ll-schedule: fix tasks removed during execution

[lyakh]

Zephyr LL-scheduler is supposed to be able to handle tasks, removed or added while the scheduler is executing. However, there is a bug in that implementation. If the task, that is currently executing, is removed, its list head, that links it into the list of tasks, is initialised. So when trying to get a pointer to the next task we obtain a pointer to the same task.

BugLink: #7084

[marc-hb]

Could we merge these two commits over two different days? So we have one run of daily test with just the first commit, which looks like a very big change.

Also, compatibility with upstream Zephyr was just broken, some daily tests are already broken: https://github.com/thesofproject/sof/actions/runs/4228168969/jobs/7343348093

/zep_workspace/sof/src/lib/dai.c: In function 'dai_get_zephyr_device':
/zep_workspace/sof/src/lib/dai.c:159:53: error: passing argument 2 of 'dai_config_get' makes pointer from integer without a cast [-Werror=int-conversion]
  159 |                 cfg = dai_config_get(zephyr_dev[i], dir);
      |                                                     ^~~
      |                                                     |
      |                                                     int
In file included from /zep_workspace/sof/src/lib/dai.c:23:
/zep_workspace/zephyr/include/zephyr/drivers/dai.h:336:53: note: expected 'struct dai_config *' but argument is of type 'int'
  336 |                                  struct dai_config *cfg,
      |                                  ~~~~~~~~~~~~~~~~~~~^~~
/zep_workspace/sof/src/lib/dai.c:159:23: error: too few arguments to function 'dai_config_get'
  159 |                 cfg = dai_config_get(zephyr_dev[i], dir);
      |                       ^~~~~~~~~~~~~~
/zep_workspace/zephyr/include/zephyr/drivers/dai.h:335:19: note: declared here
  335 | static inline int dai_config_get(const struct device *dev,
      |                   ^~~~~~~~~~~~~~

EDIT: dai_config mystery solved and fixed:

• dai: change to use new version of dai_config_get #6939 (comment)

[lyakh]

converted back to draft: tests are still failing (real tests, not this timing out CI run https://sof-ci.01.org/sofpr/PR7138/build4066/devicetest/index.html), and I cannot for now figure out why. I'll need to debug later.

[lyakh]

SOFCI TEST

[lyakh]

The single failure is a suspend-resume test https://sof-ci.01.org/sofpr/PR7138/build4166/devicetest/index.html?model=ADLP_RVP_SDW_IPC4ZPH&testcase=check-suspend-resume-with-playback

[makarukp]

Can we also remove 'schedule_task_cancel(&cd->chain_task)' in 'chain_task_pause' ? I think after this change it will not be necessary.

[lyakh]

@makarukp this should be a separate PR, reverting that change - after this is merged

[jsarha]

I run multiple-suspend-resume test number of times on my upx-i11 with this PR and "hda: chain dma: cancel task before freeing it" reverted. I got one transient failure (but no crash), but on three other times the test was successful. So I guess this is a proper fix for the ll scheduler crashing problem.