RE: [Xen-devel] [PATCH] X86 MCE: Add SRAR handler

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 06.10.11 at 20:40, "Liu, Jinsong" <jinsong.liu@xxxxxxxxx> wrote:
> Liu, Jinsong wrote:
>> Liu, Jinsong wrote:
>>> Jan Beulich wrote:
>>>>>>> On 30.09.11 at 11:42, "Liu, Jinsong" <jinsong.liu@xxxxxxxxx>
>>>>>>> wrote:
>>>>> Jan Beulich wrote:
>>>>>>>>> On 30.09.11 at 10:21, "Liu, Jinsong" <jinsong.liu@xxxxxxxxx>
>>>>>>>>> wrote:
>>>>>>> Executing guest code will not satisfy the check
>>>>>>> if ( !(gstatus & MCG_STATUS_RIPV) && !guest_mode(regs))         return
>>>>>>> -1; so it would not panic system.
>>>>>> 
>>>>>> Exactly. But it should when the prefetch was to hypervisor code.
>>>>>> 
>>>>> 
>>>>> Wouldn't processor refresh instruction prefetch queue under such
>>>>> case?
>>>> 
>>>> That's a question that you are better positioned to answer than me.
>>>> But the SRAR errors being a sub-category of uncorrected errors I
>>>> would think it can't be that simple.
>>>> 
>>> 
>>> Hmm, I will check this question internally first.
>>> BTW, we would have 7 days holiday (1/10 ~ 7/10), so email reply
>>> maybe some slow. 
>>> 
>>> Thanks,
>>> Jinsong
>> 
>> Ah, just think our talking context: the prefetched instruction would
>> have been flushed since we now at mce exception context. So I think
>> no need to overkill here, just let guest handle it --> who own, who
>> take.  
>> 
>> Thanks,
>> Jinsong
> 
> Jan,
> 
> Do you think following is OK?
> 
> if ( !(gstatus & MCG_STATUS_RIPV) && !guest_mode(regs)) 
>     return -1;

That's what we have currently, and as I said earlier I don't think using
the result of guest_mode() for any decision is valid when the EIPV bit
is clear.

Jan

> Thanks,
> Jinsong




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