lkml.org 
[lkml]   [2009]   [Jan]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [v4l-dvb-maintainer] 2.6.28: Oops at acpi_ds_exec_end_op+0x1b: kills machine
Michael Krufky wrote:
> Mark Lord wrote:
>> This happened on our Mythtv box at startup this morning:
>> The Oops happened once only, at system startup, on initial device probe.
>> Not reproduceable, so probably a rare race of some kind.
>> The xc5000 module was loaded with "init_fw=1" to force f/w load on
>> modprobe.
>>
>> 2.6.28 kernel, SMP, x86_64, two CPUs (Core2duo 1.86GHz), 2GB RAM.
>>
>> Oops text typed in by hand from screen photos
>> ( original photos: http://rtr.ca/xc5000_lockup )
>> ...
>> xc5000: firmware read 12332 bytes.
>> xc5000: firmware upload
>> BUG kernel NULL pointer dereference at 000000000000000c
>> IP: [<ffffffff8038fbcf>] acpi_ds_exec_end_op+0x1b/0x3c4
>> PGD 7c8cd067 PUD 7d411067 PMD 0
>> Oops: 0000 [#1] PREEMPT SMP
>> last sysfs file: /sys/class/firmware/i2c-2/loading
>> CPU 0
>> ...
>> Modules linked in: xc5000 au8522 tea5767 tda9887 tda8290 tea5761 fuse
>> au0828(+)
>> dvd_core usbhid tuner_simple tuner_types snd_hda_intel msp3400 ...
>> nvidia(P) ...
>> ...
>> Pid: 2383, comm: modprobe Tainted: P A 2.6.28 #9 RIP:
>> 0010:[<ffffffff8038fbcf>] [<ffffffff8038fbcf>]
>> acpi_ds_exec_end_op+0x1b/0x3c4
>> RSP: 0018:ffff88007a095ad8 EFLAGS: 00010296
>> RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
>> RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff88007f8588f0
>> R10: ffff88007f8a4d40 R11: ffff8800000bf9e0 R12: 0000000000000000
>> R13: ffff88007e43bc10 R14: ffff88007a0ac000 R15: ffff88007e43bc10
>> FS: 00007f844d5666e0(0000) GS:ffffffff8060f540(0000)
>> knlGS:0000000000000000
>> CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
>> ...
>> Process modprobe (pid: 2383, threadinfo ffff88007a094000, task
>> ffff88007fa043d0)
>> Stack:
>> ffff88007a085380 0000000000000000 ffff88007a085380 ffffffffa09cfb28
>> ffff88007e43bc10 ffffffffa09f263f ffff88007a095b18 ffff88007a085380
>> ffffc200109e6000 ffff88007a085380 000000000000f9f9 00000000000094d7
>> Call Trace:
>> xc_load_fw_and_init_tuner+0x14b/0x29c [xc5000]
>> i2c_transfer+0x8a/0x94
>> xc5000_init+0x3d/0x6f [xc5000]
>> xc5000_attach+0x22c/0x256 [xc5000]
>> __symbol_get+0x2c/0xd3
>> au0828+dvb_register+0x2ec/x5cb [au0828]
>> au0828_usb_probe+0x10c/0x140 [au0828]
>> usb_match_id+0x32/0x58 [usbcore]
>> usb_probe_interface+0xfb/0x132 [usbcore]
>> driver_probe_device+0xb5/0x159
>> __driver_attach+0x59/0x80
>> __driver_attach+0x0/0x80
>> bus_for_each_dev+0x44/0x78
>> bus_add_driver+0xac/0x1f2
>> driver_register+0xa2/0x11f
>> usb_register_driver+0x7e/0xe0 [usbcore]
>> au0828_init+0x0/0xbb [au0828]
>> au0828_init+0xa1/0xbb [au0828]
>> _stext+0x56/0x14f
>> vma_link+0x90/0xa2
>> update_curr+0x49/0xf0
>> enqueue_task_fair+0x14c/0x161
>> check_preempt_wakeup+0xf4/0x124
>> try_to_wake_up+0x16a/0x17c
>> sys_init_module+0xa0/0x1a9
>> system_call_fastpath+0x16/0x1b
>> Code: 00 66 c7 40 0c c0 00 89 d8 48 83 c4 18 5b 5d c3 41 55 41 54 55
>> 48 89
>> 48 83 ec 08 48 8b 87 08 03 00 00 4c 8b a7 00 03 00 00 <8a> 50 0c 8a
>> 40 0?
>> c8 83 f9 0a 75 25 41 0f b7 4c 24 0a 48
>> RIP [<ffffffff8038fbcf>] acpi_ds_exec_end_op+0x1b/0x3c4
>> RSP <ffff88007a095ad8>
>> CR2: 000000000000000c
>> ---[ end trace c23df576c022eb7e ]---
>>
>> Could someone perhaps explain how acpi_ds_exec_end_op even enters the
>> picture here? I'm confused by that.
>>
>> The driver was doing/completing an i2c firmware transfer at the time of
>> the oops.
>
>
> We removed the "init_fw" module option from the xc5000 module. It is no
> longer necessary, and prevents us from taking advantage of power saving
> capabilities.
..

Yeah, I noticed that, and my workaround (now permanent) is partly to remove
use of init_fw.

But also, it appeared there may be some racing between modules (au0828 and xc5000)
doing various inits. So I've also hardcoded those to be preloaded before
the ehci_hcd bus driver gets loaded.

> So, the OOPS itself is moot. How did acpi_ds_exec_end_op enter the
> picture??? That I have no idea. Can you try to reproduce that will a
> clean (not tainted) kernel?
..

Not a hope, thanks.


\
 
 \ /
  Last update: 2009-01-06 22:55    [W:0.039 / U:0.840 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site