scheduler: check task state prior to completion

[lgirdwood]

High priority IRQ handlers can potentially reschedule tasks prior to task
completion. i.e.

1. Task runs and unmasks high priority IRQ prior to completion.
2. High priority IRQ prempts task context and reschedules task (changing
   task->state).
3. Initial task resumes and then completes (clobbering task->state in 2).

Make sure this flow is supported for high priority IRQ scheduling.

Signed-off-by: Liam Girdwood liam.r.girdwood@linux.intel.com

[lgirdwood]

@ranj063 compile tested only, can you check and stress it. Better fix.

[ranj063]

@lgirdwood this looks correct. I will kick off the stress test this morning on my boards. Thanks!

[ranj063]

@lgirdwood what if the task state changes to pending because of the race condition we saw earlier. We will still end up setting it to completed and miss the next ipc task.

[lgirdwood]

Potentially if the schedule time for the next task is quite close and in this case it is immediately :

void ipc_schedule_process(struct ipc *ipc)
{
schedule_task(&ipc->ipc_task, 0, 100);
}

We need to relax this as IPC is never that high priority. Added PENDING case and pushed.

[ranj063]

@lgirdwood I will add case TASK_STATE_PENDING as well for the "do nothing" part and start the stress test.

[xiulipan]

@ranj063 There might be the case.

[ranj063]

@lgirdwood ever since we made the scheduler changes, I seem to be hitting IPC timeouts with CTX_SAVE ipc.
2 of the boards I tested on today failed within 500 iterations with the same ipc timing out.

[michalgrodzicki]

@lgirdwood could you please fix unit tests?

[jocelyn-li]

@lgirdwood This patch breaks Unit tests. Please fix the unit test failure.

[lgirdwood]

@michalgrodzicki@jocelyn-li unit tests failed due to commit f00d46b.
There is a cmocka bug that has to be fixed first #735 and some cmocka enhancements needed to allow non expert users to build cmocka tests. @jajanusz and I are already discussing.

[ranj063]

@lgirdwood this PR is good to be merged once the CI passes. Thanks!

[mengdonglin]

@lgirdwood Would you please merge this PR? Firmware team will cherry pick it to GLK release branch glk-005-drop-stable.