[xfs-masters] XFS lockdep bug (2.6.39/2.6.39-2612-g0524975) with racer

[Christoph Hellwig]

On Sat, May 21, 2011 at 01:07:41AM -0400, Erez Zadok wrote:
> I ran racer on top of xfs in 2.6.39 and v2.6.39-2612-g0524975.  Very quickly got a series of these lockdep warnings.

There's really nothing XFS specific here, except that the old IDE driver
which seems to have an issue is called by XFS in this case.

What's rather confusing about the trace is that schedule() should
never call into the block driver, but offload it to kblockd from
how I read the code in 2.6.39 final.  It used to be different before
but got fixed during the rcs.

> 
> Cheers,
> Erez.
> 
> [   45.605295] BUG: sleeping function called from invalid context at drivers/ide/ide-io.c:468
> [   45.605429] in_atomic(): 1, irqs_disabled(): 0, pid: 2464, name: dir_create.sh
> [   45.605533] 1 lock held by dir_create.sh/2464:
> [   45.605606]  #0:  (&(&ip->i_lock)->mr_lock){++++..}, at: [<d20efd34>] xfs_ilock+0x4f/0x67 [xfs]
> [   45.605848] Pid: 2464, comm: dir_create.sh Not tainted 2.6.39-linus+ #390
> [   45.605946] Call Trace:
> [   45.606021]  [<c10201eb>] __might_sleep+0xd5/0xdd
> [   45.606112]  [<c1160cc2>] do_ide_request+0x3a/0x514
> [   45.606199]  [<c111a880>] ? cfq_service_tree_add+0x1de/0x241
> [   45.606290]  [<c111a54c>] ? cfq_prio_tree_add+0x80/0x90
> [   45.606376]  [<c1110d19>] __blk_run_queue+0x14/0x16
> [   45.606458]  [<c111bee8>] cfq_insert_request+0x403/0x40b
> [   45.606542]  [<c111024e>] __elv_add_request+0x14c/0x17c
> [   45.606625]  [<c111267e>] blk_flush_plug_list+0x130/0x17a
> [   45.606714]  [<c11f7d0c>] schedule+0x22f/0x705
> [   45.606795]  [<c104710c>] ? mark_held_locks+0x3d/0x58
> [   45.606879]  [<c11fa2c5>] ? _raw_spin_unlock_irqrestore+0x36/0x59
> [   45.607041]  [<c1047226>] ? trace_hardirqs_on_caller+0xff/0x120
> [   45.607195]  [<c102039e>] ? get_parent_ip+0xb/0x31
> [   45.607337]  [<c1022eef>] ? sub_preempt_count+0x74/0x8d
> [   45.607495]  [<d20f8e6d>] _xfs_log_force_lsn+0x229/0x26c [xfs]
> [   45.607665]  [<c1047252>] ? trace_hardirqs_on+0xb/0xd
> [   45.607809]  [<c102195e>] ? try_to_wake_up+0x1db/0x1db
> [   45.607967]  [<d2104b30>] _xfs_trans_commit+0x373/0x47c [xfs]
> [   45.608137]  [<d20f5c40>] xfs_iomap_write_allocate+0x221/0x301 [xfs]
> [   45.608307]  [<d210a272>] xfs_map_blocks+0x1b9/0x1cd [xfs]
> [   45.608460]  [<c10497b4>] ? __lock_acquire+0x6ff/0x76a
> [   45.608627]  [<d210b327>] xfs_vm_writepage+0x284/0x441 [xfs]
> [   45.608782]  [<c10668b7>] __writepage+0xb/0x23
> [   45.608922]  [<c1066c28>] write_cache_pages+0x1a9/0x271
> [   45.609068]  [<c10668ac>] ? set_page_dirty+0x5a/0x5a
> [   45.609213]  [<c1066d1e>] generic_writepages+0x2e/0x43
> [   45.609408]  [<d210a4a1>] xfs_vm_writepages+0x3c/0x42 [xfs]
> [   45.640167]  [<d210a465>] ? xfs_aops_discard_page+0x14a/0x14a [xfs]
> [   45.640331]  [<c1066d4f>] do_writepages+0x1c/0x28
> [   45.640474]  [<c1061a1a>] __filemap_fdatawrite_range+0x5a/0x66
> [   45.640629]  [<c1062116>] filemap_fdatawrite_range+0x10/0x12
> [   45.640790]  [<d210ec46>] xfs_flush_pages+0x5c/0x94 [xfs]
> [   45.640948]  [<d210921c>] xfs_release+0x107/0x1cc [xfs]
> [   45.641103]  [<d210defd>] xfs_file_release+0xd/0x11 [xfs]
> [   45.641257]  [<c10858fe>] fput+0xee/0x193
> [   45.641452]  [<c1082dcd>] filp_close+0x57/0x61
> [   45.641589]  [<c1090a51>] sys_dup3+0xdd/0x101
> [   45.641722]  [<c1090af5>] sys_dup2+0x80/0x8a
> [   45.641857]  [<c11fadf0>] sysenter_do_call+0x12/0x36
> 
> _______________________________________________
> xfs-masters mailing list
> xfs-masters at oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs-masters
---end quoted text---