On Thu, Jan 13, 2022 at 8:51 AM David Dunn <daviddunn@xxxxxxxxxx> wrote:
>
> Jim,
>
> On Wed, Jan 12, 2022 at 5:15 PM Jim Mattson <jmattson@xxxxxxxxxx> wrote:
>
> > +bool is_amd_cpu(void)
> > +{
> > + return cpu_vendor_string_is("AuthenticAMD") ||
> > + cpu_vendor_string_is("AMDisbetter!");
> > +}
> > +
>
> The original code only checked for AuthenticAMD. I don't think it is
> necessary or wise to add the check for early K5 samples. Do we know
> whether they even need this KVM logic?
The original code should handle K5 CPUs in the next block:
/* On parts with <40 physical address bits, the area is fully hidden */
if (vm->pa_bits < 40)
return max_gfn;
But, you're right. K5 predates AMD support for SVM, so we don't need
to test kvm functionality on early K5 samples.
I'll remove the second disjunct in v2.
Thanks!
