Xen 2.0 / XenLinux 2.6.9 oops in handle_mm_fault

Xen 2.0 / XenLinux 2.6.9 oops in handle_mm_fault

[Chris Wedgwood]

Using XenLinux 2.6.9 trying to do very minimal work in domain0
(building a module to get networking up) I'm seeing an oops in
handle_mm_fault...  the full oops is visible at (photo, serial console
isn't working) http://stupidest.org/xen2-crash2.png

The oops is basically:

Unable to handle kernel paging request at virtual address 000112a5.
*pde = ma 00000000 55555000, with ca call chain of:

    handle_mm_fault
    do_mm_pgoff
    handle_IRQ_event
    do_IRQ
    evtch_do_upcall
    hypervisor_callback

anyone got any ideas?


  --cw


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Xen 2.0 / XenLinux 2.6.9 oops in handle_mm_fault

[Christian Limpach]

On Thu, Nov 11, 2004 at 05:04:33PM -0800, Chris Wedgwood wrote:
> Using XenLinux 2.6.9 trying to do very minimal work in domain0
> (building a module to get networking up) I'm seeing an oops in
> handle_mm_fault...  the full oops is visible at (photo, serial console
> isn't working) http://stupidest.org/xen2-crash2.png

It's actually an oops in do_timer, which fails because the regs
parameter points somewhere odd (eax=11273).

Not sure how this can happen -- could you please confirm that xen
and linux are built from the same sources using a known-good
compiler.  If you can reproduce this, then increasing the amount
of stack being printed would definitely help.

    christian



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Xen 2.0 / XenLinux 2.6.9 oops in handle_mm_fault

[Chris Wedgwood]

On Thu, Nov 11, 2004 at 05:04:33PM -0800, Chris Wedgwood wrote:

> Unable to handle kernel paging request at virtual address 000112a5.
> *pde = ma 00000000 55555000, with ca call chain of:
>
>     handle_mm_fault
>     do_mm_pgoff
>     handle_IRQ_event
>     do_IRQ
>     evtch_do_upcall
>     hypervisor_callback

As suggested I tried increasing the stack-size (to 16K) for both Xen
and the dom0 kernel without any luck...

Next I tried (I should have tried this earlier probably) was to check
CONFIG_REGPARM and sure enough having that set causes the problem.
Without that things seem to be working more or less as expected...


Thanks,
  --cw


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: Xen 2.0 / XenLinux 2.6.9 oops in handle_mm_fault

[Nuno Silva]

Chris Wedgwood wrote:
> Using XenLinux 2.6.9 trying to do very minimal work in domain0
> (building a module to get networking up) I'm seeing an oops in
> handle_mm_fault...  the full oops is visible at (photo, serial console
> isn't working) http://stupidest.org/xen2-crash2.png
> 
> The oops is basically:
> 
> Unable to handle kernel paging request at virtual address 000112a5.
> *pde = ma 00000000 55555000, with ca call chain of:
> 
>     handle_mm_fault
>     do_mm_pgoff
>     handle_IRQ_event
>     do_IRQ
>     evtch_do_upcall
>     hypervisor_callback
> 
> anyone got any ideas?

Wild guess: you recompiled with CONFIG_PREEMPT enabled. Try disabling 
PREEMPT and report back.

Regards,
Nuno Silva



-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: Xen 2.0 / XenLinux 2.6.9 oops in handle_mm_fault

[Chris Wedgwood]

On Sat, Nov 13, 2004 at 01:51:11AM +0000, Nuno Silva wrote:

> Wild guess: you recompiled with CONFIG_PREEMPT enabled.

I did.  It turns out it was CONFIG_REGPARM=y that is the problem (I
reported this to the list but since it has subscribe only b0rkedness
it's not appeared yet).

Right now I'm running with CONFIG_PREEMPT=y and don't see any
problems.  Is this purely luck right now.



  --cw


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: Xen 2.0 / XenLinux 2.6.9 oops in handle_mm_fault

[Nuno Silva]

Chris Wedgwood wrote:
> On Sat, Nov 13, 2004 at 01:51:11AM +0000, Nuno Silva wrote:
> 
> 
>>Wild guess: you recompiled with CONFIG_PREEMPT enabled.
> 
> 
> I did.  It turns out it was CONFIG_REGPARM=y that is the problem (I
> reported this to the list but since it has subscribe only b0rkedness
> it's not appeared yet).

Great! Nice to know, thanks :-)

> Right now I'm running with CONFIG_PREEMPT=y and don't see any
> problems.  Is this purely luck right now.

I don't know if PREEMPT causes some setups to oops, it was just a guess 
because many bugs are inflated by that option.

Regards,
Nuno Silva


-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

Re: Xen 2.0 / XenLinux 2.6.9 oops in handle_mm_fault

[Christian Limpach]

On Sat, Nov 13, 2004 at 01:51:11AM +0000, Nuno Silva wrote:
> Chris Wedgwood wrote:
> >Using XenLinux 2.6.9 trying to do very minimal work in domain0
> >(building a module to get networking up) I'm seeing an oops in
> >handle_mm_fault...  the full oops is visible at (photo, serial console
> >isn't working) http://stupidest.org/xen2-crash2.png
> >
> >The oops is basically:
> >
> >Unable to handle kernel paging request at virtual address 000112a5.
> >*pde = ma 00000000 55555000, with ca call chain of:
> >
> >    handle_mm_fault
> >    do_mm_pgoff
> >    handle_IRQ_event
> >    do_IRQ
> >    evtch_do_upcall
> >    hypervisor_callback
> >
> >anyone got any ideas?
> 
> Wild guess: you recompiled with CONFIG_PREEMPT enabled. Try disabling 
> PREEMPT and report back.

no -- we always build with CONFIG_PREEMPT and the non-CONFIG_PREEMPT
code paths are hardly tested at all.

    christian



-------------------------------------------------------
This SF.Net email is sponsored by: InterSystems CACHE
FREE OODBMS DOWNLOAD - A multidimensional database that combines
robust object and relational technologies, making it a perfect match
for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8

RE: Xen 2.0 / XenLinux 2.6.9 oops in handle_mm_fault

[Ian Pratt]

> Using XenLinux 2.6.9 trying to do very minimal work in 
> domain0 (building a module to get networking up) I'm seeing 
> an oops in handle_mm_fault...  the full oops is visible at 
> (photo, serial console isn't working) 
> http://stupidest.org/xen2-crash2.png

Very odd. Can you reproduce reliably? Are you using the real 2.0 rather
than a release candidate?
Did you compile your own kernel or use the supplied binaries? If the
former, what compiler are you using?

> Unable to handle kernel paging request at virtual address 000112a5.
> *pde = ma 00000000 55555000, with ca call chain of:

You haven't told us the function containing the EIP where the fault
occurred. 
Do you have xend running? What about other domains?

>     handle_mm_fault
>     do_mm_pgoff
>     handle_IRQ_event
>     do_IRQ
>     evtch_do_upcall
>     hypervisor_callback

Ian


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_idU88&alloc_id\x12065&op=click

Re: Xen 2.0 / XenLinux 2.6.9 oops in handle_mm_fault

[Steven Hand]

>> Using XenLinux 2.6.9 trying to do very minimal work in
>> domain0 (building a module to get networking up) I'm seeing
>> an oops in handle_mm_fault...  the full oops is visible at
>> (photo, serial console isn't working)
>> http://stupidest.org/xen2-crash2.png
>
>Very odd. Can you reproduce reliably? Are you using the real 2.0 rather
>than a release candidate?
>Did you compile your own kernel or use the supplied binaries? If the
>former, what compiler are you using?

Another common error is to forget to build the module with ARCH=xen 
(in which case all bets are off...)

S.



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click