The first ~500ms of HDMI audio stream are not played, if we play another stream within 5 seconds, the beggining is correctly played. This issue has been observed only with Samsung S24C650PL monitor, with and without X launched, with monitor built-in speakers or headset, on HSW, BYT-M and BDW. There is the data missing problem with Display Port only the first playback, unlike HDMI where it is reproduced each time. RTD3 is disabled. During the process, no mode set happens for the HDMI monitor (1920x1080@60.0, transcoder B, port C) and the video display is good. Reproduced with mainline kernels v3.14.0 and v3.15-rc1 We have checked the BDW gfx audio registers but found nothing wrong. So we dumped the video registers by intel_reg_dumper, and need gfx team's help to check if gfx side is also okay: (1) IRD_hdmi_only.txt … register dump before we play the audio stream over HDMI (2) IRD_hdmi_only_play.txt … register dump when we’re playing the audio stream over HDMI I only found RC6_RESIDENCY_TIME increases. The dmesg is also attached. Atm, we suspect the audio receiver part of this monitor can easily sleep and need some time to wake up and synchronize new audio stream, so there is audible audio frame loss.
Created attachment 101624 [details] Display register dump before we play the audio stream over HDMI
Created attachment 101625 [details] Display register dump when we’re playing the audio stream over HDMI
Created attachment 101626 [details] kernel message
Sorry Mengdong, we haven't looked into this yet. :( Is this still an issue with current upstream or drm-intel-nightly?
Hi Jani, This issue can still be reproduced with latest drm-intel-nightly.
Nothing in the kernel logs or reg dumps... can you get ahold of an HDMI analyzer and see if we're sending audio frames correctly right away? It might be a problem with this sink somehow. Sounds like you can work around the issue by sending two streams when starting to play new audio? That might be something you could do in the audio driver if this sink is detected.
Another try with current drm-intel-nightly would be good too. I think the audio rework on the display side fixed a bunch of odd cases.
Hopefully this went away. Please re-open with some more info if not, Mengdong.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.