Re: XFS oops in vanilla 2.6.24

Re: XFS oops in vanilla 2.6.24

[David Chinner]

On Thu, Jan 31, 2008 at 12:23:11PM +0100, Peter Zijlstra wrote:
> Lets CC the XFS maintainer..

Adding the xfs list and hch.

It might be a couple of days before I get to this - I've got a
week of backlog to catch up on after LCA....

> On Wed, 2008-01-30 at 20:23 +0000, Sven Geggus wrote:
> > Hi there,
> > 
> > I get the following with 2.6.24:
> > 
> > Ending clean XFS mount for filesystem: dm-0
> > BUG: unable to handle kernel paging request at virtual address f2134000

How long after mount does this happen? Does it happen when listing a specific
directory? i.e. do you have a reproducable test case for it?

Cheers,

Dave.

> > printing eip: c021a13a *pde = 010b5067 *pte = 32134000 
> > Oops: 0000 [#1] PREEMPT DEBUG_PAGEALLOC
> > Modules linked in: radeon drm rfcomm l2cap sym53c8xx scsi_transport_spi snd_via82xx 8139too snd_mpu401_uart snd_ens1371 snd_rawmidi snd_ac97_codec ac97_bus snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd_page_alloc via_agp agpgart
> > 
> > Pid: 3889, comm: bash Not tainted (2.6.24 #3)
> > EIP: 0060:[<c021a13a>] EFLAGS: 00010282 CPU: 0
> > EIP is at xfs_file_readdir+0xfa/0x18c
> > EAX: 00000000 EBX: 000002f5 ECX: 00000020 EDX: 00000000
> > ESI: 00000000 EDI: f2133ff8 EBP: f227ff68 ESP: f227ff10
> >  DS: 007b ES: 007b FS: 0000 GS: 0033 SS: 0068
> > Process bash (pid: 3889, ti=f227e000 task=f7205a80 task.ti=f227e000)
> > Stack: 000002f5 00000000 2c000137 00000000 00000000 c0165358 f227ff94 f221c810 
> >        f2d85e48 00000000 00000000 00000000 000002f5 00000000 f2133000 00001000 
> >        00000ff8 000002f9 00000000 c0421c80 f221c810 f1cdbe48 f227ff88 c0165543 
> > Call Trace:
> >  [<c0104da3>] show_trace_log_lvl+0x1a/0x2f
> >  [<c0104e53>] show_stack_log_lvl+0x9b/0xa3
> >  [<c0104efb>] show_registers+0xa0/0x1e2
> >  [<c010514c>] die+0x10f/0x1dd
> >  [<c0113555>] do_page_fault+0x43a/0x519
> >  [<c040fe52>] error_code+0x6a/0x70
> >  [<c0165543>] vfs_readdir+0x5d/0x89
> >  [<c01655cd>] sys_getdents64+0x5e/0xa0
> >  [<c0103f06>] syscall_call+0x7/0xb
> >  =======================
> > Code: 89 74 24 04 81 e3 ff ff ff 7f 89 1c 24 ff 55 bc 85 c0 0f 85 82 00 00 00 8b 4f 10 31 d2 83 c1 1f 83 e1 f8 29 4d d0 19 55 d4 01 cf <8b> 57 08 8b 4f 0c 89 55 d8 89 4d dc 83 7d d4 00 7f a1 7c 06 83 
> > EIP: [<c021a13a>] xfs_file_readdir+0xfa/0x18c SS:ESP 0068:f227ff10
> > ---[ end trace e518e1370efb695e ]---
> > 
> > Sven
> > 

-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

Re: XFS oops in vanilla 2.6.24

[David Chinner]

On Sun, Feb 03, 2008 at 11:20:34PM +0100, Sven Geggus wrote:
> David Chinner schrieb am Sonntag, den 03. Februar um 23:13 Uhr:
> 
> > How long after mount does this happen? Does it happen when listing a specific
> > directory? i.e. do you have a reproducable test case for it?
> 
> Well, it happens very soon after mounting the device. Typically when
> doing ls or something trivial like this. Unfortunately this is not
> really reproducable but it does happen reproducable on one specific
> device here sooner or later when booting into 2.6.24.

Ok. Can you use xfs_metadump to grab an image of the filesystem that you
are using to reproduce this and put it up somewhere on the web so I can
try to reproduce it locally?

> I run xfs-repair just to make shure that the filesystem is not
> corrupted. Works fine when booting 2.6.23.8 which I have been using
> before.

Good - that saves me from having to ask. ;)

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group