Re: [Intel-gfx] [PATCH i-g-t] [RFC] tests/drm_read: Fix subtest invalid-buffer

[Petri Latvala <petri.latvala@intel.com>]

On Fri, May 28, 2021 at 10:02:47AM +0530, Vidya Srinivas wrote:
> Using (void *)-1 directly in read is aborting on chrome systems.
> Following message is seen.
> 
> Starting subtest: invalid-buffer
> *** buffer overflow detected ***: terminated
> Received signal SIGABRT.
> Stack trace:
> Aborted (core dumped)
> 
> Patch just adds a pointer variable and uses it in read.
> 
> Signed-off-by: Vidya Srinivas <vidya.srinivas@intel.com>
> ---
>  tests/drm_read.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/tests/drm_read.c b/tests/drm_read.c
> index ccf9d822fd8d..2fdec5be4078 100644
> --- a/tests/drm_read.c
> +++ b/tests/drm_read.c
> @@ -103,10 +103,11 @@ static void teardown(int fd)
>  static void test_invalid_buffer(int in)
>  {
>  	int fd = setup(in, 0);
> +	void *add = (void *)-1;
>  
>  	alarm(1);
>  
> -	igt_assert_eq(read(fd, (void *)-1, 4096), -1);
> +	igt_assert_eq(read(fd, add, 4096), -1);
>  	igt_assert_eq(errno, EFAULT);
>  
>  	teardown(fd);

This looked weird but then I checked what glibc is actually
doing. This is FORTIFY_SOURCE in action, and read() checks the buffer
with __builtin_object_size() that it has room for the read. Which it
can only do here if the address is a literal.

Reviewed-by: Petri Latvala <petri.latvala@intel.com>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx