audio: dai-zephyr: rework calls to DMA driver, remove channel pointer

[kv2019i]

For historical reasons, dai-zephyr has somewhat complicated code to manage the DMA channel instance information. When a DMA channel is allocated, a pointer to the system DMA channel table is acquired and some additional information is stored per channel. This is however redundant as the only piece of information actually needed is the channel index.

Simplify the code by not storing the channel pointer anymore, but rather just store the channel index and use that in all calls to the DMA driver.

[kv2019i]

Similar to previously done change for host-zephyr.c in #10721

[Copilot]

Pull request overview

This PR simplifies dai-zephyr DMA channel handling by removing the stored DMA channel pointer and keeping only the DMA channel index, then updating DMA driver calls to use that index.

Changes:

• Replaced struct dai_data::chan with chan_index and updated call sites accordingly.
• Updated DAI DMA stop/release and position/status queries to use chan_index.
• Removed per-channel dev_data handling associated with the removed channel pointer.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 7 comments.

File	Description
src/ipc/ipc4/dai.c	Switches IPC4 DAI DMA release/config/position code paths to use chan_index instead of a channel pointer.
src/include/sof/lib/dai-zephyr.h	Replaces struct dai_data *chan with int chan_index.
src/audio/dai-zephyr.c	Updates DMA config/prepare/trigger/copy/status paths to use chan_index and dd->dma directly.

[softwarecki]

In my opinion we should treat any negative value as "no channel". This simplifies the code compared to checking against a specific constant, makes it safer for future changes and more consistent. The channel id value comes from sof_dma_request_channel(), and code check < 0 to detect errors, so any negative value mean no assigned channel.

[lgirdwood]

In my opinion we should treat any negative value as "no channel". This simplifies the code compared to checking against a specific constant, makes it safer for future changes and more consistent. The channel id value comes from sof_dma_request_channel(), and code check < 0 to detect errors, so any negative value mean no assigned channel.

Ack - and I think historically 0 is also a valid channel number used in the HW DMA specs so negative better.

[kv2019i]

@softwarecki wrote:

In my opinion we should treat any negative value as "no channel". This simplifies the code compared to checking against a specific constant, makes it safer for future changes and more consistent. The channel id value comes from sof_dma_request_channel(), and code check < 0 to detect errors, so any negative value mean no assigned channel.

Ack, this was actually @abonislawski 's idea to use Zephyr DMA API error values directly. But makes sense to continue and check for a negative value, will change in V2.

[kv2019i]

V2 pushed:

• addressed all review comments so far
• added commit to drop support for non-native drivers and IPC4 (no known users, pretty much everybody has moved to native drivers and those who are not, are using IPC3)
• add a comment to align host-zephyr.c to feedback from @softwarecki

[lgirdwood]

@kv2019i btw, some build errors:

modules/sof/CMakeFiles/modules_sof.dir/__w/sof/sof/sof/src/ipc/ipc3/dai.c.obj.d -o modules/sof/CMakeFiles/modules_sof.dir/__w/sof/sof/sof/src/ipc/ipc3/dai.c.obj -c /__w/sof/sof/sof/src/ipc/ipc3/dai.c
/__w/sof/sof/sof/src/ipc/ipc3/dai.c: In function 'dai_dma_release':
/__w/sof/sof/sof/src/ipc/ipc3/dai.c:314:15: error: 'struct dai_data' has no member named 'chan'
  314 |         if (dd->chan) {
      |               ^~
/__w/sof/sof/sof/src/ipc/ipc3/dai.c:316:44: error: 'struct dai_data' has no member named 'chan'
  316 |                 notifier_unregister(dev, dd->chan, NOTIFIER_ID_DMA_COPY);
      |                                            ^~
/__w/sof/sof/sof/src/ipc/ipc3/dai.c:318:39: error: 'struct dai_data' has no member named 'chan'
  318 |                 dma_release_channel(dd->chan->dma->z_dev, dd->chan->index);
      |                                       ^~
/__w/sof/sof/sof/src/ipc/ipc3/dai.c:318:61: error: 'struct dai_data' has no member named 'chan'
  318 |                 dma_release_channel(dd->chan->dma->z_dev, dd->chan->index);
      |                                                             ^~
/__w/sof/sof/sof/src/ipc/ipc3/dai.c:322:19: error: 'struct dai_data' has no member named 'chan'
  322 |                 dd->chan->dev_data = NULL;
      |                   ^~
/__w/sof/sof/sof/src/ipc/ipc3/dai.c:323:19: error: 'struct dai_data' has no member named 'chan'
  323 |                 dd->chan = NULL;
      |                   ^~
/__w/sof/sof/sof/src/ipc/ipc3/dai.c: In function 'dai_config':
/__w/sof/sof/sof/src/ipc/ipc3/dai.c:355:23: error: 'struct dai_data' has no member named 'chan'
  355 |                 if (dd->chan) {
      |                       ^~
In file included from /__w/sof/sof/zephyr/include/zephyr/logging/log.h:11,
                 from /__w/sof/sof/sof/zephyr/include/sof/trace/trace.h:10,
                 from /__w/sof/sof/sof/zephyr/include/rtos/alloc.h:12,
                 from /__w/sof/sof/sof/zephyr/include/sof/lib/dma.h:23,
                 from /__w/sof/sof/sof/src/include/sof/audio/audio_stream.h:23,
                 from /__w/sof/sof/sof/src/include/sof/audio/buffer.h:11,
                 from /__w/sof/sof/sof/src/include/sof/audio/component.h:19,
                 from /__w/sof/sof/sof/src/include/sof/audio/component_ext.h:19,
                 from /__w/sof/sof/sof/src/ipc/ipc3/dai.c:8:
/__w/sof/sof/sof/src/ipc/ipc3/dai.c:357:37: error: 'struct dai_data' has no member named 'chan'
  357 |                                   dd->chan->index);

[kv2019i]

Back to draft, this needs more work to fix IPC3 builds (dai-zephyr.c is shared and needs to work with both IPC3 and IPC4 dai code).

[kv2019i]

V3 pushed:

• Fix IPC3 builds. The IPC3 dai code has some code to interface with DMA drivers that was broken by the changes.
• With IPC3, legacy DMA/DAI drivers are still supported, but for IPC4, native drivers are mandatory.

[kv2019i]

V4 and V5 pushed:

• This turned out to be more complicated. Some of the SOF library test targets (like testbench and SOF ALSA plugin) rely on the old DAI legacy interface to build SOF IPC4 build without depending on Zephyr. By adding new dependencies to IPC4 builds, this PR broke these test targets.
• Fortunately the test targets do not actually need IPC DAI integration, so in V5 this is fixed by simply leaving this code out from library builds. Let's see if CI tests verify the solution (local builds already pass).