Sending incremental backup causes deadlock (3.14.3)

Sending incremental backup causes deadlock (3.14.3)

[Abbas]

Has anyone else experienced a deadlock while doing an incremental send
/ receive?

# uname -r
3.14.3-1.el6.elrepo.x86_64

# btrfs version
Btrfs v3.14.1

Initial send works well but sending incremental snapshots just HALTS:

# btrfs send -vvvv -p test/snap-2014-05-09-13:29:04
test/snap-2014-05-09-13:32:58 | btrfs receive -vvvv backup/test
At subvol test/snap-2014-05-09-13:32:58
At snapshot snap-2014-05-09-13:32:58
receiving snapshot snap-2014-05-09-13:32:58
uuid=8cde86f7-6ded-5046-a55c-3a0d40e7a2fc, ctransid=0
parent_uuid=1a2a9dc2-c271-524f-bd01-b671da630446, parent_ctransid=0
...

Here's the dmesg:

INFO: task btrfs:16204 blocked for more than 120 seconds.
      Not tainted 3.14.3-1.el6.elrepo.x86_64 #1
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
btrfs           D ffffffff8181d240     0 16204  23963 0x00000080
 ffff88003c8abc38 0000000000000086 ffff88003c8abbd8 ffff88003c8aa010
 0000000000014440 0000000000014440 ffff880004b40250 ffffffff81c13480
 ffff88003c8abc48 ffff880036b46960 ffff880037bf5800 ffff88003c8abc68
Call Trace:
 [<ffffffff81636649>] schedule+0x29/0x70
 [<ffffffffa011f235>] btrfs_commit_transaction+0x335/0x9e0 [btrfs]
 [<ffffffff810b2e50>] ? bit_waitqueue+0xe0/0xe0
 [<ffffffffa014fb45>] create_snapshot+0x265/0x280 [btrfs]
 [<ffffffffa0150403>] btrfs_mksubvol+0x153/0x1f0 [btrfs]
 [<ffffffffa015063c>] btrfs_ioctl_snap_create_transid+0x19c/0x1b0 [btrfs]
 [<ffffffffa01507d8>] btrfs_ioctl_snap_create_v2+0x108/0x160 [btrfs]
 [<ffffffff811d395a>] ? do_filp_open+0x4a/0xa0
 [<ffffffffa015728a>] btrfs_ioctl+0x45a/0x950 [btrfs]
 [<ffffffff8126215d>] ? inode_has_perm+0x2d/0x40
 [<ffffffff81262202>] ? file_has_perm+0x92/0xa0
 [<ffffffff811d4f53>] do_vfs_ioctl+0x73/0x380
 [<ffffffff81263e6f>] ? selinux_file_ioctl+0x4f/0xf0
 [<ffffffff811d5301>] SyS_ioctl+0xa1/0xb0
 [<ffffffff81642de9>] system_call_fastpath+0x16/0x1b

Re: Sending incremental backup causes deadlock (3.14.3)

[Josef Bacik]

On 05/09/2014 11:41 AM, Abbas wrote:
> Has anyone else experienced a deadlock while doing an incremental send
> / receive?
>
> # uname -r
> 3.14.3-1.el6.elrepo.x86_64
>
> # btrfs version
> Btrfs v3.14.1
>
> Initial send works well but sending incremental snapshots just HALTS:
>
> # btrfs send -vvvv -p test/snap-2014-05-09-13:29:04
> test/snap-2014-05-09-13:32:58 | btrfs receive -vvvv backup/test
> At subvol test/snap-2014-05-09-13:32:58
> At snapshot snap-2014-05-09-13:32:58
> receiving snapshot snap-2014-05-09-13:32:58
> uuid=8cde86f7-6ded-5046-a55c-3a0d40e7a2fc, ctransid=0
> parent_uuid=1a2a9dc2-c271-524f-bd01-b671da630446, parent_ctransid=0
> ...
>

Known problem, fixed in 3.15-rc1.  Thanks,

Josef

Re: Sending incremental backup causes deadlock (3.14.3)

[Chris Samuel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Josef,

On Fri, 9 May 2014 01:22:27 PM Josef Bacik wrote:

> Known problem, fixed in 3.15-rc1.  Thanks,

Is the fix suitable for -stable too?

All the best,
Chris
- -- 
 Chris Samuel  :  http://www.csamuel.org/  :  Melbourne, VIC

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)

iQEVAwUBU23dko1yjaOTJg85AQLPwggAw6zkv1svCb5PToUiwhjwArArlaa/hLjR
n0YgbWpfACpvL0iDrbbEKwIn6cBBOwe/wm5vSF/JoAmgRk8UWkxZrsItjBXt9XjR
Z1O/SJFUjJILxbiKNx4qjEHVKNZfBuPb5fBkcU9iRUzeAXzJUGfC5oWBx29fi187
7G+gJ/axApMdtbNkVyAzRdoS8ehDJgpb95dhxmuSv30d9sYvHlt6k+Ye6ulTO4V2
q2XUQoxTxs5HRACWToZiB5T89GQgJSppLxvRCYlHIYxOpjHvOQDkVvHnIHhVtJhl
ko0TwiQwXNVxwQ825y5T07vhlifWlRx0QHysoqBWP4Qwlfdf1cyjtw==
=xkLU
-----END PGP SIGNATURE-----

Re: Sending incremental backup causes deadlock (3.14.3)

[Josef Bacik]

On 05/10/2014 04:04 AM, Chris Samuel wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hi Josef,
>
> On Fri, 9 May 2014 01:22:27 PM Josef Bacik wrote:
>
>> Known problem, fixed in 3.15-rc1.  Thanks,
>
> Is the fix suitable for -stable too?
>

Unfortunately no, it's too big.  I had thought about shoving it back but 
it's simply too complicated and has a lot of other patches that go along 
with it to fix other things.  Thanks,

Josef