[Xenomai-help] BUG: unable to handle kernel paging request

[Xenomai-help] BUG: unable to handle kernel paging request

[Victor Shulyak]

[-- Attachment #1: Type: text/plain, Size: 416 bytes --]

Hello,

Today I upgraded my system to xenomai 2.3.0 + kernel 2.6.19.2.
I encountered the following syslog message:

BUG: unable to handle kernel paging request at virtual address 0d40004a
while no xenomai application running.

The system didn't crash, just such kernel message.


p4, nvidia geforce 4, FC6, gnome + compiz manager, gcc 4.1.1 shipped
with FC6, NVIDIA-Linux-x86-1.0-9631 nvidia driver.

Thanks, Victor

[-- Attachment #2: config.log-xenomai.gz --]
[-- Type: application/x-gzip, Size: 8947 bytes --]

[-- Attachment #3: config-kernel.gz --]
[-- Type: application/x-gzip, Size: 11324 bytes --]

[-- Attachment #4: dmesg.gz --]
[-- Type: application/x-gzip, Size: 5914 bytes --]

[-- Attachment #5: cpuinfo.gz --]
[-- Type: application/x-gzip, Size: 307 bytes --]

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Philippe Gerum]

On Fri, 2007-01-26 at 17:46 +0600, Victor Shulyak wrote:
> Hello,
> 
> Today I upgraded my system to xenomai 2.3.0 + kernel 2.6.19.2.
> I encountered the following syslog message:
> 
> BUG: unable to handle kernel paging request at virtual address 0d40004a
> while no xenomai application running.
> 
> The system didn't crash, just such kernel message.
> 
> 
> p4, nvidia geforce 4, FC6, gnome + compiz manager, gcc 4.1.1 shipped
> with FC6, NVIDIA-Linux-x86-1.0-9631 nvidia driver.
> 

> BUG: unable to handle kernel paging request at virtual address >
0d40004a
> printing eip:
> c7b84d51
> *pde = 00000000
> Oops: 0002 [#1]

I guess that you already know what kind of answer you would obtain on
the LKML when reporting faults with Nvidia modules loaded, right?
However, since there seem to be a clean backtrace, maybe we could give a
try anyway. Could you objdump your kernel image, then locate the faulty
routine containing eip=0xc7b84d51? TIA,

> Thanks, Victor
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help
-- 
Philippe.

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Dmitry Adamushko]

>
> I guess that you already know what kind of answer you would obtain on
> the LKML when reporting faults with Nvidia modules loaded, right?
> However, since there seem to be a clean backtrace, maybe we could give a
> try anyway. Could you objdump your kernel image, then locate the faulty
> routine containing eip=0xc7b84d51? TIA,

But it doesn't point to either some part of the kernel code

--->   .text : 0xc0100000 - 0xc030a212   (2088 kB)

or any kernel module (not vmalloc'ed area).

Unless some code would copy another code at this location (kmalloc'ed)
and then execute it.

Let's suppose we can believe in sanity of the call trace, then would
you, Victor, provide a disassembled code for
__ipipe_handle_exception() ?

Call Trace:
 [<c0112566>] __ipipe_handle_exception+0x58/0x11e
 [<c01050f4>] do_divide_error+0x0/0xb1
 [<c0309202>] error_code+0x3e/0x4c


Memory: 384496k/393152k available (2088k kernel code, 8108k reserved,
633k data, 180k init, 0k highmem)
virtual kernel memory layout:
    fixmap  : 0xfffb7000 - 0xfffff000   ( 288 kB)
    vmalloc : 0xd8800000 - 0xfffb5000   ( 631 MB)
    lowmem  : 0xc0000000 - 0xd7ff0000   ( 383 MB)
   .init : 0xc03ac000 - 0xc03d9000   ( 180 kB)
   .data : 0xc030a212 - 0xc03a88f4   ( 633 kB)
   .text : 0xc0100000 - 0xc030a212   (2088 kB)


-- 
Best regards,
Dmitry Adamushko

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Philippe Gerum]

On Fri, 2007-01-26 at 19:46 +0100, Dmitry Adamushko wrote:
> >
> > I guess that you already know what kind of answer you would obtain on
> > the LKML when reporting faults with Nvidia modules loaded, right?
> > However, since there seem to be a clean backtrace, maybe we could give a
> > try anyway. Could you objdump your kernel image, then locate the faulty
> > routine containing eip=0xc7b84d51? TIA,
> 
> But it doesn't point to either some part of the kernel code
> 
> --->   .text : 0xc0100000 - 0xc030a212   (2088 kB)
> 
> or any kernel module (not vmalloc'ed area).
> 
> Unless some code would copy another code at this location (kmalloc'ed)
> and then execute it.
> 
> Let's suppose we can believe in sanity of the call trace, then would
> you, Victor, provide a disassembled code for
> __ipipe_handle_exception() ?
> 

The backtrace is sane, I'm afraid. The following patch should fix the
issue:

--- arch/i386/kernel/entry.S~	2006-12-16 13:32:29.000000000 +0100
+++ arch/i386/kernel/entry.S	2007-01-26 23:44:59.000000000 +0100
@@ -1032,7 +1032,7 @@
 	RING0_INT_FRAME
 	pushl $0			# no error code
 	CFI_ADJUST_CFA_OFFSET 4
-	pushl $do_divide_error
+	PUSH_XVEC(do_divide_error)
 	CFI_ADJUST_CFA_OFFSET 4
 	jmp error_code
 	CFI_ENDPROC

> Call Trace:
>  [<c0112566>] __ipipe_handle_exception+0x58/0x11e
>  [<c01050f4>] do_divide_error+0x0/0xb1
>  [<c0309202>] error_code+0x3e/0x4c
> 
> 
> Memory: 384496k/393152k available (2088k kernel code, 8108k reserved,
> 633k data, 180k init, 0k highmem)
> virtual kernel memory layout:
>     fixmap  : 0xfffb7000 - 0xfffff000   ( 288 kB)
>     vmalloc : 0xd8800000 - 0xfffb5000   ( 631 MB)
>     lowmem  : 0xc0000000 - 0xd7ff0000   ( 383 MB)
>    .init : 0xc03ac000 - 0xc03d9000   ( 180 kB)
>    .data : 0xc030a212 - 0xc03a88f4   ( 633 kB)
>    .text : 0xc0100000 - 0xc030a212   (2088 kB)
> 
> 
-- 
Philippe.

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Philippe Gerum]

On Fri, 2007-01-26 at 23:46 +0100, Philippe Gerum wrote:
> On Fri, 2007-01-26 at 19:46 +0100, Dmitry Adamushko wrote:
> > >
> > > I guess that you already know what kind of answer you would obtain on
> > > the LKML when reporting faults with Nvidia modules loaded, right?
> > > However, since there seem to be a clean backtrace, maybe we could give a
> > > try anyway. Could you objdump your kernel image, then locate the faulty
> > > routine containing eip=0xc7b84d51? TIA,
> > 
> > But it doesn't point to either some part of the kernel code
> > 
> > --->   .text : 0xc0100000 - 0xc030a212   (2088 kB)
> > 
> > or any kernel module (not vmalloc'ed area).
> > 
> > Unless some code would copy another code at this location (kmalloc'ed)
> > and then execute it.
> > 
> > Let's suppose we can believe in sanity of the call trace, then would
> > you, Victor, provide a disassembled code for
> > __ipipe_handle_exception() ?
> > 
> 
> The backtrace is sane, I'm afraid. The following patch should fix the
> issue:
> 
> --- arch/i386/kernel/entry.S~	2006-12-16 13:32:29.000000000 +0100
> +++ arch/i386/kernel/entry.S	2007-01-26 23:44:59.000000000 +0100
> @@ -1032,7 +1032,7 @@
>  	RING0_INT_FRAME
>  	pushl $0			# no error code
>  	CFI_ADJUST_CFA_OFFSET 4
> -	pushl $do_divide_error
> +	PUSH_XVEC(do_divide_error)

s,PUSH_XVEC,PUSH_XCODE,

>  	CFI_ADJUST_CFA_OFFSET 4
>  	jmp error_code
>  	CFI_ENDPROC

> > Call Trace:
> >  [<c0112566>] __ipipe_handle_exception+0x58/0x11e
> >  [<c01050f4>] do_divide_error+0x0/0xb1
> >  [<c0309202>] error_code+0x3e/0x4c
> > 
> > 
> > Memory: 384496k/393152k available (2088k kernel code, 8108k reserved,
> > 633k data, 180k init, 0k highmem)
> > virtual kernel memory layout:
> >     fixmap  : 0xfffb7000 - 0xfffff000   ( 288 kB)
> >     vmalloc : 0xd8800000 - 0xfffb5000   ( 631 MB)
> >     lowmem  : 0xc0000000 - 0xd7ff0000   ( 383 MB)
> >    .init : 0xc03ac000 - 0xc03d9000   ( 180 kB)
> >    .data : 0xc030a212 - 0xc03a88f4   ( 633 kB)
> >    .text : 0xc0100000 - 0xc030a212   (2088 kB)
> > 
> > 
-- 
Philippe.

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Victor Shulyak]

Hello


> > > > However, since there seem to be a clean backtrace, maybe we could give a
> > > > try anyway. Could you objdump your kernel image, then locate the faulty
> > > > routine containing eip=0xc7b84d51?
I can apply objdump to kernel modules, but when this utility was applied
to kernel image, error message is displayed
objdump: /boot/vmlinuz-xenomai: File format not recognized
How could I objdump my kernel image?


> > > Let's suppose we can believe in sanity of the call trace, then would
> > > you, Victor, provide a disassembled code for
> > > __ipipe_handle_exception() ?
> > >
I will do this with pleasure, one thing to know, which tools can help
me?


> > --- arch/i386/kernel/entry.S~	2006-12-16 13:32:29.000000000 +0100
> > +++ arch/i386/kernel/entry.S	2007-01-26 23:44:59.000000000 +0100
> > @@ -1032,7 +1032,7 @@
> >  	RING0_INT_FRAME
> >  	pushl $0			# no error code
> >  	CFI_ADJUST_CFA_OFFSET 4
> > -	pushl $do_divide_error
> > +	PUSH_XVEC(do_divide_error)
> 
> s,PUSH_XVEC,PUSH_XCODE,
> 
> >  	CFI_ADJUST_CFA_OFFSET 4
> >  	jmp error_code
> >  	CFI_ENDPROC

I did so.


Thanks, Victor

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 658 bytes --]

Victor Shulyak wrote:
>>> --- arch/i386/kernel/entry.S~	2006-12-16 13:32:29.000000000 +0100
>>> +++ arch/i386/kernel/entry.S	2007-01-26 23:44:59.000000000 +0100
>>> @@ -1032,7 +1032,7 @@
>>>  	RING0_INT_FRAME
>>>  	pushl $0			# no error code
>>>  	CFI_ADJUST_CFA_OFFSET 4
>>> -	pushl $do_divide_error
>>> +	PUSH_XVEC(do_divide_error)
>> s,PUSH_XVEC,PUSH_XCODE,
>>
>>>  	CFI_ADJUST_CFA_OFFSET 4
>>>  	jmp error_code
>>>  	CFI_ENDPROC
> 
> I did so.
> 

And did this fix your problem?

It seems that applying this change solved occasional early lockups of
2.6.19.2-1.6-0x on my laptop ("occasional" - so I'm still observing...).

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Gilles Chanteperdrix]

Victor Shulyak wrote:
> Hello
> 
> 
> 
>>>>>However, since there seem to be a clean backtrace, maybe we could give a
>>>>>try anyway. Could you objdump your kernel image, then locate the faulty
>>>>>routine containing eip=0xc7b84d51?
> 
> I can apply objdump to kernel modules, but when this utility was applied
> to kernel image, error message is displayed
> objdump: /boot/vmlinuz-xenomai: File format not recognized
> How could I objdump my kernel image?

Instead of vmlinuz, you should pick vmlinux at the root of the kernel
build directory.

> 
> 
> 
>>>>Let's suppose we can believe in sanity of the call trace, then would
>>>>you, Victor, provide a disassembled code for
>>>>__ipipe_handle_exception() ?
>>>>
> 
> I will do this with pleasure, one thing to know, which tools can help
> me?

objdump as well

-- 
                                                 Gilles Chanteperdrix

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Victor Shulyak]

[-- Attachment #1: Type: text/plain, Size: 930 bytes --]

Hello

> Let's suppose we can believe in sanity of the call trace, then would
> you, Victor, provide a disassembled code for
> __ipipe_handle_exception() ?
> 
> Call Trace:
>  [<c0112566>] __ipipe_handle_exception+0x58/0x11e
>  [<c01050f4>] do_divide_error+0x0/0xb1
>  [<c0309202>] error_code+0x3e/0x4c
> 
> 
> Memory: 384496k/393152k available (2088k kernel code, 8108k reserved,
> 633k data, 180k init, 0k highmem)
> virtual kernel memory layout:
>     fixmap  : 0xfffb7000 - 0xfffff000   ( 288 kB)
>     vmalloc : 0xd8800000 - 0xfffb5000   ( 631 MB)
>     lowmem  : 0xc0000000 - 0xd7ff0000   ( 383 MB)
>    .init : 0xc03ac000 - 0xc03d9000   ( 180 kB)
>    .data : 0xc030a212 - 0xc03a88f4   ( 633 kB)
>    .text : 0xc0100000 - 0xc030a212   (2088 kB)

Attached file contains objdump -D of my kernel image with
__ipipe_handle_exception function.


How could I locate the faulty routine containing eip=0xc7b84d51?


Thanks, Victor

[-- Attachment #2: disassembled.bz2 --]
[-- Type: application/x-bzip, Size: 1979 bytes --]

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Dmitry Adamushko]

Victor,

Philippe suggested the following patch. Have you applied it /
recompiled / installed a new kernel ? So, please, give us a yes/no
answer on whether it helps to solve the problem.


--- arch/i386/kernel/entry.S~   2006-12-16 13:32:29.000000000 +0100
+++ arch/i386/kernel/entry.S    2007-01-26 23:44:59.000000000 +0100
@@ -1032,7 +1032,7 @@
       RING0_INT_FRAME
       pushl $0                        # no error code
       CFI_ADJUST_CFA_OFFSET 4
-       pushl $do_divide_error
+       PUSH_XCODE(do_divide_error)
       CFI_ADJUST_CFA_OFFSET 4
       jmp error_code
       CFI_ENDPROC


-- 
Best regards,
Dmitry Adamushko

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Gilles Chanteperdrix]

Victor Shulyak wrote:
> Attached file contains objdump -D of my kernel image with
> __ipipe_handle_exception function.
> 
> 
> How could I locate the faulty routine containing eip=0xc7b84d51?

It is probably the address of a module, so getting the disassembly is a
little bit more complicated: you have to look into /proc/modules for the
load address of the modules. Then pass it to the --adjust-vma option of
objdump.

But you should try Philippe's patch first.

-- 
                                                 Gilles Chanteperdrix

Re: [Xenomai-help] BUG: unable to handle kernel paging request

[Victor Shulyak]

Hello

> Philippe suggested the following patch. Have you applied it /
> recompiled / installed a new kernel ? So, please, give us a yes/no
> answer on whether it helps to solve the problem.

I have applied the patch. If the problem happens again I will post a
message.

Thanks, Victor