HDMI output only works on single channel on Up2

[plbossart]

When trying the latest kernel code (topic/sof-dev) along with sof/soft master, I only get sound on a single channel.

The mixer configuration is this (need to use pin6-port0 with cvt2)
amixer -Dhw:0 cset numid=28 1

and command line
speaker-test -Dhw:0,1 -c2 -r48000 -t sine

Adding as a kernel bug for now until it's sorted out.

[keyonjie]

this should be PGA default value issue, the volume of Left channel is set to 0 by default wrongly.

[plbossart]

@keyonjie what do you mean? Is this a known topology issue? or a driver issue?

[cujomalainey]

I noticed this in chrome is as well. I assumed it was a ucm issue

[ranj063]

@cujomalainey do you still have this issue on chrome? I ran into this issue earlier but couldnt reproduce it after I used the updated ucm

[cujomalainey]

I'll try and reproduce it if I have time tomorrow, I usually trigger it when I'm building the 4.18 kernel

[keyonjie]

@keyonjie what do you mean? Is this a known topology issue? or a driver issue?

I believe there is bug here, on Yorp, ucm file may hide this bug.

[plbossart]

@mengdonglin can you please assign someone to look into this and check on the validation side? Looks like a blocking P1 to me?

[cujomalainey]

@ranj063 I am unable to reproduce it currently as my patch doesn't appear to be working anymore and therefore I am getting no devices populated. It will take me some time to debug.

[markyang]

Summary:
This issue cannot be reproduced using the mixer configuration as follows:
pin5-port0 : cvt2
pin6-port0 : cvt3
pin7-port0 : cvt4

Test Steps:

1. aplay -Dhw:0,2 -r 48000 -c 2 -f s16_le -i -vv ~/Music/18_WAV_48000Hz_stereo_16bit.wav

Test env:
sof master: 43da169
sof tool: 937032f
kernel sof-dev: a5ddc3d
tplg: sof-apl-pcm512x.tplg-937032f

[juimonen]

I can't play to HDMI: aplay: set_params:1363: Unable to install hw params:
Even though I try to set the pins with alsamixer (?)
kernel: a5ddc3d
sof: 43da169
tplg: sof-apl-pcm512x.tplg

[plbossart]

I think it's a topology issue, all the HDAudio streams have the left channel muted:

numid=21,iface=MIXER,name='PGA1.0 Master Playback Volume 1'
  ; type=INTEGER,access=rw---R--,values=2,min=0,max=32,step=0
  : values=32,32
  | dBscale-min=-64.00dB,step=2.00dB,mute=1
root@ubilinux4:~# amixer -Dhw:0 cset numid=22
numid=22,iface=MIXER,name='PGA2.0 Master Playback Volume 2'
  ; type=INTEGER,access=rw---R--,values=2,min=0,max=32,step=0
  : values=0,32
  | dBscale-min=-64.00dB,step=2.00dB,mute=1
root@ubilinux4:~# amixer -Dhw:0 cset numid=23
numid=23,iface=MIXER,name='PGA3.0 Master Playback Volume 3'
  ; type=INTEGER,access=rw---R--,values=2,min=0,max=32,step=0
  : values=0,32
  | dBscale-min=-64.00dB,step=2.00dB,mute=1
root@ubilinux4:~# amixer -Dhw:0 cset numid=24
numid=24,iface=MIXER,name='PGA4.0 Master Playback Volume 4'
  ; type=INTEGER,access=rw---R--,values=2,min=0,max=32,step=0
  : values=0,32
  | dBscale-min=-64.00dB,step=2.00dB,mute=1

I absolutely didn't make this selection so it's coming from the topology or the driver. @ranj063 does this ring a bell?

[mengdonglin]

@plbossart the platform topology m4 file doesn't set the initial mixer settings for the PGA: https://github.com/thesofproject/soft/blob/master/topology/sof-apl-pcm512x.m4

And it's strange this only happens to HDMI streams, since the HDA DAI configuration m4 also doesn't set the initial value, please see: https://github.com/thesofproject/soft/blob/master/topology/m4/hda.m4

So maybe from driver side?
I tried to use the open discussion https://github.com/orgs/thesofproject/teams/sof-developers/discussions/1 to see if someone can share some tips.

[plbossart]

@ranj063 do you happen to know the mixer defaults are defined for PGA values? I just don't get how in multiple cases the max is selected - expect for the left channel.

[juimonen]

@mengdonglin I think this is not only HDMI, I've seen this with basic codec path also and I think @plbossart reported it previously in some other issue. I'm not sure do I get it with updating the kernel, sof or the whole image... (because once you set it, alsa remembers it). I can try to hunt this down, if @ranj063 doesn't immediately know the problem spot, but it will take me some time as I have to trace the control creation path in kernel.

[lgirdwood]

@juimonen also check that pusleaudio is not doping anything daft. Best to rename /usr/bin/pulseaudio to another name, delete old alsa state file then reboot to rule out userspace.

[juimonen]

@lgirdwood yep, it is not pulseaudio. This can be reproduced by renaming also alsactl. I was trying to disable it "nicely" by fiddling with systemd, but this is the easiest way. So with "first boot" all our pga volumes
are indeed 0 on the left channel (and right channel is max).

[juimonen]

actually I see this in firmware log:
0 2 unknown 74888.697917 4.166667 volume.c:387 value 0
0 2 unknown 74890.989583 2.291667 volume.c:388 value 0
0 2 unknown 74893.333333 2.343750 volume.c:387 value 1
0 2 unknown 74895.625000 2.291667 volume.c:388 value 65536

So channel 0 -> 0 volume and channel 1 -> max

[lgirdwood]

@juimonen ok, looks like bug in volume.c

[juimonen]

yep, this is the set before get:

0 2 unknown 73280.885417 2.864583 volume.c:320 value 0
0 2 unknown 73283.645833 2.760417 volume.c:321 value 0
0 2 unknown 73286.562500 2.916667 volume.c:320 value 0
0 2 unknown 73289.114583 2.552083 volume.c:321 value 0

so same channel 0 is set twice to 0. channel 1 is left untouched so it has max.

[juimonen]

and that means it is coming from kernel side with ipc ... so someone is setting it to 0 but with empty data-> as a consequence only other channel is set. And that means that default pga value is max (which is correct).

[juimonen]

and it seams it is the restore of controls that sets the value incorrectly.
sorry for dumping the debug status every minute, will stop now.

[juimonen]

#258

[plbossart]

@juimonen it's ok to share your findings live, it helps the rest of the team follow what might be happening. Thanks for looking into this!

[plbossart]

closed with PR258