On Thu, Jun 18, 2020 at 6:19 PM Pierre-Louis Bossart pierre-louis.bossart@linux.intel.com wrote:
Pierre, your thoughts on this? This has already been confirmed working.
I don't have any specific knowledge on Broadwell to comment. I also haven't had time to test this patch, I was expecting Ross to provide his Tested-by tag?
Hello,
Ross has provided his Tested-by tag already. Patch has been tested by Intel & Google side both. Given problem's impact, this fix is considered a critical one. I think we are good-to-go for quite a while now?
Czarek
I just tested speaker playback on Dell XPS13 and Samus Chromebook to double-check my UCM2 changes for SOF were indeed backwards compatible with the SST driver case. Well, my changes are fine but the kernel not so much.
With a 5.8-rc1 kernel w/ the SST driver, sounds played through pulseaudio are rendered too slowly with clicky artefacts. Using the alsa hw device works fine. In some cases, the sound rendered by PulseAudio become clear again after a while. Restarting the UI and testing degrades the audio again.
Reverting this patch - identified with git bisect - solves the issue on both devices, pulseaudio works fine again without any transient behavior. I spent 15mn monkey-testing and the audio quality was always good when this patch is reverted.
I have no idea what the fixes were, but going from a somewhat random D3 exit problem to a 100% reproducible issue is problematic. I trust both Cezary and Ross did test this patch, but could it be that pulseaudio tests were skipped?
We reverted this patch locally due to regressions and raised the issue with Cezary on Github, we got no response.
Curtis
8ec7d6043263ecf250b9b7c0dd8ade899487538a is the first bad commit commit 8ec7d6043263ecf250b9b7c0dd8ade899487538a Author: Cezary Rojewski cezary.rojewski@intel.com Date: Mon Mar 30 21:45:20 2020 +0200
ASoC: Intel: haswell: Power transition refactor