dts: fix memory allocation failed and move the pipeline to different core by joechengxperi · Pull Request #11230 · thesofproject/sof

joechengxperi · 2026-09-24T08:56:43Z

This PR is to resolve the issue Google reported that audio device would lost when there are two applications running and switch endpoint simultaneously.

Copilot

Copilot review overview

🟡 Changes recommended

Replace the undefined heap symbol with SOF_ZEPHYR_VIRTUAL_HEAP_REGION_SIZE.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 1

Open (1)

Use the correct Kconfig symbol for virtual heap size · New

What changed in this PR

This PR adds DTS-specific SoundWire topologies, configurable core placement, and a PTL heap adjustment to address endpoint switching and allocation failures.

Changes:

Adds DTS SDCA topology targets.
Applies DTS_CORE_ID to DTS pipelines and widgets.
Adds PTL virtual-heap configuration.

File	Summary
`tools/topology/topology2/production/tplg-targets-sdca-generic.cmake`	Adds DTS SDCA topology targets.
`tools/topology/topology2/platform/intel/sdw-jack-dts.conf`	Applies configurable core affinity to jack DTS pipelines.
`tools/topology/topology2/platform/intel/sdw-amp-dts.conf`	Applies configurable core affinity to amplifier DTS pipelines.
`tools/topology/topology2/cavs-sdw.conf`	Defines the default DTS core.
`app/overlays/ptl/dts_overlay.conf`	Attempts to configure the PTL virtual heap, but uses an undefined Kconfig symbol and does not change the heap size.

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

kv2019i · 2026-09-28T14:46:42Z

 CONFIG_LLEXT_HEAP_SIZE=64
 CONFIG_KCPS_DYNAMIC_CLOCK_CONTROL=n
 CONFIG_SOF_STACK_SIZE=8192
+CONFIG_SOF_ZEPHYR_VIRTUAL_HEAP_SIZE=0x80000


@joechengxperi This should change CONFIG_SOF_ZEPHYR_HEAP_SIZE due to a recent rename. See commit 5cf6f9c

@kv2019i
I've updated it now. And I have one question. I will have to create another PR to merge this to ptl-008-drop-stable. I just checked there is no this commit 5cf6f9c in ptl-008-drop-stable. Will this commit 5cf6f9c be cherry-picked to ptl-008-drop-stable? Thanks.
cc @macchian

lyakh · 2026-09-25T07:29:16Z

 			{
 				index			21
+				core_id			$DTS_CORE_ID
+				core			$DTS_CORE_ID


I think only one of them is needed and I think it's core, could you check?

Yes, it's confusing the tplg compiler doesn't catch this, but for the pipeline widget, only "core" is needed.

But I won't block because of this, "core" is the pipeline level token (SOF_TKN_SCHED_CORE for Linux kernel) and "core_id" is the widget token (SOF_TKN_COMP_CORE_ID in Linux kernel). For a pipeline, both can be set, but only "core" is needed as Linux driver will force same values.

This is confusing I admit and we should improve this for topology developers.

lyakh · 2026-09-25T07:30:23Z

+			core_id	$DTS_CORE_ID
+
+			Object.Widget.pipeline.1 {
+				core $DTS_CORE_ID


right, I think pipeline widgets use core, while processing components use core_id

lgirdwood

Patches 2/3 are good to merge now, there are some inflight memory changes now so teh fiest patch could merge then be deleted in a few weeks. @kv2019i whats the current status ?

kv2019i

Please fix the Kconfig name due to recent rename. Minor comments otherwise (other comments not blocking).

kv2019i · 2026-09-28T14:46:42Z

 CONFIG_LLEXT_HEAP_SIZE=64
 CONFIG_KCPS_DYNAMIC_CLOCK_CONTROL=n
 CONFIG_SOF_STACK_SIZE=8192
+CONFIG_SOF_ZEPHYR_VIRTUAL_HEAP_SIZE=0x80000


@joechengxperi This should change CONFIG_SOF_ZEPHYR_HEAP_SIZE due to a recent rename. See commit 5cf6f9c

kv2019i · 2026-09-28T15:39:41Z

 			{
 				index			21
+				core_id			$DTS_CORE_ID
+				core			$DTS_CORE_ID


Yes, it's confusing the tplg compiler doesn't catch this, but for the pipeline widget, only "core" is needed.

But I won't block because of this, "core" is the pipeline level token (SOF_TKN_SCHED_CORE for Linux kernel) and "core_id" is the widget token (SOF_TKN_COMP_CORE_ID in Linux kernel). For a pipeline, both can be set, but only "core" is needed as Linux driver will force same values.

This is confusing I admit and we should improve this for topology developers.

kv2019i · 2026-09-28T15:41:59Z

 SDW_AMP_FEEDBACK=false,SDW_SPK_STREAM=Playback-SmartAmp,NUM_HDMIS=0,\
 DEEP_BUF_SPK=true"

+"cavs-sdw\;sof-sdca-jack-id0-dts\;SDW_JACK_OUT_STREAM=Playback-SimpleJack,\


FYI @bardliao @ujfalusi I have no objections to this. Just noting there has been work ongoing to standardize how to extended SDCA topologies (and/or use feature topologies).

I don't think a feature topology can be used here, it cannot replace part of the topology, it can add to it a feature.
It replaces quite a bit in the topologies.

to

Similarly for the amp.

I think that for products with such changes the generic sdca topologies cannot be used, if people will start to copy, rename random topologies to 'stick', it is going to cause lots of issues and at the same time it will make debugging harder as no one knows exactly what is loaded.

Another option would be for the kernel to support sort of 'feature modifier' kernel parameter to add postfix to the feature topology name and try that first (than fall back to w/o modifier if not found).
feature_topology_type="dts" or feature_topology_type="jack:dts".

But, what if you want to add feedback also to the path which have dts? Or no DTS, but only feedback?

For products: tplg_filename=sof-<product name/code name>.tplg, my imagination let me down on the naming to imply that it is a monolithic topology for a certain configuration (and feature set).

DTS library would allocate memory failed when there are multiples apps are launched and endpoint switches at the same time. Incease the size to avoid this issue Signed-off-by: Joe Cheng <joe.cheng@xperi.com>

Add DTS_CORE_ID and let the pipeline could run on approriate core Signed-off-by: Joe Cheng <joe.cheng@xperi.com>

Enable DTS topology for SDCA and set to different cores for speaker/heapdhone pipeline to avoid resource not enough Signed-off-by: Joe Cheng <joe.cheng@xperi.com>

intel-sofci · 2026-10-08T09:30:54Z

PR 11230: test results

Run date: 2026-10-08 09:06 UTC

Tested commit: 486634e8c41e0f9b5250eb002db184c7a75e6c47

lgirdwood · 2026-10-12T15:31:50Z

@joechengxperi assume all review comment now resolved ? @kv2019i good for you ?

joechengxperi · 2026-10-13T09:03:48Z

@joechengxperi assume all review comment now resolved ? @kv2019i good for you ?

Yes, all resolved.

joechengxperi requested review from johnylin76 and macchian and a lite review from Copilot September 24, 2026 08:56

joechengxperi requested review from dbaluta, jsarha, kv2019i, lbetlej, lgirdwood, mmaka1, plbossart and ranj063 as code owners September 24, 2026 08:56

Copilot started reviewing on behalf of joechengxperi September 24, 2026 09:00

View session

Copilot AI reviewed Sep 24, 2026

View reviewed changes

lyakh reviewed Sep 25, 2026

View reviewed changes

lgirdwood reviewed Sep 25, 2026

View reviewed changes

kv2019i requested changes Sep 28, 2026

View reviewed changes

dts: increase heap size on ptl

31258f3

DTS library would allocate memory failed when there are multiples apps are launched and endpoint switches at the same time. Incease the size to avoid this issue Signed-off-by: Joe Cheng <joe.cheng@xperi.com>

joechengxperi force-pushed the dts-ptl-issue-fix-main branch from 65f79fd to 39e2cb6 Compare September 29, 2026 07:23

joechengxperi requested a review from kv2019i October 7, 2026 13:42

joechengxperi added 2 commits October 8, 2026 11:52

topology2: add DTS_CORE_ID

7d6f134

Add DTS_CORE_ID and let the pipeline could run on approriate core Signed-off-by: Joe Cheng <joe.cheng@xperi.com>

topology2: add DTS topology for SDCA

486634e

Enable DTS topology for SDCA and set to different cores for speaker/heapdhone pipeline to avoid resource not enough Signed-off-by: Joe Cheng <joe.cheng@xperi.com>

joechengxperi force-pushed the dts-ptl-issue-fix-main branch from 39e2cb6 to 486634e Compare October 8, 2026 03:53

lgirdwood approved these changes Oct 12, 2026

View reviewed changes

kv2019i approved these changes Oct 21, 2026

View reviewed changes

lgirdwood merged commit 2515800 into thesofproject:main Oct 22, 2026
46 checks passed

Conversation

joechengxperi commented Sep 24, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

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!

lgirdwood left a comment

Choose a reason for hiding this comment

Uh oh!

kv2019i left a comment

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!

intel-sofci commented Oct 8, 2026

PR 11230: test results

Uh oh!

lgirdwood commented Oct 12, 2026

Uh oh!

joechengxperi commented Oct 13, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants