On Thu, Feb 05, 2026 at 01:05:02AM +0000, Jiakai Xu wrote:
> Add RISC-V KVM selftests to verify the SBI Steal-Time Accounting (STA)
> shared memory alignment requirements.
>
> The SBI specification requires the STA shared memory GPA to be 64-byte
> aligned, or set to all-ones to explicitly disable steal-time accounting.
> This test verifies that KVM enforces the expected behavior when
> configuring the SBI STA shared memory via KVM_SET_ONE_REG.
>
> Specifically, the test checks that:
> - misaligned GPAs are rejected with -EINVAL
> - 64-byte aligned GPAs are accepted
> - INVALID_GPA correctly disables steal-time accounting
>
> Signed-off-by: Jiakai Xu <[email protected]>
> Signed-off-by: Jiakai Xu <[email protected]>
> ---
> .../selftests/kvm/include/riscv/processor.h | 4 +++
> tools/testing/selftests/kvm/steal_time.c | 33 +++++++++++++++++++
> 2 files changed, 37 insertions(+)
>
> diff --git a/tools/testing/selftests/kvm/include/riscv/processor.h
> b/tools/testing/selftests/kvm/include/riscv/processor.h
> index e58282488beb3..c3551d129d2f6 100644
> --- a/tools/testing/selftests/kvm/include/riscv/processor.h
> +++ b/tools/testing/selftests/kvm/include/riscv/processor.h
> @@ -62,6 +62,10 @@ static inline uint64_t __kvm_reg_id(uint64_t type,
> uint64_t subtype,
> KVM_REG_RISCV_SBI_SINGLE,
> \
> idx, KVM_REG_SIZE_ULONG)
>
> +#define RISCV_SBI_STA_REG(idx) __kvm_reg_id(KVM_REG_RISCV_SBI_STATE,
> \
> + KVM_REG_RISCV_SBI_STA,
> \
> + idx, KVM_REG_SIZE_ULONG)
We don't need this because...
> +
> bool __vcpu_has_ext(struct kvm_vcpu *vcpu, uint64_t ext);
>
> static inline bool __vcpu_has_isa_ext(struct kvm_vcpu *vcpu, uint64_t
> isa_ext)
> diff --git a/tools/testing/selftests/kvm/steal_time.c
> b/tools/testing/selftests/kvm/steal_time.c
> index 8edc1fca345ba..30b98d1b601c3 100644
> --- a/tools/testing/selftests/kvm/steal_time.c
> +++ b/tools/testing/selftests/kvm/steal_time.c
> @@ -209,6 +209,7 @@ static void steal_time_dump(struct kvm_vm *vm, uint32_t
> vcpu_idx)
>
> /* SBI STA shmem must have 64-byte alignment */
> #define STEAL_TIME_SIZE ((sizeof(struct sta_struct) + 63) & ~63)
> +#define INVALID_GPA (~(u64)0)
>
> static vm_paddr_t st_gpa[NR_VCPUS];
>
> @@ -301,6 +302,34 @@ static void steal_time_dump(struct kvm_vm *vm, uint32_t
> vcpu_idx)
> pr_info("\n");
> }
>
> +static void test_riscv_sta_shmem_alignment(struct kvm_vcpu *vcpu)
> +{
> + struct kvm_one_reg reg;
> + uint64_t shmem;
> + int ret;
> +
> + reg.id = RISCV_SBI_STA_REG(0);
...here we should use KVM_REG_RISCV_SBI_STA_REG(shmem_lo)
> + reg.addr = (uint64_t)&shmem;
> +
> + /* Case 1: misaligned GPA */
> + shmem = ST_GPA_BASE + 1;
> + ret = __vcpu_ioctl(vcpu, KVM_SET_ONE_REG, ®);
> + TEST_ASSERT(ret == -1 && errno == EINVAL,
> + "misaligned STA shmem should return -EINVAL");
remove the word 'should'
"misaligned STA shmem returns -EINVAL"
> +
> + /* Case 2: 64-byte aligned GPA */
> + shmem = ST_GPA_BASE;
> + ret = __vcpu_ioctl(vcpu, KVM_SET_ONE_REG, ®);
> + TEST_ASSERT(ret == 0,
> + "aligned STA shmem should succeed");
same comment about 'should'
> +
> + /* Case 3: INVALID_GPA disables STA */
> + shmem = INVALID_GPA;
> + ret = __vcpu_ioctl(vcpu, KVM_SET_ONE_REG, ®);
> + TEST_ASSERT(ret == 0,
> + "INVALID_GPA should disable STA successfully");
We're not testing that STA was successfully disabled, only that all-ones
input doesn't generate an error. So the message should be along the lines
of "all-ones for STA shmem succeeds"
> +}
> +
> #endif
>
> static void *do_steal_time(void *arg)
> @@ -369,6 +398,10 @@ int main(int ac, char **av)
> TEST_REQUIRE(is_steal_time_supported(vcpus[0]));
> ksft_set_plan(NR_VCPUS);
>
> +#ifdef __riscv
> + test_riscv_sta_shmem_alignment(vcpus[0]);
> +#endif
We like to try to avoid #ifdefs in common functions by providing stubs for
architectures that don't need them [yet]. So we should rename this to
something more generic, like
check_steal_time_uapi()
and then call it for all architectures. Actually the other architectures
can already make use of it since both x86 and arm64 do uapi tests in
their steal_time_init() functions and that's not really the right
place to do that. I suggest creating another patch that first moves those
tests into new functions (check_steal_time_uapi()) which only needs to
be called once for vcpu[0] outside the vcpu loop, as you do here. In
that patch check_steal_time_uapi() will be a stub for riscv. Then in
a second patch fill in that stub with the tests above.
Thanks,
drew
> +
> /* Run test on each VCPU */
> for (i = 0; i < NR_VCPUS; ++i) {
> steal_time_init(vcpus[i], i);
> --
> 2.34.1
>
