ASoC: Question regarding device-tree multi-lane I2S for Allwinner SoC

ASoC: Question regarding device-tree multi-lane I2S for Allwinner SoC

[Clément Péron]

Hi device-tree and sound Maintainers,

I have a question regarding multi-lane i2S representation.

On the Allwinner SoC the I2S/PCM Interface can handle up to four lanes
as output or input.
For each lane we can enable up to sixteen slots.
And for each output slot we can choose which slot to map.

The only representation I found is for Amlogic device-tree they did
the following sound node :
/* 8ch hdmi interface */
dai-link-7 {
    sound-dai = <&tdmif_b>;
    dai-format = "i2s";
    dai-tdm-slot-tx-mask-0 = <1 1>;
    dai-tdm-slot-tx-mask-1 = <1 1>;
    dai-tdm-slot-tx-mask-2 = <1 1>;
    dai-tdm-slot-tx-mask-3 = <1 1>;
    mclk-fs = <256>;

    codec {
        sound-dai = <&tohdmitx TOHDMITX_I2S_IN_B>;
    };
};

This kind of representation gives the information that 2 slots should
be enabled as TX per lane but don't give which slot to map.
I was thinking about a representation per lane but maybe it's a bit
complicated  ?

    dai-format = "dsp_a";
    dai-tdm-slot-width = <32>;
    // Lane 0 : Output 8 channels 0-7 using TDM
    dai-tdm-slot-tx-mask-0 = <1 1 1 1 1 1 1 1>;
    dai-tdm-slot-chmap-0 = <0 1 2 3 4 5 6 7>;
    // Lane 1 : Output 3 channels 5-7 using TDM
    dai-tdm-slot-tx-mask-1 = <1 1 1>;
    dai-tdm-slot-chmap-1 = <5 6 7>;

I will only start to support HDMI, so 4 I2S lanes for now it should
look like this

    dai-format = "i2s";
    dai-tdm-slot-width = <32>;
    frame-inversion;
    // Lane 0 : Output channels 0,1
    dai-tdm-slot-tx-mask-0 = <1 1>;
    dai-tdm-slot-chmap-0 = <0 1>;
    // Lane 1: Output channels 2,3
    dai-tdm-slot-tx-mask-1 = <1 1>;
    dai-tdm-slot-chmap-1 = <2 3>;
    // Lane 2: Output channels 4,5
    dai-tdm-slot-tx-mask-2 = <1 1>;
    dai-tdm-slot-chmap-2 = <4 5>;
    // Lane 3: Output channels 6,7
    dai-tdm-slot-tx-mask-3 = <1 1>;
    dai-tdm-slot-chmap-3 = <6 7>;

What do you think? Do you have any remark / idea about this ?

Thanks for your help
Clement

Re: ASoC: Question regarding device-tree multi-lane I2S for Allwinner SoC

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 2392 bytes --]

On Sun, Nov 01, 2020 at 06:30:15PM +0100, Clément Péron wrote:

Sorry about the delay here.

>     dai-tdm-slot-tx-mask-0 = <1 1>;
>     dai-tdm-slot-tx-mask-1 = <1 1>;
>     dai-tdm-slot-tx-mask-2 = <1 1>;
>     dai-tdm-slot-tx-mask-3 = <1 1>;

...

> This kind of representation gives the information that 2 slots should
> be enabled as TX per lane but don't give which slot to map.

The way the above should be interpreted is as a bitmask of slots to use,
I'm hoping that the above is for a system with 4 TX data wires each
using two slots (such designs get used for things like surround sound
amps).  

> I was thinking about a representation per lane but maybe it's a bit
> complicated  ?

>     dai-format = "dsp_a";
>     dai-tdm-slot-width = <32>;
>     // Lane 0 : Output 8 channels 0-7 using TDM
>     dai-tdm-slot-tx-mask-0 = <1 1 1 1 1 1 1 1>;
>     dai-tdm-slot-chmap-0 = <0 1 2 3 4 5 6 7>;
>     // Lane 1 : Output 3 channels 5-7 using TDM
>     dai-tdm-slot-tx-mask-1 = <1 1 1>;
>     dai-tdm-slot-chmap-1 = <5 6 7>;

I'd expect lane 1 to end up as

     dai-tdm-slot-tx-mask-1 = < 0 0 0 0 0 1 1 1 >;

>     // Lane 0 : Output channels 0,1
>     dai-tdm-slot-tx-mask-0 = <1 1>;
>     dai-tdm-slot-chmap-0 = <0 1>;
>     // Lane 1: Output channels 2,3
>     dai-tdm-slot-tx-mask-1 = <1 1>;
>     dai-tdm-slot-chmap-1 = <2 3>;
>     // Lane 2: Output channels 4,5
>     dai-tdm-slot-tx-mask-2 = <1 1>;
>     dai-tdm-slot-chmap-2 = <4 5>;
>     // Lane 3: Output channels 6,7
>     dai-tdm-slot-tx-mask-3 = <1 1>;
>     dai-tdm-slot-chmap-3 = <6 7>;

> What do you think? Do you have any remark / idea about this ?

I can see the use for the chmap binding, I can see that we might have a
system which has for example outputs labelled in some unusual order and
we'd want to remap them.  I'm less sure about defining a channel map in
a way that changes the meaning of the masks though, that seems like it'd
lead to incompatibilities.  How about making chmap also an array like
the mask is with the bits set in the mask indicating which slots in the
chmap are valid, that way your -3 would end up as:

      dai-tdm-slot-tx-mask-3 = <0 0 0 0 0 0 0 1 1>;
      dai-tdm-slot-chmap-3   = <0 0 0 0 0 0 0 6 7>;

and something that only understands the mask would at least get the
correct channels even if in a jumbled order?

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]