Re: [PATCH v3 10/12] KVM: SEV: Forcefully invalidate SNP VMSA if its backing gmem page is zapped

[Sean Christopherson]

On Wed, Jul 01, 2026, Michael Roth wrote:
> On Tue, Jun 30, 2026 at 03:26:05PM -0700, Sean Christopherson wrote:
> > ---
> >  arch/x86/include/asm/kvm-x86-ops.h |  2 +
> >  arch/x86/include/asm/kvm_host.h    |  4 ++
> >  arch/x86/kvm/mmu/mmu.c             |  5 ++
> >  arch/x86/kvm/svm/sev.c             | 79 +++++++++++++++++++++++++-----
> >  arch/x86/kvm/svm/svm.c             |  2 +
> >  arch/x86/kvm/svm/svm.h             |  2 +
> >  arch/x86/kvm/x86.c                 |  6 +++
> >  include/linux/kvm_host.h           |  1 +
> >  virt/kvm/guest_memfd.c             |  4 ++
> >  9 files changed, 94 insertions(+), 11 deletions(-)
> > 
> ...
> > diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
> > index cf7c1a437f38..123e7bf687ef 100644
> > --- a/arch/x86/kvm/svm/svm.h
> > +++ b/arch/x86/kvm/svm/svm.h
> > @@ -996,6 +996,7 @@ static inline struct page *snp_safe_alloc_page(void)
> >  {
> >  	return snp_safe_alloc_page_node(numa_node_id(), GFP_KERNEL_ACCOUNT);
> >  }
> > +void sev_snp_reload_vmsa(struct kvm_vcpu *vcpu);
> >  
> >  int sev_vcpu_create(struct kvm_vcpu *vcpu);
> >  void sev_free_vcpu(struct kvm_vcpu *vcpu);
> > @@ -1009,6 +1010,7 @@ int sev_dev_get_attr(u32 group, u64 attr, u64 *val);
> >  extern unsigned int max_sev_asid;
> >  void sev_handle_rmp_fault(struct kvm_vcpu *vcpu, gpa_t gpa, u64 error_code);
> >  int sev_gmem_prepare(struct kvm *kvm, kvm_pfn_t pfn, gfn_t gfn, int max_order);
> > +void sev_gmem_invalidate_range(struct kvm *kvm, struct kvm_gfn_range *range);
> >  void sev_gmem_reclaim_memory(kvm_pfn_t start, kvm_pfn_t end);
> >  int sev_gmem_max_mapping_level(struct kvm *kvm, kvm_pfn_t pfn, bool is_private);
> >  struct vmcb_save_area *sev_decrypt_vmsa(struct kvm_vcpu *vcpu);
> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > index bc0c3163f4a3..0bb50997c0e3 100644
> > --- a/arch/x86/kvm/x86.c
> > +++ b/arch/x86/kvm/x86.c
> > @@ -8170,6 +8170,8 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
> >  				goto out;
> >  			}
> >  		}
> > +		if (kvm_check_request(KVM_REQ_VMSA_PAGE_RELOAD, vcpu))
> > +			kvm_x86_call(reload_vmsa)(vcpu);
> 
> VMSA is SVM/SEV-specific, and while the event/handling might be SEV-specific,

It's not just SEV specific, it's SEV-ES+ specific.  Splitting hairs, but IMO
the distinction matters because while there is a "save area" for SVM and SEV
vCPUs, they don't have a dedicated, separata VMSA.

> would it make sense to make the kvm_x86_op generic at least?
> 
> 'reload_guest_save_area' maybe?

I think I'd prefer to keep the extra-ugly vmsa terminology?  We've already bled
gory details into common x86, and I don't see an easy way around that.  IMO, at
that point, any abstraction does more harm than good.

E.g. reload_guest_save_area() is more like to be misread as "reload the save state
from KVM's vCPU state".  Hmm, though by that argument, reload_vmsa_page() would
be even better.