* [PATCH 2/2] tests: kvm: Test virtualization of VMX capability MSRs
@ 2019-05-31 18:42 Aaron Lewis
2019-06-18 14:14 ` Aaron Lewis
0 siblings, 1 reply; 3+ messages in thread
From: Aaron Lewis @ 2019-05-31 18:42 UTC (permalink / raw)
To: jmattson, pshier, marcorr, kvm; +Cc: Aaron Lewis
This tests three main things:
1. Test that the VMX capability MSRs can be written to and read from
based on the rules in the SDM.
2. Test that the MSR_IA32_VMX_CR4_FIXED1 is correctly updated in
response to CPUID changes.
3. Test that the VMX preemption timer rate can be cleared when the
preemption timer is disabled.
Signed-off-by: Aaron Lewis <aaronlewis@google.com>
Reviewed-by: Peter Shier <pshier@google.com>
---
tools/testing/selftests/kvm/.gitignore | 1 +
tools/testing/selftests/kvm/Makefile | 1 +
.../testing/selftests/kvm/include/kvm_util.h | 2 +
tools/testing/selftests/kvm/lib/kvm_util.c | 5 +
.../kvm/x86_64/vmx_capability_msr_test.c | 567 ++++++++++++++++++
5 files changed, 576 insertions(+)
create mode 100644 tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c
diff --git a/tools/testing/selftests/kvm/.gitignore b/tools/testing/selftests/kvm/.gitignore
index 2689d1ea6d7a..94ca75f9f4cd 100644
--- a/tools/testing/selftests/kvm/.gitignore
+++ b/tools/testing/selftests/kvm/.gitignore
@@ -3,6 +3,7 @@
/x86_64/platform_info_test
/x86_64/set_sregs_test
/x86_64/sync_regs_test
+/x86_64/vmx_capability_msr_test
/x86_64/vmx_close_while_nested_test
/x86_64/vmx_tsc_adjust_test
/x86_64/state_test
diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile
index f8588cca2bef..a635a2c42592 100644
--- a/tools/testing/selftests/kvm/Makefile
+++ b/tools/testing/selftests/kvm/Makefile
@@ -18,6 +18,7 @@ TEST_GEN_PROGS_x86_64 += x86_64/cr4_cpuid_sync_test
TEST_GEN_PROGS_x86_64 += x86_64/state_test
TEST_GEN_PROGS_x86_64 += x86_64/evmcs_test
TEST_GEN_PROGS_x86_64 += x86_64/hyperv_cpuid
+TEST_GEN_PROGS_x86_64 += x86_64/vmx_capability_msr_test
TEST_GEN_PROGS_x86_64 += x86_64/vmx_close_while_nested_test
TEST_GEN_PROGS_x86_64 += x86_64/smm_test
TEST_GEN_PROGS_x86_64 += dirty_log_test
diff --git a/tools/testing/selftests/kvm/include/kvm_util.h b/tools/testing/selftests/kvm/include/kvm_util.h
index 07b71ad9734a..6a2e27cad877 100644
--- a/tools/testing/selftests/kvm/include/kvm_util.h
+++ b/tools/testing/selftests/kvm/include/kvm_util.h
@@ -133,6 +133,8 @@ struct kvm_vm *vm_create_default(uint32_t vcpuid, uint64_t extra_mem_size,
void *guest_code);
void vm_vcpu_add_default(struct kvm_vm *vm, uint32_t vcpuid, void *guest_code);
+int vcpu_fd(struct kvm_vm *vm, uint32_t vcpuid);
+
struct kvm_userspace_memory_region *
kvm_userspace_memory_region_find(struct kvm_vm *vm, uint64_t start,
uint64_t end);
diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
index 4ca96b228e46..3bc4d2f633c8 100644
--- a/tools/testing/selftests/kvm/lib/kvm_util.c
+++ b/tools/testing/selftests/kvm/lib/kvm_util.c
@@ -402,6 +402,11 @@ static void vm_vcpu_rm(struct kvm_vm *vm, uint32_t vcpuid)
free(vcpu);
}
+int vcpu_fd(struct kvm_vm *vm, uint32_t vcpuid)
+{
+ return vcpu_find(vm, vcpuid)->fd;
+}
+
void kvm_vm_release(struct kvm_vm *vmp)
{
int ret;
diff --git a/tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c b/tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c
new file mode 100644
index 000000000000..161a037299f0
--- /dev/null
+++ b/tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c
@@ -0,0 +1,567 @@
+/*
+ * vmx_capability_msr_test
+ *
+ * Copyright (C) 2019, Google LLC.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.
+ *
+ * Test virtualization of VMX capability MSRs.
+ */
+
+#define _GNU_SOURCE /* for program_invocation_short_name */
+#include <fcntl.h>
+#include <linux/bits.h>
+#include <linux/kernel.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/ioctl.h>
+#include <sys/stat.h>
+#include <test_util.h>
+
+#include "kvm_util.h"
+#include "processor.h"
+#include "vmx.h"
+
+#define VCPU_ID 0
+
+struct kvm_single_msr {
+ struct kvm_msrs header;
+ struct kvm_msr_entry entry;
+} __packed;
+
+struct vmx_msr {
+ uint32_t index;
+ const char *name;
+};
+#define VMX_MSR(_msr) { _msr, #_msr }
+
+struct vmx_msr vmx_msrs[] = {
+ VMX_MSR(MSR_IA32_VMX_BASIC),
+ VMX_MSR(MSR_IA32_VMX_TRUE_PINBASED_CTLS),
+ VMX_MSR(MSR_IA32_VMX_TRUE_PROCBASED_CTLS),
+ VMX_MSR(MSR_IA32_VMX_PROCBASED_CTLS2),
+ VMX_MSR(MSR_IA32_VMX_TRUE_EXIT_CTLS),
+ VMX_MSR(MSR_IA32_VMX_TRUE_ENTRY_CTLS),
+ VMX_MSR(MSR_IA32_VMX_MISC),
+ VMX_MSR(MSR_IA32_VMX_EPT_VPID_CAP),
+ VMX_MSR(MSR_IA32_VMX_CR0_FIXED0),
+ VMX_MSR(MSR_IA32_VMX_CR0_FIXED1),
+ VMX_MSR(MSR_IA32_VMX_CR4_FIXED0),
+ VMX_MSR(MSR_IA32_VMX_CR4_FIXED1),
+ VMX_MSR(MSR_IA32_VMX_VMCS_ENUM),
+};
+
+struct kvm_vm *vm;
+
+void guest_code(void)
+{
+ __asm__ __volatile__(
+ "1:"
+ "rdmsr;"
+ "outw $0;"
+ "jmp 1b;"
+ );
+}
+
+uint64_t read_msr_from_guest(uint32_t msr)
+{
+ unsigned int exit_reason, expected_exit = KVM_EXIT_IO;
+ struct kvm_regs regs;
+
+ vcpu_regs_get(vm, VCPU_ID, ®s);
+ regs.rcx = msr;
+ regs.rax = 0;
+ regs.rdx = 0;
+ vcpu_regs_set(vm, VCPU_ID, ®s);
+
+ vcpu_run(vm, VCPU_ID);
+ vcpu_regs_get(vm, VCPU_ID, ®s);
+
+ exit_reason = vcpu_state(vm, VCPU_ID)->exit_reason;
+ TEST_ASSERT(exit_reason == expected_exit,
+ "Unexpected exit reason: expected %u (%s), got %u (%s) [guest rip: %llx]",
+ expected_exit, exit_reason_str(expected_exit),
+ exit_reason, exit_reason_str(exit_reason),
+ regs.rip);
+
+ return (regs.rdx << 32) | regs.rax;
+}
+
+int try_kvm_get_msr(int vcpu_fd, uint32_t msr, uint64_t *data)
+{
+ struct kvm_single_msr buffer;
+ int rv;
+
+ memset(&buffer, 0, sizeof(buffer));
+ buffer.header.nmsrs = 1;
+ buffer.entry.index = msr;
+
+ /* Returns the number of MSRs read. */
+ rv = ioctl(vcpu_fd, KVM_GET_MSRS, &buffer);
+ *data = buffer.entry.data;
+ return rv;
+}
+
+uint64_t kvm_get_msr(int vcpu_fd, uint32_t msr)
+{
+ uint64_t data;
+ int rv;
+
+ rv = try_kvm_get_msr(vcpu_fd, msr, &data);
+ TEST_ASSERT(rv == 1, "KVM_GET_MSR of MSR %x failed. (rv %d)",
+ msr, rv);
+
+ return data;
+}
+
+void kvm_get_msr_expect_fail(int vcpu_fd, uint32_t msr)
+{
+ uint64_t data;
+ int rv;
+
+ rv = try_kvm_get_msr(vcpu_fd, msr, &data);
+ TEST_ASSERT(rv != 1, "Expected KVM_GET_MSR of MSR %x to fail.", msr);
+}
+
+int try_kvm_set_msr(int vcpu_fd, uint32_t msr, uint64_t data)
+{
+ struct kvm_single_msr buffer;
+ int rv;
+
+ memset(&buffer, 0, sizeof(buffer));
+ buffer.header.nmsrs = 1;
+ buffer.entry.index = msr;
+ buffer.entry.data = data;
+
+ /* Returns the number of MSRs written. */
+ rv = ioctl(vcpu_fd, KVM_SET_MSRS, &buffer);
+ return rv;
+}
+
+void kvm_set_msr_expect_fail(int vcpu_fd, uint32_t msr, uint64_t data)
+{
+ int rv;
+
+ rv = try_kvm_set_msr(vcpu_fd, msr, data);
+ TEST_ASSERT(rv != 1, "Expected KVM_SET_MSR of MSR %x to fail.", msr);
+}
+
+void kvm_set_msr(int vcpu_fd, uint32_t msr, uint64_t data)
+{
+ int rv;
+
+ rv = try_kvm_set_msr(vcpu_fd, msr, data);
+ TEST_ASSERT(rv == 1, "KVM_SET_MSR %x (data %lx) failed.", msr, data);
+}
+
+void test_set_vmx_msr(int vcpu_fd, uint32_t msr, uint64_t data)
+{
+ uint64_t guest, kvm;
+ uint64_t non_true;
+
+ kvm_set_msr(vcpu_fd, msr, data);
+
+ guest = read_msr_from_guest(msr);
+ kvm = kvm_get_msr(vcpu_fd, msr);
+
+ TEST_ASSERT(guest == data,
+ "KVM_SET_MSR %x not visible from guest (guest %lx, expected %lx)",
+ msr, guest, data);
+
+ TEST_ASSERT(kvm == data,
+ "KVM_SET_MSR %x not visible from host (host %lx, expected %lx)",
+ msr, kvm, data);
+
+#define IA32_VMX_PINBASED_CTLS_DEFAULT1 (BIT_ULL(1) | BIT_ULL(2) | BIT_ULL(4))
+#define IA32_VMX_PROCBASED_CTLS_DEFAULT1 (BIT_ULL(1) | BIT_ULL(4) | \
+ BIT_ULL(5) | BIT_ULL(6) | \
+ BIT_ULL(8) | BIT_ULL(13) | \
+ BIT_ULL(14) | BIT_ULL(15) | \
+ BIT_ULL(16) | BIT_ULL(26))
+#define IA32_VMX_EXIT_CTLS_DEFAULT1 (BIT_ULL(0) | BIT_ULL(1) | BIT_ULL(2) | \
+ BIT_ULL(3) | BIT_ULL(4) | BIT_ULL(5) | \
+ BIT_ULL(6) | BIT_ULL(7) | BIT_ULL(8) | \
+ BIT_ULL(10) | BIT_ULL(11) | BIT_ULL(13) | \
+ BIT_ULL(14) | BIT_ULL(16) | BIT_ULL(17))
+#define IA32_VMX_ENTRY_CTLS_DEFAULT1 (BIT_ULL(0) | BIT_ULL(1) | BIT_ULL(2) | \
+ BIT_ULL(3) | BIT_ULL(4) | BIT_ULL(5) | \
+ BIT_ULL(6) | BIT_ULL(7) | BIT_ULL(8) | \
+ BIT_ULL(12))
+ /*
+ * Make sure non-true MSRs are equal to the true MSRs with all the
+ * default1 bits set.
+ */
+ switch (msr) {
+ case MSR_IA32_VMX_TRUE_PINBASED_CTLS:
+ non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_PINBASED_CTLS);
+ TEST_ASSERT(non_true ==
+ (data | IA32_VMX_PINBASED_CTLS_DEFAULT1),
+ "Incorrect IA32_VMX_PINBASED_CTLS MSR: %lx\n",
+ non_true);
+ break;
+ case MSR_IA32_VMX_TRUE_PROCBASED_CTLS:
+ non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_PROCBASED_CTLS);
+ TEST_ASSERT(non_true ==
+ (data | IA32_VMX_PROCBASED_CTLS_DEFAULT1),
+ "Incorrect IA32_VMX_PROCBASED_CTLS MSR: %lx\n",
+ non_true);
+ break;
+ case MSR_IA32_VMX_TRUE_EXIT_CTLS:
+ non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_EXIT_CTLS);
+ TEST_ASSERT(non_true == (data | IA32_VMX_EXIT_CTLS_DEFAULT1),
+ "Incorrect IA32_VMX_EXIT_CTLS MSR: %lx\n",
+ non_true);
+ break;
+ case MSR_IA32_VMX_TRUE_ENTRY_CTLS:
+ non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_ENTRY_CTLS);
+ TEST_ASSERT(non_true == (data | IA32_VMX_ENTRY_CTLS_DEFAULT1),
+ "Incorrect IA32_VMX_ENTRY_CTLS MSR: %lx\n",
+ non_true);
+ break;
+ }
+}
+
+/*
+ * Reset the VCPU and return its file descriptor.
+ *
+ * The VCPU should be reset in between subsequent restores of the
+ * same VMX MSR.
+ */
+int reset_vcpu_fd(bool vmx_supported)
+{
+ struct kvm_cpuid2 *cpuid;
+ int fd;
+
+ if (vm)
+ kvm_vm_free(vm);
+
+ vm = vm_create_default(VCPU_ID, 0, guest_code);
+ fd = vcpu_fd(vm, VCPU_ID);
+
+ /* Enable/Disable VMX */
+ if (vmx_supported)
+ kvm_get_supported_cpuid_entry(0x1)->ecx |= (1 << 5);
+ else
+ kvm_get_supported_cpuid_entry(0x1)->ecx &= ~(1 << 5);
+ cpuid = kvm_get_supported_cpuid();
+ vcpu_set_cpuid(vm, VCPU_ID, cpuid);
+
+ return fd;
+}
+
+void test_msr_bits(uint32_t index, int high, int low, bool allowed_1)
+{
+ int fd = reset_vcpu_fd(true);
+ uint64_t value = kvm_get_msr(fd, index);
+ int i;
+
+ for (i = low; i <= high; i++) {
+ uint64_t mask = 1ull << i;
+ bool is_1 = value & mask;
+
+ fd = reset_vcpu_fd(true);
+
+ /*
+ * Following VMX nomenclature:
+ *
+ * allowed-1 means a bit is high when it can be 0 or 1. The
+ * bit is low when it must be 0.
+ *
+ * allowed-0 (!allowed_1) means a bit is low when it can be
+ * 0 or 1. The bit is high when it must be 1.
+ */
+ if (allowed_1 && is_1)
+ test_set_vmx_msr(fd, index, value & ~mask);
+ else if (!allowed_1 && !is_1)
+ test_set_vmx_msr(fd, index, value | mask);
+ else if (allowed_1 && !is_1)
+ kvm_set_msr_expect_fail(fd, index, value | mask);
+ else
+ kvm_set_msr_expect_fail(fd, index, value & ~mask);
+ }
+}
+
+void test_allowed_1_bits(uint32_t index, int high, int low)
+{
+ test_msr_bits(index, high, low, true);
+}
+#define test_allowed_1_bit(_index, _bit) \
+ test_allowed_1_bits(_index, _bit, _bit)
+
+void test_allowed_0_bits(uint32_t index, int high, int low)
+{
+ test_msr_bits(index, high, low, false);
+}
+
+uint64_t get_bits(uint64_t value, int high, int low)
+{
+ return (value & GENMASK_ULL(high, low)) >> low;
+}
+
+uint64_t set_bits(uint64_t value, int high, int low, uint64_t new_value)
+{
+ uint64_t mask = GENMASK_ULL(high, low);
+
+ return (value & ~mask) | ((new_value << low) & mask);
+}
+
+/*
+ * Test MSR[high:low].
+ *
+ * lt : KVM should support values less than the default.
+ * gt : KVM should support values greater than the default.
+ */
+void __test_msr_bits_eq(uint32_t index, int high, int low, bool lt, bool gt)
+{
+ int fd = reset_vcpu_fd(true);
+ uint64_t value = kvm_get_msr(fd, index);
+ uint64_t bits = get_bits(value, high, low);
+ uint64_t new_value;
+
+ if (bits > 0) {
+ fd = reset_vcpu_fd(true);
+ new_value = set_bits(value, high, low, bits - 1);
+ if (lt)
+ test_set_vmx_msr(fd, index, new_value);
+ else
+ kvm_set_msr_expect_fail(fd, index, new_value);
+ }
+
+ if (bits < get_bits(-1ull, high, low)) {
+ fd = reset_vcpu_fd(true);
+ new_value = set_bits(value, high, low, bits + 1);
+ if (gt)
+ test_set_vmx_msr(fd, index, new_value);
+ else
+ kvm_set_msr_expect_fail(fd, index, new_value);
+ }
+}
+
+#define test_msr_bits_eq(...) __test_msr_bits_eq(__VA_ARGS__, false, false)
+#define test_msr_bits_le(...) __test_msr_bits_eq(__VA_ARGS__, true, false)
+
+void test_rw_vmx_capability_msrs(void)
+{
+ int i;
+
+ printf("VMX MSR default values:\n");
+
+ for (i = 0; i < ARRAY_SIZE(vmx_msrs); i++) {
+ uint32_t msr = vmx_msrs[i].index;
+ uint64_t guest, kvm;
+ int vcpu_fd;
+
+ /*
+ * If VMX is not supported, test that reading the VMX MSRs
+ * fails.
+ */
+ vcpu_fd = reset_vcpu_fd(false);
+ kvm_get_msr_expect_fail(vcpu_fd, msr);
+
+ vcpu_fd = reset_vcpu_fd(true);
+ guest = read_msr_from_guest(msr);
+ kvm = kvm_get_msr(vcpu_fd, msr);
+
+ TEST_ASSERT(guest == kvm,
+ "MSR %x mismatch: guest %lx kvm %lx.",
+ msr, guest, kvm);
+
+ printf("{ %s, 0x%lx },\n", vmx_msrs[i].name, guest);
+
+ /* VM Execution Control Capability MSRs */
+ switch (msr) {
+ case MSR_IA32_VMX_BASIC:
+ test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
+ test_msr_bits_eq(msr, 30, 0); /* vmcs revision */
+ test_allowed_1_bit(msr, 31); /* reserved */
+ test_allowed_1_bit(msr, 48); /* VMCS phys bits */
+ test_allowed_1_bit(msr, 49); /* dual-monitor SMI/SMM */
+ test_allowed_1_bit(msr, 54); /* ins/outs info */
+ test_allowed_1_bit(msr, 55); /* true ctls */
+ break;
+ case MSR_IA32_VMX_MISC:
+ test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
+ test_msr_bits_eq(msr, 4, 0); /* preempt timer rate */
+ test_allowed_1_bit(msr, 5); /* store IA-32e mode */
+ test_allowed_1_bits(msr, 8, 6); /* activity states */
+ test_allowed_1_bits(msr, 13, 9); /* reserved */
+ test_allowed_1_bit(msr, 14); /* pt in vmx mode */
+ test_allowed_1_bit(msr, 15); /* smbase msr in smm */
+ test_msr_bits_le(msr, 24, 16); /* CR3-target values */
+ test_msr_bits_le(msr, 27, 25); /* MSR-store list size */
+ test_allowed_1_bit(msr, 28); /* smm_monitor_ctl 1 */
+ test_allowed_1_bit(msr, 29); /* vmwrite to any */
+ test_allowed_1_bit(msr, 30); /* insn len 0 allowed */
+ test_allowed_1_bit(msr, 31); /* reserved */
+ test_msr_bits_eq(msr, 63, 32); /* MSEG Revision */
+ break;
+ case MSR_IA32_VMX_TRUE_PINBASED_CTLS:
+ case MSR_IA32_VMX_TRUE_PROCBASED_CTLS:
+ case MSR_IA32_VMX_TRUE_EXIT_CTLS:
+ case MSR_IA32_VMX_TRUE_ENTRY_CTLS:
+ case MSR_IA32_VMX_PROCBASED_CTLS2:
+ test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
+ test_allowed_0_bits(msr, 31, 0); /* allowed-1 bits */
+ test_allowed_1_bits(msr, 63, 32); /* allowed-0 bits */
+ break;
+ case MSR_IA32_VMX_EPT_VPID_CAP:
+ test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
+ test_allowed_1_bits(msr, 63, 0); /* feature/reserved */
+ break;
+ case MSR_IA32_VMX_CR0_FIXED0:
+ case MSR_IA32_VMX_CR4_FIXED0:
+ test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
+ test_allowed_0_bits(msr, 63, 0); /* allowed-0 bits */
+ break;
+ case MSR_IA32_VMX_CR0_FIXED1:
+ case MSR_IA32_VMX_CR4_FIXED1:
+ /* These MSRs do not allow save-restore. */
+ kvm_set_msr_expect_fail(vcpu_fd, msr, guest);
+ break;
+ case MSR_IA32_VMX_PROCBASED_CTLS:
+ case MSR_IA32_VMX_PINBASED_CTLS:
+ case MSR_IA32_VMX_EXIT_CTLS:
+ case MSR_IA32_VMX_ENTRY_CTLS:
+ test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
+ break;
+ case MSR_IA32_VMX_VMCS_ENUM:
+ test_allowed_1_bit(msr, 0); /* reserved */
+ test_allowed_1_bits(msr, 63, 10); /* reserved */
+ break;
+ default:
+ TEST_ASSERT(false, "Untested MSR: 0x%x.", msr);
+ }
+ }
+}
+
+struct msr {
+ uint32_t index;
+ uint64_t data;
+};
+
+/*
+ * Test that MSR_IA32_VMX_CR4_FIXED1 is correctly updated in response to
+ * CPUID changes.
+ */
+void test_cr4_fixed1_cpuid(void)
+{
+ struct test {
+ uint64_t cr4_mask;
+ int function;
+ enum x86_register reg;
+ int cpuid_bit;
+ const char *name;
+ } tests[] = {
+ { X86_CR4_VME, 1, RDX, 1, "VME" },
+ { X86_CR4_PVI, 1, RDX, 1, "PVI" },
+ { X86_CR4_TSD, 1, RDX, 4, "TSD" },
+ { X86_CR4_DE, 1, RDX, 2, "DE" },
+ { X86_CR4_PSE, 1, RDX, 3, "PSE" },
+ { X86_CR4_PAE, 1, RDX, 6, "PAE" },
+ { X86_CR4_MCE, 1, RDX, 7, "MCE" },
+ { X86_CR4_PGE, 1, RDX, 13, "PGE" },
+ { X86_CR4_OSFXSR, 1, RDX, 24, "OSFXSR" },
+ { X86_CR4_OSXMMEXCPT, 1, RDX, 25, "OSXMMEXCPT" },
+ { X86_CR4_UMIP, 7, RCX, 2, "UMIP" },
+ { X86_CR4_VMXE, 1, RCX, 5, "VMXE" },
+ { X86_CR4_SMXE, 1, RCX, 6, "SMXE" },
+ { X86_CR4_FSGSBASE, 7, RBX, 0, "FSGSBASE" },
+ { X86_CR4_PCIDE, 1, RCX, 17, "PCIDE" },
+ { X86_CR4_OSXSAVE, 1, RCX, 26, "OSXSAVE" },
+ { X86_CR4_SMEP, 7, RBX, 7, "SMEP" },
+ { X86_CR4_SMAP, 7, RBX, 20, "SMAP" },
+ { X86_CR4_PKE, 7, RCX, 3, "PKE" },
+ };
+ int vcpu_fd = reset_vcpu_fd(true);
+ int i;
+
+ kvm_get_supported_cpuid_entry(0x1)->ecx |= (1 << 5);
+
+ for (i = 0; i < ARRAY_SIZE(tests); i++) {
+ struct test *test = &tests[i];
+ struct kvm_cpuid_entry2 *entry;
+ struct kvm_cpuid2 *cpuid = kvm_get_supported_cpuid();
+ const char *reg_name = NULL;
+ uint64_t fixed1;
+ uint32_t *reg = NULL;
+
+ entry = kvm_get_supported_cpuid_entry(test->function);
+ TEST_ASSERT(entry, "CPUID lookup failed: CPUID.%02xH",
+ test->function);
+
+ switch (test->reg) {
+ case RAX:
+ reg = &entry->eax; reg_name = "EAX";
+ break;
+ case RBX:
+ reg = &entry->ebx; reg_name = "EBX";
+ break;
+ case RCX:
+ reg = &entry->ecx; reg_name = "ECX";
+ break;
+ case RDX:
+ reg = &entry->edx; reg_name = "EDX";
+ break;
+ default:
+ TEST_ASSERT(false, "Unrecognized reg: %d", test->reg);
+ }
+
+ printf("MSR_IA32_VMX_CR4_FIXED1[%d] (%s) == CPUID.%02xH:%s[%d]\n",
+ __builtin_ffsll(test->cr4_mask) - 1, test->name,
+ test->function, reg_name, test->cpuid_bit);
+
+
+ *reg |= 1ull << test->cpuid_bit;
+ vcpu_set_cpuid(vm, VCPU_ID, cpuid);
+ fixed1 = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_CR4_FIXED1);
+ TEST_ASSERT(fixed1 & test->cr4_mask, "set bit");
+ TEST_ASSERT(fixed1 & X86_CR4_PCE, "PCE");
+
+ /*
+ * Skipping clearing support for VMX. VMX is required for
+ * MSR_IA32_VMX_CR4_FIXED1 to even exist.
+ */
+ if (test->cr4_mask == X86_CR4_VMXE)
+ continue;
+
+ *reg &= ~(1ull << test->cpuid_bit);
+ vcpu_set_cpuid(vm, VCPU_ID, cpuid);
+ fixed1 = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_CR4_FIXED1);
+ TEST_ASSERT(!(fixed1 & test->cr4_mask), "clear bit");
+ TEST_ASSERT(fixed1 & X86_CR4_PCE, "PCE");
+ };
+}
+
+void test_preemption_timer_capability(void)
+{
+ int fd = reset_vcpu_fd(true);
+ uint64_t vmx_misc = kvm_get_msr(fd, MSR_IA32_VMX_MISC);
+ uint64_t pinbased = kvm_get_msr(fd, MSR_IA32_VMX_TRUE_PINBASED_CTLS);
+
+#define VMX_PREEMPTION_TIMER_CONTROL \
+ ((uint64_t) PIN_BASED_VMX_PREEMPTION_TIMER << 32)
+#define VMX_PREEMPTION_TIMER_RATE 0x1f
+
+ TEST_ASSERT(pinbased & VMX_PREEMPTION_TIMER_CONTROL,
+ "Expected support for VMX preemption timer.");
+
+ /* Test that we can disable the VMX preemption timer. */
+ test_set_vmx_msr(fd, MSR_IA32_VMX_TRUE_PINBASED_CTLS,
+ pinbased & ~VMX_PREEMPTION_TIMER_CONTROL);
+
+ /*
+ * Test that we can clear the VMX preemption timer rate when
+ * the preemption timer is disabled.
+ */
+ test_set_vmx_msr(fd, MSR_IA32_VMX_MISC,
+ vmx_misc & ~VMX_PREEMPTION_TIMER_RATE);
+}
+
+int main(int argc, char *argv[])
+{
+ test_rw_vmx_capability_msrs();
+ test_cr4_fixed1_cpuid();
+ test_preemption_timer_capability();
+ return 0;
+}
--
2.22.0.rc1.311.g5d7573a151-goog
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH 2/2] tests: kvm: Test virtualization of VMX capability MSRs
2019-05-31 18:42 [PATCH 2/2] tests: kvm: Test virtualization of VMX capability MSRs Aaron Lewis
@ 2019-06-18 14:14 ` Aaron Lewis
2019-07-10 16:09 ` Aaron Lewis
0 siblings, 1 reply; 3+ messages in thread
From: Aaron Lewis @ 2019-06-18 14:14 UTC (permalink / raw)
To: Jim Mattson, Peter Shier, Marc Orr, kvm
On Fri, May 31, 2019 at 11:42 AM Aaron Lewis <aaronlewis@google.com> wrote:
>
> This tests three main things:
> 1. Test that the VMX capability MSRs can be written to and read from
> based on the rules in the SDM.
> 2. Test that the MSR_IA32_VMX_CR4_FIXED1 is correctly updated in
> response to CPUID changes.
> 3. Test that the VMX preemption timer rate can be cleared when the
> preemption timer is disabled.
>
> Signed-off-by: Aaron Lewis <aaronlewis@google.com>
> Reviewed-by: Peter Shier <pshier@google.com>
> ---
> tools/testing/selftests/kvm/.gitignore | 1 +
> tools/testing/selftests/kvm/Makefile | 1 +
> .../testing/selftests/kvm/include/kvm_util.h | 2 +
> tools/testing/selftests/kvm/lib/kvm_util.c | 5 +
> .../kvm/x86_64/vmx_capability_msr_test.c | 567 ++++++++++++++++++
> 5 files changed, 576 insertions(+)
> create mode 100644 tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c
>
> diff --git a/tools/testing/selftests/kvm/.gitignore b/tools/testing/selftests/kvm/.gitignore
> index 2689d1ea6d7a..94ca75f9f4cd 100644
> --- a/tools/testing/selftests/kvm/.gitignore
> +++ b/tools/testing/selftests/kvm/.gitignore
> @@ -3,6 +3,7 @@
> /x86_64/platform_info_test
> /x86_64/set_sregs_test
> /x86_64/sync_regs_test
> +/x86_64/vmx_capability_msr_test
> /x86_64/vmx_close_while_nested_test
> /x86_64/vmx_tsc_adjust_test
> /x86_64/state_test
> diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile
> index f8588cca2bef..a635a2c42592 100644
> --- a/tools/testing/selftests/kvm/Makefile
> +++ b/tools/testing/selftests/kvm/Makefile
> @@ -18,6 +18,7 @@ TEST_GEN_PROGS_x86_64 += x86_64/cr4_cpuid_sync_test
> TEST_GEN_PROGS_x86_64 += x86_64/state_test
> TEST_GEN_PROGS_x86_64 += x86_64/evmcs_test
> TEST_GEN_PROGS_x86_64 += x86_64/hyperv_cpuid
> +TEST_GEN_PROGS_x86_64 += x86_64/vmx_capability_msr_test
> TEST_GEN_PROGS_x86_64 += x86_64/vmx_close_while_nested_test
> TEST_GEN_PROGS_x86_64 += x86_64/smm_test
> TEST_GEN_PROGS_x86_64 += dirty_log_test
> diff --git a/tools/testing/selftests/kvm/include/kvm_util.h b/tools/testing/selftests/kvm/include/kvm_util.h
> index 07b71ad9734a..6a2e27cad877 100644
> --- a/tools/testing/selftests/kvm/include/kvm_util.h
> +++ b/tools/testing/selftests/kvm/include/kvm_util.h
> @@ -133,6 +133,8 @@ struct kvm_vm *vm_create_default(uint32_t vcpuid, uint64_t extra_mem_size,
> void *guest_code);
> void vm_vcpu_add_default(struct kvm_vm *vm, uint32_t vcpuid, void *guest_code);
>
> +int vcpu_fd(struct kvm_vm *vm, uint32_t vcpuid);
> +
> struct kvm_userspace_memory_region *
> kvm_userspace_memory_region_find(struct kvm_vm *vm, uint64_t start,
> uint64_t end);
> diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
> index 4ca96b228e46..3bc4d2f633c8 100644
> --- a/tools/testing/selftests/kvm/lib/kvm_util.c
> +++ b/tools/testing/selftests/kvm/lib/kvm_util.c
> @@ -402,6 +402,11 @@ static void vm_vcpu_rm(struct kvm_vm *vm, uint32_t vcpuid)
> free(vcpu);
> }
>
> +int vcpu_fd(struct kvm_vm *vm, uint32_t vcpuid)
> +{
> + return vcpu_find(vm, vcpuid)->fd;
> +}
> +
> void kvm_vm_release(struct kvm_vm *vmp)
> {
> int ret;
> diff --git a/tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c b/tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c
> new file mode 100644
> index 000000000000..161a037299f0
> --- /dev/null
> +++ b/tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c
> @@ -0,0 +1,567 @@
> +/*
> + * vmx_capability_msr_test
> + *
> + * Copyright (C) 2019, Google LLC.
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2.
> + *
> + * Test virtualization of VMX capability MSRs.
> + */
> +
> +#define _GNU_SOURCE /* for program_invocation_short_name */
> +#include <fcntl.h>
> +#include <linux/bits.h>
> +#include <linux/kernel.h>
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <string.h>
> +#include <sys/ioctl.h>
> +#include <sys/stat.h>
> +#include <test_util.h>
> +
> +#include "kvm_util.h"
> +#include "processor.h"
> +#include "vmx.h"
> +
> +#define VCPU_ID 0
> +
> +struct kvm_single_msr {
> + struct kvm_msrs header;
> + struct kvm_msr_entry entry;
> +} __packed;
> +
> +struct vmx_msr {
> + uint32_t index;
> + const char *name;
> +};
> +#define VMX_MSR(_msr) { _msr, #_msr }
> +
> +struct vmx_msr vmx_msrs[] = {
> + VMX_MSR(MSR_IA32_VMX_BASIC),
> + VMX_MSR(MSR_IA32_VMX_TRUE_PINBASED_CTLS),
> + VMX_MSR(MSR_IA32_VMX_TRUE_PROCBASED_CTLS),
> + VMX_MSR(MSR_IA32_VMX_PROCBASED_CTLS2),
> + VMX_MSR(MSR_IA32_VMX_TRUE_EXIT_CTLS),
> + VMX_MSR(MSR_IA32_VMX_TRUE_ENTRY_CTLS),
> + VMX_MSR(MSR_IA32_VMX_MISC),
> + VMX_MSR(MSR_IA32_VMX_EPT_VPID_CAP),
> + VMX_MSR(MSR_IA32_VMX_CR0_FIXED0),
> + VMX_MSR(MSR_IA32_VMX_CR0_FIXED1),
> + VMX_MSR(MSR_IA32_VMX_CR4_FIXED0),
> + VMX_MSR(MSR_IA32_VMX_CR4_FIXED1),
> + VMX_MSR(MSR_IA32_VMX_VMCS_ENUM),
> +};
> +
> +struct kvm_vm *vm;
> +
> +void guest_code(void)
> +{
> + __asm__ __volatile__(
> + "1:"
> + "rdmsr;"
> + "outw $0;"
> + "jmp 1b;"
> + );
> +}
> +
> +uint64_t read_msr_from_guest(uint32_t msr)
> +{
> + unsigned int exit_reason, expected_exit = KVM_EXIT_IO;
> + struct kvm_regs regs;
> +
> + vcpu_regs_get(vm, VCPU_ID, ®s);
> + regs.rcx = msr;
> + regs.rax = 0;
> + regs.rdx = 0;
> + vcpu_regs_set(vm, VCPU_ID, ®s);
> +
> + vcpu_run(vm, VCPU_ID);
> + vcpu_regs_get(vm, VCPU_ID, ®s);
> +
> + exit_reason = vcpu_state(vm, VCPU_ID)->exit_reason;
> + TEST_ASSERT(exit_reason == expected_exit,
> + "Unexpected exit reason: expected %u (%s), got %u (%s) [guest rip: %llx]",
> + expected_exit, exit_reason_str(expected_exit),
> + exit_reason, exit_reason_str(exit_reason),
> + regs.rip);
> +
> + return (regs.rdx << 32) | regs.rax;
> +}
> +
> +int try_kvm_get_msr(int vcpu_fd, uint32_t msr, uint64_t *data)
> +{
> + struct kvm_single_msr buffer;
> + int rv;
> +
> + memset(&buffer, 0, sizeof(buffer));
> + buffer.header.nmsrs = 1;
> + buffer.entry.index = msr;
> +
> + /* Returns the number of MSRs read. */
> + rv = ioctl(vcpu_fd, KVM_GET_MSRS, &buffer);
> + *data = buffer.entry.data;
> + return rv;
> +}
> +
> +uint64_t kvm_get_msr(int vcpu_fd, uint32_t msr)
> +{
> + uint64_t data;
> + int rv;
> +
> + rv = try_kvm_get_msr(vcpu_fd, msr, &data);
> + TEST_ASSERT(rv == 1, "KVM_GET_MSR of MSR %x failed. (rv %d)",
> + msr, rv);
> +
> + return data;
> +}
> +
> +void kvm_get_msr_expect_fail(int vcpu_fd, uint32_t msr)
> +{
> + uint64_t data;
> + int rv;
> +
> + rv = try_kvm_get_msr(vcpu_fd, msr, &data);
> + TEST_ASSERT(rv != 1, "Expected KVM_GET_MSR of MSR %x to fail.", msr);
> +}
> +
> +int try_kvm_set_msr(int vcpu_fd, uint32_t msr, uint64_t data)
> +{
> + struct kvm_single_msr buffer;
> + int rv;
> +
> + memset(&buffer, 0, sizeof(buffer));
> + buffer.header.nmsrs = 1;
> + buffer.entry.index = msr;
> + buffer.entry.data = data;
> +
> + /* Returns the number of MSRs written. */
> + rv = ioctl(vcpu_fd, KVM_SET_MSRS, &buffer);
> + return rv;
> +}
> +
> +void kvm_set_msr_expect_fail(int vcpu_fd, uint32_t msr, uint64_t data)
> +{
> + int rv;
> +
> + rv = try_kvm_set_msr(vcpu_fd, msr, data);
> + TEST_ASSERT(rv != 1, "Expected KVM_SET_MSR of MSR %x to fail.", msr);
> +}
> +
> +void kvm_set_msr(int vcpu_fd, uint32_t msr, uint64_t data)
> +{
> + int rv;
> +
> + rv = try_kvm_set_msr(vcpu_fd, msr, data);
> + TEST_ASSERT(rv == 1, "KVM_SET_MSR %x (data %lx) failed.", msr, data);
> +}
> +
> +void test_set_vmx_msr(int vcpu_fd, uint32_t msr, uint64_t data)
> +{
> + uint64_t guest, kvm;
> + uint64_t non_true;
> +
> + kvm_set_msr(vcpu_fd, msr, data);
> +
> + guest = read_msr_from_guest(msr);
> + kvm = kvm_get_msr(vcpu_fd, msr);
> +
> + TEST_ASSERT(guest == data,
> + "KVM_SET_MSR %x not visible from guest (guest %lx, expected %lx)",
> + msr, guest, data);
> +
> + TEST_ASSERT(kvm == data,
> + "KVM_SET_MSR %x not visible from host (host %lx, expected %lx)",
> + msr, kvm, data);
> +
> +#define IA32_VMX_PINBASED_CTLS_DEFAULT1 (BIT_ULL(1) | BIT_ULL(2) | BIT_ULL(4))
> +#define IA32_VMX_PROCBASED_CTLS_DEFAULT1 (BIT_ULL(1) | BIT_ULL(4) | \
> + BIT_ULL(5) | BIT_ULL(6) | \
> + BIT_ULL(8) | BIT_ULL(13) | \
> + BIT_ULL(14) | BIT_ULL(15) | \
> + BIT_ULL(16) | BIT_ULL(26))
> +#define IA32_VMX_EXIT_CTLS_DEFAULT1 (BIT_ULL(0) | BIT_ULL(1) | BIT_ULL(2) | \
> + BIT_ULL(3) | BIT_ULL(4) | BIT_ULL(5) | \
> + BIT_ULL(6) | BIT_ULL(7) | BIT_ULL(8) | \
> + BIT_ULL(10) | BIT_ULL(11) | BIT_ULL(13) | \
> + BIT_ULL(14) | BIT_ULL(16) | BIT_ULL(17))
> +#define IA32_VMX_ENTRY_CTLS_DEFAULT1 (BIT_ULL(0) | BIT_ULL(1) | BIT_ULL(2) | \
> + BIT_ULL(3) | BIT_ULL(4) | BIT_ULL(5) | \
> + BIT_ULL(6) | BIT_ULL(7) | BIT_ULL(8) | \
> + BIT_ULL(12))
> + /*
> + * Make sure non-true MSRs are equal to the true MSRs with all the
> + * default1 bits set.
> + */
> + switch (msr) {
> + case MSR_IA32_VMX_TRUE_PINBASED_CTLS:
> + non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_PINBASED_CTLS);
> + TEST_ASSERT(non_true ==
> + (data | IA32_VMX_PINBASED_CTLS_DEFAULT1),
> + "Incorrect IA32_VMX_PINBASED_CTLS MSR: %lx\n",
> + non_true);
> + break;
> + case MSR_IA32_VMX_TRUE_PROCBASED_CTLS:
> + non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_PROCBASED_CTLS);
> + TEST_ASSERT(non_true ==
> + (data | IA32_VMX_PROCBASED_CTLS_DEFAULT1),
> + "Incorrect IA32_VMX_PROCBASED_CTLS MSR: %lx\n",
> + non_true);
> + break;
> + case MSR_IA32_VMX_TRUE_EXIT_CTLS:
> + non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_EXIT_CTLS);
> + TEST_ASSERT(non_true == (data | IA32_VMX_EXIT_CTLS_DEFAULT1),
> + "Incorrect IA32_VMX_EXIT_CTLS MSR: %lx\n",
> + non_true);
> + break;
> + case MSR_IA32_VMX_TRUE_ENTRY_CTLS:
> + non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_ENTRY_CTLS);
> + TEST_ASSERT(non_true == (data | IA32_VMX_ENTRY_CTLS_DEFAULT1),
> + "Incorrect IA32_VMX_ENTRY_CTLS MSR: %lx\n",
> + non_true);
> + break;
> + }
> +}
> +
> +/*
> + * Reset the VCPU and return its file descriptor.
> + *
> + * The VCPU should be reset in between subsequent restores of the
> + * same VMX MSR.
> + */
> +int reset_vcpu_fd(bool vmx_supported)
> +{
> + struct kvm_cpuid2 *cpuid;
> + int fd;
> +
> + if (vm)
> + kvm_vm_free(vm);
> +
> + vm = vm_create_default(VCPU_ID, 0, guest_code);
> + fd = vcpu_fd(vm, VCPU_ID);
> +
> + /* Enable/Disable VMX */
> + if (vmx_supported)
> + kvm_get_supported_cpuid_entry(0x1)->ecx |= (1 << 5);
> + else
> + kvm_get_supported_cpuid_entry(0x1)->ecx &= ~(1 << 5);
> + cpuid = kvm_get_supported_cpuid();
> + vcpu_set_cpuid(vm, VCPU_ID, cpuid);
> +
> + return fd;
> +}
> +
> +void test_msr_bits(uint32_t index, int high, int low, bool allowed_1)
> +{
> + int fd = reset_vcpu_fd(true);
> + uint64_t value = kvm_get_msr(fd, index);
> + int i;
> +
> + for (i = low; i <= high; i++) {
> + uint64_t mask = 1ull << i;
> + bool is_1 = value & mask;
> +
> + fd = reset_vcpu_fd(true);
> +
> + /*
> + * Following VMX nomenclature:
> + *
> + * allowed-1 means a bit is high when it can be 0 or 1. The
> + * bit is low when it must be 0.
> + *
> + * allowed-0 (!allowed_1) means a bit is low when it can be
> + * 0 or 1. The bit is high when it must be 1.
> + */
> + if (allowed_1 && is_1)
> + test_set_vmx_msr(fd, index, value & ~mask);
> + else if (!allowed_1 && !is_1)
> + test_set_vmx_msr(fd, index, value | mask);
> + else if (allowed_1 && !is_1)
> + kvm_set_msr_expect_fail(fd, index, value | mask);
> + else
> + kvm_set_msr_expect_fail(fd, index, value & ~mask);
> + }
> +}
> +
> +void test_allowed_1_bits(uint32_t index, int high, int low)
> +{
> + test_msr_bits(index, high, low, true);
> +}
> +#define test_allowed_1_bit(_index, _bit) \
> + test_allowed_1_bits(_index, _bit, _bit)
> +
> +void test_allowed_0_bits(uint32_t index, int high, int low)
> +{
> + test_msr_bits(index, high, low, false);
> +}
> +
> +uint64_t get_bits(uint64_t value, int high, int low)
> +{
> + return (value & GENMASK_ULL(high, low)) >> low;
> +}
> +
> +uint64_t set_bits(uint64_t value, int high, int low, uint64_t new_value)
> +{
> + uint64_t mask = GENMASK_ULL(high, low);
> +
> + return (value & ~mask) | ((new_value << low) & mask);
> +}
> +
> +/*
> + * Test MSR[high:low].
> + *
> + * lt : KVM should support values less than the default.
> + * gt : KVM should support values greater than the default.
> + */
> +void __test_msr_bits_eq(uint32_t index, int high, int low, bool lt, bool gt)
> +{
> + int fd = reset_vcpu_fd(true);
> + uint64_t value = kvm_get_msr(fd, index);
> + uint64_t bits = get_bits(value, high, low);
> + uint64_t new_value;
> +
> + if (bits > 0) {
> + fd = reset_vcpu_fd(true);
> + new_value = set_bits(value, high, low, bits - 1);
> + if (lt)
> + test_set_vmx_msr(fd, index, new_value);
> + else
> + kvm_set_msr_expect_fail(fd, index, new_value);
> + }
> +
> + if (bits < get_bits(-1ull, high, low)) {
> + fd = reset_vcpu_fd(true);
> + new_value = set_bits(value, high, low, bits + 1);
> + if (gt)
> + test_set_vmx_msr(fd, index, new_value);
> + else
> + kvm_set_msr_expect_fail(fd, index, new_value);
> + }
> +}
> +
> +#define test_msr_bits_eq(...) __test_msr_bits_eq(__VA_ARGS__, false, false)
> +#define test_msr_bits_le(...) __test_msr_bits_eq(__VA_ARGS__, true, false)
> +
> +void test_rw_vmx_capability_msrs(void)
> +{
> + int i;
> +
> + printf("VMX MSR default values:\n");
> +
> + for (i = 0; i < ARRAY_SIZE(vmx_msrs); i++) {
> + uint32_t msr = vmx_msrs[i].index;
> + uint64_t guest, kvm;
> + int vcpu_fd;
> +
> + /*
> + * If VMX is not supported, test that reading the VMX MSRs
> + * fails.
> + */
> + vcpu_fd = reset_vcpu_fd(false);
> + kvm_get_msr_expect_fail(vcpu_fd, msr);
> +
> + vcpu_fd = reset_vcpu_fd(true);
> + guest = read_msr_from_guest(msr);
> + kvm = kvm_get_msr(vcpu_fd, msr);
> +
> + TEST_ASSERT(guest == kvm,
> + "MSR %x mismatch: guest %lx kvm %lx.",
> + msr, guest, kvm);
> +
> + printf("{ %s, 0x%lx },\n", vmx_msrs[i].name, guest);
> +
> + /* VM Execution Control Capability MSRs */
> + switch (msr) {
> + case MSR_IA32_VMX_BASIC:
> + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> + test_msr_bits_eq(msr, 30, 0); /* vmcs revision */
> + test_allowed_1_bit(msr, 31); /* reserved */
> + test_allowed_1_bit(msr, 48); /* VMCS phys bits */
> + test_allowed_1_bit(msr, 49); /* dual-monitor SMI/SMM */
> + test_allowed_1_bit(msr, 54); /* ins/outs info */
> + test_allowed_1_bit(msr, 55); /* true ctls */
> + break;
> + case MSR_IA32_VMX_MISC:
> + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> + test_msr_bits_eq(msr, 4, 0); /* preempt timer rate */
> + test_allowed_1_bit(msr, 5); /* store IA-32e mode */
> + test_allowed_1_bits(msr, 8, 6); /* activity states */
> + test_allowed_1_bits(msr, 13, 9); /* reserved */
> + test_allowed_1_bit(msr, 14); /* pt in vmx mode */
> + test_allowed_1_bit(msr, 15); /* smbase msr in smm */
> + test_msr_bits_le(msr, 24, 16); /* CR3-target values */
> + test_msr_bits_le(msr, 27, 25); /* MSR-store list size */
> + test_allowed_1_bit(msr, 28); /* smm_monitor_ctl 1 */
> + test_allowed_1_bit(msr, 29); /* vmwrite to any */
> + test_allowed_1_bit(msr, 30); /* insn len 0 allowed */
> + test_allowed_1_bit(msr, 31); /* reserved */
> + test_msr_bits_eq(msr, 63, 32); /* MSEG Revision */
> + break;
> + case MSR_IA32_VMX_TRUE_PINBASED_CTLS:
> + case MSR_IA32_VMX_TRUE_PROCBASED_CTLS:
> + case MSR_IA32_VMX_TRUE_EXIT_CTLS:
> + case MSR_IA32_VMX_TRUE_ENTRY_CTLS:
> + case MSR_IA32_VMX_PROCBASED_CTLS2:
> + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> + test_allowed_0_bits(msr, 31, 0); /* allowed-1 bits */
> + test_allowed_1_bits(msr, 63, 32); /* allowed-0 bits */
> + break;
> + case MSR_IA32_VMX_EPT_VPID_CAP:
> + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> + test_allowed_1_bits(msr, 63, 0); /* feature/reserved */
> + break;
> + case MSR_IA32_VMX_CR0_FIXED0:
> + case MSR_IA32_VMX_CR4_FIXED0:
> + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> + test_allowed_0_bits(msr, 63, 0); /* allowed-0 bits */
> + break;
> + case MSR_IA32_VMX_CR0_FIXED1:
> + case MSR_IA32_VMX_CR4_FIXED1:
> + /* These MSRs do not allow save-restore. */
> + kvm_set_msr_expect_fail(vcpu_fd, msr, guest);
> + break;
> + case MSR_IA32_VMX_PROCBASED_CTLS:
> + case MSR_IA32_VMX_PINBASED_CTLS:
> + case MSR_IA32_VMX_EXIT_CTLS:
> + case MSR_IA32_VMX_ENTRY_CTLS:
> + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> + break;
> + case MSR_IA32_VMX_VMCS_ENUM:
> + test_allowed_1_bit(msr, 0); /* reserved */
> + test_allowed_1_bits(msr, 63, 10); /* reserved */
> + break;
> + default:
> + TEST_ASSERT(false, "Untested MSR: 0x%x.", msr);
> + }
> + }
> +}
> +
> +struct msr {
> + uint32_t index;
> + uint64_t data;
> +};
> +
> +/*
> + * Test that MSR_IA32_VMX_CR4_FIXED1 is correctly updated in response to
> + * CPUID changes.
> + */
> +void test_cr4_fixed1_cpuid(void)
> +{
> + struct test {
> + uint64_t cr4_mask;
> + int function;
> + enum x86_register reg;
> + int cpuid_bit;
> + const char *name;
> + } tests[] = {
> + { X86_CR4_VME, 1, RDX, 1, "VME" },
> + { X86_CR4_PVI, 1, RDX, 1, "PVI" },
> + { X86_CR4_TSD, 1, RDX, 4, "TSD" },
> + { X86_CR4_DE, 1, RDX, 2, "DE" },
> + { X86_CR4_PSE, 1, RDX, 3, "PSE" },
> + { X86_CR4_PAE, 1, RDX, 6, "PAE" },
> + { X86_CR4_MCE, 1, RDX, 7, "MCE" },
> + { X86_CR4_PGE, 1, RDX, 13, "PGE" },
> + { X86_CR4_OSFXSR, 1, RDX, 24, "OSFXSR" },
> + { X86_CR4_OSXMMEXCPT, 1, RDX, 25, "OSXMMEXCPT" },
> + { X86_CR4_UMIP, 7, RCX, 2, "UMIP" },
> + { X86_CR4_VMXE, 1, RCX, 5, "VMXE" },
> + { X86_CR4_SMXE, 1, RCX, 6, "SMXE" },
> + { X86_CR4_FSGSBASE, 7, RBX, 0, "FSGSBASE" },
> + { X86_CR4_PCIDE, 1, RCX, 17, "PCIDE" },
> + { X86_CR4_OSXSAVE, 1, RCX, 26, "OSXSAVE" },
> + { X86_CR4_SMEP, 7, RBX, 7, "SMEP" },
> + { X86_CR4_SMAP, 7, RBX, 20, "SMAP" },
> + { X86_CR4_PKE, 7, RCX, 3, "PKE" },
> + };
> + int vcpu_fd = reset_vcpu_fd(true);
> + int i;
> +
> + kvm_get_supported_cpuid_entry(0x1)->ecx |= (1 << 5);
> +
> + for (i = 0; i < ARRAY_SIZE(tests); i++) {
> + struct test *test = &tests[i];
> + struct kvm_cpuid_entry2 *entry;
> + struct kvm_cpuid2 *cpuid = kvm_get_supported_cpuid();
> + const char *reg_name = NULL;
> + uint64_t fixed1;
> + uint32_t *reg = NULL;
> +
> + entry = kvm_get_supported_cpuid_entry(test->function);
> + TEST_ASSERT(entry, "CPUID lookup failed: CPUID.%02xH",
> + test->function);
> +
> + switch (test->reg) {
> + case RAX:
> + reg = &entry->eax; reg_name = "EAX";
> + break;
> + case RBX:
> + reg = &entry->ebx; reg_name = "EBX";
> + break;
> + case RCX:
> + reg = &entry->ecx; reg_name = "ECX";
> + break;
> + case RDX:
> + reg = &entry->edx; reg_name = "EDX";
> + break;
> + default:
> + TEST_ASSERT(false, "Unrecognized reg: %d", test->reg);
> + }
> +
> + printf("MSR_IA32_VMX_CR4_FIXED1[%d] (%s) == CPUID.%02xH:%s[%d]\n",
> + __builtin_ffsll(test->cr4_mask) - 1, test->name,
> + test->function, reg_name, test->cpuid_bit);
> +
> +
> + *reg |= 1ull << test->cpuid_bit;
> + vcpu_set_cpuid(vm, VCPU_ID, cpuid);
> + fixed1 = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_CR4_FIXED1);
> + TEST_ASSERT(fixed1 & test->cr4_mask, "set bit");
> + TEST_ASSERT(fixed1 & X86_CR4_PCE, "PCE");
> +
> + /*
> + * Skipping clearing support for VMX. VMX is required for
> + * MSR_IA32_VMX_CR4_FIXED1 to even exist.
> + */
> + if (test->cr4_mask == X86_CR4_VMXE)
> + continue;
> +
> + *reg &= ~(1ull << test->cpuid_bit);
> + vcpu_set_cpuid(vm, VCPU_ID, cpuid);
> + fixed1 = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_CR4_FIXED1);
> + TEST_ASSERT(!(fixed1 & test->cr4_mask), "clear bit");
> + TEST_ASSERT(fixed1 & X86_CR4_PCE, "PCE");
> + };
> +}
> +
> +void test_preemption_timer_capability(void)
> +{
> + int fd = reset_vcpu_fd(true);
> + uint64_t vmx_misc = kvm_get_msr(fd, MSR_IA32_VMX_MISC);
> + uint64_t pinbased = kvm_get_msr(fd, MSR_IA32_VMX_TRUE_PINBASED_CTLS);
> +
> +#define VMX_PREEMPTION_TIMER_CONTROL \
> + ((uint64_t) PIN_BASED_VMX_PREEMPTION_TIMER << 32)
> +#define VMX_PREEMPTION_TIMER_RATE 0x1f
> +
> + TEST_ASSERT(pinbased & VMX_PREEMPTION_TIMER_CONTROL,
> + "Expected support for VMX preemption timer.");
> +
> + /* Test that we can disable the VMX preemption timer. */
> + test_set_vmx_msr(fd, MSR_IA32_VMX_TRUE_PINBASED_CTLS,
> + pinbased & ~VMX_PREEMPTION_TIMER_CONTROL);
> +
> + /*
> + * Test that we can clear the VMX preemption timer rate when
> + * the preemption timer is disabled.
> + */
> + test_set_vmx_msr(fd, MSR_IA32_VMX_MISC,
> + vmx_misc & ~VMX_PREEMPTION_TIMER_RATE);
> +}
> +
> +int main(int argc, char *argv[])
> +{
> + test_rw_vmx_capability_msrs();
> + test_cr4_fixed1_cpuid();
> + test_preemption_timer_capability();
> + return 0;
> +}
> --
> 2.22.0.rc1.311.g5d7573a151-goog
>
ping
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH 2/2] tests: kvm: Test virtualization of VMX capability MSRs
2019-06-18 14:14 ` Aaron Lewis
@ 2019-07-10 16:09 ` Aaron Lewis
0 siblings, 0 replies; 3+ messages in thread
From: Aaron Lewis @ 2019-07-10 16:09 UTC (permalink / raw)
To: Jim Mattson, Peter Shier, Marc Orr, kvm
On Tue, Jun 18, 2019 at 7:14 AM Aaron Lewis <aaronlewis@google.com> wrote:
>
> On Fri, May 31, 2019 at 11:42 AM Aaron Lewis <aaronlewis@google.com> wrote:
> >
> > This tests three main things:
> > 1. Test that the VMX capability MSRs can be written to and read from
> > based on the rules in the SDM.
> > 2. Test that the MSR_IA32_VMX_CR4_FIXED1 is correctly updated in
> > response to CPUID changes.
> > 3. Test that the VMX preemption timer rate can be cleared when the
> > preemption timer is disabled.
> >
> > Signed-off-by: Aaron Lewis <aaronlewis@google.com>
> > Reviewed-by: Peter Shier <pshier@google.com>
> > ---
> > tools/testing/selftests/kvm/.gitignore | 1 +
> > tools/testing/selftests/kvm/Makefile | 1 +
> > .../testing/selftests/kvm/include/kvm_util.h | 2 +
> > tools/testing/selftests/kvm/lib/kvm_util.c | 5 +
> > .../kvm/x86_64/vmx_capability_msr_test.c | 567 ++++++++++++++++++
> > 5 files changed, 576 insertions(+)
> > create mode 100644 tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c
> >
> > diff --git a/tools/testing/selftests/kvm/.gitignore b/tools/testing/selftests/kvm/.gitignore
> > index 2689d1ea6d7a..94ca75f9f4cd 100644
> > --- a/tools/testing/selftests/kvm/.gitignore
> > +++ b/tools/testing/selftests/kvm/.gitignore
> > @@ -3,6 +3,7 @@
> > /x86_64/platform_info_test
> > /x86_64/set_sregs_test
> > /x86_64/sync_regs_test
> > +/x86_64/vmx_capability_msr_test
> > /x86_64/vmx_close_while_nested_test
> > /x86_64/vmx_tsc_adjust_test
> > /x86_64/state_test
> > diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile
> > index f8588cca2bef..a635a2c42592 100644
> > --- a/tools/testing/selftests/kvm/Makefile
> > +++ b/tools/testing/selftests/kvm/Makefile
> > @@ -18,6 +18,7 @@ TEST_GEN_PROGS_x86_64 += x86_64/cr4_cpuid_sync_test
> > TEST_GEN_PROGS_x86_64 += x86_64/state_test
> > TEST_GEN_PROGS_x86_64 += x86_64/evmcs_test
> > TEST_GEN_PROGS_x86_64 += x86_64/hyperv_cpuid
> > +TEST_GEN_PROGS_x86_64 += x86_64/vmx_capability_msr_test
> > TEST_GEN_PROGS_x86_64 += x86_64/vmx_close_while_nested_test
> > TEST_GEN_PROGS_x86_64 += x86_64/smm_test
> > TEST_GEN_PROGS_x86_64 += dirty_log_test
> > diff --git a/tools/testing/selftests/kvm/include/kvm_util.h b/tools/testing/selftests/kvm/include/kvm_util.h
> > index 07b71ad9734a..6a2e27cad877 100644
> > --- a/tools/testing/selftests/kvm/include/kvm_util.h
> > +++ b/tools/testing/selftests/kvm/include/kvm_util.h
> > @@ -133,6 +133,8 @@ struct kvm_vm *vm_create_default(uint32_t vcpuid, uint64_t extra_mem_size,
> > void *guest_code);
> > void vm_vcpu_add_default(struct kvm_vm *vm, uint32_t vcpuid, void *guest_code);
> >
> > +int vcpu_fd(struct kvm_vm *vm, uint32_t vcpuid);
> > +
> > struct kvm_userspace_memory_region *
> > kvm_userspace_memory_region_find(struct kvm_vm *vm, uint64_t start,
> > uint64_t end);
> > diff --git a/tools/testing/selftests/kvm/lib/kvm_util.c b/tools/testing/selftests/kvm/lib/kvm_util.c
> > index 4ca96b228e46..3bc4d2f633c8 100644
> > --- a/tools/testing/selftests/kvm/lib/kvm_util.c
> > +++ b/tools/testing/selftests/kvm/lib/kvm_util.c
> > @@ -402,6 +402,11 @@ static void vm_vcpu_rm(struct kvm_vm *vm, uint32_t vcpuid)
> > free(vcpu);
> > }
> >
> > +int vcpu_fd(struct kvm_vm *vm, uint32_t vcpuid)
> > +{
> > + return vcpu_find(vm, vcpuid)->fd;
> > +}
> > +
> > void kvm_vm_release(struct kvm_vm *vmp)
> > {
> > int ret;
> > diff --git a/tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c b/tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c
> > new file mode 100644
> > index 000000000000..161a037299f0
> > --- /dev/null
> > +++ b/tools/testing/selftests/kvm/x86_64/vmx_capability_msr_test.c
> > @@ -0,0 +1,567 @@
> > +/*
> > + * vmx_capability_msr_test
> > + *
> > + * Copyright (C) 2019, Google LLC.
> > + *
> > + * This work is licensed under the terms of the GNU GPL, version 2.
> > + *
> > + * Test virtualization of VMX capability MSRs.
> > + */
> > +
> > +#define _GNU_SOURCE /* for program_invocation_short_name */
> > +#include <fcntl.h>
> > +#include <linux/bits.h>
> > +#include <linux/kernel.h>
> > +#include <stdio.h>
> > +#include <stdlib.h>
> > +#include <string.h>
> > +#include <sys/ioctl.h>
> > +#include <sys/stat.h>
> > +#include <test_util.h>
> > +
> > +#include "kvm_util.h"
> > +#include "processor.h"
> > +#include "vmx.h"
> > +
> > +#define VCPU_ID 0
> > +
> > +struct kvm_single_msr {
> > + struct kvm_msrs header;
> > + struct kvm_msr_entry entry;
> > +} __packed;
> > +
> > +struct vmx_msr {
> > + uint32_t index;
> > + const char *name;
> > +};
> > +#define VMX_MSR(_msr) { _msr, #_msr }
> > +
> > +struct vmx_msr vmx_msrs[] = {
> > + VMX_MSR(MSR_IA32_VMX_BASIC),
> > + VMX_MSR(MSR_IA32_VMX_TRUE_PINBASED_CTLS),
> > + VMX_MSR(MSR_IA32_VMX_TRUE_PROCBASED_CTLS),
> > + VMX_MSR(MSR_IA32_VMX_PROCBASED_CTLS2),
> > + VMX_MSR(MSR_IA32_VMX_TRUE_EXIT_CTLS),
> > + VMX_MSR(MSR_IA32_VMX_TRUE_ENTRY_CTLS),
> > + VMX_MSR(MSR_IA32_VMX_MISC),
> > + VMX_MSR(MSR_IA32_VMX_EPT_VPID_CAP),
> > + VMX_MSR(MSR_IA32_VMX_CR0_FIXED0),
> > + VMX_MSR(MSR_IA32_VMX_CR0_FIXED1),
> > + VMX_MSR(MSR_IA32_VMX_CR4_FIXED0),
> > + VMX_MSR(MSR_IA32_VMX_CR4_FIXED1),
> > + VMX_MSR(MSR_IA32_VMX_VMCS_ENUM),
> > +};
> > +
> > +struct kvm_vm *vm;
> > +
> > +void guest_code(void)
> > +{
> > + __asm__ __volatile__(
> > + "1:"
> > + "rdmsr;"
> > + "outw $0;"
> > + "jmp 1b;"
> > + );
> > +}
> > +
> > +uint64_t read_msr_from_guest(uint32_t msr)
> > +{
> > + unsigned int exit_reason, expected_exit = KVM_EXIT_IO;
> > + struct kvm_regs regs;
> > +
> > + vcpu_regs_get(vm, VCPU_ID, ®s);
> > + regs.rcx = msr;
> > + regs.rax = 0;
> > + regs.rdx = 0;
> > + vcpu_regs_set(vm, VCPU_ID, ®s);
> > +
> > + vcpu_run(vm, VCPU_ID);
> > + vcpu_regs_get(vm, VCPU_ID, ®s);
> > +
> > + exit_reason = vcpu_state(vm, VCPU_ID)->exit_reason;
> > + TEST_ASSERT(exit_reason == expected_exit,
> > + "Unexpected exit reason: expected %u (%s), got %u (%s) [guest rip: %llx]",
> > + expected_exit, exit_reason_str(expected_exit),
> > + exit_reason, exit_reason_str(exit_reason),
> > + regs.rip);
> > +
> > + return (regs.rdx << 32) | regs.rax;
> > +}
> > +
> > +int try_kvm_get_msr(int vcpu_fd, uint32_t msr, uint64_t *data)
> > +{
> > + struct kvm_single_msr buffer;
> > + int rv;
> > +
> > + memset(&buffer, 0, sizeof(buffer));
> > + buffer.header.nmsrs = 1;
> > + buffer.entry.index = msr;
> > +
> > + /* Returns the number of MSRs read. */
> > + rv = ioctl(vcpu_fd, KVM_GET_MSRS, &buffer);
> > + *data = buffer.entry.data;
> > + return rv;
> > +}
> > +
> > +uint64_t kvm_get_msr(int vcpu_fd, uint32_t msr)
> > +{
> > + uint64_t data;
> > + int rv;
> > +
> > + rv = try_kvm_get_msr(vcpu_fd, msr, &data);
> > + TEST_ASSERT(rv == 1, "KVM_GET_MSR of MSR %x failed. (rv %d)",
> > + msr, rv);
> > +
> > + return data;
> > +}
> > +
> > +void kvm_get_msr_expect_fail(int vcpu_fd, uint32_t msr)
> > +{
> > + uint64_t data;
> > + int rv;
> > +
> > + rv = try_kvm_get_msr(vcpu_fd, msr, &data);
> > + TEST_ASSERT(rv != 1, "Expected KVM_GET_MSR of MSR %x to fail.", msr);
> > +}
> > +
> > +int try_kvm_set_msr(int vcpu_fd, uint32_t msr, uint64_t data)
> > +{
> > + struct kvm_single_msr buffer;
> > + int rv;
> > +
> > + memset(&buffer, 0, sizeof(buffer));
> > + buffer.header.nmsrs = 1;
> > + buffer.entry.index = msr;
> > + buffer.entry.data = data;
> > +
> > + /* Returns the number of MSRs written. */
> > + rv = ioctl(vcpu_fd, KVM_SET_MSRS, &buffer);
> > + return rv;
> > +}
> > +
> > +void kvm_set_msr_expect_fail(int vcpu_fd, uint32_t msr, uint64_t data)
> > +{
> > + int rv;
> > +
> > + rv = try_kvm_set_msr(vcpu_fd, msr, data);
> > + TEST_ASSERT(rv != 1, "Expected KVM_SET_MSR of MSR %x to fail.", msr);
> > +}
> > +
> > +void kvm_set_msr(int vcpu_fd, uint32_t msr, uint64_t data)
> > +{
> > + int rv;
> > +
> > + rv = try_kvm_set_msr(vcpu_fd, msr, data);
> > + TEST_ASSERT(rv == 1, "KVM_SET_MSR %x (data %lx) failed.", msr, data);
> > +}
> > +
> > +void test_set_vmx_msr(int vcpu_fd, uint32_t msr, uint64_t data)
> > +{
> > + uint64_t guest, kvm;
> > + uint64_t non_true;
> > +
> > + kvm_set_msr(vcpu_fd, msr, data);
> > +
> > + guest = read_msr_from_guest(msr);
> > + kvm = kvm_get_msr(vcpu_fd, msr);
> > +
> > + TEST_ASSERT(guest == data,
> > + "KVM_SET_MSR %x not visible from guest (guest %lx, expected %lx)",
> > + msr, guest, data);
> > +
> > + TEST_ASSERT(kvm == data,
> > + "KVM_SET_MSR %x not visible from host (host %lx, expected %lx)",
> > + msr, kvm, data);
> > +
> > +#define IA32_VMX_PINBASED_CTLS_DEFAULT1 (BIT_ULL(1) | BIT_ULL(2) | BIT_ULL(4))
> > +#define IA32_VMX_PROCBASED_CTLS_DEFAULT1 (BIT_ULL(1) | BIT_ULL(4) | \
> > + BIT_ULL(5) | BIT_ULL(6) | \
> > + BIT_ULL(8) | BIT_ULL(13) | \
> > + BIT_ULL(14) | BIT_ULL(15) | \
> > + BIT_ULL(16) | BIT_ULL(26))
> > +#define IA32_VMX_EXIT_CTLS_DEFAULT1 (BIT_ULL(0) | BIT_ULL(1) | BIT_ULL(2) | \
> > + BIT_ULL(3) | BIT_ULL(4) | BIT_ULL(5) | \
> > + BIT_ULL(6) | BIT_ULL(7) | BIT_ULL(8) | \
> > + BIT_ULL(10) | BIT_ULL(11) | BIT_ULL(13) | \
> > + BIT_ULL(14) | BIT_ULL(16) | BIT_ULL(17))
> > +#define IA32_VMX_ENTRY_CTLS_DEFAULT1 (BIT_ULL(0) | BIT_ULL(1) | BIT_ULL(2) | \
> > + BIT_ULL(3) | BIT_ULL(4) | BIT_ULL(5) | \
> > + BIT_ULL(6) | BIT_ULL(7) | BIT_ULL(8) | \
> > + BIT_ULL(12))
> > + /*
> > + * Make sure non-true MSRs are equal to the true MSRs with all the
> > + * default1 bits set.
> > + */
> > + switch (msr) {
> > + case MSR_IA32_VMX_TRUE_PINBASED_CTLS:
> > + non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_PINBASED_CTLS);
> > + TEST_ASSERT(non_true ==
> > + (data | IA32_VMX_PINBASED_CTLS_DEFAULT1),
> > + "Incorrect IA32_VMX_PINBASED_CTLS MSR: %lx\n",
> > + non_true);
> > + break;
> > + case MSR_IA32_VMX_TRUE_PROCBASED_CTLS:
> > + non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_PROCBASED_CTLS);
> > + TEST_ASSERT(non_true ==
> > + (data | IA32_VMX_PROCBASED_CTLS_DEFAULT1),
> > + "Incorrect IA32_VMX_PROCBASED_CTLS MSR: %lx\n",
> > + non_true);
> > + break;
> > + case MSR_IA32_VMX_TRUE_EXIT_CTLS:
> > + non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_EXIT_CTLS);
> > + TEST_ASSERT(non_true == (data | IA32_VMX_EXIT_CTLS_DEFAULT1),
> > + "Incorrect IA32_VMX_EXIT_CTLS MSR: %lx\n",
> > + non_true);
> > + break;
> > + case MSR_IA32_VMX_TRUE_ENTRY_CTLS:
> > + non_true = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_ENTRY_CTLS);
> > + TEST_ASSERT(non_true == (data | IA32_VMX_ENTRY_CTLS_DEFAULT1),
> > + "Incorrect IA32_VMX_ENTRY_CTLS MSR: %lx\n",
> > + non_true);
> > + break;
> > + }
> > +}
> > +
> > +/*
> > + * Reset the VCPU and return its file descriptor.
> > + *
> > + * The VCPU should be reset in between subsequent restores of the
> > + * same VMX MSR.
> > + */
> > +int reset_vcpu_fd(bool vmx_supported)
> > +{
> > + struct kvm_cpuid2 *cpuid;
> > + int fd;
> > +
> > + if (vm)
> > + kvm_vm_free(vm);
> > +
> > + vm = vm_create_default(VCPU_ID, 0, guest_code);
> > + fd = vcpu_fd(vm, VCPU_ID);
> > +
> > + /* Enable/Disable VMX */
> > + if (vmx_supported)
> > + kvm_get_supported_cpuid_entry(0x1)->ecx |= (1 << 5);
> > + else
> > + kvm_get_supported_cpuid_entry(0x1)->ecx &= ~(1 << 5);
> > + cpuid = kvm_get_supported_cpuid();
> > + vcpu_set_cpuid(vm, VCPU_ID, cpuid);
> > +
> > + return fd;
> > +}
> > +
> > +void test_msr_bits(uint32_t index, int high, int low, bool allowed_1)
> > +{
> > + int fd = reset_vcpu_fd(true);
> > + uint64_t value = kvm_get_msr(fd, index);
> > + int i;
> > +
> > + for (i = low; i <= high; i++) {
> > + uint64_t mask = 1ull << i;
> > + bool is_1 = value & mask;
> > +
> > + fd = reset_vcpu_fd(true);
> > +
> > + /*
> > + * Following VMX nomenclature:
> > + *
> > + * allowed-1 means a bit is high when it can be 0 or 1. The
> > + * bit is low when it must be 0.
> > + *
> > + * allowed-0 (!allowed_1) means a bit is low when it can be
> > + * 0 or 1. The bit is high when it must be 1.
> > + */
> > + if (allowed_1 && is_1)
> > + test_set_vmx_msr(fd, index, value & ~mask);
> > + else if (!allowed_1 && !is_1)
> > + test_set_vmx_msr(fd, index, value | mask);
> > + else if (allowed_1 && !is_1)
> > + kvm_set_msr_expect_fail(fd, index, value | mask);
> > + else
> > + kvm_set_msr_expect_fail(fd, index, value & ~mask);
> > + }
> > +}
> > +
> > +void test_allowed_1_bits(uint32_t index, int high, int low)
> > +{
> > + test_msr_bits(index, high, low, true);
> > +}
> > +#define test_allowed_1_bit(_index, _bit) \
> > + test_allowed_1_bits(_index, _bit, _bit)
> > +
> > +void test_allowed_0_bits(uint32_t index, int high, int low)
> > +{
> > + test_msr_bits(index, high, low, false);
> > +}
> > +
> > +uint64_t get_bits(uint64_t value, int high, int low)
> > +{
> > + return (value & GENMASK_ULL(high, low)) >> low;
> > +}
> > +
> > +uint64_t set_bits(uint64_t value, int high, int low, uint64_t new_value)
> > +{
> > + uint64_t mask = GENMASK_ULL(high, low);
> > +
> > + return (value & ~mask) | ((new_value << low) & mask);
> > +}
> > +
> > +/*
> > + * Test MSR[high:low].
> > + *
> > + * lt : KVM should support values less than the default.
> > + * gt : KVM should support values greater than the default.
> > + */
> > +void __test_msr_bits_eq(uint32_t index, int high, int low, bool lt, bool gt)
> > +{
> > + int fd = reset_vcpu_fd(true);
> > + uint64_t value = kvm_get_msr(fd, index);
> > + uint64_t bits = get_bits(value, high, low);
> > + uint64_t new_value;
> > +
> > + if (bits > 0) {
> > + fd = reset_vcpu_fd(true);
> > + new_value = set_bits(value, high, low, bits - 1);
> > + if (lt)
> > + test_set_vmx_msr(fd, index, new_value);
> > + else
> > + kvm_set_msr_expect_fail(fd, index, new_value);
> > + }
> > +
> > + if (bits < get_bits(-1ull, high, low)) {
> > + fd = reset_vcpu_fd(true);
> > + new_value = set_bits(value, high, low, bits + 1);
> > + if (gt)
> > + test_set_vmx_msr(fd, index, new_value);
> > + else
> > + kvm_set_msr_expect_fail(fd, index, new_value);
> > + }
> > +}
> > +
> > +#define test_msr_bits_eq(...) __test_msr_bits_eq(__VA_ARGS__, false, false)
> > +#define test_msr_bits_le(...) __test_msr_bits_eq(__VA_ARGS__, true, false)
> > +
> > +void test_rw_vmx_capability_msrs(void)
> > +{
> > + int i;
> > +
> > + printf("VMX MSR default values:\n");
> > +
> > + for (i = 0; i < ARRAY_SIZE(vmx_msrs); i++) {
> > + uint32_t msr = vmx_msrs[i].index;
> > + uint64_t guest, kvm;
> > + int vcpu_fd;
> > +
> > + /*
> > + * If VMX is not supported, test that reading the VMX MSRs
> > + * fails.
> > + */
> > + vcpu_fd = reset_vcpu_fd(false);
> > + kvm_get_msr_expect_fail(vcpu_fd, msr);
> > +
> > + vcpu_fd = reset_vcpu_fd(true);
> > + guest = read_msr_from_guest(msr);
> > + kvm = kvm_get_msr(vcpu_fd, msr);
> > +
> > + TEST_ASSERT(guest == kvm,
> > + "MSR %x mismatch: guest %lx kvm %lx.",
> > + msr, guest, kvm);
> > +
> > + printf("{ %s, 0x%lx },\n", vmx_msrs[i].name, guest);
> > +
> > + /* VM Execution Control Capability MSRs */
> > + switch (msr) {
> > + case MSR_IA32_VMX_BASIC:
> > + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> > + test_msr_bits_eq(msr, 30, 0); /* vmcs revision */
> > + test_allowed_1_bit(msr, 31); /* reserved */
> > + test_allowed_1_bit(msr, 48); /* VMCS phys bits */
> > + test_allowed_1_bit(msr, 49); /* dual-monitor SMI/SMM */
> > + test_allowed_1_bit(msr, 54); /* ins/outs info */
> > + test_allowed_1_bit(msr, 55); /* true ctls */
> > + break;
> > + case MSR_IA32_VMX_MISC:
> > + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> > + test_msr_bits_eq(msr, 4, 0); /* preempt timer rate */
> > + test_allowed_1_bit(msr, 5); /* store IA-32e mode */
> > + test_allowed_1_bits(msr, 8, 6); /* activity states */
> > + test_allowed_1_bits(msr, 13, 9); /* reserved */
> > + test_allowed_1_bit(msr, 14); /* pt in vmx mode */
> > + test_allowed_1_bit(msr, 15); /* smbase msr in smm */
> > + test_msr_bits_le(msr, 24, 16); /* CR3-target values */
> > + test_msr_bits_le(msr, 27, 25); /* MSR-store list size */
> > + test_allowed_1_bit(msr, 28); /* smm_monitor_ctl 1 */
> > + test_allowed_1_bit(msr, 29); /* vmwrite to any */
> > + test_allowed_1_bit(msr, 30); /* insn len 0 allowed */
> > + test_allowed_1_bit(msr, 31); /* reserved */
> > + test_msr_bits_eq(msr, 63, 32); /* MSEG Revision */
> > + break;
> > + case MSR_IA32_VMX_TRUE_PINBASED_CTLS:
> > + case MSR_IA32_VMX_TRUE_PROCBASED_CTLS:
> > + case MSR_IA32_VMX_TRUE_EXIT_CTLS:
> > + case MSR_IA32_VMX_TRUE_ENTRY_CTLS:
> > + case MSR_IA32_VMX_PROCBASED_CTLS2:
> > + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> > + test_allowed_0_bits(msr, 31, 0); /* allowed-1 bits */
> > + test_allowed_1_bits(msr, 63, 32); /* allowed-0 bits */
> > + break;
> > + case MSR_IA32_VMX_EPT_VPID_CAP:
> > + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> > + test_allowed_1_bits(msr, 63, 0); /* feature/reserved */
> > + break;
> > + case MSR_IA32_VMX_CR0_FIXED0:
> > + case MSR_IA32_VMX_CR4_FIXED0:
> > + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> > + test_allowed_0_bits(msr, 63, 0); /* allowed-0 bits */
> > + break;
> > + case MSR_IA32_VMX_CR0_FIXED1:
> > + case MSR_IA32_VMX_CR4_FIXED1:
> > + /* These MSRs do not allow save-restore. */
> > + kvm_set_msr_expect_fail(vcpu_fd, msr, guest);
> > + break;
> > + case MSR_IA32_VMX_PROCBASED_CTLS:
> > + case MSR_IA32_VMX_PINBASED_CTLS:
> > + case MSR_IA32_VMX_EXIT_CTLS:
> > + case MSR_IA32_VMX_ENTRY_CTLS:
> > + test_set_vmx_msr(vcpu_fd, msr, guest); /* identity */
> > + break;
> > + case MSR_IA32_VMX_VMCS_ENUM:
> > + test_allowed_1_bit(msr, 0); /* reserved */
> > + test_allowed_1_bits(msr, 63, 10); /* reserved */
> > + break;
> > + default:
> > + TEST_ASSERT(false, "Untested MSR: 0x%x.", msr);
> > + }
> > + }
> > +}
> > +
> > +struct msr {
> > + uint32_t index;
> > + uint64_t data;
> > +};
> > +
> > +/*
> > + * Test that MSR_IA32_VMX_CR4_FIXED1 is correctly updated in response to
> > + * CPUID changes.
> > + */
> > +void test_cr4_fixed1_cpuid(void)
> > +{
> > + struct test {
> > + uint64_t cr4_mask;
> > + int function;
> > + enum x86_register reg;
> > + int cpuid_bit;
> > + const char *name;
> > + } tests[] = {
> > + { X86_CR4_VME, 1, RDX, 1, "VME" },
> > + { X86_CR4_PVI, 1, RDX, 1, "PVI" },
> > + { X86_CR4_TSD, 1, RDX, 4, "TSD" },
> > + { X86_CR4_DE, 1, RDX, 2, "DE" },
> > + { X86_CR4_PSE, 1, RDX, 3, "PSE" },
> > + { X86_CR4_PAE, 1, RDX, 6, "PAE" },
> > + { X86_CR4_MCE, 1, RDX, 7, "MCE" },
> > + { X86_CR4_PGE, 1, RDX, 13, "PGE" },
> > + { X86_CR4_OSFXSR, 1, RDX, 24, "OSFXSR" },
> > + { X86_CR4_OSXMMEXCPT, 1, RDX, 25, "OSXMMEXCPT" },
> > + { X86_CR4_UMIP, 7, RCX, 2, "UMIP" },
> > + { X86_CR4_VMXE, 1, RCX, 5, "VMXE" },
> > + { X86_CR4_SMXE, 1, RCX, 6, "SMXE" },
> > + { X86_CR4_FSGSBASE, 7, RBX, 0, "FSGSBASE" },
> > + { X86_CR4_PCIDE, 1, RCX, 17, "PCIDE" },
> > + { X86_CR4_OSXSAVE, 1, RCX, 26, "OSXSAVE" },
> > + { X86_CR4_SMEP, 7, RBX, 7, "SMEP" },
> > + { X86_CR4_SMAP, 7, RBX, 20, "SMAP" },
> > + { X86_CR4_PKE, 7, RCX, 3, "PKE" },
> > + };
> > + int vcpu_fd = reset_vcpu_fd(true);
> > + int i;
> > +
> > + kvm_get_supported_cpuid_entry(0x1)->ecx |= (1 << 5);
> > +
> > + for (i = 0; i < ARRAY_SIZE(tests); i++) {
> > + struct test *test = &tests[i];
> > + struct kvm_cpuid_entry2 *entry;
> > + struct kvm_cpuid2 *cpuid = kvm_get_supported_cpuid();
> > + const char *reg_name = NULL;
> > + uint64_t fixed1;
> > + uint32_t *reg = NULL;
> > +
> > + entry = kvm_get_supported_cpuid_entry(test->function);
> > + TEST_ASSERT(entry, "CPUID lookup failed: CPUID.%02xH",
> > + test->function);
> > +
> > + switch (test->reg) {
> > + case RAX:
> > + reg = &entry->eax; reg_name = "EAX";
> > + break;
> > + case RBX:
> > + reg = &entry->ebx; reg_name = "EBX";
> > + break;
> > + case RCX:
> > + reg = &entry->ecx; reg_name = "ECX";
> > + break;
> > + case RDX:
> > + reg = &entry->edx; reg_name = "EDX";
> > + break;
> > + default:
> > + TEST_ASSERT(false, "Unrecognized reg: %d", test->reg);
> > + }
> > +
> > + printf("MSR_IA32_VMX_CR4_FIXED1[%d] (%s) == CPUID.%02xH:%s[%d]\n",
> > + __builtin_ffsll(test->cr4_mask) - 1, test->name,
> > + test->function, reg_name, test->cpuid_bit);
> > +
> > +
> > + *reg |= 1ull << test->cpuid_bit;
> > + vcpu_set_cpuid(vm, VCPU_ID, cpuid);
> > + fixed1 = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_CR4_FIXED1);
> > + TEST_ASSERT(fixed1 & test->cr4_mask, "set bit");
> > + TEST_ASSERT(fixed1 & X86_CR4_PCE, "PCE");
> > +
> > + /*
> > + * Skipping clearing support for VMX. VMX is required for
> > + * MSR_IA32_VMX_CR4_FIXED1 to even exist.
> > + */
> > + if (test->cr4_mask == X86_CR4_VMXE)
> > + continue;
> > +
> > + *reg &= ~(1ull << test->cpuid_bit);
> > + vcpu_set_cpuid(vm, VCPU_ID, cpuid);
> > + fixed1 = kvm_get_msr(vcpu_fd, MSR_IA32_VMX_CR4_FIXED1);
> > + TEST_ASSERT(!(fixed1 & test->cr4_mask), "clear bit");
> > + TEST_ASSERT(fixed1 & X86_CR4_PCE, "PCE");
> > + };
> > +}
> > +
> > +void test_preemption_timer_capability(void)
> > +{
> > + int fd = reset_vcpu_fd(true);
> > + uint64_t vmx_misc = kvm_get_msr(fd, MSR_IA32_VMX_MISC);
> > + uint64_t pinbased = kvm_get_msr(fd, MSR_IA32_VMX_TRUE_PINBASED_CTLS);
> > +
> > +#define VMX_PREEMPTION_TIMER_CONTROL \
> > + ((uint64_t) PIN_BASED_VMX_PREEMPTION_TIMER << 32)
> > +#define VMX_PREEMPTION_TIMER_RATE 0x1f
> > +
> > + TEST_ASSERT(pinbased & VMX_PREEMPTION_TIMER_CONTROL,
> > + "Expected support for VMX preemption timer.");
> > +
> > + /* Test that we can disable the VMX preemption timer. */
> > + test_set_vmx_msr(fd, MSR_IA32_VMX_TRUE_PINBASED_CTLS,
> > + pinbased & ~VMX_PREEMPTION_TIMER_CONTROL);
> > +
> > + /*
> > + * Test that we can clear the VMX preemption timer rate when
> > + * the preemption timer is disabled.
> > + */
> > + test_set_vmx_msr(fd, MSR_IA32_VMX_MISC,
> > + vmx_misc & ~VMX_PREEMPTION_TIMER_RATE);
> > +}
> > +
> > +int main(int argc, char *argv[])
> > +{
> > + test_rw_vmx_capability_msrs();
> > + test_cr4_fixed1_cpuid();
> > + test_preemption_timer_capability();
> > + return 0;
> > +}
> > --
> > 2.22.0.rc1.311.g5d7573a151-goog
> >
>
> ping
ping
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2019-07-10 16:10 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-05-31 18:42 [PATCH 2/2] tests: kvm: Test virtualization of VMX capability MSRs Aaron Lewis
2019-06-18 14:14 ` Aaron Lewis
2019-07-10 16:09 ` Aaron Lewis
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).