07:21 <narmstrong> xdarklight: you’re right the no_connector also breaks that

07:22 <narmstrong> xdarklight: the init code uses the internal connector (it’s not a pointer) thinking this connector will be used, so this should be reworked to be initialized differently

08:00 <xdarklight> narmstrong: do you have anything in mind on how to solve it? if not I can send an email to the drm maintainers asking for some advice

08:21 <narmstrong> xdarklight: no idea :-/ I’ll have a look at other hdmi/dp bridge to see how they do

08:24 <xdarklight> narmstrong: mtk_hdmi.c does this: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/mediatek/mtk_hdmi.c#L1396 - and then they use: https://elixir.bootlin.com/linux/latest/source/drivers/gpu/drm/mediatek/mtk_hdmi.c#L1638

08:25 <xdarklight> all other bridge drivers that I found seem to follow a similar-ish approach as dw-hdmi

08:26 <narmstrong> We shall then add a get_eld callback to the i2s & dma drivers and do the same

08:27 <narmstrong> xdarklight: I can throw a patch today

08:45 <xdarklight> narmstrong: if you have a way to also test it (it seems that I don't) then that would indeed be great

09:08 <narmstrong> xdarklight: this should do the trick https://github.com/superna9999/linux/commit/51de3699a0f6f26000b7072ddce98a07db3ae75b

09:10 <narmstrong> xdarklight: I can check if ELD is correctly reported, maybe chewitt could further

09:21 <xdarklight> narmstrong: mtk-hdmi.c uses memset(buf, 0, len); if there's no ELD. I don't know the alsa code well-enough but I'm curious if with your change we'd keep the old ELD exposed to userspace (while HDMI is disconnected - until a HDMI cable is connected again)

09:22 <xdarklight> but maybe it was like that before as well, I'm not sure :/

09:40 <narmstrong> I’ll have a look

09:42 <narmstrong> xdarklight: https://elixir.bootlin.com/linux/latest/source/sound/soc/codecs/hdmi-codec.c#L451

09:42 <narmstrong> It’s used like the dw hdmi dma

09:43 <narmstrong> So should we stop startup if hdmi not connected ? Probably not indeed

09:57 <xdarklight> narmstrong: hmm, indeed. maybe that's why mtk-hdmi uses memset (to avoid any errors)

12:15 <narmstrong> xdarklight: better https://github.com/superna9999/linux/commit/828c002903e8e4b663a411c26dab3de64cc808ab ?

12:18 <xdarklight> narmstrong: yep, from what I understand this should be functionally equivalent to what we had before

12:18 <xdarklight> narmstrong: thanks!

12:18 <narmstrong> OK, I'll send it to see what experts thinks about it !