* [PATCH] rbd: always kick acquire on "acquired" and "released" notifications
@ 2021-07-06 17:11 Ilya Dryomov
0 siblings, 0 replies; only message in thread
From: Ilya Dryomov @ 2021-07-06 17:11 UTC (permalink / raw)
To: ceph-devel; +Cc: Robin Geuze
Skipping the "lock has been released" notification if the lock owner
is not what we expect based on owner_cid can lead to I/O hangs.
One example is our own notifications: because owner_cid is cleared
in rbd_unlock(), when we get our own notification it is processed as
unexpected/duplicate and maybe_kick_acquire() isn't called. If a peer
that requested the lock then doesn't go through with acquiring it,
I/O requests that came in while the lock was being quiesced would
be stalled until another I/O request is submitted and kicks acquire
from rbd_img_exclusive_lock().
This makes the comment in rbd_release_lock() actually true: prior to
this change the canceled work was being requeued in response to the
"lock has been acquired" notification from rbd_handle_acquired_lock().
Cc: stable@vger.kernel.org # 5.3+
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
---
drivers/block/rbd.c | 20 +++++++-------------
1 file changed, 7 insertions(+), 13 deletions(-)
diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
index be033e179299..a3caecae5d6a 100644
--- a/drivers/block/rbd.c
+++ b/drivers/block/rbd.c
@@ -4203,15 +4203,11 @@ static void rbd_handle_acquired_lock(struct rbd_device *rbd_dev, u8 struct_v,
if (!rbd_cid_equal(&cid, &rbd_empty_cid)) {
down_write(&rbd_dev->lock_rwsem);
if (rbd_cid_equal(&cid, &rbd_dev->owner_cid)) {
- /*
- * we already know that the remote client is
- * the owner
- */
- up_write(&rbd_dev->lock_rwsem);
- return;
+ dout("%s rbd_dev %p cid %llu-%llu == owner_cid\n",
+ __func__, rbd_dev, cid.gid, cid.handle);
+ } else {
+ rbd_set_owner_cid(rbd_dev, &cid);
}
-
- rbd_set_owner_cid(rbd_dev, &cid);
downgrade_write(&rbd_dev->lock_rwsem);
} else {
down_read(&rbd_dev->lock_rwsem);
@@ -4236,14 +4232,12 @@ static void rbd_handle_released_lock(struct rbd_device *rbd_dev, u8 struct_v,
if (!rbd_cid_equal(&cid, &rbd_empty_cid)) {
down_write(&rbd_dev->lock_rwsem);
if (!rbd_cid_equal(&cid, &rbd_dev->owner_cid)) {
- dout("%s rbd_dev %p unexpected owner, cid %llu-%llu != owner_cid %llu-%llu\n",
+ dout("%s rbd_dev %p cid %llu-%llu != owner_cid %llu-%llu\n",
__func__, rbd_dev, cid.gid, cid.handle,
rbd_dev->owner_cid.gid, rbd_dev->owner_cid.handle);
- up_write(&rbd_dev->lock_rwsem);
- return;
+ } else {
+ rbd_set_owner_cid(rbd_dev, &rbd_empty_cid);
}
-
- rbd_set_owner_cid(rbd_dev, &rbd_empty_cid);
downgrade_write(&rbd_dev->lock_rwsem);
} else {
down_read(&rbd_dev->lock_rwsem);
--
2.19.2
^ permalink raw reply related [flat|nested] only message in thread
only message in thread, other threads:[~2021-07-06 17:12 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-07-06 17:11 [PATCH] rbd: always kick acquire on "acquired" and "released" notifications Ilya Dryomov
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).