Re: Stack dump from GIT HEAD if CONFIG_ATA_SFF is not set

Re: Stack dump from GIT HEAD if CONFIG_ATA_SFF is not set

[Andrew Morton]

(cc linux-ide)

> On Mon, 13 Oct 2008 16:25:30 -0700 Larry Finger <Larry.Finger@lwfinger.net> wrote:
> I am using kernel 2.6.27-Linus-03151-g4480f15b-dirty. On booting a system built
> with CONFIG_ATA_SFF not set, the following appears in my dmesg output. The
> "dirtiness" of the kernel is due to some code included to trace down messages
> coming from the usb system, and is not involved here. If CONFIG_ATA_SFF=y, the
> message does not appear. From Google results, it seems that this parameter
> should be set for most systems, but one should get clean runs without it.
> 
> ata1: SATA max UDMA/133 abar m8192@0xfc484000 port 0xfc484100 irq 4349
> ata2: SATA max UDMA/133 abar m8192@0xfc484000 port 0xfc484180 irq 4349
> ata3: SATA max UDMA/133 abar m8192@0xfc484000 port 0xfc484200 irq 4349
> ata4: SATA max UDMA/133 abar m8192@0xfc484000 port 0xfc484280 irq 4349
> INFO: trying to register non-static key.
> the code is fine but needs lockdep annotation.
> turning off the locking correctness validator.
> Pid: 504, comm: scsi_eh_0 Not tainted 2.6.27-Linus-03151-g4480f15b-dirty #174
> Call Trace:
>  [<ffffffff80255af9>] static_obj+0x51/0x6b
>  [<ffffffff80258d16>] __lock_acquire+0xd28/0x142f
>  [<ffffffff8022f003>] find_busiest_group+0x329/0x7da
>  [<ffffffff8025946f>] lock_acquire+0x52/0x6b
>  [<ffffffff802483de>] __cancel_work_timer+0xbf/0x1e6
>  [<ffffffff80248415>] __cancel_work_timer+0xf6/0x1e6
>  [<ffffffff802483de>] __cancel_work_timer+0xbf/0x1e6
>  [<ffffffff8025776c>] trace_hardirqs_on_caller+0xef/0x113
>  [<ffffffff8041e3e8>] _spin_unlock_irq+0x24/0x27
>  [<ffffffff80234882>] finish_task_switch+0x5f/0xc5
>  [<ffffffff80234823>] finish_task_switch+0x0/0xc5
>  [<ffffffff8041c18a>] thread_return+0x3d/0xbe
>  [<ffffffffa0034dd6>] ata_port_flush_task+0x10/0x3b [libata]
>  [<ffffffffa0041d4d>] ata_scsi_error+0x20/0x57b [libata]
>  [<ffffffff8041e3e8>] _spin_unlock_irq+0x24/0x27
>  [<ffffffff80234823>] finish_task_switch+0x0/0xc5
>  [<ffffffffa000c68a>] scsi_error_handler+0xda/0x535 [scsi_mod]
>  [<ffffffffa000c5b0>] scsi_error_handler+0x0/0x535 [scsi_mod]
>  [<ffffffff8024b2a3>] kthread+0x47/0x73
>  [<ffffffff8041e05e>] trace_hardirqs_on_thunk+0x3a/0x3f
>  [<ffffffff8020ce79>] child_rip+0xa/0x11
>  [<ffffffff8020c4af>] restore_args+0x0/0x30
>  [<ffffffff8024b237>] kthreadd+0x174/0x199
>  [<ffffffff8024b25c>] kthread+0x0/0x73
>  [<ffffffff8020ce6f>] child_rip+0x0/0x11
> input: PS/2 Mouse as /class/input/input2
> input: AlpsPS/2 ALPS GlidePoint as /class/input/input3
> ata1: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
> ata1.00: ATA-8: TOSHIBA MK2546GSX, LB014C, max UDMA/100
> 

Re: Stack dump from GIT HEAD if CONFIG_ATA_SFF is not set

[Tejun Heo]

Andrew Morton wrote:
> (cc linux-ide)
> 
>> On Mon, 13 Oct 2008 16:25:30 -0700 Larry Finger <Larry.Finger@lwfinger.net> wrote:
>> I am using kernel 2.6.27-Linus-03151-g4480f15b-dirty. On booting a system built
>> with CONFIG_ATA_SFF not set, the following appears in my dmesg output. The
>> "dirtiness" of the kernel is due to some code included to trace down messages
>> coming from the usb system, and is not involved here. If CONFIG_ATA_SFF=y, the
>> message does not appear. From Google results, it seems that this parameter
>> should be set for most systems, but one should get clean runs without it.

Yeap, working on it.  It seems SFF/BMDMA support needs to be separated
out further.

Thanks for the report.

-- 
tejun