ASoC/soundwire: add companion amp support by bardliao · Pull Request #6025 · thesofproject/linux

bardliao · 2026-11-06T08:05:26Z

Some platforms use a SoundWire companion amp, which gets its audio data
from another SoundWire peripheral (the main amp) instead of from the
host. The current machine drivers and SoundWire core assume that every
stream is exchanged with the host, so these amps cannot be described or
configured.

This series adds the common support needed for such configurations:
a way to mark companion amp endpoints in the mach table, machine driver
support for the peripheral to peripheral path, a SoundWire stream type
that fits it.

ghost

🟡 Changes recommended

Critical defects can block card registration, permanently defer probing, overlap companion-stream payload slots, and omit codec prefixes required to avoid control-name collisions and resolve routes.

5 open issues

What changed in this PR

This PR adds shared ASoC and SoundWire support for companion amplifiers that receive audio from another peripheral rather than the host.

Changes:

Adds companion endpoint metadata, grouping, and RT712/RT1320 DAI entries.
Creates Intel codec-to-codec links and adapts AMD drivers to the updated endpoint helpers.
Introduces a companion stream type and adjusts peripheral port offsets.

File	Description
`sound/soc/sof/intel/hda.c`	Excludes companion endpoints from automatic discovery.
`sound/soc/sdw_utils/soc_sdw_utils.c`	Adds companion DAI entries and endpoint grouping.
`sound/soc/intel/boards/sof_sdw.c`	Creates companion codec-to-codec links.
`sound/soc/amd/acp/acp-sdw-sof-mach.c`	Adapts SOF endpoint parsing for companion groups.
`sound/soc/amd/acp/acp-sdw-legacy-mach.c`	Adapts legacy endpoint parsing for companion groups.
`include/sound/soc-acpi.h`	Adds the companion endpoint flag.
`include/sound/soc_sdw_utils.h`	Defines companion grouping structures and helper interfaces.
`include/linux/soundwire/sdw.h`	Adds the companion stream type.
`drivers/soundwire/generic_bandwidth_allocation.c`	Resets port offsets for companion streams.

_{Review effort: Balanced}

ghost

Man you seem to start really big SoundWire changes. You should hire me haha...

Overall I didn't quite get how the new dailink is supposed to be handled, specifically how it's allocated, prepared and triggered. It's quite dependent on the host-peripheral dailink.

ghost · 2026-11-06T14:37:58Z

-	num_links = sdw_be_num + ssp_num + dmic_num + hdmi_num + bt_num + 1;
-	dai_links = devm_kcalloc(dev, num_links, sizeof(*dai_links), GFP_KERNEL);
+	num_links = sdw_be_num + ssp_num + dmic_num + hdmi_num + bt_num + num_comp_amps + 1;
+	dai_links = devm_kcalloc(dev, num_links + 1, sizeof(*dai_links), GFP_KERNEL);


you should explain why there's an additional dailink.

What's not clear specifically is how this dailink is handled, i.e. how it's allocated, prepared, triggered.

so far the device-to-device connections are not handled by the SoundWire core, so it's not clear how this might work?

That is a codec to codec dai link to connect 2 peripherals. I will add some commits from @shumingfan to this PR to show how the link might work.

There are test commits as follows:
5ad33ef
edffa25
6032198

ghost · 2026-11-06T14:40:23Z

 	SDW_STREAM_PCM = 0,
 	SDW_STREAM_PDM = 1,
 	SDW_STREAM_BPT = 2,
+	SDW_STREAM_COMPANION = 3,


yeah but it could be PDM or PCM.

It think you'd need to add a new transport type, such as MANAGER_PERIPHERAL or PERIPHERAL_PERIPHERAL.

There are other uses cases where device-to-device communication is needed, e.g. to share the amp feedback. Don't make this just for the companion case.

ghost · 2026-11-06T14:41:14Z

 			slave_total_ch += ch;
+
+			if (m_rt->stream->type == SDW_STREAM_COMPANION)
+				port_bo = t_data->block_offset;


not following this one.

This is trying to let all peripherals use the same frame slots. Like rt1320 and rt712 use the same frame slots.

ghost · 2026-11-06T14:42:25Z

+			 * the mach table if it is present.
+			 */
+			if (!is_endpoint_present(sdw_device, codec_info_list[i].dais[j].dai_type) ||
+			    (codec_info_list[i].dais[j].quirk & SOC_SDW_COMPANION_AMPS))


the commit message says the opposite of what's done in other patches where you create a new dailink.

Trying to skip the companion dais here to avoid creating the dai links that connect the companion dai to the host. And the companion link will be created later by create_sdw_comp_dailinks().

ghost

🟡 Changes recommended

Undefined DAI and stream constants prevent compilation, while a missing Kconfig dependency and critical transport, port-allocation, and teardown defects create further build and runtime risks.

8 open issues

Define the missing companion DAI type · New
Add the required sdw-utils Kconfig dependency · New
Use the defined peripheral transport type · New
Initialize the stream configuration transport · New
Keep the bookkeeping port out of hardware programming · New
Clean up streams after preparation or enable failures · New
Avoid changing microphone clocks when configuring AIF3 · New
Release stream runtimes during shutdown · New

5 resolved since last review

_{Review effort: Balanced}

+ * the SoundWire bus instead of from the host, so a codec-to-codec dai link is created between
+ * the main amp and the companion amps.
+ */
+#define SOC_SDW_COMPANION_AMPS		BIT(18)


+		retval = sdw_stream_add_slave(rt1320->sdw_slave, &stream_config,
+			&companion_config, 1, sdw_stream);


ghost

🟡 Changes recommended

Critical companion-stream conflicts can overwrite host transport configuration and change active microphone clocks without updating the corresponding SoundWire stream parameters.

6 open issues

5 resolved since last review

_{Review effort: Balanced}

+
+	sconfig.type = SDW_STREAM_PCM;
+	sconfig.transport_type = SDW_TRANSPORT_PERIPHERAL_PERIPHERAL;
+	pconfig.num = 2;  /* Use port 2 as port 0 and 1 can be used for BPT */


+	.c2c_params = &asoc_sdw_bridge_params,
+	.num_c2c_params = 1,


Add asoc_sdw_companion_amp_endpoint struct to help machine driver to create codec to codec dai links for SoundWire companion amps. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>

The sdw_stream_config structures in the manager drivers are not initialized and only some of the fields are set. Initialize them to zero so that any field added to the structure later, and not explicitly set by these drivers, gets its default value instead of stack garbage. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>

The companion stream is from peripheral to peripheral and all peripherals need to use the same SoundWire frame area to transfer the stream data. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>

The companion amp dai links should not be created by the default machine driver. We will create a mach table for the config that companion amp is used. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>

A companion amp gets its audio data from another SoundWire peripheral (the main amp) instead of from the host, so the machine driver does not open a host DMA stream for it. Add common dai ops that the companion amp codec drivers can plug into their dai to handle the peripheral to peripheral stream. Signed-off-by: Shuming Fan <shumingf@realtek.com> Co-developed-by: Bard Liao <yung-chuan.liao@linux.intel.com> Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>

Signed-off-by: Shuming Fan <shumingf@realtek.com>

Add the componion amp dai info for rt712 and rt1320. This commit is only for testing which is not a correct information. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>

ghost

🟡 Changes recommended

Unresolved controller-port conflicts, microphone-clock overwrites, and unsafe stream teardown can disrupt active audio streams and cause use-after-free.

7 open issues

2 resolved since last review

_{Review effort: Balanced}

+
+	sconfig.type = SDW_STREAM_PCM;
+	sconfig.transport_type = SDW_TRANSPORT_PERIPHERAL_PERIPHERAL;
+	pconfig.num = 2;  /* Use port 2 as port 0 and 1 can be used for BPT */


+	dev_dbg(dai->dev, "%s %s id %d", __func__, dai->name, dai->id);
+	dev_dbg(dai->dev, "%s substream stream %d", __func__, substream->stream);
+
+	sdw_shutdown_stream(substream);


+	dev_dbg(dai->dev, "%s substream stream %d dai %s id %d",
+		__func__, substream->stream, dai->name, dai->id);
+
+	ret = sdw_startup_stream(substream);


bardliao requested review from charleskeepax and shumingfan and a balanced review from Copilot November 6, 2026 08:05

bardliao requested review from dbaluta, jack-cy-yu, kv2019i, lgirdwood, oder-chiou, plbossart, ranj063 and ujfalusi as code owners November 6, 2026 08:05

Copilot started reviewing on behalf of bardliao November 6, 2026 08:05

View session

Copilot AI reviewed Nov 6, 2026

View reviewed changes

plbossart reviewed Nov 6, 2026

View reviewed changes

Copilot Bot balanced review requested due to automatic review settings November 10, 2026 09:55

bardliao force-pushed the companion_amp-machine_driver branch from e890ae0 to ced34fe Compare November 10, 2026 09:55

Copilot started reviewing on behalf of bardliao November 10, 2026 09:55

View session

Copilot AI reviewed Nov 10, 2026

View reviewed changes

Copilot Bot balanced review requested due to automatic review settings November 12, 2026 06:23

bardliao force-pushed the companion_amp-machine_driver branch from ced34fe to 79f4a06 Compare November 12, 2026 06:23

Copilot started reviewing on behalf of bardliao November 12, 2026 06:24

View session

Copilot AI reviewed Nov 12, 2026

View reviewed changes

Bard Liao and others added 8 commits November 13, 2026 16:33

ASoC: sdw_utils: add companion amp support

15207a8

Add asoc_sdw_companion_amp_endpoint struct to help machine driver to create codec to codec dai links for SoundWire companion amps. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>

soundwire: add companion stream support

0133840

The companion stream is from peripheral to peripheral and all peripherals need to use the same SoundWire frame area to transfer the stream data. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>

ASoC: SOF: Intel: skip companion amp dai

2985ca9

The companion amp dai links should not be created by the default machine driver. We will create a mach table for the config that companion amp is used. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>

ASoC: rt712: add companion input dai

e03b3d0

Signed-off-by: Shuming Fan <shumingf@realtek.com>

ASoC: rt1320: add companion dai

ef32491

Signed-off-by: Shuming Fan <shumingf@realtek.com>

TEST: ASoC: soc_sdw_utils: add componion amp dai for rt712 and rt1320

f033865

Add the componion amp dai info for rt712 and rt1320. This commit is only for testing which is not a correct information. Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com>

bardliao force-pushed the companion_amp-machine_driver branch from 79f4a06 to f033865 Compare November 13, 2026 08:37

Copilot Bot balanced review requested due to automatic review settings November 13, 2026 08:37

Copilot started reviewing on behalf of bardliao November 13, 2026 08:38

View session

Copilot AI reviewed Nov 13, 2026

View reviewed changes

		retval = sdw_stream_add_slave(rt1320->sdw_slave, &stream_config,
		&companion_config, 1, sdw_stream);

Conversation

bardliao commented Nov 6, 2026

Uh oh!

ghost left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ghost left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

ghost Nov 6, 2026

Choose a reason for hiding this comment

Uh oh!

ghost Nov 9, 2026

Choose a reason for hiding this comment

Uh oh!

ghost Nov 9, 2026

Choose a reason for hiding this comment

Uh oh!

ghost Nov 6, 2026

Choose a reason for hiding this comment

Uh oh!

ghost Nov 9, 2026

Choose a reason for hiding this comment

Uh oh!

ghost Nov 6, 2026

Choose a reason for hiding this comment

Uh oh!

ghost Nov 9, 2026

Choose a reason for hiding this comment

Uh oh!

ghost Nov 6, 2026

Choose a reason for hiding this comment

Uh oh!

ghost Nov 9, 2026

Choose a reason for hiding this comment

Uh oh!

ghost left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ghost left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

ghost left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants