diff options
Diffstat (limited to 'xlators/mgmt/glusterd/src/glusterd-handshake.c')
| -rw-r--r-- | xlators/mgmt/glusterd/src/glusterd-handshake.c | 2778 |
1 files changed, 2468 insertions, 310 deletions
diff --git a/xlators/mgmt/glusterd/src/glusterd-handshake.c b/xlators/mgmt/glusterd/src/glusterd-handshake.c index 269c6015a17..d96e35503dd 100644 --- a/xlators/mgmt/glusterd/src/glusterd-handshake.c +++ b/xlators/mgmt/glusterd/src/glusterd-handshake.c @@ -1,422 +1,2580 @@ /* - 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. -#ifndef _CONFIG_H -#define _CONFIG_H -#include "config.h" -#endif + 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. +*/ -#include "xlator.h" -#include "defaults.h" -#include "glusterfs.h" -#include "compat-errno.h" +#include <glusterfs/xlator.h> +#include <glusterfs/defaults.h> +#include <glusterfs/glusterfs.h> +#include <glusterfs/syscall.h> +#include <glusterfs/compat-errno.h> #include "glusterd.h" #include "glusterd-utils.h" #include "glusterd-op-sm.h" - +#include "glusterd-store.h" +#include "glusterd-snapshot-utils.h" +#include "glusterd-svc-mgmt.h" +#include "glusterd-snapd-svc-helper.h" +#include "glusterd-volgen.h" +#include "glusterd-quotad-svc.h" +#include "glusterd-messages.h" #include "glusterfs3.h" #include "protocol-common.h" #include "rpcsvc.h" +#include "rpc-common-xdr.h" +#include "glusterd-gfproxyd-svc-helper.h" +#include "glusterd-shd-svc-helper.h" 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; -typedef ssize_t (*gfs_serialize_t) (struct iovec outmsg, void *data); - -static size_t -build_volfile_path (const char *volname, char *path, - size_t path_len) -{ - 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; - - if (strstr (volname, "gluster/")) { - server = strchr (volname, '/') + 1; - glusterd_get_nodesvc_volfile (server, priv->workdir, - path, path_len); - ret = 1; - goto out; - } else if (volname[0] != '/') { - /* Normal behavior */ - dup_volname = gf_strdup (volname); - } else { - /* Bringing in NFS like behavior for mount command, */ - /* With this, one can mount a volume with below cmd */ - /* bash# mount -t glusterfs server:/volume /mnt/pnt */ - dup_volname = gf_strdup (&volname[1]); - } +#define TRUSTED_PREFIX "trusted-" +#define GD_PEER_ID_KEY "peer-id" - free_ptr = dup_volname; +typedef ssize_t (*gfs_serialize_t)(struct iovec outmsg, void *data); - ret = glusterd_volinfo_find (dup_volname, &volinfo); +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; + char *tmp_str_token = NULL; + char *volfile_token = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(volpath); + GF_ASSERT(volinfo); + + str_token = gf_strdup(volpath); + if (NULL == str_token) { + goto out; + } + + tmp_str_token = str_token; + + /* 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_msg(this->name, GF_LOG_ERROR, EINVAL, GD_MSG_INVALID_ENTRY, + "Invalid path: %s", volpath); + goto out; + } + + volname_token = strtok_r(NULL, "/", &save_ptr); + if (!volname_token) { + gf_msg(this->name, GF_LOG_ERROR, EINVAL, GD_MSG_INVALID_ENTRY, + "Invalid path: %s", volpath); + goto out; + } + + snap = glusterd_find_snap_by_name(snapname); + if (!snap) { + gf_msg(this->name, GF_LOG_ERROR, EINVAL, GD_MSG_SNAP_NOT_FOUND, + "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) { + gf_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_VOLINFO_GET_FAIL, + "failed to get the volinfo for the volume %s", volname_token); + + /* Get the actual volfile name. */ + volfile_token = strtok_r(NULL, "/", &save_ptr); + *volname = gf_strdup(volfile_token); + if (NULL == *volname) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRDUP_FAILED, + "Volname=%s", volfile_token, NULL); + ret = -1; + goto out; + } + + /* + * Ideally, this should succeed as volname_token now contains + * the name of the snap volume (i.e. name of the volume that + * represents the snapshot). But, if for some reason, volinfo + * for the snap volume is not found, then try to get from the + * name of the volfile. Name of the volfile is like this. + * <snap volume name>.<hostname>.<brick path>.vol + */ + ret = glusterd_snap_volinfo_find(volname_token, snap, volinfo); if (ret) { - /* Split the volume name */ - vol = strtok_r (dup_volname, ".", &tmp); - 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); - if (ret == -1) + /* Split the volume name */ + vol = strtok_r(volfile_token, ".", &save_ptr); + if (!vol) { + gf_msg(this->name, GF_LOG_ERROR, EINVAL, GD_MSG_INVALID_ENTRY, + "Invalid " + "volname (%s)", + volfile_token); 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); - ret = stat (path, &stbuf); + } + + ret = glusterd_snap_volinfo_find(vol, snap, volinfo); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SNAP_INFO_FAIL, + "Failed to " + "fetch snap volume from volname (%s)", + vol); + goto out; + } } - if ((ret == -1) && (errno == ENOENT)) { - ret = snprintf (path, path_len, "%s/vols/%s/%s-tcp.vol", - priv->workdir, volinfo->volname, volname); + } else { + /*volname_token is parent volname*/ + ret = glusterd_snap_volinfo_find_from_parent_volname(volname_token, + snap, volinfo); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SNAP_INFO_FAIL, + "Failed to " + "fetch snap volume from parent " + "volname (%s)", + volname_token); + goto out; } - ret = 1; + /* 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 (free_ptr) - GF_FREE (free_ptr); - return ret; + if (ret && NULL != *volname) { + GF_FREE(*volname); + *volname = NULL; + } + + if (tmp_str_token) + GF_FREE(tmp_str_token); + return ret; } -int -server_getspec (rpcsvc_request_t *req) +int32_t +glusterd_get_client_per_brick_volfile(glusterd_volinfo_t *volinfo, + char *filename, char *path, int path_len) { - 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; + char workdir[PATH_MAX] = { + 0, + }; + glusterd_conf_t *priv = NULL; + int32_t ret = -1; - gf_getspec_req args = {0,}; - gf_getspec_rsp rsp = {0,}; + GF_VALIDATE_OR_GOTO("glusterd", THIS, out); + priv = THIS->private; + GF_VALIDATE_OR_GOTO(THIS->name, priv, out); + GLUSTERD_GET_VOLUME_DIR(workdir, volinfo, priv); - if (!xdr_to_generic (req->msg[0], &args, (xdrproc_t)xdr_gf_getspec_req)) { - //failed to decode msg; - req->rpc_err = GARBAGE_ARGS; - goto fail; + snprintf(path, path_len, "%s/%s", workdir, filename); + + ret = 0; +out: + return ret; +} + +size_t +build_volfile_path(char *volume_id, char *path, size_t path_len, + char *trusted_str, dict_t *dict) +{ + struct stat stbuf = { + 0, + }; + int32_t ret = -1; + char *vol = NULL; + char *dup_volname = NULL; + char *save_ptr = NULL; + char *free_ptr = NULL; + char *volname = NULL; + char *volid_ptr = NULL; + char dup_volid[PATH_MAX] = { + 0, + }; + char path_prefix[PATH_MAX] = { + 0, + }; + xlator_t *this = NULL; + glusterd_volinfo_t *volinfo = NULL; + glusterd_conf_t *priv = NULL; + int32_t len = 0; + + this = THIS; + GF_ASSERT(this); + priv = this->private; + GF_ASSERT(priv); + GF_ASSERT(volume_id); + GF_ASSERT(path); + + volid_ptr = strstr(volume_id, "snapd/"); + if (volid_ptr) { + volid_ptr = strchr(volid_ptr, '/'); + if (!volid_ptr) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRCHR_FAIL, NULL); + ret = -1; + goto out; + } + volid_ptr++; + + ret = glusterd_volinfo_find(volid_ptr, &volinfo); + if (ret == -1) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL, + "Couldn't find volinfo"); + goto out; + } + glusterd_svc_build_snapd_volfile(volinfo, path, path_len); + ret = 0; + goto out; + } + + volid_ptr = strstr(volume_id, "gluster/"); + if (volid_ptr) { + volid_ptr = strchr(volid_ptr, '/'); + if (!volid_ptr) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRCHR_FAIL, NULL); + ret = -1; + goto out; + } + volid_ptr++; + + glusterd_svc_build_volfile_path(volid_ptr, priv->workdir, path, + path_len); + ret = 0; + goto out; + } + + volid_ptr = strstr(volume_id, "gfproxy-client/"); + if (volid_ptr) { + volid_ptr = strchr(volid_ptr, '/'); + if (!volid_ptr) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRCHR_FAIL, NULL); + ret = -1; + goto out; } + volid_ptr++; - volume = args.key; + ret = glusterd_volinfo_find(volid_ptr, &volinfo); + if (ret == -1) { + gf_log(this->name, GF_LOG_ERROR, "Couldn't find volinfo"); + goto out; + } - ret = build_volfile_path (volume, filename, sizeof (filename)); + glusterd_get_gfproxy_client_volfile(volinfo, path, path_len); - if (ret > 0) { - /* to allocate the proper buffer to hold the file data */ - ret = stat (filename, &stbuf); - if (ret < 0){ - gf_log ("glusterd", GF_LOG_ERROR, - "Unable to stat %s (%s)", - filename, strerror (errno)); - goto fail; - } + ret = 0; + goto out; + } + + volid_ptr = strstr(volume_id, "gfproxyd/"); + if (volid_ptr) { + volid_ptr = strchr(volid_ptr, '/'); + if (!volid_ptr) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRCHR_FAIL, NULL); + ret = -1; + goto out; + } + volid_ptr++; - spec_fd = open (filename, O_RDONLY); - if (spec_fd < 0) { - gf_log ("glusterd", GF_LOG_ERROR, - "Unable to open %s (%s)", - filename, strerror (errno)); - goto fail; - } - ret = file_len = stbuf.st_size; - } else { - op_errno = ENOENT; + ret = glusterd_volinfo_find(volid_ptr, &volinfo); + if (ret == -1) { + gf_log(this->name, GF_LOG_ERROR, "Couldn't find volinfo"); + goto out; } - if (file_len) { - rsp.spec = CALLOC (file_len+1, sizeof (char)); - if (!rsp.spec) { - ret = -1; - op_errno = ENOMEM; - goto fail; - } - ret = read (spec_fd, rsp.spec, file_len); + glusterd_svc_build_gfproxyd_volfile_path(volinfo, path, path_len); + ret = 0; + goto out; + } + + volid_ptr = strstr(volume_id, "shd/"); + if (volid_ptr) { + volid_ptr = strchr(volid_ptr, '/'); + if (!volid_ptr) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRCHR_FAIL, NULL); + ret = -1; + goto out; + } + volid_ptr++; - close (spec_fd); + ret = glusterd_volinfo_find(volid_ptr, &volinfo); + if (ret == -1) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL, + "Couldn't find volinfo for volid=%s", volid_ptr); + goto out; } - /* convert to XDR */ -fail: - rsp.op_ret = ret; + glusterd_svc_build_shd_volfile_path(volinfo, path, path_len); - if (op_errno) - rsp.op_errno = gf_errno_to_error (op_errno); - if (cookie) - rsp.op_errno = cookie; + ret = glusterd_svc_set_shd_pidfile(volinfo, dict); + if (ret == -1) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_DICT_SET_FAILED, + "Couldn't set pidfile in dict for volid=%s", volid_ptr); + goto out; + } + ret = 0; + goto out; + } - if (!rsp.spec) - rsp.spec = ""; + volid_ptr = strstr(volume_id, "/snaps/"); + if (volid_ptr) { + ret = get_snap_volname_and_volinfo(volid_ptr, &volname, &volinfo); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SNAP_INFO_FAIL, + "Failed to get snap" + " volinfo from path (%s)", + volume_id); + ret = -1; + goto out; + } - 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); + len = snprintf(path_prefix, sizeof(path_prefix), "%s/snaps/%s", + priv->workdir, volinfo->snapshot->snapname); + volid_ptr = volname; + /* this is to ensure that volname recvd from + get_snap_volname_and_volinfo is free'd */ + free_ptr = volname; + if ((len < 0) || (len >= sizeof(path_prefix))) { + ret = -1; + goto out; + } - return 0; -} + goto gotvolinfo; + } + volid_ptr = strstr(volume_id, "rebalance/"); + if (volid_ptr) { + volid_ptr = strchr(volid_ptr, '/'); + if (!volid_ptr) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRCHR_FAIL, NULL); + ret = -1; + goto out; + } + volid_ptr++; -rpcsvc_actor_t gluster_handshake_actors[] = { - [GF_HNDSK_NULL] = {"NULL", GF_HNDSK_NULL, NULL, NULL, NULL }, - [GF_HNDSK_GETSPEC] = {"GETSPEC", GF_HNDSK_GETSPEC, server_getspec, NULL, NULL }, -}; + ret = glusterd_volinfo_find(volid_ptr, &volinfo); + if (ret == -1) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL, + "Couldn't find volinfo"); + goto out; + } + glusterd_get_rebalance_volfile(volinfo, path, path_len); + ret = 0; + goto out; + } + + volid_ptr = strstr(volume_id, "client_per_brick/"); + if (volid_ptr) { + volid_ptr = strchr(volid_ptr, '/'); + if (!volid_ptr) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRCHR_FAIL, NULL); + ret = -1; + goto out; + } + volid_ptr++; + + dup_volname = gf_strdup(volid_ptr); + if (!dup_volname) { + gf_msg(this->name, GF_LOG_ERROR, ENOMEM, GD_MSG_NO_MEMORY, + "strdup failed"); + ret = -1; + goto out; + } + /* Split the volume name */ + vol = strtok_r(dup_volname, ".", &save_ptr); + if (!vol) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_SPLIT_FAIL, + "Volume name=%s", dup_volname, NULL); + ret = -1; + goto out; + } + ret = glusterd_volinfo_find(vol, &volinfo); + if (ret == -1) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL, + "Couldn't find volinfo"); + goto out; + } + ret = glusterd_get_client_per_brick_volfile(volinfo, volid_ptr, path, + path_len); + if (ret < 0) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_NO_MEMORY, + "failed to get volinfo path"); + goto out; + } -struct rpcsvc_program gluster_handshake_prog = { - .progname = "GlusterFS Handshake", - .prognum = GLUSTER_HNDSK_PROGRAM, - .progver = GLUSTER_HNDSK_VERSION, - .actors = gluster_handshake_actors, - .numactors = GF_HNDSK_MAXVALUE, -}; + ret = sys_access(path, F_OK); + goto out; + } + + if (volume_id[0] == '/') { + /* Normal behavior */ + volid_ptr = volume_id; + volid_ptr++; + + } else { + /* Bringing in NFS like behavior for mount command, */ + /* With this, one can mount a volume with below cmd */ + /* bash# mount -t glusterfs server:/volume /mnt/pnt */ + volid_ptr = volume_id; + } + + len = snprintf(path_prefix, sizeof(path_prefix), "%s/vols", priv->workdir); + if ((len < 0) || (len >= sizeof(path_prefix))) { + ret = -1; + goto out; + } + + ret = glusterd_volinfo_find(volid_ptr, &volinfo); + + if (ret) { + dup_volname = gf_strdup(volid_ptr); + if (!dup_volname) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRDUP_FAILED, + "Volume name=%s", volid_ptr, NULL); + ret = -1; + goto out; + } + /* Split the volume name */ + vol = strtok_r(dup_volname, ".", &save_ptr); + if (!vol) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_SPLIT_FAIL, + "Volume name=%s", dup_volname, NULL); + ret = -1; + goto out; + } + ret = glusterd_volinfo_find(vol, &volinfo); + if (ret) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_VOLINFO_GET_FAIL, + NULL); + goto out; + } + } + +gotvolinfo: + if (!glusterd_auth_get_username(volinfo)) + trusted_str = NULL; + + ret = snprintf(path, path_len, "%s/%s/%s.vol", path_prefix, + volinfo->volname, volid_ptr); + if (ret == -1) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_COPY_FAIL, NULL); + goto out; + } + + ret = sys_stat(path, &stbuf); + + if ((ret == -1) && (errno == ENOENT)) { + if (snprintf(dup_volid, PATH_MAX, "%s", volid_ptr) >= PATH_MAX) + goto out; + if (!strchr(dup_volid, '.')) { + switch (volinfo->transport_type) { + case GF_TRANSPORT_TCP: + strcat(dup_volid, ".tcp"); + break; + case GF_TRANSPORT_RDMA: + strcat(dup_volid, ".rdma"); + break; + case GF_TRANSPORT_BOTH_TCP_RDMA: + strcat(dup_volid, ".tcp"); + break; + default: + break; + } + } + snprintf(path, path_len, "%s/%s/%s%s-fuse.vol", path_prefix, + volinfo->volname, (trusted_str ? trusted_str : ""), dup_volid); + ret = sys_stat(path, &stbuf); + } +out: + if (dup_volname) + GF_FREE(dup_volname); + if (free_ptr) + GF_FREE(free_ptr); + return ret; +} -char *glusterd_dump_proc[GF_DUMP_MAXVALUE] = { - [GF_DUMP_NULL] = "NULL", - [GF_DUMP_DUMP] = "DUMP", -}; +/* 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. Also fetch brick_name. + */ +int32_t +glusterd_get_args_from_dict(gf_getspec_req *args, peer_info_t *peerinfo, + char **brick_name) +{ + dict_t *dict = NULL; + int client_max_op_version = 1; + int client_min_op_version = 1; + int32_t ret = -1; + xlator_t *this = NULL; + char *name = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(args); + GF_ASSERT(peerinfo); + + if (!args->xdata.xdata_len) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_INVALID_ARGUMENT, NULL); + ret = 0; + goto out; + } + + dict = dict_new(); + if (!dict) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_DICT_CREATE_FAIL, NULL); + ret = -1; + goto out; + } + + ret = dict_unserialize(args->xdata.xdata_val, args->xdata.xdata_len, &dict); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_DICT_UNSERIALIZE_FAIL, + "Failed to unserialize request dictionary"); + goto out; + } + + ret = dict_get_int32(dict, "min-op-version", &client_min_op_version); + if (ret) { + gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_DICT_GET_FAILED, + "Failed to get client-min-op-version"); + goto out; + } + + ret = dict_get_int32(dict, "max-op-version", &client_max_op_version); + if (ret) { + gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_DICT_GET_FAILED, + "Failed to get client-max-op-version"); + goto out; + } + + ret = dict_get_str(dict, "brick_name", &name); + if (ret) { + gf_msg_debug(this->name, 0, "No brick name present"); + ret = 0; + goto out; + } + *brick_name = gf_strdup(name); + if (*brick_name == NULL) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRDUP_FAILED, + "Brick_name=%s", name, NULL); + ret = -1; + goto out; + } + + gf_msg_debug(this->name, 0, "brick_name = %s", *brick_name); +out: + peerinfo->max_op_version = client_max_op_version; + peerinfo->min_op_version = client_min_op_version; -rpc_clnt_prog_t glusterd_dump_prog = { - .progname = "GLUSTERD-DUMP", - .prognum = GLUSTER_DUMP_PROGRAM, - .progver = GLUSTER_DUMP_VERSION, - .procnames = glusterd_dump_proc, -}; + if (dict) + dict_unref(dict); -static int -glusterd_event_connected_inject (glusterd_peerctx_t *peerctx) + return ret; +} + +/* Given the missed_snapinfo and snap_opinfo take the + * missed lvm snapshot + */ +int32_t +glusterd_create_missed_snap(glusterd_missed_snap_info *missed_snapinfo, + glusterd_snap_op_t *snap_opinfo) { - GF_ASSERT (peerctx); + char *device = NULL; + glusterd_conf_t *priv = NULL; + glusterd_snap_t *snap = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *volinfo = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + int32_t ret = -1; + int32_t i = 0; + uuid_t snap_uuid = { + 0, + }; + xlator_t *this = NULL; + char *mnt_device = NULL; + + this = THIS; + GF_ASSERT(this); + priv = this->private; + GF_ASSERT(priv); + GF_ASSERT(missed_snapinfo); + GF_ASSERT(snap_opinfo); + + gf_uuid_parse(missed_snapinfo->snap_uuid, snap_uuid); + + /* Find the snap-object */ + snap = glusterd_find_snap_by_id(snap_uuid); + if (!snap) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SNAP_NOT_FOUND, + "Unable to find the snap with snap_uuid %s", + missed_snapinfo->snap_uuid); + ret = -1; + goto out; + } + + /* Find the snap_vol */ + cds_list_for_each_entry(volinfo, &snap->volumes, vol_list) + { + if (!strcmp(volinfo->volname, snap_opinfo->snap_vol_id)) { + snap_vol = volinfo; + break; + } + } + + if (!snap_vol) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOL_NOT_FOUND, + "Unable to find the snap_vol(%s) " + "for snap(%s)", + snap_opinfo->snap_vol_id, snap->snapname); + ret = -1; + goto out; + } + + /* Find the missed brick in the snap volume */ + cds_list_for_each_entry(brickinfo, &snap_vol->bricks, brick_list) + { + i++; + if (i == snap_opinfo->brick_num) + break; + } + + if (brickinfo->snap_status != -1) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SNAP_STATUS_NOT_PENDING, + "The snap status of the missed " + "brick(%s) is not pending", + brickinfo->path); + goto out; + } + + /* Fetch the device path */ + mnt_device = glusterd_get_brick_mount_device(snap_opinfo->brick_path); + if (!mnt_device) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_BRICK_GET_INFO_FAIL, + "Getting device name for the" + "brick %s:%s failed", + brickinfo->hostname, snap_opinfo->brick_path); + ret = -1; + goto out; + } + + device = glusterd_build_snap_device_path(mnt_device, snap_vol->volname, + snap_opinfo->brick_num - 1); + if (!device) { + gf_msg(this->name, GF_LOG_ERROR, ENXIO, + GD_MSG_SNAP_DEVICE_NAME_GET_FAIL, + "cannot copy the snapshot " + "device name (volname: %s, snapname: %s)", + snap_vol->volname, snap->snapname); + ret = -1; + goto out; + } + if (snprintf(brickinfo->device_path, sizeof(brickinfo->device_path), "%s", + device) >= sizeof(brickinfo->device_path)) { + gf_msg(this->name, GF_LOG_ERROR, ENXIO, + GD_MSG_SNAP_DEVICE_NAME_GET_FAIL, + "cannot copy the device_path " + "(device_path: %s)", + brickinfo->device_path); + ret = -1; + goto out; + } + + /* Update the backend file-system type of snap brick in + * snap volinfo. */ + ret = glusterd_update_mntopts(snap_opinfo->brick_path, brickinfo); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_BRK_MOUNTOPTS_FAIL, + "Failed to update " + "mount options for %s brick", + brickinfo->path); + /* We should not fail snapshot operation if we fail to get + * the file-system type */ + } + + ret = glusterd_take_lvm_snapshot(brickinfo, snap_opinfo->brick_path); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_SNAPSHOT_OP_FAILED, + "Failed to take snapshot of %s", snap_opinfo->brick_path); + goto out; + } + + /* After the snapshot both the origin brick (LVM brick) and + * the snapshot brick will have the same file-system label. This + * will cause lot of problems at mount time. Therefore we must + * generate a new label for the snapshot brick + */ + ret = glusterd_update_fs_label(brickinfo); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_BRICK_SET_INFO_FAIL, + "Failed to update " + "file-system label for %s brick", + brickinfo->path); + /* Failing to update label should not cause snapshot failure. + * Currently label is updated only for XFS and ext2/ext3/ext4 + * file-system. + */ + } + + /* Create and mount the snap brick */ + ret = glusterd_snap_brick_create(snap_vol, brickinfo, + snap_opinfo->brick_num - 1, 0); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_BRICK_CREATION_FAIL, + "Failed to " + " create and mount the brick(%s) for the snap %s", + snap_opinfo->brick_path, snap_vol->snapshot->snapname); + goto out; + } + + brickinfo->snap_status = 0; + ret = glusterd_brick_start(snap_vol, brickinfo, _gf_false, _gf_false); + if (ret) { + gf_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_BRICK_DISCONNECTED, + "starting the " + "brick %s:%s for the snap %s failed", + brickinfo->hostname, brickinfo->path, snap->snapname); + goto out; + } + ret = glusterd_store_volinfo(snap_vol, GLUSTERD_VOLINFO_VER_AC_NONE); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLINFO_STORE_FAIL, + "Failed to store snapshot " + "volinfo (%s) for snap %s", + snap_vol->volname, snap->snapname); + goto out; + } - glusterd_friend_sm_event_t *event = NULL; - glusterd_probe_ctx_t *ctx = NULL; - int ret = -1; - glusterd_peerinfo_t *peerinfo = NULL; +out: + if (mnt_device) + GF_FREE(mnt_device); + if (device) + GF_FREE(device); + return ret; +} - ret = glusterd_friend_sm_new_event - (GD_FRIEND_EVENT_CONNECTED, &event); +/* Look into missed_snap_list, to see it the given brick_name, + * has any missed snap creates for the local node */ +int32_t +glusterd_take_missing_brick_snapshots(char *brick_name) +{ + char *my_node_uuid = NULL; + glusterd_conf_t *priv = NULL; + glusterd_missed_snap_info *missed_snapinfo = NULL; + glusterd_snap_op_t *snap_opinfo = NULL; + int32_t ret = -1; + gf_boolean_t update_list = _gf_false; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + priv = this->private; + GF_ASSERT(priv); + GF_ASSERT(brick_name); + + my_node_uuid = uuid_utoa(MY_UUID); + + cds_list_for_each_entry(missed_snapinfo, &priv->missed_snaps_list, + missed_snaps) + { + /* If the missed snap op is not for the local node + * then continue + */ + if (strcmp(my_node_uuid, missed_snapinfo->node_uuid)) + continue; + + cds_list_for_each_entry(snap_opinfo, &missed_snapinfo->snap_ops, + snap_ops_list) + { + /* Check if the missed snap's op is a create for + * the brick name in question + */ + if ((snap_opinfo->op == GF_SNAP_OPTION_TYPE_CREATE) && + (!strcmp(brick_name, snap_opinfo->brick_path))) { + /* Perform a snap create if the + * op is still pending + */ + if (snap_opinfo->status == GD_MISSED_SNAP_PENDING) { + ret = glusterd_create_missed_snap(missed_snapinfo, + snap_opinfo); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, + GD_MSG_MISSED_SNAP_CREATE_FAIL, + "Failed to create " + "missed snap for %s", + brick_name); + /* At this stage, we will mark + * the entry as done. Because + * of the failure other + * snapshots will not be + * affected, and neither the + * brick. Only the current snap + * brick will always remain as + * pending. + */ + } + snap_opinfo->status = GD_MISSED_SNAP_DONE; + update_list = _gf_true; + } + /* One snap-id won't have more than one missed + * create for the same brick path. Hence + * breaking in search of another missed create + * for the same brick path in the local node + */ + break; + } + } + } + if (update_list == _gf_true) { + ret = glusterd_store_update_missed_snaps(); if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to get new event"); - goto out; + gf_msg(this->name, GF_LOG_ERROR, 0, + GD_MSG_MISSED_SNAP_LIST_STORE_FAIL, + "Failed to update missed_snaps_list"); + goto out; } + } - ctx = GF_CALLOC (1, sizeof(*ctx), gf_gld_mt_probe_ctx_t); + ret = 0; +out: + return ret; +} - if (!ctx) { - ret = -1; - gf_log ("", GF_LOG_ERROR, "Memory not available"); - goto out; +/* 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_msg("glusterd", GF_LOG_INFO, ENOTSUP, GD_MSG_UNSUPPORTED_VERSION, + "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_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 *brick_name = NULL; + char *volume = NULL; + char *tmp = NULL; + 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; + xlator_t *this = NULL; + dict_t *dict = NULL; + glusterd_peerinfo_t *peer = NULL; + glusterd_conf_t *conf = NULL; + int peer_cnt = 0; + char *peer_hosts = NULL; + char *tmp_str = NULL; + char portstr[10] = { + 0, + }; + int len = 0; + + this = THIS; + GF_ASSERT(this); + + conf = this->private; + 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; + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_REQ_DECODE_FAIL, + "Failed to decode the message"); + goto fail; + } + + peerinfo = &req->trans->peerinfo; + + volume = args.key; + + if (strlen(volume) >= (NAME_MAX)) { + op_errno = EINVAL; + gf_msg(this->name, GF_LOG_ERROR, EINVAL, GD_MSG_NAME_TOO_LONG, + "volume name too long (%s)", volume); + goto fail; + } + + gf_msg(this->name, GF_LOG_INFO, 0, GD_MSG_MOUNT_REQ_RCVD, + "Received mount request for volume %s", volume); + + /* Need to strip leading '/' from volnames. This was introduced to + * support nfs style mount parameters for native gluster mount + */ + if (volume[0] == '/') + ret = snprintf(peerinfo->volname, sizeof(peerinfo->volname), "%s", + &volume[1]); + else + ret = snprintf(peerinfo->volname, sizeof(peerinfo->volname), "%s", + volume); + if (ret < 0 || ret >= sizeof(peerinfo->volname)) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VOLINFO_GET_FAIL, + "peerinfo->volname %s truncated or error occurred: " + "(ret: %d)", + peerinfo->volname, ret); + ret = -1; + goto fail; + } + + ret = glusterd_get_args_from_dict(&args, peerinfo, &brick_name); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_DICT_GET_FAILED, + "Failed to get args from dict"); + goto fail; + } + + if (!_client_supports_volume(peerinfo, &op_errno)) { + ret = -1; + goto fail; + } + + dict = dict_new(); + if (!dict) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_DICT_CREATE_FAIL, NULL); + ret = -ENOMEM; + goto fail; + } + + trans = req->trans; + /* addrstr will be empty for cli socket connections */ + ret = rpcsvc_transport_peername(trans, (char *)&addrstr, sizeof(addrstr)); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, + GD_MSG_RPC_TRANSPORT_GET_PEERNAME_FAIL, + "Failed to get the peername"); + 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. + * For unix domain socket, address will be empty. + */ + if (strlen(addrstr) == 0 || gf_is_local_addr(addrstr)) { + ret = build_volfile_path(volume, filename, sizeof(filename), + TRUSTED_PREFIX, dict); + } else { + ret = build_volfile_path(volume, filename, sizeof(filename), NULL, + dict); + } + + RCU_READ_LOCK; + cds_list_for_each_entry_rcu(peer, &conf->peers, uuid_list) + { + if (!peer->connected) + continue; + if (!peer_hosts) { + if (peer->port) { + snprintf(portstr, sizeof(portstr), "%d", peer->port); + } else { + snprintf(portstr, sizeof(portstr), "%d", GLUSTERD_DEFAULT_PORT); + } + len = strlen(peer->hostname) + strlen(portstr) + 3; + tmp_str = GF_CALLOC(1, len, gf_gld_mt_char); + snprintf(tmp_str, len, "%s%s%s%s", peer->hostname, ":", portstr, + " "); + peer_hosts = tmp_str; + } else { + len = strlen(peer_hosts) + strlen(peer->hostname) + + strlen(portstr) + 3; + tmp_str = GF_CALLOC(1, len, gf_gld_mt_char); + snprintf(tmp_str, len, "%s%s%s%s%s", peer_hosts, peer->hostname, + ":", portstr, " "); + GF_FREE(peer_hosts); + peer_hosts = tmp_str; + } + peer_cnt++; + } + RCU_READ_UNLOCK; + if (peer_cnt) { + op_ret = dict_set_str(dict, GLUSTERD_BRICK_SERVERS, peer_hosts); + if (op_ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_DICT_SET_FAILED, + "failed to set peer_host in dict"); + ret = op_ret; + goto fail; + } + } + + if (ret == 0) { + if (dict->count > 0) { + ret = dict_allocate_and_serialize(dict, &rsp.xdata.xdata_val, + &rsp.xdata.xdata_len); + if (ret) { + gf_smsg(this->name, GF_LOG_ERROR, errno, + GD_MSG_DICT_ALLOC_AND_SERL_LENGTH_GET_FAIL, NULL); + goto fail; + } } - peerinfo = peerctx->peerinfo; - ctx->hostname = gf_strdup (peerinfo->hostname); - ctx->port = peerinfo->port; - ctx->req = peerctx->args.req; + /* to allocate the proper buffer to hold the file data */ + ret = sys_stat(filename, &stbuf); + if (ret < 0) { + gf_msg("glusterd", GF_LOG_ERROR, errno, GD_MSG_FILE_OP_FAILED, + "Unable to stat %s (%s)", filename, strerror(errno)); + goto fail; + } + + spec_fd = open(filename, O_RDONLY); + if (spec_fd < 0) { + gf_msg("glusterd", GF_LOG_ERROR, errno, GD_MSG_FILE_OP_FAILED, + "Unable to open %s (%s)", filename, strerror(errno)); + goto fail; + } + ret = file_len = stbuf.st_size; + } else { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_PEER_NOT_FOUND, NULL); + op_errno = ENOENT; + goto fail; + } + + if (file_len) { + rsp.spec = CALLOC(file_len + 1, sizeof(char)); + if (!rsp.spec) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_NO_MEMORY, NULL); + ret = -1; + op_errno = ENOMEM; + goto fail; + } + ret = sys_read(spec_fd, rsp.spec, file_len); + } + + if (brick_name) { + gf_msg_debug(this->name, 0, "Look for missing snap creates for %s", + brick_name); + op_ret = glusterd_take_missing_brick_snapshots(brick_name); + if (op_ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_MISSED_SNAP_CREATE_FAIL, + "Failed to take missing brick snapshots"); + ret = -1; + goto fail; + } + } + /* convert to XDR */ +fail: + if (spec_fd >= 0) + sys_close(spec_fd); + + GF_FREE(brick_name); + + rsp.op_ret = ret; + if (rsp.op_ret < 0) + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_MOUNT_REQ_FAIL, + "Failed to mount the volume"); - event->peerinfo = peerinfo; - event->ctx = ctx; + if (op_errno) + rsp.op_errno = gf_errno_to_error(op_errno); - ret = glusterd_friend_sm_inject_event (event); + if (!rsp.spec) + rsp.spec = strdup(""); + + glusterd_submit_reply(req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_getspec_rsp); + free(args.key); // malloced by xdr + free(rsp.spec); + + if (peer_hosts) + GF_FREE(peer_hosts); + if (dict) + dict_unref(dict); + + if (args.xdata.xdata_val) + free(args.xdata.xdata_val); + + if (rsp.xdata.xdata_val) + GF_FREE(rsp.xdata.xdata_val); + + 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; + 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; + gf_smsg("glusterd", GF_LOG_ERROR, errno, GD_MSG_GARBAGE_ARGS, NULL); + goto fail; + } + + if (args.dict.dict_len) { + dict = dict_new(); + if (!dict) { + gf_smsg("glusterd", GF_LOG_ERROR, errno, GD_MSG_DICT_CREATE_FAIL, + NULL); + return ret; + } + ret = dict_unserialize(args.dict.dict_val, args.dict.dict_len, &dict); if (ret) { - gf_log ("glusterd", GF_LOG_ERROR, "Unable to inject " - "EVENT_CONNECTED ret = %d", ret); - goto out; + gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_DICT_UNSERIALIZE_FAIL, + "Failed to unserialize req"); + goto fail; } + } + + switch (args.op) { + case GF_EN_DEFRAG_STATUS: + gf_msg("glusterd", GF_LOG_INFO, 0, GD_MSG_DEFRAG_STATUS_UPDATED, + "received defrag status updated"); + if (dict) { + glusterd_defrag_event_notify_handle(dict); + need_rsp = _gf_false; + } + break; + default: + gf_msg("glusterd", GF_LOG_ERROR, EINVAL, GD_MSG_OP_UNSUPPORTED, + "Unknown op received in event " + "notify"); + gf_event(EVENT_NOTIFY_UNKNOWN_OP, "op=%d", args.op); + ret = -1; + break; + } + +fail: + rsp.op_ret = ret; + + 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_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_OP_VERSION_MISMATCH, + "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) && + !cds_list_empty(&conf->volumes)) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_OP_VERS_ADJUST_FAIL, + "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: - gf_log ("", GF_LOG_DEBUG, "returning %d", ret); - return ret; + return ret; +} + +/* Validate if glusterd can serve the management handshake request + * + * Requests are allowed if, + * - glusterd has no peers & no volumes, or + * - the request came from a known peer + * A known peer is identified using the following steps + * - the dict is checked for a peer uuid, which if present is matched with the + * peer list, else + * - the incoming request address is matched with the peer list + */ +gf_boolean_t +gd_validate_mgmt_hndsk_req(rpcsvc_request_t *req, dict_t *dict) +{ + int ret = -1; + char hostname[UNIX_PATH_MAX + 1] = { + 0, + }; + glusterd_peerinfo_t *peer = NULL; + xlator_t *this = NULL; + char *uuid_str = NULL; + uuid_t peer_uuid = { + 0, + }; + + this = THIS; + GF_ASSERT(this); + + if (!glusterd_have_peers() && !glusterd_have_volumes()) + return _gf_true; + + ret = dict_get_str(dict, GD_PEER_ID_KEY, &uuid_str); + /* Try to match uuid only if available, don't fail as older peers will + * not send a uuid + */ + if (!ret) { + gf_uuid_parse(uuid_str, peer_uuid); + RCU_READ_LOCK; + ret = (glusterd_peerinfo_find(peer_uuid, NULL) != NULL); + RCU_READ_UNLOCK; + if (ret) + return _gf_true; + } + + /* If you cannot get the hostname, you cannot authenticate */ + ret = glusterd_remote_hostname_get(req, hostname, sizeof(hostname)); + if (ret) + return _gf_false; + + /* If peer object is not found it indicates that request is from an + * unknown peer, if its found, validate whether its uuid is also + * available in the peerinfo list. There could be a case where hostname + * is available in the peerinfo list but the uuid has changed of the + * node due to a reinstall, in that case the validation should fail! + */ + RCU_READ_LOCK; + if (!uuid_str) { + ret = (glusterd_peerinfo_find(NULL, hostname) == NULL); + } else { + peer = glusterd_peerinfo_find(NULL, hostname); + if (!peer) { + ret = -1; + } else if (peer && glusterd_peerinfo_find(peer_uuid, NULL) != NULL) { + ret = 0; + } else { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_HANDSHAKE_REQ_REJECTED, + "Request from " + "peer %s has an entry in peerinfo, but uuid " + "does not match", + req->trans->peerinfo.identifier); + ret = -1; + } + } + RCU_READ_UNLOCK; + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_HANDSHAKE_REQ_REJECTED, + "Rejecting management " + "handshake request from unknown peer %s", + req->trans->peerinfo.identifier); + gf_event(EVENT_PEER_REJECT, "peer=%s", req->trans->peerinfo.identifier); + return _gf_false; + } + + return _gf_true; } int -glusterd_set_clnt_mgmt_program (glusterd_peerinfo_t *peerinfo, - gf_prog_detail *prog) +__glusterd_mgmt_hndsk_versions(rpcsvc_request_t *req) { - gf_prog_detail *trav = NULL; - int ret = -1; + 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, + }; + dict_t *args_dict = NULL; + + 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; + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_GARBAGE_ARGS, NULL); + goto out; + } + + GF_PROTOCOL_DICT_UNSERIALIZE(this, args_dict, args.hndsk.hndsk_val, + (args.hndsk.hndsk_len), ret, op_errno, out); + + /* Check if we can service the request */ + if (!gd_validate_mgmt_hndsk_req(req, args_dict)) { + ret = -1; + goto out; + } + + dict = dict_new(); + if (!dict) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_DICT_CREATE_FAIL, NULL); + goto out; + } + + ret = dict_set_int32(dict, GD_OP_VERSION_KEY, conf->op_version); + if (ret) { + gf_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_DICT_SET_FAILED, + "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_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_DICT_SET_FAILED, + "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_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_DICT_SET_FAILED, + "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: - if (!peerinfo || !prog) - goto out; + rsp.op_ret = ret; + rsp.op_errno = op_errno; - trav = prog; + glusterd_submit_reply(req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_mgmt_hndsk_rsp); - while (trav) { - /* Select 'programs' */ - 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; - } + ret = 0; - if (peerinfo->mgmt) { - 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); - } + 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); + + if (args_dict) + dict_unref(args_dict); + + 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; + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_GARBAGE_ARGS, NULL); + 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_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_DICT_GET_FAILED, + "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_msg(this->name, GF_LOG_INFO, 0, GD_MSG_VERS_INFO, + "using the op-version %d", peer_op_version); + conf->op_version = peer_op_version; + ret = glusterd_store_global_info(this); + if (ret) + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_GLOBAL_OP_VERSION_SET_FAIL, + "Failed to store op-version"); out: - return ret; + 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_peer_dump_version_cbk (struct rpc_req *req, struct iovec *iov, - int count, void *myframe) +glusterd_mgmt_hndsk_versions_ack(rpcsvc_request_t *req) { - int ret = -1; - gf_dump_rsp rsp = {0,}; - xlator_t *this = NULL; - gf_prog_detail *trav = NULL; - gf_prog_detail *next = NULL; - call_frame_t *frame = NULL; - glusterd_peerinfo_t *peerinfo = NULL; - glusterd_peerctx_t *peerctx = NULL; - - this = THIS; - 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"); - goto out; + return glusterd_big_locked_handler(req, __glusterd_mgmt_hndsk_versions_ack); +} + +int +__server_get_volume_info(rpcsvc_request_t *req) +{ + int ret = -1; + int32_t op_errno = ENOENT; + gf_get_volume_info_req vol_info_req = {{ + 0, + }}; + gf_get_volume_info_rsp vol_info_rsp = { + 0, + }; + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + dict_t *dict = NULL; + dict_t *dict_rsp = NULL; + char *volume_id_str = NULL; + int32_t flags = 0; + + xlator_t *this = THIS; + GF_ASSERT(this); + + ret = xdr_to_generic(req->msg[0], &vol_info_req, + (xdrproc_t)xdr_gf_get_volume_info_req); + if (ret < 0) { + /* failed to decode msg */ + req->rpc_err = GARBAGE_ARGS; + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_GARBAGE_ARGS, NULL); + goto out; + } + gf_smsg(this->name, GF_LOG_INFO, 0, GD_MSG_VOL_INFO_REQ_RECVD, NULL); + + if (vol_info_req.dict.dict_len) { + /* Unserialize the dictionary */ + dict = dict_new(); + if (!dict) { + gf_smsg(this->name, GF_LOG_WARNING, ENOMEM, GD_MSG_DICT_CREATE_FAIL, + NULL); + op_errno = ENOMEM; + ret = -1; + goto out; } - ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_dump_rsp); + ret = dict_unserialize(vol_info_req.dict.dict_val, + vol_info_req.dict.dict_len, &dict); if (ret < 0) { - gf_log ("", GF_LOG_ERROR, "failed to decode XDR"); - goto out; + gf_smsg(this->name, GF_LOG_ERROR, 0, GD_MSG_DICT_UNSERIALIZE_FAIL, + NULL); + op_errno = -ret; + ret = -1; + goto out; + } else { + dict->extra_stdfree = vol_info_req.dict.dict_val; } - if (-1 == rsp.op_ret) { - gf_log (frame->this->name, GF_LOG_ERROR, - "failed to get the 'versions' from remote server"); - goto out; + } + + ret = dict_get_int32(dict, "flags", &flags); + if (ret) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_DICT_GET_FAILED, + "Key=flags", NULL); + op_errno = -ret; + ret = -1; + goto out; + } + + if (!flags) { + /* Nothing to query about. Just return success */ + gf_smsg(this->name, GF_LOG_ERROR, 0, GD_MSG_NO_FLAG_SET, NULL); + ret = 0; + goto out; + } + + ret = dict_get_str(dict, "volname", &volname); + if (ret) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_DICT_GET_FAILED, + "Key=volname", NULL); + op_errno = EINVAL; + ret = -1; + goto out; + } + + ret = glusterd_volinfo_find(volname, &volinfo); + if (ret) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_VOLINFO_GET_FAIL, + "Volname=%s", volname, NULL); + op_errno = EINVAL; + ret = -1; + goto out; + } + + if (flags & (int32_t)GF_GET_VOLUME_UUID) { + volume_id_str = gf_strdup(uuid_utoa(volinfo->volume_id)); + if (!volume_id_str) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_STRDUP_FAILED, + NULL); + op_errno = ENOMEM; + ret = -1; + goto out; } - /* Make sure we assign the proper program to peer */ - ret = glusterd_set_clnt_mgmt_program (peerinfo, rsp.prog); + dict_rsp = dict_new(); + if (!dict_rsp) { + gf_smsg(this->name, GF_LOG_WARNING, ENOMEM, GD_MSG_DICT_CREATE_FAIL, + NULL); + op_errno = ENOMEM; + GF_FREE(volume_id_str); + ret = -1; + goto out; + } + ret = dict_set_dynstr(dict_rsp, "volume_id", volume_id_str); if (ret) { - gf_log ("", GF_LOG_WARNING, "failed to set the mgmt program"); - goto out; + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_DICT_SET_FAILED, + "Key=volume_id", NULL); + op_errno = -ret; + ret = -1; + goto out; } + } + ret = dict_allocate_and_serialize(dict_rsp, &vol_info_rsp.dict.dict_val, + &vol_info_rsp.dict.dict_len); + if (ret) { + gf_smsg(this->name, GF_LOG_ERROR, errno, + GD_MSG_DICT_ALLOC_AND_SERL_LENGTH_GET_FAIL, NULL); + op_errno = -ret; + ret = -1; + goto out; + } - ret = default_notify (this, GF_EVENT_CHILD_UP, NULL); +out: + vol_info_rsp.op_ret = ret; + vol_info_rsp.op_errno = op_errno; + vol_info_rsp.op_errstr = ""; + glusterd_submit_reply(req, &vol_info_rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_get_volume_info_rsp); + ret = 0; + + if (dict) { + dict_unref(dict); + } + + if (dict_rsp) { + dict_unref(dict_rsp); + } + + if (vol_info_rsp.dict.dict_val) { + GF_FREE(vol_info_rsp.dict.dict_val); + } + return ret; +} + +int +server_get_volume_info(rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler(req, __server_get_volume_info); +} + +/* + * glusterd function to get the list of snapshot names and uuids + */ +int +__server_get_snap_info(rpcsvc_request_t *req) +{ + int ret = -1; + int op_errno = ENOENT; + gf_getsnap_name_uuid_req snap_info_req = {{ + 0, + }}; + gf_getsnap_name_uuid_rsp snap_info_rsp = { + 0, + }; + dict_t *dict = NULL; + dict_t *dict_rsp = NULL; + char *volname = NULL; + + GF_ASSERT(req); + + ret = xdr_to_generic(req->msg[0], &snap_info_req, + (xdrproc_t)xdr_gf_getsnap_name_uuid_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_REQ_DECODE_FAIL, + "Failed to decode management handshake response"); + goto out; + } + + if (snap_info_req.dict.dict_len) { + dict = dict_new(); + if (!dict) { + gf_smsg("glusterd", GF_LOG_WARNING, ENOMEM, GD_MSG_DICT_CREATE_FAIL, + NULL); + op_errno = ENOMEM; + ret = -1; + goto out; + } - 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; + ret = dict_unserialize(snap_info_req.dict.dict_val, + snap_info_req.dict.dict_len, &dict); + if (ret < 0) { + gf_msg("glusterd", GF_LOG_ERROR, EINVAL, + GD_MSG_DICT_UNSERIALIZE_FAIL, + "Failed to unserialize dictionary"); + op_errno = EINVAL; + ret = -1; + goto out; } else { - gf_log ("", GF_LOG_WARNING, "unknown mode %d", - peerctx->args.mode); + dict->extra_stdfree = snap_info_req.dict.dict_val; } + } + + ret = dict_get_str(dict, "volname", &volname); + if (ret) { + op_errno = EINVAL; + gf_msg("glusterd", GF_LOG_ERROR, EINVAL, GD_MSG_DICT_GET_FAILED, + "Failed to retrieve volname"); + ret = -1; + goto out; + } + + dict_rsp = dict_new(); + if (!dict_rsp) { + gf_smsg("glusterd", GF_LOG_WARNING, ENOMEM, GD_MSG_DICT_CREATE_FAIL, + NULL); + op_errno = ENOMEM; + ret = -1; + goto out; + } + + ret = glusterd_snapshot_get_volnames_uuids(dict_rsp, volname, + &snap_info_rsp); + + if (ret) { + gf_msg("glusterd", GF_LOG_ERROR, EINVAL, GD_MSG_VOL_NOT_FOUND, + "Error getting snapshot volume names and uuids : %s", volname); + op_errno = EINVAL; + } - glusterd_friend_sm (); - glusterd_op_sm (); +out: + snap_info_rsp.op_ret = ret; + snap_info_rsp.op_errno = op_errno; + snap_info_rsp.op_errstr = ""; + glusterd_submit_reply(req, &snap_info_rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gf_getsnap_name_uuid_rsp); + + if (dict) { + dict_unref(dict); + } + + if (dict_rsp) { + dict_unref(dict_rsp); + } + + if (snap_info_rsp.dict.dict_val) { + GF_FREE(snap_info_rsp.dict.dict_val); + } + + return 0; +} + +int +server_get_snap_info(rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler(req, __server_get_snap_info); +} + +static rpcsvc_actor_t gluster_handshake_actors[GF_HNDSK_MAXVALUE] = { + [GF_HNDSK_NULL] = {"NULL", NULL, NULL, GF_HNDSK_NULL, DRC_NA, 0}, + [GF_HNDSK_GETSPEC] = {"GETSPEC", server_getspec, NULL, GF_HNDSK_GETSPEC, + DRC_NA, 0}, + [GF_HNDSK_EVENT_NOTIFY] = {"EVENTNOTIFY", server_event_notify, NULL, + GF_HNDSK_EVENT_NOTIFY, DRC_NA, 0}, + [GF_HNDSK_GET_VOLUME_INFO] = {"GETVOLUMEINFO", server_get_volume_info, NULL, + GF_HNDSK_GET_VOLUME_INFO, DRC_NA, 0}, + [GF_HNDSK_GET_SNAPSHOT_INFO] = {"GETSNAPINFO", server_get_snap_info, NULL, + GF_HNDSK_GET_SNAPSHOT_INFO, DRC_NA, 0}, +}; + +struct rpcsvc_program gluster_handshake_prog = { + .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 */ +static rpcsvc_actor_t gluster_cli_getspec_actors[GF_HNDSK_MAXVALUE] = { + [GF_HNDSK_GETSPEC] = {"GETSPEC", server_getspec, NULL, GF_HNDSK_GETSPEC, + DRC_NA, 0}, +}; + +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, +}; + +static char *glusterd_dump_proc[GF_DUMP_MAXVALUE] = { + [GF_DUMP_NULL] = "NULL", + [GF_DUMP_DUMP] = "DUMP", + [GF_DUMP_PING] = "PING", +}; + +static rpc_clnt_prog_t glusterd_dump_prog = { + .progname = "GLUSTERD-DUMP", + .prognum = GLUSTER_DUMP_PROGRAM, + .progver = GLUSTER_DUMP_VERSION, + .procnames = glusterd_dump_proc, +}; + +static rpcsvc_actor_t glusterd_mgmt_hndsk_actors[GD_MGMT_HNDSK_MAXVALUE] = { + [GD_MGMT_HNDSK_NULL] = {"NULL", NULL, NULL, GD_MGMT_HNDSK_NULL, DRC_NA, 0}, + [GD_MGMT_HNDSK_VERSIONS] = {"MGMT-VERS", glusterd_mgmt_hndsk_versions, NULL, + GD_MGMT_HNDSK_VERSIONS, DRC_NA, 0}, + [GD_MGMT_HNDSK_VERSIONS_ACK] = {"MGMT-VERS-ACK", + glusterd_mgmt_hndsk_versions_ack, NULL, + GD_MGMT_HNDSK_VERSIONS_ACK, DRC_NA, 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, +}; + +static 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", +}; + +static 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) +{ + GF_ASSERT(peerctx); + + glusterd_friend_sm_event_t *event = NULL; + glusterd_probe_ctx_t *ctx = NULL; + int ret = -1; + glusterd_peerinfo_t *peerinfo = NULL; + + ret = glusterd_friend_sm_new_event(GD_FRIEND_EVENT_CONNECTED, &event); + + if (ret) { + gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_EVENT_NEW_GET_FAIL, + "Unable to get new event"); + goto out; + } + + ctx = GF_CALLOC(1, sizeof(*ctx), gf_gld_mt_probe_ctx_t); + + if (!ctx) { + ret = -1; + gf_msg("glusterd", GF_LOG_ERROR, ENOMEM, GD_MSG_NO_MEMORY, + "Memory not available"); + goto out; + } + + RCU_READ_LOCK; + + peerinfo = glusterd_peerinfo_find_by_generation(peerctx->peerinfo_gen); + if (!peerinfo) { + RCU_READ_UNLOCK; + ret = -1; + gf_msg(THIS->name, GF_LOG_ERROR, 0, GD_MSG_PEER_NOT_FOUND, + "Could not find peer %s(%s)", peerctx->peername, + uuid_utoa(peerctx->peerid)); + GF_FREE(ctx); + goto out; + } + ctx->hostname = gf_strdup(peerinfo->hostname); + ctx->port = peerinfo->port; + ctx->req = peerctx->args.req; + ctx->dict = peerctx->args.dict; + + event->peername = gf_strdup(peerinfo->hostname); + gf_uuid_copy(event->peerid, peerinfo->uuid); + event->ctx = ctx; + + ret = glusterd_friend_sm_inject_event(event); + + RCU_READ_UNLOCK; + + if (ret) + gf_msg("glusterd", GF_LOG_ERROR, 0, GD_MSG_EVENT_INJECT_FAIL, + "Unable to inject " + "EVENT_CONNECTED ret = %d", + ret); - ret = 0; out: + gf_msg_debug("glusterd", 0, "returning %d", ret); + return ret; +} - /* don't use GF_FREE, buffer was allocated by libc */ - if (rsp.prog) { - trav = rsp.prog; - while (trav) { - next = trav->next; - free (trav->progname); - free (trav); - trav = next; - } - } +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) { + gf_smsg("glusterd", GF_LOG_WARNING, ENOMEM, GD_MSG_DICT_CREATE_FAIL, + NULL); + goto out; + } + + if (!this) { + gf_smsg("glusterd", GF_LOG_ERROR, errno, GD_MSG_XLATOR_NOT_DEFINED, + NULL); + goto out; + } + + if (!peerinfo) { + gf_smsg("glusterd", GF_LOG_ERROR, errno, GD_MSG_INVALID_ARGUMENT, NULL); + goto out; + } + + conf = this->private; + + ret = dict_get_int32(dict, GD_OP_VERSION_KEY, &peer_op_version); + if (ret) { + gf_smsg("glusterd", GF_LOG_ERROR, errno, GD_MSG_DICT_GET_FAILED, + "Key=%s", GD_OP_VERSION_KEY, NULL); + goto out; + } + + ret = dict_get_int32(dict, GD_MAX_OP_VERSION_KEY, &peer_max_op_version); + if (ret) { + gf_smsg("glusterd", GF_LOG_ERROR, errno, GD_MSG_DICT_GET_FAILED, + "Key=%s", GD_MAX_OP_VERSION_KEY, NULL); + goto out; + } + + ret = dict_get_int32(dict, GD_MIN_OP_VERSION_KEY, &peer_min_op_version); + if (ret) { + gf_smsg("glusterd", GF_LOG_ERROR, errno, GD_MSG_DICT_GET_FAILED, + "Key=%s", GD_MIN_OP_VERSION_KEY, NULL); + 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: + if (peerinfo) + gf_msg_debug((this ? this->name : "glusterd"), 0, "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; + 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[64] = { + 0, + }; + + this = THIS; + frame = myframe; + peerctx = frame->local; + + RCU_READ_LOCK; + peerinfo = glusterd_peerinfo_find_by_generation(peerctx->peerinfo_gen); + if (!peerinfo) { + gf_msg_debug(this->name, 0, "Could not find peer %s(%s)", + peerctx->peername, uuid_utoa(peerctx->peerid)); + ret = -1; + goto out; + } + + if (-1 == req->rpc_status) { + snprintf(msg, sizeof(msg), + "Error through RPC layer, retry again later"); + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_RPC_LAYER_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_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_REQ_DECODE_FAIL, "%s", msg); + peerctx->errstr = gf_strdup(msg); + goto out; + } + + if (-1 == rsp.op_ret) { + ret = -1; + snprintf(msg, sizeof(msg), + "Failed to get handshake ack from remote server"); + gf_msg(frame->this->name, GF_LOG_ERROR, 0, GD_MSG_NO_HANDSHAKE_ACK, + "%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) { + (void)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_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_UNKNOWN_MODE, + "unknown mode %d", peerctx->args.mode); + } + + ret = 0; +out: + + if (ret != 0 && peerinfo) + rpc_transport_disconnect(peerinfo->rpc->conn.trans, _gf_false); + + RCU_READ_UNLOCK; + + frame->local = NULL; + STACK_DESTROY(frame->root); + + if (rsp.hndsk.hndsk_val) + free(rsp.hndsk.hndsk_val); + + glusterd_friend_sm(); + + 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[64] = { + 0, + }; + + this = THIS; + conf = this->private; + frame = myframe; + peerctx = frame->local; + + RCU_READ_LOCK; + + peerinfo = glusterd_peerinfo_find_by_generation(peerctx->peerinfo_gen); + if (!peerinfo) { + ret = -1; + gf_msg_debug(this->name, 0, "Could not find peer %s(%s)", + peerctx->peername, uuid_utoa(peerctx->peerid)); + goto out; + } + + if (-1 == req->rpc_status) { + ret = -1; + snprintf(msg, sizeof(msg), + "Error through RPC layer, retry again later"); + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_RPC_LAYER_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_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_REQ_DECODE_FAIL, "%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_msg(this->name, GF_LOG_ERROR, op_errno, GD_MSG_VERS_GET_FAIL, + "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_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_OP_VERSION_MISMATCH, + "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_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_DICT_SET_FAILED, + "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( + 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); + STACK_DESTROY(frame->root); + if (peerinfo) + rpc_transport_disconnect(peerinfo->rpc->conn.trans, _gf_false); + } - if (ret != 0) - rpc_transport_disconnect (peerinfo->rpc->conn.trans); + RCU_READ_UNLOCK; - return 0; + 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_peer_handshake (xlator_t *this, struct rpc_clnt *rpc, - glusterd_peerctx_t *peerctx) +glusterd_mgmt_handshake(xlator_t *this, glusterd_peerctx_t *peerctx) { - call_frame_t *frame = NULL; - gf_dump_req req = {0,}; - int ret = -1; + call_frame_t *frame = NULL; + gf_mgmt_hndsk_req req = { + { + 0, + }, + }; + glusterd_peerinfo_t *peerinfo = NULL; + dict_t *req_dict = NULL; + int ret = -1; + + frame = create_frame(this, this->ctx->pool); + if (!frame) { + gf_smsg("glusterd", GF_LOG_WARNING, errno, GD_MSG_FRAME_CREATE_FAIL, + NULL); + goto out; + } + + frame->local = peerctx; + + req_dict = dict_new(); + if (!req_dict) { + gf_smsg("glusterd", GF_LOG_WARNING, ENOMEM, GD_MSG_DICT_CREATE_FAIL, + NULL); + goto out; + } + + ret = dict_set_dynstr(req_dict, GD_PEER_ID_KEY, + gf_strdup(uuid_utoa(MY_UUID))); + if (ret) { + gf_msg(this->name, GF_LOG_ERROR, errno, GD_MSG_DICT_SET_FAILED, + "failed to set peer ID in dict"); + goto out; + } + + GF_PROTOCOL_DICT_SERIALIZE(this, req_dict, (&req.hndsk.hndsk_val), + req.hndsk.hndsk_len, ret, out); + + RCU_READ_LOCK; + + peerinfo = glusterd_peerinfo_find_by_generation(peerctx->peerinfo_gen); + if (!peerinfo) { + RCU_READ_UNLOCK; + gf_msg_debug(THIS->name, 0, "Could not find peer %s(%s)", + peerctx->peername, uuid_utoa(peerctx->peerid)); + goto out; + } + + ret = glusterd_submit_request( + 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); + + RCU_READ_UNLOCK; + + ret = 0; - frame = create_frame (this, this->ctx->pool); - if (!frame) - goto out; +out: + if (req_dict) + dict_unref(req_dict); + + if (ret && frame) + STACK_DESTROY(frame->root); - frame->local = peerctx; + return ret; +} - req.gfs_id = 0xcafe; +int +glusterd_set_clnt_mgmt_program(glusterd_peerinfo_t *peerinfo, + gf_prog_detail *prog) +{ + gf_prog_detail *trav = NULL; + int ret = -1; + + if (!peerinfo || !prog) + goto out; + + trav = prog; + + while (trav) { + ret = -1; + if ((gd_mgmt_prog.prognum == trav->prognum) && + (gd_mgmt_prog.progver == trav->progver)) { + peerinfo->mgmt = &gd_mgmt_prog; + ret = 0; + } - ret = glusterd_submit_request (peerctx->peerinfo->rpc, &req, frame, - &glusterd_dump_prog, GF_DUMP_DUMP, - NULL, this, - glusterd_peer_dump_version_cbk, - (xdrproc_t)xdr_gf_dump_req); + if ((gd_peer_prog.prognum == trav->prognum) && + (gd_peer_prog.progver == trav->progver)) { + peerinfo->peer = &gd_peer_prog; + ret = 0; + } + + if (ret) { + gf_msg_debug("glusterd", 0, + "%s (%" PRId64 ":%" PRId64 ") not supported", + trav->progname, trav->prognum, trav->progver); + } + + trav = trav->next; + } + + if (peerinfo->mgmt) { + gf_msg("glusterd", GF_LOG_INFO, 0, GD_MSG_VERS_INFO, + "Using Program %s, Num (%d), Version (%d)", + peerinfo->mgmt->progname, peerinfo->mgmt->prognum, + peerinfo->mgmt->progver); + } + + if (peerinfo->peer) { + gf_msg("glusterd", GF_LOG_INFO, 0, GD_MSG_VERS_INFO, + "Using Program %s, Num (%d), Version (%d)", + peerinfo->peer->progname, peerinfo->peer->prognum, + peerinfo->peer->progver); + } + + if (peerinfo->mgmt_v3) { + gf_msg("glusterd", GF_LOG_INFO, 0, GD_MSG_VERS_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, + int count, void *myframe) +{ + int ret = -1; + gf_dump_rsp rsp = { + 0, + }; + xlator_t *this = NULL; + gf_prog_detail *trav = NULL; + gf_prog_detail *next = NULL; + 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; + + RCU_READ_LOCK; + + peerinfo = glusterd_peerinfo_find_by_generation(peerctx->peerinfo_gen); + if (!peerinfo) { + gf_msg_debug(this->name, 0, "Couldn't find peer %s(%s)", + peerctx->peername, uuid_utoa(peerctx->peerid)); + goto out; + } + + if (-1 == req->rpc_status) { + snprintf(msg, sizeof(msg), + "Error through RPC layer, retry again later"); + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_RPC_LAYER_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) { + snprintf(msg, sizeof(msg), "Failed to decode XDR"); + gf_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_REQ_DECODE_FAIL, "%s", msg); + peerctx->errstr = gf_strdup(msg); + goto out; + } + if (-1 == rsp.op_ret) { + snprintf(msg, sizeof(msg), + "Failed to get the 'versions' from remote server"); + gf_msg(frame->this->name, GF_LOG_ERROR, 0, GD_MSG_VERS_GET_FAIL, "%s", + msg); + peerctx->errstr = gf_strdup(msg); + goto out; + } + + if (_mgmt_hndsk_prog_present(rsp.prog)) { + gf_msg_debug(this->name, 0, + "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_msg(this->name, GF_LOG_ERROR, 0, GD_MSG_VERSION_UNSUPPORTED, "%s", + msg); + goto out; + } + + /* Make sure we assign the proper program to peer */ + ret = glusterd_set_clnt_mgmt_program(peerinfo, rsp.prog); + if (ret) { + gf_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_MGMT_PGM_SET_FAIL, + "failed to set the mgmt program"); + goto out; + } + + ret = default_notify(this, GF_EVENT_CHILD_UP, NULL); + + if (GD_MODE_ON == peerctx->args.mode) { + (void)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_msg(this->name, GF_LOG_WARNING, 0, GD_MSG_UNKNOWN_MODE, + "unknown mode %d", peerctx->args.mode); + } + + ret = 0; + +out: + if (ret != 0 && peerinfo) + rpc_transport_disconnect(peerinfo->rpc->conn.trans, _gf_false); + + RCU_READ_UNLOCK; + + glusterd_friend_sm(); + glusterd_op_sm(); + + /* don't use GF_FREE, buffer was allocated by libc */ + if (rsp.prog) { + trav = rsp.prog; + while (trav) { + next = trav->next; + free(trav->progname); + free(trav); + trav = next; + } + } + + frame->local = NULL; + STACK_DESTROY(frame->root); + + return 0; +} + +int +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, + }; + glusterd_peerinfo_t *peerinfo = NULL; + int ret = -1; + + frame = create_frame(this, this->ctx->pool); + if (!frame) { + gf_smsg(this->name, GF_LOG_WARNING, errno, GD_MSG_FRAME_CREATE_FAIL, + NULL); + goto out; + } + + frame->local = peerctx; + if (!peerctx) { + gf_smsg(this->name, GF_LOG_ERROR, errno, GD_MSG_INVALID_ARGUMENT, NULL); + goto out; + } + + RCU_READ_LOCK; + + peerinfo = glusterd_peerinfo_find_by_generation(peerctx->peerinfo_gen); + if (!peerinfo) { + RCU_READ_UNLOCK; + gf_msg_debug(this->name, 0, "Couldn't find peer %s(%s)", + peerctx->peername, uuid_utoa(peerctx->peerid)); + goto out; + } + + req.gfs_id = 0xcafe; + + ret = glusterd_submit_request( + peerinfo->rpc, &req, frame, &glusterd_dump_prog, GF_DUMP_DUMP, NULL, + this, glusterd_peer_dump_version_cbk, (xdrproc_t)xdr_gf_dump_req); + + RCU_READ_UNLOCK; out: - return ret; + if (ret && frame) + STACK_DESTROY(frame->root); + + return ret; } |