btrfs: use rcu_barrier() to wait for bdev puts at unmount
authorEric Sandeen <sandeen@redhat.com>
Sat, 9 Mar 2013 15:18:39 +0000 (15:18 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 20 Mar 2013 20:05:00 +0000 (13:05 -0700)
commita40a945f829a2b95d5460491d81061166817e3cb
tree829e7fb09e5576636361e52b8a54c8c0fbb122be
parent59c0110f7ee64a9b5389b8b4fd2ec5feab9c6dc8
btrfs: use rcu_barrier() to wait for bdev puts at unmount

commit bc178622d40d87e75abc131007342429c9b03351 upstream.

Doing this would reliably fail with -EBUSY for me:

# mount /dev/sdb2 /mnt/scratch; umount /mnt/scratch; mkfs.btrfs -f /dev/sdb2
...
unable to open /dev/sdb2: Device or resource busy

because mkfs.btrfs tries to open the device O_EXCL, and somebody still has it.

Using systemtap to track bdev gets & puts shows a kworker thread doing a
blkdev put after mkfs attempts a get; this is left over from the unmount
path:

btrfs_close_devices
__btrfs_close_devices
call_rcu(&device->rcu, free_device);
free_device
INIT_WORK(&device->rcu_work, __free_device);
schedule_work(&device->rcu_work);

so unmount might complete before __free_device fires & does its blkdev_put.

Adding an rcu_barrier() to btrfs_close_devices() causes unmount to wait
until all blkdev_put()s are done, and the device is truly free once
unmount completes.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/btrfs/volumes.c