[PATCH v4 3/5] KVM: SEV: Check for invalidation before warning on unassigned RMP entry

[Ackerley Tng <ackerleytng@google.com>]

When handling an RMP fault, KVM looks up the RMP entry for the backing
PFN to determine if a 2MB page needs to be split via PSMASH. If the RMP
entry is not assigned (or lookup fails), KVM logs a rate-limited warning
under the assumption that private memory should always have an assigned
RMP entry.

However, a concurrent invalidation (such as guest_memfd hole punching) can
race with RMP fault handling and transition the page to shared, unassigning
the RMP entry after KVM fetched the PFN. In such cases, not finding an
assigned RMP entry is benign.

Check mmu_invalidate_retry_gfn() under mmu_lock before warning about a
missing or unassigned RMP entry, and suppress the spurious warning if an
invalidation occurred for the faulting GFN.

Fixes: c63cf135cc99 ("KVM: SEV: Add support to handle RMP nested page faults")
Signed-off-by: Ackerley Tng <ackerleytng@google.com>
---
 arch/x86/kvm/svm/sev.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kvm/svm/sev.c b/arch/x86/kvm/svm/sev.c
index 0d027cff734cf..f9dce8acac8eb 100644
--- a/arch/x86/kvm/svm/sev.c
+++ b/arch/x86/kvm/svm/sev.c
@@ -5060,8 +5060,12 @@ void sev_handle_rmp_fault(struct kvm_vcpu *vcpu, gpa_t gpa, u64 error_code)
 
 	ret = snp_lookup_rmpentry(pfn, &assigned, &rmp_level);
 	if (ret || !assigned) {
-		pr_warn_ratelimited("SEV: Unexpected RMP fault, no assigned RMP entry found for GPA 0x%llx PFN 0x%llx error %d\n",
-				    gpa, pfn, ret);
+		guard(read_lock)(&kvm->mmu_lock);
+
+		if (!mmu_invalidate_retry_gfn(kvm, mmu_seq, gfn))
+			pr_warn_ratelimited("SEV: Unexpected RMP fault, no assigned RMP entry found for GPA 0x%llx PFN 0x%llx error %d\n",
+					    gpa, pfn, ret);
+
 		return;
 	}
 

-- 
2.55.0.887.g758fc8c411-goog