selector: convert selector module to use base config extension

[aiChaoSONG]

This patch converts the selector module to use base config extension for module init, which ensures only base_cfg and base_cfg_ext is received in module init, and align the init blob format with other modules.

Signed-off-by: Chao Song [email protected]

The selector module is using the module adapter interface, and it has one input and one output, the input format(base_cfg) is already handled with module adapter, we only need to handle the output format.
@ranj063 I suppose that the nb_input_pins/nb_output_pins in base_cfg_ext could be zero, this help us to save some memory. If it is not the case, please let me know, I will also add the input pin format in comp_data (through the input format will never be used in the fw).

[lgirdwood]

@ujfalusi @ranj063 pls review

[ujfalusi]

@aiChaoSONG, looks good!

[lgirdwood]

@ujfalusi good now ?

[aiChaoSONG]

Let's hold on merge until Libin help to verify the patch, the rimage patch is needed for the verification: thesofproject/rimage#134

[libinyang]

Let's hold on merge until Libin help to verify the patch, the rimage patch is needed for the verification: thesofproject/rimage#134

After enabling the base config extension, WoV doesn't work. But it is not this PR's issue. Selector is initialized successfully after enabling base config extension.

[aiChaoSONG]

I am able to test wov feature with this PR + thesofproject/linux#4185 + thesofproject/rimage#135 + #7106 without any hack.

@ranj063 @ujfalusi ping

[ujfalusi]

@aiChaoSONG, looks good to my eyes, thank you.

[kv2019i]

Looks good, no need to update for the whitespace issue if tests pass and other reviews are ok.

[ranj063]

@aiChaoSONG the internal CI tests are failing. Can you please check?

[aiChaoSONG]

@ranj063 Looks like the quickbuild CI issue, let me retrigger

11:37:09,286 ERROR - Step 'master>SCANS>TESTS>TRIGGER PYTHON SMOKE_IPC4 TESTS' is failed: Step is failed since the triggered build is failed, cancelled, or timed out.

[aiChaoSONG]

SOFCI TEST

[aiChaoSONG]

@wszypelt Internal CI trigger Python IPC4 smoke test failed, could you please help to take a look.

[aiChaoSONG]

Update: rebased on the latest main

[kv2019i]

test_003_01_kd_topology_host_gw test is failing. @aiChaoSONG @wszypelt is this related to PR or unrelated failure?

[aiChaoSONG]

test_003_01_kd_topology_host_gw test is failing. @aiChaoSONG @wszypelt is this related to PR or unrelated failure?

@kv2019i @wszypelt selector module is use in the wov topology, this patch changed the payload for selector, I am afraid the test case needs to be modified accordingly.

[wszypelt]

@aiChaoSONG @kv2019i I'll take a look at it, but recently the KD tests have been changed to meet the new requirements

[lgirdwood]

SOFCI TEST

[lgirdwood]

@aiChaoSONG pls check internal CI if failure persists.

[lgirdwood]

@aiChaoSONG ok, failed again, can you check internal CI.

[aiChaoSONG]

@wszypelt I have changed the payload for selector in this patch, but quickbuild python tests still construct selector payload in the old way, so I think we need to modify the test case accordingly. Could you or someone please help to do this?
Below is the payload foramt for selector initialization:

struct ipc4_base_module_extended_cfg {
	struct ipc4_base_module_cfg base_cfg;
	struct ipc4_base_module_cfg_ext base_cfg_ext;  // nb_input_pins == 1, nb_output_pins == 1
}

Below is the payload format for set_large_config:

struct ipc4_selector_coeffs_config {
	uint16_t rsvd0;
	uint16_t rsvd1;

	/** Mixing coefficients in Q10 fixed point format */
	int16_t coeffs[SEL_SINK_CHANNELS_MAX][SEL_SOURCE_CHANNELS_MAX];
};

[pblaszko]

This change is breaking compatibility to present IPC4 protocol.
Current MicSelect InitInstance structure is:
{
ipc4_base_module_cfg base_cfg;
ipc4_audio_format out_format;
}

[aiChaoSONG]

@ujfalusi @ranj063 @pblaszko After synced with Poland team, the payload base + output format still need to be supported, because Windows use it, and if we convert it directly to use base + base_ext, we will break the compatibility with windows. Howerver, if we don't do the conversion, the selector can not work on linux side, so does wov feature.
How about aiChaoSONG@2e123b5, which makes the selector be able to handle both init blob.

[lgirdwood]

@pblaszko @aiChaoSONG are you guys aligned now - do we need to make further changes ?

[aiChaoSONG]

do we need to make further change

I made a patch for the selector to support both base_config+output_format and base_config+base_config_extension initialization payload, aiChaoSONG@2e123b5, I'd like to see if Peter and Ranjani are good with this.
@ujfalusi @ranj063 Ping

[lgirdwood]

do we need to make further change

I made a patch for the selector to support both base_config+output_format and base_config+base_config_extension initialization payload, aiChaoSONG@2e123b5, I'd like to see if Peter and Ranjani are good with this. @ujfalusi @ranj063 Ping

Ok, we also need to see @pblaszko approval too.

[pblaszko]

do we need to make further change

I made a patch for the selector to support both base_config+output_format and base_config+base_config_extension initialization payload, aiChaoSONG@2e123b5, I'd like to see if Peter and Ranjani are good with this. @ujfalusi @ranj063 Ping

Ok, we also need to see @pblaszko approval too.

I'm OK with this approach. I wait for bringing changes to this PR.

[lgirdwood]

Ok, we also need to see @pblaszko approval too.

I'm OK with this approach. I wait for bringing changes to this PR.

@pblaszko ok, great - please approve yout review.