diff options
| author | Mohit Agrawal <moagrawal@redhat.com> | 2018-09-19 14:32:22 +0530 |
|---|---|---|
| committer | Atin Mukherjee <amukherj@redhat.com> | 2018-09-20 03:46:05 +0000 |
| commit | 4f6ae853ffa9d06446407f389aaef61ac0b3b424 (patch) | |
| tree | c96cf2566835a9c8513a073a94f83e2d1a5ba8fa /xlators/mgmt/glusterd/src/glusterd-op-sm.c | |
| parent | 6a49bce014ac546582cf2409fd9933115406c808 (diff) | |
glusterd: Use GF_ATOMIC to update 'blockers' counter at glusterd_conf
Problem:
Currently in glusterd code uses sync_lock/sync_unlock to update blockers
counter which could add delays to the overall transaction phase
escpecially when there's a batch of volume stop operations processed by
glusterd in brick multiplexing mode.
Solution: Use GF_ATOMIC to update blocker counter to ensure unnecessary
context switching can be avoided.
Change-Id: Ie13177dfee2af66687ae7cf5c67405c152853990
Fixes: bz#1631128
Signed-off-by: Mohit Agrawal <moagrawal@redhat.com>
Diffstat (limited to 'xlators/mgmt/glusterd/src/glusterd-op-sm.c')
| -rw-r--r-- | xlators/mgmt/glusterd/src/glusterd-op-sm.c | 5 |
1 files changed, 4 insertions, 1 deletions
diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c index 0944cc648bf..ce44c4679b1 100644 --- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c +++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c @@ -6018,9 +6018,12 @@ glusterd_op_stage_validate(glusterd_op_t op, dict_t *dict, char **op_errstr, static void glusterd_wait_for_blockers(glusterd_conf_t *priv) { - while (priv->blockers) { + uint64_t blockers = GF_ATOMIC_GET(priv->blockers); + + while (blockers) { synclock_unlock(&priv->big_lock); sleep(1); + blockers = GF_ATOMIC_GET(priv->blockers); synclock_lock(&priv->big_lock); } } |