Hi,
On 11/30/21 12:07, Hans de Goede wrote:
Hi,
On 11/25/21 13:42, Takashi Iwai wrote:
On Thu, 25 Nov 2021 12:04:41 +0100, Hans de Goede wrote:
Hi,
I've a set of Logitech Z10 USB speakers, which act as a USB soundcard.
They have this weird glitch where after turning off my PC (and their power-supply as well) and then turning things back on, they are silent until I change the PCM volume control for the speakers inside alsa-mixer.
It seems like they need some "set-volume" command to be send over the USB bus to unmute them when initially powered-up / turned on.
Is their some existing usb-audio quirk which I can try to work around this?
No such quirk is present for now.
Was it tested with 5.16-rc? There was a change in USB-audio driver initialization (commit b96681bd5827) and it might have some effect in your case.
Yes I'm at 5.16-rc3 atm but I've been seeing this for quite some time. I just never got around to reporting it. Mainly because I also never got around to getting a bit clearer picture of the problem.
I've spend some time this morning to get that clearer picture, which was insightful.
Also, it's interesting to know whether it happens also once after suspend-resume, too.
suspend-resume makes no difference, not even rebooting or powering off the machine makes a difference.
Once the speakers are in working order they stay in working order until I turn off my machine; and then flick the power-switch on the 240V AC power-bar which I use to power my laptop + dock + monitors + the speakers and turn things back on the next morning.
To be clear these speakers get their audio-data over USB (as an usb-audio device) but they have their own power-supply they are not USB powered. They also have a "soft" on/off button which turns on/off the amplifier and LCD screen parts but leaves the USB audio interface active.
So I've been experimenting with reproducing the issue and I need to do the following minimal steps to reproduce:
- Unplug USB
- Unplug power
- Re-plug power
- Re-plug USB
- speaker-test -Dfront:CARD=Speaker,DEV=0 -S1
- Turn speakers on (with the on/off button on the speakers), no audio
5 and 6 may be swapped, same result
And now that I have a reliable reproducer I've also been playing with a reliable workaround which looks like this:
- Start playing audio to the speakers
- Turn speakers on (with the on/off button on the speakers)
- Make a change to the 'PCM Playback Volume' ctrl
Where 1. and 2. may be swapped. But the 'PCM Playback Volume' ctrl change must be made while the speakers are on and playing audio !
Although I have found that this also works:
- Start playing audio to the speakers
- Turn speakers on (with the on/off button on the speakers)
- Stop playing audio
- Make a change to the 'PCM Playback Volume' ctrl
- Start playing audio to the speakers again
I then even here a brief "power-up buzz" coming from the speakers at 4.
And this sequence also works:
- Start playing audio to the speakers
- Stop playing audio
- Turn speakers on (with the on/off button on the speakers)
- Make a change to the 'PCM Playback Volume' ctrl
- Start playing audio to the speakers again
So it seems that to work (after having been unplugged form the mains) these speakers need to:
- Have had some audio send to them at least once
- After this have their 'PCM Playback Volume' ctrl poked at once while they are on (and if they are on cannot be seen from the PC side AFAICT).
Note instead of changing the 'PCM Playback Volume' ctrl toggling the associated mute ctrl works too.
TL;DR: Since getting the speakers to work involves setting a ctrl while they are on, which is something which we cannot tell from the kernel side I don't believe that there is anything we can do about this from within the kernel.
So thinking more about this I guess we could do something where we resend the last PCM volume to the device every 5 seconds *when the device is playing audio*, assuming that the resending of the same PCM volume is sufficient to fix things.
These are pretty nice speakers so getting them to work without this glitch would be nice. But it would require a significant bit of (quirk enabled) code just for this 1 model speakers.
Takashi, what do you think. Should I give the resend volume once every 5 seconds idea a try, or is it likely going to end up being too ugly to merge ?
Regards,
Hans