[alsa-devel] [PATCH] ASoC: ti: Allocate dais dynamically for TDM and audio graph card

[Peter Ujfalusi]

Hi Tony,

On 12/02/2020 16.35, Tony Lindgren wrote:
> * Peter Ujfalusi <peter.ujfalusi at ti.com> [200212 08:02]:
>>
>>
>> On 11/02/2020 19.16, Tony Lindgren wrote:
>>> We can have multiple connections on a single McBSP instance configured
>>> with audio graph card when using TDM (Time Division Multiplexing). Let's
>>> allow that by configuring dais dynamically.
>>
>> It is still one DAI...
>> If you have multiple codec connected to the same I2S lines, but the
>> codecs communicate within different time slots, you still have one DAI
>> on the CPU side, but multiple codecs (codec DAIs) with different TDM slot.
> 
> OK so subject should say "dodec DAIs" then I guess?
> 
>>> See Documentation/devicetree/bindings/sound/audio-graph-card.txt and
>>> Documentation/devicetree/bindings/graph.txt for more details for
>>> multiple endpoints.
>>
>> See the example for 'Multi DAI with DPCM' in audio-graph-card.txt
>> The PCM3168a have 2 DAIs: playback and capture, but you can have
>> multiple endpoints within a DAI.
> 
> Yes this should follow the audio-graph-card.txt example. We end up with
> mcbsp3 dts node as below on droid4:
> 
> &mcbsp3 {
>         #sound-dai-cells = <0>;
>         pinctrl-names = "default";
>         pinctrl-0 = <&mcbsp3_pins>;
>         status = "okay";
> 
>         ports {
>                 mcbsp3_port: port at 0 {
>                         #address-cells = <1>;
>                         #size-cells = <0>;
> 
>                         cpu_dai3: endpoint at 0 {
>                                 reg = <0>;
>                                 dai-format = "dsp_a";
>                                 frame-master = <&cpcap_audio_codec1>;
>                                 bitclock-master = <&cpcap_audio_codec1>;
>                                 remote-endpoint = <&cpcap_audio_codec1>;
>                         };
> 
>                         cpu_dai_mdm: endpoint at 1 {
>                                 reg = <1>;
>                                 dai-format = "dsp_a";
>                                 frame-master = <&cpcap_audio_codec1>;
>                                 bitclock-master = <&cpcap_audio_codec1>;
>                                 remote-endpoint = <&mot_mdm6600_audio_codec0>;
>                         };
>                 };
>         };
> };

According to
Documentation/devicetree/bindings/sound/audio-graph-card.txt
it should be something like this:
&mcbsp3 {
        #sound-dai-cells = <0>;
        pinctrl-names = "default";
        pinctrl-0 = <&mcbsp3_pins>;
        status = "okay";

        ports {
		#address-cells = <1>;
		#size-cells = <0>;
                port at 0 {
			reg = <0>;

			cpu_dai3: endpoint at 0 {
				dai-format = "dsp_a";
				frame-master = <&cpcap_audio_codec1>;
				bitclock-master = <&cpcap_audio_codec1>;
				remote-endpoint = <&cpcap_audio_codec1>;
			};

			cpu_dai_mdm: endpoint at 1 {
				dai-format = "dsp_a";
				frame-master = <&cpcap_audio_codec1>;
				bitclock-master = <&cpcap_audio_codec1>;
				remote-endpoint = <&mot_mdm6600_audio_codec0>;
			};
		};
	};
};

If you span out dummy DAIs got dai1+ then how you will get anything
working via endpoint1+?
There will be no ops for McBSP, so it is not going to do anything...


> That is pretty much the same as the 'Multi DAI with DPCM' example, with
> dne dai, and multiple endpoints. I think we still have just one port
> for one i2s transport on the mcbsp :)
> 
> Does the above look as what you would expect based on the binding?

The audio-graph-card.txt example shows pcm3168a which have two DAIs,
one for playback and one for capture.

I guess Morimoto-san can explain if he carries out of tree patches to
get the described setup working on top of mainline...

But, no, based on the documentation I don't ;)

>>> I've tested this with droid4 where cpcap pmic and modem voice are both
>>> both wired to mcbsp3. I've also tested this on droid4 both with and
>>> without the pending modem audio codec driver that is waiting for n_gsm
>>> serdev dependencies to clear.
>>
>> What this patch you effectively just creating dummy-dais on top of the
>> real McBSP DAI.
> 
> Yes I think this is needed for snd-soc-audio-graph-card, and this allows
> configuring whatever is needed for the i2s slot. But maybe you have some
> better way of doing it in mind?
> 
>> You also rename the DAIs, which might break ams-delta.
> 
> Oops, that's not good. So should we just keep the old naming if there's
> only one endpoint?

That's an option, yes, if we really need extra dummy McBSP DAIs at all,
again, let's hear from Morimoto-san or Mark.
 
>> We still have legacy support in
>> omap-twl4030.c
>> omap3pandora.c
>> osk5912.c
>> rx51.c
>>
>> which will break with the renamed DAI. On the other hand I think the
>> legacy support can be dropped from them.
> 
> I'm not sure what all that would take.

For some it should not be a big deal as they only boot in DT mode.
/me adds this to the TODO list.
 
>> I know it was discussed, but can not find the mail:
>> Can you brief again on the audio connection?
> 
> Below is a link to a mailing list thread where Sebastian describes
> the audio connection:
> 
> https://lkml.org/lkml/2018/3/28/881

Thanks!
 
>> Do you have branch with working code?
> 
> Yeah I have slightly older set of the patches in my droid4-pending-v5.5
> kernel.org git branch with voice calls working.

I think I should put my droid4 out and try to get it working...
Do you have a link for dummies to follow to get started? ;)

> 
> Regards,
> 
> Tony
> 

- Péter

Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki