Re: next-20081114: WARNING: at block/blk-barrier.c:246 blk_do_ordered()

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

On Tue, Nov 18 2008, Alexander Beregalov wrote:
> 2008/11/18 Alexander Beregalov <a.beregalov@gmail.com>:
> > 2008/11/17 Jens Axboe <jens.axboe@oracle.com>:
> >> On Mon, Nov 17 2008, Alexander Beregalov wrote:
> >>> 2008/11/15 Alexander Beregalov <a.beregalov@gmail.com>:
> >>> > 2008/11/15 Jens Axboe <jens.axboe@oracle.com>:
> >>> >> On Fri, Nov 14 2008, Alexander Beregalov wrote:
> >>> >>> WARNING: at block/blk-barrier.c:246 blk_do_ordered+0x22b/0x26a()
> >>> >>> Modules linked in:
> >>> >>> Pid: 0, comm: swapper Not tainted 2.6.28-rc4-next-20081114 #3
> >>> >>> Call Trace:
> >>> >>>  <IRQ>  [<ffffffff80236d63>] warn_on_slowpath+0x58/0x7d
> >>> >>>  [<ffffffff802570e3>] ? mark_lock+0x1c/0x37a
> >>> >>>  [<ffffffff803a69f9>] ? elv_rb_del+0x30/0x4b
> >>> >>>  [<ffffffff803b1092>] ? cfq_remove_request+0x180/0x1d8
> >>> >>>  [<ffffffff803afd03>] ? __cfq_slice_expired+0xd6/0xeb
> >>> >>>  [<ffffffff803afd31>] ? cfq_slice_expired+0x19/0x1b
> >>> >>>  [<ffffffff803b1832>] ? cfq_dispatch_requests+0x2c1/0x37e
> >>> >>>  [<ffffffff803aa98c>] blk_do_ordered+0x22b/0x26a
> >>> >>>  [<ffffffff803a6560>] elv_next_request+0x1d0/0x210
> >>> >>>  [<ffffffff80439c7c>] scsi_request_fn+0x9e/0x539
> >>> >>>  [<ffffffff80255c66>] ? put_lock_stats+0xe/0x27
> >>> >>>  [<ffffffff803a827e>] blk_invoke_request_fn+0x3b/0x6e
> >>> >>>  [<ffffffff803a87fd>] __blk_run_queue+0x25/0x29
> >>> >>>  [<ffffffff803a8822>] blk_run_queue+0x21/0x35
> >>> >>>  [<ffffffff80439509>] scsi_run_queue+0x2cf/0x379
> >>> >>>  [<ffffffff8043a33e>] scsi_next_command+0x36/0x46
> >>> >>>  [<ffffffff8043a509>] scsi_end_request+0x92/0xa4
> >>> >>>  [<ffffffff8043aae7>] scsi_io_completion+0x1a7/0x3ad
> >>> >>>  [<ffffffff80434551>] scsi_finish_command+0xe9/0xf2
> >>> >>>  [<ffffffff8043afd5>] scsi_softirq_done+0x10e/0x117
> >>> >>>  [<ffffffff803ac393>] blk_done_softirq+0x7f/0x8f
> >>> >>>  [<ffffffff8023c173>] __do_softirq+0x70/0x101
> >>> >>>  [<ffffffff8020cc0c>] call_softirq+0x1c/0x28
> >>> >>>  [<ffffffff8020e229>] do_softirq+0x39/0x8a
> >>> >>>  [<ffffffff8023bd78>] irq_exit+0x45/0xa2
> >>> >>>  [<ffffffff8020e53a>] do_IRQ+0x16a/0x19c
> >>> >>>  [<ffffffff8020bceb>] ret_from_intr+0x0/0xf
> >>> >>>  <EOI>  [<ffffffff80212e09>] ? mwait_idle+0x3e/0x48
> >>> >>>  [<ffffffff80212e00>] ? mwait_idle+0x35/0x48
> >>> >>>  [<ffffffff8020a940>] ? cpu_idle+0x51/0xba
> >>> >>>  [<ffffffff80514f9c>] ? rest_init+0x70/0x72
> >>> >>> ---[ end trace 115cb4be7e150bb9 ]---
> >>> >>
> >>> >> I'm assuming this is new with -next and doesn't happen with 2.6.28-rc4?
> >>> > Yes.
> >>> >> Is it reproducible in -next?
> >>> > I have seen it only once yet.
> >>>
> >>> I have the same warning on 2.6.28-rc5-next-20081117, reproducible on dbench.
> > Bisected down to 70bfe2de81b32969db1b4939bf171bbe55bcf1ef
> > "elevator prevent flushing small requests to device"
> forget: revert fixes it.

That was my main culprit, thanks a lot for testing and taking the time
to bisect! I'll take a look at it and probably ask you to test a patch.

-- 
Jens Axboe