[igt-dev] [PATCH v4] tests/kms_vblank: Turn on hardware before testing invalid vblank.

[Petri Latvala]

On Mon, Jun 21, 2021 at 06:32:57PM +0000, Mark Yacoub wrote:
> From: Mark Yacoub <markyacoub at google.com>
> 
> [Why]
> Before any hardware is on, the vblank is off and its ref counter is in
> an initialized state as each driver handles toggling it differently.
> Ioctl DRM_IOCTL_WAIT_VBLANK could return 0 such as on i915, or an invalid
> integer that doesn't mean much, such as on Zork running Kernel 5.4 due
> to the kernel workaround that increments the vblank ref count to prevent
> a get from enabling the interrupt.
> 
> [How]
> For invalid_subtest(), active the CRTCs to turn the hardware on so
> DRM_IOCTL_WAIT_VBLANK returns something meaningful.
> 
> Changes since v3:
> Turn on only 1 output instead of all.
> 
> Changes since v2:
> 1. Updated Signed-off-by.
> 
> Changes since v1:
> 1. Update Commit message
> 2. Rename variable p to pipe_number
> 
> Signed-off-by: Mark Yacoub <markyacoub at chromium.org>
> Change-Id: Id70072ea797a74d3e9563ef8feb5901daa6b511a
> ---
>  tests/kms_vblank.c | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
> 
> diff --git a/tests/kms_vblank.c b/tests/kms_vblank.c
> index 93b01eba..19284257 100644
> --- a/tests/kms_vblank.c
> +++ b/tests/kms_vblank.c
> @@ -475,8 +475,13 @@ static void invalid_subtest(data_t *data, int fd)
>  {
>  	union drm_wait_vblank vbl;
>  	unsigned long valid_flags;
> +	igt_display_t* display = &data->display;
> +	igt_output_t* output;
>  
> -	igt_display_require_output_on_pipe(&data->display, 0);
> +	igt_display_require_output_on_pipe(display, 0);
> +	igt_require(display->n_outputs);
> +	output = &display->outputs[0];
> +	prepare_crtc(data, fd, output);

prepare_crtc uses the pipe in data->pipe which is still left
uninitialized at this point.

And you can't be sure that the first output is valid for that
pipe. igt_get_single_output_for_pipe() can get you that.


-- 
Petri Latvala