Hi,
On Fri, Apr 25, 2025 at 01:27:05PM +0300, Cristian Ciocaltea wrote:
> In preparation to improve error handling throughout all test cases,
> introduce a macro to check for EDEADLK and automate the restart of the
> atomic sequence.
>
> Signed-off-by: Cristian Ciocaltea <cristian.ciocal...@collabora.com>
> ---
> drivers/gpu/drm/tests/drm_hdmi_state_helper_test.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/drivers/gpu/drm/tests/drm_hdmi_state_helper_test.c
> b/drivers/gpu/drm/tests/drm_hdmi_state_helper_test.c
> index
> c8969ee6518954ab4496d3a4398f428bf4104a36..c8bb131d63ea6d0c9e166c8d9ba5e403118cd9f1
> 100644
> --- a/drivers/gpu/drm/tests/drm_hdmi_state_helper_test.c
> +++ b/drivers/gpu/drm/tests/drm_hdmi_state_helper_test.c
> @@ -224,6 +224,16 @@ drm_kunit_helper_connector_hdmi_init(struct kunit *test,
> test_edid_hdmi_1080p_rgb_max_200mhz);
> }
>
> +#define drm_kunit_atomic_restart_on_deadlock(ret, state, ctx, start) do {
> \
> + if (ret == -EDEADLK) {
> \
> + if (state)
> \
> + drm_atomic_state_clear(state);
> \
> + ret = drm_modeset_backoff(ctx);
> \
> + if (!ret)
> \
> + goto start;
> \
> + }
> \
> +} while (0)
> +I'm not sure here either, for pretty much the same reason. As far as locking goes, I really think we should prefer something explicit even if it means a bit more boilerplate. If you still want to push this forward though, this has nothing to do with kunit so it should be made a common helper. I do think it should be done in a separate series though. Ever-expanding series are a nightmare, both to contribute and to review :) Maxime
signature.asc
Description: PGP signature
