[PATCH] drm/connector: Allow clearing hdr infoframe

Derek Foreman derek.foreman at collabora.com
Mon Dec 2 17:49:17 UTC 2024


On 2024-11-29 17:40, Dmitry Baryshkov wrote:
> On Fri, Nov 29, 2024 at 03:38:01PM -0600, Derek Foreman wrote:
>> When a display isn't presenting HDR content, the infoframe shouldn't be
>> present at all. Currently if we set the HDR_OUTPUT_METADATA blob, we'll
>> set a Dynamic Range and Mastering (DRM) infoframe. If we remove that
>> blob later, the DRM infoframe will linger in its previous state, when
>> we expect it to clear.
>>
>> Disable the infoframe when no output metadata is present to prevent
>> this from happening.
>>
>> Fixes: f378b77227bc4 ("drm/connector: hdmi: Add Infoframes generation")
>> Signed-off-by: Derek Foreman <derek.foreman at collabora.com>
>> ---
>>   drivers/gpu/drm/display/drm_hdmi_state_helper.c | 2 ++
>>   1 file changed, 2 insertions(+)
> The fix is fine, however please consider extending it to other infoframe
> types. Clear infoframe->set in the beginning of the corresponding
> function.

I thought that might be the case, but the other infoframes looked like 
they would only be cleared if there was an error, where the DRM 
infoframe was the only one that must be intentionally cleared.

I wasn't certain if it was better to leave a stale infoframe in the 
highly unlikely case that creating the new one caused an error, or to 
have the infoframe disappear entirely.

I'll make a new patch to replace this one, and include AVI, SPD, and 
VSIF infoframes this time.

Thanks,
Derek

>> diff --git a/drivers/gpu/drm/display/drm_hdmi_state_helper.c b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
>> index feb7a3a75981..1cc4b49b1013 100644
>> --- a/drivers/gpu/drm/display/drm_hdmi_state_helper.c
>> +++ b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
>> @@ -398,6 +398,8 @@ static int hdmi_generate_hdr_infoframe(const struct drm_connector *connector,
>>   		&infoframe->data.drm;
>>   	int ret;
>>   
>> +	infoframe->set = false;
>> +
>>   	if (connector->max_bpc < 10)
>>   		return 0;
>>   
>> -- 
>> 2.45.2
>>


More information about the dri-devel mailing list