Oops / BUG? (2.6.17.2 on VIA Epia CL6000)

Oops / BUG? (2.6.17.2 on VIA Epia CL6000)

[Udo van den Heuvel]

Hello,

On my otherwise stable Via EPIA CL6000 I experienced an OOPS.
Hardware should be OK. I was unable to reproduce the event, so far.
In what part of the kernel did things go wrong?
What can I do to help fix the bug? (if it is indeed a bug)

Kind regards,
Udo

Jul  2 07:11:23 epia kernel: BUG: unable to handle kernel paging request
at virtual address 52786fdd
Jul  2 07:11:23 epia kernel:  printing eip:
Jul  2 07:11:23 epia kernel: 52786fdd
Jul  2 07:11:23 epia kernel: *pde = 00000000
Jul  2 07:11:23 epia kernel: Oops: 0000 [#1]
Jul  2 07:11:23 epia kernel: PREEMPT
Jul  2 07:11:23 epia kernel: Modules linked in: eeprom sch_tbf xt_string
xt_MARK xt_length xt_tcpmss xt_mac xt_mark vt1211 hwmon_vid i2c_isa
ipt_ttl ipt_owner ip_nat_irc ip_conntrack_irc ipt_REDIRECT ipt_tos
ip_nat_ftp ip_conntrack_ftp ip_nat_h323 ip_conntrack_h323 ipt_MASQUERADE
ipt_LOG ipt_TCPMSS ipt_REJECT xt_limit xt_state ipt_TARPIT
iptable_filter ipt_TOS iptable_mangle xt_NOTRACK iptable_raw binfmt_misc
lp parport_pc parport nvram ehci_hcd uhci_hcd snd_via82xx snd_ac97_codec
snd_ac97_bus snd_seq_dummy snd_seq_oss snd_seq_midi_event snd_seq
snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd_page_alloc
snd_mpu401_uart snd_rawmidi snd_seq_device snd i2c_viapro
Jul  2 07:11:23 epia kernel: CPU:    0
Jul  2 07:11:23 epia kernel: EIP:
0060:[phys_startup_32+1382576093/-1073741824]    Not tainted VLI
Jul  2 07:11:23 epia kernel: EIP:    0060:[<52786fdd>]    Not tainted VLI
Jul  2 07:11:23 epia kernel: EFLAGS: 00010296   (2.6.17.2 #3)
Jul  2 07:11:23 epia kernel: EIP is at 0x52786fdd
Jul  2 07:11:23 epia kernel: eax: 00000000   ebx: 11d3b28e   ecx:
0000000d   edx: 00000001
Jul  2 07:11:23 epia kernel: esi: 5d21e8b6   edi: 9faa859e   ebp:
ca33023d   esp: dd071f38
Jul  2 07:11:23 epia kernel: ds: 007b   es: 007b   ss: 0068
Jul  2 07:11:23 epia kernel: Process named (pid: 1431,
threadinfo=dd070000 task=ddf61a90)
Jul  2 07:11:23 epia kernel: Stack: 2dad9f88 1b77f396 5f57be35 0e045382
fb72cd9d 9ca3ed70 29739bee 90ca7336
Jul  2 07:11:23 epia kernel:        cdb39230 d727e3ff b41a68a9 d1962a09
b9b3a8a9 1109a681 a447b354 fb7969e6
Jul  2 07:11:23 epia kernel:        73a0d873 06446b61 1be5d2eb 16872e4f
c887aede 1fcb7034 7902b644 ac07b40b
Jul  2 07:11:23 epia kernel: Call Trace:
Jul  2 07:11:30 epia kernel: Code:  Bad EIP value.
Jul  2 07:11:30 epia kernel: EIP:
[phys_startup_32+1382576093/-1073741824] 0x52786fdd SS:ESP 0068:dd071f38
Jul  2 07:11:30 epia kernel: EIP: [<52786fdd>] 0x52786fdd SS:ESP
0068:dd071f38

Re: Oops / BUG? (2.6.17.2 on VIA Epia CL6000)

[Arjan van de Ven]

On Sun, 2006-07-02 at 13:15 +0200, Udo van den Heuvel wrote:
> Hello,
> 
> On my otherwise stable Via EPIA CL6000 I experienced an OOPS.
> Hardware should be OK. I was unable to reproduce the event, so far.
> In what part of the kernel did things go wrong?
> What can I do to help fix the bug? (if it is indeed a bug)

Hi,

something really bad happened (the processor jumped into hyperspace);
however it looks like automatic symbol resolving isn't working;
could you check if CONFIG_KALLSYMS is enabled in your kernel config?
With that enabled, debugability tends to go up bigtime since at least
the backtrace becomes human readable...

Greetings,
    Arjan van de Ven

Re: Oops / BUG? (2.6.17.2 on VIA Epia CL6000)

[Udo van den Heuvel]

Arjan van de Ven wrote:
> On Sun, 2006-07-02 at 13:15 +0200, Udo van den Heuvel wrote:
>> Hello,
>>
>> On my otherwise stable Via EPIA CL6000 I experienced an OOPS.
>> Hardware should be OK. I was unable to reproduce the event, so far.
>> In what part of the kernel did things go wrong?
>> What can I do to help fix the bug? (if it is indeed a bug)
> 
> Hi,
> 
> something really bad happened (the processor jumped into hyperspace);
> however it looks like automatic symbol resolving isn't working;
> could you check if CONFIG_KALLSYMS is enabled in your kernel config?
> With that enabled, debugability tends to go up bigtime since at least
> the backtrace becomes human readable...

I have:

CONFIG_KALLSYMS=y
# CONFIG_KALLSYMS_ALL is not set
# CONFIG_KALLSYMS_EXTRA_PASS is not set


Kind regards,
Udo

Re: Oops / BUG? (2.6.17.2 on VIA Epia CL6000)

[Arjan van de Ven]

On Sun, 2006-07-02 at 13:24 +0200, Udo van den Heuvel wrote:
> Arjan van de Ven wrote:
> > On Sun, 2006-07-02 at 13:15 +0200, Udo van den Heuvel wrote:
> >> Hello,
> >>
> >> On my otherwise stable Via EPIA CL6000 I experienced an OOPS.
> >> Hardware should be OK. I was unable to reproduce the event, so far.
> >> In what part of the kernel did things go wrong?
> >> What can I do to help fix the bug? (if it is indeed a bug)
> > 
> > Hi,
> > 
> > something really bad happened (the processor jumped into hyperspace);
> > however it looks like automatic symbol resolving isn't working;
> > could you check if CONFIG_KALLSYMS is enabled in your kernel config?
> > With that enabled, debugability tends to go up bigtime since at least
> > the backtrace becomes human readable...
> 
> I have:
> 
> CONFIG_KALLSYMS=y

Hoi,

hmmmmm then something really bad happened, so bad that even the
backtrace is corrupted ;(

I'm sorry to say but unless you already have a suspect, those are really
really hard to diagnose or fix. One thing you can do is remove all
modules you don't actually use (just from a statistical pov that reduces
the risk of it repeating)... and hope maybe a next backtrace does
provide more information.

Greetings,
    Arjan van de Ven

Re: Oops / BUG? (2.6.17.2 on VIA Epia CL6000)

[Udo van den Heuvel]

Arjan van de Ven wrote:
> On Sun, 2006-07-02 at 13:24 +0200, Udo van den Heuvel wrote:
>
>> CONFIG_KALLSYMS=y
> 
> hmmmmm then something really bad happened, so bad that even the
> backtrace is corrupted ;(
> 
> I'm sorry to say but unless you already have a suspect, those are really
> really hard to diagnose or fix. One thing you can do is remove all
> modules you don't actually use (just from a statistical pov that reduces
> the risk of it repeating)... and hope maybe a next backtrace does
> provide more information.

Later another one happened with less info. Maybe the `different`
location helps?

Jul  2 09:07:30 epia kernel: BUG: unable to handle kernel paging request
at virtual address fec9bd89
Jul  2 09:07:30 epia kernel:  printing eip:
Jul  2 09:07:30 epia kernel: fec9bd89
Jul  2 09:07:30 epia kernel: *pde = 00000000
Jul  2 09:07:30 epia kernel: Oops: 0000 [#1]
Jul  2 09:07:30 epia kernel: PREEMPT
Jul  2 09:07:36 epia kernel: Modules linked in: sch_tbf xt_string
xt_MARK xt_length xt_tcpmss xt_mac xt_mark vt1211 hwmon_vid i2c_isa
ipt_ttl ipt_owner ip_nat_irc ip_conntrack_irc ipt_REDIRECT ipt_tos
ip_nat_ftp ip_conntrack_ftp ip_nat_h323 ip_conntrack_h323 ipt_MASQUERADE
ipt_LOG ipt_TCPMSS ipt_REJECT xt_limit xt_state ipt_TARPIT
iptable_filter ipt_TOS iptable_mangle xt_NOTRACK iptable_raw binfmt_misc
lp parport_pc parport nvram ehci_hcd snd_via82xx snd_ac97_codec
snd_ac97_bus snd_seq_dummy snd_seq_oss snd_seq_midi_event snd_seq
uhci_hcd snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd_page_alloc
snd_mpu401_uart snd_rawmidi snd_seq_device snd i2c_viapro
Jul  2 09:07:42 epia kernel: CPU:    0
Jul  2 09:07:42 epia kernel: EIP:    0060:[pg0+1049402761/1069728768]
 Not tainted VLI
Jul  2 09:07:42 epia kernel: EIP:    0060:[<fec9bd89>]    Not tainted VLI
Jul  2 09:07:42 epia kernel: EFLAGS: 00010296   (2.6.17.2 #3)
Jul  2 09:07:42 epia kernel: EIP is at 0xfec9bd89


Kind regards,
Udo

Re: Oops / BUG? (2.6.17.2 on VIA Epia CL6000)

[Jan Engelhardt]

>> > Hi,
>> > 
>> > something really bad happened (the processor jumped into hyperspace);
>> > however it looks like automatic symbol resolving isn't working;
>> > could you check if CONFIG_KALLSYMS is enabled in your kernel config?
>> > With that enabled, debugability tends to go up bigtime since at least
>> > the backtrace becomes human readable...
>> 
>> I have:
>> 
>> CONFIG_KALLSYMS=y
>
>Hoi,
>
>hmmmmm then something really bad happened, so bad that even the
>backtrace is corrupted ;(

How about adding this one?
	CONFIG_FRAME_POINTER=y
Compiling things without frame pointer on 
userspace (yes, userspace) makes debugging impossible, I wonder how the 
kernel can do it without FPs.
And if you got time on your hands
	CONFIG_UNWIND_INFO=y


Jan Engelhardt
-- 

Re: Oops / BUG? (2.6.17.2 on VIA Epia CL6000)

[Chuck Ebbert]

In-Reply-To: <44A7AADB.8040106@xs4all.nl>

On Sun, 02 Jul 2006 13:15:39 +0200, Udo van den Heuvel wrote:

> On my otherwise stable Via EPIA CL6000 I experienced an OOPS.
> Hardware should be OK. I was unable to reproduce the event, so far.
> In what part of the kernel did things go wrong?

Your kernel stack has been overwritten with seemingly-random data.
This is impossible to diagnose -- some module possibly scribbled
over the stack.

> Process named (pid: 1431, threadinfo=dd070000 task=ddf61a90)

> EIP: [<52786fdd>] 0x52786fdd SS:ESP 0068:dd071f38

Threadinfo is at dd070000 so end-of-stack is at dd071ff8
dd071ff8 - dd071f38 = 192 (decimal) so there are 192 bytes on the
kernel stack.  We can see 192 bytes and there is no sensible regs
information at the end, just random junk.

> What can I do to help fix the bug? (if it is indeed a bug)

Where did these modules come from?

        ipt_TARPIT
        vt1211

-- 
Chuck
 "You can't read a newspaper if you can't read."  --George W. Bush