From: Lai Jiangshan <laijs@xxxxxxxxxxxxxxxxx>
In design, the guest virtuall address is only make sense for
vcpu->arch.walk_mmu which is often the same as vcpu->arch.mmu. But
they are different semantic by design, so vcpu->arch.walk_mmu should be
used instead like other call site of kvm_mmu_invalidate_gva().
In theory, if L2's invlpg is being emulated by L0 (in practice, it
hardly happen) when nTDP is used, ->tlb_flush_gva() should be called
to flush hwTLB, but using vcpu->arch.mmu causes it incorrectly skept.
Signed-off-by: Lai Jiangshan <laijs@xxxxxxxxxxxxxxxxx>
---
arch/x86/kvm/mmu/mmu.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index f3aa91db4a7e..72ce0d78435e 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -5353,7 +5353,7 @@ void kvm_mmu_invalidate_gva(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu,
void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
{
- kvm_mmu_invalidate_gva(vcpu, vcpu->arch.mmu, gva, INVALID_PAGE);
+ kvm_mmu_invalidate_gva(vcpu, vcpu->arch.walk_mmu, gva, INVALID_PAGE);
++vcpu->stat.invlpg;
}
EXPORT_SYMBOL_GPL(kvm_mmu_invlpg);
--
2.19.1.6.gb485710b
