[igt-dev] [PATCH i-g-t] tests/kms_flip: Skip detiler fences subtest on platforms w/o aperture

[Souza, Jose]

On Thu, 2020-04-02 at 15:29 +0300, Imre Deak wrote:
> On platforms w/o an aperture subtests using fences are not relevant,
> so
> skip them. This also fixes a use-after-free error happening as a
> side-effect of running the test on such platforms.
> 
> Signed-off-by: Imre Deak <imre.deak at intel.com>
> ---
>  tests/kms_flip.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/tests/kms_flip.c b/tests/kms_flip.c
> index 9ce83c53d..506c9c142 100755
> --- a/tests/kms_flip.c
> +++ b/tests/kms_flip.c
> @@ -216,6 +216,7 @@ static void emit_fence_stress(struct test_output
> *o)
>  
>  	igt_require(bufmgr);
>  
> +	igt_assert(num_fences);
>  	bo = calloc(sizeof(*bo), num_fences);
>  	exec = calloc(sizeof(*exec), num_fences+1);
>  	for (i = 0; i < num_fences - 1; i++) {
> @@ -1234,8 +1235,10 @@ static void run_test_on_crtc_set(struct
> test_output *o, int *crtc_idxs,
>  		o->fb_width *= 2;
>  
>  	tiling = LOCAL_DRM_FORMAT_MOD_NONE;
> -	if (o->flags & TEST_FENCE_STRESS)
> +	if (o->flags & TEST_FENCE_STRESS) {
> +		igt_require(gem_available_fences(drm_fd));

Would be better to skip on the begining of run_test() so less CI time
is wasted in those tests, right after:
igt_require((flags & TEST_HANG) == 0 || !is_wedged(drm_fd));

>  		tiling = LOCAL_I915_FORMAT_MOD_X_TILED;
> +	}
>  
>  	/* 256 MB is usually the maximum mappable aperture,
>  	 * (make it 4x times that to ensure failure) */