Xen devel - Dec 2009 - pvops-dom0 console freezes after latest ballooning patch (+possible fix)

Setup: xen 3.4.1, 32-bit linux dom0 2.6.31.6 from pvops-git tree from 2009-12-03

Since the latest ballooning patch "Add hugepage support to balloon
driver" i experience dom0 console freezes. Additional to that there is a
stacktrace about a kernel-bug in drivers/xen/balloon.c in line 346, invalide
opcode 0000 (i can send the full stacktrace if someone is interested).

The cause for that seems to be that in the decrease_reservation-funtion the
PageHighMem - check is not executed on the assocciated page for the current pfn.
Its always the last page that got allocated in the for-loop before, when all the
pages are allocated.

With the attached patch that crashes dont occur anymore. But im not a
kernel-developer so im not sure that this is the correct fix.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 12/09/09 11:22, Ronny.Hegewald@online.de wrote:
> Setup: xen 3.4.1, 32-bit linux dom0 2.6.31.6 from pvops-git tree from 
> 2009-12-03
>
>
> Since the latest ballooning patch "Add hugepage support to balloon 
> driver 
>
<http://git.kernel.org/?p=linux/kernel/git/jeremy/xen.git;a=commit;h=0e898d5e9b893324435e0e77b67fe8d2f0523dbe>"
> i experience dom0 console freezes. Additional to that there is a 
> stacktrace about a kernel-bug in drivers/xen/balloon.c in line 346, 
> invalide opcode 0000 (i can send the full stacktrace if someone is 
> interested).
>
Yes, please.  Presumably its from the BUG_ON testing the hypercall 
return status?

> The cause for that seems to be that in the 
> decrease_reservation-funtion the PageHighMem - check is not executed 
> on the assocciated page for the current pfn. Its always the last page 
> that got allocated in the for-loop before, when all the pages are 
> allocated.
>
> With the attached patch that crashes dont occur anymore. But im not a 
> kernel-developer so im not sure that this is the correct fix.
It looks reasonable to me.  Dave?

     J

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Wednesday 09 December 2009, Jeremy Fitzhardinge
wrote:
> On 12/09/09 11:22, Ronny.Hegewald@online.de wrote:
> > Setup: xen 3.4.1, 32-bit linux dom0 2.6.31.6 from pvops-git tree from
> > 2009-12-03
> >
> >
> > Since the latest ballooning patch "Add hugepage support to
balloon
> > driver
> >
<http://git.kernel.org/?p=linux/kernel/git/jeremy/xen.git;a=commit;h=0e89
> >8d5e9b893324435e0e77b67fe8d2f0523dbe>" i experience dom0
console freezes.
> > Additional to that there is a stacktrace about a kernel-bug in
> > drivers/xen/balloon.c in line 346, invalide opcode 0000 (i can send
the
> > full stacktrace if someone is interested).
> 
> Yes, please.  Presumably its from the BUG_ON testing the hypercall
> return status?
> 
> > The cause for that seems to be that in the
> > decrease_reservation-funtion the PageHighMem - check is not executed
> > on the assocciated page for the current pfn. Its always the last page
> > that got allocated in the for-loop before, when all the pages are
> > allocated.
> >
> > With the attached patch that crashes dont occur anymore. But im not a
> > kernel-developer so im not sure that this is the correct fix.
> 
> It looks reasonable to me.  Dave?
> 
>      J
Yes, absolutely.  It''s the correct fix.  Good catch.

Dave McCracken
Oracle Corp.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 12/09/09 12:20, Dave McCracken wrote:
> Yes, absolutely.  It''s the correct fix.  Good catch.
>    
OK, I''ll add an acked-by for you.

Ronny, thanks for the patch and the good description of the bug it fixes 
and how; together they''re almost a perfect patch.  I added a 
Signed-off-by: line for you.

     J

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

>On 12/09/09 11:22, Ronny.Hegewald@online.de wrote:
>> Setup: xen 3.4.1, 32-bit linux dom0 2.6.31.6 from pvops-git tree from 
>> 2009-12-03
>>
>>
>> Since the latest ballooning patch "Add hugepage support to balloon
>> driver 
>> 
><http://git.kernel.org/?p=linux/kernel/git/jeremy/xen.git;a=commit;h=0e898d5e9b
>893324435e0e77b67fe8d2f0523dbe>" 
>> i experience dom0 console freezes. Additional to that there is a 
>> stacktrace about a kernel-bug in drivers/xen/balloon.c in line 346, 
>> invalide opcode 0000 (i can send the full stacktrace if someone is 
>> interested).
>>
>
>Yes, please.  Presumably its from the BUG_ON testing the hypercall 
>return status?
Attached the stacktrace.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

>On 12/09/09 11:22, Ronny.Hegewald@online.de wrote:
>> Setup: xen 3.4.1, 32-bit linux dom0 2.6.31.6 from pvops-git tree from 
>> 2009-12-03
>>
>>
>> Since the latest ballooning patch "Add hugepage support to balloon
>> driver 
>> 
><http://git.kernel.org/?p=linux/kernel/git/jeremy/xen.git;a=commit;h=0e898d5e9b
>893324435e0e77b67fe8d2f0523dbe>" 
>> i experience dom0 console freezes. Additional to that there is a 
>> stacktrace about a kernel-bug in drivers/xen/balloon.c in line 346, 
>> invalide opcode 0000 (i can send the full stacktrace if someone is 
>> interested).
>>
>
>Yes, please.  Presumably its from the BUG_ON testing the hypercall 
>return status?
>
This time really with the attachment ...

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 12/09/09 12:55, Ronny.Hegewald@online.de wrote:
>> On 12/09/09 11:22, Ronny.Hegewald@online.de wrote:
>>      
>>> Setup: xen 3.4.1, 32-bit linux dom0 2.6.31.6 from pvops-git tree
from
>>> 2009-12-03
>>>
>>>
>>> Since the latest ballooning patch "Add hugepage support to
balloon
>>> driver
>>>
>>>        
>>
<http://git.kernel.org/?p=linux/kernel/git/jeremy/xen.git;a=commit;h=0e898d5e9b
>> 893324435e0e77b67fe8d2f0523dbe>"
>>      
>>> i experience dom0 console freezes. Additional to that there is a
>>> stacktrace about a kernel-bug in drivers/xen/balloon.c in line 346,
>>> invalide opcode 0000 (i can send the full stacktrace if someone is
>>> interested).
>>>
>>>        
>> Yes, please.  Presumably its from the BUG_ON testing the hypercall
>> return status?
>>
>>      
> This time really with the attachment ...
Thanks.

BTW, it looks like I hadn''t merged the hugepage into xen/master until 
now.  Did you merge it yourself?  Or trying something else?

     J


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

>On 12/09/09 12:55, Ronny.Hegewald@online.de wrote:
>>> On 12/09/09 11:22, Ronny.Hegewald@online.de wrote:
>>>      
>>>> Setup: xen 3.4.1, 32-bit linux dom0 2.6.31.6 from pvops-git
tree from
>>>> 2009-12-03
>>>>
>>>>
>>>> Since the latest ballooning patch "Add hugepage support to
balloon
>>>> driver
>>>>
>>>>        
>Thanks.
>
>BTW, it looks like I hadn''t merged the hugepage into xen/master
until
>now.  Did you merge it yourself?  Or trying something else?
>
>     J
>
No, i used a unpatched xen/master.

That hugepage patch got merged into xen/master on 2009-11-06 (commit  
c38510904c76f910610a6e790aab594cfa5c7a2c) through xen/dom0/backend/core that was
merged with xen/balloon before.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel