diff options
| author | Rajesh Joseph <rjoseph@redhat.com> | 2016-11-22 22:25:42 +0530 |
|---|---|---|
| committer | Niels de Vos <ndevos@redhat.com> | 2016-11-29 23:52:42 -0800 |
| commit | 17d10b42fc4041442e6cd0bfda45944edea498c6 (patch) | |
| tree | 60abd1b7b5ad506e2b8d5455431188c4d6a08b5f /api/src/glfs-resolve.c | |
| parent | e34a783557e504f0793f18beb850bcebbb2af7fa (diff) | |
gfapi: glfs_subvol_done should NOT wait for graph migration.
In graph_setup function glfs_subvol_done is called which
is executed in an epoll thread. glfs_lock waits on other
thread to finish graph migration. This can lead to dead lock
if we consume all the epoll threads.
In general any call-back function executed in epoll thread
should not call any blocking call which waits on a network
reply either directly or indirectly, e.g. syncop functions
should not be called in these threads.
As a fix we should not wait for migration in the call-back path.
Change-Id: If96d0689fe1b4d74631e383048cdc30b01690dc2
BUG: 1397754
Signed-off-by: Rajesh Joseph <rjoseph@redhat.com>
Reviewed-on: http://review.gluster.org/15913
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
Smoke: Gluster Build System <jenkins@build.gluster.org>
Reviewed-by: Niels de Vos <ndevos@redhat.com>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
Diffstat (limited to 'api/src/glfs-resolve.c')
| -rw-r--r-- | api/src/glfs-resolve.c | 19 |
1 files changed, 12 insertions, 7 deletions
diff --git a/api/src/glfs-resolve.c b/api/src/glfs-resolve.c index b84e5d8f58c..f8b437bab0e 100644 --- a/api/src/glfs-resolve.c +++ b/api/src/glfs-resolve.c @@ -784,7 +784,7 @@ glfs_resolve_fd (struct glfs *fs, xlator_t *subvol, struct glfs_fd *glfd) { fd_t *fd = NULL; - glfs_lock (fs); + glfs_lock (fs, _gf_true); { fd = __glfs_resolve_fd (fs, subvol, glfd); } @@ -897,12 +897,17 @@ priv_glfs_subvol_done (struct glfs *fs, xlator_t *subvol) if (!subvol) return; - glfs_lock (fs); + /* For decrementing subvol->wind ref count we need not check/wait for + * migration-in-progress flag. + * Also glfs_subvol_done is called in call-back path therefore waiting + * fot migration-in-progress flag can lead to dead-lock. + */ + glfs_lock (fs, _gf_false); { ref = (--subvol->winds); active_subvol = fs->active_subvol; } - glfs_unlock (fs); + glfs_unlock (fs); if (ref == 0) { assert (subvol != active_subvol); @@ -919,7 +924,7 @@ priv_glfs_active_subvol (struct glfs *fs) xlator_t *subvol = NULL; xlator_t *old_subvol = NULL; - glfs_lock (fs); + glfs_lock (fs, _gf_true); { subvol = __glfs_active_subvol (fs); @@ -968,7 +973,7 @@ glfs_cwd_set (struct glfs *fs, inode_t *inode) { int ret = 0; - glfs_lock (fs); + glfs_lock (fs, _gf_true); { ret = __glfs_cwd_set (fs, inode); } @@ -1001,7 +1006,7 @@ glfs_cwd_get (struct glfs *fs) { inode_t *cwd = NULL; - glfs_lock (fs); + glfs_lock (fs, _gf_true); { cwd = __glfs_cwd_get (fs); } @@ -1041,7 +1046,7 @@ glfs_resolve_inode (struct glfs *fs, xlator_t *subvol, { inode_t *inode = NULL; - glfs_lock (fs); + glfs_lock (fs, _gf_true); { inode = __glfs_resolve_inode(fs, subvol, object); } |