selftests: KVM: aarch64: Introduce hypercall ABI test
authorRaghavendra Rao Ananta <rananta@google.com>
Mon, 2 May 2022 23:38:52 +0000 (23:38 +0000)
committerMarc Zyngier <maz@kernel.org>
Tue, 3 May 2022 20:30:20 +0000 (21:30 +0100)
Introduce a KVM selftest to check the hypercall interface
for arm64 platforms. The test validates the user-space'
[GET|SET]_ONE_REG interface to read/write the psuedo-firmware
registers as well as its effects on the guest upon certain
configurations.

Signed-off-by: Raghavendra Rao Ananta <rananta@google.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220502233853.1233742-9-rananta@google.com
tools/testing/selftests/kvm/.gitignore
tools/testing/selftests/kvm/Makefile
tools/testing/selftests/kvm/aarch64/hypercalls.c [new file with mode: 0644]

index 1bb575d..b17e464 100644 (file)
@@ -2,6 +2,7 @@
 /aarch64/arch_timer
 /aarch64/debug-exceptions
 /aarch64/get-reg-list
+/aarch64/hypercalls
 /aarch64/psci_test
 /aarch64/vcpu_width_config
 /aarch64/vgic_init
index c2cf4d3..97eef0c 100644 (file)
@@ -105,6 +105,7 @@ TEST_GEN_PROGS_x86_64 += system_counter_offset_test
 TEST_GEN_PROGS_aarch64 += aarch64/arch_timer
 TEST_GEN_PROGS_aarch64 += aarch64/debug-exceptions
 TEST_GEN_PROGS_aarch64 += aarch64/get-reg-list
+TEST_GEN_PROGS_aarch64 += aarch64/hypercalls
 TEST_GEN_PROGS_aarch64 += aarch64/psci_test
 TEST_GEN_PROGS_aarch64 += aarch64/vcpu_width_config
 TEST_GEN_PROGS_aarch64 += aarch64/vgic_init
diff --git a/tools/testing/selftests/kvm/aarch64/hypercalls.c b/tools/testing/selftests/kvm/aarch64/hypercalls.c
new file mode 100644 (file)
index 0000000..41e0210
--- /dev/null
@@ -0,0 +1,336 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+/* hypercalls: Check the ARM64's psuedo-firmware bitmap register interface.
+ *
+ * The test validates the basic hypercall functionalities that are exposed
+ * via the psuedo-firmware bitmap register. This includes the registers'
+ * read/write behavior before and after the VM has started, and if the
+ * hypercalls are properly masked or unmasked to the guest when disabled or
+ * enabled from the KVM userspace, respectively.
+ */
+
+#include <errno.h>
+#include <linux/arm-smccc.h>
+#include <asm/kvm.h>
+#include <kvm_util.h>
+
+#include "processor.h"
+
+#define FW_REG_ULIMIT_VAL(max_feat_bit) (GENMASK(max_feat_bit, 0))
+
+/* Last valid bits of the bitmapped firmware registers */
+#define KVM_REG_ARM_STD_BMAP_BIT_MAX           0
+#define KVM_REG_ARM_STD_HYP_BMAP_BIT_MAX       0
+#define KVM_REG_ARM_VENDOR_HYP_BMAP_BIT_MAX    1
+
+struct kvm_fw_reg_info {
+       uint64_t reg;           /* Register definition */
+       uint64_t max_feat_bit;  /* Bit that represents the upper limit of the feature-map */
+};
+
+#define FW_REG_INFO(r)                 \
+       {                                       \
+               .reg = r,                       \
+               .max_feat_bit = r##_BIT_MAX,    \
+       }
+
+static const struct kvm_fw_reg_info fw_reg_info[] = {
+       FW_REG_INFO(KVM_REG_ARM_STD_BMAP),
+       FW_REG_INFO(KVM_REG_ARM_STD_HYP_BMAP),
+       FW_REG_INFO(KVM_REG_ARM_VENDOR_HYP_BMAP),
+};
+
+enum test_stage {
+       TEST_STAGE_REG_IFACE,
+       TEST_STAGE_HVC_IFACE_FEAT_DISABLED,
+       TEST_STAGE_HVC_IFACE_FEAT_ENABLED,
+       TEST_STAGE_HVC_IFACE_FALSE_INFO,
+       TEST_STAGE_END,
+};
+
+static int stage = TEST_STAGE_REG_IFACE;
+
+struct test_hvc_info {
+       uint32_t func_id;
+       uint64_t arg1;
+};
+
+#define TEST_HVC_INFO(f, a1)   \
+       {                       \
+               .func_id = f,   \
+               .arg1 = a1,     \
+       }
+
+static const struct test_hvc_info hvc_info[] = {
+       /* KVM_REG_ARM_STD_BMAP */
+       TEST_HVC_INFO(ARM_SMCCC_TRNG_VERSION, 0),
+       TEST_HVC_INFO(ARM_SMCCC_TRNG_FEATURES, ARM_SMCCC_TRNG_RND64),
+       TEST_HVC_INFO(ARM_SMCCC_TRNG_GET_UUID, 0),
+       TEST_HVC_INFO(ARM_SMCCC_TRNG_RND32, 0),
+       TEST_HVC_INFO(ARM_SMCCC_TRNG_RND64, 0),
+
+       /* KVM_REG_ARM_STD_HYP_BMAP */
+       TEST_HVC_INFO(ARM_SMCCC_ARCH_FEATURES_FUNC_ID, ARM_SMCCC_HV_PV_TIME_FEATURES),
+       TEST_HVC_INFO(ARM_SMCCC_HV_PV_TIME_FEATURES, ARM_SMCCC_HV_PV_TIME_ST),
+       TEST_HVC_INFO(ARM_SMCCC_HV_PV_TIME_ST, 0),
+
+       /* KVM_REG_ARM_VENDOR_HYP_BMAP */
+       TEST_HVC_INFO(ARM_SMCCC_VENDOR_HYP_KVM_FEATURES_FUNC_ID,
+                       ARM_SMCCC_VENDOR_HYP_KVM_PTP_FUNC_ID),
+       TEST_HVC_INFO(ARM_SMCCC_VENDOR_HYP_CALL_UID_FUNC_ID, 0),
+       TEST_HVC_INFO(ARM_SMCCC_VENDOR_HYP_KVM_PTP_FUNC_ID, KVM_PTP_VIRT_COUNTER),
+};
+
+/* Feed false hypercall info to test the KVM behavior */
+static const struct test_hvc_info false_hvc_info[] = {
+       /* Feature support check against a different family of hypercalls */
+       TEST_HVC_INFO(ARM_SMCCC_TRNG_FEATURES, ARM_SMCCC_VENDOR_HYP_KVM_PTP_FUNC_ID),
+       TEST_HVC_INFO(ARM_SMCCC_ARCH_FEATURES_FUNC_ID, ARM_SMCCC_TRNG_RND64),
+       TEST_HVC_INFO(ARM_SMCCC_HV_PV_TIME_FEATURES, ARM_SMCCC_TRNG_RND64),
+};
+
+static void guest_test_hvc(const struct test_hvc_info *hc_info)
+{
+       unsigned int i;
+       struct arm_smccc_res res;
+       unsigned int hvc_info_arr_sz;
+
+       hvc_info_arr_sz =
+       hc_info == hvc_info ? ARRAY_SIZE(hvc_info) : ARRAY_SIZE(false_hvc_info);
+
+       for (i = 0; i < hvc_info_arr_sz; i++, hc_info++) {
+               memset(&res, 0, sizeof(res));
+               smccc_hvc(hc_info->func_id, hc_info->arg1, 0, 0, 0, 0, 0, 0, &res);
+
+               switch (stage) {
+               case TEST_STAGE_HVC_IFACE_FEAT_DISABLED:
+               case TEST_STAGE_HVC_IFACE_FALSE_INFO:
+                       GUEST_ASSERT_3(res.a0 == SMCCC_RET_NOT_SUPPORTED,
+                                       res.a0, hc_info->func_id, hc_info->arg1);
+                       break;
+               case TEST_STAGE_HVC_IFACE_FEAT_ENABLED:
+                       GUEST_ASSERT_3(res.a0 != SMCCC_RET_NOT_SUPPORTED,
+                                       res.a0, hc_info->func_id, hc_info->arg1);
+                       break;
+               default:
+                       GUEST_ASSERT_1(0, stage);
+               }
+       }
+}
+
+static void guest_code(void)
+{
+       while (stage != TEST_STAGE_END) {
+               switch (stage) {
+               case TEST_STAGE_REG_IFACE:
+                       break;
+               case TEST_STAGE_HVC_IFACE_FEAT_DISABLED:
+               case TEST_STAGE_HVC_IFACE_FEAT_ENABLED:
+                       guest_test_hvc(hvc_info);
+                       break;
+               case TEST_STAGE_HVC_IFACE_FALSE_INFO:
+                       guest_test_hvc(false_hvc_info);
+                       break;
+               default:
+                       GUEST_ASSERT_1(0, stage);
+               }
+
+               GUEST_SYNC(stage);
+       }
+
+       GUEST_DONE();
+}
+
+static int set_fw_reg(struct kvm_vm *vm, uint64_t id, uint64_t val)
+{
+       struct kvm_one_reg reg = {
+               .id = id,
+               .addr = (uint64_t)&val,
+       };
+
+       return _vcpu_ioctl(vm, 0, KVM_SET_ONE_REG, &reg);
+}
+
+static void get_fw_reg(struct kvm_vm *vm, uint64_t id, uint64_t *addr)
+{
+       struct kvm_one_reg reg = {
+               .id = id,
+               .addr = (uint64_t)addr,
+       };
+
+       vcpu_ioctl(vm, 0, KVM_GET_ONE_REG, &reg);
+}
+
+struct st_time {
+       uint32_t rev;
+       uint32_t attr;
+       uint64_t st_time;
+};
+
+#define STEAL_TIME_SIZE                ((sizeof(struct st_time) + 63) & ~63)
+#define ST_GPA_BASE            (1 << 30)
+
+static void steal_time_init(struct kvm_vm *vm)
+{
+       uint64_t st_ipa = (ulong)ST_GPA_BASE;
+       unsigned int gpages;
+       struct kvm_device_attr dev = {
+               .group = KVM_ARM_VCPU_PVTIME_CTRL,
+               .attr = KVM_ARM_VCPU_PVTIME_IPA,
+               .addr = (uint64_t)&st_ipa,
+       };
+
+       gpages = vm_calc_num_guest_pages(VM_MODE_DEFAULT, STEAL_TIME_SIZE);
+       vm_userspace_mem_region_add(vm, VM_MEM_SRC_ANONYMOUS, ST_GPA_BASE, 1, gpages, 0);
+
+       vcpu_ioctl(vm, 0, KVM_SET_DEVICE_ATTR, &dev);
+}
+
+static void test_fw_regs_before_vm_start(struct kvm_vm *vm)
+{
+       uint64_t val;
+       unsigned int i;
+       int ret;
+
+       for (i = 0; i < ARRAY_SIZE(fw_reg_info); i++) {
+               const struct kvm_fw_reg_info *reg_info = &fw_reg_info[i];
+
+               /* First 'read' should be an upper limit of the features supported */
+               get_fw_reg(vm, reg_info->reg, &val);
+               TEST_ASSERT(val == FW_REG_ULIMIT_VAL(reg_info->max_feat_bit),
+                       "Expected all the features to be set for reg: 0x%lx; expected: 0x%lx; read: 0x%lx\n",
+                       reg_info->reg, FW_REG_ULIMIT_VAL(reg_info->max_feat_bit), val);
+
+               /* Test a 'write' by disabling all the features of the register map */
+               ret = set_fw_reg(vm, reg_info->reg, 0);
+               TEST_ASSERT(ret == 0,
+                       "Failed to clear all the features of reg: 0x%lx; ret: %d\n",
+                       reg_info->reg, errno);
+
+               get_fw_reg(vm, reg_info->reg, &val);
+               TEST_ASSERT(val == 0,
+                       "Expected all the features to be cleared for reg: 0x%lx\n", reg_info->reg);
+
+               /*
+                * Test enabling a feature that's not supported.
+                * Avoid this check if all the bits are occupied.
+                */
+               if (reg_info->max_feat_bit < 63) {
+                       ret = set_fw_reg(vm, reg_info->reg, BIT(reg_info->max_feat_bit + 1));
+                       TEST_ASSERT(ret != 0 && errno == EINVAL,
+                       "Unexpected behavior or return value (%d) while setting an unsupported feature for reg: 0x%lx\n",
+                       errno, reg_info->reg);
+               }
+       }
+}
+
+static void test_fw_regs_after_vm_start(struct kvm_vm *vm)
+{
+       uint64_t val;
+       unsigned int i;
+       int ret;
+
+       for (i = 0; i < ARRAY_SIZE(fw_reg_info); i++) {
+               const struct kvm_fw_reg_info *reg_info = &fw_reg_info[i];
+
+               /*
+                * Before starting the VM, the test clears all the bits.
+                * Check if that's still the case.
+                */
+               get_fw_reg(vm, reg_info->reg, &val);
+               TEST_ASSERT(val == 0,
+                       "Expected all the features to be cleared for reg: 0x%lx\n",
+                       reg_info->reg);
+
+               /*
+                * Since the VM has run at least once, KVM shouldn't allow modification of
+                * the registers and should return EBUSY. Set the registers and check for
+                * the expected errno.
+                */
+               ret = set_fw_reg(vm, reg_info->reg, FW_REG_ULIMIT_VAL(reg_info->max_feat_bit));
+               TEST_ASSERT(ret != 0 && errno == EBUSY,
+               "Unexpected behavior or return value (%d) while setting a feature while VM is running for reg: 0x%lx\n",
+               errno, reg_info->reg);
+       }
+}
+
+static struct kvm_vm *test_vm_create(void)
+{
+       struct kvm_vm *vm;
+
+       vm = vm_create_default(0, 0, guest_code);
+
+       ucall_init(vm, NULL);
+       steal_time_init(vm);
+
+       return vm;
+}
+
+static struct kvm_vm *test_guest_stage(struct kvm_vm *vm)
+{
+       struct kvm_vm *ret_vm = vm;
+
+       pr_debug("Stage: %d\n", stage);
+
+       switch (stage) {
+       case TEST_STAGE_REG_IFACE:
+               test_fw_regs_after_vm_start(vm);
+               break;
+       case TEST_STAGE_HVC_IFACE_FEAT_DISABLED:
+               /* Start a new VM so that all the features are now enabled by default */
+               kvm_vm_free(vm);
+               ret_vm = test_vm_create();
+               break;
+       case TEST_STAGE_HVC_IFACE_FEAT_ENABLED:
+       case TEST_STAGE_HVC_IFACE_FALSE_INFO:
+               break;
+       default:
+               TEST_FAIL("Unknown test stage: %d\n", stage);
+       }
+
+       stage++;
+       sync_global_to_guest(vm, stage);
+
+       return ret_vm;
+}
+
+static void test_run(void)
+{
+       struct kvm_vm *vm;
+       struct ucall uc;
+       bool guest_done = false;
+
+       vm = test_vm_create();
+
+       test_fw_regs_before_vm_start(vm);
+
+       while (!guest_done) {
+               vcpu_run(vm, 0);
+
+               switch (get_ucall(vm, 0, &uc)) {
+               case UCALL_SYNC:
+                       vm = test_guest_stage(vm);
+                       break;
+               case UCALL_DONE:
+                       guest_done = true;
+                       break;
+               case UCALL_ABORT:
+                       TEST_FAIL("%s at %s:%ld\n\tvalues: 0x%lx, 0x%lx; 0x%lx, stage: %u",
+                       (const char *)uc.args[0], __FILE__, uc.args[1],
+                       uc.args[2], uc.args[3], uc.args[4], stage);
+                       break;
+               default:
+                       TEST_FAIL("Unexpected guest exit\n");
+               }
+       }
+
+       kvm_vm_free(vm);
+}
+
+int main(void)
+{
+       setbuf(stdout, NULL);
+
+       test_run();
+       return 0;
+}