hda: chain transition states fix

[makarukp]

Require maintenance in logic to handle multiple start/pause requests by one chain instance.

Signed-off-by: Piotr Makaruk piotr.makaruk@intel.com

[ujfalusi]

Created a draft PR with this and a revert to enable the ChainDMA on tgl: #6984

[jsarha]

This is a definite improvement over the situation before this PR, since before this a resume after pause did not work at all. Now it works, at least most of the time. If I run "sof-test/test-case/multiple-pause-resume.sh -r 50" on my upx-i11 it fails more often than succeeds. However, hitting the failure usually takes quite a while. When looking at FW logs, I can see several:
[ 16.733930] copier: comp:0 0x40000 failed to find dai comp or sink pipeline not running.
and even more:
[ 20.268613] dai_comp: comp:1 0x40001 dai_copy(): nothing to copy

[jsarha]

I tried adding "pm_policy_state_lock_put(PM_STATE_RUNTIME_IDLE, PM_ALL_SUBSTATES);" after "cd->chain_task.state = SOF_TASK_STATE_INIT;" but it does not appear to help (nor make the behavior worse. One example of failure in the test logs:

(3/50) pcm'Deepbuffer HDA Analog' cmd'aplay' id'5': Wait for 44 ms before resume
(3/50) pcm'HDA Analog' cmd'aplay' id'3': Wait for 49 ms before resume
(4/50) pcm'HDA Analog' cmd'aplay' id'3': Wait for 49 ms before pause
(4/50) pcm'Deepbuffer HDA Analog' cmd'aplay' id'5': Wait for 48 ms before pause
(4/50) pcm'Deepbuffer HDA Analog' cmd'aplay' id'5': Wait for 49 ms before resume
(4/50) pcm'HDA Analog' cmd'aplay' id'3': Wait for 42 ms before resume
aplay: pcm_write:2086: write error: Input/output error
aplay: pcm_write:2086: write error: Input/output error

2023-01-23 22:38:09 UTC [REMOTE_INFO] pipeline: HDA Analog with aplay
2023-01-23 22:38:09 UTC [REMOTE_INFO] pipeline: Deepbuffer HDA Analog with aplay
2023-01-23 22:38:09 UTC [REMOTE_INFO] Check expect exit status
declare -- cmd="journalctl_cmd --since=@1674513395"
2023-01-23 22:38:09 UTC [REMOTE_ERROR] pause resume PID 5752 had non-zero exit status

[jsarha]

Now after over 20 test runs of multiple-pause-resume I now saw an error also from the FW build where I had 6983 & 6656 reverted. So I guess its reasonable assume 6656 with 6983 is Ok. My review comments were also well explained. Approved.

[lgirdwood]

I will rerun CI after some regression fixes are merged today.

[marc-hb]

SOFCI TEST

[marc-hb]

topology/topology2/avs-tplg/sof-nocodec-bt-lbm.tplg build failure in https://sof-ci.01.org/sofpr/PR6983/build3565/build/tools.txt , https://sof-ci.01.org/sofpr/PR6983/build3565/buildhttps://github.com/thesofproject/sof/actions/runs/4023985041/jobs/6915487545 and many others fixed by #6999

[marc-hb]

@makarukp to fix Github Actions and retest with #6999 you must perform an empty git commit --amend -C HEAD to change the SHA1 and then force-push.

Don't rebase as this makes it very difficult to see that nothing changed. Most CIs perform a merge before testing and our Github Actions do.

[kv2019i]

@lyakh good to go?