failed"); + + if (pmap_unset (NLM_PROGRAM, NLMV1_VERSION)) + gf_log ("", GF_LOG_INFO, "De-registered NLM v1 successfully"); + else + gf_log ("", GF_LOG_ERROR, "De-registration of NLM v1 failed"); + + if (pmap_unset (ACL_PROGRAM, ACLV3_VERSION)) + gf_log ("", GF_LOG_INFO, "De-registered ACL v3 successfully"); + else + gf_log ("", GF_LOG_ERROR, "De-registration of ACL v3 failed"); } int @@ -2653,6 +4128,83 @@ glusterd_shd_stop () } int +glusterd_add_node_to_dict (char *server, dict_t *dict, int count, + dict_t *vol_opts) +{ + int ret = -1; + glusterd_conf_t *priv = THIS->private; + char pidfile[PATH_MAX] = {0,}; + gf_boolean_t running = _gf_false; + int pid = -1; + int port = 0; + char key[1024] = {0,}; + + glusterd_get_nodesvc_pidfile (server, priv->workdir, pidfile, + sizeof (pidfile)); + //Consider service to be running only when glusterd sees it Online + if (glusterd_is_nodesvc_online (server)) + running = glusterd_is_service_running (pidfile, &pid); + + /* For nfs-servers/self-heal-daemon setting + * brick<n>.hostname = "NFS Server" / "Self-heal Daemon" + * brick<n>.path = uuid + * brick<n>.port = 0 + * + * This might be confusing, but cli displays the name of + * the brick as hostname+path, so this will make more sense + * when output. + */ + snprintf (key, sizeof (key), "brick%d.hostname", count); + if (!strcmp (server, "nfs")) + ret = dict_set_str (dict, key, "NFS Server"); + else if (!strcmp (server, "glustershd")) + ret = dict_set_str (dict, key, "Self-heal Daemon"); + if (ret) + goto out; + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "brick%d.path", count); + ret = dict_set_dynstr (dict, key, gf_strdup (uuid_utoa (MY_UUID))); + if (ret) + goto out; + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "brick%d.port", count); + /* Port is available only for the NFS server. + * Self-heal daemon doesn't provide any port for access + * by entities other than gluster. + */ + if (!strcmp (server, "nfs")) { + if (dict_get (vol_opts, "nfs.port")) { + ret = dict_get_int32 (vol_opts, "nfs.port", &port); + if (ret) + goto out; + } else + port = GF_NFS3_PORT; + } + ret = dict_set_int32 (dict, key, port); + if (ret) + goto out; + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "brick%d.pid", count); + ret = dict_set_int32 (dict, key, pid); + if (ret) + goto out; + + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "brick%d.status", count); + ret = dict_set_int32 (dict, key, running); + if (ret) + goto out; + + +out: + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int glusterd_remote_hostname_get (rpcsvc_request_t *req, char *remote_host, int len) { GF_ASSERT (req); @@ -2680,8 +4232,7 @@ glusterd_remote_hostname_get (rpcsvc_request_t *req, char *remote_host, int len) out: - if (tmp_host) - GF_FREE (tmp_host); + GF_FREE (tmp_host); return ret; } @@ -2726,6 +4277,58 @@ glusterd_reconfigure_shd () } int +glusterd_reconfigure_nfs () +{ + int ret = -1; + gf_boolean_t identical = _gf_false; + + /* + * Check both OLD and NEW volfiles, if they are SAME by size + * and cksum i.e. "character-by-character". If YES, then + * NOTHING has been changed, just return. + */ + ret = glusterd_check_nfs_volfile_identical (&identical); + if (ret) + goto out; + + if (identical) { + ret = 0; + goto out; + } + + /* + * They are not identical. Find out if the topology is changed + * OR just the volume options. If just the options which got + * changed, then inform the xlator to reconfigure the options. + */ + identical = _gf_false; /* RESET the FLAG */ + ret = glusterd_check_nfs_topology_identical (&identical); + if (ret) + goto out; + + /* Topology is not changed, but just the options. But write the + * options to NFS volfile, so that NFS will be reconfigured. + */ + if (identical) { + ret = glusterd_create_nfs_volfile(); + if (ret == 0) {/* Only if above PASSES */ + ret = glusterd_fetchspec_notify (THIS); + } + goto out; + } + + /* + * NFS volfile's topology has been changed. NFS server needs + * to be RESTARTED to ACT on the changed volfile. + */ + ret = glusterd_check_generate_start_nfs (); + +out: + return ret; +} + + +int glusterd_check_generate_start_nfs () { int ret = 0; @@ -2773,8 +4376,8 @@ int glusterd_nodesvcs_start (glusterd_volinfo_t *volinfo) { return glusterd_nodesvcs_batch_op (volinfo, - glusterd_nfs_server_start, - glusterd_shd_start); + glusterd_nfs_server_start, + glusterd_shd_start); } int @@ -2849,7 +4452,7 @@ int glusterd_nodesvcs_handle_reconfigure (glusterd_volinfo_t *volinfo) { return glusterd_nodesvcs_batch_op (volinfo, - glusterd_check_generate_start_nfs, + glusterd_reconfigure_nfs, glusterd_reconfigure_shd); } @@ -2895,9 +4498,9 @@ glusterd_brickinfo_get (uuid_t uuid, char *hostname, char *path, list_for_each_entry (volinfo, &priv->volumes, vol_list) { ret = glusterd_volume_brickinfo_get (uuid, hostname, path, - volinfo, brickinfo, - GF_PATH_COMPLETE); - if (!ret) + volinfo, brickinfo); + if (ret == 0) + /*Found*/ goto out; } out: @@ -2906,7 +4509,8 @@ out: int glusterd_brick_start (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo) + glusterd_brickinfo_t *brickinfo, + gf_boolean_t wait) { int ret = -1; xlator_t *this = NULL; @@ -2923,61 +4527,83 @@ glusterd_brick_start (glusterd_volinfo_t *volinfo, if (uuid_is_null (brickinfo->uuid)) { ret = glusterd_resolve_brick (brickinfo); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, - "cannot resolve brick: %s:%s", + gf_log (this->name, GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK, brickinfo->hostname, brickinfo->path); goto out; } } - if (uuid_compare (brickinfo->uuid, conf->uuid)) { + if (uuid_compare (brickinfo->uuid, MY_UUID)) { ret = 0; goto out; } - ret = glusterd_volume_start_glusterfs (volinfo, brickinfo); + ret = glusterd_volume_start_glusterfs (volinfo, brickinfo, wait); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to start " - "glusterfs, ret: %d", ret); + gf_log (this->name, GF_LOG_ERROR, "Unable to start brick %s:%s", + brickinfo->hostname, brickinfo->path); goto out; } out: - gf_log ("", GF_LOG_DEBUG, "returning %d ", ret); + gf_log (this->name, GF_LOG_DEBUG, "returning %d ", ret); return ret; } int glusterd_restart_bricks (glusterd_conf_t *conf) { - glusterd_volinfo_t *volinfo = NULL; - glusterd_brickinfo_t *brickinfo = NULL; - int ret = 0; - gf_boolean_t start_nodesvcs = _gf_false; + int ret = 0; + glusterd_volinfo_t *volinfo = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_snap_t *snap = NULL; + gf_boolean_t start_nodesvcs = _gf_false; + xlator_t *this = NULL; - GF_ASSERT (conf); + this = THIS; + GF_ASSERT (this); list_for_each_entry (volinfo, &conf->volumes, vol_list) { - //If volume status is not started, do not proceed - if (volinfo->status == GLUSTERD_STATUS_STARTED) { + if (volinfo->status != GLUSTERD_STATUS_STARTED) + continue; + start_nodesvcs = _gf_true; + gf_log (this->name, GF_LOG_DEBUG, "starting the volume %s", + volinfo->volname); + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + glusterd_brick_start (volinfo, brickinfo, _gf_false); + } + } + + list_for_each_entry (snap, &conf->snapshots, snap_list) { + list_for_each_entry (volinfo, &snap->volumes, vol_list) { + if (volinfo->status != GLUSTERD_STATUS_STARTED) + continue; + start_nodesvcs = _gf_true; + gf_log (this->name, GF_LOG_DEBUG, "starting the snap " + "volume %s", volinfo->volname); list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { - glusterd_brick_start (volinfo, brickinfo); + glusterd_brick_start (volinfo, brickinfo, + _gf_false); } - start_nodesvcs = _gf_true; } } + if (start_nodesvcs) glusterd_nodesvcs_handle_graph_change (NULL); + return ret; } -void +int _local_gsyncd_start (dict_t *this, char *key, data_t *value, void *data) { + char *path_list = NULL; char *slave = NULL; int uuid_len = 0; + int ret = 0; char uuid_str[64] = {0}; glusterd_volinfo_t *volinfo = NULL; + char *conf_path = NULL; volinfo = data; GF_ASSERT (volinfo); @@ -2985,12 +4611,28 @@ _local_gsyncd_start (dict_t *this, char *key, data_t *value, void *data) if (slave) slave ++; else - return; + return 0; uuid_len = (slave - value->data - 1); - strncpy (uuid_str, (char*)value->data, uuid_len); - glusterd_start_gsync (volinfo, slave, uuid_str, NULL); + + ret = glusterd_get_local_brickpaths (volinfo, &path_list); + + ret = dict_get_str (this, "conf_path", &conf_path); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch conf file path."); + goto out; + } + + glusterd_start_gsync (volinfo, slave, path_list, conf_path, + uuid_str, NULL); + + GF_FREE (path_list); + path_list = NULL; + +out: + return ret; } int @@ -3014,6 +4656,15 @@ glusterd_restart_gsyncds (glusterd_conf_t *conf) return ret; } +inline int +glusterd_get_dist_leaf_count (glusterd_volinfo_t *volinfo) +{ + int rcount = volinfo->replica_count; + int scount = volinfo->stripe_count; + + return (rcount ? rcount : 1) * (scount ? scount : 1); +} + int glusterd_get_brickinfo (xlator_t *this, const char *brickname, int port, gf_boolean_t localhost, glusterd_brickinfo_t **brickinfo) @@ -3030,7 +4681,7 @@ glusterd_get_brickinfo (xlator_t *this, const char *brickname, int port, list_for_each_entry (volinfo, &priv->volumes, vol_list) { list_for_each_entry (tmpbrkinfo, &volinfo->bricks, brick_list) { - if (localhost && glusterd_is_local_addr (tmpbrkinfo->hostname)) + if (localhost && !gf_is_local_addr (tmpbrkinfo->hostname)) continue; if (!strcmp(tmpbrkinfo->path, brickname) && (tmpbrkinfo->port == port)) { @@ -3042,9 +4693,23 @@ glusterd_get_brickinfo (xlator_t *this, const char *brickname, int port, return ret; } +glusterd_brickinfo_t* +glusterd_get_brickinfo_by_position (glusterd_volinfo_t *volinfo, uint32_t pos) +{ + glusterd_brickinfo_t *tmpbrkinfo = NULL; + + list_for_each_entry (tmpbrkinfo, &volinfo->bricks, + brick_list) { + if (pos == 0) + return tmpbrkinfo; + pos--; + } + return NULL; +} + void glusterd_set_brick_status (glusterd_brickinfo_t *brickinfo, - gf_brick_status_t status) + gf_brick_status_t status) { GF_ASSERT (brickinfo); brickinfo->status = status; @@ -3087,26 +4752,28 @@ out: return -1; } +#ifdef GF_LINUX_HOST_OS int glusterd_get_brick_root (char *path, char **mount_point) { char *ptr = NULL; + char *mnt_pt = NULL; struct stat brickstat = {0}; struct stat buf = {0}; if (!path) goto err; - *mount_point = gf_strdup (path); - if (!*mount_point) + mnt_pt = gf_strdup (path); + if (!mnt_pt) goto err; - if (stat (*mount_point, &brickstat)) + if (stat (mnt_pt, &brickstat)) goto err; - while ((ptr = strrchr (*mount_point, '/')) && - ptr != *mount_point) { + while ((ptr = strrchr (mnt_pt, '/')) && + ptr != mnt_pt) { *ptr = '\0'; - if (stat (*mount_point, &buf)) { + if (stat (mnt_pt, &buf)) { gf_log (THIS->name, GF_LOG_ERROR, "error in " "stat: %s", strerror (errno)); goto err; @@ -3118,35 +4785,58 @@ glusterd_get_brick_root (char *path, char **mount_point) } } - if (ptr == *mount_point) { + if (ptr == mnt_pt) { if (stat ("/", &buf)) { gf_log (THIS->name, GF_LOG_ERROR, "error in " "stat: %s", strerror (errno)); goto err; } if (brickstat.st_dev == buf.st_dev) - strcpy (*mount_point, "/"); + strcpy (mnt_pt, "/"); } + *mount_point = mnt_pt; return 0; err: - if (*mount_point) - GF_FREE (*mount_point); + GF_FREE (mnt_pt); return -1; } +static char* +glusterd_parse_inode_size (char *stream, char *pattern) +{ + char *needle = NULL; + char *trail = NULL; + + needle = strstr (stream, pattern); + if (!needle) + goto out; + + needle = nwstrtail (needle, pattern); + + trail = needle; + while (trail && isdigit (*trail)) trail++; + if (trail) + *trail = '\0'; + +out: + return needle; +} + static int glusterd_add_inode_size_to_dict (dict_t *dict, int count) { int ret = -1; - int fd = -1; char key[1024] = {0}; char buffer[4096] = {0}; - char cmd_str[4096] = {0}; char *inode_size = NULL; char *device = NULL; char *fs_name = NULL; + char *cur_word = NULL; + char *pattern = NULL; + char *trail = NULL; + runner_t runner = {0, }; memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "brick%d.device", count); @@ -3160,26 +4850,18 @@ glusterd_add_inode_size_to_dict (dict_t *dict, int count) if (ret) goto out; + runinit (&runner); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); /* get inode size for xfs or ext2/3/4 */ if (!strcmp (fs_name, "xfs")) { - snprintf (cmd_str, sizeof (cmd_str), - "xfs_info %s | " - "grep isize | " - "cut -d ' ' -f 2- | " - "cut -d '=' -f 2 | " - "cut -d ' ' -f 1 " - "> /tmp/gf_status.txt ", - device); + runner_add_args (&runner, "xfs_info", device, NULL); + pattern = "isize="; } else if (IS_EXT_FS(fs_name)) { - snprintf (cmd_str, sizeof (cmd_str), - "tune2fs -l %s | " - "grep -i 'inode size' | " - "awk '{print $3}' " - "> /tmp/gf_status.txt ", - device); + runner_add_args (&runner, "tune2fs", "-l", device, NULL); + pattern = "Inode size:"; } else { ret = 0; @@ -3189,7 +4871,7 @@ glusterd_add_inode_size_to_dict (dict_t *dict, int count) goto out; } - ret = runcmd ("/bin/sh", "-c", cmd_str, NULL); + ret = runner_start (&runner); if (ret) { gf_log (THIS->name, GF_LOG_ERROR, "could not get inode " "size for %s : %s package missing", fs_name, @@ -3198,53 +4880,85 @@ glusterd_add_inode_size_to_dict (dict_t *dict, int count) goto out; } - fd = open ("/tmp/gf_status.txt", O_RDONLY); - unlink ("/tmp/gf_status.txt"); - if (fd < 0) { - ret = -1; + for (;;) { + if (fgets (buffer, sizeof (buffer), + runner_chio (&runner, STDOUT_FILENO)) == NULL) + break; + trail = strrchr (buffer, '\n'); + if (trail) + *trail = '\0'; + + cur_word = glusterd_parse_inode_size (buffer, pattern); + if (cur_word) + break; + } + + ret = runner_end (&runner); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "%s exited with non-zero " + "exit status", ((!strcmp (fs_name, "xfs")) ? + "xfs_info" : "tune2fs")); goto out; } - memset (buffer, 0, sizeof (buffer)); - ret = read (fd, buffer, sizeof (buffer)); - if (ret < 2) { + if (!cur_word) { ret = -1; + gf_log (THIS->name, GF_LOG_ERROR, "Unable to retrieve inode " + "size using %s", + (!strcmp (fs_name, "xfs")? "xfs_info": "tune2fs")); goto out; } + inode_size = gf_strdup (cur_word); + memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "brick%d.inode_size", count); - inode_size = get_nth_word (buffer, 1); - if (!inode_size) { - ret = -1; - goto out; - } - ret = dict_set_dynstr (dict, key, inode_size); out: - if (fd >= 0) - close (fd); if (ret) gf_log (THIS->name, GF_LOG_ERROR, "failed to get inode size"); return ret; } +struct mntent * +glusterd_get_mnt_entry_info (char *mnt_pt, FILE *mtab) +{ + struct mntent *entry = NULL; + + mtab = setmntent (_PATH_MOUNTED, "r"); + if (!mtab) + goto out; + + entry = getmntent (mtab); + + while (1) { + if (!entry) + goto out; + + if (!strcmp (entry->mnt_dir, mnt_pt) && + strcmp (entry->mnt_type, "rootfs")) + break; + entry = getmntent (mtab); + } + +out: + return entry; +} + static int glusterd_add_brick_mount_details (glusterd_brickinfo_t *brickinfo, dict_t *dict, int count) { int ret = -1; - int fd = -1; char key[1024] = {0}; char base_key[1024] = {0}; - char buffer[4096] = {0}; - char cmd_str[1024] = {0}; char *mnt_pt = NULL; char *fs_name = NULL; char *mnt_options = NULL; char *device = NULL; - runner_t runner = {0}; + struct mntent *entry = NULL; + FILE *mtab = NULL; snprintf (base_key, sizeof (base_key), "brick%d", count); @@ -3252,33 +4966,17 @@ glusterd_add_brick_mount_details (glusterd_brickinfo_t *brickinfo, if (ret) goto out; - /* get mount details of brick in back-end */ - snprintf (cmd_str, sizeof (cmd_str), " %s ", mnt_pt); - - runinit (&runner); - runner_add_args (&runner, "grep", cmd_str, "/etc/mtab", NULL); - runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); - - ret = runner_start (&runner); - if (ret) - goto out; - - if (!fgets (buffer, sizeof(buffer), - runner_chio (&runner, STDOUT_FILENO))) { + entry = glusterd_get_mnt_entry_info (mnt_pt, mtab); + if (!entry) { ret = -1; goto out; } - runner_end (&runner); - /* get device file */ memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "%s.device", base_key); - device = get_nth_word (buffer, 1); - if (!device) - goto out; - + device = gf_strdup (entry->mnt_fsname); ret = dict_set_dynstr (dict, key, device); if (ret) goto out; @@ -3287,10 +4985,7 @@ glusterd_add_brick_mount_details (glusterd_brickinfo_t *brickinfo, memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "%s.fs_name", base_key); - fs_name = get_nth_word (buffer, 3); - if (!fs_name) - goto out; - + fs_name = gf_strdup (entry->mnt_type); ret = dict_set_dynstr (dict, key, fs_name); if (ret) goto out; @@ -3299,19 +4994,57 @@ glusterd_add_brick_mount_details (glusterd_brickinfo_t *brickinfo, memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "%s.mnt_options", base_key); - mnt_options = get_nth_word (buffer, 4); - if (!mnt_options) - goto out; + mnt_options = gf_strdup (entry->mnt_opts); ret = dict_set_dynstr (dict, key, mnt_options); out: - if (mnt_pt) - GF_FREE (mnt_pt); - if (fd >= 0) - close (fd); + GF_FREE (mnt_pt); + if (mtab) + endmntent (mtab); + return ret; } +char* +glusterd_get_brick_mount_details (glusterd_brickinfo_t *brickinfo) +{ + int ret = -1; + char *mnt_pt = NULL; + char *device = NULL; + FILE *mtab = NULL; + struct mntent *entry = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (brickinfo); + + ret = glusterd_get_brick_root (brickinfo->path, &mnt_pt); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get mount point " + "for %s brick", brickinfo->path); + goto out; + } + + entry = glusterd_get_mnt_entry_info (mnt_pt, mtab); + if (NULL == entry) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get mnt entry " + "for %s mount path", mnt_pt); + goto out; + } + + /* get the fs_name/device */ + device = gf_strdup (entry->mnt_fsname); + +out: + if (NULL != mtab) { + endmntent (mtab); + } + + return device; +} +#endif + int glusterd_add_brick_detail_to_dict (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo, @@ -3326,7 +5059,9 @@ glusterd_add_brick_detail_to_dict (glusterd_volinfo_t *volinfo, char key[1024] = {0}; char base_key[1024] = {0}; struct statvfs brickstat = {0}; + xlator_t *this = NULL; + this = THIS; GF_ASSERT (volinfo); GF_ASSERT (brickinfo); GF_ASSERT (dict); @@ -3335,7 +5070,7 @@ glusterd_add_brick_detail_to_dict (glusterd_volinfo_t *volinfo, ret = statvfs (brickinfo->path, &brickstat); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "statfs error: %s ", + gf_log (this->name, GF_LOG_ERROR, "statfs error: %s ", strerror (errno)); goto out; } @@ -3382,16 +5117,16 @@ glusterd_add_brick_detail_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; } - +#ifdef GF_LINUX_HOST_OS ret = glusterd_add_brick_mount_details (brickinfo, dict, count); if (ret) goto out; ret = glusterd_add_inode_size_to_dict (dict, count); - +#endif out: if (ret) - gf_log (THIS->name, GF_LOG_DEBUG, "Error adding brick" + gf_log (this->name, GF_LOG_DEBUG, "Error adding brick" " detail to dict: %s", strerror (errno)); return ret; } @@ -3408,7 +5143,6 @@ glusterd_add_brick_to_dict (glusterd_volinfo_t *volinfo, char key[1024] = {0}; char base_key[1024] = {0}; char pidfile[PATH_MAX] = {0}; - char path[PATH_MAX] = {0}; xlator_t *this = NULL; glusterd_conf_t *priv = NULL; @@ -3441,9 +5175,7 @@ glusterd_add_brick_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; - GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv); - GLUSTERD_GET_BRICK_PIDFILE (pidfile, path, brickinfo->hostname, - brickinfo->path); + GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, brickinfo, priv); brick_online = glusterd_is_service_running (pidfile, &pid); @@ -3459,7 +5191,7 @@ glusterd_add_brick_to_dict (glusterd_volinfo_t *volinfo, out: if (ret) - gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -3531,11 +5263,14 @@ glusterd_friend_find_by_uuid (uuid_t uuid, int ret = -1; glusterd_conf_t *priv = NULL; glusterd_peerinfo_t *entry = NULL; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (peerinfo); *peerinfo = NULL; - priv = THIS->private; + priv = this->private; GF_ASSERT (priv); @@ -3545,7 +5280,7 @@ glusterd_friend_find_by_uuid (uuid_t uuid, list_for_each_entry (entry, &priv->peers, uuid_list) { if (!uuid_compare (entry->uuid, uuid)) { - gf_log ("glusterd", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "Friend found... state: %s", glusterd_friend_sm_state_name_get (entry->state.state)); *peerinfo = entry; @@ -3553,7 +5288,7 @@ glusterd_friend_find_by_uuid (uuid_t uuid, } } - gf_log ("glusterd", GF_LOG_DEBUG, "Friend with uuid: %s, not found", + gf_log (this->name, GF_LOG_DEBUG, "Friend with uuid: %s, not found", uuid_utoa (uuid)); return ret; } @@ -3573,12 +5308,15 @@ glusterd_friend_find_by_hostname (const char *hoststr, struct sockaddr_in *s4 = NULL; struct in_addr *in_addr = NULL; char hname[1024] = {0,}; + xlator_t *this = NULL; + + this = THIS; GF_ASSERT (hoststr); GF_ASSERT (peerinfo); *peerinfo = NULL; - priv = THIS->private; + priv = this->private; GF_ASSERT (priv); @@ -3586,7 +5324,7 @@ glusterd_friend_find_by_hostname (const char *hoststr, if (!strncasecmp (entry->hostname, hoststr, 1024)) { - gf_log ("glusterd", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "Friend %s found.. state: %d", hoststr, entry->state.state); *peerinfo = entry; @@ -3594,9 +5332,10 @@ glusterd_friend_find_by_hostname (const char *hoststr, } } - ret = getaddrinfo(hoststr, NULL, NULL, &addr); + ret = getaddrinfo (hoststr, NULL, NULL, &addr); if (ret != 0) { - gf_log ("", GF_LOG_ERROR, "error in getaddrinfo: %s\n", + gf_log (this->name, GF_LOG_ERROR, + "error in getaddrinfo: %s\n", gai_strerror(ret)); goto out; } @@ -3625,7 +5364,7 @@ glusterd_friend_find_by_hostname (const char *hoststr, if (!strncasecmp (entry->hostname, host, 1024) || !strncasecmp (entry->hostname,hname, 1024)) { - gf_log ("glusterd", GF_LOG_DEBUG, + gf_log (this->name, GF_LOG_DEBUG, "Friend %s found.. state: %d", hoststr, entry->state.state); *peerinfo = entry; @@ -3636,7 +5375,7 @@ glusterd_friend_find_by_hostname (const char *hoststr, } out: - gf_log ("glusterd", GF_LOG_DEBUG, "Unable to find friend: %s", hoststr); + gf_log (this->name, GF_LOG_DEBUG, "Unable to find friend: %s", hoststr); if (addr) freeaddrinfo (addr); return -1; @@ -3660,23 +5399,26 @@ glusterd_hostname_to_uuid (char *hostname, uuid_t uuid) ret = glusterd_friend_find_by_hostname (hostname, &peerinfo); if (ret) { - ret = glusterd_is_local_addr (hostname); - if (ret) + if (gf_is_local_addr (hostname)) { + uuid_copy (uuid, MY_UUID); + ret = 0; + } else { + ret = 0; goto out; - else - uuid_copy (uuid, priv->uuid); + } } else { uuid_copy (uuid, peerinfo->uuid); } out: - gf_log ("", GF_LOG_DEBUG, "returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret); return ret; } int glusterd_brick_stop (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo) + glusterd_brickinfo_t *brickinfo, + gf_boolean_t del_brick) { int ret = -1; xlator_t *this = NULL; @@ -3693,38 +5435,39 @@ glusterd_brick_stop (glusterd_volinfo_t *volinfo, if (uuid_is_null (brickinfo->uuid)) { ret = glusterd_resolve_brick (brickinfo); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, - "cannot resolve brick: %s:%s", + gf_log (this->name, GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK, brickinfo->hostname, brickinfo->path); goto out; } } - if (uuid_compare (brickinfo->uuid, conf->uuid)) { + if (uuid_compare (brickinfo->uuid, MY_UUID)) { ret = 0; + if (del_brick) + glusterd_delete_brick (volinfo, brickinfo); goto out; } - gf_log ("", GF_LOG_INFO, "About to stop glusterfs" + gf_log (this->name, GF_LOG_DEBUG, "About to stop glusterfs" " for brick %s:%s", brickinfo->hostname, brickinfo->path); - ret = glusterd_volume_stop_glusterfs (volinfo, brickinfo); + ret = glusterd_volume_stop_glusterfs (volinfo, brickinfo, del_brick); if (ret) { - gf_log ("", GF_LOG_CRITICAL, "Unable to remove" + gf_log (this->name, GF_LOG_CRITICAL, "Unable to stop" " brick: %s:%s", brickinfo->hostname, brickinfo->path); goto out; } out: - gf_log ("", GF_LOG_DEBUG, "returning %d ", ret); + gf_log (this->name, GF_LOG_DEBUG, "returning %d ", ret); return ret; } int glusterd_is_defrag_on (glusterd_volinfo_t *volinfo) { - return (volinfo->defrag != NULL); + return (volinfo->rebal.defrag != NULL); } gf_boolean_t @@ -3742,48 +5485,10 @@ glusterd_is_rb_ongoing (glusterd_volinfo_t *volinfo) } int -glusterd_is_replace_running (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo) -{ - int ret = 0; - char *src_hostname = NULL; - char *brick_hostname = NULL; - - if (volinfo->src_brick) { - src_hostname = gf_strdup (volinfo->src_brick->hostname); - if (!src_hostname) { - ret = -1; - goto out; - } - } else { - gf_log ("glusterd", GF_LOG_DEBUG, - "replace brick is not running"); - goto out; - } - - brick_hostname = gf_strdup (brickinfo->hostname); - if (!brick_hostname) { - ret = -1; - goto out; - } - if (!glusterd_is_local_addr (src_hostname) && !glusterd_is_local_addr (brick_hostname)) { - if (glusterd_is_rb_started (volinfo) || glusterd_is_rb_paused (volinfo)) - ret = -1; - } - -out: - if (src_hostname) - GF_FREE (src_hostname); - if (brick_hostname) - GF_FREE (brick_hostname); - return ret; -} - -int glusterd_new_brick_validate (char *brick, glusterd_brickinfo_t *brickinfo, char *op_errstr, size_t len) { glusterd_brickinfo_t *newbrickinfo = NULL; - glusterd_brickinfo_t *tmpbrkinfo = NULL; int ret = -1; gf_boolean_t is_allocated = _gf_false; glusterd_peerinfo_t *peerinfo = NULL; @@ -3800,7 +5505,7 @@ glusterd_new_brick_validate (char *brick, glusterd_brickinfo_t *brickinfo, GF_ASSERT (op_errstr); if (!brickinfo) { - ret = glusterd_brickinfo_from_brick (brick, &newbrickinfo); + ret = glusterd_brickinfo_new_from_brick (brick, &newbrickinfo); if (ret) goto out; is_allocated = _gf_true; @@ -3810,42 +5515,54 @@ glusterd_new_brick_validate (char *brick, glusterd_brickinfo_t *brickinfo, ret = glusterd_resolve_brick (newbrickinfo); if (ret) { - snprintf (op_errstr, len, "Host %s not a friend", - newbrickinfo->hostname); - gf_log (THIS->name, GF_LOG_ERROR, "%s", op_errstr); + snprintf(op_errstr, len, "Host %s is not in \'Peer " + "in Cluster\' state", newbrickinfo->hostname); goto out; } - if (!uuid_compare (priv->uuid, newbrickinfo->uuid)) - goto brick_validation; - ret = glusterd_friend_find_by_uuid (newbrickinfo->uuid, &peerinfo); - if (ret) - goto out; - if ((!peerinfo->connected) || - (peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED)) { - snprintf(op_errstr, len, "Host %s not connected", - newbrickinfo->hostname); - gf_log (THIS->name, GF_LOG_ERROR, "%s", op_errstr); - ret = -1; - goto out; - } -brick_validation: - ret = glusterd_brickinfo_get (newbrickinfo->uuid, - newbrickinfo->hostname, - newbrickinfo->path, &tmpbrkinfo); - if (!ret) { - snprintf(op_errstr, len, "Brick: %s already in use", - brick); - gf_log (THIS->name, GF_LOG_ERROR, "%s", op_errstr); - ret = -1; - goto out; + if (!uuid_compare (MY_UUID, newbrickinfo->uuid)) { + /* brick is local */ + if (!glusterd_is_brickpath_available (newbrickinfo->uuid, + newbrickinfo->path)) { + snprintf(op_errstr, len, "Brick: %s not available." + " Brick may be containing or be contained " + "by an existing brick", brick); + ret = -1; + goto out; + } + } else { - ret = 0; + ret = glusterd_friend_find_by_uuid (newbrickinfo->uuid, + &peerinfo); + if (ret) { + snprintf (op_errstr, len, "Failed to find host %s", + newbrickinfo->hostname); + goto out; + } + + if ((!peerinfo->connected)) { + snprintf(op_errstr, len, "Host %s not connected", + newbrickinfo->hostname); + ret = -1; + goto out; + } + + if (peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) { + snprintf(op_errstr, len, "Host %s is not in \'Peer " + "in Cluster\' state", + newbrickinfo->hostname); + ret = -1; + goto out; + } } + + ret = 0; out: - if (is_allocated && newbrickinfo) + if (is_allocated) glusterd_brickinfo_delete (newbrickinfo); - gf_log (THIS->name, GF_LOG_DEBUG, "returning %d ", ret); + if (op_errstr[0] != '\0') + gf_log (this->name, GF_LOG_ERROR, "%s", op_errstr); + gf_log (this->name, GF_LOG_DEBUG, "returning %d ", ret); return ret; } @@ -3853,8 +5570,8 @@ int glusterd_is_rb_started(glusterd_volinfo_t *volinfo) { gf_log ("", GF_LOG_DEBUG, - "is_rb_started:status=%d", volinfo->rb_status); - return (volinfo->rb_status == GF_RB_STATUS_STARTED); + "is_rb_started:status=%d", volinfo->rep_brick.rb_status); + return (volinfo->rep_brick.rb_status == GF_RB_STATUS_STARTED); } @@ -3862,9 +5579,9 @@ int glusterd_is_rb_paused ( glusterd_volinfo_t *volinfo) { gf_log ("", GF_LOG_DEBUG, - "is_rb_paused:status=%d", volinfo->rb_status); + "is_rb_paused:status=%d", volinfo->rep_brick.rb_status); - return (volinfo->rb_status == GF_RB_STATUS_PAUSED); + return (volinfo->rep_brick.rb_status == GF_RB_STATUS_PAUSED); } inline int @@ -3872,10 +5589,10 @@ glusterd_set_rb_status (glusterd_volinfo_t *volinfo, gf_rb_status_t status) { gf_log ("", GF_LOG_DEBUG, "setting status from %d to %d", - volinfo->rb_status, + volinfo->rep_brick.rb_status, status); - volinfo->rb_status = status; + volinfo->rep_brick.rb_status = status; return 0; } @@ -3883,142 +5600,192 @@ inline int glusterd_rb_check_bricks (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *src, glusterd_brickinfo_t *dst) { - if (!volinfo->src_brick || !volinfo->dst_brick) + glusterd_replace_brick_t *rb = NULL; + + GF_ASSERT (volinfo); + + rb = &volinfo->rep_brick; + + if (!rb->src_brick || !rb->dst_brick) return -1; - if (strcmp (volinfo->src_brick->hostname, src->hostname) || - strcmp (volinfo->src_brick->path, src->path)) { + if (strcmp (rb->src_brick->hostname, src->hostname) || + strcmp (rb->src_brick->path, src->path)) { gf_log("", GF_LOG_ERROR, "Replace brick src bricks differ"); return -1; } - if (strcmp (volinfo->dst_brick->hostname, dst->hostname) || - strcmp (volinfo->dst_brick->path, dst->path)) { + + if (strcmp (rb->dst_brick->hostname, dst->hostname) || + strcmp (rb->dst_brick->path, dst->path)) { gf_log ("", GF_LOG_ERROR, "Replace brick dst bricks differ"); return -1; } + return 0; } -int -glusterd_brick_create_path (char *host, char *path, uuid_t uuid, mode_t mode, - char **op_errstr) +/*path needs to be absolute; works only on gfid, volume-id*/ +static int +glusterd_is_uuid_present (char *path, char *xattr, gf_boolean_t *present) { - int ret = -1; - char msg[2048] = {0}; - struct stat st_buf = {0}; - uuid_t gfid = {0,}; - uuid_t old_uuid = {0,}; - char old_uuid_buf[64] = {0,}; - - ret = stat (path, &st_buf); - if ((!ret) && (!S_ISDIR (st_buf.st_mode))) { - snprintf (msg, sizeof (msg), "brick %s:%s, " - "path %s is not a directory", host, path, path); - gf_log ("", GF_LOG_ERROR, "%s", msg); - ret = -1; + GF_ASSERT (path); + GF_ASSERT (xattr); + GF_ASSERT (present); + + int ret = -1; + uuid_t uid = {0,}; + + if (!path || !xattr || !present) goto out; - } else if (!ret) { - goto check_xattr; - } else { - ret = mkdir (path, mode); - if (ret) { - snprintf (msg, sizeof (msg), "brick: %s:%s, path " - "creation failed, reason: %s", - host, path, strerror(errno)); - gf_log ("glusterd", GF_LOG_ERROR, "%s", msg); + + ret = sys_lgetxattr (path, xattr, &uid, 16); + + if (ret >= 0) { + *present = _gf_true; + ret = 0; + goto out; + } + + switch (errno) { +#if defined(ENODATA) + case ENODATA: /* FALLTHROUGH */ +#endif +#if defined(ENOATTR) && (ENOATTR != ENODATA) + case ENOATTR: /* FALLTHROUGH */ +#endif + case ENOTSUP: + *present = _gf_false; + ret = 0; + break; + default: + break; + } +out: + return ret; +} + +/*path needs to be absolute*/ +static int +glusterd_is_path_in_use (char *path, gf_boolean_t *in_use, char **op_errstr) +{ + int i = 0; + int ret = -1; + gf_boolean_t used = _gf_false; + char dir[PATH_MAX] = {0,}; + char *curdir = NULL; + char msg[2048] = {0}; + char *keys[3] = {GFID_XATTR_KEY, + GF_XATTR_VOL_ID_KEY, + NULL}; + + GF_ASSERT (path); + if (!path) + goto out; + + strcpy (dir, path); + curdir = dir; + do { + for (i = 0; !used && keys[i]; i++) { + ret = glusterd_is_uuid_present (curdir, keys[i], &used); + if (ret) + goto out; + } + + if (used) + break; + + curdir = dirname (curdir); + if (!strcmp (curdir, ".")) goto out; + + + } while (strcmp (curdir, "/")); + + if (!strcmp (curdir, "/")) { + for (i = 0; !used && keys[i]; i++) { + ret = glusterd_is_uuid_present (curdir, keys[i], &used); + if (ret) + goto out; + } + } + + ret = 0; + *in_use = used; +out: + if (ret) { + snprintf (msg, sizeof (msg), "Failed to get extended " + "attribute %s, reason: %s", keys[i], + strerror (errno)); + } + + if (*in_use) { + if (!strcmp (path, curdir)) { + snprintf (msg, sizeof (msg), "%s is already part of a " + "volume", path); } else { - goto check_xattr; + snprintf (msg, sizeof (msg), "parent directory %s is " + "already part of a volume", curdir); } } -/* To check if filesystem is read-only - and if it supports extended attributes */ -check_xattr: + if (strlen (msg)) { + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + } + + return ret; +} + +int +glusterd_check_and_set_brick_xattr (char *host, char *path, uuid_t uuid, + char **op_errstr, gf_boolean_t is_force) +{ + int ret = -1; + char msg[2048] = {0,}; + gf_boolean_t in_use = _gf_false; + int flags = 0; + + /* Check for xattr support in backend fs */ ret = sys_lsetxattr (path, "trusted.glusterfs.test", "working", 8, 0); if (ret) { - snprintf (msg, sizeof (msg), "glusterfs is not" + snprintf (msg, sizeof (msg), "Glusterfs is not" " supported on brick: %s:%s.\nSetting" " extended attributes failed, reason:" " %s.", host, path, strerror(errno)); - gf_log ("glusterd", GF_LOG_ERROR, "%s", msg); goto out; + } else { - /* Remove xattr *cannot* fail after setting it succeeded */ sys_lremovexattr (path, "trusted.glusterfs.test"); } - /* Now check if the export directory has some other 'gfid', - other than that of root '/' */ - ret = sys_lgetxattr (path, "trusted.gfid", gfid, 16); - if (ret == 16) { - if (!__is_root_gfid (gfid)) { - gf_log (THIS->name, GF_LOG_WARNING, - "%s: gfid (%s) is not that of glusterfs '/' ", - path, uuid_utoa (gfid)); - snprintf (msg, sizeof (msg), - "'%s:%s' gfid (%s) is not that of " - "glusterfs '/' ", host, path, uuid_utoa (gfid)); - ret = -1; - goto out; - } - } else if (ret != -1) { - /* Wrong 'gfid' is set, it should be error */ - ret = -1; - snprintf (msg, sizeof (msg), "'%s:%s' has wrong entry" - "for 'gfid'.", host, path); - goto out; - } else if ((ret == -1) && (errno != ENODATA)) { - /* Wrong 'gfid' is set, it should be error */ - snprintf (msg, sizeof (msg), "'%s:%s' has failed to fetch " - "'gfid' (%s)", host, path, strerror (errno)); + ret = glusterd_is_path_in_use (path, &in_use, op_errstr); + if (ret) goto out; - } - ret = 0; - if (!uuid) - goto out; - - /* This 'key' is set when the volume is started for the first time */ - ret = sys_lgetxattr (path, "trusted.glusterfs.volume-id", - old_uuid, 16); - if (ret == 16) { - if (uuid_compare (old_uuid, uuid)) { - uuid_utoa_r (old_uuid, old_uuid_buf); - gf_log (THIS->name, GF_LOG_WARNING, - "%s: mismatching volume-id (%s) recieved. " - "already is a part of volume %s ", - path, uuid_utoa (uuid), old_uuid_buf); - snprintf (msg, sizeof (msg), "'%s:%s' has been part of " - "a deleted volume with id %s. Please " - "re-create the brick directory.", - host, path, old_uuid_buf); - ret = -1; - goto out; - } - } else if (ret != -1) { - /* Wrong 'volume-id' is set, it should be error */ + if (in_use && !is_force) { ret = -1; - snprintf (msg, sizeof (msg), "'%s:%s' has wrong entry" - "for 'volume-id'.", host, path); - goto out; - } else if ((ret == -1) && (errno != ENODATA)) { - /* Wrong 'volume-id' is set, it should be error */ - snprintf (msg, sizeof (msg), "'%s:%s' : failed to fetch " - "'volume-id' (%s)", host, path, strerror (errno)); goto out; + } + + if (!is_force) + flags = XATTR_CREATE; + + ret = sys_lsetxattr (path, GF_XATTR_VOL_ID_KEY, uuid, 16, + flags); + if (ret) { + snprintf (msg, sizeof (msg), "Failed to set extended " + "attributes %s, reason: %s", + GF_XATTR_VOL_ID_KEY, strerror (errno)); + goto out; } - /* if 'ret == -1' then 'volume-id' not set, seems to be a fresh - directory */ ret = 0; out: - if (msg[0] != '\0') + if (strlen (msg)) *op_errstr = gf_strdup (msg); - gf_log ("", GF_LOG_DEBUG, "returning %d", ret); return ret; } @@ -4029,9 +5796,8 @@ glusterd_sm_tr_log_transition_add_to_dict (dict_t *dict, { int ret = -1; char key[512] = {0}; - char timestr[256] = {0,}; + char timestr[64] = {0,}; char *str = NULL; - struct tm tm = {0}; GF_ASSERT (dict); GF_ASSERT (log); @@ -4060,9 +5826,8 @@ glusterd_sm_tr_log_transition_add_to_dict (dict_t *dict, memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "log%d-time", count); - localtime_r ((const time_t*)&log->transitions[i].time, &tm); - memset (timestr, 0, sizeof (timestr)); - strftime (timestr, 256, "%Y-%m-%d %H:%M:%S", &tm); + gf_time_fmt (timestr, sizeof timestr, log->transitions[i].time, + gf_timefmt_FT); str = gf_strdup (timestr); ret = dict_set_dynstr (dict, key, str); if (ret) @@ -4150,8 +5915,7 @@ glusterd_sm_tr_log_delete (glusterd_sm_tr_log_t *log) { if (!log) return; - if (log->transitions) - GF_FREE (log->transitions); + GF_FREE (log->transitions); return; } @@ -4163,6 +5927,10 @@ glusterd_sm_tr_log_transition_add (glusterd_sm_tr_log_t *log, glusterd_sm_transition_t *transitions = NULL; int ret = -1; int next = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (log); if (!log) @@ -4185,18 +5953,18 @@ glusterd_sm_tr_log_transition_add (glusterd_sm_tr_log_t *log, if (log->count < log->size) log->count++; ret = 0; - gf_log ("glusterd", GF_LOG_DEBUG, "Transitioning from '%s' to '%s' " + gf_log (this->name, GF_LOG_DEBUG, "Transitioning from '%s' to '%s' " "due to event '%s'", log->state_name_get (old_state), log->state_name_get (new_state), log->event_name_get (event)); out: - gf_log ("", GF_LOG_DEBUG, "returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret); return ret; } int glusterd_peerinfo_new (glusterd_peerinfo_t **peerinfo, - glusterd_friend_sm_state_t state, - uuid_t *uuid, const char *hostname) + glusterd_friend_sm_state_t state, uuid_t *uuid, + const char *hostname, int port) { glusterd_peerinfo_t *new_peer = NULL; int ret = -1; @@ -4226,6 +5994,9 @@ glusterd_peerinfo_new (glusterd_peerinfo_t **peerinfo, if (ret) goto out; + if (new_peer->state.state == GD_FRIEND_STATE_BEFRIENDED) + new_peer->quorum_contrib = QUORUM_WAITING; + new_peer->port = port; *peerinfo = new_peer; out: if (ret && new_peer) @@ -4249,8 +6020,7 @@ glusterd_peer_destroy (glusterd_peerinfo_t *peerinfo) } list_del_init (&peerinfo->uuid_list); - if (peerinfo->hostname) - GF_FREE (peerinfo->hostname); + GF_FREE (peerinfo->hostname); glusterd_sm_tr_log_delete (&peerinfo->sm_log); GF_FREE (peerinfo); peerinfo = NULL; @@ -4277,7 +6047,7 @@ glusterd_remove_pending_entry (struct list_head *list, void *elem) } } out: - gf_log ("", GF_LOG_DEBUG, "returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "returning %d", ret); return ret; } @@ -4319,7 +6089,7 @@ glusterd_delete_volume (glusterd_volinfo_t *volinfo) ret = glusterd_volinfo_delete (volinfo); out: - gf_log ("", GF_LOG_DEBUG, "returning %d", ret); + gf_log (THIS->name, GF_LOG_DEBUG, "returning %d", ret); return ret; } @@ -4328,18 +6098,15 @@ glusterd_delete_brick (glusterd_volinfo_t* volinfo, glusterd_brickinfo_t *brickinfo) { int ret = 0; + char voldir[PATH_MAX] = {0,}; + glusterd_conf_t *priv = THIS->private; GF_ASSERT (volinfo); GF_ASSERT (brickinfo); -#ifdef DEBUG - ret = glusterd_volume_brickinfo_get (brickinfo->uuid, - brickinfo->hostname, - brickinfo->path, volinfo, - NULL, GF_PATH_COMPLETE); - GF_ASSERT (0 == ret); -#endif + GLUSTERD_GET_VOLUME_DIR(voldir, volinfo, priv); + glusterd_delete_volfile (volinfo, brickinfo); - glusterd_store_delete_brick (volinfo, brickinfo); + glusterd_store_delete_brick (brickinfo, voldir); glusterd_brickinfo_delete (brickinfo); volinfo->brick_count--; return ret; @@ -4361,28 +6128,92 @@ glusterd_delete_all_bricks (glusterd_volinfo_t* volinfo) } int -mkdir_if_missing (char *path) -{ - struct stat st = {0,}; - int ret = 0; +glusterd_get_local_brickpaths (glusterd_volinfo_t *volinfo, char **pathlist) +{ + char **path_tokens = NULL; + char *tmp_path_list = NULL; + char path[PATH_MAX] = ""; + int32_t count = 0; + int32_t pathlen = 0; + int32_t total_len = 0; + int32_t ret = 0; + int i = 0; + glusterd_brickinfo_t *brickinfo = NULL; + + if ((!volinfo) || (!pathlist)) + goto out; + + path_tokens = GF_CALLOC (sizeof(char*), volinfo->brick_count, + gf_gld_mt_charptr); + if (!path_tokens) { + gf_log ("", GF_LOG_DEBUG, "Could not allocate memory."); + ret = -1; + goto out; + } + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (uuid_compare (brickinfo->uuid, MY_UUID)) + continue; - ret = mkdir (path, 0777); - if (!ret || errno == EEXIST) - ret = stat (path, &st); - if (ret == -1 || !S_ISDIR (st.st_mode)) - gf_log ("", GF_LOG_WARNING, "Failed to create the" - " directory %s", path); + pathlen = snprintf (path, sizeof(path), + "--path=%s ", brickinfo->path); + if (pathlen < sizeof(path)) + path[pathlen] = '\0'; + else + path[sizeof(path)-1] = '\0'; + path_tokens[count] = gf_strdup (path); + if (!path_tokens[count]) { + gf_log ("", GF_LOG_DEBUG, + "Could not allocate memory."); + ret = -1; + goto out; + } + count++; + total_len += pathlen; + } + + tmp_path_list = GF_CALLOC (sizeof(char), total_len + 1, + gf_gld_mt_char); + if (!tmp_path_list) { + gf_log ("", GF_LOG_DEBUG, "Could not allocate memory."); + ret = -1; + goto out; + } + + for (i = 0; i < count; i++) + strcat (tmp_path_list, path_tokens[i]); + + if (count) + *pathlist = tmp_path_list; + + ret = count; +out: + for (i = 0; i < count; i++) { + GF_FREE (path_tokens[i]); + path_tokens[i] = NULL; + } + + GF_FREE (path_tokens); + path_tokens = NULL; + + if (ret == 0) { + gf_log ("", GF_LOG_DEBUG, "No Local Bricks Present."); + GF_FREE (tmp_path_list); + tmp_path_list = NULL; + } + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } int glusterd_start_gsync (glusterd_volinfo_t *master_vol, char *slave, - char *glusterd_uuid_str, char **op_errstr) + char *path_list, char *conf_path, + char *glusterd_uuid_str, + char **op_errstr) { int32_t ret = 0; int32_t status = 0; - char buf[PATH_MAX] = {0,}; char uuid_str [64] = {0}; runner_t runner = {0,}; xlator_t *this = NULL; @@ -4394,48 +6225,46 @@ glusterd_start_gsync (glusterd_volinfo_t *master_vol, char *slave, priv = this->private; GF_ASSERT (priv); - uuid_utoa_r (priv->uuid, uuid_str); - if (strcmp (uuid_str, glusterd_uuid_str)) - goto out; + uuid_utoa_r (MY_UUID, uuid_str); - ret = gsync_status (master_vol->volname, slave, &status); - if (status == 0) - goto out; - - snprintf (buf, PATH_MAX, "%s/"GEOREP"/%s", priv->workdir, master_vol->volname); - ret = mkdir_if_missing (buf); - if (ret) { - errcode = -1; + if (!path_list) { + ret = 0; + gf_log ("", GF_LOG_DEBUG, "No Bricks in this node." + " Not starting gsyncd."); goto out; } - snprintf (buf, PATH_MAX, DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/%s", - master_vol->volname); - ret = mkdir_if_missing (buf); - if (ret) { - errcode = -1; + ret = gsync_status (master_vol->volname, slave, conf_path, &status); + if (status == 0) goto out; - } uuid_utoa_r (master_vol->volume_id, uuid_str); runinit (&runner); - runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL); - runner_argprintf (&runner, "%s/"GSYNC_CONF, priv->workdir); + runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", + path_list, "-c", NULL); + runner_argprintf (&runner, "%s", conf_path); runner_argprintf (&runner, ":%s", master_vol->volname); runner_add_args (&runner, slave, "--config-set", "session-owner", uuid_str, NULL); + synclock_unlock (&priv->big_lock); ret = runner_run (&runner); + synclock_lock (&priv->big_lock); if (ret == -1) { errcode = -1; goto out; } runinit (&runner); - runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "--monitor", "-c", NULL); - runner_argprintf (&runner, "%s/"GSYNC_CONF, priv->workdir); + runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", + path_list, "--monitor", "-c", NULL); + runner_argprintf (&runner, "%s", conf_path); runner_argprintf (&runner, ":%s", master_vol->volname); + runner_argprintf (&runner, "--glusterd-uuid=%s", + uuid_utoa (priv->uuid)); runner_add_arg (&runner, slave); + synclock_unlock (&priv->big_lock); ret = runner_run (&runner); + synclock_lock (&priv->big_lock); if (ret == -1) { gf_asprintf (op_errstr, GEOREP" start failed for %s %s", master_vol->volname, slave); @@ -4447,7 +6276,7 @@ glusterd_start_gsync (glusterd_volinfo_t *master_vol, char *slave, out: if ((ret != 0) && errcode == -1) { if (op_errstr) - *op_errstr = gf_strdup ("internal error, cannot start" + *op_errstr = gf_strdup ("internal error, cannot start " "the " GEOREP " session"); } @@ -4456,17 +6285,38 @@ out: } int32_t -glusterd_recreate_bricks (glusterd_conf_t *conf) +glusterd_recreate_volfiles (glusterd_conf_t *conf) { glusterd_volinfo_t *volinfo = NULL; int ret = 0; + int op_ret = 0; GF_ASSERT (conf); list_for_each_entry (volinfo, &conf->volumes, vol_list) { ret = generate_brick_volfiles (volinfo); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, "Failed to " + "regenerate brick volfiles for %s", + volinfo->volname); + op_ret = ret; + } + ret = generate_client_volfiles (volinfo, GF_CLIENT_TRUSTED); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, "Failed to " + "regenerate trusted client volfiles for %s", + volinfo->volname); + op_ret = ret; + } + ret = generate_client_volfiles (volinfo, GF_CLIENT_OTHER); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, "Failed to " + "regenerate client volfiles for %s", + volinfo->volname); + op_ret = ret; + } } - return ret; + return op_ret; } int32_t @@ -4476,7 +6326,8 @@ glusterd_handle_upgrade_downgrade (dict_t *options, glusterd_conf_t *conf) char *type = NULL; gf_boolean_t upgrade = _gf_false; gf_boolean_t downgrade = _gf_false; - gf_boolean_t regenerate_brick_volfiles = _gf_false; + gf_boolean_t regenerate_volfiles = _gf_false; + gf_boolean_t terminate = _gf_false; ret = dict_get_str (options, "upgrade", &type); if (!ret) { @@ -4488,7 +6339,7 @@ glusterd_handle_upgrade_downgrade (dict_t *options, glusterd_conf_t *conf) goto out; } if (_gf_true == upgrade) - regenerate_brick_volfiles = _gf_true; + regenerate_volfiles = _gf_true; } ret = dict_get_str (options, "downgrade", &type); @@ -4511,10 +6362,14 @@ glusterd_handle_upgrade_downgrade (dict_t *options, glusterd_conf_t *conf) if (!upgrade && !downgrade) ret = 0; - if (regenerate_brick_volfiles) { - ret = glusterd_recreate_bricks (conf); + else + terminate = _gf_true; + if (regenerate_volfiles) { + ret = glusterd_recreate_volfiles (conf); } out: + if (terminate && (ret == 0)) + kill (getpid(), SIGTERM); return ret; } @@ -4551,7 +6406,7 @@ glusterd_set_dump_options (char *dumpoptions_path, char *options, goto out; } dup_options = gf_strdup (options); - gf_log ("", GF_LOG_INFO, "Recieved following statedump options: %s", + gf_log ("", GF_LOG_INFO, "Received following statedump options: %s", dup_options); option = strtok_r (dup_options, " ", &tmpptr); while (option) { @@ -4572,8 +6427,7 @@ glusterd_set_dump_options (char *dumpoptions_path, char *options, out: if (fp) fclose (fp); - if (dup_options) - GF_FREE (dup_options); + GF_FREE (dup_options); return ret; } @@ -4586,7 +6440,6 @@ glusterd_brick_statedump (glusterd_volinfo_t *volinfo, xlator_t *this = NULL; glusterd_conf_t *conf = NULL; char pidfile_path[PATH_MAX] = {0,}; - char path[PATH_MAX] = {0,}; char dumpoptions_path[PATH_MAX] = {0,}; FILE *pidfile = NULL; pid_t pid = -1; @@ -4606,14 +6459,12 @@ glusterd_brick_statedump (glusterd_volinfo_t *volinfo, } } - if (uuid_compare (brickinfo->uuid, conf->uuid)) { + if (uuid_compare (brickinfo->uuid, MY_UUID)) { ret = 0; goto out; } - GLUSTERD_GET_VOLUME_DIR (path, volinfo, conf); - GLUSTERD_GET_BRICK_PIDFILE (pidfile_path, path, brickinfo->hostname, - brickinfo->path); + GLUSTERD_GET_BRICK_PIDFILE (pidfile_path, volinfo, brickinfo, conf); pidfile = fopen (pidfile_path, "r"); if (!pidfile) { @@ -4631,7 +6482,7 @@ glusterd_brick_statedump (glusterd_volinfo_t *volinfo, } snprintf (dumpoptions_path, sizeof (dumpoptions_path), - "/tmp/glusterdump.%d.options", pid); + DEFAULT_VAR_RUN_DIRECTORY"/glusterdump.%d.options", pid); ret = glusterd_set_dump_options (dumpoptions_path, options, option_cnt); if (ret < 0) { gf_log ("", GF_LOG_ERROR, "error while parsing the statedump " @@ -4704,7 +6555,7 @@ glusterd_nfs_statedump (char *options, int option_cnt, char **op_errstr) } snprintf (dumpoptions_path, sizeof (dumpoptions_path), - "/tmp/glusterdump.%d.options", pid); + DEFAULT_VAR_RUN_DIRECTORY"/glusterdump.%d.options", pid); ret = glusterd_set_dump_options (dumpoptions_path, options, option_cnt); if (ret < 0) { gf_log ("", GF_LOG_ERROR, "error while parsing the statedump " @@ -4725,8 +6576,7 @@ out: if (pidfile) fclose (pidfile); unlink (dumpoptions_path); - if (dup_options) - GF_FREE (dup_options); + GF_FREE (dup_options); return ret; } @@ -4813,3 +6663,2232 @@ glusterd_chk_peers_connected_befriended (uuid_t skip_uuid) (ret?"TRUE":"FALSE")); return ret; } + +void +glusterd_get_client_filepath (char *filepath, glusterd_volinfo_t *volinfo, + gf_transport_type type) +{ + char path[PATH_MAX] = {0,}; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + + GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv); + + if ((volinfo->transport_type == GF_TRANSPORT_BOTH_TCP_RDMA) && + (type == GF_TRANSPORT_RDMA)) + snprintf (filepath, PATH_MAX, "%s/%s.rdma-fuse.vol", + path, volinfo->volname); + else + snprintf (filepath, PATH_MAX, "%s/%s-fuse.vol", + path, volinfo->volname); +} + +void +glusterd_get_trusted_client_filepath (char *filepath, + glusterd_volinfo_t *volinfo, + gf_transport_type type) +{ + char path[PATH_MAX] = {0,}; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + + GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv); + + if ((volinfo->transport_type == GF_TRANSPORT_BOTH_TCP_RDMA) && + (type == GF_TRANSPORT_RDMA)) + snprintf (filepath, PATH_MAX, + "%s/trusted-%s.rdma-fuse.vol", + path, volinfo->volname); + else + snprintf (filepath, PATH_MAX, + "%s/trusted-%s-fuse.vol", + path, volinfo->volname); +} + +int +glusterd_volume_defrag_restart (glusterd_volinfo_t *volinfo, char *op_errstr, + size_t len, int cmd, defrag_cbk_fn_t cbk) +{ + glusterd_conf_t *priv = NULL; + char pidfile[PATH_MAX]; + int ret = -1; + pid_t pid; + + priv = THIS->private; + if (!priv) + return ret; + + GLUSTERD_GET_DEFRAG_PID_FILE(pidfile, volinfo, priv); + + if (!glusterd_is_service_running (pidfile, &pid)) { + glusterd_handle_defrag_start (volinfo, op_errstr, len, cmd, + cbk, volinfo->rebal.op); + } else { + glusterd_rebalance_rpc_create (volinfo, priv, cmd); + } + + return ret; +} + +int +glusterd_restart_rebalance (glusterd_conf_t *conf) +{ + glusterd_volinfo_t *volinfo = NULL; + int ret = 0; + char op_errstr[256]; + + list_for_each_entry (volinfo, &conf->volumes, vol_list) { + if (!volinfo->rebal.defrag_cmd) + continue; + glusterd_volume_defrag_restart (volinfo, op_errstr, 256, + volinfo->rebal.defrag_cmd, NULL); + } + return ret; +} + +void +glusterd_volinfo_reset_defrag_stats (glusterd_volinfo_t *volinfo) +{ + glusterd_rebalance_t *rebal = NULL; + GF_ASSERT (volinfo); + + rebal = &volinfo->rebal; + rebal->rebalance_files = 0; + rebal->rebalance_data = 0; + rebal->lookedup_files = 0; + rebal->rebalance_failures = 0; + rebal->rebalance_time = 0; + rebal->skipped_files = 0; + +} + +/* Return hostname for given uuid if it exists + * else return NULL + */ +char * +glusterd_uuid_to_hostname (uuid_t uuid) +{ + char *hostname = NULL; + glusterd_conf_t *priv = NULL; + glusterd_peerinfo_t *entry = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + + if (!uuid_compare (MY_UUID, uuid)) { + hostname = gf_strdup ("localhost"); + } + if (!list_empty (&priv->peers)) { + list_for_each_entry (entry, &priv->peers, uuid_list) { + if (!uuid_compare (entry->uuid, uuid)) { + hostname = gf_strdup (entry->hostname); + break; + } + } + } + + return hostname; +} + +gf_boolean_t +glusterd_is_local_brick (xlator_t *this, glusterd_volinfo_t *volinfo, + glusterd_brickinfo_t *brickinfo) +{ + gf_boolean_t local = _gf_false; + int ret = 0; + glusterd_conf_t *conf = NULL; + + if (uuid_is_null (brickinfo->uuid)) { + ret = glusterd_resolve_brick (brickinfo); + if (ret) + goto out; + } + conf = this->private; + local = !uuid_compare (brickinfo->uuid, MY_UUID); +out: + return local; +} +int +glusterd_validate_volume_id (dict_t *op_dict, glusterd_volinfo_t *volinfo) +{ + int ret = -1; + char *volid_str = NULL; + uuid_t vol_uid = {0, }; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + ret = dict_get_str (op_dict, "vol-id", &volid_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volume id for " + "volume %s", volinfo->volname); + goto out; + } + ret = uuid_parse (volid_str, vol_uid); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to parse volume id " + "for volume %s", volinfo->volname); + goto out; + } + + if (uuid_compare (vol_uid, volinfo->volume_id)) { + gf_log (this->name, GF_LOG_ERROR, "Volume ids of volume %s - %s" + " and %s - are different. Possibly a split brain among " + "peers.", volinfo->volname, volid_str, + uuid_utoa (volinfo->volume_id)); + ret = -1; + goto out; + } + +out: + return ret; +} + +int +glusterd_defrag_volume_status_update (glusterd_volinfo_t *volinfo, + dict_t *rsp_dict) +{ + int ret = 0; + uint64_t files = 0; + uint64_t size = 0; + uint64_t lookup = 0; + gf_defrag_status_t status = GF_DEFRAG_STATUS_NOT_STARTED; + uint64_t failures = 0; + uint64_t skipped = 0; + xlator_t *this = NULL; + double run_time = 0; + + this = THIS; + + ret = dict_get_uint64 (rsp_dict, "files", &files); + if (ret) + gf_log (this->name, GF_LOG_TRACE, + "failed to get file count"); + + ret = dict_get_uint64 (rsp_dict, "size", &size); + if (ret) + gf_log (this->name, GF_LOG_TRACE, + "failed to get size of xfer"); + + ret = dict_get_uint64 (rsp_dict, "lookups", &lookup); + if (ret) + gf_log (this->name, GF_LOG_TRACE, + "failed to get lookedup file count"); + + ret = dict_get_int32 (rsp_dict, "status", (int32_t *)&status); + if (ret) + gf_log (this->name, GF_LOG_TRACE, + "failed to get status"); + + ret = dict_get_uint64 (rsp_dict, "failures", &failures); + if (ret) + gf_log (this->name, GF_LOG_TRACE, + "failed to get failure count"); + + ret = dict_get_uint64 (rsp_dict, "skipped", &skipped); + if (ret) + gf_log (this->name, GF_LOG_TRACE, + "failed to get skipped count"); + + ret = dict_get_double (rsp_dict, "run-time", &run_time); + if (ret) + gf_log (this->name, GF_LOG_TRACE, + "failed to get run-time"); + + if (files) + volinfo->rebal.rebalance_files = files; + if (size) + volinfo->rebal.rebalance_data = size; + if (lookup) + volinfo->rebal.lookedup_files = lookup; + if (status) + volinfo->rebal.defrag_status = status; + if (failures) + volinfo->rebal.rebalance_failures = failures; + if (skipped) + volinfo->rebal.skipped_files = skipped; + if (run_time) + volinfo->rebal.rebalance_time = run_time; + + return ret; +} + +int +glusterd_check_topology_identical (const char *filename1, + const char *filename2, + gf_boolean_t *identical) +{ + int ret = -1; /* FAILURE */ + xlator_t *this = NULL; + FILE *fp1 = NULL; + FILE *fp2 = NULL; + glusterfs_graph_t *grph1 = NULL; + glusterfs_graph_t *grph2 = NULL; + + if ((!filename1) || (!filename2) || (!identical)) + goto out; + + this = THIS; + + errno = 0; /* RESET the errno */ + + /* fopen() the volfile1 to create the graph */ + fp1 = fopen (filename1, "r"); + if (fp1 == NULL) { + gf_log (this->name, GF_LOG_ERROR, "fopen() on file: %s failed " + "(%s)", filename1, strerror (errno)); + goto out; + } + + /* fopen() the volfile2 to create the graph */ + fp2 = fopen (filename2, "r"); + if (fp2 == NULL) { + gf_log (this->name, GF_LOG_ERROR, "fopen() on file: %s failed " + "(%s)", filename2, strerror (errno)); + goto out; + } + + /* create the graph for filename1 */ + grph1 = glusterfs_graph_construct(fp1); + if (grph1 == NULL) + goto out; + + /* create the graph for filename2 */ + grph2 = glusterfs_graph_construct(fp2); + if (grph2 == NULL) + goto out; + + /* compare the graph topology */ + *identical = is_graph_topology_equal(grph1, grph2); + ret = 0; /* SUCCESS */ +out: + if (fp1) + fclose(fp1); + if (fp2) + fclose(fp2); + if (grph1) + glusterfs_graph_destroy(grph1); + if (grph2) + glusterfs_graph_destroy(grph2); + + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); + return ret; +} + +int +glusterd_check_files_identical (char *filename1, char *filename2, + gf_boolean_t *identical) +{ + int ret = -1; + struct stat buf1 = {0,}; + struct stat buf2 = {0,}; + uint32_t cksum1 = 0; + uint32_t cksum2 = 0; + xlator_t *this = NULL; + + GF_ASSERT (filename1); + GF_ASSERT (filename2); + GF_ASSERT (identical); + + this = THIS; + + ret = stat (filename1, &buf1); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "stat on file: %s failed " + "(%s)", filename1, strerror (errno)); + goto out; + } + + ret = stat (filename2, &buf2); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "stat on file: %s failed " + "(%s)", filename2, strerror (errno)); + goto out; + } + + if (buf1.st_size != buf2.st_size) { + *identical = _gf_false; + goto out; + } + + ret = get_checksum_for_path (filename1, &cksum1); + if (ret) + goto out; + + + ret = get_checksum_for_path (filename2, &cksum2); + if (ret) + goto out; + + if (cksum1 != cksum2) + *identical = _gf_false; + else + *identical = _gf_true; + +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); + return ret; +} + +int +glusterd_volset_help (dict_t *dict, char **op_errstr) +{ + int ret = -1; + gf_boolean_t xml_out = _gf_false; + xlator_t *this = NULL; + + this = THIS; + + if (!dict) { + if (!(dict = glusterd_op_get_ctx ())) { + ret = 0; + goto out; + } + } + + if (dict_get (dict, "help" )) { + xml_out = _gf_false; + + } else if (dict_get (dict, "help-xml" )) { + xml_out = _gf_true; +#if (HAVE_LIB_XML) + ret = 0; +#else + gf_log (this->name, GF_LOG_ERROR, + "libxml not present in the system"); + if (op_errstr) + *op_errstr = gf_strdup ("Error: xml libraries not " + "present to produce " + "xml-output"); + goto out; +#endif + + } else { + goto out; + } + + ret = glusterd_get_volopt_content (dict, xml_out); + if (ret && op_errstr) + *op_errstr = gf_strdup ("Failed to get volume options help"); + out: + + gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterd_to_cli (rpcsvc_request_t *req, gf_cli_rsp *arg, struct iovec *payload, + int payloadcount, struct iobref *iobref, xdrproc_t xdrproc, + dict_t *dict) +{ + int ret = -1; + char *cmd = NULL; + int op_ret = 0; + char *op_errstr = NULL; + int op_errno = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + op_ret = arg->op_ret; + op_errstr = arg->op_errstr; + op_errno = arg->op_errno; + + ret = dict_get_str (dict, "cmd-str", &cmd); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "Failed to get command " + "string"); + + if (cmd) { + if (op_ret) + gf_cmd_log ("", "%s : FAILED %s %s", cmd, + (op_errstr)? ":" : " ", + (op_errstr)? op_errstr : " "); + else + gf_cmd_log ("", "%s : SUCCESS", cmd); + } + + glusterd_submit_reply (req, arg, payload, payloadcount, iobref, + (xdrproc_t) xdrproc); + if (dict) + dict_unref (dict); + + return ret; +} + +static int32_t +glusterd_append_gsync_status (dict_t *dst, dict_t *src) +{ + int ret = 0; + char *stop_msg = NULL; + + ret = dict_get_str (src, "gsync-status", &stop_msg); + if (ret) { + ret = 0; + goto out; + } + + ret = dict_set_dynstr (dst, "gsync-status", gf_strdup (stop_msg)); + if (ret) { + gf_log ("glusterd", GF_LOG_WARNING, "Unable to set the stop" + "message in the ctx dictionary"); + goto out; + } + + ret = 0; + out: + gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + return ret; + +} + +static int32_t +glusterd_append_status_dicts (dict_t *dst, dict_t *src) +{ + int dst_count = 0; + int src_count = 0; + int i = 0; + int ret = 0; + char mst[PATH_MAX] = {0,}; + char slv[PATH_MAX] = {0, }; + char sts[PATH_MAX] = {0, }; + char nds[PATH_MAX] = {0, }; + char *mst_val = NULL; + char *slv_val = NULL; + char *sts_val = NULL; + char *nds_val = NULL; + + GF_ASSERT (dst); + + if (src == NULL) + goto out; + + ret = dict_get_int32 (dst, "gsync-count", &dst_count); + if (ret) + dst_count = 0; + + ret = dict_get_int32 (src, "gsync-count", &src_count); + if (ret || !src_count) { + gf_log ("", GF_LOG_DEBUG, "Source brick empty"); + ret = 0; + goto out; + } + + for (i = 1; i <= src_count; i++) { + snprintf (nds, sizeof(nds), "node%d", i); + snprintf (mst, sizeof(mst), "master%d", i); + snprintf (slv, sizeof(slv), "slave%d", i); + snprintf (sts, sizeof(sts), "status%d", i); + + ret = dict_get_str (src, nds, &nds_val); + if (ret) + goto out; + + ret = dict_get_str (src, mst, &mst_val); + if (ret) + goto out; + + ret = dict_get_str (src, slv, &slv_val); + if (ret) + goto out; + + ret = dict_get_str (src, sts, &sts_val); + if (ret) + goto out; + + snprintf (nds, sizeof(nds), "node%d", i+dst_count); + snprintf (mst, sizeof(mst), "master%d", i+dst_count); + snprintf (slv, sizeof(slv), "slave%d", i+dst_count); + snprintf (sts, sizeof(sts), "status%d", i+dst_count); + + ret = dict_set_dynstr (dst, nds, gf_strdup (nds_val)); + if (ret) + goto out; + + ret = dict_set_dynstr (dst, mst, gf_strdup (mst_val)); + if (ret) + goto out; + + ret = dict_set_dynstr (dst, slv, gf_strdup (slv_val)); + if (ret) + goto out; + + ret = dict_set_dynstr (dst, sts, gf_strdup (sts_val)); + if (ret) + goto out; + + } + + ret = dict_set_int32 (dst, "gsync-count", dst_count+src_count); + + out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; + +} + +int32_t +glusterd_gsync_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict, char *op_errstr) +{ + dict_t *ctx = NULL; + int ret = 0; + char *conf_path = NULL; + + if (aggr) { + ctx = aggr; + + } else { + ctx = glusterd_op_get_ctx (); + if (!ctx) { + gf_log ("", GF_LOG_ERROR, + "Operation Context is not present"); + GF_ASSERT (0); + } + } + + if (rsp_dict) { + ret = glusterd_append_status_dicts (ctx, rsp_dict); + if (ret) + goto out; + + ret = glusterd_append_gsync_status (ctx, rsp_dict); + if (ret) + goto out; + + ret = dict_get_str (rsp_dict, "conf_path", &conf_path); + if (!ret && conf_path) { + ret = dict_set_dynstr (ctx, "conf_path", + gf_strdup(conf_path)); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store conf path."); + goto out; + } + } + } + if ((op_errstr) && (strcmp ("", op_errstr))) { + ret = dict_set_dynstr (ctx, "errstr", gf_strdup(op_errstr)); + if (ret) + goto out; + } + + ret = 0; + out: + gf_log ("", GF_LOG_DEBUG, "Returning %d ", ret); + return ret; +} + +int32_t +glusterd_rb_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict) +{ + int32_t src_port = 0; + int32_t dst_port = 0; + int ret = 0; + dict_t *ctx = NULL; + + + if (aggr) { + ctx = aggr; + + } else { + ctx = glusterd_op_get_ctx (); + if (!ctx) { + gf_log ("", GF_LOG_ERROR, + "Operation Context is not present"); + GF_ASSERT (0); + } + } + + if (rsp_dict) { + ret = dict_get_int32 (rsp_dict, "src-brick-port", &src_port); + if (ret == 0) { + gf_log ("", GF_LOG_DEBUG, + "src-brick-port=%d found", src_port); + } + + ret = dict_get_int32 (rsp_dict, "dst-brick-port", &dst_port); + if (ret == 0) { + gf_log ("", GF_LOG_DEBUG, + "dst-brick-port=%d found", dst_port); + } + + } + + if (src_port) { + ret = dict_set_int32 (ctx, "src-brick-port", + src_port); + if (ret) { + gf_log ("", GF_LOG_DEBUG, + "Could not set src-brick"); + goto out; + } + } + + if (dst_port) { + ret = dict_set_int32 (ctx, "dst-brick-port", + dst_port); + if (ret) { + gf_log ("", GF_LOG_DEBUG, + "Could not set dst-brick"); + goto out; + } + + } + +out: + return ret; + +} + +int32_t +glusterd_sync_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict) +{ + int ret = 0; + + GF_ASSERT (rsp_dict); + + if (!rsp_dict) { + goto out; + } + + ret = glusterd_import_friend_volumes (rsp_dict); +out: + return ret; + +} + +static int +_profile_volume_add_friend_rsp (dict_t *this, char *key, data_t *value, + void *data) +{ + char new_key[256] = {0}; + glusterd_pr_brick_rsp_conv_t *rsp_ctx = NULL; + data_t *new_value = NULL; + int brick_count = 0; + char brick_key[256]; + + if (strcmp (key, "count") == 0) + return 0; + sscanf (key, "%d%s", &brick_count, brick_key); + rsp_ctx = data; + new_value = data_copy (value); + GF_ASSERT (new_value); + snprintf (new_key, sizeof (new_key), "%d%s", + rsp_ctx->count + brick_count, brick_key); + dict_set (rsp_ctx->dict, new_key, new_value); + return 0; +} + +int +glusterd_profile_volume_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict) +{ + int ret = 0; + glusterd_pr_brick_rsp_conv_t rsp_ctx = {0}; + int32_t brick_count = 0; + int32_t count = 0; + dict_t *ctx_dict = NULL; + glusterd_op_t op = GD_OP_NONE; + + GF_ASSERT (rsp_dict); + + ret = dict_get_int32 (rsp_dict, "count", &brick_count); + if (ret) { + ret = 0; //no bricks in the rsp + goto out; + } + + op = glusterd_op_get_op (); + GF_ASSERT (GD_OP_PROFILE_VOLUME == op); + if (aggr) { + ctx_dict = aggr; + + } else { + ctx_dict = glusterd_op_get_ctx (); + } + + ret = dict_get_int32 (ctx_dict, "count", &count); + rsp_ctx.count = count; + rsp_ctx.dict = ctx_dict; + dict_foreach (rsp_dict, _profile_volume_add_friend_rsp, &rsp_ctx); + dict_del (ctx_dict, "count"); + ret = dict_set_int32 (ctx_dict, "count", count + brick_count); +out: + return ret; +} + +static int +glusterd_volume_status_add_peer_rsp (dict_t *this, char *key, data_t *value, + void *data) +{ + glusterd_status_rsp_conv_t *rsp_ctx = NULL; + data_t *new_value = NULL; + char brick_key[1024] = {0,}; + char new_key[1024] = {0,}; + int32_t index = 0; + int32_t ret = 0; + + /* Skip the following keys, they are already present in the ctx_dict */ + if (!strcmp (key, "count") || !strcmp (key, "cmd") || + !strcmp (key, "brick-index-max") || !strcmp (key, "other-count")) + return 0; + + rsp_ctx = data; + new_value = data_copy (value); + GF_ASSERT (new_value); + + sscanf (key, "brick%d.%s", &index, brick_key); + + if (index > rsp_ctx->brick_index_max) { + snprintf (new_key, sizeof (new_key), "brick%d.%s", + index + rsp_ctx->other_count, brick_key); + } else { + strncpy (new_key, key, sizeof (new_key)); + new_key[sizeof (new_key) - 1] = 0; + } + + ret = dict_set (rsp_ctx->dict, new_key, new_value); + if (ret) + gf_log ("", GF_LOG_ERROR, "Unable to set key: %s in dict", + key); + + return 0; +} + +int +glusterd_volume_status_copy_to_op_ctx_dict (dict_t *aggr, dict_t *rsp_dict) +{ + int ret = 0; + glusterd_status_rsp_conv_t rsp_ctx = {0}; + int32_t cmd = GF_CLI_STATUS_NONE; + int32_t node_count = 0; + int32_t other_count = 0; + int32_t brick_index_max = -1; + int32_t rsp_node_count = 0; + int32_t rsp_other_count = 0; + int vol_count = -1; + int i = 0; + dict_t *ctx_dict = NULL; + char key[PATH_MAX] = {0,}; + char *volname = NULL; + + GF_ASSERT (rsp_dict); + + if (aggr) { + ctx_dict = aggr; + + } else { + ctx_dict = glusterd_op_get_ctx (GD_OP_STATUS_VOLUME); + + } + + ret = dict_get_int32 (ctx_dict, "cmd", &cmd); + if (ret) + goto out; + + if (cmd & GF_CLI_STATUS_ALL && is_origin_glusterd (ctx_dict)) { + ret = dict_get_int32 (rsp_dict, "vol_count", &vol_count); + if (ret == 0) { + ret = dict_set_int32 (ctx_dict, "vol_count", + vol_count); + if (ret) + goto out; + + for (i = 0; i < vol_count; i++) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "vol%d", i); + ret = dict_get_str (rsp_dict, key, &volname); + if (ret) + goto out; + + ret = dict_set_str (ctx_dict, key, volname); + if (ret) + goto out; + } + } + } + + if ((cmd & GF_CLI_STATUS_TASKS) != 0) { + dict_copy (rsp_dict, aggr); + ret = 0; + goto out; + } + + ret = dict_get_int32 (rsp_dict, "count", &rsp_node_count); + if (ret) { + ret = 0; //no bricks in the rsp + goto out; + } + + ret = dict_get_int32 (rsp_dict, "other-count", &rsp_other_count); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to get other count from rsp_dict"); + goto out; + } + + ret = dict_get_int32 (ctx_dict, "count", &node_count); + ret = dict_get_int32 (ctx_dict, "other-count", &other_count); + if (!dict_get (ctx_dict, "brick-index-max")) { + ret = dict_get_int32 (rsp_dict, "brick-index-max", &brick_index_max); + if (ret) + goto out; + ret = dict_set_int32 (ctx_dict, "brick-index-max", brick_index_max); + if (ret) + goto out; + + } else { + ret = dict_get_int32 (ctx_dict, "brick-index-max", &brick_index_max); + } + + rsp_ctx.count = node_count; + rsp_ctx.brick_index_max = brick_index_max; + rsp_ctx.other_count = other_count; + rsp_ctx.dict = ctx_dict; + + dict_foreach (rsp_dict, glusterd_volume_status_add_peer_rsp, &rsp_ctx); + + ret = dict_set_int32 (ctx_dict, "count", node_count + rsp_node_count); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to update node count"); + goto out; + } + + ret = dict_set_int32 (ctx_dict, "other-count", + (other_count + rsp_other_count)); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to update other-count"); +out: + return ret; +} + +int +glusterd_volume_rebalance_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict) +{ + char key[256] = {0,}; + char *node_uuid = NULL; + char *node_uuid_str = NULL; + char *volname = NULL; + dict_t *ctx_dict = NULL; + double elapsed_time = 0; + glusterd_conf_t *conf = NULL; + glusterd_op_t op = GD_OP_NONE; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = 0; + int32_t index = 0; + int32_t count = 0; + int32_t current_index = 2; + int32_t value32 = 0; + uint64_t value = 0; + char *peer_uuid_str = NULL; + + GF_ASSERT (rsp_dict); + conf = THIS->private; + + op = glusterd_op_get_op (); + GF_ASSERT ((GD_OP_REBALANCE == op) || + (GD_OP_DEFRAG_BRICK_VOLUME == op)); + + if (aggr) { + ctx_dict = aggr; + + } else { + ctx_dict = glusterd_op_get_ctx (op); + + } + + if (!ctx_dict) + goto out; + + ret = dict_get_str (ctx_dict, "volname", &volname); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + + if (ret) + goto out; + + ret = dict_get_int32 (rsp_dict, "count", &index); + if (ret) + gf_log ("", GF_LOG_ERROR, "failed to get index"); + + memset (key, 0, 256); + snprintf (key, 256, "node-uuid-%d", index); + ret = dict_get_str (rsp_dict, key, &node_uuid); + if (!ret) { + node_uuid_str = gf_strdup (node_uuid); + + /* Finding the index of the node-uuid in the peer-list */ + list_for_each_entry (peerinfo, &conf->peers, uuid_list) { + peer_uuid_str = gd_peer_uuid_str (peerinfo); + if (strcmp (peer_uuid_str, node_uuid_str) == 0) + break; + + current_index++; + } + + /* Setting the largest index value as the total count. */ + ret = dict_get_int32 (ctx_dict, "count", &count); + if (count < current_index) { + ret = dict_set_int32 (ctx_dict, "count", current_index); + if (ret) + gf_log ("", GF_LOG_ERROR, "Failed to set count"); + } + + /* Setting the same index for the node, as is in the peerlist.*/ + memset (key, 0, 256); + snprintf (key, 256, "node-uuid-%d", current_index); + ret = dict_set_dynstr (ctx_dict, key, node_uuid_str); + if (ret) { + gf_log (THIS->name, GF_LOG_DEBUG, + "failed to set node-uuid"); + } + } + + snprintf (key, 256, "files-%d", index); + ret = dict_get_uint64 (rsp_dict, key, &value); + if (!ret) { + memset (key, 0, 256); + snprintf (key, 256, "files-%d", current_index); + ret = dict_set_uint64 (ctx_dict, key, value); + if (ret) { + gf_log (THIS->name, GF_LOG_DEBUG, + "failed to set the file count"); + } + } + + memset (key, 0, 256); + snprintf (key, 256, "size-%d", index); + ret = dict_get_uint64 (rsp_dict, key, &value); + if (!ret) { + memset (key, 0, 256); + snprintf (key, 256, "size-%d", current_index); + ret = dict_set_uint64 (ctx_dict, key, value); + if (ret) { + gf_log (THIS->name, GF_LOG_DEBUG, + "failed to set the size of migration"); + } + } + + memset (key, 0, 256); + snprintf (key, 256, "lookups-%d", index); + ret = dict_get_uint64 (rsp_dict, key, &value); + if (!ret) { + memset (key, 0, 256); + snprintf (key, 256, "lookups-%d", current_index); + ret = dict_set_uint64 (ctx_dict, key, value); + if (ret) { + gf_log (THIS->name, GF_LOG_DEBUG, + "failed to set lookuped file count"); + } + } + + memset (key, 0, 256); + snprintf (key, 256, "status-%d", index); + ret = dict_get_int32 (rsp_dict, key, &value32); + if (!ret) { + memset (key, 0, 256); + snprintf (key, 256, "status-%d", current_index); + ret = dict_set_int32 (ctx_dict, key, value32); + if (ret) { + gf_log (THIS->name, GF_LOG_DEBUG, + "failed to set status"); + } + } + + memset (key, 0, 256); + snprintf (key, 256, "failures-%d", index); + ret = dict_get_uint64 (rsp_dict, key, &value); + if (!ret) { + memset (key, 0, 256); + snprintf (key, 256, "failures-%d", current_index); + ret = dict_set_uint64 (ctx_dict, key, value); + if (ret) { + gf_log (THIS->name, GF_LOG_DEBUG, + "failed to set failure count"); + } + } + + memset (key, 0, 256); + snprintf (key, 256, "skipped-%d", index); + ret = dict_get_uint64 (rsp_dict, key, &value); + if (!ret) { + memset (key, 0, 256); + snprintf (key, 256, "skipped-%d", current_index); + ret = dict_set_uint64 (ctx_dict, key, value); + if (ret) { + gf_log (THIS->name, GF_LOG_DEBUG, + "failed to set skipped count"); + } + } + memset (key, 0, 256); + snprintf (key, 256, "run-time-%d", index); + ret = dict_get_double (rsp_dict, key, &elapsed_time); + if (!ret) { + memset (key, 0, 256); + snprintf (key, 256, "run-time-%d", current_index); + ret = dict_set_double (ctx_dict, key, elapsed_time); + if (ret) { + gf_log (THIS->name, GF_LOG_DEBUG, + "failed to set run-time"); + } + } + + ret = 0; + +out: + return ret; +} + +int +glusterd_snap_config_use_rsp_dict (dict_t *dst, dict_t *src) +{ + char buf[PATH_MAX] = ""; + char *volname = NULL; + int ret = -1; + int config_command = 0; + uint64_t i = 0; + uint64_t value = 0; + uint64_t voldisplaycount = 0; + + if (!dst || !src) { + gf_log ("", GF_LOG_ERROR, "Source or Destination " + "dict is empty."); + goto out; + } + + ret = dict_get_int32 (dst, "config-command", &config_command); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "failed to get config-command type"); + goto out; + } + + switch (config_command) { + case GF_SNAP_CONFIG_DISPLAY: + ret = dict_get_uint64 (src, "snap-max-hard-limit", &value); + if (!ret) { + ret = dict_set_uint64 (dst, "snap-max-hard-limit", value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set snap_max_hard_limit"); + goto out; + } + } else { + /* Received dummy response from other nodes */ + ret = 0; + goto out; + } + + ret = dict_get_uint64 (src, "snap-max-soft-limit", &value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get snap_max_soft_limit"); + goto out; + } + + ret = dict_set_uint64 (dst, "snap-max-soft-limit", value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set snap_max_soft_limit"); + goto out; + } + + ret = dict_get_uint64 (src, "voldisplaycount", + &voldisplaycount); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get voldisplaycount"); + goto out; + } + + ret = dict_set_uint64 (dst, "voldisplaycount", + voldisplaycount); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set voldisplaycount"); + goto out; + } + + for (i = 0; i < voldisplaycount; i++) { + snprintf (buf, sizeof(buf), "volume%ld-volname", i); + ret = dict_get_str (src, buf, &volname); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get %s", buf); + goto out; + } + ret = dict_set_str (dst, buf, volname); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-hard-limit", i); + ret = dict_get_uint64 (src, buf, &value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get %s", buf); + goto out; + } + ret = dict_set_uint64 (dst, buf, value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-active-hard-limit", i); + ret = dict_get_uint64 (src, buf, &value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get %s", buf); + goto out; + } + ret = dict_set_uint64 (dst, buf, value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-soft-limit", i); + ret = dict_get_uint64 (src, buf, &value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get %s", buf); + goto out; + } + ret = dict_set_uint64 (dst, buf, value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set %s", buf); + goto out; + } + } + + break; + default: + break; + } + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +/* Aggregate missed_snap_counts from different nodes and save it * + * in the req_dict of the originator node */ +int +glusterd_snap_create_use_rsp_dict (dict_t *dst, dict_t *src) +{ + char *buf = NULL; + char *tmp_str = NULL; + char name_buf[PATH_MAX] = ""; + int32_t i = -1; + int32_t ret = -1; + int32_t src_missed_snap_count = -1; + int32_t dst_missed_snap_count = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + if (!dst || !src) { + gf_log (this->name, GF_LOG_ERROR, "Source or Destination " + "dict is empty."); + goto out; + } + + ret = dict_get_int32 (src, "missed_snap_count", + &src_missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "No missed snaps"); + ret = 0; + goto out; + } + + ret = dict_get_int32 (dst, "missed_snap_count", + &dst_missed_snap_count); + if (ret) { + /* Initialize dst_missed_count for the first time */ + dst_missed_snap_count = 0; + } + + for (i = 0; i < src_missed_snap_count; i++) { + snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d", + i); + ret = dict_get_str (src, name_buf, &buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch %s", name_buf); + goto out; + } + + snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d", + dst_missed_snap_count); + + tmp_str = gf_strdup (buf); + if (!tmp_str) { + ret = -1; + goto out; + } + + ret = dict_set_dynstr (dst, name_buf, tmp_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set %s", name_buf); + goto out; + } + + tmp_str = NULL; + dst_missed_snap_count++; + } + + ret = dict_set_int32 (dst, "missed_snap_count", dst_missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set dst_missed_snap_count"); + goto out; + } + +out: + if (ret && tmp_str) + GF_FREE(tmp_str); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_snap_use_rsp_dict (dict_t *dst, dict_t *src) +{ + int ret = -1; + int32_t snap_command = 0; + + if (!dst || !src) { + gf_log ("", GF_LOG_ERROR, "Source or Destination " + "dict is empty."); + goto out; + } + + ret = dict_get_int32 (dst, "type", &snap_command); + if (ret) { + gf_log ("", GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case GF_SNAP_OPTION_TYPE_CREATE: + case GF_SNAP_OPTION_TYPE_DELETE: + ret = glusterd_snap_create_use_rsp_dict (dst, src); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to use rsp dict"); + goto out; + } + break; + case GF_SNAP_OPTION_TYPE_CONFIG: + ret = glusterd_snap_config_use_rsp_dict (dst, src); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to use rsp dict"); + goto out; + } + break; + default: + // copy the response dictinary's contents to the dict to be + // sent back to the cli + dict_copy (src, dst); + break; + } + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterd_sys_exec_output_rsp_dict (dict_t *dst, dict_t *src) +{ + char output_name[PATH_MAX] = ""; + char *output = NULL; + int ret = 0; + int i = 0; + int len = 0; + int src_output_count = 0; + int dst_output_count = 0; + + if (!dst || !src) { + gf_log ("", GF_LOG_ERROR, "Source or Destination " + "dict is empty."); + goto out; + } + + ret = dict_get_int32 (dst, "output_count", &dst_output_count); + + ret = dict_get_int32 (src, "output_count", &src_output_count); + if (ret) { + gf_log ("", GF_LOG_DEBUG, "No output from source"); + ret = 0; + goto out; + } + + for (i = 1; i <= src_output_count; i++) { + len = snprintf (output_name, sizeof(output_name) - 1, + "output_%d", i); + output_name[len] = '\0'; + ret = dict_get_str (src, output_name, &output); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to fetch %s", + output_name); + goto out; + } + + len = snprintf (output_name, sizeof(output_name) - 1, + "output_%d", i+dst_output_count); + output_name[len] = '\0'; + ret = dict_set_dynstr (dst, output_name, gf_strdup (output)); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to set %s", + output_name); + goto out; + } + } + + ret = dict_set_int32 (dst, "output_count", + dst_output_count+src_output_count); +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterd_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict) +{ + int ret = 0; + glusterd_op_t op = GD_OP_NONE; + + op = glusterd_op_get_op (); + GF_ASSERT (aggr); + GF_ASSERT (rsp_dict); + + if (!aggr) + goto out; + dict_copy (rsp_dict, aggr); +out: + return ret; +} + +int +glusterd_volume_heal_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict) +{ + int ret = 0; + dict_t *ctx_dict = NULL; + glusterd_op_t op = GD_OP_NONE; + + GF_ASSERT (rsp_dict); + + op = glusterd_op_get_op (); + GF_ASSERT (GD_OP_HEAL_VOLUME == op); + + if (aggr) { + ctx_dict = aggr; + + } else { + ctx_dict = glusterd_op_get_ctx (op); + } + + if (!ctx_dict) + goto out; + dict_copy (rsp_dict, ctx_dict); +out: + return ret; +} + +int +_profile_volume_add_brick_rsp (dict_t *this, char *key, data_t *value, + void *data) +{ + char new_key[256] = {0}; + glusterd_pr_brick_rsp_conv_t *rsp_ctx = NULL; + data_t *new_value = NULL; + + rsp_ctx = data; + new_value = data_copy (value); + GF_ASSERT (new_value); + snprintf (new_key, sizeof (new_key), "%d-%s", rsp_ctx->count, key); + dict_set (rsp_ctx->dict, new_key, new_value); + return 0; +} + +int +glusterd_profile_volume_brick_rsp (void *pending_entry, + dict_t *rsp_dict, dict_t *op_ctx, + char **op_errstr, gd_node_type type) +{ + int ret = 0; + glusterd_pr_brick_rsp_conv_t rsp_ctx = {0}; + int32_t count = 0; + char brick[PATH_MAX+1024] = {0}; + char key[256] = {0}; + char *full_brick = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + + GF_ASSERT (rsp_dict); + GF_ASSERT (op_ctx); + GF_ASSERT (op_errstr); + GF_ASSERT (pending_entry); + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_int32 (op_ctx, "count", &count); + if (ret) { + count = 1; + } else { + count++; + } + snprintf (key, sizeof (key), "%d-brick", count); + if (type == GD_NODE_BRICK) { + brickinfo = pending_entry; + snprintf (brick, sizeof (brick), "%s:%s", brickinfo->hostname, + brickinfo->path); + } else if (type == GD_NODE_NFS) { + snprintf (brick, sizeof (brick), "%s", uuid_utoa (MY_UUID)); + } + full_brick = gf_strdup (brick); + GF_ASSERT (full_brick); + ret = dict_set_dynstr (op_ctx, key, full_brick); + + rsp_ctx.count = count; + rsp_ctx.dict = op_ctx; + dict_foreach (rsp_dict, _profile_volume_add_brick_rsp, &rsp_ctx); + dict_del (op_ctx, "count"); + ret = dict_set_int32 (op_ctx, "count", count); + return ret; +} + +//input-key: <replica-id>:<child-id>-* +//output-key: <brick-id>-* +int +_heal_volume_add_shd_rsp (dict_t *this, char *key, data_t *value, void *data) +{ + char new_key[256] = {0,}; + char int_str[16] = {0}; + data_t *new_value = NULL; + char *rxl_end = NULL; + char *rxl_child_end = NULL; + glusterd_volinfo_t *volinfo = NULL; + int rxl_id = 0; + int rxl_child_id = 0; + int brick_id = 0; + int int_len = 0; + int ret = 0; + glusterd_heal_rsp_conv_t *rsp_ctx = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + + rsp_ctx = data; + rxl_end = strchr (key, '-'); + if (!rxl_end) + goto out; + + int_len = strlen (key) - strlen (rxl_end); + strncpy (int_str, key, int_len); + int_str[int_len] = '\0'; + ret = gf_string2int (int_str, &rxl_id); + if (ret) + goto out; + + rxl_child_end = strchr (rxl_end + 1, '-'); + if (!rxl_child_end) + goto out; + + int_len = strlen (rxl_end) - strlen (rxl_child_end) - 1; + strncpy (int_str, rxl_end + 1, int_len); + int_str[int_len] = '\0'; + ret = gf_string2int (int_str, &rxl_child_id); + if (ret) + goto out; + + volinfo = rsp_ctx->volinfo; + brick_id = rxl_id * volinfo->replica_count + rxl_child_id; + + if (!strcmp (rxl_child_end, "-status")) { + brickinfo = glusterd_get_brickinfo_by_position (volinfo, + brick_id); + if (!brickinfo) + goto out; + if (!glusterd_is_local_brick (rsp_ctx->this, volinfo, + brickinfo)) + goto out; + } + new_value = data_copy (value); + snprintf (new_key, sizeof (new_key), "%d%s", brick_id, rxl_child_end); + dict_set (rsp_ctx->dict, new_key, new_value); + +out: + return 0; +} + +int +_heal_volume_add_shd_rsp_of_statistics (dict_t *this, char *key, data_t + *value, void *data) +{ + char new_key[256] = {0,}; + char int_str[16] = {0,}; + char key_begin_string[128] = {0,}; + data_t *new_value = NULL; + char *rxl_end = NULL; + char *rxl_child_end = NULL; + glusterd_volinfo_t *volinfo = NULL; + char *key_begin_str = NULL; + int rxl_id = 0; + int rxl_child_id = 0; + int brick_id = 0; + int int_len = 0; + int ret = 0; + glusterd_heal_rsp_conv_t *rsp_ctx = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + + rsp_ctx = data; + key_begin_str = strchr (key, '-'); + if (!key_begin_str) + goto out; + + int_len = strlen (key) - strlen (key_begin_str); + strncpy (key_begin_string, key, int_len); + key_begin_string[int_len] = '\0'; + + rxl_end = strchr (key_begin_str + 1, '-'); + if (!rxl_end) + goto out; + + int_len = strlen (key_begin_str) - strlen (rxl_end) - 1; + strncpy (int_str, key_begin_str + 1, int_len); + int_str[int_len] = '\0'; + ret = gf_string2int (int_str, &rxl_id); + if (ret) + goto out; + + + rxl_child_end = strchr (rxl_end + 1, '-'); + if (!rxl_child_end) + goto out; + + int_len = strlen (rxl_end) - strlen (rxl_child_end) - 1; + strncpy (int_str, rxl_end + 1, int_len); + int_str[int_len] = '\0'; + ret = gf_string2int (int_str, &rxl_child_id); + if (ret) + goto out; + + volinfo = rsp_ctx->volinfo; + brick_id = rxl_id * volinfo->replica_count + rxl_child_id; + + brickinfo = glusterd_get_brickinfo_by_position (volinfo, brick_id); + if (!brickinfo) + goto out; + if (!glusterd_is_local_brick (rsp_ctx->this, volinfo, brickinfo)) + goto out; + + new_value = data_copy (value); + snprintf (new_key, sizeof (new_key), "%s-%d%s", key_begin_string, + brick_id, rxl_child_end); + dict_set (rsp_ctx->dict, new_key, new_value); + +out: + return 0; + +} + +int +glusterd_heal_volume_brick_rsp (dict_t *req_dict, dict_t *rsp_dict, + dict_t *op_ctx, char **op_errstr) +{ + int ret = 0; + glusterd_heal_rsp_conv_t rsp_ctx = {0}; + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + int heal_op = -1; + + GF_ASSERT (rsp_dict); + GF_ASSERT (op_ctx); + GF_ASSERT (op_errstr); + + ret = dict_get_str (req_dict, "volname", &volname); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + goto out; + } + + ret = dict_get_int32 (req_dict, "heal-op", &heal_op); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get heal_op"); + goto out; + } + + + ret = glusterd_volinfo_find (volname, &volinfo); + + if (ret) + goto out; + + rsp_ctx.dict = op_ctx; + rsp_ctx.volinfo = volinfo; + rsp_ctx.this = THIS; + if (heal_op == GF_AFR_OP_STATISTICS) + dict_foreach (rsp_dict, _heal_volume_add_shd_rsp_of_statistics, + &rsp_ctx); + else + dict_foreach (rsp_dict, _heal_volume_add_shd_rsp, &rsp_ctx); + + +out: + return ret; +} + +int +_status_volume_add_brick_rsp (dict_t *this, char *key, data_t *value, + void *data) +{ + char new_key[256] = {0,}; + data_t *new_value = 0; + glusterd_pr_brick_rsp_conv_t *rsp_ctx = NULL; + + rsp_ctx = data; + new_value = data_copy (value); + snprintf (new_key, sizeof (new_key), "brick%d.%s", rsp_ctx->count, key); + dict_set (rsp_ctx->dict, new_key, new_value); + + return 0; +} + +int +glusterd_status_volume_brick_rsp (dict_t *rsp_dict, dict_t *op_ctx, + char **op_errstr) +{ + int ret = 0; + glusterd_pr_brick_rsp_conv_t rsp_ctx = {0}; + int32_t count = 0; + int index = 0; + + GF_ASSERT (rsp_dict); + GF_ASSERT (op_ctx); + GF_ASSERT (op_errstr); + + ret = dict_get_int32 (op_ctx, "count", &count); + if (ret) { + count = 0; + } else { + count++; + } + ret = dict_get_int32 (rsp_dict, "index", &index); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Couldn't get node index"); + goto out; + } + dict_del (rsp_dict, "index"); + + rsp_ctx.count = index; + rsp_ctx.dict = op_ctx; + dict_foreach (rsp_dict, _status_volume_add_brick_rsp, &rsp_ctx); + ret = dict_set_int32 (op_ctx, "count", count); + +out: + return ret; +} + +int +glusterd_defrag_volume_node_rsp (dict_t *req_dict, dict_t *rsp_dict, + dict_t *op_ctx) +{ + int ret = 0; + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + char key[256] = {0,}; + int32_t i = 0; + char buf[1024] = {0,}; + char *node_str = NULL; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (req_dict); + + ret = dict_get_str (req_dict, "volname", &volname); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + + if (ret) + goto out; + + if (rsp_dict) { + ret = glusterd_defrag_volume_status_update (volinfo, + rsp_dict); + } + + if (!op_ctx) { + dict_copy (rsp_dict, op_ctx); + goto out; + } + + ret = dict_get_int32 (op_ctx, "count", &i); + i++; + + ret = dict_set_int32 (op_ctx, "count", i); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, "Failed to set count"); + + snprintf (buf, 1024, "%s", uuid_utoa (MY_UUID)); + node_str = gf_strdup (buf); + + snprintf (key, 256, "node-uuid-%d",i); + ret = dict_set_dynstr (op_ctx, key, node_str); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "failed to set node-uuid"); + + memset (key, 0 , 256); + snprintf (key, 256, "files-%d", i); + ret = dict_set_uint64 (op_ctx, key, volinfo->rebal.rebalance_files); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "failed to set file count"); + + memset (key, 0 , 256); + snprintf (key, 256, "size-%d", i); + ret = dict_set_uint64 (op_ctx, key, volinfo->rebal.rebalance_data); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "failed to set size of xfer"); + + memset (key, 0 , 256); + snprintf (key, 256, "lookups-%d", i); + ret = dict_set_uint64 (op_ctx, key, volinfo->rebal.lookedup_files); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "failed to set lookedup file count"); + + memset (key, 0 , 256); + snprintf (key, 256, "status-%d", i); + ret = dict_set_int32 (op_ctx, key, volinfo->rebal.defrag_status); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "failed to set status"); + + memset (key, 0 , 256); + snprintf (key, 256, "failures-%d", i); + ret = dict_set_uint64 (op_ctx, key, volinfo->rebal.rebalance_failures); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "failed to set failure count"); + + memset (key, 0 , 256); + snprintf (key, 256, "skipped-%d", i); + ret = dict_set_uint64 (op_ctx, key, volinfo->rebal.skipped_files); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "failed to set skipped count"); + + memset (key, 0, 256); + snprintf (key, 256, "run-time-%d", i); + ret = dict_set_double (op_ctx, key, volinfo->rebal.rebalance_time); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "failed to set run-time"); + +out: + return ret; +} +int32_t +glusterd_handle_node_rsp (dict_t *req_dict, void *pending_entry, + glusterd_op_t op, dict_t *rsp_dict, dict_t *op_ctx, + char **op_errstr, gd_node_type type) +{ + int ret = 0; + + GF_ASSERT (op_errstr); + + switch (op) { + case GD_OP_PROFILE_VOLUME: + ret = glusterd_profile_volume_brick_rsp (pending_entry, + rsp_dict, op_ctx, + op_errstr, type); + break; + case GD_OP_STATUS_VOLUME: + ret = glusterd_status_volume_brick_rsp (rsp_dict, op_ctx, + op_errstr); + break; + + case GD_OP_DEFRAG_BRICK_VOLUME: + glusterd_defrag_volume_node_rsp (req_dict, + rsp_dict, op_ctx); + break; + + case GD_OP_HEAL_VOLUME: + ret = glusterd_heal_volume_brick_rsp (req_dict, rsp_dict, + op_ctx, op_errstr); + break; + default: + break; + } + + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +/* Should be used only when an operation is in progress, as that is the only + * time a lock_owner is set + */ +gf_boolean_t +is_origin_glusterd (dict_t *dict) +{ + gf_boolean_t ret = _gf_false; + uuid_t lock_owner = {0,}; + uuid_t *originator_uuid = NULL; + + GF_ASSERT (dict); + + ret = dict_get_bin (dict, "originator_uuid", + (void **) &originator_uuid); + if (ret) { + /* If not originator_uuid has been set, then the command + * has been originated from a glusterd running on older version + * Hence fetching the lock owner */ + ret = glusterd_get_lock_owner (&lock_owner); + if (ret) { + ret = _gf_false; + goto out; + } + ret = !uuid_compare (MY_UUID, lock_owner); + } else + ret = !uuid_compare (MY_UUID, *originator_uuid); + +out: + return ret; +} + +int +glusterd_generate_and_set_task_id (dict_t *dict, char *key) +{ + int ret = -1; + uuid_t task_id = {0,}; + char *uuid_str = NULL; + xlator_t *this = NULL; + + GF_ASSERT (dict); + + this = THIS; + GF_ASSERT (this); + + uuid_generate (task_id); + uuid_str = gf_strdup (uuid_utoa (task_id)); + if (!uuid_str) { + ret = -1; + goto out; + } + + ret = dict_set_dynstr (dict, key, uuid_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set %s in dict", + key); + goto out; + } + gf_log (this->name, GF_LOG_INFO, "Generated task-id %s for key %s", + uuid_str, key); + +out: + if (ret) + GF_FREE (uuid_str); + return ret; +} + +int +glusterd_copy_uuid_to_dict (uuid_t uuid, dict_t *dict, char *key) +{ + int ret = -1; + char tmp_str[40] = {0,}; + char *task_id_str = NULL; + + GF_ASSERT (dict); + GF_ASSERT (key); + + uuid_unparse (uuid, tmp_str); + task_id_str = gf_strdup (tmp_str); + if (!task_id_str) + return -1; + + ret = dict_set_dynstr (dict, key, task_id_str); + if (ret) { + GF_FREE (task_id_str); + gf_log (THIS->name, GF_LOG_ERROR, + "Error setting uuid in dict with key %s", key); + } + + return 0; +} + +int +_update_volume_op_versions (dict_t *this, char *key, data_t *value, void *data) +{ + int op_version = 0; + glusterd_volinfo_t *ctx = NULL; + gf_boolean_t enabled = _gf_true; + int ret = -1; + + GF_ASSERT (data); + ctx = data; + + op_version = glusterd_get_op_version_for_key (key); + + if (gd_is_xlator_option (key) || gd_is_boolean_option (key)) { + ret = gf_string2boolean (value->data, &enabled); + if (ret) + return 0; + + if (!enabled) + return 0; + } + + if (op_version > ctx->op_version) + ctx->op_version = op_version; + + if (gd_is_client_option (key) && + (op_version > ctx->client_op_version)) + ctx->client_op_version = op_version; + + return 0; +} + +void +gd_update_volume_op_versions (glusterd_volinfo_t *volinfo) +{ + glusterd_conf_t *conf = NULL; + gf_boolean_t ob_enabled = _gf_false; + + GF_ASSERT (volinfo); + + conf = THIS->private; + GF_ASSERT (conf); + + /* Reset op-versions to minimum */ + volinfo->op_version = 1; + volinfo->client_op_version = 1; + + dict_foreach (volinfo->dict, _update_volume_op_versions, volinfo); + + /* Special case for open-behind + * If cluster op-version >= 2 and open-behind hasn't been explicitly + * disabled, volume op-versions must be updated to account for it + */ + + /* TODO: Remove once we have a general way to update automatically + * enabled features + */ + if (conf->op_version >= 2) { + ob_enabled = dict_get_str_boolean (volinfo->dict, + "performance.open-behind", + _gf_true); + if (ob_enabled) { + + if (volinfo->op_version < 2) + volinfo->op_version = 2; + if (volinfo->client_op_version < 2) + volinfo->client_op_version = 2; + } + } + + return; +} + +/* A task is committed/completed once the task-id for it is cleared */ +gf_boolean_t +gd_is_remove_brick_committed (glusterd_volinfo_t *volinfo) +{ + GF_ASSERT (volinfo); + + if ((GD_OP_REMOVE_BRICK == volinfo->rebal.op) && + !uuid_is_null (volinfo->rebal.rebalance_id)) + return _gf_false; + + return _gf_true; +} + +gf_boolean_t +glusterd_are_vol_all_peers_up (glusterd_volinfo_t *volinfo, + struct list_head *peers, + char **down_peerstr) +{ + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + gf_boolean_t ret = _gf_false; + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (!uuid_compare (brickinfo->uuid, MY_UUID)) + continue; + + list_for_each_entry (peerinfo, peers, uuid_list) { + if (uuid_compare (peerinfo->uuid, brickinfo->uuid)) + continue; + + /*Found peer who owns the brick, return false + * if peer is not connected or not friend */ + if (!(peerinfo->connected) || + (peerinfo->state.state != + GD_FRIEND_STATE_BEFRIENDED)) { + *down_peerstr = gf_strdup (peerinfo->hostname); + gf_log ("", GF_LOG_DEBUG, "Peer %s is down. ", + peerinfo->hostname); + goto out; + } + } + } + + ret = _gf_true; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +gf_boolean_t +glusterd_is_status_tasks_op (glusterd_op_t op, dict_t *dict) +{ + int ret = -1; + uint32_t cmd = GF_CLI_STATUS_NONE; + gf_boolean_t is_status_tasks = _gf_false; + + if (op != GD_OP_STATUS_VOLUME) + goto out; + + ret = dict_get_uint32 (dict, "cmd", &cmd); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to get opcode"); + goto out; + } + + if (cmd & GF_CLI_STATUS_TASKS) + is_status_tasks = _gf_true; + +out: + return is_status_tasks; +} + +int +glusterd_compare_snap_time(struct list_head *list1, struct list_head *list2) +{ + glusterd_snap_t *snap1 = NULL; + glusterd_snap_t *snap2 = NULL; + double diff_time = 0; + + GF_ASSERT (list1); + GF_ASSERT (list2); + + snap1 = list_entry(list1, glusterd_snap_t, snap_list); + snap2 = list_entry(list2, glusterd_snap_t, snap_list); + diff_time = difftime(snap1->time_stamp, snap2->time_stamp); + + return ((int)diff_time); +} + +int +glusterd_compare_snap_vol_time(struct list_head *list1, struct list_head *list2) +{ + glusterd_volinfo_t *snapvol1 = NULL; + glusterd_volinfo_t *snapvol2 = NULL; + double diff_time = 0; + + GF_ASSERT (list1); + GF_ASSERT (list2); + + snapvol1 = list_entry(list1, glusterd_volinfo_t, snapvol_list); + snapvol2 = list_entry(list2, glusterd_volinfo_t, snapvol_list); + diff_time = difftime(snapvol1->snapshot->time_stamp, + snapvol2->snapshot->time_stamp); + + return ((int)diff_time); +} + +int32_t +glusterd_missed_snapinfo_new (glusterd_missed_snap_info **missed_snapinfo) +{ + glusterd_missed_snap_info *new_missed_snapinfo = NULL; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (missed_snapinfo); + + new_missed_snapinfo = GF_CALLOC (1, sizeof(*new_missed_snapinfo), + gf_gld_mt_missed_snapinfo_t); + + if (!new_missed_snapinfo) + goto out; + + new_missed_snapinfo->node_snap_info = NULL; + INIT_LIST_HEAD (&new_missed_snapinfo->missed_snaps); + INIT_LIST_HEAD (&new_missed_snapinfo->snap_ops); + + *missed_snapinfo = new_missed_snapinfo; + + ret = 0; + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_missed_snap_op_new (glusterd_snap_op_t **snap_op) +{ + glusterd_snap_op_t *new_snap_op = NULL; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (snap_op); + + new_snap_op = GF_CALLOC (1, sizeof(*new_snap_op), + gf_gld_mt_missed_snapinfo_t); + + if (!new_snap_op) + goto out; + + new_snap_op->brick_path = NULL; + new_snap_op->brick_num = -1; + new_snap_op->op = -1; + new_snap_op->status = -1; + INIT_LIST_HEAD (&new_snap_op->snap_ops_list); + + *snap_op = new_snap_op; + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h index ba1b8f678..56bb799bf 100644 --- a/xlators/mgmt/glusterd/src/glusterd-utils.h +++ b/xlators/mgmt/glusterd/src/glusterd-utils.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _GLUSTERD_UTILS_H #define _GLUSTERD_UTILS_H_ @@ -38,24 +28,23 @@ #include "rpc-clnt.h" #include "protocol-common.h" -/* For brick search using path: partial or complete */ -typedef enum { - GF_PATH_COMPLETE = 0, - GF_PATH_PARTIAL -} gf_path_match_t; +#define GLUSTERD_SOCK_DIR "/var/run" struct glusterd_lock_ { uuid_t owner; time_t timestamp; }; -typedef struct glusterd_voldict_ctx_ { +typedef struct glusterd_dict_ctx_ { dict_t *dict; - int count; int opt_count; char *key_name; char *val_name; -} glusterd_voldict_ctx_t; + char *prefix; +} glusterd_dict_ctx_t; + +int +glusterd_compare_lines (const void *a, const void *b); typedef int (*glusterd_condition_func) (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo, @@ -80,6 +69,11 @@ glusterd_submit_reply (rpcsvc_request_t *req, void *arg, struct iobref *iobref, xdrproc_t xdrproc); int +glusterd_to_cli (rpcsvc_request_t *req, gf_cli_rsp *arg, struct iovec *payload, + int payloadcount, struct iobref *iobref, xdrproc_t xdrproc, + dict_t *dict); + +int glusterd_submit_request (struct rpc_clnt *rpc, void *req, call_frame_t *frame, rpc_clnt_prog_t *prog, int procnum, struct iobref *iobref, @@ -87,6 +81,26 @@ glusterd_submit_request (struct rpc_clnt *rpc, void *req, int32_t glusterd_volinfo_new (glusterd_volinfo_t **volinfo); +int32_t +glusterd_volinfo_dup (glusterd_volinfo_t *volinfo, + glusterd_volinfo_t **dup_volinfo, + gf_boolean_t set_userauth); + +char * +glusterd_auth_get_username (glusterd_volinfo_t *volinfo); + +char * +glusterd_auth_get_password (glusterd_volinfo_t *volinfo); + +int32_t +glusterd_auth_set_username (glusterd_volinfo_t *volinfo, char *username); + +int32_t +glusterd_auth_set_password (glusterd_volinfo_t *volinfo, char *password); + +void +glusterd_auth_cleanup (glusterd_volinfo_t *volinfo); + gf_boolean_t glusterd_check_volume_exists (char *volname); @@ -94,7 +108,7 @@ int32_t glusterd_brickinfo_new (glusterd_brickinfo_t **brickinfo); int32_t -glusterd_brickinfo_from_brick (char *brick, glusterd_brickinfo_t **brickinfo); +glusterd_brickinfo_new_from_brick (char *brick, glusterd_brickinfo_t **brickinfo); int32_t glusterd_friend_cleanup (glusterd_peerinfo_t *peerinfo); @@ -106,8 +120,23 @@ int32_t glusterd_peer_hostname_new (char *hostname, glusterd_peer_hostname_t **name); int32_t +glusterd_snap_volinfo_find (char *volname, glusterd_snap_t *snap, + glusterd_volinfo_t **volinfo); +int32_t +glusterd_snap_volinfo_find_from_parent_volname (char *origin_volname, + glusterd_snap_t *snap, + glusterd_volinfo_t **volinfo); + +int32_t glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo); +int +glusterd_volinfo_find_by_volume_id (uuid_t volume_id, glusterd_volinfo_t **volinfo); + +int +glusterd_snap_volinfo_find_by_volume_id (uuid_t volume_id, + glusterd_volinfo_t **volinfo); + int32_t glusterd_service_stop(const char *service, char *pidfile, int sig, gf_boolean_t force_kill); @@ -117,11 +146,13 @@ glusterd_resolve_brick (glusterd_brickinfo_t *brickinfo); int32_t glusterd_volume_start_glusterfs (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo); + glusterd_brickinfo_t *brickinfo, + gf_boolean_t wait); int32_t glusterd_volume_stop_glusterfs (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo); + glusterd_brickinfo_t *brickinfo, + gf_boolean_t del_brick); int32_t glusterd_volinfo_delete (glusterd_volinfo_t *volinfo); @@ -135,16 +166,13 @@ glusterd_is_cli_op_req (int32_t op); int32_t glusterd_volume_brickinfo_get_by_brick (char *brick, glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t **brickinfo, - gf_path_match_t path_match); -int32_t -glusterd_is_local_addr (char *hostname); + glusterd_brickinfo_t **brickinfo); int32_t glusterd_build_volume_dict (dict_t **vols); int32_t -glusterd_compare_friend_data (dict_t *vols, int32_t *status); +glusterd_compare_friend_data (dict_t *vols, int32_t *status, char *hostname); int glusterd_volume_compute_cksum (glusterd_volinfo_t *volinfo); @@ -177,27 +205,30 @@ glusterd_shd_start (); int32_t glusterd_shd_stop (); +void +glusterd_set_socket_filepath (char *sock_filepath, char *sockpath, size_t len); + int32_t -glusterd_shd_set_socket_filepath (char *rundir, uuid_t uuid, - char *socketpath, int len); +glusterd_nodesvc_set_socket_filepath (char *rundir, uuid_t uuid, + char *socketpath, int len); struct rpc_clnt* glusterd_pending_node_get_rpc (glusterd_pending_node_t *pending_node); struct rpc_clnt* -glusterd_shd_get_rpc (void); +glusterd_nodesvc_get_rpc (char *server); int32_t -glusterd_shd_set_rpc (struct rpc_clnt *rpc); +glusterd_nodesvc_set_rpc (char *server, struct rpc_clnt *rpc); int32_t -glusterd_shd_connect (char *socketpath); +glusterd_nodesvc_connect (char *server, char *socketpath); void -glusterd_shd_set_running (gf_boolean_t status); +glusterd_nodesvc_set_online_status (char *server, gf_boolean_t status); gf_boolean_t -glusterd_shd_is_running (); +glusterd_is_nodesvc_online (char *server); int glusterd_remote_hostname_get (rpcsvc_request_t *req, @@ -231,8 +262,8 @@ int32_t glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, dict_t *dict, int32_t count); int -glusterd_get_brickinfo (xlator_t *this, const char *brickname, - int port, gf_boolean_t localhost, +glusterd_get_brickinfo (xlator_t *this, const char *brickname, + int port, gf_boolean_t localhost, glusterd_brickinfo_t **brickinfo); void @@ -256,16 +287,19 @@ glusterd_all_volume_cond_check (glusterd_condition_func func, int status, void *ctx); int glusterd_brick_start (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo); + glusterd_brickinfo_t *brickinfo, + gf_boolean_t wait); int glusterd_brick_stop (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo); + glusterd_brickinfo_t *brickinfo, + gf_boolean_t del_brick); int glusterd_is_defrag_on (glusterd_volinfo_t *volinfo); int32_t glusterd_volinfo_bricks_delete (glusterd_volinfo_t *volinfo); + int glusterd_friend_find_by_uuid (uuid_t uuid, glusterd_peerinfo_t **peerinfo); @@ -274,11 +308,12 @@ glusterd_new_brick_validate (char *brick, glusterd_brickinfo_t *brickinfo, char *op_errstr, size_t len); int32_t glusterd_volume_brickinfos_delete (glusterd_volinfo_t *volinfo); + int32_t glusterd_volume_brickinfo_get (uuid_t uuid, char *hostname, char *path, glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t **brickinfo, - gf_path_match_t path_match); + glusterd_brickinfo_t **brickinfo); + int glusterd_brickinfo_get (uuid_t uuid, char *hostname, char *path, glusterd_brickinfo_t **brickinfo); @@ -295,22 +330,26 @@ gf_boolean_t glusterd_is_rb_ongoing (glusterd_volinfo_t *volinfo); int -glusterd_is_replace_running (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo); +glusterd_rb_check_bricks (glusterd_volinfo_t *volinfo, + glusterd_brickinfo_t *src_brick, + glusterd_brickinfo_t *dst_brick); int -glusterd_rb_check_bricks (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *src_brick, glusterd_brickinfo_t *dst_brick); +glusterd_check_and_set_brick_xattr (char *host, char *path, uuid_t uuid, + char **op_errstr, gf_boolean_t is_force); + int -glusterd_brick_create_path (char *host, char *path, uuid_t uuid, mode_t mode, - char **op_errstr); +glusterd_validate_and_create_brickpath (glusterd_brickinfo_t *brickinfo, + uuid_t volume_id, char **op_errstr, + gf_boolean_t is_force); int glusterd_sm_tr_log_transition_add (glusterd_sm_tr_log_t *log, int old_state, int new_state, int event); int glusterd_peerinfo_new (glusterd_peerinfo_t **peerinfo, - glusterd_friend_sm_state_t state, - uuid_t *uuid, const char *hostname); + glusterd_friend_sm_state_t state, uuid_t *uuid, + const char *hostname, int port); int glusterd_sm_tr_log_init (glusterd_sm_tr_log_t *log, char * (*state_name_get) (int), @@ -330,7 +369,7 @@ gf_boolean_t glusterd_peerinfo_is_uuid_unknown (glusterd_peerinfo_t *peerinfo); int32_t glusterd_brick_connect (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo); + glusterd_brickinfo_t *brickinfo, char *socketpath); int32_t glusterd_brick_disconnect (glusterd_brickinfo_t *brickinfo); int32_t @@ -338,22 +377,31 @@ glusterd_delete_volume (glusterd_volinfo_t *volinfo); int32_t glusterd_delete_brick (glusterd_volinfo_t* volinfo, glusterd_brickinfo_t *brickinfo); + int32_t glusterd_delete_all_bricks (glusterd_volinfo_t* volinfo); + +int +glusterd_spawn_daemons (void *opaque); + int glusterd_restart_gsyncds (glusterd_conf_t *conf); + int glusterd_start_gsync (glusterd_volinfo_t *master_vol, char *slave, - char *glusterd_uuid_str, char **op_errstr); + char *path_list, char *conf_path, + char *glusterd_uuid_str, + char **op_errstr); +int +glusterd_get_local_brickpaths (glusterd_volinfo_t *volinfo, + char **pathlist); + int32_t glusterd_recreate_bricks (glusterd_conf_t *conf); int32_t glusterd_handle_upgrade_downgrade (dict_t *options, glusterd_conf_t *conf); int -glusterd_get_brick_root (char *path, char **mount_point); - -int glusterd_add_brick_detail_to_dict (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo, dict_t *dict, int32_t count); @@ -388,4 +436,199 @@ glusterd_friend_remove_cleanup_vols (uuid_t uuid); gf_boolean_t glusterd_chk_peers_connected_befriended (uuid_t skip_uuid); + +void +glusterd_get_client_filepath (char *filepath, + glusterd_volinfo_t *volinfo, + gf_transport_type type); +void +glusterd_get_trusted_client_filepath (char *filepath, + glusterd_volinfo_t *volinfo, + gf_transport_type type); +int +glusterd_restart_rebalance (glusterd_conf_t *conf); + +int32_t +glusterd_add_bricks_hname_path_to_dict (dict_t *dict, + glusterd_volinfo_t *volinfo); + +int +glusterd_add_node_to_dict (char *server, dict_t *dict, int count, + dict_t *vol_opts); + +char * +glusterd_uuid_to_hostname (uuid_t uuid); + +int +glusterd_get_dist_leaf_count (glusterd_volinfo_t *volinfo); + +glusterd_brickinfo_t* +glusterd_get_brickinfo_by_position (glusterd_volinfo_t *volinfo, uint32_t pos); + +gf_boolean_t +glusterd_is_local_brick (xlator_t *this, glusterd_volinfo_t *volinfo, + glusterd_brickinfo_t *brickinfo); +int +glusterd_validate_volume_id (dict_t *op_dict, glusterd_volinfo_t *volinfo); + +int +glusterd_defrag_volume_status_update (glusterd_volinfo_t *volinfo, + dict_t *rsp_dict); + +int +glusterd_check_files_identical (char *filename1, char *filename2, + gf_boolean_t *identical); + +int +glusterd_check_topology_identical (const char *filename1, + const char *filename2, + gf_boolean_t *identical); + +void +glusterd_volinfo_reset_defrag_stats (glusterd_volinfo_t *volinfo); +int +glusterd_volset_help (dict_t *dict, char **op_errstr); + +int32_t +glusterd_sync_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int32_t +glusterd_gsync_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict, char *op_errstr); +int32_t +glusterd_rb_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int +glusterd_profile_volume_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int +glusterd_volume_status_copy_to_op_ctx_dict (dict_t *aggr, dict_t *rsp_dict); +int +glusterd_volume_rebalance_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int +glusterd_volume_heal_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int +glusterd_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int +glusterd_sys_exec_output_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int +glusterd_snap_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int32_t +glusterd_handle_node_rsp (dict_t *req_ctx, void *pending_entry, + glusterd_op_t op, dict_t *rsp_dict, dict_t *op_ctx, + char **op_errstr, gd_node_type type); +int +glusterd_volume_rebalance_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int +glusterd_volume_heal_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int +_profile_volume_add_brick_rsp (dict_t *this, char *key, data_t *value, + void *data); +int +glusterd_profile_volume_brick_rsp (void *pending_entry, + dict_t *rsp_dict, dict_t *op_ctx, + char **op_errstr, gd_node_type type); + +gf_boolean_t +glusterd_are_vol_all_peers_up (glusterd_volinfo_t *volinfo, + struct list_head *peers, + char **down_peerstr); + +/* Should be used only when an operation is in progress, as that is the only + * time a lock_owner is set + */ +gf_boolean_t +is_origin_glusterd (dict_t *dict); + +gf_boolean_t +glusterd_is_quorum_changed (dict_t *options, char *option, char *value); + +int +glusterd_do_quorum_action (); + +int +glusterd_get_quorum_cluster_counts (xlator_t *this, int *active_count, + int *quorum_count); + +int +glusterd_get_next_global_opt_version_str (dict_t *opts, char **version_str); +gf_boolean_t +glusterd_is_quorum_option (char *option); +gf_boolean_t +glusterd_is_volume_in_server_quorum (glusterd_volinfo_t *volinfo); +gf_boolean_t +glusterd_is_any_volume_in_server_quorum (xlator_t *this); +gf_boolean_t +does_gd_meet_server_quorum (xlator_t *this); + +int +glusterd_generate_and_set_task_id (dict_t *dict, char *key); + +int +glusterd_copy_uuid_to_dict (uuid_t uuid, dict_t *dict, char *key); + +gf_boolean_t +glusterd_is_same_address (char *name1, char *name2); + +void +gd_update_volume_op_versions (glusterd_volinfo_t *volinfo); + +char* +gd_peer_uuid_str (glusterd_peerinfo_t *peerinfo); + +gf_boolean_t +gd_is_remove_brick_committed (glusterd_volinfo_t *volinfo); + +gf_boolean_t +glusterd_are_vol_all_peers_up (glusterd_volinfo_t *volinfo, + struct list_head *peers, + char **down_peerstr); + +int +glusterd_get_slave_details_confpath (glusterd_volinfo_t *volinfo, dict_t *dict, + char **slave_ip, char **slave_vol, + char **conf_path, char **op_errstr); + +int +glusterd_check_restart_gsync_session (glusterd_volinfo_t *volinfo, char *slave, + dict_t *resp_dict, char *path_list, + char *conf_path, gf_boolean_t is_force); + +int +glusterd_check_gsync_running_local (char *master, char *slave, + char *conf_path, + gf_boolean_t *is_run); + +gf_boolean_t +glusterd_is_status_tasks_op (glusterd_op_t op, dict_t *dict); + +#ifdef GF_LINUX_HOST_OS +char* +glusterd_get_brick_mount_details (glusterd_brickinfo_t *brickinfo); +struct mntent * +glusterd_get_mnt_entry_info (char *mnt_pt, FILE *mtab); +int +glusterd_get_brick_root (char *path, char **mount_point); +#endif //LINUX_HOST + +int +glusterd_compare_snap_time(struct list_head *, struct list_head *); + +int +glusterd_compare_snap_vol_time(struct list_head *, struct list_head *); + +int32_t +glusterd_snap_volinfo_restore (dict_t *rsp_dict, + glusterd_volinfo_t *new_volinfo, + glusterd_volinfo_t *snap_volinfo); +int32_t +glusterd_lvm_snapshot_remove (dict_t *rsp_dict, glusterd_volinfo_t *snap_vol); + +int32_t +glusterd_missed_snapinfo_new (glusterd_missed_snap_info **missed_snapinfo); + +int32_t +glusterd_missed_snap_op_new (glusterd_snap_op_t **snap_op); + +int32_t +glusterd_add_missed_snaps_to_dict (dict_t *rsp_dict, char *snap_uuid, + glusterd_brickinfo_t *brickinfo, + int32_t brick_number, int32_t op); + #endif diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.c b/xlators/mgmt/glusterd/src/glusterd-volgen.c index f83a052ed..6f3c69e7d 100644 --- a/xlators/mgmt/glusterd/src/glusterd-volgen.c +++ b/xlators/mgmt/glusterd/src/glusterd-volgen.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H @@ -25,6 +15,7 @@ #include <fnmatch.h> #include <sys/wait.h> +#include <dlfcn.h> #if (HAVE_LIB_XML) #include <libxml/encoding.h> @@ -37,6 +28,8 @@ #include "logging.h" #include "dict.h" #include "graph-utils.h" +#include "glusterd-store.h" +#include "glusterd-hooks.h" #include "trie.h" #include "glusterd-mem-types.h" #include "cli1-xdr.h" @@ -44,165 +37,9 @@ #include "glusterd-op-sm.h" #include "glusterd-utils.h" #include "run.h" +#include "options.h" - -/* dispatch table for VOLUME SET - * ----------------------------- - * - * Format of entries: - * - * First field is the <key>, for the purpose of looking it up - * in volume dictionary. Each <key> is of the format "<domain>.<specifier>". - * - * Second field is <voltype>. - * - * Third field is <option>, if its unset, it's assumed to be - * the same as <specifier>. - * - * Fourth field is <value>. In this context they are used to specify - * a default. That is, even the volume dict doesn't have a value, - * we procced as if the default value were set for it. - * - * There are two type of entries: basic and special. - * - * - Basic entries are the ones where the <option> does _not_ start with - * the bang! character ('!'). - * - * In their case, <option> is understood as an option for an xlator of - * type <voltype>. Their effect is to copy over the volinfo->dict[<key>] - * value to all graph nodes of type <voltype> (if such a value is set). - * - * You are free to add entries of this type, they will become functional - * just by being present in the table. - * - * - Special entries where the <option> starts with the bang!. - * - * They are not applied to all graphs during generation, and you cannot - * extend them in a trivial way which could be just picked up. Better - * not touch them unless you know what you do. - * - * "NODOC" entries are not part of the public interface and are subject - * to change at any time. - * - * Another kind of grouping for options, according to visibility: - * - * - Exported: one which is used in the code. These are characterized by - * being used a macro as <key> (of the format VKEY_..., defined in - * glusterd-volgen.h - * - * - Non-exported: the rest; these have string literal <keys>. - * - * Adhering to this policy, option name changes shall be one-liners. - * - */ -typedef enum { DOC, NO_DOC, GLOBAL_DOC, GLOBAL_NO_DOC } option_type_t; - - -struct volopt_map_entry { - char *key; - char *voltype; - char *option; - char *value; - option_type_t type; - uint32_t flags; -}; - -static struct volopt_map_entry glusterd_volopt_map[] = { - - {"cluster.lookup-unhashed", "cluster/distribute", NULL, NULL, NO_DOC, 0 }, - {"cluster.min-free-disk", "cluster/distribute", NULL, NULL, NO_DOC, 0 }, - {"cluster.min-free-inodes", "cluster/distribute", NULL, NULL, NO_DOC, 0 }, - - {"cluster.entry-change-log", "cluster/replicate", NULL, NULL, NO_DOC, 0 }, - {"cluster.read-subvolume", "cluster/replicate", NULL, NULL, NO_DOC, 0 }, - {"cluster.background-self-heal-count", "cluster/replicate", NULL, NULL, NO_DOC, 0 }, - {"cluster.metadata-self-heal", "cluster/replicate", NULL, NULL, NO_DOC, 0 }, - {"cluster.data-self-heal", "cluster/replicate", NULL, NULL, NO_DOC, 0 }, - {"cluster.entry-self-heal", "cluster/replicate", NULL, NULL, NO_DOC, 0 }, - {"cluster.self-heal-daemon", "cluster/replicate", "!self-heal-daemon" , NULL, NO_DOC, 0 }, - {"cluster.strict-readdir", "cluster/replicate", NULL, NULL, NO_DOC, 0 }, - {"cluster.self-heal-window-size", "cluster/replicate", "data-self-heal-window-size", NULL, DOC, 0}, - {"cluster.data-change-log", "cluster/replicate", NULL, NULL, NO_DOC, 0 }, - {"cluster.metadata-change-log", "cluster/replicate", NULL, NULL, NO_DOC, 0 }, - {"cluster.data-self-heal-algorithm", "cluster/replicate", "data-self-heal-algorithm", NULL,DOC, 0}, - {"cluster.quorum-type", "cluster/replicate", "quorum-type", NULL, NO_DOC, 0}, - {"cluster.quorum-count", "cluster/replicate", "quorum-count", NULL, NO_DOC, 0}, - - {"cluster.stripe-block-size", "cluster/stripe", "block-size", NULL, DOC, 0}, - - {VKEY_DIAG_LAT_MEASUREMENT, "debug/io-stats", "latency-measurement", "off", NO_DOC, 0 }, - {"diagnostics.dump-fd-stats", "debug/io-stats", NULL, NULL, NO_DOC, 0 }, - {VKEY_DIAG_CNT_FOP_HITS, "debug/io-stats", "count-fop-hits", "off", NO_DOC, 0 }, - - {"diagnostics.brick-log-level", "debug/io-stats", "!brick-log-level", NULL, DOC, 0}, - {"diagnostics.client-log-level", "debug/io-stats", "!client-log-level", NULL, DOC, 0}, - {"diagnostics.brick-sys-log-level", "debug/io-stats", "!sys-log-level", NULL, DOC, 0}, - {"diagnostics.client-sys-log-level", "debug/io-stats", "!sys-log-level", NULL, DOC, 0}, - - {"performance.cache-max-file-size", "performance/io-cache", "max-file-size", NULL, DOC, 0}, - {"performance.cache-min-file-size", "performance/io-cache", "min-file-size", NULL, DOC, 0}, - {"performance.cache-refresh-timeout", "performance/io-cache", "cache-timeout", NULL, DOC, 0}, - {"performance.cache-priority", "performance/io-cache", "priority", NULL, DOC, 0}, - {"performance.cache-size", "performance/io-cache", NULL, NULL, NO_DOC, 0 }, - {"performance.cache-size", "performance/quick-read", NULL, NULL, NO_DOC, 0 }, - {"performance.flush-behind", "performance/write-behind", "flush-behind", NULL, DOC, 0}, - - {"performance.io-thread-count", "performance/io-threads", "thread-count", DOC, 0}, - - {"performance.disk-usage-limit", "performance/quota", NULL, NULL, NO_DOC, 0 }, - {"performance.min-free-disk-limit", "performance/quota", NULL, NULL, NO_DOC, 0 }, - - {"performance.write-behind-window-size", "performance/write-behind", "cache-size", NULL, DOC}, - - {"network.frame-timeout", "protocol/client", NULL, NULL, NO_DOC, 0 }, - {"network.ping-timeout", "protocol/client", NULL, NULL, NO_DOC, 0 }, - {"network.inode-lru-limit", "protocol/server", NULL, NULL, NO_DOC, 0 }, - - {"auth.allow", "protocol/server", "!server-auth", "*", DOC, 0}, - {"auth.reject", "protocol/server", "!server-auth", NULL, DOC, 0}, - - {"transport.keepalive", "protocol/server", "transport.socket.keepalive", NULL, NO_DOC, 0}, - {"server.allow-insecure", "protocol/server", "rpc-auth-allow-insecure", NULL, NO_DOC, 0}, - - {"performance.write-behind", "performance/write-behind", "!perf", "on", NO_DOC, 0}, - {"performance.read-ahead", "performance/read-ahead", "!perf", "on", NO_DOC, 0}, - {"performance.io-cache", "performance/io-cache", "!perf", "on", NO_DOC, 0}, - {"performance.quick-read", "performance/quick-read", "!perf", "on", NO_DOC, 0}, - {VKEY_PERF_STAT_PREFETCH, "performance/stat-prefetch", "!perf", "on", NO_DOC, 0}, - {"performance.client-io-threads", "performance/io-threads", "!perf", "off", NO_DOC, 0}, - {VKEY_MARKER_XTIME, "features/marker", "xtime", "off", NO_DOC, OPT_FLAG_FORCE}, - {VKEY_MARKER_XTIME, "features/marker", "!xtime", "off", NO_DOC, OPT_FLAG_FORCE}, - - {"nfs.enable-ino32", "nfs/server", "nfs.enable-ino32", NULL, GLOBAL_DOC, 0}, - {"nfs.mem-factor", "nfs/server", "nfs.mem-factor", NULL, GLOBAL_DOC, 0}, - {"nfs.export-dirs", "nfs/server", "nfs3.export-dirs", NULL, GLOBAL_DOC, 0}, - {"nfs.export-volumes", "nfs/server", "nfs3.export-volumes", NULL, GLOBAL_DOC, 0}, - {"nfs.addr-namelookup", "nfs/server", "rpc-auth.addr.namelookup", NULL, GLOBAL_DOC, 0}, - {"nfs.dynamic-volumes", "nfs/server", "nfs.dynamic-volumes", NULL, GLOBAL_NO_DOC, 0}, - {"nfs.register-with-portmap", "nfs/server", "rpc.register-with-portmap", NULL, GLOBAL_DOC, 0}, - {"nfs.port", "nfs/server", "nfs.port", NULL, GLOBAL_DOC, 0}, - - {"nfs.rpc-auth-unix", "nfs/server", "!rpc-auth.auth-unix.*", NULL, DOC, 0}, - {"nfs.rpc-auth-null", "nfs/server", "!rpc-auth.auth.null.*", NULL, DOC, 0}, - {"nfs.rpc-auth-allow", "nfs/server", "!rpc-auth.addr.*.allow", NULL, DOC, 0}, - {"nfs.rpc-auth-reject", "nfs/server", "!rpc-auth.addr.*.reject", NULL, DOC, 0}, - {"nfs.ports-insecure", "nfs/server", "!rpc-auth.ports.*.insecure", NULL, DOC, 0}, - {"nfs.transport-type", "nfs/server", "!nfs.transport-type", NULL, DOC, 0}, - - {"nfs.trusted-sync", "nfs/server", "!nfs3.*.trusted-sync", NULL, DOC, 0}, - {"nfs.trusted-write", "nfs/server", "!nfs3.*.trusted-write", NULL, DOC, 0}, - {"nfs.volume-access", "nfs/server", "!nfs3.*.volume-access", NULL, DOC, 0}, - {"nfs.export-dir", "nfs/server", "!nfs3.*.export-dir", NULL, DOC, 0}, - {"nfs.disable", "nfs/server", "!nfs-disable", NULL, DOC, 0}, - - {VKEY_FEATURES_QUOTA, "features/marker", "quota", "off", NO_DOC, OPT_FLAG_FORCE}, - {VKEY_FEATURES_LIMIT_USAGE, "features/quota", "limit-set", NULL, NO_DOC, 0}, - {"features.quota-timeout", "features/quota", "timeout", "0", DOC, 0}, - {"server.statedump-path", "protocol/server", "statedump-path", NULL, NO_DOC, 0}, - {NULL, } -}; - - +extern struct volopt_map_entry glusterd_volopt_map[]; /********************************************* * @@ -252,13 +89,14 @@ xlator_instantiate_va (const char *type, const char *format, va_list arg) xl->name = volname; INIT_LIST_HEAD (&xl->volume_options); + xl->ctx = THIS->ctx; + return xl; error: gf_log ("", GF_LOG_ERROR, "creating xlator of type %s failed", type); - if (volname) - GF_FREE (volname); + GF_FREE (volname); if (xl) xlator_destroy (xl); @@ -581,8 +419,7 @@ volopt_trie (char *key, char **hint) } out: - if (patt[0]) - GF_FREE (patt[0]); + GF_FREE (patt[0]); if (ret) *hint = NULL; @@ -614,26 +451,21 @@ struct opthandler_data { void *param; }; -#define pattern_match_options 0 - - -static void -process_option (dict_t *dict, char *key, data_t *value, void *param) +static int +process_option (char *key, data_t *value, void *param) { struct opthandler_data *odt = param; struct volopt_map_entry vme = {0,}; if (odt->rv) - return; -#if pattern_match_options - if (fnmatch (odt->vme->key, key, 0) != 0) - return; -#endif + return 0; odt->found = _gf_true; vme.key = key; vme.voltype = odt->vme->voltype; vme.option = odt->vme->option; + vme.op_version = odt->vme->op_version; + if (!vme.option) { vme.option = strrchr (key, '.'); if (vme.option) @@ -647,6 +479,7 @@ process_option (dict_t *dict, char *key, data_t *value, void *param) vme.value = value->data; odt->rv = odt->handler (odt->graph, &vme, odt->param); + return 0; } static int @@ -667,14 +500,10 @@ volgen_graph_set_options_generic (volgen_graph_t *graph, dict_t *dict, odt.found = _gf_false; odt.data_t_fake = _gf_false; -#if pattern_match_options - dict_foreach (dict, process_option, &odt); -#else data = dict_get (dict, vme->key); if (data) - process_option (dict, vme->key, data, &odt); -#endif + process_option (vme->key, data, &odt); if (odt.rv) return odt.rv; @@ -688,8 +517,7 @@ volgen_graph_set_options_generic (volgen_graph_t *graph, dict_t *dict, * in this context */ odt.data_t_fake = _gf_true; - process_option (NULL, vme->key, (data_t *)vme->value, - &odt); + process_option (vme->key, (data_t *)vme->value, &odt); if (odt.rv) return odt.rv; } @@ -749,6 +577,108 @@ optget_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme, return 0; } +static glusterd_server_xlator_t +get_server_xlator (char *xlator) +{ + glusterd_server_xlator_t subvol = GF_XLATOR_NONE; + + if (strcmp (xlator, "posix") == 0) + subvol = GF_XLATOR_POSIX; + if (strcmp (xlator, "acl") == 0) + subvol = GF_XLATOR_ACL; + if (strcmp (xlator, "locks") == 0) + subvol = GF_XLATOR_LOCKS; + if (strcmp (xlator, "io-threads") == 0) + subvol = GF_XLATOR_IOT; + if (strcmp (xlator, "index") == 0) + subvol = GF_XLATOR_INDEX; + if (strcmp (xlator, "marker") == 0) + subvol = GF_XLATOR_MARKER; + if (strcmp (xlator, "io-stats") == 0) + subvol = GF_XLATOR_IO_STATS; + if (strcmp (xlator, "bd") == 0) + subvol = GF_XLATOR_BD; + + return subvol; +} + +static glusterd_client_xlator_t +get_client_xlator (char *xlator) +{ + glusterd_client_xlator_t subvol = GF_CLNT_XLATOR_NONE; + + if (strcmp (xlator, "client") == 0) + subvol = GF_CLNT_XLATOR_FUSE; + + return subvol; +} + +static int +debugxl_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme, + void *param) +{ + char *volname = NULL; + gf_boolean_t enabled = _gf_false; + + volname = param; + + if (strcmp (vme->option, "!debug") != 0) + return 0; + + if (!strcmp (vme->key , "debug.trace") || + !strcmp (vme->key, "debug.error-gen")) { + if (get_server_xlator (vme->value) == GF_XLATOR_NONE && + get_client_xlator (vme->value) == GF_CLNT_XLATOR_NONE) + return 0; + else + goto add_graph; + } + + if (gf_string2boolean (vme->value, &enabled) == -1) + return -1; + if (!enabled) + return 0; + +add_graph: + if (volgen_graph_add (graph, vme->voltype, volname)) + return 0; + else + return -1; +} + +int +check_and_add_debug_xl (volgen_graph_t *graph, dict_t *set_dict, char *volname, + char *xlname) +{ + int ret = 0; + char *value_str = NULL; + + ret = dict_get_str (set_dict, "debug.trace", &value_str); + if (!ret) { + if (strcmp (xlname, value_str) == 0) { + ret = volgen_graph_set_options_generic (graph, set_dict, volname, + &debugxl_option_handler); + if (ret) + goto out; + } + } + + ret = dict_get_str (set_dict, "debug.error-gen", &value_str); + if (!ret) { + if (strcmp (xlname, value_str) == 0) { + ret = volgen_graph_set_options_generic (graph, set_dict, volname, + &debugxl_option_handler); + if (ret) + goto out; + } + } + + ret = 0; + +out: + return ret; +} + /* This getter considers defaults also. */ static int volgen_dict_get (dict_t *dict, char *key, char **value) @@ -917,20 +847,20 @@ glusterd_check_voloption (char *key) int glusterd_check_option_exists (char *key, char **completion) { - dict_t *dict = NULL; struct volopt_map_entry vme = {0,}; struct volopt_map_entry *vmep = NULL; int ret = 0; + xlator_t *this = THIS; (void)vme; (void)vmep; - (void)dict; if (!strchr (key, '.')) { if (completion) { ret = option_complete (key, completion); if (ret) { - gf_log ("", GF_LOG_ERROR, "Out of memory"); + gf_log (this->name, GF_LOG_ERROR, + "Out of memory"); return -1; } @@ -943,42 +873,12 @@ glusterd_check_option_exists (char *key, char **completion) return 0; } -#if !pattern_match_options for (vmep = glusterd_volopt_map; vmep->key; vmep++) { if (strcmp (vmep->key, key) == 0) { ret = 1; break; } } -#else - vme.key = key; - - /* We are getting a bit anal here to avoid typing - * fnmatch one more time. Orthogonality foremost! - * The internal logic of looking up in the volopt_map table - * should be coded exactly once. - * - * [[Ha-ha-ha, so now if I ever change the internals then I'll - * have to update the fnmatch in this comment also :P ]] - */ - dict = get_new_dict (); - if (!dict || dict_set_str (dict, key, "")) { - gf_log ("", GF_LOG_ERROR, "Out of memory"); - - return -1; - } - - ret = volgen_graph_set_options_generic (NULL, dict, &vme, - &optget_option_handler); - dict_destroy (dict); - if (ret) { - gf_log ("", GF_LOG_ERROR, "Out of memory"); - - return -1; - } - - ret = !!vme.value; -#endif if (ret || !completion) return ret; @@ -986,7 +886,7 @@ glusterd_check_option_exists (char *key, char **completion) trie: ret = volopt_trie (key, completion); if (ret) { - gf_log ("", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Some error occurred during keyword hinting"); } @@ -1115,8 +1015,12 @@ free_fp: static int volgen_write_volfile (volgen_graph_t *graph, char *filename) { - char *ftmp = NULL; - FILE *f = NULL; + char *ftmp = NULL; + FILE *f = NULL; + int fd = 0; + xlator_t *this = NULL; + + this = THIS; if (gf_asprintf (&ftmp, "%s.tmp", filename) == -1) { ftmp = NULL; @@ -1124,15 +1028,35 @@ volgen_write_volfile (volgen_graph_t *graph, char *filename) goto error; } + fd = creat (ftmp, S_IRUSR | S_IWUSR); + if (fd < 0) { + gf_log (this->name, GF_LOG_ERROR, "%s", + strerror (errno)); + goto error; + } + + close (fd); + f = fopen (ftmp, "w"); if (!f) goto error; if (glusterfs_graph_print_file (f, &graph->graph) == -1) - goto error; + goto error; - if (fclose (f) == -1) + if (fclose (f) != 0) { + gf_log (THIS->name, GF_LOG_ERROR, "fclose on the file %s " + "failed (%s)", ftmp, strerror (errno)); + /* + * Even though fclose has failed here, we have to set f to NULL. + * Otherwise when the code path goes to error, there again we + * try to close it which might cause undefined behavior such as + * process crash. + */ + f = NULL; goto error; + } + f = NULL; if (rename (ftmp, filename) == -1) @@ -1146,12 +1070,12 @@ volgen_write_volfile (volgen_graph_t *graph, char *filename) error: - if (ftmp) - GF_FREE (ftmp); + GF_FREE (ftmp); if (f) fclose (f); - gf_log ("", GF_LOG_ERROR, "failed to create volfile %s", filename); + gf_log (this->name, GF_LOG_ERROR, + "failed to create volfile %s", filename); return -1; } @@ -1482,21 +1406,28 @@ static int server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, dict_t *set_dict, void *param) { - char *volname = NULL; - char *path = NULL; - int pump = 0; - xlator_t *xl = NULL; - xlator_t *txl = NULL; - xlator_t *rbxl = NULL; - char transt[16] = {0,}; - char *ptranst = NULL; - char volume_id[64] = {0,}; - char tstamp_file[PATH_MAX] = {0,}; - int ret = 0; - char *xlator = NULL; - char *loglevel = NULL; - - path = param; + char *volname = NULL; + char *path = NULL; + int pump = 0; + xlator_t *xl = NULL; + xlator_t *txl = NULL; + xlator_t *rbxl = NULL; + char transt[16] = {0,}; + char *ptranst = NULL; + char volume_id[64] = {0,}; + char tstamp_file[PATH_MAX] = {0,}; + int ret = 0; + char *xlator = NULL; + char *loglevel = NULL; + char *username = NULL; + char *password = NULL; + char index_basepath[PATH_MAX] = {0}; + char key[1024] = {0}; + glusterd_brickinfo_t *brickinfo = NULL; + char changelog_basepath[PATH_MAX] = {0,}; + + brickinfo = param; + path = brickinfo->path; volname = volinfo->volname; get_vol_transport_type (volinfo, transt); @@ -1525,23 +1456,82 @@ server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, if (ret) return -1; + ret = check_and_add_debug_xl (graph, set_dict, volname, + "posix"); + if (ret) + return -1; +#ifdef HAVE_BD_XLATOR + if (*brickinfo->vg != '\0') { + /* Now add BD v2 xlator if volume is BD type */ + xl = volgen_graph_add (graph, "storage/bd", volname); + if (!xl) + return -1; + + ret = xlator_set_option (xl, "device", "vg"); + if (ret) + return -1; + ret = xlator_set_option (xl, "export", brickinfo->vg); + if (ret) + return -1; + + ret = check_and_add_debug_xl (graph, set_dict, volname, "bd"); + if (ret) + return -1; + + } +#endif + + xl = volgen_graph_add (graph, "features/changelog", volname); + if (!xl) + return -1; + + ret = xlator_set_option (xl, "changelog-brick", path); + if (ret) + return -1; + + snprintf (changelog_basepath, sizeof (changelog_basepath), + "%s/%s", path, ".glusterfs/changelogs"); + ret = xlator_set_option (xl, "changelog-dir", changelog_basepath); + if (ret) + return -1; + + ret = check_and_add_debug_xl (graph, set_dict, volname, "changelog"); + if (ret) + return -1; + xl = volgen_graph_add (graph, "features/access-control", volname); if (!xl) return -1; + ret = check_and_add_debug_xl (graph, set_dict, volname, "acl"); + if (ret) + return -1; + xl = volgen_graph_add (graph, "features/locks", volname); if (!xl) return -1; + ret = check_and_add_debug_xl (graph, set_dict, volname, "locks"); + if (ret) + return -1; + xl = volgen_graph_add (graph, "performance/io-threads", volname); if (!xl) return -1; + ret = check_and_add_debug_xl (graph, set_dict, volname, "io-threads"); + if (ret) + return -1; + ret = dict_get_int32 (volinfo->dict, "enable-pump", &pump); if (ret == -ENOENT) ret = pump = 0; if (ret) return -1; + + username = glusterd_auth_get_username (volinfo); + password = glusterd_auth_get_password (volinfo); + if (pump) { txl = first_of (graph); @@ -1554,6 +1544,18 @@ server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, if (NULL == ptranst) return -1; + if (username) { + ret = xlator_set_option (rbxl, "username", username); + if (ret) + return -1; + } + + if (password) { + ret = xlator_set_option (rbxl, "password", password); + if (ret) + return -1; + } + ret = xlator_set_option (rbxl, "transport-type", ptranst); GF_FREE (ptranst); if (ret) @@ -1571,9 +1573,25 @@ server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, return -1; } + xl = volgen_graph_add (graph, "features/index", volname); + if (!xl) + return -1; + + snprintf (index_basepath, sizeof (index_basepath), "%s/%s", + path, ".glusterfs/indices"); + ret = xlator_set_option (xl, "index-base", index_basepath); + if (ret) + return -1; + + ret = check_and_add_debug_xl (graph, set_dict, volname, + "index"); + if (ret) + return -1; + xl = volgen_graph_add (graph, "features/marker", volname); if (!xl) return -1; + uuid_unparse (volinfo->volume_id, volume_id); ret = xlator_set_option (xl, "volume-uuid", volume_id); if (ret) @@ -1583,6 +1601,49 @@ server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, if (ret) return -1; + ret = check_and_add_debug_xl (graph, set_dict, volname, "marker"); + if (ret) + return -1; + + if (dict_get_str_boolean (set_dict, "features.read-only", 0) && + dict_get_str_boolean (set_dict, "features.worm",0)) { + gf_log (THIS->name, GF_LOG_ERROR, + "read-only and worm cannot be set together"); + ret = -1; + goto out; + } + + /* Check for read-only volume option, and add it to the graph */ + if (dict_get_str_boolean (set_dict, "features.read-only", 0) + || volinfo -> is_snap_volume) { + xl = volgen_graph_add (graph, "features/read-only", volname); + if (!xl) { + ret = -1; + goto out; + } + } + + /* Check for worm volume option, and add it to the graph */ + if (dict_get_str_boolean (set_dict, "features.worm", 0)) { + xl = volgen_graph_add (graph, "features/worm", volname); + if (!xl) { + ret = -1; + goto out; + } + } + + /* Check for compress volume option, and add it to the graph on server side */ + if (dict_get_str_boolean (set_dict, "features.compress", 0)) { + xl = volgen_graph_add (graph, "features/cdc", volname); + if (!xl) { + ret = -1; + goto out; + } + ret = dict_set_str (set_dict, "compress.mode", "server"); + if (ret) + goto out; + } + xl = volgen_graph_add_as (graph, "debug/io-stats", path); if (!xl) return -1; @@ -1594,6 +1655,35 @@ server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, if (ret) return -1; + /*In the case of running multiple glusterds on a single machine, + * we should ensure that bricks don't listen on all IPs on that + * machine and break the IP based separation being brought about.*/ + if (dict_get (THIS->options, "transport.socket.bind-address")) { + ret = xlator_set_option (xl, "transport.socket.bind-address", + brickinfo->hostname); + if (ret) + return -1; + } + + if (username) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "auth.login.%s.allow", path); + + ret = xlator_set_option (xl, key, username); + if (ret) + return -1; + } + + if (password) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "auth.login.%s.password", + username); + + ret = xlator_set_option (xl, key, password); + if (ret) + return -1; + } + ret = volgen_graph_set_options_generic (graph, set_dict, (xlator && loglevel) ? (void *)set_dict : volinfo, (xlator && loglevel) ? &server_spec_extended_option_handler : @@ -1607,9 +1697,9 @@ server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, /* builds a graph for server role , with option overrides in mod_dict */ static int build_server_graph (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, - dict_t *mod_dict, char *path) + dict_t *mod_dict, glusterd_brickinfo_t *brickinfo) { - return build_graph_generic (graph, volinfo, mod_dict, path, + return build_graph_generic (graph, volinfo, mod_dict, brickinfo, &server_graph_builder); } @@ -1617,12 +1707,42 @@ static int perfxl_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme, void *param) { + gf_boolean_t enabled = _gf_false; + glusterd_volinfo_t *volinfo = NULL; + + GF_ASSERT (param); + volinfo = param; + + if (strcmp (vme->option, "!perf") != 0) + return 0; + + if (gf_string2boolean (vme->value, &enabled) == -1) + return -1; + if (!enabled) + return 0; + + /* Check op-version before adding the 'open-behind' xlator in the graph + */ + if (!strcmp (vme->key, "performance.open-behind") && + (vme->op_version > volinfo->client_op_version)) + return 0; + + if (volgen_graph_add (graph, vme->voltype, volinfo->volname)) + return 0; + else + return -1; +} + +static int +nfsperfxl_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme, + void *param) +{ char *volname = NULL; gf_boolean_t enabled = _gf_false; volname = param; - if (strcmp (vme->option, "!perf") != 0) + if (strcmp (vme->option, "!nfsperf") != 0) return 0; if (gf_string2boolean (vme->value, &enabled) == -1) @@ -1693,8 +1813,7 @@ init_sethelp_xml_doc (xmlTextWriterPtr *writer, xmlBufferPtr *buf) goto out; } - ret = xmlTextWriterStartElement(*writer, - (xmlChar *)"volumeOptionsDefaults"); + ret = xmlTextWriterStartElement(*writer, (xmlChar *)"options"); if (ret < 0) { gf_log ("glusterd", GF_LOG_ERROR, "Could not create an " "xmlElemetnt"); @@ -1720,7 +1839,7 @@ xml_add_volset_element (xmlTextWriterPtr writer, const char *name, GF_ASSERT (name); - ret = xmlTextWriterStartElement(writer, (xmlChar *) "volumeOption"); + ret = xmlTextWriterStartElement(writer, (xmlChar *) "option"); if (ret < 0) { gf_log ("glusterd", GF_LOG_ERROR, "Could not create an " "xmlElemetnt"); @@ -1773,7 +1892,7 @@ xml_add_volset_element (xmlTextWriterPtr writer, const char *name, #endif static int -get_key_from_volopt ( struct volopt_map_entry *vme, char **key) +_get_xlator_opt_key_from_vme ( struct volopt_map_entry *vme, char **key) { int ret = 0; @@ -1781,25 +1900,32 @@ get_key_from_volopt ( struct volopt_map_entry *vme, char **key) GF_ASSERT (key); - if (vme->option) { - if (vme->option[0] == '!') { - *key = vme->option + 1; - if (!*key[0]) - ret = -1; + if (!strcmp (vme->key, AUTH_ALLOW_MAP_KEY)) + *key = gf_strdup (AUTH_ALLOW_OPT_KEY); + else if (!strcmp (vme->key, AUTH_REJECT_MAP_KEY)) + *key = gf_strdup (AUTH_REJECT_OPT_KEY); + else if (!strcmp (vme->key, NFS_DISABLE_MAP_KEY)) + *key = gf_strdup (NFS_DISABLE_OPT_KEY); + else { + if (vme->option) { + if (vme->option[0] == '!') { + *key = vme->option + 1; + if (!*key[0]) + ret = -1; + } else { + *key = vme->option; + } } else { - *key = vme->option; - } - } else { - *key = strchr (vme->key, '.'); - if (*key) { - (*key) ++; - if (!*key[0]) + *key = strchr (vme->key, '.'); + if (*key) { + (*key) ++; + if (!*key[0]) + ret = -1; + } else { ret = -1; - } else { - ret = -1; + } } } - if (ret) gf_log ("glusterd", GF_LOG_ERROR, "Wrong entry found in " "glusterd_volopt_map entry %s", vme->key); @@ -1809,23 +1935,32 @@ get_key_from_volopt ( struct volopt_map_entry *vme, char **key) return ret; } +static void +_free_xlator_opt_key (char *key) +{ + GF_ASSERT (key); + + if (!strcmp (key, AUTH_ALLOW_OPT_KEY) || + !strcmp (key, AUTH_REJECT_OPT_KEY) || + !strcmp (key, NFS_DISABLE_OPT_KEY)) + GF_FREE (key); + + return; +} int -glusterd_get_volopt_content (gf_boolean_t xml_out) +glusterd_get_volopt_content (dict_t * ctx, gf_boolean_t xml_out) { - - char *xlator_type = NULL; void *dl_handle = NULL; - volume_opt_list_t vol_opt_handle = {{0},}; + volume_opt_list_t vol_opt_handle = {{0},}; char *key = NULL; struct volopt_map_entry *vme = NULL; int ret = -1; char *def_val = NULL; char *descr = NULL; - char output_string[8192] = {0, }; + char output_string[25600] = {0, }; char *output = NULL; - char tmp_str[1024] = {0, }; - dict_t *ctx = NULL; + char tmp_str[2048] = {0, }; #if (HAVE_LIB_XML) xmlTextWriterPtr writer = NULL; xmlBufferPtr buf = NULL; @@ -1837,51 +1972,73 @@ glusterd_get_volopt_content (gf_boolean_t xml_out) } #endif - ctx = glusterd_op_get_ctx (); - - if (!ctx) { - /*extract the vol-set-help output only in host glusterd*/ - ret = 0; - goto out; - } - INIT_LIST_HEAD (&vol_opt_handle.list); for (vme = &glusterd_volopt_map[0]; vme->key; vme++) { - if ( ( vme->type == NO_DOC) || (vme->type == GLOBAL_NO_DOC) ) + if ((vme->type == NO_DOC) || (vme->type == GLOBAL_NO_DOC)) continue; - if (get_key_from_volopt (vme, &key)) - goto out; /*Some error while getin key*/ + if (vme->description) { + descr = vme->description; + def_val = vme->value; + } else { + if (_get_xlator_opt_key_from_vme (vme, &key)) { + gf_log ("glusterd", GF_LOG_DEBUG, "Failed to " + "get %s key from volume option entry", + vme->key); + goto out; /*Some error while geting key*/ + } - if (!xlator_type || strcmp (vme->voltype, xlator_type)){ ret = xlator_volopt_dynload (vme->voltype, - &dl_handle, - &vol_opt_handle); - if (ret) - continue; - } + &dl_handle, + &vol_opt_handle); - ret = xlator_option_info_list (&vol_opt_handle, key, - &def_val, &descr); - if (ret) /*Swallow Error i.e if option not found*/ - continue; + if (ret) { + gf_log ("glusterd", GF_LOG_DEBUG, + "xlator_volopt_dynload error(%d)", ret); + ret = 0; + goto cont; + } + + ret = xlator_option_info_list (&vol_opt_handle, key, + &def_val, &descr); + if (ret) { /*Swallow Error i.e if option not found*/ + gf_log ("glusterd", GF_LOG_DEBUG, + "Failed to get option for %s key", key); + ret = 0; + goto cont; + } + } if (xml_out) { #if (HAVE_LIB_XML) if (xml_add_volset_element (writer,vme->key, - def_val, descr)) - goto out; + def_val, descr)) { + ret = -1; + goto cont; + } #else gf_log ("glusterd", GF_LOG_ERROR, "Libxml not present"); #endif } else { - snprintf (tmp_str, 1024, "Option: %s\nDefault " + snprintf (tmp_str, sizeof (tmp_str), "Option: %s\nDefault " "Value: %s\nDescription: %s\n\n", vme->key, def_val, descr); strcat (output_string, tmp_str); } +cont: + if (dl_handle) { + dlclose (dl_handle); + dl_handle = NULL; + vol_opt_handle.given_opt = NULL; + } + if (key) { + _free_xlator_opt_key (key); + key = NULL; + } + if (ret) + goto out; } #if (HAVE_LIB_XML) @@ -1908,7 +2065,7 @@ glusterd_get_volopt_content (gf_boolean_t xml_out) } ret = dict_set_dynstr (ctx, "help-str", output); - out: +out: gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); return ret; @@ -1920,10 +2077,14 @@ volgen_graph_build_clients (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, { int i = 0; int ret = -1; + uint32_t client_type = GF_CLIENT_OTHER; char transt[16] = {0,}; char *volname = NULL; - glusterd_brickinfo_t *brick = NULL; - xlator_t *xl = NULL; + char *str = NULL; + glusterd_brickinfo_t *brick = NULL; + xlator_t *xl = NULL; + char *ssl_str = NULL; + gf_boolean_t ssl_bool; volname = volinfo->volname; @@ -1966,8 +2127,45 @@ volgen_graph_build_clients (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, ret = xlator_set_option (xl, "transport-type", transt); if (ret) goto out; + + ret = dict_get_uint32 (set_dict, "trusted-client", + &client_type); + + if (!ret && client_type == GF_CLIENT_TRUSTED) { + str = NULL; + str = glusterd_auth_get_username (volinfo); + if (str) { + ret = xlator_set_option (xl, "username", + str); + if (ret) + goto out; + } + + str = glusterd_auth_get_password (volinfo); + if (str) { + ret = xlator_set_option (xl, "password", + str); + if (ret) + goto out; + } + } + + if (dict_get_str(set_dict,"client.ssl",&ssl_str) == 0) { + if (gf_string2boolean(ssl_str,&ssl_bool) == 0) { + if (ssl_bool) { + ret = xlator_set_option(xl, + "transport.socket.ssl-enabled", + "true"); + if (ret) { + goto out; + } + } + } + } + i++; } + if (i != volinfo->brick_count) { gf_log ("", GF_LOG_ERROR, "volume inconsistency: actual number of bricks (%d) " @@ -2018,6 +2216,7 @@ volgen_graph_build_clusters (volgen_graph_t *graph, if (trav == txl) break; + i++; } @@ -2128,13 +2327,33 @@ volgen_graph_build_dht_cluster (volgen_graph_t *graph, int ret = -1; char *decommissioned_children = NULL; xlator_t *dht = NULL; + char *voltype = "cluster/distribute"; + + /* NUFA and Switch section */ + if (dict_get_str_boolean (volinfo->dict, "cluster.nufa", 0) && + dict_get_str_boolean (volinfo->dict, "cluster.switch", 0)) { + gf_log (THIS->name, GF_LOG_ERROR, + "nufa and switch cannot be set together"); + ret = -1; + goto out; + } + + /* Check for NUFA volume option, and change the voltype */ + if (dict_get_str_boolean (volinfo->dict, "cluster.nufa", 0)) + voltype = "cluster/nufa"; + + /* Check for switch volume option, and change the voltype */ + if (dict_get_str_boolean (volinfo->dict, "cluster.switch", 0)) + voltype = "cluster/switch"; - GF_ASSERT (child_count > 1); clusters = volgen_graph_build_clusters (graph, volinfo, - "cluster/distribute", "%s-dht", - child_count, child_count); + voltype, + "%s-dht", + child_count, + child_count); if (clusters < 0) goto out; + dht = first_of (graph); ret = _graph_get_decommissioned_children (dht, volinfo, &decommissioned_children); @@ -2148,8 +2367,7 @@ volgen_graph_build_dht_cluster (volgen_graph_t *graph, } ret = 0; out: - if (decommissioned_children) - GF_FREE (decommissioned_children); + GF_FREE (decommissioned_children); return ret; } @@ -2222,24 +2440,51 @@ volume_volgen_graph_build_clusters (volgen_graph_t *graph, build_distribute: dist_count = volinfo->brick_count / volinfo->dist_leaf_count; - if (dist_count > 1) { - ret = volgen_graph_build_dht_cluster (graph, volinfo, - dist_count); - if (ret) - goto out; + if (!dist_count) { + ret = -1; + goto out; } + + ret = volgen_graph_build_dht_cluster (graph, volinfo, + dist_count); + if (ret == -1) + goto out; + ret = 0; out: return ret; } +static int client_graph_set_perf_options(volgen_graph_t *graph, + glusterd_volinfo_t *volinfo, + dict_t *set_dict) +{ + data_t *tmp_data = NULL; + char *volname = NULL; + + /* + * Logic to make sure NFS doesn't have performance translators by + * default for a volume + */ + volname = volinfo->volname; + tmp_data = dict_get (set_dict, "nfs-volume-file"); + if (!tmp_data) + return volgen_graph_set_options_generic(graph, set_dict, + volname, + &perfxl_option_handler); + else + return volgen_graph_set_options_generic(graph, set_dict, + volname, + &nfsperfxl_option_handler); +} + static int client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, dict_t *set_dict, void *param) { - int ret = 0; - xlator_t *xl = NULL; - char *volname = NULL; + int ret = 0; + xlator_t *xl = NULL; + char *volname = NULL; volname = volinfo->volname; ret = volgen_graph_build_clients (graph, volinfo, set_dict, param); @@ -2247,9 +2492,34 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, goto out; ret = volume_volgen_graph_build_clusters (graph, volinfo); - if (ret) + if (ret == -1) goto out; + /* Check for compress volume option, and add it to the graph on client side */ + if (dict_get_str_boolean (set_dict, "features.compress", 0)) { + xl = volgen_graph_add (graph, "features/cdc", volname); + if (!xl) { + ret = -1; + goto out; + } + ret = dict_set_str (set_dict, "compress.mode", "client"); + if (ret) + goto out; + + } + + ret = glusterd_volinfo_get_boolean (volinfo, "features.encryption"); + if (ret == -1) + goto out; + if (ret) { + xl = volgen_graph_add (graph, "encryption/crypt", volname); + + if (!xl) { + ret = -1; + goto out; + } + } + ret = glusterd_volinfo_get_boolean (volinfo, VKEY_FEATURES_QUOTA); if (ret == -1) goto out; @@ -2262,11 +2532,29 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, } } - ret = volgen_graph_set_options_generic (graph, set_dict, volname, - &perfxl_option_handler); + + ret = glusterd_volinfo_get_boolean (volinfo, "features.file-snapshot"); + if (ret == -1) + goto out; + if (ret) { + xl = volgen_graph_add (graph, "features/qemu-block", volname); + + if (!xl) { + ret = -1; + goto out; + } + } + + ret = client_graph_set_perf_options(graph, volinfo, set_dict); if (ret) goto out; + /* add debug translators depending on the options */ + ret = check_and_add_debug_xl (graph, set_dict, volname, + "client"); + if (ret) + return -1; + ret = -1; xl = volgen_graph_add_as (graph, "debug/io-stats", volname); if (!xl) @@ -2275,9 +2563,15 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, ret = volgen_graph_set_options_generic (graph, set_dict, "client", &loglevel_option_handler); - if (!ret) - ret = volgen_graph_set_options_generic (graph, set_dict, "client", - &sys_loglevel_option_handler); + if (ret) + gf_log (THIS->name, GF_LOG_WARNING, "changing client log level" + " failed"); + + ret = volgen_graph_set_options_generic (graph, set_dict, "client", + &sys_loglevel_option_handler); + if (ret) + gf_log (THIS->name, GF_LOG_WARNING, "changing client syslog " + "level failed"); out: return ret; } @@ -2292,20 +2586,35 @@ build_client_graph (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, &client_graph_builder); } +char *gd_shd_options[] = { + "!self-heal-daemon", + "!heal-timeout", + NULL +}; + +char* +gd_get_matching_option (char **options, char *option) +{ + while (*options && strcmp (*options, option)) + options++; + return *options; +} + static int shd_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme, - void *param) + void *param) { int ret = 0; struct volopt_map_entry new_vme = {0}; - int shd = 0; + char *shd_option = NULL; - shd = !strcmp (vme->option, "!self-heal-daemon"); - if ((vme->option[0] == '!') && !shd) + shd_option = gd_get_matching_option (gd_shd_options, vme->option); + if ((vme->option[0] == '!') && !shd_option) goto out; new_vme = *vme; - if (shd) - new_vme.option = "self-heal-daemon"; + if (shd_option) { + new_vme.option = shd_option + 1;//option with out '!' + } ret = no_filter_option_handler (graph, &new_vme, param); out: @@ -2329,7 +2638,7 @@ nfs_option_handler (volgen_graph_t *graph, ret = xlator_set_option (xl, vme->key, vme->value); }*/ - if ( !volinfo || !volinfo->volname) + if (!volinfo || (volinfo->volname[0] == '\0')) return 0; if (! strcmp (vme->option, "!rpc-auth.addr.*.allow")) { @@ -2359,7 +2668,7 @@ nfs_option_handler (volgen_graph_t *graph, } if (! strcmp (vme->option, "!rpc-auth.auth-unix.*")) { - ret = gf_asprintf (&aa, "rpc-auth.auth.unix.%s", + ret = gf_asprintf (&aa, "rpc-auth.auth-unix.%s", volinfo->volname); if (ret != -1) { @@ -2370,8 +2679,8 @@ nfs_option_handler (volgen_graph_t *graph, if (ret) return -1; } - if (! strcmp (vme->option, "!rpc-auth.auth.null.*")) { - ret = gf_asprintf (&aa, "rpc-auth.auth.null.%s", + if (! strcmp (vme->option, "!rpc-auth.auth-null.*")) { + ret = gf_asprintf (&aa, "rpc-auth.auth-null.%s", volinfo->volname); if (ret != -1) { @@ -2427,6 +2736,10 @@ nfs_option_handler (volgen_graph_t *graph, volinfo->volname); if (ret != -1) { + ret = gf_canonicalize_path (vme->value); + if (ret) + return -1; + ret = xlator_set_option (xl, aa, vme->value); GF_FREE (aa); } @@ -2489,6 +2802,23 @@ nfs_option_handler (volgen_graph_t *graph, } static int +volgen_graph_set_iam_shd (volgen_graph_t *graph) +{ + xlator_t *trav; + int ret = 0; + + for (trav = first_of (graph); trav; trav = trav->next) { + if (strcmp (trav->type, "cluster/replicate") != 0) + continue; + + ret = xlator_set_option (trav, "iam-self-heal-daemon", "yes"); + if (ret) + break; + } + return ret; +} + +static int build_shd_graph (volgen_graph_t *graph, dict_t *mod_dict) { volgen_graph_t cgraph = {0}; @@ -2499,8 +2829,9 @@ build_shd_graph (volgen_graph_t *graph, dict_t *mod_dict) int ret = 0; gf_boolean_t valid_config = _gf_false; xlator_t *iostxl = NULL; - int rclusters = 0; + int rclusters = 0; int replica_count = 0; + gf_boolean_t graph_check = _gf_false; this = THIS; priv = this->private; @@ -2511,6 +2842,7 @@ build_shd_graph (volgen_graph_t *graph, dict_t *mod_dict) goto out; } + graph_check = dict_get_str_boolean (mod_dict, "graph-check", 0); iostxl = volgen_graph_add_as (graph, "debug/io-stats", "glustershd"); if (!iostxl) { ret = -1; @@ -2518,7 +2850,8 @@ build_shd_graph (volgen_graph_t *graph, dict_t *mod_dict) } list_for_each_entry (voliter, &priv->volumes, vol_list) { - if (voliter->status != GLUSTERD_STATUS_STARTED) + if (!graph_check && + (voliter->status != GLUSTERD_STATUS_STARTED)) continue; if (!glusterd_is_volume_replicate (voliter)) @@ -2532,6 +2865,11 @@ build_shd_graph (volgen_graph_t *graph, dict_t *mod_dict) if (ret) goto out; + ret = dict_set_uint32 (set_dict, "trusted-client", + GF_CLIENT_TRUSTED); + if (ret) + goto out; + dict_copy (voliter->dict, set_dict); if (mod_dict) dict_copy (mod_dict, set_dict); @@ -2557,10 +2895,29 @@ build_shd_graph (volgen_graph_t *graph, dict_t *mod_dict) if (ret) goto out; + ret = volgen_graph_set_iam_shd (&cgraph); + if (ret) + goto out; + ret = volgen_graph_merge_sub (graph, &cgraph, rclusters); if (ret) goto out; + ret = volgen_graph_set_options_generic (graph, set_dict, + "client", + &loglevel_option_handler); + + if (ret) + gf_log (THIS->name, GF_LOG_WARNING, "changing loglevel " + "of self-heal daemon failed"); + + ret = volgen_graph_set_options_generic (graph, set_dict, + "client", + &sys_loglevel_option_handler); + if (ret) + gf_log (THIS->name, GF_LOG_WARNING, "changing syslog " + "level of self-heal daemon failed"); + ret = dict_reset (set_dict); if (ret) goto out; @@ -2586,6 +2943,8 @@ build_nfs_graph (volgen_graph_t *graph, dict_t *mod_dict) char *skey = NULL; int ret = 0; char nfs_xprt[16] = {0,}; + char *volname = NULL; + data_t *data = NULL; this = THIS; GF_ASSERT (this); @@ -2607,6 +2966,14 @@ build_nfs_graph (volgen_graph_t *graph, dict_t *mod_dict) if (ret) goto out; + ret = xlator_set_option (nfsxl, "nfs.nlm", "on"); + if (ret) + goto out; + + ret = xlator_set_option (nfsxl, "nfs.drc", "on"); + if (ret) + goto out; + list_for_each_entry (voliter, &priv->volumes, vol_list) { if (voliter->status != GLUSTERD_STATUS_STARTED) continue; @@ -2644,16 +3011,35 @@ build_nfs_graph (volgen_graph_t *graph, dict_t *mod_dict) else get_transport_type (voliter, voliter->dict, nfs_xprt, _gf_true); - ret = dict_set_str (set_dict, VKEY_PERF_STAT_PREFETCH, "off"); + ret = dict_set_str (set_dict, "performance.stat-prefetch", "off"); if (ret) goto out; - ret = dict_set_str (set_dict, "performance.client-io-threads", "off"); + ret = dict_set_str (set_dict, "performance.client-io-threads", + "off"); if (ret) goto out; ret = dict_set_str (set_dict, "client-transport-type", nfs_xprt); + if (ret) + goto out; + + ret = dict_set_uint32 (set_dict, "trusted-client", + GF_CLIENT_TRUSTED); + if (ret) + goto out; + + ret = dict_set_str (set_dict, "nfs-volume-file", "yes"); + if (ret) + goto out; + + if (mod_dict && (data = dict_get (mod_dict, "volume-name"))) { + volname = data->data; + if (strcmp (volname, voliter->volname) == 0) + dict_copy (mod_dict, set_dict); + } + ret = build_client_graph (&cgraph, voliter, set_dict); if (ret) goto out; @@ -2736,16 +3122,21 @@ glusterd_is_valid_volfpath (char *volname, char *brick) glusterd_brickinfo_t *brickinfo = NULL; glusterd_volinfo_t *volinfo = NULL; int32_t ret = 0; + xlator_t *this = NULL; - ret = glusterd_brickinfo_from_brick (brick, &brickinfo); + this = THIS; + GF_ASSERT (this); + + ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo); if (ret) { - gf_log ("", GF_LOG_WARNING, "brick path validation failed"); + gf_log (this->name, GF_LOG_WARNING, "Failed to create brickinfo" + " for brick %s", brick ); ret = 0; goto out; } ret = glusterd_volinfo_new (&volinfo); if (ret) { - gf_log ("", GF_LOG_WARNING, "brick path validation failed"); + gf_log (this->name, GF_LOG_WARNING, "Failed to create volinfo"); ret = 0; goto out; } @@ -2775,7 +3166,7 @@ glusterd_generate_brick_volfile (glusterd_volinfo_t *volinfo, get_brick_filepath (filename, volinfo, brickinfo); - ret = build_server_graph (&graph, volinfo, NULL, brickinfo->path); + ret = build_server_graph (&graph, volinfo, NULL, brickinfo); if (!ret) ret = volgen_write_volfile (&graph, filename); @@ -2784,8 +3175,6 @@ glusterd_generate_brick_volfile (glusterd_volinfo_t *volinfo, return ret; } - - static void get_vol_tstamp_file (char *filename, glusterd_volinfo_t *volinfo) { @@ -2812,7 +3201,7 @@ generate_brick_volfiles (glusterd_volinfo_t *volinfo) get_vol_tstamp_file (tstamp_file, volinfo); if (ret) { - ret = open (tstamp_file, O_WRONLY|O_CREAT|O_EXCL, 0644); + ret = open (tstamp_file, O_WRONLY|O_CREAT|O_EXCL, 0600); if (ret == -1 && errno == EEXIST) { gf_log ("", GF_LOG_DEBUG, "timestamp file exist"); ret = -2; @@ -2869,25 +3258,6 @@ generate_single_transport_client_volfile (glusterd_volinfo_t *volinfo, return ret; } -void -get_client_filepath (char *filepath, glusterd_volinfo_t *volinfo, gf_transport_type type) -{ - char path[PATH_MAX] = {0,}; - glusterd_conf_t *priv = NULL; - - priv = THIS->private; - - GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv); - - if ((volinfo->transport_type == GF_TRANSPORT_BOTH_TCP_RDMA) && - (type == GF_TRANSPORT_RDMA)) - snprintf (filepath, PATH_MAX, "%s/%s.rdma-fuse.vol", - path, volinfo->volname); - else - snprintf (filepath, PATH_MAX, "%s/%s-fuse.vol", - path, volinfo->volname); -} - static void enumerate_transport_reqs (gf_transport_type type, char **types) { @@ -2905,8 +3275,9 @@ enumerate_transport_reqs (gf_transport_type type, char **types) } } -static int -generate_client_volfiles (glusterd_volinfo_t *volinfo) +int +generate_client_volfiles (glusterd_volinfo_t *volinfo, + glusterd_client_type_t client_type) { char filepath[PATH_MAX] = {0,}; int ret = -1; @@ -2925,7 +3296,21 @@ generate_client_volfiles (glusterd_volinfo_t *volinfo) if (ret) goto out; type = transport_str_to_type (types[i]); - get_client_filepath (filepath, volinfo, type); + + ret = dict_set_uint32 (dict, "trusted-client", client_type); + if (ret) + goto out; + + if (client_type == GF_CLIENT_TRUSTED) { + glusterd_get_trusted_client_filepath (filepath, + volinfo, + type); + } else { + glusterd_get_client_filepath (filepath, + volinfo, + type); + } + ret = generate_single_transport_client_volfile (volinfo, filepath, dict); @@ -2946,7 +3331,7 @@ glusterd_create_rb_volfiles (glusterd_volinfo_t *volinfo, ret = glusterd_generate_brick_volfile (volinfo, brickinfo); if (!ret) - ret = generate_client_volfiles (volinfo); + ret = generate_client_volfiles (volinfo, GF_CLIENT_TRUSTED); if (!ret) ret = glusterd_fetchspec_notify (THIS); @@ -2956,23 +3341,33 @@ glusterd_create_rb_volfiles (glusterd_volinfo_t *volinfo, int glusterd_create_volfiles_and_notify_services (glusterd_volinfo_t *volinfo) { - int ret = -1; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; ret = generate_brick_volfiles (volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "Could not generate volfiles for bricks"); goto out; } - ret = generate_client_volfiles (volinfo); + ret = generate_client_volfiles (volinfo, GF_CLIENT_TRUSTED); if (ret) { - gf_log ("", GF_LOG_ERROR, - "Could not generate volfile for client"); + gf_log (this->name, GF_LOG_ERROR, + "Could not generate trusted client volfiles"); goto out; } - ret = glusterd_fetchspec_notify (THIS); + ret = generate_client_volfiles (volinfo, GF_CLIENT_OTHER); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not generate client volfiles"); + goto out; + } + + ret = glusterd_fetchspec_notify (this); out: return ret; @@ -3027,6 +3422,14 @@ glusterd_create_shd_volfile () if (ret) goto out; + ret = dict_set_str (mod_dict, "cluster.metadata-self-heal", "on"); + if (ret) + goto out; + + ret = dict_set_str (mod_dict, "cluster.entry-self-heal", "on"); + if (ret) + goto out; + glusterd_get_nodesvc_volfile ("glustershd", conf->workdir, filepath, sizeof (filepath)); ret = glusterd_create_global_volfile (build_shd_graph, filepath, @@ -3038,6 +3441,106 @@ out: } int +glusterd_check_nfs_topology_identical (gf_boolean_t *identical) +{ + char nfsvol[PATH_MAX] = {0,}; + char tmpnfsvol[PATH_MAX] = {0,}; + glusterd_conf_t *conf = NULL; + xlator_t *this = THIS; + int ret = -1; + int tmpclean = 0; + int tmpfd = -1; + + if ((!identical) || (!this) || (!this->private)) + goto out; + + conf = (glusterd_conf_t *) this->private; + + /* Fetch the original NFS volfile */ + glusterd_get_nodesvc_volfile ("nfs", conf->workdir, + nfsvol, sizeof (nfsvol)); + + /* Create the temporary NFS volfile */ + snprintf (tmpnfsvol, sizeof (tmpnfsvol), "/tmp/gnfs-XXXXXX"); + tmpfd = mkstemp (tmpnfsvol); + if (tmpfd < 0) { + gf_log (this->name, GF_LOG_WARNING, + "Unable to create temp file %s: (%s)", + tmpnfsvol, strerror (errno)); + goto out; + } + + tmpclean = 1; /* SET the flag to unlink() tmpfile */ + + ret = glusterd_create_global_volfile (build_nfs_graph, + tmpnfsvol, NULL); + if (ret) + goto out; + + /* Compare the topology of volfiles */ + ret = glusterd_check_topology_identical (nfsvol, tmpnfsvol, + identical); +out: + if (tmpfd >= 0) + close (tmpfd); + if (tmpclean) + unlink (tmpnfsvol); + return ret; +} + +int +glusterd_check_nfs_volfile_identical (gf_boolean_t *identical) +{ + char nfsvol[PATH_MAX] = {0,}; + char tmpnfsvol[PATH_MAX] = {0,}; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + int ret = -1; + int need_unlink = 0; + int tmp_fd = -1; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (identical); + + conf = this->private; + + glusterd_get_nodesvc_volfile ("nfs", conf->workdir, + nfsvol, sizeof (nfsvol)); + + snprintf (tmpnfsvol, sizeof (tmpnfsvol), "/tmp/gnfs-XXXXXX"); + + tmp_fd = mkstemp (tmpnfsvol); + if (tmp_fd < 0) { + gf_log ("", GF_LOG_WARNING, "Unable to create temp file %s: " + "(%s)", tmpnfsvol, strerror (errno)); + goto out; + } + + need_unlink = 1; + + ret = glusterd_create_global_volfile (build_nfs_graph, + tmpnfsvol, NULL); + if (ret) + goto out; + + ret = glusterd_check_files_identical (nfsvol, tmpnfsvol, + identical); + if (ret) + goto out; + +out: + if (need_unlink) + unlink (tmpnfsvol); + + if (tmp_fd >= 0) + close (tmp_fd); + + return ret; +} + +int glusterd_delete_volfile (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo) { @@ -3056,6 +3559,35 @@ glusterd_delete_volfile (glusterd_volinfo_t *volinfo, } int +validate_shdopts (glusterd_volinfo_t *volinfo, + dict_t *val_dict, + char **op_errstr) +{ + volgen_graph_t graph = {0,}; + int ret = -1; + + graph.errstr = op_errstr; + + if (!glusterd_is_volume_replicate (volinfo)) { + ret = 0; + goto out; + } + ret = dict_set_str (val_dict, "graph-check", "on"); + if (ret) + goto out; + ret = build_shd_graph (&graph, val_dict); + if (!ret) + ret = graph_reconf_validateopt (&graph.graph, op_errstr); + + volgen_graph_free (&graph); + + gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); +out: + dict_del (val_dict, "graph-check"); + return ret; +} + +int validate_nfsopts (glusterd_volinfo_t *volinfo, dict_t *val_dict, char **op_errstr) @@ -3065,6 +3597,9 @@ validate_nfsopts (glusterd_volinfo_t *volinfo, char transport_type[16] = {0,}; char *tt = NULL; char err_str[4096] = {0,}; + xlator_t *this = THIS; + + GF_ASSERT (this); graph.errstr = op_errstr; @@ -3075,7 +3610,7 @@ validate_nfsopts (glusterd_volinfo_t *volinfo, snprintf (err_str, sizeof (err_str), "Changing nfs " "transport type is allowed only for volumes " "of transport type tcp,rdma"); - gf_log ("", GF_LOG_ERROR, "%s", err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); *op_errstr = gf_strdup (err_str); ret = -1; goto out; @@ -3089,6 +3624,12 @@ validate_nfsopts (glusterd_volinfo_t *volinfo, } } + ret = dict_set_str (val_dict, "volume-name", volinfo->volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set volume name"); + goto out; + } + ret = build_nfs_graph (&graph, val_dict); if (!ret) ret = graph_reconf_validateopt (&graph.graph, op_errstr); @@ -3096,17 +3637,20 @@ validate_nfsopts (glusterd_volinfo_t *volinfo, volgen_graph_free (&graph); out: - gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret); + if (dict_get (val_dict, "volume-name")) + dict_del (val_dict, "volume-name"); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } + int validate_clientopts (glusterd_volinfo_t *volinfo, - dict_t *val_dict, - char **op_errstr) + dict_t *val_dict, + char **op_errstr) { volgen_graph_t graph = {0,}; - int ret = -1; + int ret = -1; GF_ASSERT (volinfo); @@ -3124,18 +3668,18 @@ validate_clientopts (glusterd_volinfo_t *volinfo, int validate_brickopts (glusterd_volinfo_t *volinfo, - char *brickinfo_path, + glusterd_brickinfo_t *brickinfo, dict_t *val_dict, char **op_errstr) { volgen_graph_t graph = {0,}; - int ret = -1; + int ret = -1; GF_ASSERT (volinfo); graph.errstr = op_errstr; - ret = build_server_graph (&graph, volinfo, val_dict, brickinfo_path); + ret = build_server_graph (&graph, volinfo, val_dict, brickinfo); if (!ret) ret = graph_reconf_validateopt (&graph.graph, op_errstr); @@ -3157,28 +3701,30 @@ glusterd_validate_brickreconf (glusterd_volinfo_t *volinfo, gf_log ("", GF_LOG_DEBUG, "Validating %s", brickinfo->hostname); - ret = validate_brickopts (volinfo, brickinfo->path, val_dict, + ret = validate_brickopts (volinfo, brickinfo, val_dict, op_errstr); if (ret) goto out; } ret = 0; -out: +out: return ret; } -static void +static int _check_globalopt (dict_t *this, char *key, data_t *value, void *ret_val) { int *ret = NULL; ret = ret_val; if (*ret) - return; + return 0; if (!glusterd_check_globaloption (key)) *ret = 1; + + return 0; } int @@ -3208,23 +3754,34 @@ glusterd_validate_globalopts (glusterd_volinfo_t *volinfo, } ret = validate_nfsopts (volinfo, val_dict, op_errstr); + if (ret) { + gf_log ("", GF_LOG_DEBUG, "Could not Validate nfs"); + goto out; + } + ret = validate_shdopts (volinfo, val_dict, op_errstr); + if (ret) { + gf_log ("", GF_LOG_DEBUG, "Could not Validate self-heald"); + goto out; + } out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - return ret; + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; } -static void +static int _check_localopt (dict_t *this, char *key, data_t *value, void *ret_val) { int *ret = NULL; ret = ret_val; if (*ret) - return; + return 0; if (!glusterd_check_localoption (key)) *ret = 1; + + return 0; } int @@ -3254,9 +3811,255 @@ glusterd_validate_reconfopts (glusterd_volinfo_t *volinfo, dict_t *val_dict, } ret = validate_nfsopts (volinfo, val_dict, op_errstr); + if (ret) { + gf_log ("", GF_LOG_DEBUG, "Could not Validate nfs"); + goto out; + } + + + ret = validate_shdopts (volinfo, val_dict, op_errstr); + if (ret) { + gf_log ("", GF_LOG_DEBUG, "Could not Validate self-heald"); + goto out; + } out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - return ret; + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +static struct volopt_map_entry * +_gd_get_vmep (char *key) { + char *completion = NULL; + struct volopt_map_entry *vmep = NULL; + int ret = 0; + + COMPLETE_OPTION ((char *)key, completion, ret); + for (vmep = glusterd_volopt_map; vmep->key; vmep++) { + if (strcmp (vmep->key, key) == 0) + return vmep; + } + + return NULL; +} + +uint32_t +glusterd_get_op_version_for_key (char *key) +{ + struct volopt_map_entry *vmep = NULL; + + GF_ASSERT (key); + + vmep = _gd_get_vmep (key); + if (vmep) + return vmep->op_version; + + return 0; +} + +gf_boolean_t +gd_is_client_option (char *key) +{ + struct volopt_map_entry *vmep = NULL; + + GF_ASSERT (key); + + vmep = _gd_get_vmep (key); + if (vmep && (vmep->flags & OPT_FLAG_CLIENT_OPT)) + return _gf_true; + + return _gf_false; +} + +gf_boolean_t +gd_is_xlator_option (char *key) +{ + struct volopt_map_entry *vmep = NULL; + + GF_ASSERT (key); + + vmep = _gd_get_vmep (key); + if (vmep && (vmep->flags & OPT_FLAG_XLATOR_OPT)) + return _gf_true; + + return _gf_false; +} + +volume_option_type_t +_gd_get_option_type (char *key) +{ + struct volopt_map_entry *vmep = NULL; + void *dl_handle = NULL; + volume_opt_list_t vol_opt_list = {{0},}; + int ret = -1; + volume_option_t *opt = NULL; + char *xlopt_key = NULL; + volume_option_type_t opt_type = GF_OPTION_TYPE_MAX; + + GF_ASSERT (key); + + vmep = _gd_get_vmep (key); + + if (vmep) { + INIT_LIST_HEAD (&vol_opt_list.list); + ret = xlator_volopt_dynload (vmep->voltype, &dl_handle, + &vol_opt_list); + if (ret) + goto out; + + if (_get_xlator_opt_key_from_vme (vmep, &xlopt_key)) + goto out; + + opt = xlator_volume_option_get_list (&vol_opt_list, xlopt_key); + _free_xlator_opt_key (xlopt_key); + + if (opt) + opt_type = opt->type; + } + +out: + if (dl_handle) { + dlclose (dl_handle); + dl_handle = NULL; + } + + return opt_type; +} + +gf_boolean_t +gd_is_boolean_option (char *key) +{ + GF_ASSERT (key); + + if (GF_OPTION_TYPE_BOOL == _gd_get_option_type (key)) + return _gf_true; + + return _gf_false; +} + +/* This function will restore origin volume to it's snap. + * The restore operation will simply replace the Gluster origin + * volume with the snap volume. + * TODO: Multi-volume delete to be done. + * Cleanup in case of restore failure is pending. + * + * @param orig_vol volinfo of origin volume + * @param snap_vol volinfo of snapshot volume + * + * @return 0 on success and negative value on error + */ +int +gd_restore_snap_volume (dict_t *rsp_dict, + glusterd_volinfo_t *orig_vol, + glusterd_volinfo_t *snap_vol) +{ + int ret = -1; + glusterd_volinfo_t *new_volinfo = NULL; + glusterd_snap_t *snap = NULL; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *temp_volinfo = NULL; + glusterd_volinfo_t *voliter = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + conf = this->private; + GF_ASSERT (conf); + + GF_VALIDATE_OR_GOTO (this->name, orig_vol, out); + GF_VALIDATE_OR_GOTO (this->name, snap_vol, out); + snap = snap_vol->snapshot; + GF_VALIDATE_OR_GOTO (this->name, snap, out); + + /* Snap volume must be stoped before performing the + * restore operation. + */ + ret = glusterd_stop_volume (snap_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to stop " + "snap volume"); + goto out; + } + + /* Create a new volinfo for the restored volume */ + ret = glusterd_volinfo_dup (snap_vol, &new_volinfo, _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to create volinfo"); + goto out; + } + + /* Following entries need to be derived from origin volume. */ + strcpy (new_volinfo->volname, orig_vol->volname); + uuid_copy (new_volinfo->volume_id, orig_vol->volume_id); + new_volinfo->snap_count = orig_vol->snap_count; + new_volinfo->snap_max_hard_limit = orig_vol->snap_max_hard_limit; + new_volinfo->is_volume_restored = _gf_true; + + /* Bump the version of the restored volume, so that nodes * + * which are done can sync during handshake */ + new_volinfo->version = orig_vol->version; + + list_for_each_entry_safe (voliter, temp_volinfo, + &orig_vol->snap_volumes, snapvol_list) { + list_add_tail (&voliter->snapvol_list, + &new_volinfo->snap_volumes); + } + /* Copy the snap vol info to the new_volinfo.*/ + ret = glusterd_snap_volinfo_restore (rsp_dict, new_volinfo, snap_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to restore snap"); + (void)glusterd_volinfo_delete (new_volinfo); + goto out; + } + + /* If the orig_vol is already restored then we should delete + * the backend LVMs */ + if (orig_vol->is_volume_restored) { + ret = glusterd_lvm_snapshot_remove (rsp_dict, orig_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to remove " + "LVM backend"); + (void)glusterd_volinfo_delete (new_volinfo); + goto out; + } + } + + /* Once the new_volinfo is completely constructed then delete + * the orinal volinfo + */ + ret = glusterd_volinfo_delete (orig_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to delete volinfo"); + (void)glusterd_volinfo_delete (new_volinfo); + goto out; + } + + /* New volinfo always shows the status as created. Therefore + * set the status to stop. */ + glusterd_set_volume_status (new_volinfo, GLUSTERD_STATUS_STOPPED); + + list_add_tail (&new_volinfo->vol_list, &conf->volumes); + + /* Now delete the snap entry. As a first step delete the snap + * volume information stored in store. */ + ret = glusterd_snap_remove (rsp_dict, snap, _gf_false, _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Failed to delete " + "snap %s", snap->snapname); + goto out; + } + + ret = glusterd_store_volinfo (new_volinfo, + GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to store volinfo"); + goto out; + } + + ret = 0; +out: + + return ret; } diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.h b/xlators/mgmt/glusterd/src/glusterd-volgen.h index 974aed934..fcbaaf93e 100644 --- a/xlators/mgmt/glusterd/src/glusterd-volgen.h +++ b/xlators/mgmt/glusterd/src/glusterd-volgen.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _GLUSTERD_VOLGEN_H_ #define _GLUSTERD_VOLGEN_H_ @@ -33,8 +23,22 @@ #define VKEY_DIAG_LAT_MEASUREMENT "diagnostics.latency-measurement" #define VKEY_FEATURES_LIMIT_USAGE "features.limit-usage" #define VKEY_MARKER_XTIME GEOREP".indexing" +#define VKEY_MARKER_XTIME_FORCE GEOREP".ignore-pid-check" +#define VKEY_CHANGELOG "changelog.changelog" #define VKEY_FEATURES_QUOTA "features.quota" -#define VKEY_PERF_STAT_PREFETCH "performance.stat-prefetch" + +#define AUTH_ALLOW_MAP_KEY "auth.allow" +#define AUTH_REJECT_MAP_KEY "auth.reject" +#define NFS_DISABLE_MAP_KEY "nfs.disable" +#define AUTH_ALLOW_OPT_KEY "auth.addr.*.allow" +#define AUTH_REJECT_OPT_KEY "auth.addr.*.reject" +#define NFS_DISABLE_OPT_KEY "nfs.*.disable" + + +typedef enum { + GF_CLIENT_TRUSTED, + GF_CLIENT_OTHER +} glusterd_client_type_t; #define COMPLETE_OPTION(key, completion, ret) \ do { \ @@ -58,9 +62,54 @@ typedef enum gd_volopt_flags_ { OPT_FLAG_NONE, - OPT_FLAG_FORCE = 1, + OPT_FLAG_FORCE = 0x01, // option needs force to be reset + OPT_FLAG_XLATOR_OPT = 0x02, // option enables/disables xlators + OPT_FLAG_CLIENT_OPT = 0x04, // option affects clients } gd_volopt_flags_t; +typedef enum { + GF_XLATOR_POSIX = 0, + GF_XLATOR_ACL, + GF_XLATOR_LOCKS, + GF_XLATOR_IOT, + GF_XLATOR_INDEX, + GF_XLATOR_MARKER, + GF_XLATOR_IO_STATS, + GF_XLATOR_BD, + GF_XLATOR_NONE, +} glusterd_server_xlator_t; + +/* As of now debug xlators can be loaded only below fuse in the client + * graph via cli. More xlators can be added below when the cli option + * for adding debug xlators anywhere in the client graph has to be made + * available. + */ +typedef enum { + GF_CLNT_XLATOR_FUSE = 0, + GF_CLNT_XLATOR_NONE, +} glusterd_client_xlator_t; + +typedef enum { DOC, NO_DOC, GLOBAL_DOC, GLOBAL_NO_DOC } option_type_t; + +typedef int (*vme_option_validation) (dict_t *dict, char *key, char *value, + char **op_errstr); + +struct volopt_map_entry { + char *key; + char *voltype; + char *option; + char *value; + option_type_t type; + uint32_t flags; + uint32_t op_version; + char *description; + vme_option_validation validate_fn; + /* If client_option is true, the option affects clients. + * this is used to calculate client-op-version of volumes + */ + //gf_boolean_t client_option; +}; + int glusterd_create_rb_volfiles (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo); @@ -75,6 +124,10 @@ int glusterd_create_shd_volfile (); int glusterd_delete_volfile (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo); +int +glusterd_delete_snap_volfile (glusterd_volinfo_t *volinfo, + glusterd_volinfo_t *snap_volinfo, + glusterd_brickinfo_t *brickinfo); int glusterd_volinfo_get (glusterd_volinfo_t *volinfo, char *key, char **value); int glusterd_volinfo_get_boolean (glusterd_volinfo_t *volinfo, char *key); @@ -88,7 +141,36 @@ glusterd_check_voloption_flags (char *key, int32_t flags); gf_boolean_t glusterd_is_valid_volfpath (char *volname, char *brick); int generate_brick_volfiles (glusterd_volinfo_t *volinfo); -int glusterd_get_volopt_content (gf_boolean_t xml_out); +int generate_snap_brick_volfiles (glusterd_volinfo_t *volinfo, + glusterd_volinfo_t *snap_volinfo); +int generate_client_volfiles (glusterd_volinfo_t *volinfo, + glusterd_client_type_t client_type); +int +generate_snap_client_volfiles (glusterd_volinfo_t *actual_volinfo, + glusterd_volinfo_t *snap_volinfo, + glusterd_client_type_t client_type, + gf_boolean_t vol_restore); +int glusterd_get_volopt_content (dict_t *dict, gf_boolean_t xml_out); char* glusterd_get_trans_type_rb (gf_transport_type ttype); +int +glusterd_check_nfs_volfile_identical (gf_boolean_t *identical); +int +glusterd_check_nfs_topology_identical (gf_boolean_t *identical); + +uint32_t +glusterd_get_op_version_for_key (char *key); + +gf_boolean_t +gd_is_client_option (char *key); + +gf_boolean_t +gd_is_xlator_option (char *key); + +gf_boolean_t +gd_is_boolean_option (char *key); + +int gd_restore_snap_volume (dict_t *rsp_dict, + glusterd_volinfo_t *orig_vol, + glusterd_volinfo_t *snap_vol); #endif diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c index 1389f78f9..0d322b9ad 100644 --- a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c +++ b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c @@ -1,28 +1,23 @@ /* - Copyright (c) 2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" #endif +#ifdef HAVE_BD_XLATOR +#include <lvm2app.h> +#endif + #include "common-utils.h" +#include "syscall.h" #include "cli1-xdr.h" #include "xdr-generic.h" #include "glusterd.h" @@ -35,20 +30,16 @@ #define glusterd_op_start_volume_args_get(dict, volname, flags) \ glusterd_op_stop_volume_args_get (dict, volname, flags) + int -glusterd_handle_create_volume (rpcsvc_request_t *req) +__glusterd_handle_create_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; dict_t *dict = NULL; - glusterd_brickinfo_t *brickinfo = NULL; - char *brick = NULL; char *bricks = NULL; char *volname = NULL; - int brick_count = 0; - char *tmpptr = NULL; - int i = 0; - char *brick_list = NULL; + int brick_count = 0; void *cli_rsp = NULL; char err_str[2048] = {0,}; gf_cli_rsp rsp = {0,}; @@ -56,27 +47,27 @@ glusterd_handle_create_volume (rpcsvc_request_t *req) char *free_ptr = NULL; char *trans_type = NULL; uuid_t volume_id = {0,}; - glusterd_brickinfo_t *tmpbrkinfo = NULL; - glusterd_volinfo_t tmpvolinfo = {{0},}; - int32_t type = 0; + uuid_t tmp_uuid = {0}; + int32_t type = 0; + char *username = NULL; + char *password = NULL; GF_ASSERT (req); - INIT_LIST_HEAD (&tmpvolinfo.bricks); - this = THIS; GF_ASSERT(this); ret = -1; - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { - //failed to decode msg; + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { req->rpc_err = GARBAGE_ARGS; - snprintf (err_str, sizeof (err_str), "Garbage args received"); + snprintf (err_str, sizeof (err_str), "Failed to decode request " + "received from cli"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } - gf_log ("glusterd", GF_LOG_INFO, "Received create volume req"); + gf_log (this->name, GF_LOG_DEBUG, "Received create volume req"); if (cli_req.dict.dict_len) { /* Unserialize the dictionary */ @@ -86,11 +77,11 @@ glusterd_handle_create_volume (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); snprintf (err_str, sizeof (err_str), "Unable to decode " - "the buffer"); + "the command"); goto out; } else { dict->extra_stdfree = cli_req.dict.dict_val; @@ -100,47 +91,52 @@ glusterd_handle_create_volume (rpcsvc_request_t *req) ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); snprintf (err_str, sizeof (err_str), "Unable to get volume " "name"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } - gf_cmd_log ("Volume create", "on volname: %s attempted", volname); if ((ret = glusterd_check_volume_exists (volname))) { - snprintf(err_str, 2048, "Volume %s already exists", volname); - gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str); + snprintf (err_str, sizeof (err_str), "Volume %s already exists", + volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } ret = dict_get_int32 (dict, "count", &brick_count); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get count"); - snprintf (err_str, sizeof (err_str), "Unable to get volume " - "brick count"); + snprintf (err_str, sizeof (err_str), "Unable to get brick count" + " for volume %s", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } ret = dict_get_int32 (dict, "type", &type); - /*if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get type"); - snprintf (err_str, sizeof (err_str), "Unable to get volume " - "type"); + if (ret) { + snprintf (err_str, sizeof (err_str), "Unable to get type of " + "volume %s", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; - }*/ + } ret = dict_get_str (dict, "transport", &trans_type); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get transport-type"); - snprintf (err_str, sizeof (err_str), "Unable to get volume " - "transport-type"); + snprintf (err_str, sizeof (err_str), "Unable to get " + "transport-type of volume %s", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } ret = dict_get_str (dict, "bricks", &bricks); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get bricks"); - snprintf (err_str, sizeof (err_str), "Unable to get volume " - "bricks"); + snprintf (err_str, sizeof (err_str), "Unable to get bricks for " + "volume %s", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + goto out; + } + + if (!dict_get (dict, "force")) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get 'force' flag"); goto out; } @@ -148,102 +144,82 @@ glusterd_handle_create_volume (rpcsvc_request_t *req) free_ptr = gf_strdup (uuid_utoa (volume_id)); ret = dict_set_dynstr (dict, "volume-id", free_ptr); if (ret) { - gf_log ("", GF_LOG_ERROR, "unable to set volume-id"); snprintf (err_str, sizeof (err_str), "Unable to set volume " - "id"); + "id of volume %s", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } free_ptr = NULL; - if (bricks) { - brick_list = gf_strdup (bricks); - free_ptr = brick_list; - } - - gf_cmd_log ("Volume create", "on volname: %s type:%s count:%d bricks:%s", - volname, ((type == 0)? "DEFAULT": - ((type == 1)? "STRIPE":"REPLICATE")), brick_count, bricks); - + /* generate internal username and password */ - while ( i < brick_count) { - i++; - brick= strtok_r (brick_list, " \n", &tmpptr); - brick_list = tmpptr; - ret = glusterd_brickinfo_from_brick (brick, &brickinfo); - if (ret) { - snprintf (err_str, sizeof (err_str), "Unable to get " - "brick info from brick %s", brick); - goto out; - } + uuid_generate (tmp_uuid); + username = gf_strdup (uuid_utoa (tmp_uuid)); + ret = dict_set_dynstr (dict, "internal-username", username); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set username for " + "volume %s", volname); + goto out; + } - ret = glusterd_new_brick_validate (brick, brickinfo, err_str, - sizeof (err_str)); - if (ret) - goto out; - ret = glusterd_volume_brickinfo_get (brickinfo->uuid, - brickinfo->hostname, - brickinfo->path, - &tmpvolinfo, &tmpbrkinfo, - GF_PATH_PARTIAL); - if (!ret) { - ret = -1; - snprintf (err_str, sizeof (err_str), "Brick: %s:%s, %s" - " one of the bricks contain the other", - tmpbrkinfo->hostname, tmpbrkinfo->path, brick); - goto out; - } - list_add_tail (&brickinfo->brick_list, &tmpvolinfo.bricks); - brickinfo = NULL; + uuid_generate (tmp_uuid); + password = gf_strdup (uuid_utoa (tmp_uuid)); + ret = dict_set_dynstr (dict, "internal-password", password); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set password for " + "volume %s", volname); + goto out; } - ret = glusterd_op_begin (req, GD_OP_CREATE_VOLUME, dict); - gf_cmd_log ("Volume create", "on volname: %s %s", volname, - (ret != 0) ? "FAILED": "SUCCESS"); + ret = glusterd_op_begin_synctask (req, GD_OP_CREATE_VOLUME, dict); out: if (ret) { - if (dict) - dict_unref (dict); rsp.op_ret = -1; rsp.op_errno = 0; if (err_str[0] == '\0') - snprintf (err_str, sizeof (err_str), "Operation failed"); + snprintf (err_str, sizeof (err_str), + "Operation failed"); rsp.op_errstr = err_str; cli_rsp = &rsp; - glusterd_submit_reply(req, cli_rsp, NULL, 0, NULL, - (xdrproc_t)xdr_gf_cli_rsp); - + glusterd_to_cli (req, cli_rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_cli_rsp, dict); ret = 0; //Client response sent, prevent second response } - if (free_ptr) - GF_FREE(free_ptr); - - glusterd_volume_brickinfos_delete (&tmpvolinfo); - if (brickinfo) - glusterd_brickinfo_delete (brickinfo); - - glusterd_friend_sm (); - glusterd_op_sm (); + GF_FREE(free_ptr); return ret; } int -glusterd_handle_cli_start_volume (rpcsvc_request_t *req) +glusterd_handle_create_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_create_volume); +} + +int +__glusterd_handle_cli_start_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; - char *dup_volname = NULL; + char *volname = NULL; dict_t *dict = NULL; glusterd_op_t cli_op = GD_OP_START_VOLUME; + char errstr[2048] = {0,}; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { - //failed to decode msg; + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + snprintf (errstr, sizeof (errstr), "Failed to decode message " + "received from cli"); req->rpc_err = GARBAGE_ARGS; + gf_log (this->name, sizeof (errstr), "%s", errstr); goto out; } @@ -255,58 +231,68 @@ glusterd_handle_cli_start_volume (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (errstr, sizeof (errstr), "Unable to decode " + "the command"); goto out; } } - ret = dict_get_str (dict, "volname", &dup_volname); + ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "dict get failed"); + snprintf (errstr, sizeof (errstr), "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "%s", errstr); goto out; } - gf_log ("glusterd", GF_LOG_INFO, "Received start vol req" - "for volume %s", dup_volname); - ret = glusterd_op_begin (req, GD_OP_START_VOLUME, dict); + gf_log (this->name, GF_LOG_DEBUG, "Received start vol req" + " for volume %s", volname); - gf_cmd_log ("volume start","on volname: %s %s", dup_volname, - ((ret == 0) ? "SUCCESS": "FAILED")); + ret = glusterd_op_begin_synctask (req, GD_OP_START_VOLUME, dict); out: - if (ret && dict) - dict_unref (dict); - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val); //its malloced by xdr + free (cli_req.dict.dict_val); //its malloced by xdr - glusterd_friend_sm (); - glusterd_op_sm (); - - if (ret) + if (ret) { + if(errstr[0] == '\0') + snprintf (errstr, sizeof (errstr), "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, errstr); + } return ret; } +int +glusterd_handle_cli_start_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_start_volume); +} int -glusterd_handle_cli_stop_volume (rpcsvc_request_t *req) +__glusterd_handle_cli_stop_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; char *dup_volname = NULL; dict_t *dict = NULL; glusterd_op_t cli_op = GD_OP_STOP_VOLUME; + xlator_t *this = NULL; + char err_str[2048] = {0,}; + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { - //failed to decode msg; + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + snprintf (err_str, sizeof (err_str), "Failed to decode message " + "received from cli"); req->rpc_err = GARBAGE_ARGS; + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } if (cli_req.dict.dict_len) { @@ -317,9 +303,11 @@ glusterd_handle_cli_stop_volume (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); goto out; } } @@ -327,48 +315,59 @@ glusterd_handle_cli_stop_volume (rpcsvc_request_t *req) ret = dict_get_str (dict, "volname", &dup_volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + snprintf (err_str, sizeof (err_str), "Failed to get volume " + "name"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } - gf_log ("glusterd", GF_LOG_INFO, "Received stop vol req" + gf_log (this->name, GF_LOG_DEBUG, "Received stop vol req " "for volume %s", dup_volname); - ret = glusterd_op_begin (req, GD_OP_STOP_VOLUME, dict); - gf_cmd_log ("Volume stop","on volname: %s %s", dup_volname, - ((ret)?"FAILED":"SUCCESS")); + ret = glusterd_op_begin_synctask (req, GD_OP_STOP_VOLUME, dict); out: - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val); //its malloced by xdr - - glusterd_friend_sm (); - glusterd_op_sm (); + free (cli_req.dict.dict_val); //its malloced by xdr if (ret) { - if (dict) - dict_unref (dict); + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, err_str); } return ret; } int -glusterd_handle_cli_delete_volume (rpcsvc_request_t *req) +glusterd_handle_cli_stop_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_stop_volume); +} + +int +__glusterd_handle_cli_delete_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,},}; glusterd_op_t cli_op = GD_OP_DELETE_VOLUME; dict_t *dict = NULL; char *volname = NULL; + char err_str[2048]= {0,}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { - //failed to decode msg; + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + snprintf (err_str, sizeof (err_str), "Failed to decode request " + "received from cli"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); req->rpc_err = GARBAGE_ARGS; goto out; } @@ -381,65 +380,74 @@ glusterd_handle_cli_delete_volume (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); goto out; } } ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, - "Failed to get volname"); + snprintf (err_str, sizeof (err_str), "Failed to get volume " + "name"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); req->rpc_err = GARBAGE_ARGS; goto out; } - gf_cmd_log ("Volume delete","on volname: %s attempted", volname); - - gf_log ("glusterd", GF_LOG_INFO, "Received delete vol req" + gf_log (this->name, GF_LOG_DEBUG, "Received delete vol req" "for volume %s", volname); - ret = glusterd_op_begin (req, GD_OP_DELETE_VOLUME, dict); - gf_cmd_log ("Volume delete", "on volname: %s %s", volname, - ((ret) ? "FAILED" : "SUCCESS")); + ret = glusterd_op_begin_synctask (req, GD_OP_DELETE_VOLUME, dict); out: - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val); //its malloced by xdr - if (ret && dict) - dict_unref (dict); - - glusterd_friend_sm (); - glusterd_op_sm (); + free (cli_req.dict.dict_val); //its malloced by xdr if (ret) { + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, err_str); } return ret; } int -glusterd_handle_cli_heal_volume (rpcsvc_request_t *req) +glusterd_handle_cli_delete_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_delete_volume); +} + +int +__glusterd_handle_cli_heal_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; dict_t *dict = NULL; glusterd_op_t cli_op = GD_OP_HEAL_VOLUME; char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + xlator_t *this = NULL; + char op_errstr[2048] = {0,}; GF_ASSERT (req); - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { //failed to decode msg; req->rpc_err = GARBAGE_ARGS; goto out; } + this = THIS; + GF_ASSERT (this); + if (cli_req.dict.dict_len) { /* Unserialize the dictionary */ dict = dict_new (); @@ -448,44 +456,67 @@ glusterd_handle_cli_heal_volume (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (op_errstr, sizeof (op_errstr), + "Unable to decode the command"); goto out; + } else { + dict->extra_stdfree = cli_req.dict.dict_val; } } ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to get volname"); + snprintf (op_errstr, sizeof (op_errstr), "Unable to find " + "volume name"); + gf_log (this->name, GF_LOG_ERROR, "%s", op_errstr); goto out; } - gf_log ("glusterd", GF_LOG_INFO, "Received heal vol req" + + gf_log (this->name, GF_LOG_INFO, "Received heal vol req " "for volume %s", volname); - ret = glusterd_op_begin (req, GD_OP_HEAL_VOLUME, dict); + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (op_errstr, sizeof (op_errstr), + "Volume %s does not exist", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", op_errstr); + goto out; + } - gf_cmd_log ("volume heal","on volname: %s %s", volname, - ((ret == 0) ? "SUCCESS": "FAILED")); + ret = glusterd_add_bricks_hname_path_to_dict (dict, volinfo); + if (ret) + goto out; -out: - if (ret && dict) - dict_unref (dict); - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val); //its malloced by xdr + ret = dict_set_int32 (dict, "count", volinfo->brick_count); + if (ret) + goto out; - glusterd_friend_sm (); - glusterd_op_sm (); + ret = glusterd_op_begin_synctask (req, GD_OP_HEAL_VOLUME, dict); - if (ret) +out: + if (ret) { + if (op_errstr[0] == '\0') + snprintf (op_errstr, sizeof (op_errstr), + "operation failed"); ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, - NULL, "operation failed"); + dict, op_errstr); + } return ret; } int -glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req) +glusterd_handle_cli_heal_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_heal_volume); +} + +int +__glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; @@ -493,12 +524,18 @@ glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req) char *options = NULL; dict_t *dict = NULL; int32_t option_cnt = 0; + glusterd_op_t cli_op = GD_OP_STATEDUMP_VOLUME; + char err_str[2048] = {0,}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); GF_ASSERT (req); ret = -1; - if (!xdr_to_generic (req->msg[0], &cli_req, - (xdrproc_t)xdr_gf_cli_req)) { + ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { req->rpc_err = GARBAGE_ARGS; goto out; } @@ -510,50 +547,158 @@ glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req) cli_req.dict.dict_len, &dict); if (ret < 0) { - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "failed to " "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to " + "decode the command"); goto out; } } ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to get volname"); + snprintf (err_str, sizeof (err_str), "Unable to get the volume " + "name"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } ret = dict_get_str (dict, "options", &options); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to get options"); + snprintf (err_str, sizeof (err_str), "Unable to get options"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } ret = dict_get_int32 (dict, "option_cnt", &option_cnt); if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "failed to get option cnt"); + snprintf (err_str , sizeof (err_str), "Unable to get option " + "count"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); goto out; } - gf_log ("glusterd", GF_LOG_INFO, "Recieved statedump request for " + gf_log (this->name, GF_LOG_INFO, "Received statedump request for " "volume %s with options %s", volname, options); - ret = glusterd_op_begin (req, GD_OP_STATEDUMP_VOLUME, dict); - - gf_cmd_log ("statedump", "on volume %s %s", volname, - ((0 == ret) ? "SUCCEEDED" : "FAILED")); + ret = glusterd_op_begin_synctask (req, GD_OP_STATEDUMP_VOLUME, dict); out: - if (ret && dict) - dict_unref (dict); - if (cli_req.dict.dict_val) - free (cli_req.dict.dict_val); - glusterd_friend_sm (); - glusterd_op_sm(); + if (ret) { + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); + ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, + dict, err_str); + } + free (cli_req.dict.dict_val); return ret; } +int +glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_cli_statedump_volume); +} + +#ifdef HAVE_BD_XLATOR +/* + * Validates if given VG in the brick exists or not. Also checks if VG has + * GF_XATTR_VOL_ID_KEY tag set to avoid using same VG for multiple bricks. + * Tag is checked only during glusterd_op_stage_create_volume. Tag is set during + * glusterd_validate_and_create_brickpath(). + * @brick - brick info, @check_tag - check for VG tag or not + * @msg - Error message to return to caller + */ +int +glusterd_is_valid_vg (glusterd_brickinfo_t *brick, int check_tag, char *msg) +{ + lvm_t handle = NULL; + vg_t vg = NULL; + char *vg_name = NULL; + int retval = 0; + char *p = NULL; + char *ptr = NULL; + struct dm_list *dm_lvlist = NULL; + struct dm_list *dm_seglist = NULL; + struct lvm_lv_list *lv_list = NULL; + struct lvm_property_value prop = {0, }; + struct lvm_lvseg_list *seglist = NULL; + struct dm_list *taglist = NULL; + struct lvm_str_list *strl = NULL; + + handle = lvm_init (NULL); + if (!handle) { + sprintf (msg, "lvm_init failed, could not validate vg"); + return -1; + } + if (*brick->vg == '\0') { /* BD xlator has vg in brick->path */ + p = gf_strdup (brick->path); + vg_name = strtok_r (p, "/", &ptr); + } else + vg_name = brick->vg; + + vg = lvm_vg_open (handle, vg_name, "r", 0); + if (!vg) { + sprintf (msg, "no such vg: %s", vg_name); + retval = -1; + goto out; + } + if (!check_tag) + goto next; + + taglist = lvm_vg_get_tags (vg); + if (!taglist) + goto next; + + dm_list_iterate_items (strl, taglist) { + if (!strncmp(strl->str, GF_XATTR_VOL_ID_KEY, + strlen (GF_XATTR_VOL_ID_KEY))) { + sprintf (msg, "VG %s is already part of" + " a brick", vg_name); + retval = -1; + goto out; + } + } +next: + + brick->caps = CAPS_BD | CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT; + + dm_lvlist = lvm_vg_list_lvs (vg); + if (!dm_lvlist) + goto out; + + dm_list_iterate_items (lv_list, dm_lvlist) { + dm_seglist = lvm_lv_list_lvsegs (lv_list->lv); + dm_list_iterate_items (seglist, dm_seglist) { + prop = lvm_lvseg_get_property (seglist->lvseg, + "segtype"); + if (!prop.is_valid || !prop.value.string) + continue; + if (!strcmp (prop.value.string, "thin-pool")) { + brick->caps |= CAPS_THIN; + gf_log (THIS->name, GF_LOG_INFO, "Thin Pool " + "\"%s\" will be used for thin LVs", + lvm_lv_get_name (lv_list->lv)); + break; + } + } + } + + retval = 0; +out: + if (vg) + lvm_vg_close (vg); + lvm_quit (handle); + if (p) + GF_FREE (p); + return retval; +} +#endif + /* op-sm */ int glusterd_op_stage_create_volume (dict_t *dict, char **op_errstr) @@ -569,73 +714,68 @@ glusterd_op_stage_create_volume (dict_t *dict, char **op_errstr) int32_t i = 0; char *brick = NULL; char *tmpptr = NULL; - char cmd_str[1024]; xlator_t *this = NULL; glusterd_conf_t *priv = NULL; char msg[2048] = {0}; uuid_t volume_uuid; char *volume_uuid_str; + gf_boolean_t is_force = _gf_false; this = THIS; - if (!this) { - gf_log ("glusterd", GF_LOG_ERROR, - "this is NULL"); - goto out; - } - + GF_ASSERT (this); priv = this->private; - if (!priv) { - gf_log ("glusterd", GF_LOG_ERROR, - "priv is NULL"); - goto out; - } + GF_ASSERT (priv); ret = dict_get_str (dict, "volname", &volname); - if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } exists = glusterd_check_volume_exists (volname); - if (exists) { snprintf (msg, sizeof (msg), "Volume %s already exists", volname); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); ret = -1; goto out; } else { ret = 0; } + ret = dict_get_int32 (dict, "count", &brick_count); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get count"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get brick count " + "for volume %s", volname); goto out; } + ret = dict_get_str (dict, "volume-id", &volume_uuid_str); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume id"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume id of " + "volume %s", volname); goto out; } + ret = uuid_parse (volume_uuid_str, volume_uuid); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to parse volume id"); + gf_log (this->name, GF_LOG_ERROR, "Unable to parse volume id of" + " volume %s", volname); goto out; } ret = dict_get_str (dict, "bricks", &bricks); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get bricks"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get bricks for " + "volume %s", volname); goto out; } + is_force = dict_get_str_boolean (dict, "force", _gf_false); + if (bricks) { brick_list = gf_strdup (bricks); if (!brick_list) { ret = -1; - gf_log ("", GF_LOG_ERROR, "Out of memory"); goto out; } else { free_ptr = brick_list; @@ -651,30 +791,38 @@ glusterd_op_stage_create_volume (dict_t *dict, char **op_errstr) !glusterd_is_valid_volfpath (volname, brick)) { snprintf (msg, sizeof (msg), "brick path %s is too " "long.", brick); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - ret = -1; goto out; } - ret = glusterd_brickinfo_from_brick (brick, &brick_info); + ret = glusterd_brickinfo_new_from_brick (brick, &brick_info); if (ret) goto out; - snprintf (cmd_str, 1024, "%s", brick_info->path); + + ret = glusterd_new_brick_validate (brick, brick_info, msg, + sizeof (msg)); + if (ret) + goto out; + ret = glusterd_resolve_brick (brick_info); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "cannot resolve " - "brick: %s:%s", brick_info->hostname, - brick_info->path); + gf_log (this->name, GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK, + brick_info->hostname, brick_info->path); goto out; } - if (!uuid_compare (brick_info->uuid, priv->uuid)) { - ret = glusterd_brick_create_path (brick_info->hostname, - brick_info->path, - volume_uuid, - 0777, op_errstr); + if (!uuid_compare (brick_info->uuid, MY_UUID)) { + +#ifdef HAVE_BD_XLATOR + if (brick_info->vg[0]) { + ret = glusterd_is_valid_vg (brick_info, 1, msg); + if (ret) + goto out; + } +#endif + ret = glusterd_validate_and_create_brickpath (brick_info, + volume_uuid, op_errstr, + is_force); if (ret) goto out; brick_list = tmpptr; @@ -683,11 +831,15 @@ glusterd_op_stage_create_volume (dict_t *dict, char **op_errstr) brick_info = NULL; } out: - if (free_ptr) - GF_FREE (free_ptr); + GF_FREE (free_ptr); if (brick_info) glusterd_brickinfo_delete (brick_info); - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + + if (msg[0] != '\0') { + gf_log (this->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + } + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -696,19 +848,23 @@ int glusterd_op_stop_volume_args_get (dict_t *dict, char** volname, int *flags) { int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); if (!dict || !volname || !flags) goto out; ret = dict_get_str (dict, "volname", volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } ret = dict_get_int32 (dict, "flags", flags); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get flags"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get flags"); goto out; } out: @@ -757,14 +913,16 @@ glusterd_op_stage_start_volume (dict_t *dict, char **op_errstr) glusterd_brickinfo_t *brickinfo = NULL; char msg[2048]; glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + uuid_t volume_id = {0,}; + char volid[50] = {0,}; + char xattr_volid[50] = {0,}; + int caps = 0; - priv = THIS->private; - if (!priv) { - gf_log ("glusterd", GF_LOG_ERROR, - "priv is NULL"); - ret = -1; - goto out; - } + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); ret = glusterd_op_start_volume_args_get (dict, &volname, &flags); if (ret) @@ -773,53 +931,110 @@ glusterd_op_stage_start_volume (dict_t *dict, char **op_errstr) exists = glusterd_check_volume_exists (volname); if (!exists) { - snprintf (msg, sizeof (msg), "Volume %s does not exist", volname); - gf_log ("", GF_LOG_ERROR, "%s", - msg); - *op_errstr = gf_strdup (msg); + snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname); ret = -1; - } else { - ret = 0; + goto out; } ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS, + volname); + goto out; + } + + ret = glusterd_validate_volume_id (dict, volinfo); if (ret) goto out; + if (!(flags & GF_CLI_FLAG_OP_FORCE)) { + if (glusterd_is_volume_started (volinfo)) { + snprintf (msg, sizeof (msg), "Volume %s already " + "started", volname); + ret = -1; + goto out; + } + } + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { ret = glusterd_resolve_brick (brickinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, - "Unable to resolve brick %s:%s", + gf_log (this->name, GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK, brickinfo->hostname, brickinfo->path); goto out; } - if (!uuid_compare (brickinfo->uuid, priv->uuid)) { - ret = glusterd_brick_create_path (brickinfo->hostname, - brickinfo->path, - volinfo->volume_id, - 0777, op_errstr); - if (ret) - goto out; + if (uuid_compare (brickinfo->uuid, MY_UUID)) + continue; + + ret = gf_lstat_dir (brickinfo->path, NULL); + if (ret && (flags & GF_CLI_FLAG_OP_FORCE)) { + continue; + } else if (ret) { + snprintf (msg, sizeof (msg), "Failed to find " + "brick directory %s for volume %s. " + "Reason : %s", brickinfo->path, + volname, strerror (errno)); + goto out; } - - if (!(flags & GF_CLI_FLAG_OP_FORCE)) { - if (glusterd_is_volume_started (volinfo)) { - snprintf (msg, sizeof (msg), "Volume %s already" - " started", volname); - gf_log ("glusterd", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - ret = -1; + ret = sys_lgetxattr (brickinfo->path, GF_XATTR_VOL_ID_KEY, + volume_id, 16); + if (ret < 0 && (!(flags & GF_CLI_FLAG_OP_FORCE))) { + snprintf (msg, sizeof (msg), "Failed to get " + "extended attribute %s for brick dir %s. " + "Reason : %s", GF_XATTR_VOL_ID_KEY, + brickinfo->path, strerror (errno)); + ret = -1; + goto out; + } else if (ret < 0) { + ret = sys_lsetxattr (brickinfo->path, + GF_XATTR_VOL_ID_KEY, + volinfo->volume_id, 16, + XATTR_CREATE); + if (ret) { + snprintf (msg, sizeof (msg), "Failed to set " + "extended attribute %s on %s. Reason: " + "%s", GF_XATTR_VOL_ID_KEY, + brickinfo->path, strerror (errno)); goto out; + } else { + continue; } } + if (uuid_compare (volinfo->volume_id, volume_id)) { + snprintf (msg, sizeof (msg), "Volume id mismatch for " + "brick %s:%s. Expected volume id %s, " + "volume id %s found", brickinfo->hostname, + brickinfo->path, + uuid_utoa_r (volinfo->volume_id, volid), + uuid_utoa_r (volume_id, xattr_volid)); + ret = -1; + goto out; + } +#ifdef HAVE_BD_XLATOR + if (brickinfo->vg[0]) + caps = CAPS_BD | CAPS_THIN | + CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT; + /* Check for VG/thin pool if its BD volume */ + if (brickinfo->vg[0]) { + ret = glusterd_is_valid_vg (brickinfo, 0, msg); + if (ret) + goto out; + /* if anyone of the brick does not have thin support, + disable it for entire volume */ + caps &= brickinfo->caps; + } else + caps = 0; +#endif } + volinfo->caps = caps; ret = 0; out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - + if (ret && (msg[0] != '\0')) { + gf_log (this->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + } return ret; } @@ -833,7 +1048,10 @@ glusterd_op_stage_stop_volume (dict_t *dict, char **op_errstr) gf_boolean_t is_run = _gf_false; glusterd_volinfo_t *volinfo = NULL; char msg[2048] = {0}; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); ret = glusterd_op_stop_volume_args_get (dict, &volname, &flags); if (ret) @@ -842,16 +1060,20 @@ glusterd_op_stage_stop_volume (dict_t *dict, char **op_errstr) exists = glusterd_check_volume_exists (volname); if (!exists) { - snprintf (msg, sizeof (msg), "Volume %s does not exist", volname); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); + snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); ret = -1; goto out; - } else { - ret = 0; } ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); + goto out; + } + + ret = glusterd_validate_volume_id (dict, volinfo); if (ret) goto out; @@ -862,48 +1084,55 @@ glusterd_op_stage_stop_volume (dict_t *dict, char **op_errstr) if (_gf_false == glusterd_is_volume_started (volinfo)) { snprintf (msg, sizeof(msg), "Volume %s " "is not in the started state", volname); - gf_log ("", GF_LOG_ERROR, "Volume %s " - "has not been started", volname); - *op_errstr = gf_strdup (msg); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); ret = -1; goto out; } ret = glusterd_check_gsync_running (volinfo, &is_run); if (ret && (is_run == _gf_false)) - gf_log ("", GF_LOG_WARNING, "Unable to get the status" + gf_log (this->name, GF_LOG_WARNING, "Unable to get the status" " of active "GEOREP" session"); if (is_run) { - gf_log ("", GF_LOG_WARNING, GEOREP" sessions active" + gf_log (this->name, GF_LOG_WARNING, GEOREP" sessions active" "for the volume %s ", volname); snprintf (msg, sizeof(msg), GEOREP" sessions are active " "for the volume '%s'.\nUse 'volume "GEOREP" " - "status' command for more info. Use 'force'" - "option to ignore and stop stop the volume", + "status' command for more info. Use 'force' " + "option to ignore and stop the volume.", volname); - *op_errstr = gf_strdup (msg); ret = -1; goto out; } + + if (glusterd_is_rb_ongoing (volinfo)) { + snprintf (msg, sizeof (msg), "Replace brick is in progress on " + "volume %s. Please retry after replace-brick " + "operation is committed or aborted", volname); + gf_log (this->name, GF_LOG_WARNING, "replace-brick in progress " + "on volume %s", volname); + ret = -1; + goto out; + } + if (glusterd_is_defrag_on (volinfo)) { snprintf (msg, sizeof(msg), "rebalance session is " "in progress for the volume '%s'", volname); - gf_log (THIS->name, GF_LOG_WARNING, "%s", msg); - *op_errstr = gf_strdup (msg); + gf_log (this->name, GF_LOG_WARNING, "%s", msg); ret = -1; goto out; } - if (volinfo->rb_status != GF_RB_STATUS_NONE) { + if (volinfo->rep_brick.rb_status != GF_RB_STATUS_NONE) { snprintf (msg, sizeof(msg), "replace-brick session is " "in progress for the volume '%s'", volname); - gf_log (THIS->name, GF_LOG_WARNING, "%s", msg); - *op_errstr = gf_strdup (msg); + gf_log (this->name, GF_LOG_WARNING, "%s", msg); ret = -1; goto out; } out: - - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + if (msg[0] != 0) + *op_errstr = gf_strdup (msg); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -916,21 +1145,20 @@ glusterd_op_stage_delete_volume (dict_t *dict, char **op_errstr) gf_boolean_t exists = _gf_false; glusterd_volinfo_t *volinfo = NULL; char msg[2048] = {0}; + xlator_t *this = NULL; - ret = dict_get_str (dict, "volname", &volname); + this = THIS; + GF_ASSERT (this); + ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } exists = glusterd_check_volume_exists (volname); - if (!exists) { - snprintf (msg, sizeof (msg), "Volume %s does not exist", - volname); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); + snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname); ret = -1; goto out; } else { @@ -938,7 +1166,12 @@ glusterd_op_stage_delete_volume (dict_t *dict, char **op_errstr) } ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname); + goto out; + } + ret = glusterd_validate_volume_id (dict, volinfo); if (ret) goto out; @@ -946,8 +1179,16 @@ glusterd_op_stage_delete_volume (dict_t *dict, char **op_errstr) snprintf (msg, sizeof (msg), "Volume %s has been started." "Volume needs to be stopped before deletion.", volname); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); + ret = -1; + goto out; + } + + if (volinfo->snap_count > 0 || !list_empty(&volinfo->snap_volumes)) { + snprintf (msg, sizeof (msg), "Cannot delete Volume %s ," + "as it has %ld snapshots. " + "To delete the volume, " + "first delete all the snapshots under it.", + volname, volinfo->snap_count); ret = -1; goto out; } @@ -955,7 +1196,11 @@ glusterd_op_stage_delete_volume (dict_t *dict, char **op_errstr) ret = 0; out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + if (msg[0] != '\0') { + gf_log (this->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + } + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -970,11 +1215,14 @@ glusterd_op_stage_heal_volume (dict_t *dict, char **op_errstr) char msg[2048]; glusterd_conf_t *priv = NULL; dict_t *opt_dict = NULL; + gf_xl_afr_op_t heal_op = GF_AFR_OP_INVALID; + xlator_t *this = NULL; - priv = THIS->private; + this = THIS; + priv = this->private; if (!priv) { ret = -1; - gf_log (THIS->name, GF_LOG_ERROR, + gf_log (this->name, GF_LOG_ERROR, "priv is NULL"); goto out; } @@ -989,17 +1237,21 @@ glusterd_op_stage_heal_volume (dict_t *dict, char **op_errstr) if (ret) { ret = -1; snprintf (msg, sizeof (msg), "Volume %s does not exist", volname); - gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + gf_log (this->name, GF_LOG_ERROR, "%s", msg); *op_errstr = gf_strdup (msg); goto out; } + ret = glusterd_validate_volume_id (dict, volinfo); + if (ret) + goto out; + if (!glusterd_is_volume_replicate (volinfo)) { ret = -1; - snprintf (msg, sizeof (msg), "Volume %s is not a replicate " - "type volume", volname); + snprintf (msg, sizeof (msg), "Volume %s is not of type " + "replicate", volname); *op_errstr = gf_strdup (msg); - gf_log (THIS->name, GF_LOG_WARNING, "%s", msg); + gf_log (this->name, GF_LOG_WARNING, "%s", msg); goto out; } @@ -1025,20 +1277,37 @@ glusterd_op_stage_heal_volume (dict_t *dict, char **op_errstr) snprintf (msg, sizeof (msg), "Self-heal-daemon is " "disabled. Heal will not be triggered on volume %s", volname); - gf_log (THIS->name, GF_LOG_WARNING, "%s", msg); + gf_log (this->name, GF_LOG_WARNING, "%s", msg); *op_errstr = gf_strdup (msg); goto out; } - if (!glusterd_shd_is_running ()) { + ret = dict_get_int32 (dict, "heal-op", (int32_t*)&heal_op); + if (ret || (heal_op == GF_AFR_OP_INVALID)) { ret = -1; - snprintf (msg, sizeof (msg), "Self-heal daemon is not " - "running. Check self-heal daemon log file."); - *op_errstr = gf_strdup (msg); - gf_log (THIS->name, GF_LOG_WARNING, "%s", msg); + *op_errstr = gf_strdup("Invalid heal-op"); + gf_log (this->name, GF_LOG_WARNING, "%s", "Invalid heal-op"); goto out; } + switch (heal_op) { + case GF_AFR_OP_INDEX_SUMMARY: + case GF_AFR_OP_STATISTICS_HEAL_COUNT: + case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA: + break; + default: + if (!glusterd_is_nodesvc_online("glustershd")){ + ret = -1; + *op_errstr = gf_strdup ("Self-heal daemon is " + "not running. Check self-heal " + "daemon log file."); + gf_log (this->name, GF_LOG_WARNING, "%s", + "Self-heal daemon is not running." + "Check self-heal daemon log file."); + goto out; + } + } + ret = 0; out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); @@ -1071,6 +1340,10 @@ glusterd_op_stage_statedump_volume (dict_t *dict, char **op_errstr) goto out; } + ret = glusterd_validate_volume_id (dict, volinfo); + if (ret) + goto out; + is_running = glusterd_is_volume_started (volinfo); if (!is_running) { snprintf (msg, sizeof(msg), "Volume %s is not in a started" @@ -1087,6 +1360,76 @@ out: } int +glusterd_op_stage_clearlocks_volume (dict_t *dict, char **op_errstr) +{ + int ret = -1; + char *volname = NULL; + char *path = NULL; + char *type = NULL; + char *kind = NULL; + glusterd_volinfo_t *volinfo = NULL; + char msg[2048] = {0,}; + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + snprintf (msg, sizeof(msg), "Failed to get volume name"); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + goto out; + } + + ret = dict_get_str (dict, "path", &path); + if (ret) { + snprintf (msg, sizeof(msg), "Failed to get path"); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + goto out; + } + + ret = dict_get_str (dict, "kind", &kind); + if (ret) { + snprintf (msg, sizeof(msg), "Failed to get kind"); + gf_log ("", GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + goto out; + } + + ret = dict_get_str (dict, "type", &type); + if (ret) { + snprintf (msg, sizeof(msg), "Failed to get type"); + gf_log ("", GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (msg, sizeof(msg), "Volume %s does not exist", + volname); + gf_log ("", GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + goto out; + } + + ret = glusterd_validate_volume_id (dict, volinfo); + if (ret) + goto out; + + if (!glusterd_is_volume_started (volinfo)) { + snprintf (msg, sizeof(msg), "Volume %s is not started", + volname); + gf_log ("", GF_LOG_ERROR, "%s", msg); + *op_errstr = gf_strdup (msg); + goto out; + } + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int glusterd_op_create_volume (dict_t *dict, char **op_errstr) { int ret = 0; @@ -1105,6 +1448,10 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) char *saveptr = NULL; char *trans_type = NULL; char *str = NULL; + char *username = NULL; + char *password = NULL; + int caps = 0; + char msg[1024] __attribute__((unused)) = {0, }; this = THIS; GF_ASSERT (this); @@ -1115,14 +1462,15 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) ret = glusterd_volinfo_new (&volinfo); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to allocate memory"); + gf_log (this->name, GF_LOG_ERROR, + "Unable to allocate memory for volinfo"); goto out; } ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } @@ -1131,19 +1479,21 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) ret = dict_get_int32 (dict, "type", &volinfo->type); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get type"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get type of volume" + " %s", volname); goto out; } ret = dict_get_int32 (dict, "count", &volinfo->brick_count); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get count"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get brick count of" + " volume %s", volname); goto out; } ret = dict_get_int32 (dict, "port", &volinfo->port); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get port"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get port"); goto out; } @@ -1151,7 +1501,8 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) ret = dict_get_str (dict, "bricks", &bricks); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get bricks"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get bricks for " + "volume %s", volname); goto out; } @@ -1163,28 +1514,44 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) if (GF_CLUSTER_TYPE_REPLICATE == volinfo->type) { ret = dict_get_int32 (dict, "replica-count", &volinfo->replica_count); - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "replica count for volume %s", volname); goto out; + } } else if (GF_CLUSTER_TYPE_STRIPE == volinfo->type) { ret = dict_get_int32 (dict, "stripe-count", &volinfo->stripe_count); - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get stripe" + " count for volume %s", volname); goto out; + } } else if (GF_CLUSTER_TYPE_STRIPE_REPLICATE == volinfo->type) { ret = dict_get_int32 (dict, "stripe-count", &volinfo->stripe_count); - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get stripe" + " count for volume %s", volname); goto out; + } ret = dict_get_int32 (dict, "replica-count", &volinfo->replica_count); - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "replica count for volume %s", volname); goto out; + } } /* dist-leaf-count is the count of brick nodes for a given subvolume of distribute */ - volinfo->dist_leaf_count = (volinfo->stripe_count * - volinfo->replica_count); + volinfo->dist_leaf_count = glusterd_get_dist_leaf_count (volinfo); + + /* subvol_count is the count of number of subvolumes present + for a given distribute volume */ + volinfo->subvol_count = (volinfo->brick_count / + volinfo->dist_leaf_count); /* Keep sub-count same as earlier, for the sake of backward compatibility */ @@ -1193,20 +1560,41 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) ret = dict_get_str (dict, "transport", &trans_type); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get transport"); + gf_log (this->name, GF_LOG_ERROR, + "Unable to get transport type of volume %s", volname); goto out; } ret = dict_get_str (dict, "volume-id", &str); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume-id"); + gf_log (this->name, GF_LOG_ERROR, + "Unable to get volume-id of volume %s", volname); goto out; } ret = uuid_parse (str, volinfo->volume_id); if (ret) { - gf_log ("", GF_LOG_ERROR, "unable to parse uuid %s", str); + gf_log (this->name, GF_LOG_ERROR, + "unable to parse uuid %s of volume %s", str, volname); + goto out; + } + + ret = dict_get_str (dict, "internal-username", &username); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "unable to get internal username of volume %s", + volname); + goto out; + } + glusterd_auth_set_username (volinfo, username); + + ret = dict_get_str (dict, "internal-password", &password); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "unable to get internal password of volume %s", + volname); goto out; } + glusterd_auth_set_password (volinfo, password); if (strcasecmp (trans_type, "rdma") == 0) { volinfo->transport_type = GF_TRANSPORT_RDMA; @@ -1226,20 +1614,49 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) if (count) brick = strtok_r (brick_list+1, " \n", &saveptr); + caps = CAPS_BD | CAPS_THIN | CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT; while ( i <= count) { - ret = glusterd_brickinfo_from_brick (brick, &brickinfo); + ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo); if (ret) goto out; ret = glusterd_resolve_brick (brickinfo); - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK, + brickinfo->hostname, brickinfo->path); goto out; + } + +#ifdef HAVE_BD_XLATOR + if (!uuid_compare (brickinfo->uuid, MY_UUID)) { + if (brickinfo->vg[0]) { + ret = glusterd_is_valid_vg (brickinfo, 0, msg); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + msg); + goto out; + } + + /* if anyone of the brick does not have thin + support, disable it for entire volume */ + caps &= brickinfo->caps; + + + } else + caps = 0; + } +#endif + list_add_tail (&brickinfo->brick_list, &volinfo->bricks); brick = strtok_r (NULL, " \n", &saveptr); i++; } + gd_update_volume_op_versions (volinfo); + + volinfo->caps = caps; + ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); if (ret) { glusterd_store_delete_volume (volinfo); @@ -1253,18 +1670,12 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) goto out; } - ret = glusterd_volume_compute_cksum (volinfo); - if (ret) { - *op_errstr = gf_strdup ("Failed to compute checksum of volume"); - goto out; - } - - volinfo->defrag_status = 0; + volinfo->rebal.defrag_status = 0; list_add_tail (&volinfo->vol_list, &priv->volumes); vol_added = _gf_true; + out: - if (free_ptr) - GF_FREE(free_ptr); + GF_FREE(free_ptr); if (!vol_added && volinfo) glusterd_volinfo_delete (volinfo); return ret; @@ -1278,18 +1689,28 @@ glusterd_op_start_volume (dict_t *dict, char **op_errstr) int flags = 0; glusterd_volinfo_t *volinfo = NULL; glusterd_brickinfo_t *brickinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); ret = glusterd_op_start_volume_args_get (dict, &volname, &flags); if (ret) goto out; ret = glusterd_volinfo_find (volname, &volinfo); - - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS, + volname); goto out; + } + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { - ret = glusterd_brick_start (volinfo, brickinfo); - if (ret) + ret = glusterd_brick_start (volinfo, brickinfo, _gf_true); + /* If 'force' try to start all bricks regardless of success or + * failure + */ + if (!(flags & GF_CLI_FLAG_OP_FORCE) && ret) goto out; } @@ -1302,7 +1723,48 @@ glusterd_op_start_volume (dict_t *dict, char **op_errstr) ret = glusterd_nodesvcs_handle_graph_change (volinfo); out: - gf_log ("", GF_LOG_DEBUG, "returning %d ", ret); + gf_log (this->name, GF_LOG_DEBUG, "returning %d ", ret); + return ret; +} + +int +glusterd_stop_volume (glusterd_volinfo_t *volinfo) +{ + int ret = -1; + glusterd_brickinfo_t *brickinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + GF_VALIDATE_OR_GOTO (this->name, volinfo, out); + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + ret = glusterd_brick_stop (volinfo, brickinfo, _gf_false); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to stop " + "brick (%s)", brickinfo->path); + goto out; + } + } + + glusterd_set_volume_status (volinfo, GLUSTERD_STATUS_STOPPED); + + ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to store volinfo of " + "%s volume", volinfo->volname); + goto out; + } + + ret = glusterd_nodesvcs_handle_graph_change (volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to notify graph " + "change for %s volume", volinfo->volname); + goto out; + } + +out: return ret; } @@ -1314,30 +1776,28 @@ glusterd_op_stop_volume (dict_t *dict) int flags = 0; char *volname = NULL; glusterd_volinfo_t *volinfo = NULL; - glusterd_brickinfo_t *brickinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); ret = glusterd_op_stop_volume_args_get (dict, &volname, &flags); if (ret) goto out; ret = glusterd_volinfo_find (volname, &volinfo); - - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS, + volname); goto out; - - list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { - ret = glusterd_brick_stop (volinfo, brickinfo); - if (ret) - goto out; } - glusterd_set_volume_status (volinfo, GLUSTERD_STATUS_STOPPED); - - ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); - if (ret) + ret = glusterd_stop_volume (volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to stop %s volume", + volname); goto out; - - ret = glusterd_nodesvcs_handle_graph_change (volinfo); + } out: return ret; } @@ -1358,18 +1818,20 @@ glusterd_op_delete_volume (dict_t *dict) ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); goto out; } ret = glusterd_volinfo_find (volname, &volinfo); - - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS, + volname); goto out; + } ret = glusterd_delete_volume (volinfo); out: - gf_log ("", GF_LOG_DEBUG, "returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret); return ret; } @@ -1411,12 +1873,353 @@ glusterd_op_statedump_volume (dict_t *dict, char **op_errstr) ret = glusterd_brick_statedump (volinfo, brickinfo, options, option_cnt, op_errstr); + /* Let us take the statedump of other bricks instead of + * exiting, if statedump of this brick fails. + */ if (ret) - goto out; + gf_log (THIS->name, GF_LOG_WARNING, "could not " + "take the statedump of the brick %s:%s." + " Proceeding to other bricks", + brickinfo->hostname, brickinfo->path); + } + } + +out: + return ret; +} + +int +glusterd_clearlocks_send_cmd (glusterd_volinfo_t *volinfo, char *cmd, + char *path, char *result, char *errstr, + int err_len, char *mntpt) +{ + int ret = -1; + glusterd_conf_t *priv = NULL; + char abspath[PATH_MAX] = {0, }; + + priv = THIS->private; + + snprintf (abspath, sizeof (abspath), "%s/%s", mntpt, path); + ret = sys_lgetxattr (abspath, cmd, result, PATH_MAX); + if (ret < 0) { + snprintf (errstr, err_len, "clear-locks getxattr command " + "failed. Reason: %s", strerror (errno)); + gf_log (THIS->name, GF_LOG_DEBUG, "%s", errstr); + goto out; + } + + ret = 0; +out: + return ret; +} + +int +glusterd_clearlocks_rmdir_mount (glusterd_volinfo_t *volinfo, char *mntpt) +{ + int ret = -1; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + + ret = rmdir (mntpt); + if (ret) { + gf_log (THIS->name, GF_LOG_DEBUG, "rmdir failed"); + goto out; + } + + ret = 0; +out: + return ret; +} + +void +glusterd_clearlocks_unmount (glusterd_volinfo_t *volinfo, char *mntpt) +{ + glusterd_conf_t *priv = NULL; + runner_t runner = {0,}; + int ret = 0; + + priv = THIS->private; + + /*umount failures are ignored. Using stat we could have avoided + * attempting to unmount a non-existent filesystem. But a failure of + * stat() on mount can be due to network failures.*/ + + runinit (&runner); + runner_add_args (&runner, "/bin/umount", "-f", NULL); + runner_argprintf (&runner, "%s", mntpt); + + synclock_unlock (&priv->big_lock); + ret = runner_run (&runner); + synclock_lock (&priv->big_lock); + if (ret) { + ret = 0; + gf_log ("", GF_LOG_DEBUG, + "umount failed on maintenance client"); + } + + return; +} + +int +glusterd_clearlocks_create_mount (glusterd_volinfo_t *volinfo, char **mntpt) +{ + int ret = -1; + glusterd_conf_t *priv = NULL; + char template[PATH_MAX] = {0,}; + char *tmpl = NULL; + + priv = THIS->private; + + snprintf (template, sizeof (template), "/tmp/%s.XXXXXX", + volinfo->volname); + tmpl = mkdtemp (template); + if (!tmpl) { + gf_log (THIS->name, GF_LOG_DEBUG, "Couldn't create temporary " + "mount directory. Reason %s", strerror (errno)); + goto out; + } + + *mntpt = gf_strdup (tmpl); + ret = 0; +out: + return ret; +} + +int +glusterd_clearlocks_mount (glusterd_volinfo_t *volinfo, char **xl_opts, + char *mntpt) +{ + int ret = -1; + int i = 0; + glusterd_conf_t *priv = NULL; + runner_t runner = {0,}; + char client_volfpath[PATH_MAX] = {0,}; + char self_heal_opts[3][1024] = {"*replicate*.data-self-heal=off", + "*replicate*.metadata-self-heal=off", + "*replicate*.entry-self-heal=off"}; + + priv = THIS->private; + + runinit (&runner); + glusterd_get_trusted_client_filepath (client_volfpath, volinfo, + volinfo->transport_type); + runner_add_args (&runner, SBIN_DIR"/glusterfs", "-f", NULL); + runner_argprintf (&runner, "%s", client_volfpath); + runner_add_arg (&runner, "-l"); + runner_argprintf (&runner, DEFAULT_LOG_FILE_DIRECTORY + "/%s-clearlocks-mnt.log", volinfo->volname); + if (volinfo->memory_accounting) + runner_add_arg (&runner, "--mem-accounting"); + + for (i = 0; i < volinfo->brick_count && xl_opts[i]; i++) { + runner_add_arg (&runner, "--xlator-option"); + runner_argprintf (&runner, "%s", xl_opts[i]); + } + + for (i = 0; i < 3; i++) { + runner_add_args (&runner, "--xlator-option", + self_heal_opts[i], NULL); + } + + runner_argprintf (&runner, "%s", mntpt); + synclock_unlock (&priv->big_lock); + ret = runner_run (&runner); + synclock_lock (&priv->big_lock); + if (ret) { + gf_log (THIS->name, GF_LOG_DEBUG, + "Could not start glusterfs"); + goto out; + } + gf_log (THIS->name, GF_LOG_DEBUG, + "Started glusterfs successfully"); + +out: + return ret; +} + +int +glusterd_clearlocks_get_local_client_ports (glusterd_volinfo_t *volinfo, + char **xl_opts) +{ + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_conf_t *priv = NULL; + int index = 0; + int ret = -1; + int i = 0; + int port = 0; + + GF_ASSERT (xl_opts); + if (!xl_opts) { + gf_log (THIS->name, GF_LOG_DEBUG, "Should pass non-NULL " + "xl_opts"); + goto out; + } + + priv = THIS->private; + + index = -1; + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + index++; + if (uuid_compare (brickinfo->uuid, MY_UUID)) + continue; + + port = pmap_registry_search (THIS, brickinfo->path, + GF_PMAP_PORT_BRICKSERVER); + if (!port) { + ret = -1; + gf_log (THIS->name, GF_LOG_DEBUG, "Couldn't get port " + " for brick %s:%s", brickinfo->hostname, + brickinfo->path); + goto out; + } + + ret = gf_asprintf (&xl_opts[i], "%s-client-%d.remote-port=%d", + volinfo->volname, index, port); + if (ret == -1) { + xl_opts[i] = NULL; + goto out; } + i++; } + ret = 0; out: return ret; } +int +glusterd_op_clearlocks_volume (dict_t *dict, char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + int i = 0; + char *volname = NULL; + char *path = NULL; + char *kind = NULL; + char *type = NULL; + char *opts = NULL; + char *cmd_str = NULL; + char *free_ptr = NULL; + char msg[PATH_MAX] = {0,}; + char result[PATH_MAX] = {0,}; + char *mntpt = NULL; + char **xl_opts = NULL; + glusterd_volinfo_t *volinfo = NULL; + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to get volume name"); + goto out; + } + gf_log ("", GF_LOG_DEBUG, "Performing clearlocks on volume %s", volname); + + ret = dict_get_str (dict, "path", &path); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to get path"); + goto out; + } + + ret = dict_get_str (dict, "kind", &kind); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to get kind"); + goto out; + } + + ret = dict_get_str (dict, "type", &type); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to get type"); + goto out; + } + + ret = dict_get_str (dict, "opts", &opts); + if (ret) + ret = 0; + + gf_log (THIS->name, GF_LOG_INFO, "Received clear-locks request for " + "volume %s with kind %s type %s and options %s", volname, + kind, type, opts); + + if (opts) + ret = gf_asprintf (&cmd_str, GF_XATTR_CLRLK_CMD".t%s.k%s.%s", + type, kind, opts); + else + ret = gf_asprintf (&cmd_str, GF_XATTR_CLRLK_CMD".t%s.k%s", + type, kind); + if (ret == -1) + goto out; + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (msg, sizeof (msg), "Volume %s doesn't exist.", + volname); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + goto out; + } + + xl_opts = GF_CALLOC (volinfo->brick_count+1, sizeof (char*), + gf_gld_mt_charptr); + if (!xl_opts) + goto out; + + ret = glusterd_clearlocks_get_local_client_ports (volinfo, xl_opts); + if (ret) { + snprintf (msg, sizeof (msg), "Couldn't get port numbers of " + "local bricks"); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + goto out; + } + + ret = glusterd_clearlocks_create_mount (volinfo, &mntpt); + if (ret) { + snprintf (msg, sizeof (msg), "Creating mount directory " + "for clear-locks failed."); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + goto out; + } + + ret = glusterd_clearlocks_mount (volinfo, xl_opts, mntpt); + if (ret) { + snprintf (msg, sizeof (msg), "Failed to mount clear-locks " + "maintenance client."); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + goto out; + } + + ret = glusterd_clearlocks_send_cmd (volinfo, cmd_str, path, result, + msg, sizeof (msg), mntpt); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + goto umount; + } + + free_ptr = gf_strdup(result); + if (dict_set_dynstr (rsp_dict, "lk-summary", free_ptr)) { + GF_FREE (free_ptr); + snprintf (msg, sizeof (msg), "Failed to set clear-locks " + "result"); + gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); + } + +umount: + glusterd_clearlocks_unmount (volinfo, mntpt); + + if (glusterd_clearlocks_rmdir_mount (volinfo, mntpt)) + gf_log (THIS->name, GF_LOG_WARNING, "Couldn't unmount " + "clear-locks mount point"); + +out: + if (ret) + *op_errstr = gf_strdup (msg); + + if (xl_opts) { + for (i = 0; i < volinfo->brick_count && xl_opts[i]; i++) + GF_FREE (xl_opts[i]); + GF_FREE (xl_opts); + } + + GF_FREE (cmd_str); + + GF_FREE (mntpt); + + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-set.c b/xlators/mgmt/glusterd/src/glusterd-volume-set.c new file mode 100644 index 000000000..665a8b298 --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-volume-set.c @@ -0,0 +1,1452 @@ +/* + Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "glusterd-volgen.h" +#include "glusterd-utils.h" + +static int +check_dict_key_value (dict_t *dict, char *key, char *value) +{ + glusterd_conf_t *priv = NULL; + int ret = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + if (!dict) { + gf_log (this->name, GF_LOG_ERROR, "Received Empty Dict."); + ret = -1; + goto out; + } + + if (!key) { + gf_log (this->name, GF_LOG_ERROR, "Received Empty Key."); + ret = -1; + goto out; + } + + if (!value) { + gf_log (this->name, GF_LOG_ERROR, "Received Empty Value."); + ret = -1; + goto out; + } + +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + + return ret; +} + +static int +get_volname_volinfo (dict_t *dict, char **volname, glusterd_volinfo_t **volinfo) +{ + glusterd_conf_t *priv = NULL; + int ret = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_str (dict, "volname", volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); + goto out; + } + + ret = glusterd_volinfo_find (*volname, volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to allocate memory"); + goto out; + } + +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + + return ret; +} + +static int +validate_cache_max_min_size (dict_t *dict, char *key, char *value, + char **op_errstr) +{ + char *current_max_value = NULL; + char *current_min_value = NULL; + char errstr[2048] = ""; + char *volname = NULL; + glusterd_conf_t *priv = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = 0; + uint64_t max_value = 0; + uint64_t min_value = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + ret = check_dict_key_value (dict, key, value); + if (ret) + goto out; + + ret = get_volname_volinfo (dict, &volname, &volinfo); + if (ret) + goto out; + + if ((!strcmp (key, "performance.cache-min-file-size")) || + (!strcmp (key, "cache-min-file-size"))) { + glusterd_volinfo_get (volinfo, + "performance.cache-max-file-size", + ¤t_max_value); + if (current_max_value) { + gf_string2bytesize (current_max_value, &max_value); + gf_string2bytesize (value, &min_value); + current_min_value = value; + } + } else if ((!strcmp (key, "performance.cache-max-file-size")) || + (!strcmp (key, "cache-max-file-size"))) { + glusterd_volinfo_get (volinfo, + "performance.cache-min-file-size", + ¤t_min_value); + if (current_min_value) { + gf_string2bytesize (current_min_value, &min_value); + gf_string2bytesize (value, &max_value); + current_max_value = value; + } + } + + if (min_value > max_value) { + snprintf (errstr, sizeof (errstr), + "cache-min-file-size (%s) is greater than " + "cache-max-file-size (%s)", + current_min_value, current_max_value); + gf_log (this->name, GF_LOG_ERROR, "%s", errstr); + *op_errstr = gf_strdup (errstr); + ret = -1; + goto out; + } + +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + + return ret; +} + +static int +validate_quota (dict_t *dict, char *key, char *value, + char **op_errstr) +{ + char errstr[2048] = ""; + char *volname = NULL; + glusterd_conf_t *priv = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + ret = check_dict_key_value (dict, key, value); + if (ret) + goto out; + + ret = get_volname_volinfo (dict, &volname, &volinfo); + if (ret) + goto out; + + ret = glusterd_volinfo_get_boolean (volinfo, VKEY_FEATURES_QUOTA); + if (ret == -1) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get the quota status"); + goto out; + } + + if (ret == _gf_false) { + snprintf (errstr, sizeof (errstr), + "Cannot set %s. Enable quota first.", key); + gf_log (this->name, GF_LOG_ERROR, "%s", errstr); + *op_errstr = gf_strdup (errstr); + ret = -1; + goto out; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + + return ret; +} + +static int +validate_stripe (dict_t *dict, char *key, char *value, char **op_errstr) +{ + char errstr[2048] = ""; + char *volname = NULL; + glusterd_conf_t *priv = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + ret = check_dict_key_value (dict, key, value); + if (ret) + goto out; + + ret = get_volname_volinfo (dict, &volname, &volinfo); + if (ret) + goto out; + + if (volinfo->stripe_count == 1) { + snprintf (errstr, sizeof (errstr), + "Cannot set %s for a non-stripe volume.", key); + gf_log (this->name, GF_LOG_ERROR, "%s", errstr); + *op_errstr = gf_strdup (errstr); + ret = -1; + goto out; + } + +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + + return ret; +} + +static int +validate_subvols_per_directory (dict_t *dict, char *key, char *value, + char **op_errstr) +{ + char errstr[2048] = ""; + char *volname = NULL; + glusterd_conf_t *priv = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = 0; + int subvols = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + ret = check_dict_key_value (dict, key, value); + if (ret) + goto out; + + ret = get_volname_volinfo (dict, &volname, &volinfo); + if (ret) + goto out; + + subvols = atoi(value); + + /* Checking if the subvols-per-directory exceed the total + number of subvolumes. */ + if (subvols > volinfo->subvol_count) { + snprintf (errstr, sizeof(errstr), + "subvols-per-directory(%d) is greater " + "than the number of subvolumes(%d).", + subvols, volinfo->subvol_count); + gf_log (this->name, GF_LOG_ERROR, + "%s.", errstr); + *op_errstr = gf_strdup (errstr); + ret = -1; + goto out; + } + +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + + return ret; +} + + +/* dispatch table for VOLUME SET + * ----------------------------- + * + * Format of entries: + * + * First field is the <key>, for the purpose of looking it up + * in volume dictionary. Each <key> is of the format "<domain>.<specifier>". + * + * Second field is <voltype>. + * + * Third field is <option>, if its unset, it's assumed to be + * the same as <specifier>. + * + * Fourth field is <value>. In this context they are used to specify + * a default. That is, even the volume dict doesn't have a value, + * we procced as if the default value were set for it. + * + * Fifth field is <doctype>, which decides if the option is public and available + * in "set help" or not. "NO_DOC" entries are not part of the public interface + * and are subject to change at any time. This also decides if an option is + * global (apllies to all volumes) or normal (applies to only specified volume). + * + * Sixth field is <flags>. + * + * Seventh field is <op-version>. + * + * Eight field is description of option: If NULL, tried to fetch from + * translator code's xlator_options table. + * + * Nineth field is validation function: If NULL, xlator's option specific + * validation will be tried, otherwise tried at glusterd code itself. + * + * There are two type of entries: basic and special. + * + * - Basic entries are the ones where the <option> does _not_ start with + * the bang! character ('!'). + * + * In their case, <option> is understood as an option for an xlator of + * type <voltype>. Their effect is to copy over the volinfo->dict[<key>] + * value to all graph nodes of type <voltype> (if such a value is set). + * + * You are free to add entries of this type, they will become functional + * just by being present in the table. + * + * - Special entries where the <option> starts with the bang!. + * + * They are not applied to all graphs during generation, and you cannot + * extend them in a trivial way which could be just picked up. Better + * not touch them unless you know what you do. + * + * + * Another kind of grouping for options, according to visibility: + * + * - Exported: one which is used in the code. These are characterized by + * being used a macro as <key> (of the format VKEY_..., defined in + * glusterd-volgen.h + * + * - Non-exported: the rest; these have string literal <keys>. + * + * Adhering to this policy, option name changes shall be one-liners. + * + */ + +struct volopt_map_entry glusterd_volopt_map[] = { + /* DHT xlator options */ + { .key = "cluster.lookup-unhashed", + .voltype = "cluster/distribute", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.min-free-disk", + .voltype = "cluster/distribute", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.min-free-inodes", + .voltype = "cluster/distribute", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.rebalance-stats", + .voltype = "cluster/distribute", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.subvols-per-directory", + .voltype = "cluster/distribute", + .option = "directory-layout-spread", + .op_version = 2, + .validate_fn = validate_subvols_per_directory, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.readdir-optimize", + .voltype = "cluster/distribute", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.rsync-hash-regex", + .voltype = "cluster/distribute", + .type = NO_DOC, + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.extra-hash-regex", + .voltype = "cluster/distribute", + .type = NO_DOC, + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.dht-xattr-name", + .voltype = "cluster/distribute", + .option = "xattr-name", + .type = NO_DOC, + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* NUFA xlator options (Distribute special case) */ + { .key = "cluster.nufa", + .voltype = "cluster/distribute", + .option = "!nufa", + .type = NO_DOC, + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.local-volume-name", + .voltype = "cluster/nufa", + .option = "local-volume-name", + .type = NO_DOC, + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* Switch xlator options (Distribute special case) */ + { .key = "cluster.switch", + .voltype = "cluster/distribute", + .option = "!switch", + .type = NO_DOC, + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.switch-pattern", + .voltype = "cluster/switch", + .option = "pattern.switch.case", + .type = NO_DOC, + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* AFR xlator options */ + { .key = "cluster.entry-change-log", + .voltype = "cluster/replicate", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.read-subvolume", + .voltype = "cluster/replicate", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.read-subvolume-index", + .voltype = "cluster/replicate", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.read-hash-mode", + .voltype = "cluster/replicate", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.background-self-heal-count", + .voltype = "cluster/replicate", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.metadata-self-heal", + .voltype = "cluster/replicate", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.data-self-heal", + .voltype = "cluster/replicate", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.entry-self-heal", + .voltype = "cluster/replicate", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.self-heal-daemon", + .voltype = "cluster/replicate", + .option = "!self-heal-daemon", + .op_version = 1 + }, + { .key = "cluster.heal-timeout", + .voltype = "cluster/replicate", + .option = "!heal-timeout", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.strict-readdir", + .voltype = "cluster/replicate", + .type = NO_DOC, + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.self-heal-window-size", + .voltype = "cluster/replicate", + .option = "data-self-heal-window-size", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.data-change-log", + .voltype = "cluster/replicate", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.metadata-change-log", + .voltype = "cluster/replicate", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.data-self-heal-algorithm", + .voltype = "cluster/replicate", + .option = "data-self-heal-algorithm", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.eager-lock", + .voltype = "cluster/replicate", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.quorum-type", + .voltype = "cluster/replicate", + .option = "quorum-type", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.quorum-count", + .voltype = "cluster/replicate", + .option = "quorum-count", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.choose-local", + .voltype = "cluster/replicate", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.self-heal-readdir-size", + .voltype = "cluster/replicate", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.post-op-delay-secs", + .voltype = "cluster/replicate", + .type = NO_DOC, + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.readdir-failover", + .voltype = "cluster/replicate", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.ensure-durability", + .voltype = "cluster/replicate", + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* Stripe xlator options */ + { .key = "cluster.stripe-block-size", + .voltype = "cluster/stripe", + .option = "block-size", + .op_version = 1, + .validate_fn = validate_stripe, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "cluster.stripe-coalesce", + .voltype = "cluster/stripe", + .option = "coalesce", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* IO-stats xlator options */ + { .key = VKEY_DIAG_LAT_MEASUREMENT, + .voltype = "debug/io-stats", + .option = "latency-measurement", + .value = "off", + .op_version = 1 + }, + { .key = "diagnostics.dump-fd-stats", + .voltype = "debug/io-stats", + .op_version = 1 + }, + { .key = VKEY_DIAG_CNT_FOP_HITS, + .voltype = "debug/io-stats", + .option = "count-fop-hits", + .value = "off", + .type = NO_DOC, + .op_version = 1 + }, + { .key = "diagnostics.brick-log-level", + .voltype = "debug/io-stats", + .option = "!brick-log-level", + .op_version = 1 + }, + { .key = "diagnostics.client-log-level", + .voltype = "debug/io-stats", + .option = "!client-log-level", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "diagnostics.brick-sys-log-level", + .voltype = "debug/io-stats", + .option = "!sys-log-level", + .op_version = 1 + }, + { .key = "diagnostics.client-sys-log-level", + .voltype = "debug/io-stats", + .option = "!sys-log-level", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* IO-cache xlator options */ + { .key = "performance.cache-max-file-size", + .voltype = "performance/io-cache", + .option = "max-file-size", + .op_version = 1, + .validate_fn = validate_cache_max_min_size, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.cache-min-file-size", + .voltype = "performance/io-cache", + .option = "min-file-size", + .op_version = 1, + .validate_fn = validate_cache_max_min_size, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.cache-refresh-timeout", + .voltype = "performance/io-cache", + .option = "cache-timeout", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.cache-priority", + .voltype = "performance/io-cache", + .option = "priority", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.cache-size", + .voltype = "performance/io-cache", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* IO-threads xlator options */ + { .key = "performance.io-thread-count", + .voltype = "performance/io-threads", + .option = "thread-count", + .op_version = 1 + }, + { .key = "performance.high-prio-threads", + .voltype = "performance/io-threads", + .op_version = 1 + }, + { .key = "performance.normal-prio-threads", + .voltype = "performance/io-threads", + .op_version = 1 + }, + { .key = "performance.low-prio-threads", + .voltype = "performance/io-threads", + .op_version = 1 + }, + { .key = "performance.least-prio-threads", + .voltype = "performance/io-threads", + .op_version = 1 + }, + { .key = "performance.enable-least-priority", + .voltype = "performance/io-threads", + .op_version = 1 + }, + { .key = "performance.least-rate-limit", + .voltype = "performance/io-threads", + .op_version = 2 + }, + + /* Other perf xlators' options */ + { .key = "performance.cache-size", + .voltype = "performance/quick-read", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.flush-behind", + .voltype = "performance/write-behind", + .option = "flush-behind", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.write-behind-window-size", + .voltype = "performance/write-behind", + .option = "cache-size", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.strict-o-direct", + .voltype = "performance/write-behind", + .option = "strict-O_DIRECT", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.strict-write-ordering", + .voltype = "performance/write-behind", + .option = "strict-write-ordering", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.lazy-open", + .voltype = "performance/open-behind", + .option = "lazy-open", + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.read-ahead-page-count", + .voltype = "performance/read-ahead", + .option = "page-count", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.md-cache-timeout", + .voltype = "performance/md-cache", + .option = "md-cache-timeout", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* Crypt xlator options */ + + { .key = "features.encryption", + .voltype = "encryption/crypt", + .option = "!feat", + .value = "off", + .op_version = 3, + .description = "enable/disable client-side encryption for " + "the volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + + { .key = "encryption.master-key", + .voltype = "encryption/crypt", + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "encryption.data-key-size", + .voltype = "encryption/crypt", + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "encryption.block-size", + .voltype = "encryption/crypt", + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* Client xlator options */ + { .key = "network.frame-timeout", + .voltype = "protocol/client", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "network.ping-timeout", + .voltype = "protocol/client", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "network.tcp-window-size", + .voltype = "protocol/client", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "features.lock-heal", + .voltype = "protocol/client", + .option = "lk-heal", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "features.grace-timeout", + .voltype = "protocol/client", + .option = "grace-timeout", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "client.ssl", + .voltype = "protocol/client", + .option = "transport.socket.ssl-enabled", + .type = NO_DOC, + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "network.remote-dio", + .voltype = "protocol/client", + .option = "filter-O_DIRECT", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* Server xlator options */ + { .key = "network.tcp-window-size", + .voltype = "protocol/server", + .op_version = 1 + }, + { .key = "network.inode-lru-limit", + .voltype = "protocol/server", + .op_version = 1 + }, + { .key = AUTH_ALLOW_MAP_KEY, + .voltype = "protocol/server", + .option = "!server-auth", + .value = "*", + .op_version = 1 + }, + { .key = AUTH_REJECT_MAP_KEY, + .voltype = "protocol/server", + .option = "!server-auth", + .op_version = 1 + }, + { .key = "transport.keepalive", + .voltype = "protocol/server", + .option = "transport.socket.keepalive", + .type = NO_DOC, + .op_version = 1 + }, + { .key = "server.allow-insecure", + .voltype = "protocol/server", + .option = "rpc-auth-allow-insecure", + .type = NO_DOC, + .op_version = 1 + }, + { .key = "server.root-squash", + .voltype = "protocol/server", + .option = "root-squash", + .op_version = 2 + }, + { .key = "server.statedump-path", + .voltype = "protocol/server", + .option = "statedump-path", + .op_version = 1 + }, + { .key = "server.outstanding-rpc-limit", + .voltype = "protocol/server", + .option = "rpc.outstanding-rpc-limit", + .type = GLOBAL_DOC, + .op_version = 3 + }, + { .key = "features.lock-heal", + .voltype = "protocol/server", + .option = "lk-heal", + .type = NO_DOC, + .op_version = 1 + }, + { .key = "features.grace-timeout", + .voltype = "protocol/server", + .option = "grace-timeout", + .type = NO_DOC, + .op_version = 1 + }, + { .key = "server.ssl", + .voltype = "protocol/server", + .option = "transport.socket.ssl-enabled", + .type = NO_DOC, + .op_version = 2 + }, + + /* Performance xlators enable/disbable options */ + { .key = "performance.write-behind", + .voltype = "performance/write-behind", + .option = "!perf", + .value = "on", + .op_version = 1, + .description = "enable/disable write-behind translator in the " + "volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.read-ahead", + .voltype = "performance/read-ahead", + .option = "!perf", + .value = "on", + .op_version = 1, + .description = "enable/disable read-ahead translator in the volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.readdir-ahead", + .voltype = "performance/readdir-ahead", + .option = "!perf", + .value = "off", + .op_version = 3, + .description = "enable/disable readdir-ahead translator in the volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + + { .key = "performance.io-cache", + .voltype = "performance/io-cache", + .option = "!perf", + .value = "on", + .op_version = 1, + .description = "enable/disable io-cache translator in the volume.", + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "performance.quick-read", + .voltype = "performance/quick-read", + .option = "!perf", + .value = "on", + .op_version = 1, + .description = "enable/disable quick-read translator in the volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + + }, + { .key = "performance.open-behind", + .voltype = "performance/open-behind", + .option = "!perf", + .value = "on", + .op_version = 2, + .description = "enable/disable open-behind translator in the volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + + }, + { .key = "performance.stat-prefetch", + .voltype = "performance/md-cache", + .option = "!perf", + .value = "on", + .op_version = 1, + .description = "enable/disable meta-data caching translator in the " + "volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.client-io-threads", + .voltype = "performance/io-threads", + .option = "!perf", + .value = "off", + .op_version = 1, + .description = "enable/disable io-threads translator in the client " + "graph of volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.nfs.write-behind", + .voltype = "performance/write-behind", + .option = "!nfsperf", + .value = "on", + .type = NO_DOC, + .op_version = 1, + .flags = OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.nfs.read-ahead", + .voltype = "performance/read-ahead", + .option = "!nfsperf", + .value = "off", + .type = NO_DOC, + .op_version = 1, + .flags = OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.nfs.io-cache", + .voltype = "performance/io-cache", + .option = "!nfsperf", + .value = "off", + .type = NO_DOC, + .op_version = 1, + .flags = OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.nfs.quick-read", + .voltype = "performance/quick-read", + .option = "!nfsperf", + .value = "off", + .type = NO_DOC, + .op_version = 1, + .flags = OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.nfs.stat-prefetch", + .voltype = "performance/md-cache", + .option = "!nfsperf", + .value = "off", + .type = NO_DOC, + .op_version = 1, + .flags = OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.nfs.io-threads", + .voltype = "performance/io-threads", + .option = "!nfsperf", + .value = "off", + .type = NO_DOC, + .op_version = 1, + .flags = OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.force-readdirp", + .voltype = "performance/md-cache", + .option = "force-readdirp", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* Feature translators */ + { .key = "features.file-snapshot", + .voltype = "features/qemu-block", + .option = "!feat", + .value = "off", + .op_version = 3, + .description = "enable/disable file-snapshot feature in the " + "volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + +#ifdef HAVE_LIB_Z + /* Compressor-decompressor xlator options + * defaults used from xlator/feature/compress/src/cdc.h + */ + { .key = "features.compress", + .voltype = "features/cdc", + .option = "!compress", + .value = "off", + .type = NO_DOC, + .op_version = 2, + .description = "enable/disable compression translator" + }, + { .key = "compress.mode", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "compress.window-size", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "compress.mem-level", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "compress.min-size", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "compress.compression-level", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "compress.debug", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + #endif + + /* Quota xlator options */ + { .key = VKEY_FEATURES_LIMIT_USAGE, + .voltype = "features/quota", + .option = "limit-set", + .type = NO_DOC, + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "features.quota-timeout", + .voltype = "features/quota", + .option = "timeout", + .value = "0", + .op_version = 1, + .validate_fn = validate_quota, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "features.quota-deem-statfs", + .voltype = "features/quota", + .option = "deem-statfs", + .value = "off", + .type = DOC, + .op_version = 3, + .validate_fn = validate_quota, + .flags = OPT_FLAG_CLIENT_OPT + }, + + /* Marker xlator options */ + { .key = VKEY_MARKER_XTIME, + .voltype = "features/marker", + .option = "xtime", + .value = "off", + .type = NO_DOC, + .flags = OPT_FLAG_FORCE, + .op_version = 1 + }, + { .key = VKEY_MARKER_XTIME, + .voltype = "features/marker", + .option = "!xtime", + .value = "off", + .type = NO_DOC, + .flags = OPT_FLAG_FORCE, + .op_version = 1 + }, + { .key = VKEY_MARKER_XTIME_FORCE, + .voltype = "features/marker", + .option = "gsync-force-xtime", + .value = "off", + .type = NO_DOC, + .flags = OPT_FLAG_FORCE, + .op_version = 2 + }, + { .key = VKEY_MARKER_XTIME_FORCE, + .voltype = "features/marker", + .option = "!gsync-force-xtime", + .value = "off", + .type = NO_DOC, + .flags = OPT_FLAG_FORCE, + .op_version = 2 + }, + { .key = VKEY_FEATURES_QUOTA, + .voltype = "features/marker", + .option = "quota", + .value = "off", + .type = NO_DOC, + .flags = OPT_FLAG_FORCE, + .op_version = 1 + }, + + /* Debug xlators options */ + { .key = "debug.trace", + .voltype = "debug/trace", + .option = "!debug", + .value = "off", + .type = NO_DOC, + .op_version = 1, + .flags = OPT_FLAG_XLATOR_OPT + }, + { .key = "debug.log-history", + .voltype = "debug/trace", + .option = "log-history", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "debug.log-file", + .voltype = "debug/trace", + .option = "log-file", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "debug.exclude-ops", + .voltype = "debug/trace", + .option = "exclude-ops", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "debug.include-ops", + .voltype = "debug/trace", + .option = "include-ops", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "debug.error-gen", + .voltype = "debug/error-gen", + .option = "!debug", + .value = "off", + .type = NO_DOC, + .op_version = 1, + .flags = OPT_FLAG_XLATOR_OPT + }, + { .key = "debug.error-failure", + .voltype = "debug/error-gen", + .option = "failure", + .type = NO_DOC, + .op_version = 3 + }, + { .key = "debug.error-number", + .voltype = "debug/error-gen", + .option = "error-no", + .type = NO_DOC, + .op_version = 3 + }, + { .key = "debug.random-failure", + .voltype = "debug/error-gen", + .option = "random-failure", + .type = NO_DOC, + .op_version = 3 + }, + { .key = "debug.error-fops", + .voltype = "debug/error-gen", + .option = "enable", + .type = NO_DOC, + .op_version = 3 + }, + + + /* NFS xlator options */ + { .key = "nfs.enable-ino32", + .voltype = "nfs/server", + .option = "nfs.enable-ino32", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.mem-factor", + .voltype = "nfs/server", + .option = "nfs.mem-factor", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.export-dirs", + .voltype = "nfs/server", + .option = "nfs3.export-dirs", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.export-volumes", + .voltype = "nfs/server", + .option = "nfs3.export-volumes", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.addr-namelookup", + .voltype = "nfs/server", + .option = "rpc-auth.addr.namelookup", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.dynamic-volumes", + .voltype = "nfs/server", + .option = "nfs.dynamic-volumes", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.register-with-portmap", + .voltype = "nfs/server", + .option = "rpc.register-with-portmap", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.outstanding-rpc-limit", + .voltype = "nfs/server", + .option = "rpc.outstanding-rpc-limit", + .type = GLOBAL_DOC, + .op_version = 3 + }, + { .key = "nfs.port", + .voltype = "nfs/server", + .option = "nfs.port", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.rpc-auth-unix", + .voltype = "nfs/server", + .option = "!rpc-auth.auth-unix.*", + .op_version = 1 + }, + { .key = "nfs.rpc-auth-null", + .voltype = "nfs/server", + .option = "!rpc-auth.auth-null.*", + .op_version = 1 + }, + { .key = "nfs.rpc-auth-allow", + .voltype = "nfs/server", + .option = "!rpc-auth.addr.*.allow", + .op_version = 1 + }, + { .key = "nfs.rpc-auth-reject", + .voltype = "nfs/server", + .option = "!rpc-auth.addr.*.reject", + .op_version = 1 + }, + { .key = "nfs.ports-insecure", + .voltype = "nfs/server", + .option = "!rpc-auth.ports.*.insecure", + .op_version = 1 + }, + { .key = "nfs.transport-type", + .voltype = "nfs/server", + .option = "!nfs.transport-type", + .value = "tcp", + .op_version = 1, + .description = "Specifies the nfs transport type. Valid " + "transport types are 'tcp' and 'rdma'." + }, + { .key = "nfs.trusted-sync", + .voltype = "nfs/server", + .option = "!nfs3.*.trusted-sync", + .op_version = 1 + }, + { .key = "nfs.trusted-write", + .voltype = "nfs/server", + .option = "!nfs3.*.trusted-write", + .op_version = 1 + }, + { .key = "nfs.volume-access", + .voltype = "nfs/server", + .option = "!nfs3.*.volume-access", + .op_version = 1 + }, + { .key = "nfs.export-dir", + .voltype = "nfs/server", + .option = "!nfs3.*.export-dir", + .op_version = 1 + }, + { .key = NFS_DISABLE_MAP_KEY, + .voltype = "nfs/server", + .option = "!nfs-disable", + .op_version = 1 + }, + { .key = "nfs.nlm", + .voltype = "nfs/server", + .option = "nfs.nlm", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.acl", + .voltype = "nfs/server", + .option = "nfs.acl", + .type = GLOBAL_DOC, + .op_version = 3 + }, + { .key = "nfs.mount-udp", + .voltype = "nfs/server", + .option = "nfs.mount-udp", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.mount-rmtab", + .voltype = "nfs/server", + .option = "nfs.mount-rmtab", + .type = GLOBAL_DOC, + .op_version = 1 + }, + { .key = "nfs.server-aux-gids", + .voltype = "nfs/server", + .option = "nfs.server-aux-gids", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "nfs.drc", + .voltype = "nfs/server", + .option = "nfs.drc", + .type = GLOBAL_DOC, + .op_version = 3 + }, + { .key = "nfs.drc-size", + .voltype = "nfs/server", + .option = "nfs.drc-size", + .type = GLOBAL_DOC, + .op_version = 3 + }, + { .key = "nfs.read-size", + .voltype = "nfs/server", + .option = "nfs3.read-size", + .type = GLOBAL_DOC, + .op_version = 3 + }, + { .key = "nfs.write-size", + .voltype = "nfs/server", + .option = "nfs3.write-size", + .type = GLOBAL_DOC, + .op_version = 3 + }, + { .key = "nfs.readdir-size", + .voltype = "nfs/server", + .option = "nfs3.readdir-size", + .type = GLOBAL_DOC, + .op_version = 3 + }, + + /* Other options which don't fit any place above */ + { .key = "features.read-only", + .voltype = "features/read-only", + .option = "!read-only", + .value = "off", + .op_version = 1, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + { .key = "features.worm", + .voltype = "features/worm", + .option = "!worm", + .value = "off", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + { .key = "storage.linux-aio", + .voltype = "storage/posix", + .op_version = 1 + }, + { .key = "storage.batch-fsync-mode", + .voltype = "storage/posix", + .op_version = 3 + }, + { .key = "storage.batch-fsync-delay-usec", + .voltype = "storage/posix", + .op_version = 3 + }, + { .key = "storage.owner-uid", + .voltype = "storage/posix", + .option = "brick-uid", + .op_version = 1 + }, + { .key = "storage.owner-gid", + .voltype = "storage/posix", + .option = "brick-gid", + .op_version = 1 + }, + { .key = "storage.node-uuid-pathinfo", + .voltype = "storage/posix", + .op_version = 3 + }, + { .key = "storage.health-check-interval", + .voltype = "storage/posix", + .op_version = 3 + }, + { .key = "storage.bd-aio", + .voltype = "storage/bd", + .op_version = 3 + }, + { .key = "config.memory-accounting", + .voltype = "configuration", + .option = "!config", + .op_version = 2, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "config.transport", + .voltype = "configuration", + .option = "!config", + .op_version = 2 + }, + { .key = GLUSTERD_QUORUM_TYPE_KEY, + .voltype = "mgmt/glusterd", + .value = "off", + .op_version = 2 + }, + { .key = GLUSTERD_QUORUM_RATIO_KEY, + .voltype = "mgmt/glusterd", + .value = "0", + .op_version = 2 + }, + /* changelog translator - global tunables */ + { .key = "changelog.changelog", + .voltype = "features/changelog", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "changelog.changelog-dir", + .voltype = "features/changelog", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "changelog.encoding", + .voltype = "features/changelog", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "changelog.rollover-time", + .voltype = "features/changelog", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "changelog.fsync-interval", + .voltype = "features/changelog", + .type = NO_DOC, + .op_version = 2 + }, + { .key = NULL + } +}; diff --git a/xlators/mgmt/glusterd/src/glusterd.c b/xlators/mgmt/glusterd/src/glusterd.c index af6381bbf..59288ada0 100644 --- a/xlators/mgmt/glusterd/src/glusterd.c +++ b/xlators/mgmt/glusterd/src/glusterd.c @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2013 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _CONFIG_H #define _CONFIG_H @@ -44,20 +34,29 @@ #include "glusterd-sm.h" #include "glusterd-op-sm.h" #include "glusterd-store.h" +#include "glusterd-hooks.h" #include "glusterd-utils.h" +#include "glusterd-locks.h" #include "common-utils.h" #include "run.h" +#include "syncop.h" + #include "glusterd-mountbroker.h" -static uuid_t glusterd_uuid; extern struct rpcsvc_program gluster_handshake_prog; +extern struct rpcsvc_program gluster_cli_getspec_prog; extern struct rpcsvc_program gluster_pmap_prog; extern glusterd_op_info_t opinfo; extern struct rpcsvc_program gd_svc_mgmt_prog; +extern struct rpcsvc_program gd_svc_mgmt_v3_prog; extern struct rpcsvc_program gd_svc_peer_prog; extern struct rpcsvc_program gd_svc_cli_prog; +extern struct rpcsvc_program gd_svc_cli_prog_ro; extern struct rpc_clnt_program gd_brick_prog; +extern struct rpcsvc_program glusterd_mgmt_hndsk_prog; + +extern char snap_mount_folder[PATH_MAX]; rpcsvc_cbk_program_t glusterd_cbk_prog = { .progname = "Gluster Callback", @@ -65,6 +64,57 @@ rpcsvc_cbk_program_t glusterd_cbk_prog = { .progver = GLUSTER_CBK_VERSION, }; +struct rpcsvc_program *gd_inet_programs[] = { + &gd_svc_peer_prog, + &gd_svc_cli_prog_ro, + &gd_svc_mgmt_prog, + &gd_svc_mgmt_v3_prog, + &gluster_pmap_prog, + &gluster_handshake_prog, + &glusterd_mgmt_hndsk_prog, +}; +int gd_inet_programs_count = (sizeof (gd_inet_programs) / + sizeof (gd_inet_programs[0])); + +struct rpcsvc_program *gd_uds_programs[] = { + &gd_svc_cli_prog, + &gluster_cli_getspec_prog, +}; +int gd_uds_programs_count = (sizeof (gd_uds_programs) / + sizeof (gd_uds_programs[0])); + +const char *gd_op_list[GD_OP_MAX + 1] = { + [GD_OP_NONE] = "Invalid op", + [GD_OP_CREATE_VOLUME] = "Create", + [GD_OP_START_BRICK] = "Start Brick", + [GD_OP_STOP_BRICK] = "Stop Brick", + [GD_OP_DELETE_VOLUME] = "Delete", + [GD_OP_START_VOLUME] = "Start", + [GD_OP_STOP_VOLUME] = "Stop", + [GD_OP_DEFRAG_VOLUME] = "Rebalance", + [GD_OP_ADD_BRICK] = "Add brick", + [GD_OP_REMOVE_BRICK] = "Remove brick", + [GD_OP_REPLACE_BRICK] = "Replace brick", + [GD_OP_SET_VOLUME] = "Set", + [GD_OP_RESET_VOLUME] = "Reset", + [GD_OP_SYNC_VOLUME] = "Sync", + [GD_OP_LOG_ROTATE] = "Log rotate", + [GD_OP_GSYNC_SET] = "Geo-replication", + [GD_OP_PROFILE_VOLUME] = "Profile", + [GD_OP_QUOTA] = "Quota", + [GD_OP_STATUS_VOLUME] = "Status", + [GD_OP_REBALANCE] = "Rebalance", + [GD_OP_HEAL_VOLUME] = "Heal", + [GD_OP_STATEDUMP_VOLUME] = "Statedump", + [GD_OP_LIST_VOLUME] = "Lists", + [GD_OP_CLEARLOCKS_VOLUME] = "Clear locks", + [GD_OP_DEFRAG_BRICK_VOLUME] = "Rebalance", + [GD_OP_COPY_FILE] = "Copy File", + [GD_OP_SYS_EXEC] = "Execute system commands", + [GD_OP_GSYNC_CREATE] = "Geo-replication Create", + [GD_OP_SNAP] = "Snapshot", + [GD_OP_MAX] = "Invalid op" +}; static int glusterd_opinfo_init () @@ -76,35 +126,30 @@ glusterd_opinfo_init () return ret; } -static int -glusterd_uuid_init (int flag) + +int +glusterd_uuid_init () { int ret = -1; + xlator_t *this = NULL; glusterd_conf_t *priv = NULL; - priv = THIS->private; + this = THIS; + GF_ASSERT (this); + priv = this->private; - if (!flag) { - ret = glusterd_retrieve_uuid (); - if (!ret) { - uuid_copy (glusterd_uuid, priv->uuid); - gf_log ("glusterd", GF_LOG_INFO, - "retrieved UUID: %s", uuid_utoa (priv->uuid)); - return 0; - } + ret = glusterd_retrieve_uuid (); + if (ret == 0) { + gf_log (this->name, GF_LOG_INFO, + "retrieved UUID: %s", uuid_utoa (priv->uuid)); + return 0; } - uuid_generate (glusterd_uuid); - - gf_log ("glusterd", GF_LOG_INFO, - "generated UUID: %s", uuid_utoa (glusterd_uuid)); - uuid_copy (priv->uuid, glusterd_uuid); - - ret = glusterd_store_uuid (); + ret = glusterd_uuid_generate_save (); if (ret) { gf_log ("glusterd", GF_LOG_ERROR, - "Unable to store generated UUID"); + "Unable to generate and save new UUID"); return ret; } @@ -112,6 +157,63 @@ glusterd_uuid_init (int flag) } int +glusterd_uuid_generate_save () +{ + int ret = -1; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + uuid_generate (priv->uuid); + + gf_log (this->name, GF_LOG_INFO, "generated UUID: %s", + uuid_utoa (priv->uuid)); + + ret = glusterd_store_global_info (this); + + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to store the generated uuid %s", + uuid_utoa (priv->uuid)); + + return ret; +} + +int +glusterd_options_init (xlator_t *this) +{ + int ret = -1; + glusterd_conf_t *priv = NULL; + char *initial_version = "0"; + + priv = this->private; + + priv->opts = dict_new (); + if (!priv->opts) + goto out; + + ret = glusterd_store_retrieve_options (this); + if (ret == 0) + goto out; + + ret = dict_set_str (priv->opts, GLUSTERD_GLOBAL_OPT_VERSION, + initial_version); + if (ret) + goto out; + ret = glusterd_store_options (this, priv->opts); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to store version"); + return ret; + } +out: + + return 0; +} +int glusterd_fetchspec_notify (xlator_t *this) { int ret = -1; @@ -120,10 +222,15 @@ glusterd_fetchspec_notify (xlator_t *this) priv = this->private; - list_for_each_entry (trans, &priv->xprt_list, list) { - rpcsvc_callback_submit (priv->rpc, trans, &glusterd_cbk_prog, - GF_CBK_FETCHSPEC, NULL, 0); + pthread_mutex_lock (&priv->xprt_lock); + { + list_for_each_entry (trans, &priv->xprt_list, list) { + rpcsvc_callback_submit (priv->rpc, trans, + &glusterd_cbk_prog, + GF_CBK_FETCHSPEC, NULL, 0); + } } + pthread_mutex_unlock (&priv->xprt_lock); ret = 0; @@ -181,12 +288,16 @@ glusterd_rpcsvc_notify (rpcsvc_t *rpc, void *xl, rpcsvc_event_t event, { INIT_LIST_HEAD (&xprt->list); + pthread_mutex_lock (&priv->xprt_lock); list_add_tail (&xprt->list, &priv->xprt_list); + pthread_mutex_unlock (&priv->xprt_lock); break; } case RPCSVC_EVENT_DISCONNECT: { + pthread_mutex_lock (&priv->xprt_lock); list_del (&xprt->list); + pthread_mutex_unlock (&priv->xprt_lock); pmap_registry_remove (this, 0, NULL, GF_PMAP_PORT_NONE, xprt); break; } @@ -243,18 +354,6 @@ out: return ret; } -/* defined in usterd-utils.c -- no - * glusterd header where it would be - * appropriate to put to, and too - * accidental routine to place in - * libglusterfs. - * - * (Indeed, XXX: we'd rather need a general - * "mkdir -p" like routine in - * libglusterfs) - */ -extern int mkdir_if_missing (char *path); - #if SYNCDAEMON_COMPILE static int glusterd_check_gsync_present (int *valid_state) @@ -347,13 +446,13 @@ glusterd_crt_georep_folders (char *georepdir, glusterd_conf_t *conf) if (strlen (conf->workdir)+2 > PATH_MAX-strlen(GEOREP)) { ret = -1; gf_log ("glusterd", GF_LOG_CRITICAL, - "Unable to create "GEOREP" directory %s", - georepdir); + "directory path %s/"GEOREP" is longer than PATH_MAX", + conf->workdir); goto out; } snprintf (georepdir, PATH_MAX, "%s/"GEOREP, conf->workdir); - ret = mkdir_if_missing (georepdir); + ret = mkdir_p (georepdir, 0777, _gf_true); if (-1 == ret) { gf_log ("glusterd", GF_LOG_CRITICAL, "Unable to create "GEOREP" directory %s", @@ -364,31 +463,49 @@ glusterd_crt_georep_folders (char *georepdir, glusterd_conf_t *conf) if (strlen (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP) >= PATH_MAX) { ret = -1; gf_log ("glusterd", GF_LOG_CRITICAL, - "Unable to create "GEOREP" directory %s", - georepdir); + "directory path "DEFAULT_LOG_FILE_DIRECTORY"/" + GEOREP" is longer than PATH_MAX"); goto out; } - ret = mkdir_if_missing (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP); + ret = mkdir_p (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP, 0777, _gf_true); if (-1 == ret) { gf_log ("glusterd", GF_LOG_CRITICAL, "Unable to create "GEOREP" log directory"); goto out; } + /* Slave log file directory */ if (strlen(DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves") >= PATH_MAX) { ret = -1; gf_log ("glusterd", GF_LOG_CRITICAL, - "Unable to create "GEOREP" directory %s", - georepdir); + "directory path "DEFAULT_LOG_FILE_DIRECTORY"/" + GEOREP"-slaves"" is longer than PATH_MAX"); goto out; } - ret = mkdir_if_missing (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves"); + ret = mkdir_p (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves", 0777, + _gf_true); if (-1 == ret) { gf_log ("glusterd", GF_LOG_CRITICAL, "Unable to create "GEOREP" slave log directory"); goto out; } + /* MountBroker log file directory */ + if (strlen(DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/mbr") >= PATH_MAX) { + ret = -1; + gf_log ("glusterd", GF_LOG_CRITICAL, + "directory path "DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP + "-slaves/mbr"" is longer than PATH_MAX"); + goto out; + } + ret = mkdir_p (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/mbr", 0777, + _gf_true); + if (-1 == ret) { + gf_log ("glusterd", GF_LOG_CRITICAL, + "Unable to create "GEOREP" mountbroker slave log directory"); + goto out; + } + ret = dict_get_str (THIS->options, GEOREP"-log-group", &greplg_s); if (ret) ret = 0; @@ -406,20 +523,22 @@ glusterd_crt_georep_folders (char *georepdir, glusterd_conf_t *conf) if (ret == 0) ret = group_write_allow (DEFAULT_LOG_FILE_DIRECTORY"/" GEOREP"-slaves", gr->gr_gid); + if (ret == 0) + ret = group_write_allow (DEFAULT_LOG_FILE_DIRECTORY"/" + GEOREP"-slaves/mbr", gr->gr_gid); } out: gf_log("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } -#endif static void runinit_gsyncd_setrx (runner_t *runner, glusterd_conf_t *conf) { runinit (runner); runner_add_args (runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL); - runner_argprintf (runner, "%s/"GSYNC_CONF,conf->workdir); + runner_argprintf (runner, "%s/"GSYNC_CONF_TEMPLATE, conf->workdir); runner_add_arg (runner, "--config-set-rx"); } @@ -436,7 +555,6 @@ configure_syncdaemon (glusterd_conf_t *conf) } while (0) { int ret = 0; -#if SYNCDAEMON_COMPILE runner_t runner = {0,}; char georepdir[PATH_MAX] = {0,}; int valid_state = 0; @@ -469,8 +587,8 @@ configure_syncdaemon (glusterd_conf_t *conf) RUN_GSYNCD_CMD; runinit_gsyncd_setrx (&runner, conf); - runner_add_args (&runner, "remote-gsyncd", - "/usr/local/libexec/glusterfs/gsyncd", ".", "^ssh:", NULL); + runner_add_args (&runner, "remote-gsyncd", "/nonexistent/gsyncd", + ".", "^ssh:", NULL); RUN_GSYNCD_CMD; /* gluster-command-dir */ @@ -482,7 +600,7 @@ configure_syncdaemon (glusterd_conf_t *conf) /* gluster-params */ runinit_gsyncd_setrx (&runner, conf); runner_add_args (&runner, "gluster-params", - "xlator-option=*-dht.assert-no-child-down=true", + "aux-gfid-mount xlator-option=*-dht.assert-no-child-down=true", ".", ".", NULL); RUN_GSYNCD_CMD; @@ -499,17 +617,45 @@ configure_syncdaemon (glusterd_conf_t *conf) /* pid-file */ runinit_gsyncd_setrx (&runner, conf); runner_add_arg (&runner, "pid-file"); - runner_argprintf (&runner, "%s/${mastervol}/${eSlave}.pid", georepdir); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.pid", georepdir); runner_add_args (&runner, ".", ".", NULL); RUN_GSYNCD_CMD; /* state-file */ runinit_gsyncd_setrx (&runner, conf); runner_add_arg (&runner, "state-file"); - runner_argprintf (&runner, "%s/${mastervol}/${eSlave}.status", georepdir); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.status", georepdir); runner_add_args (&runner, ".", ".", NULL); RUN_GSYNCD_CMD; + /* state-detail-file */ + runinit_gsyncd_setrx (&runner, conf); + runner_add_arg (&runner, "state-detail-file"); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}-detail.status", + georepdir); + runner_add_args (&runner, ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* state-detail-file */ + runinit_gsyncd_setrx (&runner, conf); + runner_add_arg (&runner, "state-detail-file"); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}-detail.status", + georepdir); + runner_add_args (&runner, ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* state-socket */ + runinit_gsyncd_setrx (&runner, conf); + runner_add_arg (&runner, "state-socket-unencoded"); + runner_argprintf (&runner, "%s/${mastervol}/${eSlave}.socket", georepdir); + runner_add_args (&runner, ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* socketdir */ + runinit_gsyncd_setrx (&runner, conf); + runner_add_args (&runner, "socketdir", GLUSTERD_SOCK_DIR, ".", ".", NULL); + RUN_GSYNCD_CMD; + /* log-file */ runinit_gsyncd_setrx (&runner, conf); runner_add_args (&runner, @@ -522,10 +668,32 @@ configure_syncdaemon (glusterd_conf_t *conf) runinit_gsyncd_setrx (&runner, conf); runner_add_args (&runner, "gluster-log-file", - DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/${mastervol}/${eSlave}.gluster.log", + DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/${mastervol}/${eSlave}${local_id}.gluster.log", ".", ".", NULL); RUN_GSYNCD_CMD; + /* ignore-deletes */ + runinit_gsyncd_setrx (&runner, conf); + runner_add_args (&runner, "ignore-deletes", "true", ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* special-sync-mode */ + runinit_gsyncd_setrx (&runner, conf); + runner_add_args (&runner, "special-sync-mode", "partial", ".", ".", NULL); + RUN_GSYNCD_CMD; + + /* change-detector == changelog */ + runinit_gsyncd_setrx (&runner, conf); + runner_add_args(&runner, "change-detector", "changelog", ".", ".", NULL); + RUN_GSYNCD_CMD; + + runinit_gsyncd_setrx (&runner, conf); + runner_add_arg(&runner, "working-dir"); + runner_argprintf(&runner, "%s/${mastervol}/${eSlave}", + DEFAULT_VAR_RUN_DIRECTORY); + runner_add_args (&runner, ".", ".", NULL); + RUN_GSYNCD_CMD; + /************ * slave pre-configuration ************/ @@ -539,7 +707,7 @@ configure_syncdaemon (glusterd_conf_t *conf) /* gluster-params */ runinit_gsyncd_setrx (&runner, conf); runner_add_args (&runner, "gluster-params", - "xlator-option=*-dht.assert-no-child-down=true", + "aux-gfid-mount xlator-option=*-dht.assert-no-child-down=true", ".", NULL); RUN_GSYNCD_CMD; @@ -551,6 +719,14 @@ configure_syncdaemon (glusterd_conf_t *conf) ".", NULL); RUN_GSYNCD_CMD; + /* MountBroker log-file */ + runinit_gsyncd_setrx (&runner, conf); + runner_add_args (&runner, + "log-file-mbr", + DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/mbr/${session_owner}:${eSlave}.log", + ".", NULL); + RUN_GSYNCD_CMD; + /* gluster-log-file */ runinit_gsyncd_setrx (&runner, conf); runner_add_args (&runner, @@ -560,12 +736,17 @@ configure_syncdaemon (glusterd_conf_t *conf) RUN_GSYNCD_CMD; out: -#else - (void)conf; -#endif return ret ? -1 : 0; } #undef RUN_GSYNCD_CMD +#else /* SYNCDAEMON_COMPILE */ +static int +configure_syncdaemon (glusterd_conf_t *conf) +{ + return 0; +} +#endif /* !SYNCDAEMON_COMPILE */ + static int check_prepare_mountbroker_root (char *mountbroker_root) @@ -597,6 +778,11 @@ check_prepare_mountbroker_root (char *mountbroker_root) ret = -1; goto out; } + if (!(st.st_mode & (S_IXGRP|S_IXOTH))) { + gf_log ("", GF_LOG_WARNING, + "permissions on mountbroker-root directory %s are " + "probably too strict", mountbroker_root); + } dfd0 = dup (dfd); @@ -625,6 +811,11 @@ check_prepare_mountbroker_root (char *mountbroker_root) ret = -1; goto out; } + if (!(st.st_mode & (S_IXGRP|S_IXOTH))) { + gf_log ("", GF_LOG_WARNING, + "permissions on ancestors of mountbroker-root " + "directory are probably too strict"); + } close (dfd); dfd = dfd2; @@ -647,14 +838,17 @@ check_prepare_mountbroker_root (char *mountbroker_root) ret = 0; out: - close (dfd0); - close (dfd); - close (dfd2); + if (dfd0 != -1) + close (dfd0); + if (dfd != -1) + close (dfd); + if (dfd2 != -1) + close (dfd2); return ret; } -static void +static int _install_mount_spec (dict_t *opts, char *key, data_t *value, void *data) { glusterd_conf_t *priv = THIS->private; @@ -663,15 +857,11 @@ _install_mount_spec (dict_t *opts, char *key, data_t *value, void *data) gf_boolean_t ghadoop = _gf_false; char *pdesc = value->data; char *volname = NULL; - int *ret = data; int rv = 0; gf_mount_spec_t *mspec = NULL; char *user = NULL; char *volfile_server = NULL; - if (*ret == -1) - return; - label = strtail (key, "mountbroker."); /* check for presence of geo-rep/hadoop label */ @@ -679,14 +869,15 @@ _install_mount_spec (dict_t *opts, char *key, data_t *value, void *data) label = strtail (key, "mountbroker-"GEOREP"."); if (label) georep = _gf_true; - - label = strtail (key, "mountbroker-"GHADOOP"."); - if (label) - ghadoop = _gf_true; + else { + label = strtail (key, "mountbroker-"GHADOOP"."); + if (label) + ghadoop = _gf_true; + } } if (!label) - return; + return 0; mspec = GF_CALLOC (1, sizeof (*mspec), gf_gld_mt_mount_spec); if (!mspec) @@ -725,14 +916,222 @@ _install_mount_spec (dict_t *opts, char *key, data_t *value, void *data) list_add_tail (&mspec->speclist, &priv->mount_specs); - return; + return 0; err: gf_log ("", GF_LOG_ERROR, "adding %smount spec failed: label: %s desc: %s", georep ? GEOREP" " : "", label, pdesc); - *ret = -1; + return -1; +} + + +static int +gd_default_synctask_cbk (int ret, call_frame_t *frame, void *opaque) +{ + glusterd_conf_t *priv = THIS->private; + synclock_unlock (&priv->big_lock); + return ret; +} + +static void +glusterd_launch_synctask (synctask_fn_t fn, void *opaque) +{ + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int ret = -1; + + this = THIS; + priv = this->private; + + synclock_lock (&priv->big_lock); + ret = synctask_new (this->ctx->env, fn, gd_default_synctask_cbk, NULL, + opaque); + if (ret) + gf_log (this->name, GF_LOG_CRITICAL, "Failed to spawn bricks" + " and other volume related services"); +} + +int +glusterd_uds_rpcsvc_notify (rpcsvc_t *rpc, void *xl, rpcsvc_event_t event, + void *data) +{ + /* glusterd_rpcsvc_notify() does stuff that calls coming in from the + * unix domain socket don't need. This is just an empty function to be + * used for the uds listener. This will be used later if required. + */ + return 0; +} + +/* The glusterd unix domain socket listener only listens for cli */ +rpcsvc_t * +glusterd_init_uds_listener (xlator_t *this) +{ + int ret = -1; + dict_t *options = NULL; + rpcsvc_t *rpc = NULL; + data_t *sock_data = NULL; + char sockfile[PATH_MAX+1] = {0,}; + int i = 0; + + + GF_ASSERT (this); + + sock_data = dict_get (this->options, "glusterd-sockfile"); + if (!sock_data) { + strncpy (sockfile, DEFAULT_GLUSTERD_SOCKFILE, PATH_MAX); + } else { + strncpy (sockfile, sock_data->data, PATH_MAX); + } + + options = dict_new (); + if (!options) + goto out; + + ret = rpcsvc_transport_unix_options_build (&options, sockfile); + if (ret) + goto out; + + rpc = rpcsvc_init (this, this->ctx, options, 8); + if (rpc == NULL) { + ret = -1; + goto out; + } + + ret = rpcsvc_register_notify (rpc, glusterd_uds_rpcsvc_notify, + this); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Failed to register notify function"); + goto out; + } + + ret = rpcsvc_create_listeners (rpc, options, this->name); + if (ret != 1) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to create listener"); + goto out; + } + ret = 0; + + for (i = 0; i < gd_uds_programs_count; i++) { + ret = glusterd_program_register (this, rpc, gd_uds_programs[i]); + if (ret) { + i--; + for (; i >= 0; i--) + rpcsvc_program_unregister (rpc, + gd_uds_programs[i]); + + goto out; + } + } + +out: + if (options) + dict_unref (options); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to start glusterd " + "unix domain socket listener."); + if (rpc) { + GF_FREE (rpc); + rpc = NULL; + } + } + return rpc; +} + +void +glusterd_stop_uds_listener (xlator_t *this) +{ + glusterd_conf_t *conf = NULL; + rpcsvc_listener_t *listener = NULL; + rpcsvc_listener_t *next = NULL; + + GF_ASSERT (this); + conf = this->private; + + (void) rpcsvc_program_unregister (conf->uds_rpc, &gd_svc_cli_prog); + (void) rpcsvc_program_unregister (conf->uds_rpc, &gluster_handshake_prog); + + list_for_each_entry_safe (listener, next, &conf->uds_rpc->listeners, + list) { + rpcsvc_listener_destroy (listener); + } + + (void) rpcsvc_unregister_notify (conf->uds_rpc, glusterd_rpcsvc_notify, + this); + + unlink (DEFAULT_GLUSTERD_SOCKFILE); + + GF_FREE (conf->uds_rpc); + conf->uds_rpc = NULL; + + return; +} + +static int +glusterd_init_snap_folder (xlator_t *this) +{ + int ret = -1; + struct stat buf = {0,}; + + GF_ASSERT (this); + + /* Snapshot volumes are mounted under /var/run/gluster/snaps folder. + * But /var/run is normally a symbolic link to /run folder, which + * creates problems as the entry point in the mtab for the mount point + * and glusterd maintained entry point will be different. Therefore + * identify the correct run folder and use it for snap volume mounting. + */ + ret = lstat (GLUSTERD_VAR_RUN_DIR, &buf); + if (ret != 0) { + gf_log (this->name, GF_LOG_ERROR, + "stat fails on %s, exiting. (errno = %d)", + GLUSTERD_VAR_RUN_DIR, errno); + goto out; + } + + /* If /var/run is symlink then use /run folder */ + if (S_ISLNK (buf.st_mode)) { + strcpy (snap_mount_folder, GLUSTERD_RUN_DIR); + } else { + strcpy (snap_mount_folder, GLUSTERD_VAR_RUN_DIR); + } + + strcat (snap_mount_folder, GLUSTERD_DEFAULT_SNAPS_BRICK_DIR); + + ret = stat (snap_mount_folder, &buf); + if ((ret != 0) && (ENOENT != errno)) { + gf_log (this->name, GF_LOG_ERROR, + "stat fails on %s, exiting. (errno = %d)", + snap_mount_folder, errno); + ret = -1; + goto out; + } + + if ((!ret) && (!S_ISDIR(buf.st_mode))) { + gf_log (this->name, GF_LOG_CRITICAL, + "Provided snap path %s is not a directory," + "exiting", snap_mount_folder); + ret = -1; + goto out; + } + + if ((-1 == ret) && (ENOENT == errno)) { + /* Create missing folders */ + ret = mkdir_p (snap_mount_folder, 0777, _gf_false); + + if (-1 == ret) { + gf_log (this->name, GF_LOG_CRITICAL, + "Unable to create directory %s" + " ,errno = %d", snap_mount_folder, errno); + goto out; + } + } + +out: + return ret; } /* @@ -746,57 +1145,67 @@ init (xlator_t *this) { int32_t ret = -1; rpcsvc_t *rpc = NULL; + rpcsvc_t *uds_rpc = NULL; glusterd_conf_t *conf = NULL; data_t *dir_data = NULL; struct stat buf = {0,}; - char voldir [PATH_MAX] = {0,}; - char dirname [PATH_MAX]; + char storedir [PATH_MAX] = {0,}; + char workdir [PATH_MAX] = {0,}; + char hooks_dir [PATH_MAX] = {0,}; char cmd_log_filename [PATH_MAX] = {0,}; int first_time = 0; char *mountbroker_root = NULL; - -#ifdef DEBUG + int i = 0; char *valgrind_str = NULL; -#endif + dir_data = dict_get (this->options, "working-directory"); if (!dir_data) { //Use default working dir - strncpy (dirname, GLUSTERD_DEFAULT_WORKDIR, PATH_MAX); + strncpy (workdir, GLUSTERD_DEFAULT_WORKDIR, PATH_MAX); } else { - strncpy (dirname, dir_data->data, PATH_MAX); + strncpy (workdir, dir_data->data, PATH_MAX); } - ret = stat (dirname, &buf); + ret = stat (workdir, &buf); if ((ret != 0) && (ENOENT != errno)) { gf_log (this->name, GF_LOG_ERROR, "stat fails on %s, exiting. (errno = %d)", - dirname, errno); + workdir, errno); exit (1); } if ((!ret) && (!S_ISDIR(buf.st_mode))) { gf_log (this->name, GF_LOG_CRITICAL, "Provided working area %s is not a directory," - "exiting", dirname); + "exiting", workdir); exit (1); } if ((-1 == ret) && (ENOENT == errno)) { - ret = mkdir (dirname, 0777); + ret = mkdir (workdir, 0777); if (-1 == ret) { gf_log (this->name, GF_LOG_CRITICAL, "Unable to create directory %s" - " ,errno = %d", dirname, errno); + " ,errno = %d", workdir, errno); exit (1); } + first_time = 1; } + setenv ("GLUSTERD_WORKING_DIR", workdir, 1); gf_log (this->name, GF_LOG_INFO, "Using %s as working directory", - dirname); + workdir); + + ret = glusterd_init_snap_folder (this); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, "Unable to create " + "snap backend folder"); + exit (1); + } snprintf (cmd_log_filename, PATH_MAX,"%s/.cmd_log_history", DEFAULT_LOG_FILE_DIRECTORY); @@ -808,59 +1217,68 @@ init (xlator_t *this) exit (1); } - snprintf (voldir, PATH_MAX, "%s/vols", dirname); + snprintf (storedir, PATH_MAX, "%s/vols", workdir); - ret = mkdir (voldir, 0777); + ret = mkdir (storedir, 0777); if ((-1 == ret) && (errno != EEXIST)) { gf_log (this->name, GF_LOG_CRITICAL, "Unable to create volume directory %s" - " ,errno = %d", voldir, errno); + " ,errno = %d", storedir, errno); exit (1); } - snprintf (voldir, PATH_MAX, "%s/peers", dirname); + snprintf (storedir, PATH_MAX, "%s/peers", workdir); - ret = mkdir (voldir, 0777); + ret = mkdir (storedir, 0777); if ((-1 == ret) && (errno != EEXIST)) { gf_log (this->name, GF_LOG_CRITICAL, "Unable to create peers directory %s" - " ,errno = %d", voldir, errno); + " ,errno = %d", storedir, errno); exit (1); } - snprintf (voldir, PATH_MAX, "%s/bricks", DEFAULT_LOG_FILE_DIRECTORY); - ret = mkdir (voldir, 0777); + snprintf (storedir, PATH_MAX, "%s/bricks", DEFAULT_LOG_FILE_DIRECTORY); + ret = mkdir (storedir, 0777); if ((-1 == ret) && (errno != EEXIST)) { gf_log (this->name, GF_LOG_CRITICAL, "Unable to create logs directory %s" - " ,errno = %d", voldir, errno); + " ,errno = %d", storedir, errno); exit (1); } - snprintf (voldir, PATH_MAX, "%s/nfs", dirname); - ret = mkdir (voldir, 0777); + snprintf (storedir, PATH_MAX, "%s/nfs", workdir); + ret = mkdir (storedir, 0777); if ((-1 == ret) && (errno != EEXIST)) { gf_log (this->name, GF_LOG_CRITICAL, "Unable to create nfs directory %s" - " ,errno = %d", voldir, errno); + " ,errno = %d", storedir, errno); exit (1); } - snprintf (voldir, PATH_MAX, "%s/glustershd", dirname); - ret = mkdir (voldir, 0777); + snprintf (storedir, PATH_MAX, "%s/glustershd", workdir); + ret = mkdir (storedir, 0777); if ((-1 == ret) && (errno != EEXIST)) { gf_log (this->name, GF_LOG_CRITICAL, "Unable to create glustershd directory %s" - " ,errno = %d", voldir, errno); + " ,errno = %d", storedir, errno); + exit (1); + } + + snprintf (storedir, PATH_MAX, "%s/groups", workdir); + ret = mkdir (storedir, 0777); + if ((-1 == ret) && (errno != EEXIST)) { + gf_log (this->name, GF_LOG_CRITICAL, + "Unable to create glustershd directory %s" + " ,errno = %d", storedir, errno); exit (1); } ret = glusterd_rpcsvc_options_build (this->options); if (ret) goto out; - rpc = rpcsvc_init (this, this->ctx, this->options); + rpc = rpcsvc_init (this, this->ctx, this->options, 64); if (rpc == NULL) { gf_log (this->name, GF_LOG_ERROR, "failed to init rpc"); @@ -886,38 +1304,26 @@ init (xlator_t *this) goto out; } - ret = glusterd_program_register (this, rpc, &gd_svc_peer_prog); - if (ret) { - goto out; - } - - ret = glusterd_program_register (this, rpc, &gd_svc_cli_prog); - if (ret) { - rpcsvc_program_unregister (rpc, &gd_svc_peer_prog); - goto out; - } + for (i = 0; i < gd_inet_programs_count; i++) { + ret = glusterd_program_register (this, rpc, + gd_inet_programs[i]); + if (ret) { + i--; + for (; i >= 0; i--) + rpcsvc_program_unregister (rpc, + gd_inet_programs[i]); - ret = glusterd_program_register (this, rpc, &gd_svc_mgmt_prog); - if (ret) { - rpcsvc_program_unregister (rpc, &gd_svc_peer_prog); - rpcsvc_program_unregister (rpc, &gd_svc_cli_prog); - goto out; - } - - ret = glusterd_program_register (this, rpc, &gluster_pmap_prog); - if (ret) { - rpcsvc_program_unregister (rpc, &gd_svc_peer_prog); - rpcsvc_program_unregister (rpc, &gd_svc_cli_prog); - rpcsvc_program_unregister (rpc, &gd_svc_mgmt_prog); - goto out; + goto out; + } } - ret = glusterd_program_register (this, rpc, &gluster_handshake_prog); - if (ret) { - rpcsvc_program_unregister (rpc, &gd_svc_peer_prog); - rpcsvc_program_unregister (rpc, &gluster_pmap_prog); - rpcsvc_program_unregister (rpc, &gd_svc_cli_prog); - rpcsvc_program_unregister (rpc, &gd_svc_mgmt_prog); + /* Start a unix domain socket listener just for cli commands + * This should prevent ports from being wasted by being in TIMED_WAIT + * when cli commands are done continuously + */ + uds_rpc = glusterd_init_uds_listener (this); + if (uds_rpc == NULL) { + ret = -1; goto out; } @@ -927,15 +1333,30 @@ init (xlator_t *this) conf->shd = GF_CALLOC (1, sizeof (nodesrv_t), gf_gld_mt_nodesrv_t); GF_VALIDATE_OR_GOTO(this->name, conf->shd, out); + conf->nfs = GF_CALLOC (1, sizeof (nodesrv_t), + gf_gld_mt_nodesrv_t); + GF_VALIDATE_OR_GOTO(this->name, conf->nfs, out); INIT_LIST_HEAD (&conf->peers); INIT_LIST_HEAD (&conf->volumes); + INIT_LIST_HEAD (&conf->snapshots); + INIT_LIST_HEAD (&conf->missed_snaps_list); + pthread_mutex_init (&conf->mutex, NULL); conf->rpc = rpc; + conf->uds_rpc = uds_rpc; conf->gfs_mgmt = &gd_brick_prog; - strncpy (conf->workdir, dirname, PATH_MAX); + strncpy (conf->workdir, workdir, PATH_MAX); + synclock_init (&conf->big_lock); + pthread_mutex_init (&conf->xprt_lock, NULL); INIT_LIST_HEAD (&conf->xprt_list); + + glusterd_friend_sm_init (); + glusterd_op_sm_init (); + glusterd_opinfo_init (); + glusterd_mgmt_v3_lock_init (); + glusterd_txn_opinfo_dict_init (); ret = glusterd_sm_tr_log_init (&conf->op_sm_log, glusterd_op_sm_state_name_get, glusterd_op_sm_event_name_get, @@ -943,31 +1364,42 @@ init (xlator_t *this) if (ret) goto out; + conf->base_port = GF_IANA_PRIV_PORTS_START; + if (dict_get_uint32(this->options, "base-port", &conf->base_port) == 0) { + gf_log (this->name, GF_LOG_INFO, + "base-port override: %d", conf->base_port); + } + /* Set option to run bricks on valgrind if enabled in glusterd.vol */ -#ifdef DEBUG conf->valgrind = _gf_false; - ret = dict_get_str (this->options, "brick-with-valgrind", &valgrind_str); + ret = dict_get_str (this->options, "run-with-valgrind", &valgrind_str); if (ret < 0) { - gf_log (THIS->name, GF_LOG_ERROR, - "cannot get brick-with-valgrind value"); + gf_log (this->name, GF_LOG_DEBUG, + "cannot get run-with-valgrind value"); } if (valgrind_str) { if (gf_string2boolean (valgrind_str, &(conf->valgrind))) { - gf_log (THIS->name, GF_LOG_WARNING, - "brick-with-valgrind value not a boolean string"); + gf_log (this->name, GF_LOG_WARNING, + "run-with-valgrind value not a boolean string"); } } -#endif + this->private = conf; - (void) glusterd_shd_set_running (_gf_false); - /* this->ctx->top = this;*/ + (void) glusterd_nodesvc_set_online_status ("glustershd", _gf_false); - ret = glusterd_uuid_init (first_time); - if (ret < 0) - goto out; + GLUSTERD_GET_HOOKS_DIR (hooks_dir, GLUSTERD_HOOK_VER, conf); + if (stat (hooks_dir, &buf)) { + ret = glusterd_hooks_create_hooks_directory (conf->workdir); + if (-1 == ret) { + gf_log (this->name, GF_LOG_CRITICAL, + "Unable to create hooks directory "); + exit (1); + } + } INIT_LIST_HEAD (&conf->mount_specs); - dict_foreach (this->options, _install_mount_spec, &ret); + + ret = dict_foreach (this->options, _install_mount_spec, NULL); if (ret) goto out; ret = dict_get_str (this->options, "mountbroker-root", @@ -987,18 +1419,25 @@ init (xlator_t *this) if (ret < 0) goto out; - glusterd_friend_sm_init (); - glusterd_op_sm_init (); - glusterd_opinfo_init (); + /* If there are no 'friends', this would be the best time to + * spawn process/bricks that may need (re)starting since last + * time (this) glusterd was up.*/ + + if (list_empty (&conf->peers)) { + glusterd_launch_synctask (glusterd_spawn_daemons, NULL); + } + ret = glusterd_options_init (this); + if (ret < 0) + goto out; ret = glusterd_handle_upgrade_downgrade (this->options, conf); if (ret) goto out; - glusterd_restart_bricks (conf); - ret = glusterd_restart_gsyncds (conf); + ret = glusterd_hooks_spawn_worker (this); if (ret) goto out; + ret = 0; out: if (ret < 0) { @@ -1030,12 +1469,17 @@ fini (xlator_t *this) goto out; conf = this->private; - if (conf->pmap) - FREE (conf->pmap); + + glusterd_stop_uds_listener (this); + + FREE (conf->pmap); if (conf->handle) - glusterd_store_handle_destroy (conf->handle); + gf_store_handle_destroy (conf->handle); glusterd_sm_tr_log_delete (&conf->op_sm_log); + glusterd_mgmt_v3_lock_fini (); + glusterd_txn_opinfo_dict_fini (); GF_FREE (conf); + this->private = NULL; out: return; @@ -1073,11 +1517,9 @@ notify (xlator_t *this, int32_t event, void *data, ...) } -struct xlator_fops fops = { -}; +struct xlator_fops fops; -struct xlator_cbks cbks = { -}; +struct xlator_cbks cbks; struct xlator_dumpops dumpops = { .priv = glusterd_priv, @@ -1124,8 +1566,32 @@ struct volume_options options[] = { { .key = {GEOREP"-log-group"}, .type = GF_OPTION_TYPE_ANY, }, - { .key = {"brick-with-valgrind"}, + { .key = {"run-with-valgrind"}, .type = GF_OPTION_TYPE_BOOL, }, + { .key = {"server-quorum-type"}, + .type = GF_OPTION_TYPE_STR, + .value = { "none", "server"}, + .description = "If set to server, enables the specified " + "volume to participate in quorum." + }, + { .key = {"server-quorum-ratio"}, + .type = GF_OPTION_TYPE_PERCENT, + .description = "Sets the quorum percentage for the trusted " + "storage pool." + }, + { .key = {"glusterd-sockfile"}, + .type = GF_OPTION_TYPE_PATH, + .description = "The socket file on which glusterd should listen for " + "cli requests. Default is "DEFAULT_GLUSTERD_SOCKFILE "." + }, + { .key = {"base-port"}, + .type = GF_OPTION_TYPE_INT, + .description = "Sets the base port for portmap query" + }, + { .key = {"snap-brick-path"}, + .type = GF_OPTION_TYPE_STR, + .description = "directory where the bricks for the snapshots will be created" + }, { .key = {NULL} }, }; diff --git a/xlators/mgmt/glusterd/src/glusterd.h b/xlators/mgmt/glusterd/src/glusterd.h index fe2402ec2..3aa395ebc 100644 --- a/xlators/mgmt/glusterd/src/glusterd.h +++ b/xlators/mgmt/glusterd/src/glusterd.h @@ -1,22 +1,12 @@ /* - Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com> - This file is part of GlusterFS. - - GlusterFS is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published - by the Free Software Foundation; either version 3 of the License, - or (at your option) any later version. - - GlusterFS is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program. If not, see - <http://www.gnu.org/licenses/>. -*/ + Copyright (c) 2006-2013 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ #ifndef _GLUSTERD_H_ #define _GLUSTERD_H_ @@ -46,13 +36,47 @@ #include "protocol-common.h" #include "glusterd-pmap.h" #include "cli1-xdr.h" +#include "syncop.h" +#include "store.h" #define GLUSTERD_MAX_VOLUME_NAME 1000 -#define DEFAULT_LOG_FILE_DIRECTORY DATADIR "/log/glusterfs" #define GLUSTERD_TR_LOG_SIZE 50 #define GLUSTERD_NAME "glusterd" #define GLUSTERD_SOCKET_LISTEN_BACKLOG 128 +#define GLUSTERD_QUORUM_TYPE_KEY "cluster.server-quorum-type" +#define GLUSTERD_QUORUM_RATIO_KEY "cluster.server-quorum-ratio" +#define GLUSTERD_GLOBAL_OPT_VERSION "global-option-version" +#define GLUSTERD_COMMON_PEM_PUB_FILE "/geo-replication/common_secret.pem.pub" +#define GEO_CONF_MAX_OPT_VALS 5 +#define GLUSTERD_CREATE_HOOK_SCRIPT "/hooks/1/gsync-create/post/" \ + "S56glusterd-geo-rep-create-post.sh" + + +#define GLUSTERD_SNAPS_MAX_HARD_LIMIT 256 +#define GLUSTERD_SNAPS_DEF_SOFT_LIMIT_PERCENT 90 +#define GLUSTERD_SNAPS_MAX_SOFT_LIMIT_PERCENT 100 +#define GLUSTERD_SERVER_QUORUM "server" + +#define FMTSTR_CHECK_VOL_EXISTS "Volume %s does not exist" +#define FMTSTR_RESOLVE_BRICK "Could not find peer on which brick %s:%s resides" + +#define LOGSTR_FOUND_BRICK "Found brick %s:%s in volume %s" +#define LOGSTR_BUILD_PAYLOAD "Failed to build payload for operation 'Volume %s'" +#define LOGSTR_STAGE_FAIL "Staging of operation 'Volume %s' failed on %s %s %s" +#define LOGSTR_COMMIT_FAIL "Commit of operation 'Volume %s' failed on %s %s %s" + +#define OPERRSTR_BUILD_PAYLOAD "Failed to build payload. Please check the log "\ + "file for more details." +#define OPERRSTR_STAGE_FAIL "Staging failed on %s. Please check the log file " \ + "for more details." +#define OPERRSTR_COMMIT_FAIL "Commit failed on %s. Please check the log file "\ + "for more details." +struct glusterd_volinfo_; +typedef struct glusterd_volinfo_ glusterd_volinfo_t; + +struct glusterd_snap_; +typedef struct glusterd_snap_ glusterd_snap_t; typedef enum glusterd_op_ { GD_OP_NONE = 0, @@ -77,68 +101,98 @@ typedef enum glusterd_op_ { GD_OP_REBALANCE, GD_OP_HEAL_VOLUME, GD_OP_STATEDUMP_VOLUME, + GD_OP_LIST_VOLUME, + GD_OP_CLEARLOCKS_VOLUME, + GD_OP_DEFRAG_BRICK_VOLUME, + GD_OP_COPY_FILE, + GD_OP_SYS_EXEC, + GD_OP_GSYNC_CREATE, + GD_OP_SNAP, GD_OP_MAX, } glusterd_op_t; - -struct glusterd_store_iter_ { - int fd; - FILE *file; - char filepath[PATH_MAX]; -}; - -typedef struct glusterd_store_iter_ glusterd_store_iter_t; +extern const char * gd_op_list[]; struct glusterd_volgen { dict_t *dict; }; + typedef struct { - struct rpc_clnt *rpc; - gf_boolean_t running; + struct rpc_clnt *rpc; + gf_boolean_t online; } nodesrv_t; typedef struct { - struct _volfile_ctx *volfile; - pthread_mutex_t mutex; - struct list_head peers; -// struct list_head pending_peers; - gf_boolean_t verify_volfile_checksum; - gf_boolean_t trace; - uuid_t uuid; - char workdir[PATH_MAX]; - rpcsvc_t *rpc; - nodesrv_t *shd; - struct pmap_registry *pmap; - struct list_head volumes; - struct list_head xprt_list; - glusterd_store_handle_t *handle; - gf_timer_t *timer; - glusterd_sm_tr_log_t op_sm_log; - struct rpc_clnt_program *gfs_mgmt; - struct list_head mount_specs; -#ifdef DEBUG - gf_boolean_t valgrind; -#endif + gf_boolean_t quorum; + double quorum_ratio; + uint64_t gl_opt_version; +} gd_global_opts_t; + +typedef struct { + struct _volfile_ctx *volfile; + pthread_mutex_t mutex; + struct list_head peers; + struct list_head xaction_peers; + gf_boolean_t verify_volfile_checksum; + gf_boolean_t trace; + uuid_t uuid; + char workdir[PATH_MAX]; + rpcsvc_t *rpc; + nodesrv_t *shd; + nodesrv_t *nfs; + struct pmap_registry *pmap; + struct list_head volumes; + struct list_head snapshots; /*List of snap volumes */ + pthread_mutex_t xprt_lock; + struct list_head xprt_list; + gf_store_handle_t *handle; + gf_timer_t *timer; + glusterd_sm_tr_log_t op_sm_log; + struct rpc_clnt_program *gfs_mgmt; + + struct list_head mount_specs; + gf_boolean_t valgrind; + pthread_t brick_thread; + void *hooks_priv; + /* need for proper handshake_t */ + int op_version; /* Starts with 1 for 3.3.0 */ + xlator_t *xl; /* Should be set to 'THIS' before creating thread */ + gf_boolean_t pending_quorum_action; + dict_t *opts; + synclock_t big_lock; + gf_boolean_t restart_done; + rpcsvc_t *uds_rpc; /* RPCSVC for the unix domain socket */ + uint32_t base_port; + uint64_t snap_max_hard_limit; + uint64_t snap_max_soft_limit; + char *snap_bricks_directory; + gf_store_handle_t *missed_snaps_list_shandle; + struct list_head missed_snaps_list; } glusterd_conf_t; + typedef enum gf_brick_status { GF_BRICK_STOPPED, GF_BRICK_STARTED, } gf_brick_status_t; struct glusterd_brickinfo { - char hostname[1024]; - char path[PATH_MAX]; - struct list_head brick_list; - uuid_t uuid; - int port; - int rdma_port; - char *logfile; - gf_boolean_t signed_in; - glusterd_store_handle_t *shandle; - gf_brick_status_t status; - struct rpc_clnt *rpc; - int decommissioned; + char hostname[1024]; + char path[PATH_MAX]; + char device_path[PATH_MAX]; + struct list_head brick_list; + uuid_t uuid; + int port; + int rdma_port; + char *logfile; + gf_boolean_t signed_in; + gf_store_handle_t *shandle; + gf_brick_status_t status; + struct rpc_clnt *rpc; + int decommissioned; + char vg[PATH_MAX]; /* FIXME: Use max size for length of vg */ + int caps; /* Capability */ + int32_t snap_status; }; typedef struct glusterd_brickinfo glusterd_brickinfo_t; @@ -149,9 +203,6 @@ struct gf_defrag_brickinfo_ { int size; }; -struct glusterd_volinfo_; -typedef struct glusterd_volinfo_ glusterd_volinfo_t; - typedef int (*defrag_cbk_fn_t) (glusterd_volinfo_t *volinfo, gf_defrag_status_t status); @@ -159,9 +210,13 @@ struct glusterd_defrag_info_ { uint64_t total_files; uint64_t total_data; uint64_t num_files_lookedup; + uint64_t total_failures; gf_lock_t lock; int cmd; pthread_t th; + gf_defrag_status_t defrag_status; + struct rpc_clnt *rpc; + uint32_t connected; char mount[1024]; char databuf[131072]; struct gf_defrag_brickinfo_ *bricks; /* volinfo->brick_count */ @@ -186,56 +241,168 @@ typedef enum gf_rb_status_ { GF_RB_STATUS_PAUSED, } gf_rb_status_t; +struct _auth { + char *username; + char *password; +}; + +typedef struct _auth auth_t; + +/* Capabilities of xlator */ +#define CAPS_BD 0x00000001 +#define CAPS_THIN 0x00000002 +#define CAPS_OFFLOAD_COPY 0x00000004 +#define CAPS_OFFLOAD_SNAPSHOT 0x00000008 + +struct glusterd_rebalance_ { + gf_defrag_status_t defrag_status; + uint64_t rebalance_files; + uint64_t rebalance_data; + uint64_t lookedup_files; + uint64_t skipped_files; + glusterd_defrag_info_t *defrag; + gf_cli_defrag_type defrag_cmd; + uint64_t rebalance_failures; + uuid_t rebalance_id; + double rebalance_time; + glusterd_op_t op; + dict_t *dict; /* Dict to store misc information + * like list of bricks being removed */ +}; + +typedef struct glusterd_rebalance_ glusterd_rebalance_t; + +struct glusterd_replace_brick_ { + gf_rb_status_t rb_status; + glusterd_brickinfo_t *src_brick; + glusterd_brickinfo_t *dst_brick; + uuid_t rb_id; +}; + +typedef struct glusterd_replace_brick_ glusterd_replace_brick_t; + struct glusterd_volinfo_ { - char volname[GLUSTERD_MAX_VOLUME_NAME]; - int type; - int brick_count; - struct list_head vol_list; - struct list_head bricks; - glusterd_volume_status status; - int sub_count; /* backward compatibility */ - int stripe_count; - int replica_count; - int dist_leaf_count; /* Number of bricks in one + gf_lock_t lock; + char volname[GLUSTERD_MAX_VOLUME_NAME]; + gf_boolean_t is_snap_volume; + glusterd_snap_t *snapshot; + gf_boolean_t is_volume_restored; + char parent_volname[GLUSTERD_MAX_VOLUME_NAME]; + /* In case of a snap volume + i.e (is_snap_volume == TRUE) this + field will contain the name of + the volume which is snapped. In + case of a non-snap volume, this + field will be initialized as N/A */ + int type; + int brick_count; + uint64_t snap_count; + uint64_t snap_max_hard_limit; + struct list_head vol_list; + /* In case of a snap volume + i.e (is_snap_volume == TRUE) this + is linked to glusterd_snap_t->volumes. + In case of a non-snap volume, this is + linked to glusterd_conf_t->volumes */ + struct list_head snapvol_list; + /* This is a current pointer for + glusterd_volinfo_t->snap_volumes */ + struct list_head bricks; + struct list_head snap_volumes; + /* TODO : Need to remove this, as this + * is already part of snapshot object. + */ + glusterd_volume_status status; + int sub_count; /* backward compatibility */ + int stripe_count; + int replica_count; + int subvol_count; /* Number of subvolumes in a + distribute volume */ + int dist_leaf_count; /* Number of bricks in one distribute subvolume */ - int port; - glusterd_store_handle_t *shandle; - glusterd_store_handle_t *rb_shandle; + int port; + gf_store_handle_t *shandle; + gf_store_handle_t *rb_shandle; + gf_store_handle_t *node_state_shandle; /* Defrag/rebalance related */ - gf_defrag_status_t defrag_status; - uint64_t rebalance_files; - uint64_t rebalance_data; - uint64_t lookedup_files; - glusterd_defrag_info_t *defrag; + glusterd_rebalance_t rebal; /* Replace brick status */ - gf_rb_status_t rb_status; - glusterd_brickinfo_t *src_brick; - glusterd_brickinfo_t *dst_brick; + glusterd_replace_brick_t rep_brick; - int version; - uint32_t cksum; - gf_transport_type transport_type; - gf_transport_type nfs_transport_type; + int version; + uint32_t cksum; + gf_transport_type transport_type; + gf_transport_type nfs_transport_type; - dict_t *dict; + dict_t *dict; - uuid_t volume_id; - char *logdir; + uuid_t volume_id; + auth_t auth; + char *logdir; - dict_t *gsync_slaves; + dict_t *gsync_slaves; - int decommission_in_progress; - xlator_t *xl; + int decommission_in_progress; + xlator_t *xl; + + gf_boolean_t memory_accounting; + int caps; /* Capability */ + + int op_version; + int client_op_version; }; +typedef enum gd_snap_status_ { + GD_SNAP_STATUS_NONE, + GD_SNAP_STATUS_INIT, + GD_SNAP_STATUS_IN_USE, + GD_SNAP_STATUS_DECOMMISSION, + GD_SNAP_STATUS_RESTORED, +} gd_snap_status_t; + +struct glusterd_snap_ { + gf_lock_t lock; + struct list_head volumes; + struct list_head snap_list; + char snapname[GLUSTERD_MAX_SNAP_NAME]; + uuid_t snap_id; + char *description; + time_t time_stamp; + gf_boolean_t snap_restored; + gd_snap_status_t snap_status; + gf_store_handle_t *shandle; +}; + +typedef struct glusterd_snap_op_ { + int32_t brick_num; + char *brick_path; + int32_t op; + int32_t status; + struct list_head snap_ops_list; +} glusterd_snap_op_t; + +typedef struct glusterd_missed_snap_ { + char *node_snap_info; + struct list_head missed_snaps; + struct list_head snap_ops; +} glusterd_missed_snap_info; + typedef enum gd_node_type_ { GD_NODE_NONE, GD_NODE_BRICK, - GD_NODE_SHD + GD_NODE_SHD, + GD_NODE_REBALANCE, + GD_NODE_NFS, } gd_node_type; +typedef enum missed_snap_stat { + GD_MISSED_SNAP_NONE, + GD_MISSED_SNAP_PENDING, + GD_MISSED_SNAP_DONE, +} missed_snap_stat; + typedef struct glusterd_pending_node_ { struct list_head list; void *node; @@ -243,6 +410,13 @@ typedef struct glusterd_pending_node_ { int32_t index; } glusterd_pending_node_t; +struct gsync_config_opt_vals_ { + char *op_name; + int no_of_pos_vals; + gf_boolean_t case_sensitive; + char *values[GEO_CONF_MAX_OPT_VALS]; +}; + enum glusterd_op_ret { GLUSTERD_CONNECTION_AWAITED = 100, }; @@ -254,15 +428,24 @@ enum glusterd_vol_comp_status_ { GLUSTERD_VOL_COMP_RJT, }; -#define GLUSTERD_DEFAULT_WORKDIR "/etc/glusterd" +#define GLUSTERD_DEFAULT_WORKDIR "/var/lib/glusterd" #define GLUSTERD_DEFAULT_PORT GF_DEFAULT_BASE_PORT #define GLUSTERD_INFO_FILE "glusterd.info" #define GLUSTERD_VOLUME_DIR_PREFIX "vols" #define GLUSTERD_PEER_DIR_PREFIX "peers" #define GLUSTERD_VOLUME_INFO_FILE "info" +#define GLUSTERD_SNAP_INFO_FILE "info" #define GLUSTERD_VOLUME_RBSTATE_FILE "rbstate" #define GLUSTERD_BRICK_INFO_DIR "bricks" #define GLUSTERD_CKSUM_FILE "cksum" +#define GLUSTERD_TRASH "trash" +#define GLUSTERD_NODE_STATE_FILE "node_state.info" +#define GLUSTERD_MISSED_SNAPS_LIST_FILE "missed_snaps_list" +#define GLUSTERD_VOL_SNAP_DIR_PREFIX "snaps" + +#define GLUSTERD_DEFAULT_SNAPS_BRICK_DIR "/gluster/snaps" +#define GLUSTERD_VAR_RUN_DIR "/var/run" +#define GLUSTERD_RUN_DIR "/run" /* definitions related to replace brick */ #define RB_CLIENT_MOUNTPOINT "rb_mount" @@ -275,14 +458,29 @@ enum glusterd_vol_comp_status_ { typedef ssize_t (*gd_serialize_t) (struct iovec outmsg, void *args); -#define GLUSTERD_GET_VOLUME_DIR(path, volinfo, priv) \ - snprintf (path, PATH_MAX, "%s/vols/%s", priv->workdir,\ - volinfo->volname); +#define GLUSTERD_GET_VOLUME_DIR(path, volinfo, priv) \ + if (volinfo->is_snap_volume) { \ + snprintf (path, PATH_MAX, "%s/snaps/%s/%s", priv->workdir, \ + volinfo->snapshot->snapname, volinfo->volname); \ + } else { \ + snprintf (path, PATH_MAX, "%s/vols/%s", priv->workdir, \ + volinfo->volname); \ + } -#define GLUSTERD_GET_BRICK_DIR(path, volinfo, priv) \ - snprintf (path, PATH_MAX, "%s/%s/%s/%s", priv->workdir,\ - GLUSTERD_VOLUME_DIR_PREFIX, volinfo->volname, \ - GLUSTERD_BRICK_INFO_DIR); +#define GLUSTERD_GET_SNAP_DIR(path, snap, priv) \ + snprintf (path, PATH_MAX, "%s/snaps/%s", priv->workdir, \ + snap->snapname); + +#define GLUSTERD_GET_BRICK_DIR(path, volinfo, priv) \ + if (volinfo->is_snap_volume) { \ + snprintf (path, PATH_MAX, "%s/snaps/%s/%s/%s", priv->workdir, \ + volinfo->snapshot->snapname, volinfo->volname, \ + GLUSTERD_BRICK_INFO_DIR); \ + } else { \ + snprintf (path, PATH_MAX, "%s/%s/%s/%s", priv->workdir, \ + GLUSTERD_VOLUME_DIR_PREFIX, volinfo->volname, \ + GLUSTERD_BRICK_INFO_DIR); \ + } #define GLUSTERD_GET_NFS_DIR(path, priv) \ snprintf (path, PATH_MAX, "%s/nfs", priv->workdir); @@ -296,12 +494,14 @@ typedef ssize_t (*gd_serialize_t) (struct iovec outmsg, void *args); } \ } while (0) -#define GLUSTERD_GET_BRICK_PIDFILE(pidfile,volpath,hostname,brickpath) { \ - char exp_path[PATH_MAX] = {0,}; \ - GLUSTERD_REMOVE_SLASH_FROM_PATH (brickpath, exp_path); \ - snprintf (pidfile, PATH_MAX, "%s/run/%s-%s.pid", \ - volpath, hostname, exp_path); \ - } +#define GLUSTERD_GET_BRICK_PIDFILE(pidfile,volinfo,brickinfo, priv) do { \ + char exp_path[PATH_MAX] = {0,}; \ + char volpath[PATH_MAX] = {0,}; \ + GLUSTERD_GET_VOLUME_DIR (volpath, volinfo, priv); \ + GLUSTERD_REMOVE_SLASH_FROM_PATH (brickinfo->path, exp_path); \ + snprintf (pidfile, PATH_MAX, "%s/run/%s-%s.pid", \ + volpath, brickinfo->hostname, exp_path); \ + } while (0) #define GLUSTERD_GET_NFS_PIDFILE(pidfile,nfspath) { \ snprintf (pidfile, PATH_MAX, "%s/run/nfs.pid", \ @@ -309,19 +509,81 @@ typedef ssize_t (*gd_serialize_t) (struct iovec outmsg, void *args); } #define GLUSTERD_STACK_DESTROY(frame) do {\ - frame->local = NULL; \ - STACK_DESTROY (frame->root);\ - } while (0) + frame->local = NULL; \ + STACK_DESTROY (frame->root); \ + } while (0) + +#define GLUSTERD_GET_DEFRAG_DIR(path, volinfo, priv) do { \ + char vol_path[PATH_MAX]; \ + GLUSTERD_GET_VOLUME_DIR(vol_path, volinfo, priv); \ + snprintf (path, PATH_MAX, "%s/rebalance",vol_path); \ + } while (0) + +#define GLUSTERD_GET_DEFRAG_SOCK_FILE(path, volinfo, priv) do { \ + char defrag_path[PATH_MAX]; \ + GLUSTERD_GET_DEFRAG_DIR(defrag_path, volinfo, priv); \ + snprintf (path, PATH_MAX, "%s/%s.sock", defrag_path, \ + uuid_utoa(MY_UUID)); \ + } while (0) + +#define GLUSTERD_GET_DEFRAG_PID_FILE(path, volinfo, priv) do { \ + char defrag_path[PATH_MAX]; \ + GLUSTERD_GET_DEFRAG_DIR(defrag_path, volinfo, priv); \ + snprintf (path, PATH_MAX, "%s/%s.pid", defrag_path, \ + uuid_utoa(MY_UUID)); \ + } while (0) + +#define GLUSTERD_GET_UUID_NOHYPHEN(ret_string, uuid) do { \ + char *snap_volname_ptr = ret_string; \ + char *snap_volid_ptr = uuid_utoa(uuid); \ + while (*snap_volid_ptr) { \ + if (*snap_volid_ptr == '-') { \ + snap_volid_ptr++; \ + } else { \ + (*snap_volname_ptr++) = \ + (*snap_volid_ptr++); \ + } \ + } \ + *snap_volname_ptr = '\0'; \ + } while (0) + +int glusterd_uuid_init(); + +int glusterd_uuid_generate_save (); + +#define MY_UUID (__glusterd_uuid()) + +static inline unsigned char * +__glusterd_uuid() +{ + glusterd_conf_t *priv = THIS->private; + + if (uuid_is_null (priv->uuid)) + glusterd_uuid_init(); + return &priv->uuid[0]; +} + +int glusterd_big_locked_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, + void *data, rpc_clnt_notify_t notify_fn); + +int +glusterd_big_locked_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe, fop_cbk_fn_t fn); + +int glusterd_big_locked_handler (rpcsvc_request_t *req, rpcsvc_actor actor_fn); int32_t glusterd_brick_from_brickinfo (glusterd_brickinfo_t *brickinfo, char **new_brick); int -glusterd_probe_begin (rpcsvc_request_t *req, const char *hoststr, int port); +glusterd_probe_begin (rpcsvc_request_t *req, const char *hoststr, int port, + dict_t *dict); int -glusterd_xfer_friend_add_resp (rpcsvc_request_t *req, char *hostname, - int port, int32_t op_ret, int32_t op_errno); +glusterd_xfer_friend_add_resp (rpcsvc_request_t *req, char *myhostname, + char *remote_hostname, int port, int32_t op_ret, + int32_t op_errno); int glusterd_friend_find (uuid_t uuid, char *hostname, @@ -330,12 +592,13 @@ glusterd_friend_find (uuid_t uuid, char *hostname, int glusterd_friend_add (const char *hoststr, int port, glusterd_friend_sm_state_t state, - uuid_t *uuid, struct rpc_clnt *rpc, - glusterd_peerinfo_t **friend, - gf_boolean_t restore, - glusterd_peerctx_args_t *args); + uuid_t *uuid, glusterd_peerinfo_t **friend, + gf_boolean_t restore, glusterd_peerctx_args_t *args); int +glusterd_friend_rpc_create (xlator_t *this, glusterd_peerinfo_t *peerinfo, + glusterd_peerctx_args_t *args); +int glusterd_friend_remove (uuid_t uuid, char *hostname); int @@ -345,6 +608,14 @@ int glusterd_op_unlock_send_resp (rpcsvc_request_t *req, int32_t status); int +glusterd_op_mgmt_v3_lock_send_resp (rpcsvc_request_t *req, + uuid_t *txn_id, int32_t status); + +int +glusterd_op_mgmt_v3_unlock_send_resp (rpcsvc_request_t *req, + uuid_t *txn_id, int32_t status); + +int glusterd_op_stage_send_resp (rpcsvc_request_t *req, int32_t op, int32_t status, char *op_errstr, dict_t *rsp_dict); @@ -388,7 +659,8 @@ glusterd_handle_defrag_volume_v2 (rpcsvc_request_t *req); int glusterd_xfer_cli_probe_resp (rpcsvc_request_t *req, int32_t op_ret, - int32_t op_errno, char *hostname, int port); + int32_t op_errno, char *op_errstr, char *hostname, + int port, dict_t *dict); int glusterd_op_commit_send_resp (rpcsvc_request_t *req, @@ -400,7 +672,7 @@ glusterd_xfer_friend_remove_resp (rpcsvc_request_t *req, char *hostname, int por int glusterd_deprobe_begin (rpcsvc_request_t *req, const char *hoststr, int port, - uuid_t uuid); + uuid_t uuid, dict_t *dict); int glusterd_handle_cli_deprobe (rpcsvc_request_t *req); @@ -475,6 +747,12 @@ int glusterd_handle_reset_volume (rpcsvc_request_t *req); int +glusterd_handle_copy_file (rpcsvc_request_t *req); + +int +glusterd_handle_sys_exec (rpcsvc_request_t *req); + +int glusterd_handle_gsync_set (rpcsvc_request_t *req); int @@ -485,7 +763,8 @@ glusterd_handle_fsm_log (rpcsvc_request_t *req); int glusterd_xfer_cli_deprobe_resp (rpcsvc_request_t *req, int32_t op_ret, - int32_t op_errno, char *hostname); + int32_t op_errno, char *op_errstr, + char *hostname, dict_t *dict); int glusterd_fetchspec_notify (xlator_t *this); @@ -502,8 +781,8 @@ glusterd_volume_txn (rpcsvc_request_t *req, char *volname, int flags, glusterd_op_t op); int -glusterd_peer_handshake (xlator_t *this, struct rpc_clnt *rpc, - glusterd_peerctx_t *peerctx); +glusterd_peer_dump_version (xlator_t *this, struct rpc_clnt *rpc, + glusterd_peerctx_t *peerctx); int glusterd_validate_reconfopts (glusterd_volinfo_t *volinfo, dict_t *val_dict, char **op_errstr); @@ -524,8 +803,8 @@ glusterd_brick_rpc_notify (struct rpc_clnt *rpc, void *mydata, rpc_clnt_event_t event, void *data); int -glusterd_shd_rpc_notify (struct rpc_clnt *rpc, void *mydata, - rpc_clnt_event_t event, void *data); +glusterd_nodesvc_rpc_notify (struct rpc_clnt *rpc, void *mydata, + rpc_clnt_event_t event, void *data); int glusterd_rpc_create (struct rpc_clnt **rpc, dict_t *options, @@ -533,22 +812,40 @@ glusterd_rpc_create (struct rpc_clnt **rpc, dict_t *options, /* handler functions */ -int32_t glusterd_op_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx); +int32_t glusterd_op_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx, + char *err_str, size_t size); /* removed other definitions as they have been defined elsewhere in this file*/ int glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req); +int glusterd_handle_cli_clearlocks_volume (rpcsvc_request_t *req); int glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr, - size_t len, int cmd, defrag_cbk_fn_t cbk); + size_t len, int cmd, defrag_cbk_fn_t cbk, + glusterd_op_t op); +int +glusterd_rebalance_rpc_create (glusterd_volinfo_t *volinfo, + glusterd_conf_t *priv, int cmd); + int glusterd_handle_cli_heal_volume (rpcsvc_request_t *req); +int glusterd_handle_cli_list_volume (rpcsvc_request_t *req); + +int +glusterd_handle_snapshot (rpcsvc_request_t *req); + /* op-sm functions */ int glusterd_op_stage_heal_volume (dict_t *dict, char **op_errstr); int glusterd_op_heal_volume (dict_t *dict, char **op_errstr); int glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr); int glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict); -int glusterd_op_quota (dict_t *dict, char **op_errstr); +int glusterd_op_stage_copy_file (dict_t *dict, char **op_errstr); +int glusterd_op_copy_file (dict_t *dict, char **op_errstr); +int glusterd_op_stage_sys_exec (dict_t *dict, char **op_errstr); +int glusterd_op_sys_exec (dict_t *dict, char **op_errstr, dict_t *rsp_dict); +int glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr); +int glusterd_op_gsync_create (dict_t *dict, char **op_errstr, dict_t *rsp_dict); +int glusterd_op_quota (dict_t *dict, char **op_errstr, dict_t *rsp_dict); int glusterd_op_stage_quota (dict_t *dict, char **op_errstr); int glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, dict_t *rsp_dict); @@ -575,6 +872,10 @@ int glusterd_op_rebalance (dict_t *dict, char **op_errstr, dict_t *rsp_dict); int glusterd_op_stage_statedump_volume (dict_t *dict, char **op_errstr); int glusterd_op_statedump_volume (dict_t *dict, char **op_errstr); +int glusterd_op_stage_clearlocks_volume (dict_t *dict, char **op_errstr); +int glusterd_op_clearlocks_volume (dict_t *dict, char **op_errstr, + dict_t *rsp_dict); + /* misc */ void glusterd_do_replace_brick (void *data); int glusterd_op_perform_remove_brick (glusterd_volinfo_t *volinfo, char *brick, @@ -583,4 +884,69 @@ int glusterd_op_stop_volume_args_get (dict_t *dict, char** volname, int *flags); int glusterd_op_statedump_volume_args_get (dict_t *dict, char **volname, char **options, int *option_cnt); +int glusterd_op_gsync_args_get (dict_t *dict, char **op_errstr, + char **master, char **slave, char **host_uuid); +int glusterd_stop_volume (glusterd_volinfo_t *volinfo); + +/* Synctask part */ +int32_t glusterd_op_begin_synctask (rpcsvc_request_t *req, glusterd_op_t op, + void *dict); +int32_t +glusterd_defrag_event_notify_handle (dict_t *dict); + +int32_t +glusterd_txn_opinfo_dict_init (); + +void +glusterd_txn_opinfo_dict_fini (); + +void +glusterd_txn_opinfo_init (); + +/* snapshot */ +glusterd_snap_t* +glusterd_new_snap_object(); + +int32_t +glusterd_list_add_snapvol (glusterd_volinfo_t *origin_vol, + glusterd_volinfo_t *snap_vol); + +glusterd_snap_t* +glusterd_remove_snap_by_id (uuid_t snap_id); + +glusterd_snap_t* +glusterd_remove_snap_by_name (char *snap_name); + +glusterd_snap_t* +glusterd_find_snap_by_name (char *snap_name); + +glusterd_snap_t* +glusterd_find_snap_by_id (uuid_t snap_id); + +int +glusterd_snapshot_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict); +int +glusterd_snapshot_brickop (dict_t *dict, char **op_errstr, dict_t *rsp_dict); +int +glusterd_snapshot (dict_t *dict, char **op_errstr, dict_t *rsp_dict); +int +glusterd_snapshot_postvalidate (dict_t *dict, int32_t op_ret, char **op_errstr, + dict_t *rsp_dict); +char * +glusterd_build_snap_device_path (char *device, char *snapname); +int32_t +glusterd_snap_remove (dict_t *rsp_dict, glusterd_snap_t *snap, + gf_boolean_t remove_lvm, gf_boolean_t force); +int32_t +glusterd_snapshot_cleanup (dict_t *dict, char **op_errstr, dict_t *rsp_dict); + +int32_t +glusterd_add_missed_snaps_to_list (dict_t *dict, int32_t missed_snap_count); + +int32_t +glusterd_store_missed_snaps_list (char *missed_info, int32_t brick_num, + char *brick_path, int32_t snap_op, + int32_t snap_status); + #endif |