Re: [PATCH 0/11] Per-bdi writeback flusher threads v8

[Jens Axboe <jens.axboe@oracle.com>]

On Wed, May 27 2009, Theodore Tso wrote:
> On Wed, May 27, 2009 at 10:47:54AM -0400, Theodore Tso wrote:
> > 
> > I'll retry the test with your stock writeback-v8 git branch w/o any
> > ext4 patches planned the next mere window mainline to see if I get the
> > same soft lockup, but I thought I should give you an early heads up.
> 
> Confirmed.  I had to run fsstress twice, but I was able to trigger a
> soft hangup with just the per-bdi v8 patches using ext4.
> 
> With ext3, fsstress didn't cause a soft lockup while it was running
> --- but after the test, when I tried to unmount the filesystem,
> /sbin/umount hung:
> 
> [ 2040.893469] INFO: task umount:7154 blocked for more than 120 seconds.
> [ 2040.893487] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> [ 2040.893503] umount        D 000001ba  2600  7154   5885
> [ 2040.893531]  ec408db8 00000046 ba2bff0b 000001ba c0be7148 c0e68bc8 c0163ebd c0a78700
> [ 2040.893572]  c0a78700 ec408d74 c0164e28 e95c0000 e95c027c c2d13700 00000000 ba2d9a13
> [ 2040.893612]  000001ba c0165031 00000006 e95c0000 c05e9594 00000002 ec408d9c e95c027c
> [ 2040.893652] Call Trace:
> [ 2040.893683]  [<c0163ebd>] ? lock_release_holdtime+0x30/0x131
> [ 2040.893702]  [<c0164e28>] ? mark_lock+0x1e/0x1e4
> [ 2040.893720]  [<c0165031>] ? mark_held_locks+0x43/0x5b
> [ 2040.893742]  [<c05e9594>] ? _spin_unlock_irqrestore+0x3c/0x48
> [ 2040.893761]  [<c01652ba>] ? trace_hardirqs_on+0xb/0xd
> [ 2040.893782]  [<c05e79ff>] schedule+0x8/0x17
> [ 2040.893801]  [<c01d7009>] bdi_sched_wait+0x8/0xc
> [ 2040.893818]  [<c05e7ee8>] __wait_on_bit+0x36/0x5d
> [ 2040.893836]  [<c01d7001>] ? bdi_sched_wait+0x0/0xc
> [ 2040.893854]  [<c05e7fba>] out_of_line_wait_on_bit+0xab/0xb3
> [ 2040.893872]  [<c01d7001>] ? bdi_sched_wait+0x0/0xc
> [ 2040.893892]  [<c01577ae>] ? wake_bit_function+0x0/0x43
> [ 2040.893911]  [<c01d618e>] wait_on_bit+0x20/0x2c
> [ 2040.893929]  [<c01d6d06>] bdi_writeback_all+0x161/0x18e
> [ 2040.893951]  [<c0199f63>] ? wait_on_page_writeback_range+0x9d/0xdc
> [ 2040.894052]  [<c01d6e47>] generic_sync_sb_inodes+0x2f/0xcc
> [ 2040.894079]  [<c01d6f52>] sync_inodes_sb+0x6e/0x76
> [ 2040.894107]  [<c01c1aa0>] __fsync_super+0x63/0x66
> [ 2040.894131]  [<c01c1aae>] fsync_super+0xb/0x19
> [ 2040.894149]  [<c01c1d16>] generic_shutdown_super+0x1c/0xde
> [ 2040.894167]  [<c01c1df5>] kill_block_super+0x1d/0x31
> [ 2040.894186]  [<c01f0a85>] ? vfs_quota_off+0x0/0x12
> [ 2040.894204]  [<c01c2350>] deactivate_super+0x57/0x6b
> [ 2040.894223]  [<c01d2156>] mntput_no_expire+0xca/0xfb
> [ 2040.894242]  [<c01d2633>] sys_umount+0x28f/0x2b4
> [ 2040.894262]  [<c01d2665>] sys_oldumount+0xd/0xf
> [ 2040.894281]  [<c011c264>] sysenter_do_call+0x12/0x38
> [ 2040.894297] 1 lock held by umount/7154:
> [ 2040.894307]  #0:  (&type->s_umount_key#31){++++..}, at: [<c01c234b>] deactivate_super+0x52/0x6b
> 
> 
> Given that the ext4 hangs were also related to s_umount being taken by
> sync_inodes(), there seems to be something going on there:

You didn't happen to catch a sysrq-t of the bdi-* threads as well, did
you? That would confirm the suspicion on this bug, but I'm pretty sure I
know what it is (see the Jan Kara reply). I'll move the super sync to a
silly thread for now, then we can later take care of that with per-bdi
super syncing instead.

-- 
Jens Axboe