Re: [Xen-devel] PTE Write access is not set

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 14.12.11 at 15:59, Liwei <xieliwei@xxxxxxxxx> wrote:
> On 14 December 2011 22:52, Jan Beulich <JBeulich@xxxxxxxx> wrote:
>>
>> So this suggests that the virtual firmware has a problem coming back
>> up with the passed through device still active (and the IOMMU may in
>> fact be shielding the rest of the system from bad effects of this).
>>
>> This by itself is a problem, and the sudden death of the guest is
>> another (particularly if indeed there's no trace of that event in any
>> logs). Debugging this may be difficult for anyone who can't
>> reproduce it, so we'll have to rely on you providing more insight.
>> Enable more debugging/verbosity in the hypervisor and try again is
>> probably all I can suggest at this point.
> 
> Currently I have earlyprintk=xen, loglvl=all and guest_loglvl=all.

"earlyprintk=xen" is completely irrelevant here. For the others, as
I see them commonly misplaced, I'll take the risk of embarrassing you
by asking whether they are on the Xen command line (and not the
kernel one).

> Other than adding loglevel=10 and debug initcall_debug, what other
> options and/or procedures do you suggest?

"hvm_debug=" comes to mind (grep for the various DBG_LEVEL_*
defines and uses to see which one might provide greatest benefit,
and be prepared to go through a lot of data afterwards, though
presumably just the tail is going to be interesting if you make the
guest not immediately restart after death).

Jan


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