Re: [PATCH 3/3] selftests/seccomp: Check ENOSYS under tracing

[Kees Cook]

On Sat, Jul 04, 2020 at 11:12:32PM -0700, Kees Cook wrote:
> There should be no difference between -1 and other negative syscalls
> while tracing.
> 
> Cc: Andy Lutomirski <luto@xxxxxxxxxxxxxx>
> Cc: Will Drewry <wad@xxxxxxxxxxxx>
> Cc: Will Deacon <will@xxxxxxxxxx>
> Cc: Keno Fischer <keno@xxxxxxxxxxxxxxxxxx>
> Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
> ---
>  tools/testing/selftests/seccomp/seccomp_bpf.c | 26 +++++++++++++++++++
>  1 file changed, 26 insertions(+)
> 
> diff --git a/tools/testing/selftests/seccomp/seccomp_bpf.c b/tools/testing/selftests/seccomp/seccomp_bpf.c
> index 966dec340ea8..bf6aa06c435c 100644
> --- a/tools/testing/selftests/seccomp/seccomp_bpf.c
> +++ b/tools/testing/selftests/seccomp/seccomp_bpf.c
> @@ -1973,6 +1973,32 @@ FIXTURE_TEARDOWN(TRACE_syscall)
>  	teardown_trace_fixture(_metadata, self->tracer);
>  }
>  
> +TEST(negative_ENOSYS)
> +{
> +	/* Untraced negative syscalls should return ENOSYS. */
> +	errno = 0;
> +	EXPECT_EQ(-1, syscall(-1));
> +	EXPECT_EQ(errno, ENOSYS);
> +	errno = 0;
> +	EXPECT_EQ(-1, syscall(-101));
> +	EXPECT_EQ(errno, ENOSYS);
> +}
> +
> +TEST_F(TRACE_syscall, negative_ENOSYS)
> +{
> +	/*
> +	 * There should be no difference between an "internal" skip
> +	 * and userspace asking for syscall "-1".
> +	 */
> +	errno = 0;
> +	EXPECT_EQ(-1, syscall(-1));
> +	EXPECT_EQ(errno, ENOSYS);
> +	/* And no difference for "still not valid but not -1". */
> +	errno = 0;
> +	EXPECT_EQ(-1, syscall(-101));
> +	EXPECT_EQ(errno, ENOSYS);
> +}
> +

I realized after sending this that the second function could just be:

+TEST_F(TRACE_syscall, negative_ENOSYS)
+{
+	negative_ENOSYS(_metadata);
+}

:)

>  TEST_F(TRACE_syscall, syscall_allowed)
>  {
>  	/* getppid works as expected (no changes). */
> -- 
> 2.25.1
> 

-- 
Kees Cook