Re: [igt-dev] [PATCH i-g-t v3] tests/kms_concurrent: Move bandwidth calculation to igt_fixture

["Kahola, Mika" <mika.kahola@intel.com>]

-----Original Message-----
From: Latvala, Petri <petri.latvala@intel.com> 
Sent: Tuesday, March 31, 2020 9:45 AM
To: Kahola, Mika <mika.kahola@intel.com>
Cc: igt-dev@lists.freedesktop.org; juhapekka.heikkila@gmail.com; Lisovskiy, Stanislav <stanislav.lisovskiy@intel.com>
Subject: Re: [PATCH i-g-t v3] tests/kms_concurrent: Move bandwidth calculation to igt_fixture

On Mon, Mar 30, 2020 at 02:02:29PM +0300, Mika Kahola wrote:
> The commit 153b34b5353df8c18a87d ("tests/kms_concurrent:
> Test maximum number of planes supported by the platform") caused 
> regression on HSW pipe B testing such as.
> 
> IGT-Version: 1.25-gfd8248084 (x86_64) (Linux: 
> 5.6.0-rc7-CI-CI_DRM_8194+ x86_64) Starting subtest: pipe-B Testing 
> resolution with connector VGA-1 using pipe B with seed 1585245074 
> child 0 died with signal 11, Segmentation fault Subtest pipe-B: FAIL 
> (0.330s)
> 
> To fix this, we need move bandwidth calculation routines into part of 
> igt_fixture instead of calculating it just before actual testing. The 
> patch takes the minimum of those maximum number of planes for given output.
> 
> v2: Limit bandwidth check only gen11+ (CI)
> v3: Loop child process 5x longer when running test with 1 iteration 
> (CI)
> 
> Signed-off-by: Mika Kahola <mika.kahola@intel.com>
> ---
>  tests/kms_concurrent.c | 30 ++++++++++++++++--------------
>  1 file changed, 16 insertions(+), 14 deletions(-)
> 
> diff --git a/tests/kms_concurrent.c b/tests/kms_concurrent.c index 
> 1403e990..31c5620a 100644
> --- a/tests/kms_concurrent.c
> +++ b/tests/kms_concurrent.c
> @@ -36,6 +36,7 @@ typedef struct {
>  	igt_display_t display;
>  	igt_plane_t **plane;
>  	struct igt_fb *fb;
> +	int max_planes;
>  } data_t;
>  
>  /* Command line parameters. */
> @@ -223,13 +224,13 @@ test_plane_position_with_output(data_t *data, enum pipe pipe, int max_planes,
>  				igt_output_t *output)
>  {
>  	int i;
> -	int iterations = opt.iterations < 1 ? 1 : opt.iterations;
> +	int iterations = opt.iterations < 5 ? 1 : opt.iterations;

This changes nothing though for CI. You still use 1. In fact this just limits the iterations to 1 for everyone who specifies more than 1, until they use 5 or more.

That's a typo.

And still, this patch needs an actual explanation why the crash happened and why doing the same calculation in a fixture help. Otherwise there's no guarantee it won't happen again. Until proven otherwise, avoiding the crash is accidental.

CI was clean this time around even though nothing has changed.

--
Petri Latvala
_______________________________________________
igt-dev mailing list
igt-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/igt-dev