Commit 26b69a33 authored by Ilya Dryomov's avatar Ilya Dryomov Committed by Greg Kroah-Hartman

rbd: avoid a deadlock on header_rwsem when flushing notifies

[ Upstream commit 0e4e1de5 ]

rbd_unregister_watch() flushes notifies and therefore cannot be called
under header_rwsem because a header update notify takes header_rwsem to
synchronize with "rbd map".  If mapping an image fails after the watch
is established and a header update notify sneaks in, we deadlock when
erroring out from rbd_dev_image_probe().

Move watch registration and unregistration out of the critical section.
The only reason they were put there was to make header_rwsem management
slightly more obvious.

Fixes: 811c6688 ("rbd: fix rbd map vs notify races")
Signed-off-by: default avatarIlya Dryomov <idryomov@gmail.com>
Reviewed-by: default avatarJason Dillaman <dillaman@redhat.com>
Signed-off-by: default avatarSasha Levin <sashal@kernel.org>
parent adff7c6c
...@@ -3427,6 +3427,10 @@ static void cancel_tasks_sync(struct rbd_device *rbd_dev) ...@@ -3427,6 +3427,10 @@ static void cancel_tasks_sync(struct rbd_device *rbd_dev)
cancel_work_sync(&rbd_dev->unlock_work); cancel_work_sync(&rbd_dev->unlock_work);
} }
/*
* header_rwsem must not be held to avoid a deadlock with
* rbd_dev_refresh() when flushing notifies.
*/
static void rbd_unregister_watch(struct rbd_device *rbd_dev) static void rbd_unregister_watch(struct rbd_device *rbd_dev)
{ {
WARN_ON(waitqueue_active(&rbd_dev->lock_waitq)); WARN_ON(waitqueue_active(&rbd_dev->lock_waitq));
...@@ -5732,6 +5736,9 @@ static void rbd_dev_image_release(struct rbd_device *rbd_dev) ...@@ -5732,6 +5736,9 @@ static void rbd_dev_image_release(struct rbd_device *rbd_dev)
* device. If this image is the one being mapped (i.e., not a * device. If this image is the one being mapped (i.e., not a
* parent), initiate a watch on its header object before using that * parent), initiate a watch on its header object before using that
* object to get detailed information about the rbd image. * object to get detailed information about the rbd image.
*
* On success, returns with header_rwsem held for write if called
* with @depth == 0.
*/ */
static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth) static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth)
{ {
...@@ -5764,6 +5771,9 @@ static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth) ...@@ -5764,6 +5771,9 @@ static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth)
} }
} }
if (!depth)
down_write(&rbd_dev->header_rwsem);
ret = rbd_dev_header_info(rbd_dev); ret = rbd_dev_header_info(rbd_dev);
if (ret) if (ret)
goto err_out_watch; goto err_out_watch;
...@@ -5814,6 +5824,8 @@ static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth) ...@@ -5814,6 +5824,8 @@ static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth)
err_out_probe: err_out_probe:
rbd_dev_unprobe(rbd_dev); rbd_dev_unprobe(rbd_dev);
err_out_watch: err_out_watch:
if (!depth)
up_write(&rbd_dev->header_rwsem);
if (!depth) if (!depth)
rbd_unregister_watch(rbd_dev); rbd_unregister_watch(rbd_dev);
err_out_format: err_out_format:
...@@ -5872,12 +5884,9 @@ static ssize_t do_rbd_add(struct bus_type *bus, ...@@ -5872,12 +5884,9 @@ static ssize_t do_rbd_add(struct bus_type *bus,
goto err_out_rbd_dev; goto err_out_rbd_dev;
} }
down_write(&rbd_dev->header_rwsem);
rc = rbd_dev_image_probe(rbd_dev, 0); rc = rbd_dev_image_probe(rbd_dev, 0);
if (rc < 0) { if (rc < 0)
up_write(&rbd_dev->header_rwsem);
goto err_out_rbd_dev; goto err_out_rbd_dev;
}
/* If we are mapping a snapshot it must be marked read-only */ /* If we are mapping a snapshot it must be marked read-only */
if (rbd_dev->spec->snap_id != CEPH_NOSNAP) if (rbd_dev->spec->snap_id != CEPH_NOSNAP)
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment