Dne sreda, 31. julij 2019 ob 14:29:53 CEST je Maxime Ripard napisal(a): > On Tue, Jul 30, 2019 at 07:57:10PM +0200, Jernej Škrabec wrote: > > Dne torek, 04. junij 2019 ob 11:38:44 CEST je Code Kipper napisal(a): > > > On Tue, 4 Jun 2019 at 11:02, Christopher Obbard <chris@xxxxxxxxxxxx> wrote: > > > > On Tue, 4 Jun 2019 at 09:43, Code Kipper <codekipper@xxxxxxxxx> wrote: > > > > > On Tue, 4 Jun 2019 at 09:58, Maxime Ripard > > > > > <maxime.ripard@xxxxxxxxxxx> > > > > wrote: > > > > > > On Mon, Jun 03, 2019 at 07:47:32PM +0200, codekipper@xxxxxxxxx wrote: > > > > > > > From: Marcus Cooper <codekipper@xxxxxxxxx> > > > > > > > > > > > > > > The i2s block supports multi-lane i2s output however this > > > > > > > functionality > > > > > > > is only possible in earlier SoCs where the pins are exposed and > > > > > > > for > > > > > > > the i2s block used for HDMI audio on the later SoCs. > > > > > > > > > > > > > > To enable this functionality, an optional property has been > > > > > > > added to > > > > > > > the bindings. > > > > > > > > > > > > > > Signed-off-by: Marcus Cooper <codekipper@xxxxxxxxx> > > > > > > > > > > > > I'd like to have Mark's input on this, but I'm really worried > > > > > > about > > > > > > the interaction with the proper TDM support. > > > > > > > > > > > > Our fundamental issue is that the controller can have up to 8 > > > > > > channels, but either on 4 lines (instead of 1), or 8 channels on 1 > > > > > > (like proper TDM) (or any combination between the two, but that > > > > > > should > > > > > > be pretty rare). > > > > > > > > > > I understand...maybe the TDM needs to be extended to support this to > > > > > consider channel mapping and multiple transfer lines. I was thinking > > > > > about the later when someone was requesting support on IIRC a while > > > > > ago, I thought masking might of been a solution. These can wait as > > > > > the > > > > > only consumer at the moment is LibreELEC and we can patch it there. > > > > > > > > Hi Marcus, > > > > > > > > FWIW, the TI McASP driver has support for TDM & (i think?) multiple > > > > transfer lines which are called serializers. > > > > Maybe this can help with inspiration? > > > > see > > > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tre > > > > e/s > > > > ound/soc/ti/davinci-mcasp.c sample DTS: > > > > > > > > &mcasp0 { > > > > > > > > #sound-dai-cells = <0>; > > > > status = "okay"; > > > > pinctrl-names = "default"; > > > > pinctrl-0 = <&mcasp0_pins>; > > > > > > > > op-mode = <0>; > > > > tdm-slots = <8>; > > > > serial-dir = < > > > > > > > > 2 0 1 0 > > > > 0 0 0 0 > > > > 0 0 0 0 > > > > 0 0 0 0 > > > > > > > > > >; > > > > > > > > tx-num-evt = <1>; > > > > rx-num-evt = <1>; > > > > > > > > }; > > > > > > > > Cheers! > > > > > > Thanks, this looks good. > > > > I would really like to see this issue resolved, because HDMI audio support > > in mainline Linux for those SoCs is long overdue. > > > > However, there is a possibility to still add HDMI audio suport (stereo > > only) even if this issue is not completely solved. If we agree that > > configuration of channels would be solved with additional property as > > Christopher suggested, support for >2 channels can be left for a later > > time when support for that property would be implemented. Currently, > > stereo HDMI audio support can be added with a few patches. > > > > I think all I2S cores are really the same, no matter if internally > > connected to HDMI controller or routed to pins, so it would make sense to > > use same compatible for all of them. It's just that those I2S cores which > > are routed to pins can use only one lane and >2 channels can be used only > > in TDM mode of operation, if I understand this correctly. > > > > New property would have to be optional, so it's omission would result in > > same channel configuration as it is already present, to preserve > > compatibility with old device trees. And this mode is already sufficient > > for stereo HDMI audio support. > > Yeah, it looks like a good plan. > > > Side note: HDMI audio with current sun4i-i2s driver has a delay (about a > > second), supposedly because DW HDMI controller automatically generates CTS > > value based on I2S clock (auto CTS value generation is enabled per > > DesignWare recomendation for DW HDMI I2S interface). > > Is that a constant delay during the audio playback, or only at startup? I think it's just at startup, e.g. if you're watching movie, audio is in sync, it's just that first second or so is silent. > > > I2S driver from BSP Linux solves that by having I2S clock output > > enabled all the time. Should this be flagged with some additional > > property in DT? > > I'd say that if the codec has that requirement, then it should be > between the codec and the DAI, the DT doesn't really have anything to > do with this. Ok, but how to communicate that fact to I2S driver then? BSP driver solves that by using different compatible, but as I said before, I2S cores are not really different, so this seems wrong. Best regards, Jernej _______________________________________________ Alsa-devel mailing list Alsa-devel@xxxxxxxxxxxxxxxx https://mailman.alsa-project.org/mailman/listinfo/alsa-devel
