Re: [PATCH 05/12] KVM: x86: Use KVM-governed feature framework to track "XSAVES enabled"

From: Sean Christopherson
Date: Wed Feb 22 2023 - 13:56:10 EST


On Tue, Feb 21, 2023, Yu Zhang wrote:
> On Fri, Feb 17, 2023 at 03:10:15PM -0800, Sean Christopherson wrote:
> > Use the governed feature framework to track if XSAVES is "enabled", i.e.
> > if XSAVES can be used by the guest. Add a comment in the SVM code to
> > explain the very unintuitive logic of deliberately NOT checking if XSAVES
> > is enumerated in the guest CPUID model.
> >
> > No functional change intended.
>
> xsaves_enabled in struct kvm_vcpu_arch is no longer used. But instead of
> just deleting it, maybe we could move 'bool load_eoi_exitmap_pending' to
> its place, so 7 bytes can be saved for each struct kvm_vcpu_arch:

I prefer leaving load_eoi_exitmap_pending where it is so that it's co-located with
ioapic_handled_vectors. I agree wasting 7 bytes is unfortunate, but I don't want
to take an ad hoc approach to shrinking per-vCPU structs. See the link below for
more discussion.

https://lore.kernel.org/all/20230213163351.30704-1-minipli@xxxxxxxxxxxxxx

>
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index cd660de02f7b..0eef5469c165 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -740,7 +740,6 @@ struct kvm_vcpu_arch {
> u64 efer;
> u64 apic_base;
> struct kvm_lapic *apic; /* kernel irqchip context */
> - bool load_eoi_exitmap_pending;
> DECLARE_BITMAP(ioapic_handled_vectors, 256);
> unsigned long apic_attention;
> int32_t apic_arb_prio;
> @@ -750,7 +749,7 @@ struct kvm_vcpu_arch {
> u64 smi_count;
> bool at_instruction_boundary;
> bool tpr_access_reporting;
> - bool xsaves_enabled;
> + bool load_eoi_exitmap_pending;
> bool xfd_no_write_intercept;
> u64 ia32_xss;
> u64 microcode_version;
>
> B.R.
> Yu
>