[PATCH v8 0/4] tests/kms_async_flips: Create subtest for overlay planes

Melissa Wen mwen at igalia.com
Wed Apr 9 23:52:51 UTC 2025



On 09/04/2025 12:22, André Almeida wrote:
> This patchset creates a subtest for overlay planes. This is supported by amdgpu
> and is merged in the kernel:
>
> https://lore.kernel.org/lkml/173948734065.719858.7405160715916126757.b4-ty@linaro.org/
Hi,

I just waited a while for CI results and they look good.
So I applied this series to master.

Thanks!

Melissa

>
> v8:
>   - Make "Async page flip" name consistent in patch 1
>   - Clarify documentation about asserting with a bad fd in igt_has_drm_cap()
>
> v7:
>   - igt_has_drm_cap() asserts only if it was a bad fd. The rest is up to the
>   caller to assert or require.
>   - we now have different errors for when the cap is not supported by the driver,
>   and when the cap doesn't exist in the kernel
>
> v6:
>   - Added missing `test_init_ops()` for "overlay-atomic" test
>   - Reworked `test_init_fbs()`: it destroys all buffers when the modifier
>   changes, and recreate then if they are need for the next test. It also destroys
>   overlay buffers if they are not needed anymore
>
> v5:
>   - Added a new patch "lib/ioctl_wrappers: let the caller handle capability check
>     result" by Melissa Wen
>   - This patch solved an issue that the test was exiting instead of skipping when
>     the driver did not support ATOMIC_ASYNC_PAGE_FLIP
>   - test_finish() is now called test_init_ops() and is called in the beginning of
>     every test. This was done because some asserts prevented the test to reach
>     the final of the run_test() function.
>   - Fixed a typo to correctly set the overlay fb to the overlay plane
>   - Added code to remove the overlay fb by the end of the test
>   - Fixed a bug that test_init_fbs() was skipping the fb creation for overlay
>     planes
>   - Fixed require_overlay_flip_support() by changing the fb to be flipped with
>     the async flip
>
> v4:
>   - Created a patch to check for DRM_CAP_ATOMIC_ASYNC_PAGE_FLIP, so drivers
>   without support for it, skips instead of failing (Melissa)
>   - Encapsulated all test paths options into a function to easy maintenance (Melissa)
>   - Brought back test description for GitLab compilation
>
> v3:
>   - Fixed a bug that was mixing `alternate_sync_async = true` with the overlay path
>   - Refactored how the test parameters are being set/unset (patch 1/2)
>   - Fixed a bug where a primary buffer was being used with an overlay plane
>
> v2: Add test description for GitLab compilation
>
> André Almeida (4):
>    lib/ioctl_wrappers: let the caller handle capability check result
>    tests/kms_async_flips: Check for atomic async flip cap
>    kms_async_flips: Refactor data options
>    tests/kms_async_flips: Create subtest for overlay planes
>
>   lib/ioctl_wrappers.c    |  21 ++++---
>   lib/ioctl_wrappers.h    |   2 +-
>   tests/kms_async_flips.c | 131 ++++++++++++++++++++++++++++++++++------
>   3 files changed, 128 insertions(+), 26 deletions(-)
>



More information about the igt-dev mailing list