Re: [PATCH i-g-t 00/12] Prepare for returning NULL from igt_crtc_for_pipe()

[Jani Nikula <jani.nikula@intel.com>]

On Mon, 16 Mar 2026, Ville Syrjälä <ville.syrjala@linux.intel.com> wrote:
> On Mon, Mar 16, 2026 at 01:19:02PM +0200, Jani Nikula wrote:
>> We want to return NULL for invalid pipes from igt_crtc_for_pipe(), so
>> that we don't need to have igt_crtc_t objects with crtc->valid == false
>> around.
>> 
>> The primary way to prepare for this is to avoid using
>> igt_crtc_for_pipe() altogether. We've done a *lot* of that already, and
>> continue to do so here.
>> 
>> For the remaining ones, start off by ensuring crtc->valid == true, one
>> way or other. This ensures the test logic is sane and actually operates
>> on valid CRTCs. Once this prep is done, we can actually return NULL from
>> igt_crtc_for_pipe(), and check for crtc != NULL instead of crtc->valid
>> == true.
>
> I was hoping to make igt_crtc_for_pipe() just skip if the CRTC
> doesn't exist. But maybe there are still too many places that
> assume they can ask for any pipe and not skip?

At the very least the joiner logic has to gracefully check if a pipe
exists or not, without impacting test execution status.

The alternative would be to have igt_crtc_for_pipe() skip, but introduce
igt_pipe_exists() or somesuch for the check. That might lead to some
places requiring both, though.

Partially I'm also interested in the test results here; are there tests
that have been using invalid CRTCs...


BR,
Jani.


>
>> 
>> BR,
>> Jani.
>> 
>> 
>> Jani Nikula (12):
>>   tests/amdgpu/amd_mode_switch: Prepare for returning NULL from
>>     igt_crtc_for_pipe()
>>   tests/amdgpu/amd_multidisplay_modeset: Prepare for returning NULL from
>>     igt_crtc_for_pipe()
>>   tests/intel/kms_cdclk: use igt_next_crtc()
>>   tests/intel/kms_joiner: Prepare for returning NULL from
>>     igt_crtc_for_pipe()
>>   tests/intel/kms_joiner_helper: Prepare for returning NULL from
>>     igt_crtc_for_pipe()
>>   tests/intel/kms_pipe_b_c_ivb: Prepare for returning NULL from
>>     igt_crtc_for_pipe()
>>   tests/intel/kms_pipe_stress: Prepare for returning NULL from
>>     igt_crtc_for_pipe()
>>   tests/kms_bw: fix misleading indent
>>   tests/kms_bw: clarify what's going on with proper variable naming
>>   tests/kms_bw: avoid igt_crtc_for_pipe() and crtc->pipe usage
>>   tests/kms_explicit_fence: store igt_crtc_t* instead of pipe in data
>>   lib/kms: convert igt_crtc_for_pipe() into a proper function
>> 
>>  lib/igt_kms.c                           |  8 +++
>>  lib/igt_kms.h                           |  8 +--
>>  tests/amdgpu/amd_mode_switch.c          |  2 +
>>  tests/amdgpu/amd_multidisplay_modeset.c |  2 +
>>  tests/intel/kms_cdclk.c                 | 10 ++--
>>  tests/intel/kms_joiner.c                | 66 ++++++++++++++++++-------
>>  tests/intel/kms_joiner_helper.c         | 16 +++---
>>  tests/intel/kms_pipe_b_c_ivb.c          | 28 +++++++----
>>  tests/intel/kms_pipe_stress.c           |  4 ++
>>  tests/kms_bw.c                          | 48 +++++++++---------
>>  tests/kms_explicit_fence.c              | 14 +++---
>>  11 files changed, 131 insertions(+), 75 deletions(-)
>> 
>> -- 
>> 2.47.3

-- 
Jani Nikula, Intel