XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Oliver Pinter]

http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00775.jpg
http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00776.jpg
http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00777.jpg
http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00778.jpg
http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00779.jpg
http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00780.jpg
http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00781.jpg

tainted with: nvidia and madwifi
kernel: 2.6.25.6 + queue patches (queue head on
aa8edcb997ff605bd1424630356866b78ef06cdc)

-- 
Thanks,
Oliver

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Eric Sandeen]

Oliver Pinter wrote:
> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00775.jpg
> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00776.jpg
> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00777.jpg
> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00778.jpg
> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00779.jpg
> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00780.jpg
> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00781.jpg
> 
> tainted with: nvidia and madwifi
> kernel: 2.6.25.6 + queue patches (queue head on
> aa8edcb997ff605bd1424630356866b78ef06cdc)
> 

What sort of remount were you doing?

-Eric

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Oliver Pinter]

this mistake shutdownkor comes out, remount launched because he
panics, only Alt+Sysrq+foo works

On 6/15/08, Eric Sandeen <sandeen@sandeen.net> wrote:
> Oliver Pinter wrote:
>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00775.jpg
>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00776.jpg
>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00777.jpg
>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00778.jpg
>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00779.jpg
>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00780.jpg
>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00781.jpg
>>
>> tainted with: nvidia and madwifi
>> kernel: 2.6.25.6 + queue patches (queue head on
>> aa8edcb997ff605bd1424630356866b78ef06cdc)
>>
>
> What sort of remount were you doing?
>
> -Eric
>


-- 
Thanks,
Oliver

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Eric Sandeen]

Oliver Pinter wrote:
> this mistake shutdownkor comes out, remount launched because he
> panics, only Alt+Sysrq+foo works

I'm not sure I follow.

Assuming that :
http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00775.jpg

is the first event of interest, it is hitting an assertion on the
remount path:

        ASSERT_ALWAYS(atomic_read(&mp->m_active_trans) == 0);

in xfs_attr_quiesce.

Are you saying something else happen before this?  If so, what?

-Eric

> On 6/15/08, Eric Sandeen <sandeen@sandeen.net> wrote:
>> Oliver Pinter wrote:
>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00775.jpg
>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00776.jpg
>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00777.jpg
>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00778.jpg
>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00779.jpg
>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00780.jpg
>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00781.jpg
>>>
>>> tainted with: nvidia and madwifi
>>> kernel: 2.6.25.6 + queue patches (queue head on
>>> aa8edcb997ff605bd1424630356866b78ef06cdc)
>>>
>> What sort of remount were you doing?
>>
>> -Eric
>>
> 
> 

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Oliver Pinter]

this is the debian way of the unmounting the root fs:
do_stop () {
        [ "$VERBOSE" = no ] || log_action_begin_msg "Mounting root filesystem re
ad-only"
        MOUNT_FORCE_OPT=
        [ "$(uname -s)" = "GNU/kFreeBSD" ] && MOUNT_FORCE_OPT=-f
        # This:
        #     mount -n -o remount,ro /
        # will act on a bind mount of / if there is one.
        # See #339023 and the comment in checkroot.sh
        mount    $MOUNT_FORCE_OPT -n -o remount,ro -t dummytype dummydev / 2>/de
v/null \
        || mount $MOUNT_FORCE_OPT -n -o remount,ro              dummydev / 2>/de
v/null \
        || mount $MOUNT_FORCE_OPT -n -o remount,ro                       /
        ES=$?
        [ "$VERBOSE" = no ] || log_action_end_msg $ES
}


On 6/15/08, Eric Sandeen <sandeen@sandeen.net> wrote:
> Oliver Pinter wrote:
>> this mistake shutdownkor comes out, remount launched because he
>> panics, only Alt+Sysrq+foo works
>
> I'm not sure I follow.
>
> Assuming that :
> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00775.jpg
>
> is the first event of interest, it is hitting an assertion on the
> remount path:
>
>         ASSERT_ALWAYS(atomic_read(&mp->m_active_trans) == 0);
>
> in xfs_attr_quiesce.
>
> Are you saying something else happen before this?  If so, what?
>
> -Eric
>
>> On 6/15/08, Eric Sandeen <sandeen@sandeen.net> wrote:
>>> Oliver Pinter wrote:
>>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00775.jpg
>>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00776.jpg
>>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00777.jpg
>>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00778.jpg
>>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00779.jpg
>>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00780.jpg
>>>> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00781.jpg
>>>>
>>>> tainted with: nvidia and madwifi
>>>> kernel: 2.6.25.6 + queue patches (queue head on
>>>> aa8edcb997ff605bd1424630356866b78ef06cdc)
>>>>
>>> What sort of remount were you doing?
>>>
>>> -Eric
>>>
>>
>>
>
>


-- 
Thanks,
Oliver

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Eric Sandeen]

Oliver Pinter wrote:
> this is the debian way of the unmounting the root fs:

Ok we're slowly getting to useful information.

So you were shutting down the system when the panic happened, correct?

-Eric

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Oliver Pinter]

sorry, but I did not learn English, is going a little bit slowly in this manner.

yeah, under poweroff (init 0) panics the system

On 6/15/08, Eric Sandeen <sandeen@sandeen.net> wrote:
> Oliver Pinter wrote:
>> this is the debian way of the unmounting the root fs:
>
> Ok we're slowly getting to useful information.
>
> So you were shutting down the system when the panic happened, correct?
>
> -Eric
>
>


-- 
Thanks,
Oliver

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Oliver Pinter]

hmm, the system have two XFS partition the / (sda4) and /home (sda3)

/dev/sda3 on /home type xfs (rw,noexec,nosuid,nodev)
/dev/sda4 on / type xfs (rw)

On 6/15/08, Oliver Pinter <oliver.pntr@gmail.com> wrote:
> sorry, but I did not learn English, is going a little bit slowly in this
> manner.
>
> yeah, under poweroff (init 0) panics the system
>
> On 6/15/08, Eric Sandeen <sandeen@sandeen.net> wrote:
>> Oliver Pinter wrote:
>>> this is the debian way of the unmounting the root fs:
>>
>> Ok we're slowly getting to useful information.
>>
>> So you were shutting down the system when the panic happened, correct?
>>
>> -Eric
>>
>>
>
>
> --
> Thanks,
> Oliver
>


-- 
Thanks,
Oliver

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Oliver Pinter]

but now it was him that you say the antecedents apt-get update &
upgrade and git-update-tree (is is a script, that updated 3 git tree
(linux-2.6, linux-2.6.25.y and stable-queue)) , that is heavy disc i/o
with much small files

On 6/15/08, Oliver Pinter <oliver.pntr@gmail.com> wrote:
> hmm, the system have two XFS partition the / (sda4) and /home (sda3)
>
> /dev/sda3 on /home type xfs (rw,noexec,nosuid,nodev)
> /dev/sda4 on / type xfs (rw)
>
> On 6/15/08, Oliver Pinter <oliver.pntr@gmail.com> wrote:
>> sorry, but I did not learn English, is going a little bit slowly in this
>> manner.
>>
>> yeah, under poweroff (init 0) panics the system
>>
>> On 6/15/08, Eric Sandeen <sandeen@sandeen.net> wrote:
>>> Oliver Pinter wrote:
>>>> this is the debian way of the unmounting the root fs:
>>>
>>> Ok we're slowly getting to useful information.
>>>
>>> So you were shutting down the system when the panic happened, correct?
>>>
>>> -Eric
>>>
>>>
>>
>>
>> --
>> Thanks,
>> Oliver
>>
>
>
> --
> Thanks,
> Oliver
>


-- 
Thanks,
Oliver

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Eric Sandeen]

Oliver Pinter wrote:
> sorry, but I did not learn English, is going a little bit slowly in this manner.

That's, ok, it's just useful to know what you were doing when it
panicked.  I'm sure you speak english better than I speak you native
language ;)

> yeah, under poweroff (init 0) panics the system

Hm, interesting.

> but now it was him that you say the antecedents apt-get update &
> upgrade and git-update-tree (is is a script, that updated 3 git tree
> (linux-2.6, linux-2.6.25.y and stable-queue)) , that is heavy disc i/o
> with much small files

So, you had a lot of IO just before shutdown?

-Eric

> On 6/15/08, Eric Sandeen <sandeen@sandeen.net> wrote:
>> Oliver Pinter wrote:
>>> this is the debian way of the unmounting the root fs:
>> Ok we're slowly getting to useful information.
>>
>> So you were shutting down the system when the panic happened, correct?
>>
>> -Eric
>>
>>
> 
> 

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Dave Chinner]

On Sun, Jun 15, 2008 at 10:58:09AM -0500, Eric Sandeen wrote:
> Oliver Pinter wrote:
> > this mistake shutdownkor comes out, remount launched because he
> > panics, only Alt+Sysrq+foo works
> 
> I'm not sure I follow.
> 
> Assuming that :
> http://students.zipernowsky.hu/~oliverp/upload/xfs_error/dsc00775.jpg
> 
> is the first event of interest, it is hitting an assertion on the
> remount path:
> 
>         ASSERT_ALWAYS(atomic_read(&mp->m_active_trans) == 0);
> 
> in xfs_attr_quiesce.

That's caused by a race condition in the VFS w.r.t setting the MS_RDONLY
flag. It should be fixed in 2.6.26.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: XFS errors / BUG / panic, with 2.6.25.7-queue (and older 2.6.25.y)

[Christoph Hellwig]

On Mon, Jun 16, 2008 at 02:52:21PM +1000, Dave Chinner wrote:
> > is the first event of interest, it is hitting an assertion on the
> > remount path:
> > 
> >         ASSERT_ALWAYS(atomic_read(&mp->m_active_trans) == 0);
> > 
> > in xfs_attr_quiesce.
> 
> That's caused by a race condition in the VFS w.r.t setting the MS_RDONLY
> flag. It should be fixed in 2.6.26.

I was claiming it would be fixed in 2.6.26 but unfortunately the part of
the the per-mount r/o patches that fixes various bits in remount isn't
there yet.