Hi Dmitry,
On 2025/5/29 0:09, Dmitry Baryshkov wrote:
@@ -595,16 +546,41 @@ static bool cdn_dp_check_link_status(struct cdn_dp_device
*dp)
static void cdn_dp_audio_handle_plugged_change(struct cdn_dp_device *dp,
bool plugged)
{
- if (dp->codec_dev)
- dp->plugged_cb(dp->codec_dev, plugged);
+ if (dp->sink_has_audio)
+ drm_connector_hdmi_audio_plugged_notify(dp->connector, plugged);
I'd say, notify always and let userspace figure it out via the ELD. Then
you shouldn't need sink_has_audio. This would match the behaviour of
HDMI drivers.
Oh, I find that there are similar usages in qcom msm driver. Is there
any more progress?
For msm driver it is required as DSP requires HDMI to be plugged for
the audio path to work.
I see, will fix in v4.
@@ -705,8 +681,6 @@ static int cdn_dp_encoder_atomic_check(struct drm_encoder
*encoder,
static const struct drm_encoder_helper_funcs cdn_dp_encoder_helper_funcs = {
.mode_set = cdn_dp_encoder_mode_set,
- .enable = cdn_dp_encoder_enable,
- .disable = cdn_dp_encoder_disable,
.atomic_check = cdn_dp_encoder_atomic_check,
Nit: for the future cleanup, it should probably be possible to get rid
of these encoder ops too by moving them to the bridge ops.
Interesting, have these patches been submitted upstream yet?
Everything is already there, see drm_bridge_funcs::mode_set() and
drm_bridge_funcs::atomic_check().
Thanks for the clarification. I will move mode_set() to bridge ops.
And for the drm_encoder_helper_funcs::atomic_check(), most Rockchip
drivers will set some Rockchip-specific properties here so that the VOP
driver can process them. In the future, we may integrate a new encoder
driver to process these private properties. So, I prefer to keep this as
it is.
