[igt-dev] [PATCH i-g-t] tests/i915/kms_big_fb: Move intel_ibb_destroy out of test_cleanup

[Zbigniew Kempczyński]

On Mon, Jun 13, 2022 at 11:54:47AM +0530, Karthik B S wrote:
> All the clean up functions were moved to fixture in patch "2aff4179".
> But this causes an assert in intel_bb_destory() after the test passes,
> on configurations having eDP connected.
> 
> So moving the intel_bb_destroy() call back to the original function.

Yes, calling intel_bb_destroy() in fixture is problematic. When test completes
we clean the allocator and initialize it from scratch for next test. 

With intel-bb and softpin during intel_bb_create() we open allocator (get ahnd)
which is not valid after test completes. So intel_bb_destroy() in fixture 
will try to free all offsets for attached intel_bufs. Unfortunately allocator
was already reinitialized and there's no ahnd in it.

Acked-by: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>

--
Zbigniew

> 
> Fixes: 2aff41793e5f ("tests/i915/kms_big_fb: Move cleanup code to fixture")
> 
> Signed-off-by: Karthik B S <karthik.b.s at intel.com>
> ---
>  tests/i915/kms_big_fb.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/tests/i915/kms_big_fb.c b/tests/i915/kms_big_fb.c
> index abf974bb..d50fde45 100644
> --- a/tests/i915/kms_big_fb.c
> +++ b/tests/i915/kms_big_fb.c
> @@ -457,6 +457,8 @@ static bool test_pipe(data_t *data)
>  	if (data->format == DRM_FORMAT_C8)
>  		unset_lut(data);
>  
> +	intel_bb_destroy(data->ibb);
> +
>  	return ret;
>  }
>  
> @@ -549,6 +551,8 @@ max_hw_stride_async_flip_test(data_t *data)
>  	}
>  	igt_reset_timeout();
>  
> +	intel_bb_destroy(data->ibb);
> +
>  	return true;
>  }
>  
> @@ -763,7 +767,6 @@ static void test_cleanup(data_t *data)
>  	igt_remove_fb(data->drm_fd, &data->big_fb_flip[1]);
>  	igt_remove_fb(data->drm_fd, &data->small_fb);
>  
> -	intel_bb_destroy(data->ibb);
>  	data->output = NULL;
>  }
>  
> -- 
> 2.22.0
>