Messages in this thread | | | Date | Thu, 9 Sep 2021 15:22:51 +0200 | From | Joerg Roedel <> | Subject | Re: [PATCH v2 1/4] KVM: SVM: Get rid of *ghcb_msr_bits() functions |
| |
Hi Sean,
On Wed, Sep 01, 2021 at 09:31:52PM +0000, Sean Christopherson wrote: > On Wed, Sep 01, 2021, Sean Christopherson wrote: > > control->ghcb_gpa = MAKE_GHCB_MSR_RESP(cpuid_reg, cpuid_value);
Made that change, but kept the set_ghcb_msr_cpuid_resp() and renamed it to ghcb_msr_cpuid_resp(). It now returns the MSR value for the CPUID response.
I like the keep the more complicated response setters as functions and not macros for readability.
> case GHCB_MSR_SEV_INFO_REQ: > control->ghcb_gpa = GHCB_MSR_SEV_INFO(GHCB_VERSION_MAX, > GHCB_VERSION_MIN, > sev_enc_bit)); > break; > > and drop set_ghcb_msr() altogether.
Makes sense, I replaced the set_ghcb_msr() calls with the above.
> Side topic, what about renaming control->ghcb_gpa => control->ghcb_msr so that > the code for the MSR protocol is a bit more self-documenting? The APM defines > the field as "Guest physical address of GHCB", so it's not exactly prescribing a > specific name.
No strong opinion here, I let this up to the AMD engineers to decide. If we change the name I can add a separate patch for this.
Regards,
Joerg
| |