KVM: VMX: Mark VCPU_EXREG_CR3 dirty when !CR0_PG -> CR0_PG if EPT + !URG

When !CR0_PG -> CR0_PG, vcpu->arch.cr3 becomes active, but GUEST_CR3 is
still vmx->ept_identity_map_addr if EPT + !URG.  So VCPU_EXREG_CR3 is
considered to be dirty and GUEST_CR3 needs to be updated in this case.

Reported-by: Maxim Levitsky <mlevitsk@redhat.com>
Suggested-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Lai Jiangshan <laijs@linux.alibaba.com>
Message-Id: <20211216021938.11752-4-jiangshanlai@gmail.com>
Fixes: c62c7bd ("KVM: VMX: Update vmcs.GUEST_CR3 only when the guest CR3 is dirty")
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

arch/x86/kvm/vmx/vmx.c

@@ -3071,6 +3071,13 @@ void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 		/* Note, vmx_set_cr4() consumes the new vcpu->arch.cr0. */
 		if ((old_cr0_pg ^ cr0) & X86_CR0_PG)
 			vmx_set_cr4(vcpu, kvm_read_cr4(vcpu));
+
+		/*
+		 * When !CR0_PG -> CR0_PG, vcpu->arch.cr3 becomes active, but
+		 * GUEST_CR3 is still vmx->ept_identity_map_addr if EPT + !URG.
+		 */
+		if (!(old_cr0_pg & X86_CR0_PG) && (cr0 & X86_CR0_PG))
+			kvm_register_mark_dirty(vcpu, VCPU_EXREG_CR3);
 	}
 
 	/* depends on vcpu->arch.cr0 to be set to a new value */