diff options
Diffstat (limited to 'xlators/cluster/afr/src/afr.c')
| -rw-r--r-- | xlators/cluster/afr/src/afr.c | 4015 |
1 files changed, 1187 insertions, 2828 deletions
diff --git a/xlators/cluster/afr/src/afr.c b/xlators/cluster/afr/src/afr.c index 33cd70cd478..df7366f0a65 100644 --- a/xlators/cluster/afr/src/afr.c +++ b/xlators/cluster/afr/src/afr.c @@ -1,20 +1,11 @@ /* - Copyright (c) 2007-2009 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) 2008-2012 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. */ #include <libgen.h> @@ -24,2962 +15,1330 @@ #include <stdlib.h> #include <signal.h> -#ifndef _CONFIG_H -#define _CONFIG_H -#include "config.h" -#endif - -#include "glusterfs.h" -#include "afr.h" -#include "dict.h" -#include "xlator.h" -#include "hashfn.h" -#include "logging.h" -#include "stack.h" -#include "list.h" -#include "call-stub.h" -#include "defaults.h" -#include "common-utils.h" -#include "compat-errno.h" -#include "compat.h" -#include "byte-order.h" -#include "statedump.h" - -#include "fd.h" - -#include "afr-inode-read.h" -#include "afr-inode-write.h" -#include "afr-dir-read.h" -#include "afr-dir-write.h" -#include "afr-transaction.h" - -#include "afr-self-heal.h" - -#define AFR_ICTX_OPENDIR_DONE_MASK 0x0000000200000000ULL -#define AFR_ICTX_SPLIT_BRAIN_MASK 0x0000000100000000ULL -#define AFR_ICTX_READ_CHILD_MASK 0x00000000FFFFFFFFULL - - -uint64_t -afr_is_split_brain (xlator_t *this, inode_t *inode) -{ - int ret = 0; - - uint64_t ctx = 0; - uint64_t split_brain = 0; - - VALIDATE_OR_GOTO (inode, out); - - LOCK (&inode->lock); - { - ret = __inode_ctx_get (inode, this, &ctx); - - if (ret < 0) - goto unlock; - - split_brain = ctx & AFR_ICTX_SPLIT_BRAIN_MASK; - } -unlock: - UNLOCK (&inode->lock); - -out: - return split_brain; -} - - -void -afr_set_split_brain (xlator_t *this, inode_t *inode) -{ - uint64_t ctx = 0; - int ret = 0; - - VALIDATE_OR_GOTO (inode, out); - - LOCK (&inode->lock); - { - ret = __inode_ctx_get (inode, this, &ctx); - - if (ret < 0) { - ctx = 0; - } - - ctx = (~AFR_ICTX_SPLIT_BRAIN_MASK & ctx) - | (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_SPLIT_BRAIN_MASK); - - __inode_ctx_put (inode, this, ctx); - } - UNLOCK (&inode->lock); -out: - return; -} - - -uint64_t -afr_is_opendir_done (xlator_t *this, inode_t *inode) -{ - int ret = 0; - - uint64_t ctx = 0; - uint64_t opendir_done = 0; - - VALIDATE_OR_GOTO (inode, out); - - LOCK (&inode->lock); - { - ret = __inode_ctx_get (inode, this, &ctx); - - if (ret < 0) - goto unlock; - - opendir_done = ctx & AFR_ICTX_OPENDIR_DONE_MASK; - } -unlock: - UNLOCK (&inode->lock); - -out: - return opendir_done; -} - - -void -afr_set_opendir_done (xlator_t *this, inode_t *inode) -{ - uint64_t ctx = 0; - int ret = 0; - - VALIDATE_OR_GOTO (inode, out); - - LOCK (&inode->lock); - { - ret = __inode_ctx_get (inode, this, &ctx); - - if (ret < 0) { - ctx = 0; - } - - ctx = (~AFR_ICTX_OPENDIR_DONE_MASK & ctx) - | (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_OPENDIR_DONE_MASK); - - __inode_ctx_put (inode, this, ctx); - } - UNLOCK (&inode->lock); -out: - return; -} - - -uint64_t -afr_read_child (xlator_t *this, inode_t *inode) -{ - int ret = 0; - - uint64_t ctx = 0; - uint64_t read_child = 0; - - VALIDATE_OR_GOTO (inode, out); - - LOCK (&inode->lock); - { - ret = __inode_ctx_get (inode, this, &ctx); - - if (ret < 0) - goto unlock; - - read_child = ctx & AFR_ICTX_READ_CHILD_MASK; - } -unlock: - UNLOCK (&inode->lock); - -out: - return read_child; -} - - -void -afr_set_read_child (xlator_t *this, inode_t *inode, int32_t read_child) -{ - uint64_t ctx = 0; - int ret = 0; - - VALIDATE_OR_GOTO (inode, out); - - LOCK (&inode->lock); - { - ret = __inode_ctx_get (inode, this, &ctx); - - if (ret < 0) { - ctx = 0; - } - - ctx = (~AFR_ICTX_READ_CHILD_MASK & ctx) - | (AFR_ICTX_READ_CHILD_MASK & read_child); - - __inode_ctx_put (inode, this, ctx); - } - UNLOCK (&inode->lock); - -out: - return; -} - - -/** - * afr_local_cleanup - cleanup everything in frame->local - */ +#include "afr-common.c" +#include "afr-messages.h" -void -afr_local_sh_cleanup (afr_local_t *local, xlator_t *this) -{ - afr_self_heal_t *sh = NULL; - afr_private_t *priv = NULL; - int i = 0; - - - sh = &local->self_heal; - priv = this->private; - - if (sh->buf) - FREE (sh->buf); - - if (sh->xattr) { - for (i = 0; i < priv->child_count; i++) { - if (sh->xattr[i]) { - dict_unref (sh->xattr[i]); - sh->xattr[i] = NULL; - } - } - FREE (sh->xattr); - } - - if (sh->child_errno) - FREE (sh->child_errno); - - if (sh->pending_matrix) { - for (i = 0; i < priv->child_count; i++) { - FREE (sh->pending_matrix[i]); - } - FREE (sh->pending_matrix); - } - - if (sh->delta_matrix) { - for (i = 0; i < priv->child_count; i++) { - FREE (sh->delta_matrix[i]); - } - FREE (sh->delta_matrix); - } - - if (sh->sources) - FREE (sh->sources); - - if (sh->success) - FREE (sh->success); - - if (sh->locked_nodes) - FREE (sh->locked_nodes); - - if (sh->healing_fd && !sh->healing_fd_opened) { - fd_unref (sh->healing_fd); - sh->healing_fd = NULL; - } - - if (sh->linkname) - FREE (sh->linkname); - - loc_wipe (&sh->parent_loc); -} - - -void -afr_local_cleanup (afr_local_t *local, xlator_t *this) -{ - int i; - afr_private_t * priv = NULL; - - if (!local) - return; - - afr_local_sh_cleanup (local, this); - - FREE (local->child_errno); - - priv = this->private; - - for (i = 0; i < priv->child_count; i++) { - if (local->pending && local->pending[i]) - FREE (local->pending[i]); - } - - FREE (local->pending); - - loc_wipe (&local->loc); - loc_wipe (&local->newloc); - - FREE (local->transaction.locked_nodes); - FREE (local->transaction.child_errno); - - FREE (local->transaction.basename); - FREE (local->transaction.new_basename); - - loc_wipe (&local->transaction.parent_loc); - loc_wipe (&local->transaction.new_parent_loc); - - if (local->fd) - fd_unref (local->fd); - - if (local->xattr_req) - dict_unref (local->xattr_req); - - FREE (local->child_up); - - { /* lookup */ - if (local->cont.lookup.xattrs) { - for (i = 0; i < priv->child_count; i++) { - if (local->cont.lookup.xattrs[i]) { - dict_unref (local->cont.lookup.xattrs[i]); - local->cont.lookup.xattrs[i] = NULL; - } - } - FREE (local->cont.lookup.xattrs); - local->cont.lookup.xattrs = NULL; - } - - if (local->cont.lookup.xattr) { - dict_unref (local->cont.lookup.xattr); - } - } - - { /* getxattr */ - if (local->cont.getxattr.name) - FREE (local->cont.getxattr.name); - } - - { /* lk */ - if (local->cont.lk.locked_nodes) - FREE (local->cont.lk.locked_nodes); - } - - { /* checksum */ - if (local->cont.checksum.file_checksum) - FREE (local->cont.checksum.file_checksum); - if (local->cont.checksum.dir_checksum) - FREE (local->cont.checksum.dir_checksum); - } - - { /* create */ - if (local->cont.create.fd) - fd_unref (local->cont.create.fd); - } - - { /* writev */ - FREE (local->cont.writev.vector); - } - - { /* setxattr */ - if (local->cont.setxattr.dict) - dict_unref (local->cont.setxattr.dict); - } - - { /* removexattr */ - FREE (local->cont.removexattr.name); - } - - { /* symlink */ - FREE (local->cont.symlink.linkpath); - } - - { /* opendir */ - if (local->cont.opendir.checksum) - FREE (local->cont.opendir.checksum); - } -} - - -int -afr_frame_return (call_frame_t *frame) -{ - afr_local_t *local = NULL; - int call_count = 0; - - local = frame->local; - - LOCK (&frame->lock); - { - call_count = --local->call_count; - } - UNLOCK (&frame->lock); - - return call_count; -} - - -/** - * up_children_count - return the number of children that are up - */ - -int -afr_up_children_count (int child_count, unsigned char *child_up) -{ - int i = 0; - int ret = 0; - - for (i = 0; i < child_count; i++) - if (child_up[i]) - ret++; - return ret; -} - - -int -afr_locked_nodes_count (unsigned char *locked_nodes, int child_count) -{ - int ret = 0; - int i; - - for (i = 0; i < child_count; i++) - if (locked_nodes[i]) - ret++; - - return ret; -} +struct volume_options options[]; +static char *afr_favorite_child_policies[AFR_FAV_CHILD_POLICY_MAX + 1] = { + [AFR_FAV_CHILD_NONE] = "none", + [AFR_FAV_CHILD_BY_SIZE] = "size", + [AFR_FAV_CHILD_BY_CTIME] = "ctime", + [AFR_FAV_CHILD_BY_MTIME] = "mtime", + [AFR_FAV_CHILD_BY_MAJORITY] = "majority", + [AFR_FAV_CHILD_POLICY_MAX] = NULL, +}; -ino64_t -afr_itransform (ino64_t ino, int child_count, int child_index) +int32_t +notify(xlator_t *this, int32_t event, void *data, ...) { - ino64_t scaled_ino = -1; - - if (ino == ((uint64_t) -1)) { - scaled_ino = ((uint64_t) -1); - goto out; - } + int ret = -1; + va_list ap; + void *data2 = NULL; - scaled_ino = (ino * child_count) + child_index; + va_start(ap, data); + data2 = va_arg(ap, dict_t *); + va_end(ap); + ret = afr_notify(this, event, data, data2); -out: - return scaled_ino; + return ret; } - -int -afr_deitransform_orig (ino64_t ino, int child_count) +int32_t +mem_acct_init(xlator_t *this) { - int index = -1; + int ret = -1; - index = ino % child_count; + if (!this) + return ret; - return index; -} + ret = xlator_mem_acct_init(this, gf_afr_mt_end + 1); + if (ret != 0) { + return ret; + } -int -afr_deitransform (ino64_t ino, int child_count) -{ - return 0; + return ret; } - int -afr_self_heal_lookup_unwind (call_frame_t *frame, xlator_t *this) -{ - afr_local_t *local = NULL; - - local = frame->local; - - if (local->govinda_gOvinda) { - afr_set_split_brain (this, local->cont.lookup.inode); - } - - AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno, - local->cont.lookup.inode, - &local->cont.lookup.buf, - local->cont.lookup.xattr, - &local->cont.lookup.postparent); - - return 0; -} - - -static void -afr_lookup_collect_xattr (afr_local_t *local, xlator_t *this, - int child_index, dict_t *xattr) -{ - uint32_t open_fd_count = 0; - uint32_t inodelk_count = 0; - uint32_t entrylk_count = 0; - - int ret = 0; - - if (afr_sh_has_metadata_pending (xattr, child_index, this)) - local->self_heal.need_metadata_self_heal = _gf_true; - - if (afr_sh_has_entry_pending (xattr, child_index, this)) - local->self_heal.need_entry_self_heal = _gf_true; - - if (afr_sh_has_data_pending (xattr, child_index, this)) - local->self_heal.need_data_self_heal = _gf_true; - - ret = dict_get_uint32 (xattr, GLUSTERFS_OPEN_FD_COUNT, - &open_fd_count); - if (ret == 0) - local->open_fd_count += open_fd_count; - - ret = dict_get_uint32 (xattr, GLUSTERFS_INODELK_COUNT, - &inodelk_count); - if (ret == 0) - local->inodelk_count += inodelk_count; - - ret = dict_get_uint32 (xattr, GLUSTERFS_ENTRYLK_COUNT, - &entrylk_count); - if (ret == 0) - local->entrylk_count += entrylk_count; -} - - -static void -afr_lookup_self_heal_check (afr_local_t *local, struct stat *buf, - struct stat *lookup_buf) +xlator_subvolume_index(xlator_t *this, xlator_t *subvol) { - if (FILETYPE_DIFFERS (buf, lookup_buf)) { - /* mismatching filetypes with same name - -- Govinda !! GOvinda !!! - */ - - gf_log ("afr", GF_LOG_TRACE, - "file %s is govinda!", local->loc.path); - - local->govinda_gOvinda = 1; - } + int index = -1; + int i = 0; + xlator_list_t *list = NULL; - if (PERMISSION_DIFFERS (buf, lookup_buf)) { - /* mismatching permissions */ - local->self_heal.need_metadata_self_heal = _gf_true; - } + list = this->children; - if (OWNERSHIP_DIFFERS (buf, lookup_buf)) { - /* mismatching permissions */ - local->self_heal.need_metadata_self_heal = _gf_true; - } - - if (SIZE_DIFFERS (buf, lookup_buf) - && S_ISREG (buf->st_mode)) { - local->self_heal.need_data_self_heal = _gf_true; + while (list) { + if (subvol == list->xlator || + strcmp(subvol->name, list->xlator->name) == 0) { + index = i; + break; } + list = list->next; + i++; + } + return index; } - static void -afr_lookup_done (call_frame_t *frame, xlator_t *this, struct stat *lookup_buf) +fix_quorum_options(xlator_t *this, afr_private_t *priv, char *qtype, + dict_t *options) { - int unwind = 1; - int source = -1; - - afr_local_t *local = NULL; - - local = frame->local; - - local->cont.lookup.postparent.st_ino = local->cont.lookup.parent_ino; - - if (local->cont.lookup.ino) { - local->cont.lookup.buf.st_ino = local->cont.lookup.ino; - local->cont.lookup.buf.st_dev = local->cont.lookup.gen; - } - - if (local->op_ret == 0) { - /* KLUDGE: assuming DHT will not itransform in - revalidate */ - if (local->cont.lookup.inode->ino) { - local->cont.lookup.buf.st_ino = - local->cont.lookup.inode->ino; - local->cont.lookup.buf.st_dev = - local->cont.lookup.inode->generation; - } - } - - if (local->success_count && local->enoent_count) { - local->self_heal.need_metadata_self_heal = _gf_true; - local->self_heal.need_data_self_heal = _gf_true; - local->self_heal.need_entry_self_heal = _gf_true; - } - - if (local->success_count) { - /* check for split-brain case in previous lookup */ - if (afr_is_split_brain (this, - local->cont.lookup.inode)) - local->self_heal.need_data_self_heal = _gf_true; - } - - if ((local->self_heal.need_metadata_self_heal - || local->self_heal.need_data_self_heal - || local->self_heal.need_entry_self_heal) - && ((!local->cont.lookup.is_revalidate) - || (local->op_ret != -1))) { - - if (local->open_fd_count - || local->inodelk_count - || local->entrylk_count) { - - /* Someone else is doing self-heal on this file. - So just make a best effort to set the read-subvolume - and return */ - - if (S_ISREG (local->cont.lookup.inode->st_mode)) { - source = afr_self_heal_get_source (this, local, local->cont.lookup.xattrs); - - if (source >= 0) { - afr_set_read_child (this, - local->cont.lookup.inode, - source); - } - } - } else { - if (!local->cont.lookup.inode->st_mode) { - /* fix for RT #602 */ - local->cont.lookup.inode->st_mode = - lookup_buf->st_mode; - } - - local->self_heal.background = _gf_true; - local->self_heal.mode = local->cont.lookup.buf.st_mode; - local->self_heal.unwind = afr_self_heal_lookup_unwind; - - unwind = 0; - - afr_self_heal (frame, this); - } - } - - if (unwind) { - AFR_STACK_UNWIND (lookup, frame, local->op_ret, - local->op_errno, - local->cont.lookup.inode, - &local->cont.lookup.buf, - local->cont.lookup.xattr, - &local->cont.lookup.postparent); - } + if (dict_get_sizen(options, "quorum-type") == NULL) { + /* If user doesn't configure anything enable auto-quorum if the + * replica has more than two subvolumes */ + if (priv->child_count > 2) + qtype = "auto"; + } + + if (priv->quorum_count && strcmp(qtype, "fixed")) { + gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_QUORUM_OVERRIDE, + "quorum-type %s overriding quorum-count %u", qtype, + priv->quorum_count); + } + + if (!strcmp(qtype, "none")) { + priv->quorum_count = 0; + } else if (!strcmp(qtype, "auto")) { + priv->quorum_count = AFR_QUORUM_AUTO; + } } - -/* - * During a lookup, some errors are more "important" than - * others in that they must be given higher priority while - * returning to the user. - * - * The hierarchy is ESTALE > ENOENT > others - * - */ - -static gf_boolean_t -__error_more_important (int32_t old_errno, int32_t new_errno) +int +afr_set_favorite_child_policy(afr_private_t *priv, char *policy) { - gf_boolean_t ret = _gf_true; + int index = -1; - /* Nothing should ever overwrite ESTALE */ - if (old_errno == ESTALE) - ret = _gf_false; + index = gf_get_index_by_elem(afr_favorite_child_policies, policy); + if (index < 0 || index >= AFR_FAV_CHILD_POLICY_MAX) + return -1; - /* Nothing should overwrite ENOENT, except ESTALE */ - else if ((old_errno == ENOENT) && (new_errno != ESTALE)) - ret = _gf_false; + priv->fav_child_policy = index; - return ret; + return 0; } - -int -afr_fresh_lookup_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno, - inode_t *inode, struct stat *buf, dict_t *xattr, - struct stat *postparent) +static void +set_data_self_heal_algorithm(afr_private_t *priv, char *algo) { - afr_local_t * local = NULL; - afr_private_t * priv = NULL; - struct stat * lookup_buf = NULL; - - int call_count = -1; - int child_index = -1; - int first_up_child = -1; - - child_index = (long) cookie; - priv = this->private; - - LOCK (&frame->lock); - { - local = frame->local; - - lookup_buf = &local->cont.lookup.buf; - - if (op_ret == -1) { - if (op_errno == ENOENT) - local->enoent_count++; - - if (__error_more_important (local->op_errno, op_errno)) - local->op_errno = op_errno; - - if (local->op_errno == ESTALE) { - local->op_ret = -1; - } - - goto unlock; - } - - afr_lookup_collect_xattr (local, this, child_index, xattr); - - first_up_child = afr_first_up_child (priv); - - if (child_index == first_up_child) { - local->cont.lookup.ino = - afr_itransform (buf->st_ino, - priv->child_count, - first_up_child); - local->cont.lookup.gen = buf->st_dev; - } - - if (local->success_count == 0) { - if (local->op_errno != ESTALE) - local->op_ret = op_ret; - - local->cont.lookup.inode = inode; - local->cont.lookup.xattr = dict_ref (xattr); - local->cont.lookup.xattrs[child_index] = dict_ref (xattr); - local->cont.lookup.postparent = *postparent; - - *lookup_buf = *buf; - - lookup_buf->st_ino = afr_itransform (buf->st_ino, - priv->child_count, - child_index); - if (priv->read_child >= 0) { - afr_set_read_child (this, - local->cont.lookup.inode, - priv->read_child); - } else { - afr_set_read_child (this, - local->cont.lookup.inode, - child_index); - } - - } else { - afr_lookup_self_heal_check (local, buf, lookup_buf); - - if (child_index == local->read_child_index) { - /* - lookup has succeeded on the read child. - So use its inode number - */ - if (local->cont.lookup.xattr) - dict_unref (local->cont.lookup.xattr); - - local->cont.lookup.xattr = dict_ref (xattr); - - local->cont.lookup.inode = inode; - local->cont.lookup.xattrs[child_index] = dict_ref (xattr); - local->cont.lookup.postparent = *postparent; - - *lookup_buf = *buf; - - if (priv->read_child >= 0) { - afr_set_read_child (this, - local->cont.lookup.inode, - priv->read_child); - } else { - afr_set_read_child (this, - local->cont.lookup.inode, - local->read_child_index); - } - } - - } - - local->success_count++; - } -unlock: - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); - - if (call_count == 0) { - afr_lookup_done (frame, this, lookup_buf); - } - - return 0; + if (!algo) { + priv->data_self_heal_algorithm = AFR_SELFHEAL_DATA_DYNAMIC; + } else if (strcmp(algo, "full") == 0) { + priv->data_self_heal_algorithm = AFR_SELFHEAL_DATA_FULL; + } else if (strcmp(algo, "diff") == 0) { + priv->data_self_heal_algorithm = AFR_SELFHEAL_DATA_DIFF; + } else { + priv->data_self_heal_algorithm = AFR_SELFHEAL_DATA_DYNAMIC; + } } - -int -afr_revalidate_lookup_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno, - inode_t *inode, struct stat *buf, dict_t *xattr, - struct stat *postparent) +void +afr_handle_anon_inode_options(afr_private_t *priv, dict_t *options) { - afr_local_t * local = NULL; - afr_private_t * priv = NULL; - struct stat * lookup_buf = NULL; - - int call_count = -1; - int child_index = -1; - int first_up_child = -1; - - child_index = (long) cookie; - priv = this->private; - - LOCK (&frame->lock); - { - local = frame->local; - - lookup_buf = &local->cont.lookup.buf; - - if (op_ret == -1) { - if (op_errno == ENOENT) - local->enoent_count++; - - if (__error_more_important (local->op_errno, op_errno)) - local->op_errno = op_errno; - - if (local->op_errno == ESTALE) { - local->op_ret = -1; - } - - goto unlock; - } - - afr_lookup_collect_xattr (local, this, child_index, xattr); - - first_up_child = afr_first_up_child (priv); - - if (child_index == first_up_child) { - local->cont.lookup.ino = - afr_itransform (buf->st_ino, - priv->child_count, - first_up_child); - local->cont.lookup.gen = buf->st_dev; - } - - /* in case of revalidate, we need to send stat of the - * child whose stat was sent during the first lookup. - * (so that time stamp does not vary with revalidate. - * in case it is down, stat of the fist success will - * be replied */ - - /* inode number should be preserved across revalidates */ - - if (local->success_count == 0) { - if (local->op_errno != ESTALE) - local->op_ret = op_ret; - - local->cont.lookup.inode = inode; - local->cont.lookup.xattr = dict_ref (xattr); - local->cont.lookup.xattrs[child_index] = dict_ref (xattr); - local->cont.lookup.postparent = *postparent; - - *lookup_buf = *buf; - - lookup_buf->st_ino = afr_itransform (buf->st_ino, - priv->child_count, - child_index); - - if (priv->read_child >= 0) { - afr_set_read_child (this, - local->cont.lookup.inode, - priv->read_child); - } else { - afr_set_read_child (this, - local->cont.lookup.inode, - child_index); - } - - } else { - afr_lookup_self_heal_check (local, buf, lookup_buf); - - if (child_index == local->read_child_index) { - - /* - lookup has succeeded on the read child. - So use its inode number - */ - - if (local->cont.lookup.xattr) - dict_unref (local->cont.lookup.xattr); - - local->cont.lookup.inode = inode; - local->cont.lookup.xattr = dict_ref (xattr); - local->cont.lookup.xattrs[child_index] = dict_ref (xattr); - local->cont.lookup.postparent = *postparent; - - *lookup_buf = *buf; - - if (priv->read_child >= 0) { - afr_set_read_child (this, - local->cont.lookup.inode, - priv->read_child); - } else { - afr_set_read_child (this, - local->cont.lookup.inode, - local->read_child_index); - } - } - - } - - local->success_count++; - } -unlock: - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); + char *volfile_id_str = NULL; + uuid_t anon_inode_gfid = {0}; - if (call_count == 0) { - afr_lookup_done (frame, this, lookup_buf); - } - - return 0; + /*If volume id is not present don't enable anything*/ + if (dict_get_str(options, "volume-id", &volfile_id_str)) + return; + GF_ASSERT(strlen(AFR_ANON_DIR_PREFIX) + strlen(volfile_id_str) <= NAME_MAX); + /*anon_inode_name is not supposed to change once assigned*/ + if (!priv->anon_inode_name[0]) { + snprintf(priv->anon_inode_name, sizeof(priv->anon_inode_name), "%s-%s", + AFR_ANON_DIR_PREFIX, volfile_id_str); + gf_uuid_parse(volfile_id_str, anon_inode_gfid); + /*Flip a bit to make sure volfile-id and anon-gfid are not same*/ + anon_inode_gfid[0] ^= 1; + uuid_utoa_r(anon_inode_gfid, priv->anon_gfid_str); + } } - int -afr_lookup (call_frame_t *frame, xlator_t *this, - loc_t *loc, dict_t *xattr_req) +reconfigure(xlator_t *this, dict_t *options) { - afr_private_t *priv = NULL; - afr_local_t *local = NULL; - int ret = -1; - int i = 0; + afr_private_t *priv = NULL; + xlator_t *read_subvol = NULL; + int read_subvol_index = -1; + int timeout_old = 0; + int ret = -1; + int index = -1; + char *qtype = NULL; + char *fav_child_policy = NULL; + char *data_self_heal = NULL; + char *data_self_heal_algorithm = NULL; + char *locking_scheme = NULL; + gf_boolean_t consistent_io = _gf_false; + gf_boolean_t choose_local_old = _gf_false; + gf_boolean_t enabled_old = _gf_false; - fop_lookup_cbk_t callback; + priv = this->private; - int call_count = 0; + GF_OPTION_RECONF("metadata-splitbrain-forced-heal", + priv->metadata_splitbrain_forced_heal, options, bool, out); - uint64_t ctx; + GF_OPTION_RECONF("background-self-heal-count", + priv->background_self_heal_count, options, uint32, out); - int32_t op_errno = 0; + GF_OPTION_RECONF("heal-wait-queue-length", priv->heal_wait_qlen, options, + uint32, out); - priv = this->private; + GF_OPTION_RECONF("metadata-self-heal", priv->metadata_self_heal, options, + bool, out); - ALLOC_OR_GOTO (local, afr_local_t, out); + GF_OPTION_RECONF("data-self-heal", data_self_heal, options, str, out); + if (gf_string2boolean(data_self_heal, &priv->data_self_heal) == -1) + goto out; - local->op_ret = -1; + GF_OPTION_RECONF("entry-self-heal", priv->entry_self_heal, options, bool, + out); - frame->local = local; + GF_OPTION_RECONF("data-self-heal-window-size", + priv->data_self_heal_window_size, options, uint32, out); - if (!strcmp (loc->path, "/" GF_REPLICATE_TRASH_DIR)) { - op_errno = ENOENT; - goto out; - } - - loc_copy (&local->loc, loc); - - ret = inode_ctx_get (loc->inode, this, &ctx); - if (ret == 0) { - /* lookup is a revalidate */ + GF_OPTION_RECONF("data-self-heal-algorithm", data_self_heal_algorithm, + options, str, out); + set_data_self_heal_algorithm(priv, data_self_heal_algorithm); - callback = afr_revalidate_lookup_cbk; + GF_OPTION_RECONF("halo-enabled", priv->halo_enabled, options, bool, out); - local->cont.lookup.is_revalidate = _gf_true; - local->read_child_index = afr_read_child (this, - loc->inode); - } else { - callback = afr_fresh_lookup_cbk; - - LOCK (&priv->read_child_lock); - { - local->read_child_index = (++priv->read_child_rr) - % (priv->child_count); - } - UNLOCK (&priv->read_child_lock); - } + GF_OPTION_RECONF("halo-shd-max-latency", priv->shd.halo_max_latency_msec, + options, uint32, out); - local->child_up = memdup (priv->child_up, priv->child_count); + GF_OPTION_RECONF("halo-nfsd-max-latency", priv->nfsd.halo_max_latency_msec, + options, uint32, out); - local->cont.lookup.xattrs = CALLOC (priv->child_count, - sizeof (*local->cont.lookup.xattr)); + GF_OPTION_RECONF("halo-max-latency", priv->halo_max_latency_msec, options, + uint32, out); - local->call_count = afr_up_children_count (priv->child_count, - local->child_up); - call_count = local->call_count; + GF_OPTION_RECONF("halo-max-replicas", priv->halo_max_replicas, options, + uint32, out); - if (local->call_count == 0) { - ret = -1; - op_errno = ENOTCONN; - goto out; - } - - /* By default assume ENOTCONN. On success it will be set to 0. */ - local->op_errno = ENOTCONN; - - if (xattr_req == NULL) - local->xattr_req = dict_new (); - else - local->xattr_req = dict_ref (xattr_req); - - for (i = 0; i < priv->child_count; i++) { - ret = dict_set_uint64 (local->xattr_req, priv->pending_key[i], - 3 * sizeof(int32_t)); - - /* 3 = data+metadata+entry */ - } + GF_OPTION_RECONF("halo-min-replicas", priv->halo_min_replicas, options, + uint32, out); - ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_OPEN_FD_COUNT, 0); - ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0); - ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0); - - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND_COOKIE (frame, callback, (void *) (long) i, - priv->children[i], - priv->children[i]->fops->lookup, - loc, local->xattr_req); - if (!--call_count) - break; - } - } - - ret = 0; -out: - if (ret == -1) - AFR_STACK_UNWIND (lookup, frame, -1, op_errno, - NULL, NULL, NULL, NULL); + GF_OPTION_RECONF("read-subvolume", read_subvol, options, xlator, out); - return 0; -} + choose_local_old = priv->choose_local; + GF_OPTION_RECONF("choose-local", priv->choose_local, options, bool, out); + if (choose_local_old != priv->choose_local) { + priv->read_child = -1; + if (choose_local_old == _gf_false) + priv->did_discovery = _gf_false; + } -/* {{{ open */ + GF_OPTION_RECONF("read-hash-mode", priv->hash_mode, options, uint32, out); -int -afr_fd_ctx_set (xlator_t *this, fd_t *fd) -{ - afr_private_t * priv = NULL; - - int op_ret = 0; - int ret = 0; - - uint64_t ctx; - afr_fd_ctx_t * fd_ctx = NULL; - - VALIDATE_OR_GOTO (this->private, out); - VALIDATE_OR_GOTO (fd, out); - - priv = this->private; - - LOCK (&fd->lock); - { - ret = __fd_ctx_get (fd, this, &ctx); - - if (ret == 0) - goto unlock; - - fd_ctx = CALLOC (1, sizeof (afr_fd_ctx_t)); - if (!fd_ctx) { - gf_log (this->name, GF_LOG_ERROR, - "Out of memory"); - - op_ret = -ENOMEM; - goto unlock; - } - - fd_ctx->pre_op_done = CALLOC (sizeof (*fd_ctx->pre_op_done), - priv->child_count); - if (!fd_ctx->pre_op_done) { - gf_log (this->name, GF_LOG_ERROR, - "Out of memory"); - op_ret = -ENOMEM; - goto unlock; - } - - fd_ctx->opened_on = CALLOC (sizeof (*fd_ctx->opened_on), - priv->child_count); - if (!fd_ctx->opened_on) { - gf_log (this->name, GF_LOG_ERROR, - "Out of memory"); - op_ret = -ENOMEM; - goto unlock; - } - - fd_ctx->child_failed = CALLOC (sizeof (*fd_ctx->child_failed), - priv->child_count); - - if (!fd_ctx->child_failed) { - gf_log (this->name, GF_LOG_ERROR, - "Out of memory"); - - op_ret = -ENOMEM; - goto unlock; - } - - fd_ctx->up_count = priv->up_count; - fd_ctx->down_count = priv->down_count; - - ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx); - if (ret < 0) { - op_ret = ret; - } + if (read_subvol) { + index = xlator_subvolume_index(this, read_subvol); + if (index == -1) { + gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL, + "%s not a subvolume", read_subvol->name); + goto out; } -unlock: - UNLOCK (&fd->lock); -out: - return ret; -} - -/* {{{ flush */ - -int -afr_flush_unwind (call_frame_t *frame, xlator_t *this) -{ - afr_local_t * local = NULL; - afr_private_t * priv = NULL; - call_frame_t *main_frame = NULL; - - local = frame->local; - priv = this->private; - - LOCK (&frame->lock); - { - if (local->transaction.main_frame) - main_frame = local->transaction.main_frame; - local->transaction.main_frame = NULL; - } - UNLOCK (&frame->lock); - - if (main_frame) { - AFR_STACK_UNWIND (flush, main_frame, - local->op_ret, local->op_errno); - } - - return 0; -} - - -int -afr_flush_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, - int32_t op_ret, int32_t op_errno) -{ - afr_local_t * local = NULL; - afr_private_t * priv = NULL; - - int call_count = -1; - int child_index = (long) cookie; - int need_unwind = 0; - - local = frame->local; - priv = this->private; - - LOCK (&frame->lock); - { - if (afr_fop_failed (op_ret, op_errno)) - afr_transaction_fop_failed (frame, this, child_index); - - if (op_ret != -1) { - if (local->success_count == 0) { - local->op_ret = op_ret; - } - local->success_count++; - - if (local->success_count == priv->wait_count) { - need_unwind = 1; - } - } - - local->op_errno = op_errno; - } - UNLOCK (&frame->lock); - - if (need_unwind) - afr_flush_unwind (frame, this); - - call_count = afr_frame_return (frame); - - if (call_count == 0) { - local->transaction.resume (frame, this); - } - - return 0; -} - - -int -afr_flush_wind (call_frame_t *frame, xlator_t *this) -{ - afr_local_t *local = NULL; - afr_private_t *priv = NULL; - - int i = 0; - int call_count = -1; - - local = frame->local; - priv = this->private; - - call_count = afr_up_children_count (priv->child_count, local->child_up); - - if (call_count == 0) { - local->transaction.resume (frame, this); - return 0; - } - - local->call_count = call_count; - - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND_COOKIE (frame, afr_flush_wind_cbk, - (void *) (long) i, - priv->children[i], - priv->children[i]->fops->flush, - local->fd); - - if (!--call_count) - break; - } - } - - return 0; -} - - -int -afr_flush_done (call_frame_t *frame, xlator_t *this) -{ - afr_local_t *local = NULL; - - local = frame->local; - - local->transaction.unwind (frame, this); - - AFR_STACK_DESTROY (frame); - - return 0; -} - - -int -afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd) -{ - afr_private_t * priv = NULL; - afr_local_t * local = NULL; - - call_frame_t * transaction_frame = NULL; - - int ret = -1; - - int op_ret = -1; - int op_errno = 0; - - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); - - priv = this->private; - - transaction_frame = copy_frame (frame); - if (!transaction_frame) { - op_errno = ENOMEM; - gf_log (this->name, GF_LOG_ERROR, - "Out of memory."); - goto out; + priv->read_child = index; + } + + GF_OPTION_RECONF("read-subvolume-index", read_subvol_index, options, int32, + out); + + if (read_subvol_index > -1) { + index = read_subvol_index; + if (index >= priv->child_count) { + gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL, + "%d not a subvolume-index", index); + goto out; } - - ALLOC_OR_GOTO (local, afr_local_t, out); - - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } - - transaction_frame->local = local; - - local->op = GF_FOP_FLUSH; - - local->transaction.fop = afr_flush_wind; - local->transaction.done = afr_flush_done; - local->transaction.unwind = afr_flush_unwind; - - local->fd = fd_ref (fd); - - local->transaction.main_frame = frame; - local->transaction.start = LLONG_MAX - 1; - local->transaction.len = 0; - - afr_transaction (transaction_frame, this, AFR_FLUSH_TRANSACTION); - - op_ret = 0; + priv->read_child = index; + } + + GF_OPTION_RECONF("pre-op-compat", priv->pre_op_compat, options, bool, out); + GF_OPTION_RECONF("locking-scheme", locking_scheme, options, str, out); + priv->granular_locks = (strcmp(locking_scheme, "granular") == 0); + GF_OPTION_RECONF("full-lock", priv->full_lock, options, bool, out); + GF_OPTION_RECONF("granular-entry-heal", priv->esh_granular, options, bool, + out); + + GF_OPTION_RECONF("eager-lock", priv->eager_lock, options, bool, out); + GF_OPTION_RECONF("optimistic-change-log", priv->optimistic_change_log, + options, bool, out); + GF_OPTION_RECONF("quorum-type", qtype, options, str, out); + GF_OPTION_RECONF("quorum-count", priv->quorum_count, options, uint32, out); + fix_quorum_options(this, priv, qtype, options); + if (priv->quorum_count && !afr_has_quorum(priv->child_up, this, NULL)) + gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_QUORUM_FAIL, + "Client-quorum is not met"); + + GF_OPTION_RECONF("post-op-delay-secs", priv->post_op_delay_secs, options, + uint32, out); + + GF_OPTION_RECONF(AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, options, + size_uint64, out); + /* Reset this so we re-discover in case the topology changed. */ + GF_OPTION_RECONF("ensure-durability", priv->ensure_durability, options, + bool, out); + + enabled_old = priv->shd.enabled; + GF_OPTION_RECONF("self-heal-daemon", priv->shd.enabled, options, bool, out); + + GF_OPTION_RECONF("iam-self-heal-daemon", priv->shd.iamshd, options, bool, + out); + + timeout_old = priv->shd.timeout; + GF_OPTION_RECONF("heal-timeout", priv->shd.timeout, options, int32, out); + + GF_OPTION_RECONF("consistent-metadata", priv->consistent_metadata, options, + bool, out); + + GF_OPTION_RECONF("shd-max-threads", priv->shd.max_threads, options, uint32, + out); + + GF_OPTION_RECONF("shd-wait-qlength", priv->shd.wait_qlength, options, + uint32, out); + + GF_OPTION_RECONF("favorite-child-policy", fav_child_policy, options, str, + out); + if (afr_set_favorite_child_policy(priv, fav_child_policy) == -1) + goto out; + + priv->did_discovery = _gf_false; + + GF_OPTION_RECONF("consistent-io", consistent_io, options, bool, out); + if (priv->quorum_count != 0) + consistent_io = _gf_false; + priv->consistent_io = consistent_io; + + afr_handle_anon_inode_options(priv, options); + + GF_OPTION_RECONF("use-anonymous-inode", priv->use_anon_inode, options, bool, + out); + if (priv->shd.enabled) { + if ((priv->shd.enabled != enabled_old) || + (timeout_old != priv->shd.timeout)) + afr_selfheal_childup(this, priv); + } + + ret = 0; out: - if (op_ret == -1) { - if (transaction_frame) - AFR_STACK_DESTROY (transaction_frame); - - AFR_STACK_UNWIND (flush, frame, op_ret, op_errno); - } - - return 0; + return ret; } -/* }}} */ - - -int -afr_release (xlator_t *this, fd_t *fd) +static int +afr_pending_xattrs_init(afr_private_t *priv, xlator_t *this) { - uint64_t ctx; - afr_fd_ctx_t * fd_ctx; - - int ret = 0; - - ret = fd_ctx_get (fd, this, &ctx); - - if (ret < 0) + int ret = -1; + int i = 0; + char *ptr = NULL; + char *ptr1 = NULL; + char *xattrs_list = NULL; + xlator_list_t *trav = NULL; + int child_count = -1; + + trav = this->children; + child_count = priv->child_count; + if (priv->thin_arbiter_count) { + /* priv->pending_key[THIN_ARBITER_BRICK_INDEX] is used as the + * name of the thin arbiter file for persistence across add/ + * removal of DHT subvols.*/ + child_count++; + } + + GF_OPTION_INIT("afr-pending-xattr", xattrs_list, str, out); + priv->pending_key = GF_CALLOC(sizeof(*priv->pending_key), child_count, + gf_afr_mt_char); + if (!priv->pending_key) { + ret = -ENOMEM; + goto out; + } + if (!xattrs_list) { + gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_NO_CHANGELOG, + "Unable to fetch afr-pending-xattr option from volfile." + " Falling back to using client translator names. "); + + while (i < child_count) { + ret = gf_asprintf(&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX, + trav->xlator->name); + if (ret == -1) { + ret = -ENOMEM; goto out; - - fd_ctx = (afr_fd_ctx_t *)(long) ctx; - - if (fd_ctx) { - if (fd_ctx->child_failed) - FREE (fd_ctx->child_failed); - - FREE (fd_ctx); + } + trav = trav->next; + i++; } - -out: - return 0; -} - - -/* {{{ fsync */ - -int -afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, - int32_t op_ret, int32_t op_errno, struct stat *prebuf, - struct stat *postbuf) -{ - afr_local_t *local = NULL; - - int call_count = -1; - - int child_index = (long) cookie; - int read_child = 0; - - local = frame->local; - - read_child = afr_read_child (this, local->fd->inode); - - LOCK (&frame->lock); - { - if (child_index == read_child) { - local->read_child_returned = _gf_true; - } - - if (op_ret == 0) { - local->op_ret = 0; - - if (local->success_count == 0) { - local->cont.fsync.prebuf = *prebuf; - local->cont.fsync.postbuf = *postbuf; - } - - if (child_index == read_child) { - local->cont.fsync.prebuf = *prebuf; - local->cont.fsync.postbuf = *postbuf; - } - - local->success_count++; - } - - local->op_errno = op_errno; - } - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); - - if (call_count == 0) { - local->cont.fsync.prebuf.st_ino = local->cont.fsync.ino; - local->cont.fsync.postbuf.st_ino = local->cont.fsync.ino; - - AFR_STACK_UNWIND (fsync, frame, local->op_ret, local->op_errno, - &local->cont.fsync.prebuf, - &local->cont.fsync.postbuf); + ret = 0; + goto out; + } + + ptr = ptr1 = gf_strdup(xattrs_list); + if (!ptr) { + ret = -ENOMEM; + goto out; + } + for (i = 0, ptr = strtok(ptr, ","); ptr; ptr = strtok(NULL, ",")) { + ret = gf_asprintf(&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX, + ptr); + if (ret == -1) { + ret = -ENOMEM; + goto out; } + i++; + } + ret = 0; - return 0; -} - - -int -afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, - int32_t datasync) -{ - afr_private_t *priv = NULL; - afr_local_t *local = NULL; - - int ret = -1; - - int i = 0; - int32_t call_count = 0; - int32_t op_ret = -1; - int32_t op_errno = 0; - - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); - - priv = this->private; - - ALLOC_OR_GOTO (local, afr_local_t, out); - - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } - - call_count = local->call_count; - frame->local = local; - - local->fd = fd_ref (fd); - local->cont.fsync.ino = fd->inode->ino; - - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND_COOKIE (frame, afr_fsync_cbk, - (void *) (long) i, - priv->children[i], - priv->children[i]->fops->fsync, - fd, datasync); - if (!--call_count) - break; - } - } - - op_ret = 0; out: - if (op_ret == -1) { - AFR_STACK_UNWIND (fsync, frame, op_ret, op_errno, NULL, NULL); - } - return 0; + GF_FREE(ptr1); + return ret; } -/* }}} */ - -/* {{{ fsync */ - -int32_t -afr_fsyncdir_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno) -{ - afr_local_t *local = NULL; - - int call_count = -1; - - local = frame->local; - - LOCK (&frame->lock); - { - if (op_ret == 0) - local->op_ret = 0; - - local->op_errno = op_errno; - } - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); - - if (call_count == 0) - AFR_STACK_UNWIND (fsyncdir, frame, local->op_ret, - local->op_errno); - - return 0; -} - - -int32_t -afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, - int32_t datasync) -{ - afr_private_t *priv = NULL; - afr_local_t *local = NULL; - - int ret = -1; - - int i = 0; - int32_t call_count = 0; - int32_t op_ret = -1; - int32_t op_errno = 0; - - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); - - priv = this->private; - - ALLOC_OR_GOTO (local, afr_local_t, out); - - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } - - call_count = local->call_count; - frame->local = local; - - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND (frame, afr_fsyncdir_cbk, - priv->children[i], - priv->children[i]->fops->fsyncdir, - fd, datasync); - if (!--call_count) - break; - } - } - - op_ret = 0; -out: - if (op_ret == -1) { - AFR_STACK_UNWIND (fsyncdir, frame, op_ret, op_errno); - } - return 0; -} - -/* }}} */ - -/* {{{ xattrop */ - -int32_t -afr_xattrop_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno, - dict_t *xattr) -{ - afr_local_t *local = NULL; - - int call_count = -1; - - local = frame->local; - - LOCK (&frame->lock); - { - if (op_ret == 0) - local->op_ret = 0; - - local->op_errno = op_errno; - } - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); - - if (call_count == 0) - AFR_STACK_UNWIND (xattrop, frame, local->op_ret, local->op_errno, - xattr); - - return 0; -} - - -int32_t -afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc, - gf_xattrop_flags_t optype, dict_t *xattr) -{ - afr_private_t *priv = NULL; - afr_local_t *local = NULL; - - int ret = -1; - - int i = 0; - int32_t call_count = 0; - int32_t op_ret = -1; - int32_t op_errno = 0; - - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); - - priv = this->private; - - ALLOC_OR_GOTO (local, afr_local_t, out); - - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } - - call_count = local->call_count; - frame->local = local; - - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND (frame, afr_xattrop_cbk, - priv->children[i], - priv->children[i]->fops->xattrop, - loc, optype, xattr); - if (!--call_count) - break; - } - } - - op_ret = 0; -out: - if (op_ret == -1) { - AFR_STACK_UNWIND (xattrop, frame, op_ret, op_errno, NULL); - } - return 0; -} - -/* }}} */ - -/* {{{ fxattrop */ - -int32_t -afr_fxattrop_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno, - dict_t *xattr) -{ - afr_local_t *local = NULL; - - int call_count = -1; - - local = frame->local; - - LOCK (&frame->lock); - { - if (op_ret == 0) - local->op_ret = 0; - - local->op_errno = op_errno; - } - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); - - if (call_count == 0) - AFR_STACK_UNWIND (fxattrop, frame, local->op_ret, local->op_errno, - xattr); - - return 0; -} - - -int32_t -afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd, - gf_xattrop_flags_t optype, dict_t *xattr) -{ - afr_private_t *priv = NULL; - afr_local_t *local = NULL; - - int ret = -1; - - int i = 0; - int32_t call_count = 0; - int32_t op_ret = -1; - int32_t op_errno = 0; - - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); - - priv = this->private; - - ALLOC_OR_GOTO (local, afr_local_t, out); - - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } - - call_count = local->call_count; - frame->local = local; - - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND (frame, afr_fxattrop_cbk, - priv->children[i], - priv->children[i]->fops->fxattrop, - fd, optype, xattr); - if (!--call_count) - break; - } - } - - op_ret = 0; -out: - if (op_ret == -1) { - AFR_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, NULL); - } - return 0; -} - -/* }}} */ - - -int32_t -afr_inodelk_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno) - -{ - afr_local_t *local = NULL; - - int call_count = -1; - - local = frame->local; - - LOCK (&frame->lock); - { - if (op_ret == 0) - local->op_ret = 0; - - local->op_errno = op_errno; - } - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); - - if (call_count == 0) - AFR_STACK_UNWIND (inodelk, frame, local->op_ret, - local->op_errno); - - return 0; -} - - -int32_t -afr_inodelk (call_frame_t *frame, xlator_t *this, - const char *volume, loc_t *loc, int32_t cmd, struct flock *flock) -{ - afr_private_t *priv = NULL; - afr_local_t *local = NULL; - - int ret = -1; - - int i = 0; - int32_t call_count = 0; - int32_t op_ret = -1; - int32_t op_errno = 0; - - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); - - priv = this->private; - - ALLOC_OR_GOTO (local, afr_local_t, out); - - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } - - call_count = local->call_count; - frame->local = local; - - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND (frame, afr_inodelk_cbk, - priv->children[i], - priv->children[i]->fops->inodelk, - volume, loc, cmd, flock); - - if (!--call_count) - break; - } - } - - op_ret = 0; -out: - if (op_ret == -1) { - AFR_STACK_UNWIND (inodelk, frame, op_ret, op_errno); - } - return 0; -} - - -int32_t -afr_finodelk_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno) - -{ - afr_local_t *local = NULL; - - int call_count = -1; - - local = frame->local; - - LOCK (&frame->lock); - { - if (op_ret == 0) - local->op_ret = 0; - - local->op_errno = op_errno; - } - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); - - if (call_count == 0) - AFR_STACK_UNWIND (finodelk, frame, local->op_ret, - local->op_errno); - - return 0; -} - - -int32_t -afr_finodelk (call_frame_t *frame, xlator_t *this, - const char *volume, fd_t *fd, int32_t cmd, struct flock *flock) -{ - afr_private_t *priv = NULL; - afr_local_t *local = NULL; - - int ret = -1; - - int i = 0; - int32_t call_count = 0; - int32_t op_ret = -1; - int32_t op_errno = 0; - - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); - - priv = this->private; - - ALLOC_OR_GOTO (local, afr_local_t, out); - - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } - - call_count = local->call_count; - frame->local = local; - - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND (frame, afr_finodelk_cbk, - priv->children[i], - priv->children[i]->fops->finodelk, - volume, fd, cmd, flock); - - if (!--call_count) - break; - } - } - - op_ret = 0; -out: - if (op_ret == -1) { - AFR_STACK_UNWIND (finodelk, frame, op_ret, op_errno); - } - return 0; -} - - -int32_t -afr_entrylk_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno) - -{ - afr_local_t *local = NULL; - - int call_count = -1; - - local = frame->local; - - LOCK (&frame->lock); - { - if (op_ret == 0) - local->op_ret = 0; - - local->op_errno = op_errno; - } - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); - - if (call_count == 0) - AFR_STACK_UNWIND (entrylk, frame, local->op_ret, - local->op_errno); - - return 0; -} - - -int32_t -afr_entrylk (call_frame_t *frame, xlator_t *this, - const char *volume, loc_t *loc, - const char *basename, entrylk_cmd cmd, entrylk_type type) -{ - afr_private_t *priv = NULL; - afr_local_t *local = NULL; - - int ret = -1; - - int i = 0; - int32_t call_count = 0; - int32_t op_ret = -1; - int32_t op_errno = 0; - - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); - - priv = this->private; - - ALLOC_OR_GOTO (local, afr_local_t, out); - - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } - - call_count = local->call_count; - frame->local = local; - - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND (frame, afr_entrylk_cbk, - priv->children[i], - priv->children[i]->fops->entrylk, - volume, loc, basename, cmd, type); - - if (!--call_count) - break; - } - } - - op_ret = 0; -out: - if (op_ret == -1) { - AFR_STACK_UNWIND (entrylk, frame, op_ret, op_errno); - } - return 0; -} - - - -int32_t -afr_fentrylk_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno) - +void +afr_ta_init(afr_private_t *priv) { - afr_local_t *local = NULL; - - int call_count = -1; - - local = frame->local; - - LOCK (&frame->lock); - { - if (op_ret == 0) - local->op_ret = 0; - - local->op_errno = op_errno; - } - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); - - if (call_count == 0) - AFR_STACK_UNWIND (fentrylk, frame, local->op_ret, - local->op_errno); - - return 0; + priv->thin_arbiter_count = 1; + priv->child_count--; + priv->ta_child_up = 0; + priv->ta_bad_child_index = AFR_CHILD_UNKNOWN; + priv->ta_notify_dom_lock_offset = 0; + priv->ta_in_mem_txn_count = 0; + priv->ta_on_wire_txn_count = 0; + priv->release_ta_notify_dom_lock = _gf_false; + INIT_LIST_HEAD(&priv->ta_waitq); + INIT_LIST_HEAD(&priv->ta_onwireq); + gf_uuid_clear(priv->ta_gfid); } - int32_t -afr_fentrylk (call_frame_t *frame, xlator_t *this, - const char *volume, fd_t *fd, - const char *basename, entrylk_cmd cmd, entrylk_type type) +init(xlator_t *this) { - afr_private_t *priv = NULL; - afr_local_t *local = NULL; - - int ret = -1; - - int i = 0; - int32_t call_count = 0; - int32_t op_ret = -1; - int32_t op_errno = 0; - - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); + afr_private_t *priv = NULL; + int child_count = 0; + xlator_list_t *trav = NULL; + int i = 0; + int ret = -1; + GF_UNUSED int op_errno = 0; + xlator_t *read_subvol = NULL; + int read_subvol_index = -1; + char *qtype = NULL; + char *fav_child_policy = NULL; + char *thin_arbiter = NULL; + char *data_self_heal = NULL; + char *locking_scheme = NULL; + char *data_self_heal_algorithm = NULL; + + if (!this->children) { + gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_CHILD_MISCONFIGURED, + "replicate translator needs more than one " + "subvolume defined."); + return -1; + } + + if (!this->parents) { + gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_VOL_MISCONFIGURED, + "Volume is dangling."); + } + + this->private = GF_CALLOC(1, sizeof(afr_private_t), + gf_afr_mt_afr_private_t); + if (!this->private) + goto out; + + priv = this->private; + INIT_LIST_HEAD(&priv->saved_locks); + INIT_LIST_HEAD(&priv->lk_healq); + LOCK_INIT(&priv->lock); + + child_count = xlator_subvolume_count(this); + + priv->child_count = child_count; + + priv->read_child = -1; + + GF_OPTION_INIT("arbiter-count", priv->arbiter_count, uint32, out); + GF_OPTION_INIT("thin-arbiter", thin_arbiter, str, out); + if (thin_arbiter && strlen(thin_arbiter) > 0) { + afr_ta_init(priv); + } + INIT_LIST_HEAD(&priv->healing); + INIT_LIST_HEAD(&priv->heal_waiting); + + priv->spb_choice_timeout = AFR_DEFAULT_SPB_CHOICE_TIMEOUT; + + GF_OPTION_INIT("afr-dirty-xattr", priv->afr_dirty, str, out); + + GF_OPTION_INIT("metadata-splitbrain-forced-heal", + priv->metadata_splitbrain_forced_heal, bool, out); + + GF_OPTION_INIT("read-subvolume", read_subvol, xlator, out); + if (read_subvol) { + priv->read_child = xlator_subvolume_index(this, read_subvol); + if (priv->read_child == -1) { + gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL, + "%s not a subvolume", read_subvol->name); + goto out; + } + } + GF_OPTION_INIT("read-subvolume-index", read_subvol_index, int32, out); + if (read_subvol_index > -1) { + if (read_subvol_index >= priv->child_count) { + gf_msg(this->name, GF_LOG_ERROR, 0, AFR_MSG_INVALID_SUBVOL, + "%d not a subvolume-index", read_subvol_index); + goto out; + } + priv->read_child = read_subvol_index; + } + GF_OPTION_INIT("choose-local", priv->choose_local, bool, out); - priv = this->private; + priv->pending_reads = GF_CALLOC(sizeof(*priv->pending_reads), + priv->child_count, gf_afr_mt_atomic_t); - ALLOC_OR_GOTO (local, afr_local_t, out); + GF_OPTION_INIT("read-hash-mode", priv->hash_mode, uint32, out); - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } + priv->favorite_child = -1; - call_count = local->call_count; - frame->local = local; + GF_OPTION_INIT("favorite-child-policy", fav_child_policy, str, out); + if (afr_set_favorite_child_policy(priv, fav_child_policy) == -1) + goto out; - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND (frame, afr_fentrylk_cbk, - priv->children[i], - priv->children[i]->fops->fentrylk, - volume, fd, basename, cmd, type); + GF_OPTION_INIT("shd-max-threads", priv->shd.max_threads, uint32, out); - if (!--call_count) - break; - } - } + GF_OPTION_INIT("shd-wait-qlength", priv->shd.wait_qlength, uint32, out); - op_ret = 0; -out: - if (op_ret == -1) { - AFR_STACK_UNWIND (fentrylk, frame, op_ret, op_errno); - } - return 0; -} + GF_OPTION_INIT("background-self-heal-count", + priv->background_self_heal_count, uint32, out); + GF_OPTION_INIT("heal-wait-queue-length", priv->heal_wait_qlen, uint32, out); -int32_t -afr_checksum_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno, - uint8_t *file_checksum, uint8_t *dir_checksum) - -{ - afr_local_t *local = NULL; - - int call_count = -1; + GF_OPTION_INIT("data-self-heal", data_self_heal, str, out); + if (gf_string2boolean(data_self_heal, &priv->data_self_heal) == -1) + goto out; - local = frame->local; + GF_OPTION_INIT("data-self-heal-algorithm", data_self_heal_algorithm, str, + out); + set_data_self_heal_algorithm(priv, data_self_heal_algorithm); - LOCK (&frame->lock); - { - if (op_ret == 0 && (local->op_ret != 0)) { - local->op_ret = 0; + GF_OPTION_INIT("data-self-heal-window-size", + priv->data_self_heal_window_size, uint32, out); - local->cont.checksum.file_checksum = MALLOC (NAME_MAX); - memcpy (local->cont.checksum.file_checksum, file_checksum, - NAME_MAX); + GF_OPTION_INIT("metadata-self-heal", priv->metadata_self_heal, bool, out); - local->cont.checksum.dir_checksum = MALLOC (NAME_MAX); - memcpy (local->cont.checksum.dir_checksum, dir_checksum, - NAME_MAX); + GF_OPTION_INIT("entry-self-heal", priv->entry_self_heal, bool, out); - } + GF_OPTION_INIT("halo-shd-max-latency", priv->shd.halo_max_latency_msec, + uint32, out); - local->op_errno = op_errno; - } - UNLOCK (&frame->lock); + GF_OPTION_INIT("halo-max-latency", priv->halo_max_latency_msec, uint32, + out); + GF_OPTION_INIT("halo-max-replicas", priv->halo_max_replicas, uint32, out); + GF_OPTION_INIT("halo-min-replicas", priv->halo_min_replicas, uint32, out); - call_count = afr_frame_return (frame); + GF_OPTION_INIT("halo-enabled", priv->halo_enabled, bool, out); - if (call_count == 0) - AFR_STACK_UNWIND (checksum, frame, local->op_ret, local->op_errno, - local->cont.checksum.file_checksum, - local->cont.checksum.dir_checksum); + GF_OPTION_INIT("halo-nfsd-max-latency", priv->nfsd.halo_max_latency_msec, + uint32, out); - return 0; -} + GF_OPTION_INIT("iam-nfs-daemon", priv->nfsd.iamnfsd, bool, out); + GF_OPTION_INIT("optimistic-change-log", priv->optimistic_change_log, bool, + out); -int32_t -afr_checksum (call_frame_t *frame, xlator_t *this, loc_t *loc, - int32_t flag) -{ - afr_private_t *priv = NULL; - afr_local_t *local = NULL; + GF_OPTION_INIT("pre-op-compat", priv->pre_op_compat, bool, out); + GF_OPTION_INIT("locking-scheme", locking_scheme, str, out); + priv->granular_locks = (strcmp(locking_scheme, "granular") == 0); + GF_OPTION_INIT("full-lock", priv->full_lock, bool, out); + GF_OPTION_INIT("granular-entry-heal", priv->esh_granular, bool, out); - int ret = -1; + GF_OPTION_INIT("eager-lock", priv->eager_lock, bool, out); + GF_OPTION_INIT("quorum-type", qtype, str, out); + GF_OPTION_INIT("quorum-count", priv->quorum_count, uint32, out); + GF_OPTION_INIT(AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, size_uint64, + out); + fix_quorum_options(this, priv, qtype, this->options); - int i = 0; - int32_t call_count = 0; - int32_t op_ret = -1; - int32_t op_errno = 0; + GF_OPTION_INIT("post-op-delay-secs", priv->post_op_delay_secs, uint32, out); + GF_OPTION_INIT("ensure-durability", priv->ensure_durability, bool, out); - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); + GF_OPTION_INIT("self-heal-daemon", priv->shd.enabled, bool, out); - priv = this->private; + GF_OPTION_INIT("iam-self-heal-daemon", priv->shd.iamshd, bool, out); + GF_OPTION_INIT("heal-timeout", priv->shd.timeout, int32, out); - ALLOC_OR_GOTO (local, afr_local_t, out); + GF_OPTION_INIT("consistent-metadata", priv->consistent_metadata, bool, out); + GF_OPTION_INIT("consistent-io", priv->consistent_io, bool, out); + afr_handle_anon_inode_options(priv, this->options); - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } + GF_OPTION_INIT("use-anonymous-inode", priv->use_anon_inode, bool, out); + if (priv->quorum_count != 0) + priv->consistent_io = _gf_false; - call_count = local->call_count; - frame->local = local; + priv->wait_count = 1; - for (i = 0; i < priv->child_count; i++) { - if (local->child_up[i]) { - STACK_WIND (frame, afr_checksum_cbk, - priv->children[i], - priv->children[i]->fops->checksum, - loc, flag); + priv->local = GF_CALLOC(sizeof(unsigned char), child_count, gf_afr_mt_char); + if (!priv->local) { + ret = -ENOMEM; + goto out; + } - if (!--call_count) - break; - } - } + priv->anon_inode = GF_CALLOC(sizeof(unsigned char), child_count, + gf_afr_mt_char); - op_ret = 0; -out: - if (op_ret == -1) { - AFR_STACK_UNWIND (checksum, frame, op_ret, op_errno, - NULL, NULL); - } - return 0; -} + priv->child_up = GF_CALLOC(sizeof(unsigned char), child_count, + gf_afr_mt_char); + priv->child_latency = GF_MALLOC(sizeof(*priv->child_latency) * child_count, + gf_afr_mt_child_latency_t); + priv->halo_child_up = GF_CALLOC(sizeof(unsigned char), child_count, + gf_afr_mt_char); -int32_t -afr_statfs_cbk (call_frame_t *frame, void *cookie, - xlator_t *this, int32_t op_ret, int32_t op_errno, - struct statvfs *statvfs) -{ - afr_local_t *local = NULL; - - int call_count = 0; - - LOCK (&frame->lock); - { - local = frame->local; - - if (op_ret == 0) { - local->op_ret = op_ret; - - if (local->cont.statfs.buf_set) { - if (statvfs->f_bavail < local->cont.statfs.buf.f_bavail) - local->cont.statfs.buf = *statvfs; - } else { - local->cont.statfs.buf = *statvfs; - local->cont.statfs.buf_set = 1; - } - } - - if (op_ret == -1) - local->op_errno = op_errno; - - } - UNLOCK (&frame->lock); - - call_count = afr_frame_return (frame); - - if (call_count == 0) - AFR_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno, - &local->cont.statfs.buf); + if (!priv->child_up || !priv->child_latency || !priv->halo_child_up || + !priv->anon_inode) { + ret = -ENOMEM; + goto out; + } + /*Initialize to -ve ping timeout so that they are not considered + * in child-up events until ping-event comes*/ + for (i = 0; i < child_count; i++) + priv->child_latency[i] = -1; + + priv->children = GF_CALLOC(sizeof(xlator_t *), child_count, + gf_afr_mt_xlator_t); + if (!priv->children) { + ret = -ENOMEM; + goto out; + } + + ret = afr_pending_xattrs_init(priv, this); + if (ret) + goto out; + + trav = this->children; + i = 0; + while (i < child_count) { + priv->children[i] = trav->xlator; + trav = trav->next; + i++; + } + + ret = gf_asprintf(&priv->sh_domain, AFR_SH_DATA_DOMAIN_FMT, this->name); + if (-1 == ret) { + ret = -ENOMEM; + goto out; + } + + priv->last_event = GF_CALLOC(child_count, sizeof(*priv->last_event), + gf_afr_mt_int32_t); + if (!priv->last_event) { + ret = -ENOMEM; + goto out; + } + + this->itable = inode_table_new(SHD_INODE_LRU_LIMIT, this); + if (!this->itable) { + ret = -ENOMEM; + goto out; + } + + if (priv->shd.iamshd) { + ret = afr_selfheal_daemon_init(this); + if (ret) { + ret = -ENOMEM; + goto out; + } + } - return 0; -} + /* keep more local here as we may need them for self-heal etc */ + this->local_pool = mem_pool_new(afr_local_t, 512); + if (!this->local_pool) { + ret = -1; + goto out; + } + priv->root_inode = NULL; -int32_t -afr_statfs (call_frame_t *frame, xlator_t *this, - loc_t *loc) -{ - afr_private_t * priv = NULL; - int child_count = 0; - afr_local_t * local = NULL; - int i = 0; - - int ret = -1; - int call_count = 0; - int32_t op_ret = -1; - int32_t op_errno = 0; - - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); - VALIDATE_OR_GOTO (loc, out); - - priv = this->private; - child_count = priv->child_count; - - ALLOC_OR_GOTO (local, afr_local_t, out); - - ret = AFR_LOCAL_INIT (local, priv); - if (ret < 0) { - op_errno = -ret; - goto out; - } - - frame->local = local; - call_count = local->call_count; - - for (i = 0; i < child_count; i++) { - if (local->child_up[i]) { - STACK_WIND (frame, afr_statfs_cbk, - priv->children[i], - priv->children[i]->fops->statfs, - loc); - if (!--call_count) - break; - } - } - - op_ret = 0; + ret = 0; out: - if (op_ret == -1) { - AFR_STACK_UNWIND (statfs, frame, op_ret, op_errno, NULL); - } - return 0; -} - - -int32_t -afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this, - int32_t op_ret, int32_t op_errno, struct flock *lock) -{ - afr_local_t * local = NULL; - - int call_count = -1; - - local = frame->local; - call_count = afr_frame_return (frame); - - if (call_count == 0) - AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno, - lock); - - return 0; -} - - -int32_t -afr_lk_unlock (call_frame_t *frame, xlator_t *this) -{ - afr_local_t * local = NULL; - afr_private_t * priv = NULL; - - int i; - int call_count = 0; - - local = frame->local; - priv = this->private; - - call_count = afr_locked_nodes_count (local->cont.lk.locked_nodes, - priv->child_count); - - if (call_count == 0) { - AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno, - &local->cont.lk.flock); - return 0; - } - - local->call_count = call_count; - - local->cont.lk.flock.l_type = F_UNLCK; - - for (i = 0; i < priv->child_count; i++) { - if (local->cont.lk.locked_nodes[i]) { - STACK_WIND (frame, afr_lk_unlock_cbk, - priv->children[i], - priv->children[i]->fops->lk, - local->fd, F_SETLK, - &local->cont.lk.flock); - - if (!--call_count) - break; - } - } - - return 0; + return ret; } - - -int32_t -afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this, - int32_t op_ret, int32_t op_errno, struct flock *lock) -{ - afr_local_t *local = NULL; - afr_private_t *priv = NULL; - - int call_count = -1; - int child_index = -1; - - local = frame->local; - priv = this->private; - - child_index = (long) cookie; - - call_count = --local->call_count; - - if (!child_went_down (op_ret, op_errno) && (op_ret == -1)) { - local->op_ret = -1; - local->op_errno = op_errno; - - afr_lk_unlock (frame, this); - return 0; - } - - if (op_ret == 0) { - local->op_ret = 0; - local->op_errno = 0; - local->cont.lk.locked_nodes[child_index] = 1; - } - - child_index++; - - if (child_index < priv->child_count) { - STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) child_index, - priv->children[child_index], - priv->children[child_index]->fops->lk, - local->fd, local->cont.lk.cmd, - &local->cont.lk.flock); - } else if (local->op_ret == -1) { - /* all nodes have gone down */ - - AFR_STACK_UNWIND (lk, frame, -1, ENOTCONN, &local->cont.lk.flock); - } else { - /* locking has succeeded on all nodes that are up */ - - AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno, - &local->cont.lk.flock); - } - - return 0; -} - - -int -afr_lk (call_frame_t *frame, xlator_t *this, - fd_t *fd, int32_t cmd, - struct flock *flock) +void +afr_destroy_healer_object(xlator_t *this, struct subvol_healer *healer) { - afr_private_t *priv = NULL; - afr_local_t *local = NULL; - - int i = 0; + int ret = -1; - int32_t op_ret = -1; - int32_t op_errno = 0; - - VALIDATE_OR_GOTO (frame, out); - VALIDATE_OR_GOTO (this, out); - VALIDATE_OR_GOTO (this->private, out); - - priv = this->private; - - ALLOC_OR_GOTO (local, afr_local_t, out); - AFR_LOCAL_INIT (local, priv); - - frame->local = local; - - local->cont.lk.locked_nodes = CALLOC (priv->child_count, - sizeof (*local->cont.lk.locked_nodes)); - - if (!local->cont.lk.locked_nodes) { - gf_log (this->name, GF_LOG_ERROR, "Out of memory"); - op_errno = ENOMEM; - goto out; - } - - local->fd = fd_ref (fd); - local->cont.lk.cmd = cmd; - local->cont.lk.flock = *flock; - - STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) 0, - priv->children[i], - priv->children[i]->fops->lk, - fd, cmd, flock); - - op_ret = 0; -out: - if (op_ret == -1) { - AFR_STACK_UNWIND (lk, frame, op_ret, op_errno, NULL); - } - return 0; -} + if (!healer) + return; -int -afr_priv_dump (xlator_t *this) -{ - afr_private_t *priv = NULL; - char key_prefix[GF_DUMP_MAX_BUF_LEN]; - char key[GF_DUMP_MAX_BUF_LEN]; - int i = 0; - - - assert(this); - priv = this->private; - - assert(priv); - snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name); - gf_proc_dump_add_section(key_prefix); - gf_proc_dump_build_key(key, key_prefix, "child_count"); - gf_proc_dump_write(key, "%u", priv->child_count); - gf_proc_dump_build_key(key, key_prefix, "read_child_rr"); - gf_proc_dump_write(key, "%u", priv->read_child_rr); - for (i = 0; i < priv->child_count; i++) { - gf_proc_dump_build_key(key, key_prefix, "child_up[%d]", i); - gf_proc_dump_write(key, "%d", priv->child_up[i]); - gf_proc_dump_build_key(key, key_prefix, - "pending_key[%d]", i); - gf_proc_dump_write(key, "%s", priv->pending_key[i]); - } - gf_proc_dump_build_key(key, key_prefix, "data_self_heal"); - gf_proc_dump_write(key, "%d", priv->data_self_heal); - gf_proc_dump_build_key(key, key_prefix, "metadata_self_heal"); - gf_proc_dump_write(key, "%d", priv->metadata_self_heal); - gf_proc_dump_build_key(key, key_prefix, "entry_self_heal"); - gf_proc_dump_write(key, "%d", priv->entry_self_heal); - gf_proc_dump_build_key(key, key_prefix, "data_change_log"); - gf_proc_dump_write(key, "%d", priv->data_change_log); - gf_proc_dump_build_key(key, key_prefix, "metadata_change_log"); - gf_proc_dump_write(key, "%d", priv->metadata_change_log); - gf_proc_dump_build_key(key, key_prefix, "entry_change_log"); - gf_proc_dump_write(key, "%d", priv->entry_change_log); - gf_proc_dump_build_key(key, key_prefix, "read_child"); - gf_proc_dump_write(key, "%d", priv->read_child); - gf_proc_dump_build_key(key, key_prefix, "favorite_child"); - gf_proc_dump_write(key, "%u", priv->favorite_child); - gf_proc_dump_build_key(key, key_prefix, "data_lock_server_count"); - gf_proc_dump_write(key, "%u", priv->data_lock_server_count); - gf_proc_dump_build_key(key, key_prefix, "metadata_lock_server_count"); - gf_proc_dump_write(key, "%u", priv->metadata_lock_server_count); - gf_proc_dump_build_key(key, key_prefix, "entry_lock_server_count"); - gf_proc_dump_write(key, "%u", priv->entry_lock_server_count); - gf_proc_dump_build_key(key, key_prefix, "wait_count"); - gf_proc_dump_write(key, "%u", priv->wait_count); - - return 0; + if (healer->running) { + /* + * If there are any resources to cleanup, We need + * to do that gracefully using pthread_cleanup_push + */ + ret = gf_thread_cleanup_xint(healer->thread); + if (ret) + gf_msg(this->name, GF_LOG_WARNING, 0, AFR_MSG_SELF_HEAL_FAILED, + "Failed to clean up healer threads."); + healer->thread = 0; + } + pthread_cond_destroy(&healer->cond); + pthread_mutex_destroy(&healer->mutex); } - -/** - * find_child_index - find the child's index in the array of subvolumes - * @this: AFR - * @child: child - */ - -static int -find_child_index (xlator_t *this, xlator_t *child) +void +afr_selfheal_daemon_fini(xlator_t *this) { - afr_private_t *priv = NULL; + struct subvol_healer *healer = NULL; + afr_self_heald_t *shd = NULL; + afr_private_t *priv = NULL; + int i = 0; - int i = -1; - - priv = this->private; + priv = this->private; + if (!priv) + return; - for (i = 0; i < priv->child_count; i++) { - if ((xlator_t *) child == priv->children[i]) - break; - } + shd = &priv->shd; + if (!shd->iamshd) + return; - return i; + for (i = 0; i < priv->child_count; i++) { + healer = &shd->index_healers[i]; + afr_destroy_healer_object(this, healer); + + healer = &shd->full_healers[i]; + afr_destroy_healer_object(this, healer); + + if (shd->statistics[i]) + eh_destroy(shd->statistics[i]); + } + GF_FREE(shd->index_healers); + GF_FREE(shd->full_healers); + GF_FREE(shd->statistics); + if (shd->split_brain) + eh_destroy(shd->split_brain); } - - -int32_t -notify (xlator_t *this, int32_t event, - void *data, ...) +void +fini(xlator_t *this) { - afr_private_t * priv = NULL; - unsigned char * child_up = NULL; - - int i = -1; - int up_children = 0; - - priv = this->private; + afr_private_t *priv = NULL; - if (!priv) - return 0; + priv = this->private; - child_up = priv->child_up; + afr_selfheal_daemon_fini(this); + GF_ASSERT(list_empty(&priv->saved_locks)); - switch (event) { - case GF_EVENT_CHILD_UP: - i = find_child_index (this, data); + LOCK(&priv->lock); + if (priv->timer != NULL) { + gf_timer_call_cancel(this->ctx, priv->timer); + priv->timer = NULL; + } + UNLOCK(&priv->lock); - child_up[i] = 1; + if (this->local_pool != NULL) { + mem_pool_destroy(this->local_pool); + this->local_pool = NULL; + } - LOCK (&priv->lock); - { - priv->up_count++; - } - UNLOCK (&priv->lock); + this->private = NULL; + afr_priv_destroy(priv); + if (this->itable) { + inode_table_destroy(this->itable); + this->itable = NULL; + } - /* - if all the children were down, and one child came up, - send notify to parent - */ - - for (i = 0; i < priv->child_count; i++) - if (child_up[i]) - up_children++; - - if (up_children == 1) { - gf_log (this->name, GF_LOG_NORMAL, - "Subvolume '%s' came back up; " - "going online.", ((xlator_t *)data)->name); - - default_notify (this, event, data); - } - - break; - - case GF_EVENT_CHILD_DOWN: - i = find_child_index (this, data); - - child_up[i] = 0; - - LOCK (&priv->lock); - { - priv->down_count++; - } - UNLOCK (&priv->lock); - - /* - if all children are down, and this was the last to go down, - send notify to parent - */ - - for (i = 0; i < priv->child_count; i++) - if (child_up[i]) - up_children++; - - if (up_children == 0) { - gf_log (this->name, GF_LOG_ERROR, - "All subvolumes are down. Going offline " - "until atleast one of them comes back up."); - - default_notify (this, event, data); - } - - break; - - default: - default_notify (this, event, data); - } - - return 0; -} - - -static const char *favorite_child_warning_str = "You have specified subvolume '%s' " - "as the 'favorite child'. This means that if a discrepancy in the content " - "or attributes (ownership, permission, etc.) of a file is detected among " - "the subvolumes, the file on '%s' will be considered the definitive " - "version and its contents will OVERWRITE the contents of the file on other " - "subvolumes. All versions of the file except that on '%s' " - "WILL BE LOST."; - -static const char *no_lock_servers_warning_str = "You have set lock-server-count = 0. " - "This means correctness is NO LONGER GUARANTEED in all cases. If two or more " - "applications write to the same region of a file, there is a possibility that " - "its copies will be INCONSISTENT. Set it to a value greater than 0 unless you " - "are ABSOLUTELY SURE of what you are doing and WILL NOT HOLD GlusterFS " - "RESPONSIBLE for inconsistent data. If you are in doubt, set it to a value " - "greater than 0."; - -int32_t -init (xlator_t *this) -{ - afr_private_t * priv = NULL; - int child_count = 0; - xlator_list_t * trav = NULL; - int i = 0; - int ret = -1; - int op_errno = 0; - - char * read_subvol = NULL; - char * fav_child = NULL; - char * self_heal = NULL; - char * algo = NULL; - char * change_log = NULL; - - int32_t background_count = 0; - int32_t lock_server_count = 1; - int32_t window_size; - - int fav_ret = -1; - int read_ret = -1; - int dict_ret = -1; - - if (!this->children) { - gf_log (this->name, GF_LOG_ERROR, - "replicate translator needs more than one " - "subvolume defined."); - return -1; - } - - if (!this->parents) { - gf_log (this->name, GF_LOG_WARNING, - "Volume is dangling."); - } - - ALLOC_OR_GOTO (this->private, afr_private_t, out); - - priv = this->private; - - read_ret = dict_get_str (this->options, "read-subvolume", &read_subvol); - priv->read_child = -1; - - fav_ret = dict_get_str (this->options, "favorite-child", &fav_child); - priv->favorite_child = -1; - - priv->background_self_heal_count = 16; - - dict_ret = dict_get_int32 (this->options, "background-self-heal-count", - &background_count); - if (dict_ret == 0) { - gf_log (this->name, GF_LOG_DEBUG, - "Setting background self-heal count to %d.", - window_size); - - priv->background_self_heal_count = background_count; - } - - /* Default values */ - - priv->data_self_heal = 1; - priv->metadata_self_heal = 1; - priv->entry_self_heal = 1; - - dict_ret = dict_get_str (this->options, "data-self-heal", &self_heal); - if (dict_ret == 0) { - ret = gf_string2boolean (self_heal, &priv->data_self_heal); - if (ret < 0) { - gf_log (this->name, GF_LOG_WARNING, - "Invalid 'option data-self-heal %s'. " - "Defaulting to data-self-heal as 'on'", - self_heal); - priv->data_self_heal = 1; - } - } - - priv->data_self_heal_algorithm = ""; - - dict_ret = dict_get_str (this->options, "data-self-heal-algorithm", - &algo); - if (dict_ret == 0) { - priv->data_self_heal_algorithm = strdup (algo); - } - - - priv->data_self_heal_window_size = 16; - - dict_ret = dict_get_int32 (this->options, "data-self-heal-window-size", - &window_size); - if (dict_ret == 0) { - gf_log (this->name, GF_LOG_DEBUG, - "Setting data self-heal window size to %d.", - window_size); - - priv->data_self_heal_window_size = window_size; - } - - dict_ret = dict_get_str (this->options, "metadata-self-heal", - &self_heal); - if (dict_ret == 0) { - ret = gf_string2boolean (self_heal, &priv->metadata_self_heal); - if (ret < 0) { - gf_log (this->name, GF_LOG_WARNING, - "Invalid 'option metadata-self-heal %s'. " - "Defaulting to metadata-self-heal as 'on'.", - self_heal); - priv->metadata_self_heal = 1; - } - } - - dict_ret = dict_get_str (this->options, "entry-self-heal", &self_heal); - if (dict_ret == 0) { - ret = gf_string2boolean (self_heal, &priv->entry_self_heal); - if (ret < 0) { - gf_log (this->name, GF_LOG_WARNING, - "Invalid 'option entry-self-heal %s'. " - "Defaulting to entry-self-heal as 'on'.", - self_heal); - priv->entry_self_heal = 1; - } - } - - /* Change log options */ - - priv->data_change_log = 1; - priv->metadata_change_log = 0; - priv->entry_change_log = 1; - - dict_ret = dict_get_str (this->options, "data-change-log", - &change_log); - if (dict_ret == 0) { - ret = gf_string2boolean (change_log, &priv->data_change_log); - if (ret < 0) { - gf_log (this->name, GF_LOG_WARNING, - "Invalid 'option data-change-log %s'. " - "Defaulting to data-change-log as 'on'.", - change_log); - priv->data_change_log = 1; - } - } - - dict_ret = dict_get_str (this->options, "metadata-change-log", - &change_log); - if (dict_ret == 0) { - ret = gf_string2boolean (change_log, - &priv->metadata_change_log); - if (ret < 0) { - gf_log (this->name, GF_LOG_WARNING, - "Invalid 'option metadata-change-log %s'. " - "Defaulting to metadata-change-log as 'off'.", - change_log); - priv->metadata_change_log = 0; - } - } - - dict_ret = dict_get_str (this->options, "entry-change-log", - &change_log); - if (dict_ret == 0) { - ret = gf_string2boolean (change_log, &priv->entry_change_log); - if (ret < 0) { - gf_log (this->name, GF_LOG_WARNING, - "Invalid 'option entry-change-log %s'. " - "Defaulting to entry-change-log as 'on'.", - change_log); - priv->entry_change_log = 1; - } - } - - /* Locking options */ - - priv->data_lock_server_count = 1; - priv->metadata_lock_server_count = 0; - priv->entry_lock_server_count = 1; - - dict_ret = dict_get_int32 (this->options, "data-lock-server-count", - &lock_server_count); - if (dict_ret == 0) { - gf_log (this->name, GF_LOG_DEBUG, - "Setting data lock server count to %d.", - lock_server_count); - - if (lock_server_count == 0) - gf_log (this->name, GF_LOG_WARNING, "%s", - no_lock_servers_warning_str); - - priv->data_lock_server_count = lock_server_count; - } - - - dict_ret = dict_get_int32 (this->options, - "metadata-lock-server-count", - &lock_server_count); - if (dict_ret == 0) { - gf_log (this->name, GF_LOG_DEBUG, - "Setting metadata lock server count to %d.", - lock_server_count); - priv->metadata_lock_server_count = lock_server_count; - } - - - dict_ret = dict_get_int32 (this->options, "entry-lock-server-count", - &lock_server_count); - if (dict_ret == 0) { - gf_log (this->name, GF_LOG_DEBUG, - "Setting entry lock server count to %d.", - lock_server_count); - - priv->entry_lock_server_count = lock_server_count; - } - - trav = this->children; - while (trav) { - if (!read_ret && !strcmp (read_subvol, trav->xlator->name)) { - gf_log (this->name, GF_LOG_DEBUG, - "Subvolume '%s' specified as read child.", - trav->xlator->name); - - priv->read_child = child_count; - } - - if (fav_ret == 0 && !strcmp (fav_child, trav->xlator->name)) { - gf_log (this->name, GF_LOG_WARNING, - favorite_child_warning_str, trav->xlator->name, - trav->xlator->name, trav->xlator->name); - priv->favorite_child = child_count; - } - - child_count++; - trav = trav->next; - } - - priv->wait_count = 1; - - priv->child_count = child_count; - - LOCK_INIT (&priv->lock); - LOCK_INIT (&priv->read_child_lock); - - priv->child_up = CALLOC (sizeof (unsigned char), child_count); - if (!priv->child_up) { - gf_log (this->name, GF_LOG_ERROR, - "Out of memory."); - op_errno = ENOMEM; - goto out; - } - - priv->children = CALLOC (sizeof (xlator_t *), child_count); - if (!priv->children) { - gf_log (this->name, GF_LOG_ERROR, - "Out of memory."); - op_errno = ENOMEM; - goto out; - } - - priv->pending_key = CALLOC (sizeof (*priv->pending_key), child_count); - if (!priv->pending_key) { - gf_log (this->name, GF_LOG_ERROR, - "Out of memory."); - op_errno = ENOMEM; - goto out; - } - - trav = this->children; - i = 0; - while (i < child_count) { - priv->children[i] = trav->xlator; - - ret = asprintf (&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX, - trav->xlator->name); - if (-1 == ret) { - gf_log (this->name, GF_LOG_ERROR, - "asprintf failed to set pending key"); - op_errno = ENOMEM; - goto out; - } - - trav = trav->next; - i++; - } - - ret = 0; -out: - return ret; -} - - -int -fini (xlator_t *this) -{ - return 0; + return; } - struct xlator_fops fops = { - .lookup = afr_lookup, - .open = afr_open, - .lk = afr_lk, - .flush = afr_flush, - .statfs = afr_statfs, - .fsync = afr_fsync, - .fsyncdir = afr_fsyncdir, - .xattrop = afr_xattrop, - .fxattrop = afr_fxattrop, - .inodelk = afr_inodelk, - .finodelk = afr_finodelk, - .entrylk = afr_entrylk, - .fentrylk = afr_fentrylk, - .checksum = afr_checksum, - - /* inode read */ - .access = afr_access, - .stat = afr_stat, - .fstat = afr_fstat, - .readlink = afr_readlink, - .getxattr = afr_getxattr, - .readv = afr_readv, - - /* inode write */ - .writev = afr_writev, - .truncate = afr_truncate, - .ftruncate = afr_ftruncate, - .setxattr = afr_setxattr, - .setattr = afr_setattr, - .fsetattr = afr_fsetattr, - .removexattr = afr_removexattr, - - /* dir read */ - .opendir = afr_opendir, - .readdir = afr_readdir, - .readdirp = afr_readdirp, - .getdents = afr_getdents, - - /* dir write */ - .create = afr_create, - .mknod = afr_mknod, - .mkdir = afr_mkdir, - .unlink = afr_unlink, - .rmdir = afr_rmdir, - .link = afr_link, - .symlink = afr_symlink, - .rename = afr_rename, - .setdents = afr_setdents, -}; - - -struct xlator_mops mops = { + .lookup = afr_lookup, + .lk = afr_lk, + .flush = afr_flush, + .statfs = afr_statfs, + .fsyncdir = afr_fsyncdir, + .inodelk = afr_inodelk, + .finodelk = afr_finodelk, + .entrylk = afr_entrylk, + .fentrylk = afr_fentrylk, + .ipc = afr_ipc, + .lease = afr_lease, + + /* inode read */ + .access = afr_access, + .stat = afr_stat, + .fstat = afr_fstat, + .readlink = afr_readlink, + .getxattr = afr_getxattr, + .fgetxattr = afr_fgetxattr, + .readv = afr_readv, + .seek = afr_seek, + + /* inode write */ + .writev = afr_writev, + .truncate = afr_truncate, + .ftruncate = afr_ftruncate, + .setxattr = afr_setxattr, + .fsetxattr = afr_fsetxattr, + .setattr = afr_setattr, + .fsetattr = afr_fsetattr, + .removexattr = afr_removexattr, + .fremovexattr = afr_fremovexattr, + .fallocate = afr_fallocate, + .discard = afr_discard, + .zerofill = afr_zerofill, + .xattrop = afr_xattrop, + .fxattrop = afr_fxattrop, + .fsync = afr_fsync, + + /*inode open*/ + .opendir = afr_opendir, + .open = afr_open, + + /* dir read */ + .readdir = afr_readdir, + .readdirp = afr_readdirp, + + /* dir write */ + .create = afr_create, + .mknod = afr_mknod, + .mkdir = afr_mkdir, + .unlink = afr_unlink, + .rmdir = afr_rmdir, + .link = afr_link, + .symlink = afr_symlink, + .rename = afr_rename, }; struct xlator_dumpops dumpops = { - .priv = afr_priv_dump, + .priv = afr_priv_dump, }; - struct xlator_cbks cbks = { - .release = afr_release, + .release = afr_release, + .releasedir = afr_releasedir, + .forget = afr_forget, }; - struct volume_options options[] = { - { .key = {"read-subvolume" }, - .type = GF_OPTION_TYPE_XLATOR - }, - { .key = {"favorite-child"}, - .type = GF_OPTION_TYPE_XLATOR - }, - { .key = {"background-self-heal-count"}, - .type = GF_OPTION_TYPE_INT, - .min = 0 - }, - { .key = {"data-self-heal"}, - .type = GF_OPTION_TYPE_BOOL - }, - { .key = {"data-self-heal-algorithm"}, - .type = GF_OPTION_TYPE_STR - }, - { .key = {"data-self-heal-window-size"}, - .type = GF_OPTION_TYPE_INT, - .min = 1, - .max = 1024 - }, - { .key = {"metadata-self-heal"}, - .type = GF_OPTION_TYPE_BOOL - }, - { .key = {"entry-self-heal"}, - .type = GF_OPTION_TYPE_BOOL - }, - { .key = {"data-change-log"}, - .type = GF_OPTION_TYPE_BOOL - }, - { .key = {"metadata-change-log"}, - .type = GF_OPTION_TYPE_BOOL - }, - { .key = {"entry-change-log"}, - .type = GF_OPTION_TYPE_BOOL - }, - { .key = {"data-lock-server-count"}, - .type = GF_OPTION_TYPE_INT, - .min = 0 - }, - { .key = {"metadata-lock-server-count"}, - .type = GF_OPTION_TYPE_INT, - .min = 0 - }, - { .key = {"entry-lock-server-count"}, - .type = GF_OPTION_TYPE_INT, - .min = 0 - }, - { .key = {NULL} }, + {.key = {"read-subvolume"}, + .type = GF_OPTION_TYPE_XLATOR, + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "inode-read fops happen only on one of the bricks in " + "replicate. Afr will prefer the one specified using " + "this option if it is not stale. Option value must be " + "one of the xlator names of the children. " + "Ex: <volname>-client-0 till " + "<volname>-client-<number-of-bricks - 1>"}, + {.key = {"read-subvolume-index"}, + .type = GF_OPTION_TYPE_INT, + .default_value = "-1", + .op_version = {2}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "inode-read fops happen only on one of the bricks in " + "replicate. AFR will prefer the one specified using " + "this option if it is not stale. allowed options" + " include -1 till replica-count - 1"}, + {.key = {"read-hash-mode"}, + .type = GF_OPTION_TYPE_INT, + .min = 0, + .max = 5, + .default_value = "1", + .op_version = {2}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = + "inode-read fops happen only on one of the bricks in " + "replicate. AFR will prefer the one computed using " + "the method specified using this option.\n" + "0 = first readable child of AFR, starting from 1st child.\n" + "1 = hash by GFID of file (all clients use " + "same subvolume).\n" + "2 = hash by GFID of file and client PID.\n" + "3 = brick having the least outstanding read requests.\n" + "4 = brick having the least network ping latency.\n" + "5 = Hybrid mode between 3 and 4, ie least value among " + "network-latency multiplied by outstanding-read-requests."}, + { + .key = {"choose-local"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "true", + .op_version = {2}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "Choose a local subvolume (i.e. Brick) to read from" + " if read-subvolume is not explicitly set.", + }, + {.key = {"background-self-heal-count"}, + .type = GF_OPTION_TYPE_INT, + .min = 0, + .max = 256, + .default_value = "8", + .validate = GF_OPT_VALIDATE_MIN, + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "This specifies the number of per client self-heal " + "jobs that can perform parallel heals in the " + "background."}, + {.key = {"halo-shd-max-latency"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = 99999, + .default_value = "99999", + .op_version = {GD_OP_VERSION_3_11_0}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate", "halo"}, + .description = "Maximum latency for shd halo replication in msec."}, + {.key = {"halo-enabled"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "False", + .op_version = {GD_OP_VERSION_3_11_0}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate", "halo"}, + .description = "Enable Halo (geo) replication mode."}, + {.key = {"halo-nfsd-max-latency"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = 99999, + .default_value = "5", + .op_version = {GD_OP_VERSION_3_11_0}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate", "halo"}, + .description = "Maximum latency for nfsd halo replication in msec."}, + {.key = {"halo-max-latency"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = AFR_HALO_MAX_LATENCY, + .default_value = "5", + .op_version = {GD_OP_VERSION_3_11_0}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate", "halo"}, + .description = "Maximum latency for halo replication in msec."}, + {.key = {"halo-max-replicas"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = 99999, + .default_value = "99999", + .op_version = {GD_OP_VERSION_3_11_0}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate", "halo"}, + .description = "The maximum number of halo replicas; replicas" + " beyond this value will be written asynchronously" + "via the SHD."}, + {.key = {"halo-min-replicas"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = 99999, + .default_value = "2", + .op_version = {GD_OP_VERSION_3_11_0}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate", "halo"}, + .description = "The minimmum number of halo replicas, before adding " + "out of region replicas."}, + {.key = {"heal-wait-queue-length"}, + .type = GF_OPTION_TYPE_INT, + .min = 0, + .max = 10000, /*Around 100MB with sizeof(afr_local_t)= 10496 bytes*/ + .default_value = "128", + .validate = GF_OPT_VALIDATE_MIN, + .op_version = {GD_OP_VERSION_3_7_10}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "This specifies the number of heals that can be queued" + " for the parallel background self heal jobs."}, + {.key = {"data-self-heal"}, + .type = GF_OPTION_TYPE_STR, + .value = {"1", "on", "yes", "true", "enable", "0", "off", "no", "false", + "disable", "open"}, + .default_value = "off", + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "Using this option we can enable/disable data " + "self-heal on the file. \"open\" means data " + "self-heal action will only be triggered by file " + "open operations."}, + {.key = {"data-self-heal-algorithm"}, + .type = GF_OPTION_TYPE_STR, + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "Select between \"full\", \"diff\". The " + "\"full\" algorithm copies the entire file from " + "source to sink. The \"diff\" algorithm copies to " + "sink only those blocks whose checksums don't match " + "with those of source. If no option is configured " + "the option is chosen dynamically as follows: " + "If the file does not exist on one of the sinks " + "or empty file exists or if the source file size is " + "about the same as page size the entire file will " + "be read and written i.e \"full\" algo, " + "otherwise \"diff\" algo is chosen.", + .value = {"diff", "full"}}, + {.key = {"data-self-heal-window-size"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = 1024, + .default_value = "1", + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "Maximum number blocks per file for which self-heal " + "process would be applied simultaneously."}, + {.key = {"metadata-self-heal"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + /*.validate_fn = validate_replica*/ + .description = "Using this option we can enable/disable metadata " + "i.e. Permissions, ownerships, xattrs self-heal on " + "the file/directory."}, + {.key = {"entry-self-heal"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + /*.validate_fn = validate_replica*/ + .description = "Using this option we can enable/disable entry " + "self-heal on the directory."}, + {.key = {"data-change-log"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "This option exists only for backward compatibility " + "and configuring it doesn't have any effect"}, + {.key = {"metadata-change-log"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "This option exists only for backward compatibility " + "and configuring it doesn't have any effect"}, + {.key = {"entry-change-log"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "This option exists only for backward compatibility " + "and configuring it doesn't have any effect"}, + {.key = {"optimistic-change-log"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "Entry/Metadata fops will not perform " + "pre fop changelog operations in afr transaction " + "if this option is enabled."}, + {.key = {"inodelk-trace"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "Enabling this option logs inode lock/unlocks"}, + {.key = {"entrylk-trace"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "Enabling this option logs entry lock/unlocks"}, + {.key = {"pre-op-compat"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .description = "Use separate pre-op xattrop() FOP rather than " + "overloading xdata of the OP"}, + {.key = {"eager-lock"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = + "Enable/Disable eager lock for replica volume. " + "Lock phase of a transaction has two sub-phases. " + "First is an attempt to acquire locks in parallel by " + "broadcasting non-blocking lock requests. If lock " + "acquisition fails on any server, then the held locks " + "are unlocked and we revert to a blocking locks mode " + "sequentially on one server after another. If this " + "option is enabled the initial broadcasting lock " + "request attempts to acquire a full lock on the entire file. " + "If this fails, we revert back to the sequential " + "\"regional\" blocking locks as before. In the case " + "where such an \"eager\" lock is granted in the " + "non-blocking phase, it gives rise to an opportunity " + "for optimization. i.e, if the next write transaction " + "on the same FD arrives before the unlock phase of " + "the first transaction, it \"takes over\" the full " + "file lock. Similarly if yet another data transaction " + "arrives before the unlock phase of the \"optimized\" " + "transaction, that in turn \"takes over\" the lock as " + "well. The actual unlock now happens at the end of " + "the last \"optimized\" transaction." + + }, + {.key = {"self-heal-daemon"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "on", + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE, + .tags = {"replicate"}, + /*.validate_fn = validate_replica_heal_enable_disable*/ + .description = "This option applies to only self-heal-daemon. " + "Index directory crawl and automatic healing of files " + "will not be performed if this option is turned off."}, + {.key = {"iam-self-heal-daemon"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "This option differentiates if the replicate " + "translator is running as part of self-heal-daemon " + "or not."}, + {.key = {"iam-nfs-daemon"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + .description = "This option differentiates if the replicate " + "translator is running as part of an NFS daemon " + "or not."}, + { + .key = {"quorum-type"}, + .type = GF_OPTION_TYPE_STR, + .value = {"none", "auto", "fixed"}, + .default_value = "none", + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + /*.option = quorum-type*/ + .description = "If value is \"fixed\" only allow writes if " + "quorum-count bricks are present. If value is " + "\"auto\" only allow writes if more than half of " + "bricks, or exactly half including the first, are " + "present.", + }, + { + .key = {"quorum-count"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = INT_MAX, + .default_value = 0, + .op_version = {1}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + /*.option = quorum-count*/ + /*.validate_fn = validate_quorum_count*/ + .description = "If quorum-type is \"fixed\" only allow writes if " + "this many bricks are present. Other quorum types " + "will OVERWRITE this value.", + }, + { + .key = {"quorum-reads"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "no", + .op_version = {GD_OP_VERSION_3_7_0}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "This option has been removed. Reads are not allowed " + "if quorum is not met.", + }, + { + .key = {"node-uuid"}, + .type = GF_OPTION_TYPE_STR, + .description = "Local glusterd uuid string, used in starting " + "self-heal-daemon so that it can crawl only on " + "local index directories.", + }, + { + .key = {"post-op-delay-secs"}, + .type = GF_OPTION_TYPE_INT, + .min = 0, + .max = INT_MAX, + .default_value = "1", + .op_version = {2}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "Time interval induced artificially before " + "post-operation phase of the transaction to " + "enhance overlap of adjacent write operations.", + }, + { + .key = {AFR_SH_READDIR_SIZE_KEY}, + .type = GF_OPTION_TYPE_SIZET, + .description = "readdirp size for performing entry self-heal", + .min = 1024, + .max = 131072, + .op_version = {2}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE, + .tags = {"replicate"}, + .default_value = "1KB", + }, + { + .key = {"ensure-durability"}, + .type = GF_OPTION_TYPE_BOOL, + .op_version = {3}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "Afr performs fsyncs for transactions if this " + "option is on to make sure the changelogs/data is " + "written to the disk", + .default_value = "on", + }, + { + .key = {"afr-dirty-xattr"}, + .type = GF_OPTION_TYPE_STR, + .default_value = AFR_DIRTY_DEFAULT, + }, + {.key = {"afr-pending-xattr"}, + .type = GF_OPTION_TYPE_STR, + .description = "Comma separated list of xattrs that are used to " + "capture information on pending heals."}, + { + .key = {"metadata-splitbrain-forced-heal"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "off", + }, + {.key = {"heal-timeout"}, + .type = GF_OPTION_TYPE_INT, + .min = 5, + .max = INT_MAX, + .default_value = "600", + .op_version = {2}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "time interval for checking the need to self-heal " + "in self-heal-daemon"}, + { + .key = {"consistent-metadata"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "no", + .op_version = {GD_OP_VERSION_3_7_0}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "If this option is enabled, readdirp will force " + "lookups on those entries read whose read child is " + "not the same as that of the parent. This will " + "guarantee that all read operations on a file serve " + "attributes from the same subvol as long as it holds " + " a good copy of the file/dir.", + }, + {.key = {"arbiter-count"}, + .type = GF_OPTION_TYPE_INT, + .description = "subset of child_count. Has to be 0 or 1."}, + { + .key = {"thin-arbiter"}, + .type = GF_OPTION_TYPE_STR, + .op_version = {GD_OP_VERSION_4_1_0}, + .flags = OPT_FLAG_SETTABLE, + .tags = {"replicate"}, + .description = "contains host:path of thin abriter brick", + }, + {.key = {"shd-max-threads"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = 64, + .default_value = "1", + .op_version = {GD_OP_VERSION_3_7_12}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "Maximum number of parallel heals SHD can do per " + "local brick. This can substantially lower heal times" + ", but can also crush your bricks if you don't have " + "the storage hardware to support this."}, + { + .key = {"shd-wait-qlength"}, + .type = GF_OPTION_TYPE_INT, + .min = 1, + .max = 655536, + .default_value = "1024", + .op_version = {GD_OP_VERSION_3_7_12}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "This option can be used to control number of heals" + " that can wait in SHD per subvolume", + }, + { + .key = {"locking-scheme"}, + .type = GF_OPTION_TYPE_STR, + .value = {"full", "granular"}, + .default_value = "full", + .op_version = {GD_OP_VERSION_3_7_12}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "If this option is set to granular, self-heal will " + "stop being compatible with afr-v1, which helps afr " + "be more granular while self-healing", + }, + {.key = {"full-lock"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "yes", + .op_version = {GD_OP_VERSION_3_13_2}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE, + .tags = {"replicate"}, + .description = "If this option is disabled, then the IOs will take " + "range locks same as versions till 3.13.1."}, + { + .key = {"granular-entry-heal"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "no", + .op_version = {GD_OP_VERSION_3_8_0}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "If this option is enabled, self-heal will resort to " + "granular way of recording changelogs and doing entry " + "self-heal.", + }, + { + .key = {"favorite-child-policy"}, + .type = GF_OPTION_TYPE_STR, + .value = {"none", "size", "ctime", "mtime", "majority"}, + .default_value = "none", + .op_version = {GD_OP_VERSION_3_7_12}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "This option can be used to automatically resolve " + "split-brains using various policies without user " + "intervention. \"size\" picks the file with the " + "biggest size as the source. \"ctime\" and \"mtime\" " + "pick the file with the latest ctime and mtime " + "respectively as the source. \"majority\" picks a file" + " with identical mtime and size in more than half the " + "number of bricks in the replica.", + }, + { + .key = {"consistent-io"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "no", + .description = "If this option is enabled, i/o will fail even if " + "one of the bricks is down in the replicas", + }, + {.key = {"use-compound-fops"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "no", + .op_version = {GD_OP_VERSION_3_8_4}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE | OPT_FLAG_DOC, + .tags = {"replicate"}, + .description = "This option exists only for backward compatibility " + "and configuring it doesn't have any effect"}, + {.key = {"use-anonymous-inode"}, + .type = GF_OPTION_TYPE_BOOL, + .default_value = "no", + .op_version = {GD_OP_VERSION_8_0}, + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_SETTABLE, + .tags = {"replicate"}, + .description = "Setting this option heals directory renames efficiently"}, + + {.key = {NULL}}, +}; + +xlator_api_t xlator_api = { + .init = init, + .fini = fini, + .notify = notify, + .reconfigure = reconfigure, + .mem_acct_init = mem_acct_init, + .op_version = {1}, /* Present from the initial version */ + .dumpops = &dumpops, + .fops = &fops, + .cbks = &cbks, + .options = options, + .identifier = "replicate", + .category = GF_MAINTAINED, }; |