Re: BUG at drivers/xen/balloon.c:353

[David Vrabel <david.vrabel@citrix.com>]

On 10/09/13 17:52, Boris Ostrovsky wrote:
> On 09/10/2013 12:30 PM, David Vrabel wrote:
>> On 10/09/13 17:12, Ian Jackson wrote:
>>> Now that we are on Linux 3.4.y, I am once again trying to commission
>>> my pair of weevils.  Things are better than they were but it's still
>>> totally broken.
>>>
>>> Below you can see the logs from my adhoc osstest flight 19169.  I
>>> looked at one of the save/restore failures (test-amd64-i386-xl) and
>>> see this:
>>>
>>> Sep 10 00:03:09.073066 [  240.142970] kernel BUG at
>>> drivers/xen/balloon.c:353!
>>>
>>> I'd appreciate any help available with debugging this.  The logs are
>>> here:
>>>
>>>   
>>> http://www.chiark.greenend.org.uk/~xensrcts/logs/19169/test-amd64-i386-xl/info.html
>>>
>>>
>>> I looked at the code in balloon.c:
>>>
>>>         BUG_ON(!xen_feature(XENFEAT_auto_translated_physmap) &&
>>>                phys_to_machine_mapping_valid(pfn));
>>>
>>> Ian.
>>>
>>> Sep 10 00:03:08.629133 [  240.142955] ------------[ cut here
>>> ]------------
>>> Sep 10 00:03:09.073066 [  240.142970] kernel BUG at
>>> drivers/xen/balloon.c:353!
>>> Sep 10 00:03:09.073102 [  240.142974] invalid opcode: 0000 [#1] SMP
>>> Sep 10 00:03:09.081064 [  240.142978] Modules linked in:
>>> xen_acpi_processor xen_gntalloc ext4 jbd2 mbcache e1000e
>>> Sep 10 00:03:09.093059 [  240.142987] CPU: 6 PID: 847 Comm:
>>> kworker/6:1 Not tainted 3.11.0+ #1
>> There's a know bug in this area with 3.11+ (3.12-rc0) and there is
>> pending pull request for the fix.
> 
> If you are thinking about preempt count bug then its' likely a different
> one.

No, I'm thinking of Wei's "xen/balloon: remove BUG_ON in
increase_reservation" which actually doesn't appear to be queued up.

David