[PATCH v3 19/20] KVM: x86: selftests: Add APX state and ABI test

["Chang S. Bae" <chang.seok.bae@intel.com>]

Test APX-specific behavior and ABI interactions as these are unique
comparing to other state components:

  * EGPR state is saved on VM entry assembly (unlike other components).
  * The saved state is retained even if the guest disables APX.
  * EGPR state is exposed via the XSAVE ABI while not stored in the
    kernel XSAVE buffer.

Signed-off-by: Chang S. Bae <chang.seok.bae@intel.com>
---
V2 -> V3: New patch
---
 tools/testing/selftests/kvm/Makefile.kvm      |   1 +
 .../selftests/kvm/include/x86/processor.h     | 120 +++++++++++
 tools/testing/selftests/kvm/x86/apx_test.c    | 192 ++++++++++++++++++
 3 files changed, 313 insertions(+)
 create mode 100644 tools/testing/selftests/kvm/x86/apx_test.c

diff --git a/tools/testing/selftests/kvm/Makefile.kvm b/tools/testing/selftests/kvm/Makefile.kvm
index 075eecab8a19..45b4329543f1 100644
--- a/tools/testing/selftests/kvm/Makefile.kvm
+++ b/tools/testing/selftests/kvm/Makefile.kvm
@@ -158,6 +158,7 @@ TEST_GEN_PROGS_x86 += rseq_test
 TEST_GEN_PROGS_x86 += steal_time
 TEST_GEN_PROGS_x86 += system_counter_offset_test
 TEST_GEN_PROGS_x86 += pre_fault_memory_test
+TEST_GEN_PROGS_x86 += x86/apx_test
 
 # Compiled outputs used by test targets
 TEST_GEN_PROGS_EXTENDED_x86 += x86/nx_huge_pages_test
diff --git a/tools/testing/selftests/kvm/include/x86/processor.h b/tools/testing/selftests/kvm/include/x86/processor.h
index d8634a760a60..e30e23a15b52 100644
--- a/tools/testing/selftests/kvm/include/x86/processor.h
+++ b/tools/testing/selftests/kvm/include/x86/processor.h
@@ -90,6 +90,7 @@ struct xstate {
 #define XFEATURE_MASK_LBR		BIT_ULL(15)
 #define XFEATURE_MASK_XTILE_CFG		BIT_ULL(17)
 #define XFEATURE_MASK_XTILE_DATA	BIT_ULL(18)
+#define XFEATURE_MASK_APX		BIT_ULL(19)
 
 #define XFEATURE_MASK_AVX512		(XFEATURE_MASK_OPMASK | \
 					 XFEATURE_MASK_ZMM_Hi256 | \
@@ -177,6 +178,7 @@ struct kvm_x86_cpu_feature {
 #define	X86_FEATURE_SPEC_CTRL		KVM_X86_CPU_FEATURE(0x7, 0, EDX, 26)
 #define	X86_FEATURE_ARCH_CAPABILITIES	KVM_X86_CPU_FEATURE(0x7, 0, EDX, 29)
 #define	X86_FEATURE_PKS			KVM_X86_CPU_FEATURE(0x7, 0, ECX, 31)
+#define X86_FEATURE_APX			KVM_X86_CPU_FEATURE(0x7, 1, EDX, 21)
 #define	X86_FEATURE_XTILECFG		KVM_X86_CPU_FEATURE(0xD, 0, EAX, 17)
 #define	X86_FEATURE_XTILEDATA		KVM_X86_CPU_FEATURE(0xD, 0, EAX, 18)
 #define	X86_FEATURE_XSAVES		KVM_X86_CPU_FEATURE(0xD, 1, EAX, 3)
@@ -860,6 +862,124 @@ static inline void write_sse_reg(int reg, const sse128_t *data)
 	}
 }
 
+static inline unsigned long read_egpr(int reg)
+{
+	unsigned long data = 0;
+
+	/* mov %r16..%r31, %rax */
+	switch (reg) {
+	case 16:
+		asm(".byte 0xd5, 0x48, 0x89, 0xc0" : "=a"(data));
+		break;
+	case 17:
+		asm(".byte 0xd5, 0x48, 0x89, 0xc8" : "=a"(data));
+		break;
+	case 18:
+		asm(".byte 0xd5, 0x48, 0x89, 0xd0" : "=a"(data));
+		break;
+	case 19:
+		asm(".byte 0xd5, 0x48, 0x89, 0xd8" : "=a"(data));
+		break;
+	case 20:
+		asm(".byte 0xd5, 0x48, 0x89, 0xe0" : "=a"(data));
+		break;
+	case 21:
+		asm(".byte 0xd5, 0x48, 0x89, 0xe8" : "=a"(data));
+		break;
+	case 22:
+		asm(".byte 0xd5, 0x48, 0x89, 0xf0" : "=a"(data));
+		break;
+	case 23:
+		asm(".byte 0xd5, 0x48, 0x89, 0xf8" : "=a"(data));
+		break;
+	case 24:
+		asm(".byte 0xd5, 0x4c, 0x89, 0xc0" : "=a"(data));
+		break;
+	case 25:
+		asm(".byte 0xd5, 0x4c, 0x89, 0xc8" : "=a"(data));
+		break;
+	case 26:
+		asm(".byte 0xd5, 0x4c, 0x89, 0xd0" : "=a"(data));
+		break;
+	case 27:
+		asm(".byte 0xd5, 0x4c, 0x89, 0xd8" : "=a"(data));
+		break;
+	case 28:
+		asm(".byte 0xd5, 0x4c, 0x89, 0xe0" : "=a"(data));
+		break;
+	case 29:
+		asm(".byte 0xd5, 0x4c, 0x89, 0xe8" : "=a"(data));
+		break;
+	case 30:
+		asm(".byte 0xd5, 0x4c, 0x89, 0xf0" : "=a"(data));
+		break;
+	case 31:
+		asm(".byte 0xd5, 0x4c, 0x89, 0xf8" : "=a"(data));
+		break;
+	default:
+		BUG();
+	}
+
+	return data;
+}
+
+static inline void write_egpr(int reg, unsigned long data)
+{
+	/* mov %%rax, %r16...%r31*/
+	switch (reg) {
+	case 16:
+		asm(".byte 0xd5, 0x18, 0x89, 0xc0" : : "a"(data));
+		break;
+	case 17:
+		asm(".byte 0xd5, 0x18, 0x89, 0xc1" : : "a"(data));
+		break;
+	case 18:
+		asm(".byte 0xd5, 0x18, 0x89, 0xc2" : : "a"(data));
+		break;
+	case 19:
+		asm(".byte 0xd5, 0x18, 0x89, 0xc3" : : "a"(data));
+		break;
+	case 20:
+		asm(".byte 0xd5, 0x18, 0x89, 0xc4" : : "a"(data));
+		break;
+	case 21:
+		asm(".byte 0xd5, 0x18, 0x89, 0xc5" : : "a"(data));
+		break;
+	case 22:
+		asm(".byte 0xd5, 0x18, 0x89, 0xc6" : : "a"(data));
+		break;
+	case 23:
+		asm(".byte 0xd5, 0x18, 0x89, 0xc7" : : "a"(data));
+		break;
+	case 24:
+		asm(".byte 0xd5, 0x19, 0x89, 0xc0" : : "a"(data));
+		break;
+	case 25:
+		asm(".byte 0xd5, 0x19, 0x89, 0xc1" : : "a"(data));
+		break;
+	case 26:
+		asm(".byte 0xd5, 0x19, 0x89, 0xc2" : : "a"(data));
+		break;
+	case 27:
+		asm(".byte 0xd5, 0x19, 0x89, 0xc3" : : "a"(data));
+		break;
+	case 28:
+		asm(".byte 0xd5, 0x19, 0x89, 0xc4" : : "a"(data));
+		break;
+	case 29:
+		asm(".byte 0xd5, 0x19, 0x89, 0xc5" : : "a"(data));
+		break;
+	case 30:
+		asm(".byte 0xd5, 0x19, 0x89, 0xc6" : : "a"(data));
+		break;
+	case 31:
+		asm(".byte 0xd5, 0x19, 0x89, 0xc7" : : "a"(data));
+		break;
+	default:
+		BUG();
+	}
+}
+
 static inline void cpu_relax(void)
 {
 	asm volatile("rep; nop" ::: "memory");
diff --git a/tools/testing/selftests/kvm/x86/apx_test.c b/tools/testing/selftests/kvm/x86/apx_test.c
new file mode 100644
index 000000000000..0327b32e1042
--- /dev/null
+++ b/tools/testing/selftests/kvm/x86/apx_test.c
@@ -0,0 +1,192 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include "processor.h"
+
+enum stages {
+	GUEST_UPDATE,
+	USERSPACE_UPDATE,
+	GUEST_APXOFF,
+};
+
+enum egpr_ops {
+	EGPRS_WRITE,
+	EGPRS_CHECK,
+};
+
+#define for_each_egpr(reg)	for (reg = 16; reg <= 31; reg++)
+
+/*
+ * Deterministic per-stage test values for EGPRs so that guest and
+ * userspace can validate state transitions.
+ */
+static inline unsigned long egpr_data(enum stages stage, int reg)
+{
+	switch (stage) {
+	case GUEST_UPDATE:
+		return 0xabcd + reg;
+	case USERSPACE_UPDATE:
+		return 0xbcde + reg;
+	case GUEST_APXOFF:
+		return 0xcdef + reg;
+	default:
+		return 0;
+	}
+}
+
+/*
+ * Read/write or validate EGPR values either directly via registers
+ * (guest context) or via a provided buffer (userspace XSAVE).
+ */
+static bool handle_egprs(enum egpr_ops ops, unsigned long *egprs, enum stages stage)
+{
+	unsigned long data;
+	int reg;
+
+	for_each_egpr(reg) {
+		data = egpr_data(stage, reg);
+
+		if (ops == EGPRS_WRITE) {
+			if (egprs)
+				egprs[reg - 16] = data;
+			else
+				write_egpr(reg, data);
+			continue;
+		}
+
+		if (ops != EGPRS_CHECK)
+			return false;
+
+		if (egprs) {
+			if (egprs[reg - 16] != data)
+				return false;
+			continue;
+		}
+
+		if (read_egpr(reg) != data)
+			return false;
+	}
+
+	return true;
+}
+
+static void write_egprs(enum stages stage)
+{
+	handle_egprs(EGPRS_WRITE, NULL, stage);
+}
+
+static bool validate_egprs(enum stages stage)
+{
+	return handle_egprs(EGPRS_CHECK, NULL, stage);
+}
+
+static void test_guest_update(void)
+{
+	write_egprs(GUEST_UPDATE);
+	GUEST_SYNC(GUEST_UPDATE);
+	GUEST_ASSERT(validate_egprs(GUEST_UPDATE));
+}
+
+static void test_userspace_update(void)
+{
+	/* Userspace updates EGPR state via the KVM XSAVE ABI */
+	GUEST_SYNC(USERSPACE_UPDATE);
+	GUEST_ASSERT(validate_egprs(USERSPACE_UPDATE));
+}
+
+static void test_guest_apxoff(void)
+{
+	write_egprs(GUEST_APXOFF);
+	/* Disable APX to verify state is preserved */
+	GUEST_ASSERT(!xsetbv_safe(0, this_cpu_supported_xcr0() & ~XFEATURE_MASK_APX));
+	GUEST_SYNC(GUEST_APXOFF);
+	GUEST_ASSERT(!xsetbv_safe(0, this_cpu_supported_xcr0()));
+	GUEST_ASSERT(validate_egprs(GUEST_APXOFF));
+}
+
+static void guest_code(void)
+{
+	set_cr4(get_cr4() | X86_CR4_OSXSAVE);
+	GUEST_ASSERT(!xsetbv_safe(0, this_cpu_supported_xcr0()));
+
+	test_guest_update();
+	test_userspace_update();
+	test_guest_apxoff();
+
+	GUEST_DONE();
+}
+
+
+#define X86_PROPERTY_XSTATE_APX_OFFSET	KVM_X86_CPU_PROPERTY(0xd, 19, EBX, 0, 31)
+#define XSAVE_HDR_OFFSET		512
+
+static inline unsigned long *xsave_egprs(void *xsave)
+{
+	return xsave + kvm_cpu_property(X86_PROPERTY_XSTATE_APX_OFFSET);
+}
+
+static inline void xstatebv_set(void *xsave, uint64_t mask)
+{
+	*(uint64_t *)(xsave + XSAVE_HDR_OFFSET) |= mask;
+}
+
+static void write_xsave_egprs(void *xsave, enum stages stage)
+{
+	handle_egprs(EGPRS_WRITE, xsave_egprs(xsave), stage);
+	xstatebv_set(xsave, XFEATURE_MASK_APX);
+}
+
+static bool validate_xsave_egprs(void *xsave, enum stages stage)
+{
+	return handle_egprs(EGPRS_CHECK, xsave_egprs(xsave), stage);
+}
+
+int main(int argc, char *argv[])
+{
+	struct kvm_xsave *xsave;
+	struct kvm_vcpu *vcpu;
+	struct kvm_vm *vm;
+	enum stages stage;
+	struct ucall uc;
+	int xsave_size;
+
+	TEST_REQUIRE(kvm_cpu_has(X86_FEATURE_XSAVE));
+	TEST_REQUIRE(kvm_cpu_has(X86_FEATURE_APX));
+
+	vm = vm_create_with_one_vcpu(&vcpu, guest_code);
+	xsave_size = vm_check_cap(vcpu->vm, KVM_CAP_XSAVE2);
+	TEST_ASSERT(xsave_size, "KVM_CAP_XSAVE2 not supported");
+	xsave = malloc(xsave_size);
+	TEST_ASSERT(xsave, "Failed to allocate XSAVE buffer");
+
+	while (1) {
+		vcpu_run(vcpu);
+		TEST_ASSERT_KVM_EXIT_REASON(vcpu, KVM_EXIT_IO);
+
+		switch (get_ucall(vcpu, &uc)) {
+		case UCALL_ABORT:
+			REPORT_GUEST_ASSERT(uc);
+			break;
+		case UCALL_SYNC: {
+			stage = uc.args[1];
+			vcpu_xsave_get(vcpu, xsave);
+			if (stage == USERSPACE_UPDATE) {
+				write_xsave_egprs(xsave, stage);
+			} else {
+				TEST_ASSERT(validate_xsave_egprs(xsave, stage),
+					    "EGPR state mismatch in userspace XSAVE buffer");
+			}
+			vcpu_xsave_set(vcpu, xsave);
+			break;
+		}
+		case UCALL_DONE:
+			goto done;
+		default:
+			TEST_FAIL("Unknown ucall %lu", uc.cmd);
+		}
+	}
+
+done:
+	free(xsave);
+	kvm_vm_free(vm);
+	return 0;
+}
-- 
2.51.0