On 21.05.2024 12:00, Sergiy Kibrik wrote:
> 21.05.24 09:05, Jan Beulich:
>>> This is because MCG_LMCE_P is defined in arch/x86/cpu/mcheck/x86_mca.h
>>> -- so either MCG_LMCE_P (+ a bunch of MCG_* declarations) has to be
>>> moved to common header to be accessible, or local x86_mca.h got to be
>>> included from common arch/x86/include/asm/mce.h.
>>>
>>> As for the MCG_* declarations movement I didn't think there's a good
>>> enough reason to do it; as for the inclusion of x86_mca.h it didn't look
>>> nice at all.
>> I'm afraid I don't follow the latter: Why's including x86_mca.h any worse
>> than what you do right now?
> 
> To include x86_mca.h from asm/mce.h something like this line would be 
> needed:
> 
> #include "../../cpu/mcheck/x86_mca.h"
> 
> I've found only two include-s of such kind, so I presume they're not common.

Indeed, and I have to apologize for not reading your earlier reply quite
right.

Jan

> Besides xen/sched.h includes asm/mce.h before declaration of struct 
> vcpu, so body of vmce_has_lmce(const struct vcpu *v) can't really be 
> compiled in asm/mce.h
> 
>    -Sergiy


Reply via email to