Re: [PATCH] x86/MCE: sanitize domain/vcpu ID handling

[Boris Ostrovsky <boris.ostrovsky@oracle.com>]

On 02/23/2017 05:05 AM, Jan Beulich wrote:
>>>> On 23.02.17 at 11:01, <haozhong.zhang@intel.com> wrote:
>> On 02/23/17 02:28 -0700, Jan Beulich wrote:
>>> --- a/xen/arch/x86/cpu/mcheck/mcaction.c
>>> +++ b/xen/arch/x86/cpu/mcheck/mcaction.c
>>> @@ -100,7 +100,8 @@ mc_memerr_dhandler(struct mca_binfo *bin
>>>                      goto vmce_failed;
>>>                  }
>>>  
>>> -                if ( boot_cpu_data.x86_vendor == X86_VENDOR_INTEL )
>>> +                if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL ||
>>> +                    global->mc_vcpuid == XEN_MC_VCPUID_INVALID)
>>>                      vmce_vcpuid = VMCE_INJECT_BROADCAST;
>>>                  else
>>>                      vmce_vcpuid = global->mc_vcpuid;
>> If an invalid vcpuid is got on AMD machine, should we report error
>> or inject to a default vcpu (vcpu0?) ?
> Well, broadcasting in that case seems the best option to me,
> but let's add AMD maintainers to Cc.

Yes, I think we should broadcast if we don't know VCPU identity.

OTOH, is mc_memerr_dhandler() even called on AMD? The only caller I see
is intel_memerr_dhandler().

-boris


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel