13:18 <mort> does anyone know anything about i2s on the rk3588?

13:19 <mort> I *think* I enabled i2s1 in the devicetree? But there's still only one card in alsamixer. But that card is called "card 0: rockchiphdmi0 [rockchip-hdmi0], device 0: rockchip-hdmi0 i2s-hifi-0 [rockchip-hdmi0 i2s-hifi-0]"

13:19 <mort> how can card 0's device 0 be both rockchip-hdmi0 and i2s-hifi-0

14:20 <wens> a card can have multiple pcm streams

14:22 <wens> plus, you would need to add descriptions to DT on how to tie i2s1 to a codec, using something like simple-audio-card. then you would get another card in the system

14:26 <wens> minute: all PCIe peripherals (both ethernet controllers, M.2 E key, M.2 B key) work fine for me on Orange Pi 5 Plus on next-20240215

14:53 <mort> wens: that sounds like exactly what I'm missing, but I don't know how I'd go about that

14:58 <mort> there's an hdmi0-sound node which refers to the hdmi@fde80000 node as its 'rockchip,codec', I guess I need something similar which refers to the i2s@fe480000 node as its codec

15:00 <mort> ...but I also have no idea how to really do that, so help would be greatly appreciated

15:31 <wens> mort: something like this maybe? https://github.com/wens/dt-overlays/blob/master/bpi-m3-pifi-dac-plus.dts

15:31 <wens> adjust for your codec

15:32 <mort> I don't think I understand exactly how this audio stuff fits together, e.g what a "codec" is and what a "dai" is and what a "card" is

15:32 <mort> or what a "cpu" is

15:38 <mort> but okay, in your case, the i2s node is the "cpu" and the i2c node contains the "codec"

15:43 <mort> wens: I don't understand why you're setting codec's compatible to ti,pcm5122 or its reg to 0x4d

15:46 <mort> isn't it a bit weird that something as basic as "output sound on this linux computer" requires digging so extremely deep into devicetree stuff?

15:47 <wens> mort: well, i2s is just a digital audio interface. you need a codec to convert that to analog

15:48 <mort> that's an external dac connected via i2s, which gets configured via SPI in userspace

15:48 <mort> I literally just wanna send digital sound over the i2s gpio pins

15:48 <CounterPillow> i2s is a non-self-enumerating bus so yes device tree is a requirement, otherwise how would you describe the hardware?

15:49 <wens> mort: as far as ALSA knows, I2S is only part of a card, you need to give it the other parts

15:49 <mort> CounterPillow: right but I just want to say "send digital sound via i2s1", I get that some level of devicetree stuff is required for that but this seems like a lot

15:50 <mort> wens: can I give it "dummy" versions of those other cards somehow..?

15:50 <mort> things like volume control will also happen via spi in userspace

15:50 <wens> for now, just use "ti,pcm5102a" for the codec. It's a dumb codec.

15:52 <mort> and the reg?

15:52 <mort> also you put it as a sub-node under i2c2, where am I supposed to put it?

15:52 <wens> no reg. and move that node up to the root node, at the same level as the audio card

15:52 <mort> alright

15:53 <wens> for the kernel, you'll need CONFIG_SND_SOC_PCM5102A and CONFIG_SND_SIMPLE_CARD for those two

15:55 <mort> SND_SIMPLE_CARD is enabled, SND_SOC_PCM5102A isn't even there (it was presumably added after 5.10?)

15:57 <mort> very weird that there's no way to tell alsa, "just send digital audio data through this digital audio interface you know exists"

16:03 <wens> it could be blocked behind some other option

16:05 <mort> it's not like we need the PCM5102A anyway right, we just need to find another codec driver which doesn't mess stuff up too badly by being too smart

16:06 <mort> how about the TS3A227E, that's =y

16:08 <mort> or SND_SOC_DUMMY_CODEC sounds relevant?

16:10 <wens> I've not tried anything else

16:29 <mort> wens: why does your overlay work even though https://github.com/torvalds/linux/blob/master/Documentation/devicetree/bindings/sound/pcm512x.txt claims that there are 3 required power supply related properties that you didn't provide?

16:37 <mort> ok hang on

16:38 <mort> I've spent the past 20 minutes trying to find a codec driver that's 1) enabled in my kernel config and 2) stupid enough to work even though the thing it's driving doesn't actually exist

16:38 <mort> and the only reason I need this is to trick Linux into letting me send digital audio data through the digital audio interface that Linux already knows exists

16:39 <mort> this *can't* be the right solution, there *has* to be a better way

17:22 <mort> so, realistically, what do I do

17:26 <minute> wens: thank you very much, i also wrote here that i made it work by feeding another refclk into port1 of pcie3. i only had one fed into port0 but it looks like the IP needs both to work

19:46 <mort> wens: alright so I found the linux,spdif-dit codec which is supposed to be pretty dumb. I now have this in the devicetree: https://p.mort.coffee/Xva and a fe480000.i2s-dit-hifi card is now available

19:46 <mort> however, when I play audio through it, only the sclk pin has data on it; none of the others seem to do anything

19:46 <mort> not the dto pins, not the lrclk pin