Re: [PATCH v2 1/4] xen/libxc: Allow changes to hypervisor CPUID leaf from config file

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

On 03/11/2014 04:45 AM, Jan Beulich wrote:
>>>> On 11.03.14 at 04:54, Boris Ostrovsky <boris.ostrovsky@oracle.com> wrote:
>> Currently only "real" cpuid leaves can be overwritten by users via
>> 'cpuid' option in the configuration file. This patch provides ability to
>> do the same for hypervisor leaves (those in the 0x40000000 range).
> And honestly I'm not certain we want to go that far. Limiting the
> number of leaves seems reasonable (even CPU vendors had to
> introduce this), but altering other hypervisor CPUID output seems
> to only call for trouble.


If we do this I suspect we can get rid of the sysctl altogether. 
Alternatively we can do this as part of policy in libxc but preserve 
ability to change the policy (and keep sysctl). I slightly prefer the 
latter as I think it's a useful feature but I can see reasons for not 
doing it.


>
>> +struct xen_sysctl_cpuid {
>> +     uint32_t input[2];
>> +     uint32_t eax;
>> +     uint32_t ebx;
>> +     uint32_t ecx;
>> +     uint32_t edx;
>> +};
> Having just the four register fields here would be enough - eax
> and ecx would simply be IN/OUT (and if need be in the future,
> ebx/edx could become IN/OUT too without altering the structure
> layout).

Right. I just blindly copied xen_domctl_cpuid here.

Thanks.
-boris