KVM: selftests: evmcs_test: Check that VMCS12 is alway properly synced to eVMCS after restore

Add a test for the regression, introduced by commit f2c7ef3ba955
("KVM: nSVM: cancel KVM_REQ_GET_NESTED_STATE_PAGES on nested vmexit"). When
L2->L1 exit is forced immediately after restoring nested state,
KVM_REQ_GET_NESTED_STATE_PAGES request is cleared and VMCS12 changes
(e.g. fresh RIP) are not reflected to eVMCS. The consequent nested
vCPU run gets broken.

Utilize NMI injection to do the job.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-Id: <20210505151823.1341678-3-vkuznets@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

1 files changed, 55 insertions, 11 deletions

diff --git a/tools/testing/selftests/kvm/x86_64/evmcs_test.c b/tools/testing/selftests/kvm/x86_64/evmcs_test.c
index b01f64ac6ce3..63096cea26c6 100644
--- a/tools/testing/selftests/kvm/x86_64/evmcs_test.c
+++ b/tools/testing/selftests/kvm/x86_64/evmcs_test.c
@@ -18,30 +18,52 @@
 #include "vmx.h"
 
 #define VCPU_ID		5
+#define NMI_VECTOR	2
 
 static int ud_count;
 
+void enable_x2apic(void)
+{
+	uint32_t spiv_reg = APIC_BASE_MSR + (APIC_SPIV >> 4);
+
+	wrmsr(MSR_IA32_APICBASE, rdmsr(MSR_IA32_APICBASE) |
+	      MSR_IA32_APICBASE_ENABLE | MSR_IA32_APICBASE_EXTD);
+	wrmsr(spiv_reg, rdmsr(spiv_reg) | APIC_SPIV_APIC_ENABLED);
+}
+
 static void guest_ud_handler(struct ex_regs *regs)
 {
 	ud_count++;
 	regs->rip += 3; /* VMLAUNCH */
 }
 
+static void guest_nmi_handler(struct ex_regs *regs)
+{
+}
+
 void l2_guest_code(void)
 {
 	GUEST_SYNC(7);
 
 	GUEST_SYNC(8);
 
+	/* Forced exit to L1 upon restore */
+	GUEST_SYNC(9);
+
 	/* Done, exit to L1 and never come back.  */
 	vmcall();
 }
 
-void l1_guest_code(struct vmx_pages *vmx_pages)
+void guest_code(struct vmx_pages *vmx_pages)
 {
 #define L2_GUEST_STACK_SIZE 64
 	unsigned long l2_guest_stack[L2_GUEST_STACK_SIZE];
 
+	enable_x2apic();
+
+	GUEST_SYNC(1);
+	GUEST_SYNC(2);
+
 	enable_vp_assist(vmx_pages->vp_assist_gpa, vmx_pages->vp_assist);
 
 	GUEST_ASSERT(vmx_pages->vmcs_gpa);
@@ -63,21 +85,22 @@ void l1_guest_code(struct vmx_pages *vmx_pages)
 	current_evmcs->revision_id = EVMCS_VERSION;
 	GUEST_SYNC(6);
 
+	current_evmcs->pin_based_vm_exec_control |=
+		PIN_BASED_NMI_EXITING;
 	GUEST_ASSERT(!vmlaunch());
 	GUEST_ASSERT(vmptrstz() == vmx_pages->enlightened_vmcs_gpa);
-	GUEST_SYNC(9);
+
+	/*
+	 * NMI forces L2->L1 exit, resuming L2 and hope that EVMCS is
+	 * up-to-date (RIP points where it should and not at the beginning
+	 * of l2_guest_code(). GUEST_SYNC(9) checkes that.
+	 */
 	GUEST_ASSERT(!vmresume());
-	GUEST_ASSERT(vmreadz(VM_EXIT_REASON) == EXIT_REASON_VMCALL);
-	GUEST_SYNC(10);
-}
 
-void guest_code(struct vmx_pages *vmx_pages)
-{
-	GUEST_SYNC(1);
-	GUEST_SYNC(2);
+	GUEST_SYNC(10);
 
-	if (vmx_pages)
-		l1_guest_code(vmx_pages);
+	GUEST_ASSERT(vmreadz(VM_EXIT_REASON) == EXIT_REASON_VMCALL);
+	GUEST_SYNC(11);
 
 	/* Try enlightened vmptrld with an incorrect GPA */
 	evmcs_vmptrld(0xdeadbeef, vmx_pages->enlightened_vmcs);
@@ -86,6 +109,18 @@ void guest_code(struct vmx_pages *vmx_pages)
 	GUEST_DONE();
 }
 
+void inject_nmi(struct kvm_vm *vm)
+{
+	struct kvm_vcpu_events events;
+
+	vcpu_events_get(vm, VCPU_ID, &events);
+
+	events.nmi.pending = 1;
+	events.flags |= KVM_VCPUEVENT_VALID_NMI_PENDING;
+
+	vcpu_events_set(vm, VCPU_ID, &events);
+}
+
 int main(int argc, char *argv[])
 {
 	vm_vaddr_t vmx_pages_gva = 0;
@@ -120,6 +155,9 @@ int main(int argc, char *argv[])
 	vm_init_descriptor_tables(vm);
 	vcpu_init_descriptor_tables(vm, VCPU_ID);
 	vm_handle_exception(vm, UD_VECTOR, guest_ud_handler);
+	vm_handle_exception(vm, NMI_VECTOR, guest_nmi_handler);
+
+	pr_info("Running L1 which uses EVMCS to run L2\n");
 
 	for (stage = 1;; stage++) {
 		_vcpu_run(vm, VCPU_ID);
@@ -166,6 +204,12 @@ int main(int argc, char *argv[])
 		TEST_ASSERT(!memcmp(&regs1, &regs2, sizeof(regs2)),
 			    "Unexpected register values after vcpu_load_state; rdi: %lx rsi: %lx",
 			    (ulong) regs2.rdi, (ulong) regs2.rsi);
+
+		/* Force immediate L2->L1 exit before resuming */
+		if (stage == 8) {
+			pr_info("Injecting NMI into L1 before L2 had a chance to run after restore\n");
+			inject_nmi(vm);
+		}
 	}
 
 done: