[Xen-devel] Re: kernel BUG at arch/x86/xen/mmu.c:1860!

[MaoXiaoyun <tinnycloud@xxxxxxxxxxx>]

Hi:
     Unfortunately I met the exactly same bug today. With pvops kernel 2.6.32.36, and xen 4.0.1.
     Kernel Panic and serial log attached. 
 
     Our test cases is quite simple, on a single physical host, we start 12 HVMS(windows 2003),
each of the HVM reboot every 10minutes.
 
     The bug is easy to hit on our 48G machine(in hours). But We haven't hit the bug in our 24G
machine(we have three 24G machine, all works fine.)  -----Is is possible related to Memory capacity?
 
Taking a look at the serial output,  the Dom0 code is attempting to pin what it thins
is a "PGT_l3_page_table", however the hypervisor returns -EINVAL because it actually  is a "PGT_writable_page".
 
(XEN) mm.c:2364:d0 Bad type (saw 7400000000000001 != exp 4000 0000 0000 0000) for mfn 898a41 (pfn 9ca41)
(XEN) mm.c:2733:d0 Error while pinning mfn 898a41
 
And  before that quite a lot abnormal grant table log like :
 
(XEN) grant_table.c:1717:d0 Bad grant reference 4294965983
(XEN) grant_table.c:266:d0 Bad flags (0) or dom (0). (expected dom 0)
(XEN) grant_table.c:266:d0 Bad flags (0) or dom (0). (expected dom 0)
(XEN) grant_table.c:266:d0 Bad flags (0) or dom (0). (expected dom 0)
(XEN) grant_table.c:1717:d0 Bad grant reference 4294965888
(XEN) grant_table.c:1717:d0 Bad grant reference 4294965983
 
It looks like something wrong with grant table.
 
Many thanks.
 
> From: Jeremy Fitzhardinge <jeremy@xxxxxxxx>
> Subject: Re: [Xen-devel] [SPAM] Re: kernel BUG at
> arch/x86/xen/mmu.c:1860! - ideas.
> To: Ian Campbell <Ian.Campbell@xxxxxxxxxx>
> Cc: Dave Hunter <dave@xxxxxxxxxx>, Teck Choon Giam
> <giamteckchoon@xxxxxxxxx>, "xen-devel@xxxxxxxxxxxxxxxxxxx"
> xen-devel@xxxxxxxxxxxxxxxxxxx
 
> On 04/06/2011 12:53 AM, Ian Campbell wrote:
> > Please don't top post.
> >
> > On Wed, 2011-04-06 at 00:20 +0100, Dave Hunter wrote:
> >> Is it likely that Debian would release an updated kernel in squeeze with
> >> this configuration? (sorry, this might not be the place to ask).
> > I doubt they will, enabling DEBUG_PAGEALLOC seems very much like a
> > workaround not a solution to me.
>
> Yes, it will impose a pretty large performance overhead.>
> J
>
> 

Attachment: kernel.txt
Description: Text document

Attachment: serial.txt
Description: Text document

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