andy/glusterfs
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

cluster/afr: Perform post-op in entry selfheal inside locks

Browse Source 
Take entrylks in xlator domain before doing post-op (undo-pending) in
entry self-heal. This is to prevent a parallel name self-heal on
an entry under @fd->inode from reading pending xattrs while it is
being modified by SHD after entry sh below, given that
name self-heal takes locks ONLY in xlator domain and is free to read
pending changelog in the absence of the following locking.

Change-Id: Ie083ceab10155c460447f04bdce7688480f1ac4f
BUG: 1128721
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: http://review.gluster.org/9020
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
Tested-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
This commit is contained in:
Krutika Dhananjay 2014-10-31 12:51:15 +05:30 committed by Vijay Bellur
parent 756949556b
commit 927b9f3f48
1 changed files with 31 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
xlators/cluster/afr/src/afr-self-heal-entry.c
View File

@ -565,6 +565,7 @@ __afr_selfheal_entry (call_frame_t *frame, xlator_t *this, fd_t *fd,
unsigned char *sources = NULL;
unsigned char *sinks = NULL;
unsigned char *data_lock = NULL;
unsigned char *postop_lock = NULL;
unsigned char *healed_sinks = NULL;
struct afr_reply *locked_replies = NULL;
afr_private_t *priv = NULL;
@ -575,6 +576,7 @@ __afr_selfheal_entry (call_frame_t *frame, xlator_t *this, fd_t *fd,
sinks = alloca0 (priv->child_count);
healed_sinks = alloca0 (priv->child_count);
data_lock = alloca0 (priv->child_count);
postop_lock = alloca0 (priv->child_count);
locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
@ -606,9 +608,35 @@ unlock:
if (ret)
goto out;
ret = afr_selfheal_undo_pending (frame, this, fd->inode, sources, sinks,
healed_sinks, AFR_ENTRY_TRANSACTION,
locked_replies, data_lock);
/* Take entrylks in xlator domain before doing post-op (undo-pending) in
* entry self-heal. This is to prevent a parallel name self-heal on
* an entry under @fd->inode from reading pending xattrs while it is
* being modified by SHD after entry sh below, given that
* name self-heal takes locks ONLY in xlator domain and is free to read
* pending changelog in the absence of the following locking.
*/
ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name, NULL,
postop_lock);
{
if (AFR_CMP (data_lock, postop_lock, priv->child_count) != 0) {
gf_log (this->name, GF_LOG_DEBUG, "%s: Skipping "
"post-op after entry self-heal as %d "
"sub-volumes, as opposed to %d, could be locked"
" in %s domain", uuid_utoa (fd->inode->gfid),
ret, AFR_COUNT (data_lock, priv->child_count),
this->name);
ret = -ENOTCONN;
goto postop_unlock;
}
ret = afr_selfheal_undo_pending (frame, this, fd->inode,
sources, sinks, healed_sinks,
AFR_ENTRY_TRANSACTION,
locked_replies, postop_lock);
}
postop_unlock:
afr_selfheal_unentrylk (frame, this, fd->inode, this->name, NULL,
postop_lock);
out:
afr_log_selfheal (fd->inode->gfid, this, ret, "entry", source,
healed_sinks);