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):

quoted

On Tue, 4 Jun 2019 at 11:02, Christopher Obbard [off-list ref] wrote:

quoted

On Tue, 4 Jun 2019 at 09:43, Code Kipper [off-list ref] wrote:

quoted

On Tue, 4 Jun 2019 at 09:58, Maxime Ripard [off-list ref]

wrote:

quoted

quoted

quoted

quoted

On Mon, Jun 03, 2019 at 07:47:32PM +0200, codekipper@gmail.com wrote:

quoted

From: Marcus Cooper <redacted>

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 <redacted>

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/tree/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?

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.

Maxime

--
Maxime Ripard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel