[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] vMCE vs migration

>>> On 03.02.12 at 13:34, "Liu, Jinsong" <jinsong.liu@xxxxxxxxx> wrote:
> Jan Beulich wrote:
>>>>> On 03.02.12 at 08:18, "Liu, Jinsong" <jinsong.liu@xxxxxxxxx> wrote:
>>> How about use static bank size, say 256, which architecture
>>> (IA32_MCG_CAP MSR) defined max bank number?
>> The maximum, if any, is 32 (beyond which collision with other defined
>> MSRs would arise). But with the possibility of having a discontiguous
>> or relocated MSR space, I don't think any static maximum would be a
>> good idea.
>> Jan
> The advantages of static max bank is simple, any disadvantages of static 
> solution?

Lack of forward compatibility. I may even be that the use of uint8_t
wasn't really a good choice of mine.

> or, can we use bank_entry listed in vmce->impact_header for mci_ctl, like 
> what 
> mci_status/mci_addr/mci_misc do?

I don't think so - the control register (obviously) must be accessible
independent of any ongoing (latched) MCE.

> I just think the patch may be too complicated for a minor issue.

Is a guest crash minor? After all there is no guarantee that the
accesses to these MSRs are exception handler protected in a
way similar to what current Linux does.


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.