Re: [PATCH 2/2] Xen/vMCE: bugfix to remove problematic is_vmce_ready check

[Christoph Egger <chegger@amazon.de>]

On 13.05.13 15:35, Liu, Jinsong wrote:
> Christoph Egger wrote:
>> On 13.05.13 12:44, Liu, Jinsong wrote:
>>>>> Please refer to the description of patch 2/2, especially
>>>>>
>>>>>     * For dom0, if really need check, it should check whether vMCE
>>>>>     injection for dom0 ready (say, exception trap bounce check,
>>>>>     which has been done at inject_vmce()), not check dom0 mcelog
>>>>>     ready (which has been done at mce_softirq() before send global
>>>>> virq to dom0). 
>>>>>
>>>>> Which means before hypervisor send error log via virq to dom0,
>>>>> current code has checked whether mcelog ready at dom0 or not -->
>>>>> that's the right place for your concern, and it has indeed done
>>>>> check.
>>>>
>>>> I think, I do not understand the patch description.
>>>> Let me rephrase if I do now due to this discussion:
>>>>
>>>> The mcelog driver in Dom0 registers itself to the virq handler to
>>>> provide the machine check logging service.
>>>
>>> Yes.
>>>
>>>> Xen checks if a virq handler has been registered
>>>
>>> Yes.
>>>
>>>> but does not check
>>>> if the dom0 handler is actually ready to take the errors.
>>>> This patch fixes this.
>>>>
>>>
>>> I'm not clear your question 'does not check if the dom0 handler
>>> is actually ready to take the errors'. Could you elaborate more your
>>> concern at this point? 
>>
>> Yes, this is exactly my question. You got it.
>>
>> Christoph
> 
> Hmm, seems you misunderstand my word. What I meant is,
> I don't know what you are asking by 'does not check if the dom0 handler is actually ready to take the errors'.
> Could you elaborate more your question?

I reread your patch description:

> * For dom0, if really need check, it should check whether vMCE
                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> injection for dom0 ready (say, exception trap bounce check, which
  ^^^^^^^^^^^^^^^^^^^^^^^^
> has been done at inject_vmce()), not check dom0 mcelog ready (which
>                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^
> has been done at mce_softirq() before send global virq to dom0).

My question is: Is it possible when mcelog driver registers
the virq handler that it cannot deal with machine check errors immediately?

Christoph