wov: [CI TEST] initial fixes for wov regressions on several targets. by lgirdwood · Pull Request #11309 · thesofproject/sof

lgirdwood · 2026-10-14T14:28:46Z

Testing, expect some patches/changes not required.

The DP lock is taken by the LL thread on every tick in scheduler_dp_ll_tick() and by each DP thread after every run. It was a k_sem, which has no priority inheritance: when a DP thread holding it was preempted by other DP threads, the cooperative high priority LL thread blocked until the holder got CPU time again. With several busy DP modules (3 MFCC + 3 MWW on cAVS 2.5) single LL ticks took up to ~40 ms, overrunning the DMIC DMAs of unrelated streams. With a k_mutex the holder inherits the LL priority: LL max drops from ~1.5M to <30k timer cycles, with no LL overruns or DMA xruns. Co-developed-by: Liam Girdwood <liam.r.girdwood@linux.intel.com> Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com>

CONFIG_ZEPHYR_DP_SCHEDULER can be enabled on cAVS, but only the ACE platform_init() registered the DP scheduler. Creating a DP module on cAVS then failed with "unable to bind scheduler to task (type 3)" and module init returned -ENODEV. Call scheduler_dp_init() after the LL scheduler is set up, as ACE does. Co-developed-by: Liam Girdwood <liam.r.girdwood@linux.intel.com> Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com>

lgirdwood · 2026-10-14T14:40:50Z

 DECLARE_TR_CTX(dp_tr, SOF_UUID(dp_sched_uuid), LOG_LEVEL_INFO);

-#define DP_LOCK_INIT(i, _)	Z_SEM_INITIALIZER(dp_lock[i], 1, 1)
+/* k_mutex, not k_sem: priority inheritance keeps a preempted DP holder from stalling LL */


@kv2019i assuming this patch wont be compatible with userspace.
@ujfalusi I suspect this wont be needed with clang. We can revert test when time permits.

lgirdwood · 2026-10-14T14:42:05Z

 	default y
 	help
 	  Select for KPB component
-


not needed.

lgirdwood · 2026-10-14T14:42:26Z

 	if (buffer->secondary_buffer_sink && buffer->secondary_buffer_sink->ops->reset)
 		buffer->secondary_buffer_sink->ops->reset(buffer->secondary_buffer_sink);

+#ifdef CONFIG_DP_TO_DP_BIND


this looks like it belongs in another patch.

lgirdwood · 2026-10-14T14:43:20Z

 	assert(!ret);
 }
-#else /* CONFIG_LIBRARY */
+#else


this looks like a completely separate patch too.

lgirdwood · 2026-10-14T14:45:00Z

+
+[[maybe_unused]] void *sof_operator_new(size_t size)
+{
+	return malloc(size);


how can we map this to mod_alloc ? We could use current thread context here to get metadata in order to alloc.

lgirdwood · 2026-10-14T14:47:50Z

 	default 0x40000 if VIRTUAL_HEAP
 	default 0xF0000 if SOC_ACE15_MTPM || SOC_ACE20_LNL
-	default 0x100000 if SOC_ACE30 || SOC_ACE40
+	default 0xD0000 if SOC_ACE30 || SOC_ACE40


is this needed for the C++ new ?

lgirdwood · 2026-10-14T14:48:12Z

         ext_init payload. If the stack size requested in the IPC is
         smaller than this, then the value defined here takes over.

+config DP_TO_DP_BIND


seems unrelated, this should be form DP to DP patch.

lgirdwood · 2026-10-14T14:49:35Z


 	err = comp_prepare(current);
-	if (err < 0 || err == PPL_STATUS_PATH_STOP)
+	if (err < 0 || err == PPL_STATUS_PATH_STOP) {


this looks like a separate patch

lgirdwood · 2026-10-14T14:50:12Z

-		dai_common_free(cd->dd[i]);
-		mod_free(mod, cd->dd[i]->gain_data);
-		mod_free(mod, cd->dd[i]);
+		if (cd->dd[i]) {


this should be its own patch.

lgirdwood · 2026-10-14T14:52:04Z

 {
 	assert_can_be_cold();

+	if (!dd)


I suspect a lot of these are really not needed

When freeing an EDF scheduled task, ensure any pending delayed work associated with the task is cancelled before marking the task free and clearing callbacks. This prevents delayed work from triggering on freed task data during pipeline resets or teardown. Co-developed-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com> Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

In copier_dai_free(), guard cd->dd[i] against NULL before accessing cd->dd[i]->gain_data and calling dai_common_free(). This prevents a NULL pointer dereference if endpoint allocation failed during initialization or during teardown of partially configured copier instances. Also reset pointers to NULL after freeing. Co-developed-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com> Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

When comp_prepare() returns a negative error code in pipeline_comp_prepare(), emit a pipe_err message including the component ID and error code. This greatly accelerates diagnosing topology and IPC parameter configuration failures during pipeline setup. Co-developed-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com> Signed-off-by: Liam Girdwood <liam.r.girdwood@linux.intel.com>

Copilot

🟡 Changes recommended

EDF task cleanup can race with an already-running delayed-work callback.

1 open finding

Delayable work cancellation can race with in-flight callback · New

What changed in this PR

Addresses scheduling and cleanup regressions affecting WoV targets.

Changes:

Improves DP scheduler initialization and priority inheritance.
Adds preparation failure diagnostics.
Hardens EDF and DAI resource cleanup.

File	Description
`zephyr/edf_schedule.c`	Cancels delayed work during task cleanup.
`src/schedule/zephyr_dp_schedule.c`	Replaces DP semaphores with priority-inheriting mutexes.
`src/platform/intel/cavs/platform.c`	Initializes the DP scheduler on cAVS.
`src/audio/pipeline/pipeline-params.c`	Logs component preparation failures.
`src/audio/copier/copier_dai.c`	Adds guarded, idempotent DAI cleanup.

🧠 Review effort: Balanced

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.


 static int schedule_edf_task_free(void *data, struct task *task)
 {
+	k_work_cancel_delayable(&task->z_delayed_work);


ujfalusi added 2 commits October 14, 2026 14:35

lgirdwood commented Oct 14, 2026

View reviewed changes

lrgirdwo added 3 commits October 14, 2026 17:26

lgirdwood force-pushed the upstream-pr1-core-sched branch from 91ea041 to 610f318 Compare October 14, 2026 16:27

lgirdwood requested a balanced review from Copilot October 14, 2026 19:06

Copilot started reviewing on behalf of lgirdwood October 14, 2026 19:06

View session

Copilot AI reviewed Oct 14, 2026

View reviewed changes

Comment thread zephyr/edf_schedule.c

static int schedule_edf_task_free(void *data, struct task *task)

{

k_work_cancel_delayable(&task->z_delayed_work);

Navigation Menu

Repository navigation

wov: [CI TEST] initial fixes for wov regressions on several targets. - #11309

wov: [CI TEST] initial fixes for wov regressions on several targets.#11309
lgirdwood wants to merge 5 commits into
thesofproject:mainfrom
lgirdwood:upstream-pr1-core-sched

lgirdwood commented Oct 14, 2026

Uh oh!

lgirdwood Oct 14, 2026

Uh oh!

lgirdwood Oct 14, 2026

Uh oh!

lgirdwood Oct 14, 2026

Uh oh!

lgirdwood Oct 14, 2026

Uh oh!

lgirdwood Oct 14, 2026

Uh oh!

lgirdwood Oct 14, 2026

Uh oh!

lgirdwood Oct 14, 2026

Uh oh!

lgirdwood Oct 14, 2026

Uh oh!

lgirdwood Oct 14, 2026

Uh oh!

lgirdwood Oct 14, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

lgirdwood commented Oct 14, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants