Latest sofproject driver (untested)

[maaarghk]

No description provided.

[maaarghk]

i'm not going to merge, just telling people to git checkout latest-sofproject if on 5.8

[szbergeron]

Have been testing this for the past day or so.
Issues observed:

• Sometimes power usage locks itself at about 18w, seems new with this driver. CPU usage isn't high or anything, power draw just goes up. Haven't done a proper A/B test with/without the driver though so haven't fully confirmed this one.
• Headphone plug/unplug doesn't switch sinks, need to change what the fallback sink is from pavucontrol to switch

Otherwise all seems to be working with a minimum of fuss. Using on arch didn't require using the git versions of anything pulseaudio, but perhaps this is the source of the earlier issues.

[maaarghk]

Hey, how you measuring power usage? I will check it on my machine. When you say with this driver do you mean you were using the 5.7 version before?

[szbergeron]

I've been measuring using powertop and some other power management tools I've made that read battery drain. I was not using any driver before, I had just resigned myself to using a dongle whenever I needed audio

[divoxx]

I am currently trying to build this branch against 5.8.13, attempting to get my Dell XPS 9700 to work and following this post: https://blog.fts.scot/2020/07/04/dell-xps-2020-how-to-get-audio-working-on-linux/

After trying to build the module using dkms build ..., it fails due to this error:

ERROR: modpost: "sdw_startup_stream" [/var/lib/dkms/soundwire/1.4.0/build/soc/intel/boards/snd-soc-sof-sdw.ko] undefined!
ERROR: modpost: "sdw_shutdown_stream" [/var/lib/dkms/soundwire/1.4.0/build/soc/intel/boards/snd-soc-sof-sdw.ko] undefined!
ERROR: modpost: "sdw_bread_no_pm_unlocked" [/var/lib/dkms/soundwire/1.4.0/build/soundwire/soundwire-cadence.ko] undefined!
ERROR: modpost: "sdw_rows" [/var/lib/dkms/soundwire/1.4.0/build/soundwire/soundwire-generic-allocation.ko] undefined!
ERROR: modpost: "sdw_cols" [/var/lib/dkms/soundwire/1.4.0/build/soundwire/soundwire-generic-allocation.ko] undefined!

The full log is available here: make.log

I see those functions defined in my local copy of the kernel tree, so I think I need to enable something so that they are compiled and available to be linked?

Sorry for the support like request/comment, but also wanted to bring up to your attention in case it's something that needs to be addressed.

Edit: hit enter too soon, added more details after posting

[szbergeron]

@divoxx did you install sof-firmware? I seem to remember something similar before when I forgot to install it

[fjalvingh]

Thanks a lot for your effort. I got my sound to work using this version of the driver on my xps 9700.

[maaarghk]

Hi @fjalvingh - are you using sof-firmware 1.5.1 or 1.6? I have issues with 1.6 only but trying to rule out a hardware problem :)

[fjalvingh]

Hi @fjalvingh - are you using sof-firmware 1.5.1 or 1.6? I have issues with 1.6 only but trying to rule out a hardware problem :)

I use 1.5.1.

[maaarghk]

Ok, I suggest to stay on it for now :) Would you mind if I @'ed you and requested you to upgrade to 1.6 in the future? It might or might not be required, depending how tracking down this issue goes.

[szbergeron]

@maaarghk did you have weird crackling start with that update and silence on the headphone jack?

[maaarghk]

No, everything works perfectly but intermittently the left channel will become distorted.

[szbergeron]

I think I had something slightly more severe but similar, only some weird crackling and distortion on the left side. Do you set any configs in pulse for sample rate/format or around realtime scheduling? I just commented everything in .pulse/daemon.conf that I had that was correcting for an incorrect default sample rate on my fiio btr3, and that fixed it for me.

[maaarghk]

I did not make any changes like that, so not sure if there's anything I can try there...

[szbergeron]

Nvm apparently changing those settings just made it less common, it just happened for me again lol

[fjalvingh]

@maaarghk I won't mind trying an update to 1.6, I can always get back to 1.5.1 if I needed to. I'm not using sound that much on the laptop though, so I will not notice issues that take long to appear..

[divoxx]

@szbergeron sorry for the delay responding to you. Yes, I have sof-firmware 1.5.1 installed

[maaarghk]

@divoxx you probably have a kernel config option disabled that needs enabled, if you see those functions exported they are probably inside #ifdefs

[szbergeron]

Has anyone tested if this PR works with 5.9? Arch has updated core to push 5.9 as of today.

[mark-hayward]

This does work on 5.9 but my speakers crackle and sound very tinny. They seem ok for a short while after a reboot but then it quickly becomes unlistenable. When I play audio via an HDMI output it works fine but the laptop speaker is terrible. This didn't happen on the old kernel.

[maaarghk]

I'm on 5.9.1 and it works fine. Do you have the latest firmware? Try downgrading to 1.5.1 if yes.

[mark-hayward]

Yes I'm using 5.9.1. I used the latest-sofproject branch. Will 1.5.1 work on 5.9.1? I'll try it later this evening.

I want to do a bit more debugging which I'll do later. There must be something triggering the change because it does work for a little while after a reboot.

[maaarghk]

To be clear, this is the driver not the firmware, use the latest-sofproject branch here and v1.5.1 on the firmware (easiest is via your distro sof-firmware package)

[mark-hayward]

I'm using 1.6 (arch linux). I'll try downgrading this evening. Thanks!

[mark-hayward]

Weirdly, only the left channel is effected.

[maaarghk]

yeah, downgrading to 1.5.1 will definitely resolve that - i had the same thing.

[ruffson]

@divoxx Did you figure out how to get around this? I have the same issue on 5.8.16 (and on 5.8.17).

[maaarghk]

@ruffson you have a kconfig option needing set - or find some #IFDEFs and get rid of them

[ruffson]

@ruffson you have a kconfig option needing set - or find some #IFDEFs and get rid of them

Mh no, these functions are defined in include/dkms/linux/soundwire/sdw.h and there is no #ifdef around them.

//Edit: Export happens here without any #ifdef pragma around.

[maaarghk]

hello everyone in the thread, 5.10 works fine without this dkms module so you can remove that before installing your distro's 5.10 thing. in arch that would be linux 5.10.1.arch2 or something. in manjaro linux-5.10.1-3. firmware 1.6 now also works well

[m5r]

Hello, thanks for keeping us up to date with the latest changes.

After installing kernel version 5.10.0-1-MANJARO, I don't have any sound on my XPS 9700.
Downgrading back to kernel version 5.8.18-1-MANJARO with this dkms module brings it back.

I'm not sure what I've done wrong, could you share more details on how you got it working please ?

[maaarghk]

hi, you need to be on unstable branch, the minimum version with support is 5.10.1-3-MANJARO. best to wait for now I suppose

[mark-hayward]

Since upgrading to linux 5.11.7.arch1-1 this issue has come back for me... anybody else?

[maaarghk]

I'm on manjaro 5.11.6 just now and it seems fine, did you run that on arch?