[Lguest] lguest: unhandled trap 13

Avi Kivity avi at redhat.com
Thu Feb 11 18:25:33 EST 2010


On 02/11/2010 08:47 AM, Rusty Russell wrote:
> On Thu, 11 Feb 2010 05:07:49 pm Avi Kivity wrote:
>    
>> On 02/11/2010 08:33 AM, Rusty Russell wrote:
>>      
>>> I don't have my manuals on me,
>>> but It Used To Work.
>>>
>>>        
>> IIRC we used to inject a #GP in that case.  But that's not guaranteed by
>> the documentation.  We could change kvm, but we can't change other
>> hypervisors.
>>      
> Yes, but from the lguest POV it's a pain; we perform differently under VT
> than on real hardware, yet we shouldn't care AFAICT.
>    

That's because you use an instruction that operates differently under VT 
vs. real hardware.

> It is definitely weird to swallow it silently.

It's not silent, we return -ESOMETHING.  The idea is we might want 
userspace-invoked hypercalls one day.

> A #UD or #GP seems logical
> (we handle both).  But I'll revert to our own hypercalls using int 0x1f.
>    

Er.  int 0x1f is reserved.

-- 
Do not meddle in the internals of kernels, for they are subtle and quick to panic.



More information about the Lguest mailing list