[plbossart]

This is an early POC to see if we can support a 'nocodec' mode for SoundWire.

The following PR is the companion of thesofproject/acpi-scripts#12, which registers pretend devices so that we can load a machine driver.

This loads a card but fails since transactions on the bus fail: there is nothing connected so that's quite normal.
I tested this with the sof-cml-rt711-rt1308-mono-rt715.tplg copied as sof-cnl-sdw-null.tplg
The case with the dual amps failed with a kernel oops, but that should be for another day.

Additional work is required to avoid sending any commands to the codec, or alternatively ignore -ENODATA results.

@ranj063 @bardliao this could be useful to debug driver/topology changes without having any actual device connected.

[plbossart]

@bardliao @ranj063 I figured out how to do a 'playback', the data seems to be consumed at the right rate (still an issue in the resume case)
If you have an UpExtreme you can start checking this out, but it looks promising.

[ranj063]

@bardliao @ranj063 I figured out how to do a 'playback', the data seems to be consumed at the right rate (still an issue in the resume case)
If you have an UpExtreme you can start checking this out, but it looks promising.

@plbossart this is awesome. Btw which topology do I use for testing?

[bardliao]

@plbossart @ranj063 I did a quick test with sof-cnl-sdw-null.m4.txt tplg.
But I got soundwire sdw-master-0: Slave frame_ctrl reg write failed in playback.
See sdw-nocodec-playback.txt

[bardliao]

BTW, there is no 0xded0 id on my Up Xtreme board. So I use 0x0700 instead.

[bardliao]

@plbossart @ranj063 I did a quick test with sof-cnl-sdw-null.m4.txt tplg.
But I got soundwire sdw-master-0: Slave frame_ctrl reg write failed in playback.
See sdw-nocodec-playback.txt

I can work around it by

diff --git a/drivers/soundwire/bus.c b/drivers/soundwire/bus.c
index ef7b7b976b0b..83ed0c4b8ab4 100644
--- a/drivers/soundwire/bus.c
+++ b/drivers/soundwire/bus.c
@@ -215,7 +215,7 @@ static inline int do_transfer(struct sdw_bus *bus, struct sdw_msg *msg)

                /* if cmd is ok or ignored return */
                if (ret == 0 || ret == -ENODATA)
-                       return ret;
+                       return 0;
        }

        return ret;

Now, I can run about 2 sec playback and end up with IO error.
dmesg_sdw_nocodec_playback.txt

[plbossart]

@plbossart @ranj063 I did a quick test with sof-cnl-sdw-null.m4.txt tplg.

let's use an existing topology for now.

But I got soundwire sdw-master-0: Slave frame_ctrl reg write failed in playback.
See sdw-nocodec-playback.txt

I can work around it by

diff --git a/drivers/soundwire/bus.c b/drivers/soundwire/bus.c
index ef7b7b976b0b..83ed0c4b8ab4 100644
--- a/drivers/soundwire/bus.c
+++ b/drivers/soundwire/bus.c
@@ -215,7 +215,7 @@ static inline int do_transfer(struct sdw_bus *bus, struct sdw_msg *msg)

                /* if cmd is ok or ignored return */
                if (ret == 0 || ret == -ENODATA)
-                       return ret;
+                       return 0;
        }

        return ret;

Now, I can run about 2 sec playback and end up with IO error.
dmesg_sdw_nocodec_playback.txt

BTW, there is no 0xded0 id on my Up Xtreme board. So I use 0x0700 instead.

See above, this assumes that the ACPI table in thesofproject/acpi-scripts#12 is used.

I used the usual CML mono topology and copied it over. If I used the two amp version there is a kernel oops that I will need to debug with @ranj063
We will need a new topology with more streams but for now I am only trying to replace the existing topologies and make them work, but without any codec connected.

[plbossart]

@bardliao @RanderWang @ranj063 I updated the code to work with CML topologies, we can either use the 1amp or the 2-amp aggregated mode (that part doesn't work due to a regression in topic/sof-dev).

Comments and feedback welcome.

[ranj063]

@bardliao @RanderWang @ranj063 I updated the code to work with CML topologies, we can either use the 1amp or the 2-amp aggregated mode (that part doesn't work due to a regression in topic/sof-dev).

Comments and feedback welcome.

@plbossart this fix works for me #2832

[bardliao]

@plbossart Now I can do playback/capture with below patch. Note that capture can only get silence data as expected.

diff --git a/drivers/soundwire/stream.c b/drivers/soundwire/stream.c
index 1eaedaaba094..92fc033ad246 100644
--- a/drivers/soundwire/stream.c
+++ b/drivers/soundwire/stream.c
@@ -133,6 +133,8 @@ static int sdw_program_slave_port_params(struct sdw_bus *bus,
        int ret;
        u8 wbuf;

+       if (s_rt->slave->is_null_device)
+               return 0;
        dpn_prop = sdw_get_slave_dpn_prop(s_rt->slave,
                                          s_rt->direction,
                                          t_params->port_num);
@@ -698,7 +700,7 @@ static int sdw_bank_switch(struct sdw_bus *bus, int m_rt_count)
        else
                ret = sdw_transfer(bus, wr_msg);

-       if (ret < 0) {
+       if (ret < 0 && ret != -ENODATA) {
                dev_err(bus->dev, "Slave frame_ctrl reg write failed\n");
                goto error;
        }

[plbossart]

@plbossart Now I can do playback/capture with below patch. Note that capture can only get silence data as expected.

@bardliao I don't get why I am able to play/record without your patch. @ranj063 also tested yesterday. Weird.

[bardliao]

@plbossart Now I can do playback/capture with below patch. Note that capture can only get silence data as expected.

@bardliao I don't get why I am able to play/record without your patch. @ranj063 also tested yesterday. Weird.

@plbossart I will get below error without my patch.

[  127.267120] soundwire sdw-master-0: Slave frame_ctrl reg write failed
[  127.267192] soundwire sdw-master-0: Bank switch failed: -61
[  127.267205] soundwire sdw-master-0: Bank switch failed: -61
[  127.267217]  SDW0-Playback: ASoC: error at snd_soc_link_prepare on SDW0-Playback: -61
[  127.267234]  SDW0-Playback: ASoC: soc_pcm_prepare() failed (-61)
[  127.267247]  Jack Out: ASoC: dpcm_be_dai_prepare() failed (-61)
[  127.267260]  Jack Out: ASoC: dpcm_fe_dai_prepare() failed (-61)

See dmesg_sdw_nocodec.txt

[RanderWang]

great! we can even add loop-back feature for nocodec mode, just set the offset of playback & capture with the same value in frame.

[RanderWang]

commit message "a) For some historical reason, we set this CMD_ACCEPT bitfield during
the initialization, but we don't during a resume from a clock-stoppped
state." ?

[plbossart]

commit message "a) For some historical reason, we set this CMD_ACCEPT bitfield during
the initialization, but we don't during a resume from a clock-stoppped
state." ?

I meant "I do not know why we are doing things differently for probe and resume". The reason is lost in history.

[plbossart]

@bardliao @ranj063 with the update I can test the mockup even on a device with existing codecs. The mockups are added first in the ACPI match tables, so when they are present they will be detected and a different topology chosen.

The only think that will not work is that we cannot enable mockup devices on a link that is not enabled physically in the hardware. We would still try to transmit commands and that could generate conflicts at the PHY level. In other words, we can discard existing codecs or add new codecs on enabled links, not change which links are enabled.

[lyakh]

at least the '-' wants to be fixed, not sure if we find a better solution for the header. Maybe wait for some more opinions

[plbossart]

@bardliao I'll let you merge if you approve.

[bardliao]

@plbossart The code looks good to me. I am just wondering should if (value > pdi_out_num) belong to 98cf1cd?

[RanderWang]

LGTM

[bardliao]

Thanks @plbossart I will merge it when we have second approval.