search

raspberrypi / linux

Kernel source tree for Raspberry Pi-provided kernel builds. Issues unrelated to the linux kernel should be posted on the community forum at https://forums.raspberrypi.com/
Other
11.15k stars 5k forks source link

bcm2835_audio bcm2835_audio: Failed to get DT property 'brcm,pwm-channels' #3005

Open jeffersonchua opened 5 years ago

jeffersonchua commented 5 years ago

dmesg

bcm2835_audio bcm2835_audio: Failed to get DT property 'brcm,pwm-channels'

... happens on Linux 5.2.0-rc3 with latest commit 060a2d44b2e6f0bc62137e11c89404e0a3c71a4f

How to get rid of this ...

Thanks, Jeff

dxmann commented 5 years ago

I have the same problem on 5.2.2. Any news?

pelwell commented 5 years ago

It's a consequence of upstream's push to make VCHIQ's clients platform drivers, which removes the ability to use Device Tree properties and forces the use of module parameters.

@lategoodbye can explain why this is a good thing.

dxmann commented 5 years ago

ok, so how can I use fe-pi-audio? usually it is sufficent to load the dtoverlay on config.txt: dtoverlay=fe-pi-audio dtoverlay=i2s-mmap

pelwell commented 5 years ago

This error message comes from the driver that controls audio output via HDMI or the headphone socket - Fe-Pi is an I2S-based soundcard that should be completely unaffected.

pelwell commented 5 years ago

Also, the i2s-mmap overlay no longer exists because it isn't needed.

Notwithstanding my earlier comment, in the case of the bcm2835-audio driver, the DT support that has been reverted back into downstream kernels since 5.0 doesn't seem to be doing anything useful (the brcm,pwm-channels property has the default value of 8), so it is doing more harm than good. Unless @popcornmix knows of any reason not to I'll get rid of them.

popcornmix commented 5 years ago

@pelwell not entirely sure on the reasons for that revert, but if its doing no good then feel free to drop it.

pelwell commented 5 years ago

It might have been picked up (again) as one of the mass "lostcommits" cherry-picks. I'll blacklist it to ensure it doesn't happen again.

dxmann commented 5 years ago

so, is the error below related to missing audio devices?

[    4.283470] snd_bcm2835: module is from the staging directory, the quality is unknown, you have been warned.
[    4.285271] bcm2835_audio bcm2835_audio: Failed to get DT property 'brcm,pwm-channels'
[    4.296313] bcm2835_audio: probe of bcm2835_audio failed with error -22
pelwell commented 5 years ago

It will break the analogue and (non-VC4) HDMI audio, but external soundcards should be unaffected.

dxmann commented 5 years ago

ok, I'll check elsewhere

pelwell commented 5 years ago

Ah - Fe-Pi uses the sgtl5000 codec, which means you'll be needing f6cc589, which hasn't yet been picked up by one of the mass "lostcommits" cherry-picks.

dxmann commented 5 years ago

In the kernel checkout that I use, the fe-pi-audio-overlay has been patched with the f6cc589 patch.

pelwell commented 5 years ago

In that case I have no idea.

dxmann commented 5 years ago

me too :) even with the kernel 4.19.60 I can't see any audio device. I'm using the rpi-kernel with voidlinux. I have installed alsa-lib alsa-utils and alsa-firmware. I don't know if I have missed any package. but it is not related to this issue....

pelwell commented 5 years ago

Time for some systematic debugging: 1) Read the output of "sudo vcdbg log message" and check for errors with the overlay loading. 2) Run "raspi-gpio get" and confirm that GPIOs 28-31 are configured for I2S. 3) Run "dmesg -l err" and check for errors. 4) run "lsmod" and look for the codec and soundcard drivers.

pelwell commented 5 years ago

Going back to the original topic for a moment, although the error message is fatal for the device instance, it doesn't break audio output. The reason is that two instances are being created - one via device tree that (with the revert patch) can find brcm,pwm-channels, and one created explicitly by VCHIQ as a platform device. We could undo the reversion of the upstream DT-exorcism commit and stop instantiating bcm2835-audio from DT, which is what upstream intends, but then we would lose dtparam=audio=on/off support.

dxmann commented 5 years ago

ok, thanks :)

lategoodbye commented 5 years ago

It's a consequence of upstream's push to make VCHIQ's clients platform drivers, which removes the ability to use Device Tree properties and forces the use of module parameters.

@lategoodbye can explain why this is a good thing.

You already found one reason. Another one is that VCHIQ is a software interface to the firmware, so devicetree bindings will have a little chance to be accepted.

FunWithFlaggs commented 4 years ago

Did i overread something or how can the original problem be solved? bcm2835_audio bcm2835_audio: Failed to get DT property 'brcm,pwm-channels'

I try to get hwaccel video decoding + audio over hdmi but i can only get either of them, not both at the same time by either compiling the sources from here (to get hwaccel working, need /dev/vcsm-cma) or from https://git.archlinux.org/linux.git to get audio working.