Moving Linux Kernel to 5.10

[dom]

The first post says in this thread says
Rather than dtoverlay=vc4-fkms-v3d, switch to dtoverlay=vc4-kms-v3d-pi4 on a pi4, or dtoverlay=vc4-kms-v3d on Pi2/3 (not recommended on Pi0/1)

I'm not seeing that (either someone else has edited my post or you are not referring to the first post in this thread).
But you are correct the "-pi4" is redundant since device tree aliases were added long ago.

[graysky]

I was sure that I read elsewhere that the -pi4 was redundant, but now I'm not sure what I've been using :/

I understand that is true so long as you're using a recent firmware.

[dom]

Full KMS vs Fake KMS, and the ARM-side HDMI ALSA driver

Can anyone suggest if and when I would see or measure any differences between the full KMS and HDMI audio drivers over the previous firmware-based "Fake KMS/fkms" setup and Broadcom audio?

• Lower CPU?
• Lower latency possible?
• Faster anything?
• Different APIs supported?

Lower latency is quite likely as the route to output audio is much shorter (firmware based drivers require an additional queue of audio data to the GPU).
Arm CPU is probably higher (as spdif formatting occurs on arm rather than gpu), but that should be negligible on pi2/3/4.
They both use the alsa api, although the arm driver driver can support HBR passthrough (TrueHD/DTS-HD) on pi4, but the firmware driver does not.

[dom]

Latest update includes a fix (well reverting upstream changes that appear to have broken it) to vchiq which hopefully restores hardware video decode.

[busywait]

(... you are not referring to the first post in this thread).
But you are correct the "-pi4" is redundant since device tree aliases were added long ago.

Thank you - looks like I saw that old text in the "moving to kernel 5.4 thread", which I was looking at to confirm my audio setup.

[busywait]

Full KMS vs Fake KMS, and the ARM-side HDMI ALSA driver

Lower latency is quite likely as the route to output audio is much shorter (firmware based drivers require an additional queue of audio data to the GPU).
Arm CPU is probably higher (as spdif formatting occurs on arm rather than gpu), but that should be negligible on pi2/3/4.
They both use the alsa api, although the arm driver driver can support HBR passthrough (TrueHD/DTS-HD) on pi4, but the firmware driver does not.

Thanks Dom!

And what are the likely differences between fkms and kms for video?

[dom]

And what are the likely differences between fkms and kms for video?

Similar to audio. kms drives the display hardware directly from the arm and fkms sends remote messages to gpu.
fkms was really a kludge to make stuff work while the kms driver was developed (it's been a lot of work but it's getting close now).

Latency will be better with kms. cpu usage is negligible in both cases.
fkms has an issue with tearing in full screen apps that kms fixes.

[busywait]

Latest update includes a fix (well reverting upstream changes that appear to have broken it) to vchiq which hopefully restores hardware video decode.

I didn't notice any problems with hardware decode before :/ However, I've just now used the Screen Layout Editor to set the desktop refresh rate to 60Hz, and then used Chromium to play some YouTube 1080p 60Hz content with the stats on. Some frame dropped each second, but looked and sounded pretty smooth with vc4-kms-v3d+ARM-HDMI audio.

I'm not quite sure why my desktop had "reset" itself to 50Hz not 60Hz, and it won't remember my selection from the Preferences > Screen Configuration application (Screen Layout Editor) any more. Could that be kernel 5.10 related?

[chrisy]

Latest update includes a fix (well reverting upstream changes that appear to have broken it) to vchiq which hopefully restores hardware video decode.

I updated to the latest and Twitch is still causing the Pi to completely freeze. (db075f137fe12748c85a36fb1c52aacab61bce8a)

[Double L]

This evening I updated to kernel 5.10.2 and YouTube still causes the Pi4 to completely freeze after a few seconds of playing. I tried both FKMS and KMS. Same behaviour.

[elpeh]

RPi 400 with latest Raspberry Pi OS (10.7).
Bluetooth is not working with 5.10 kernels from rpi-upgrade (5.10.0, 5.10.1 and 5.1.12). bluthtoothctl is hanging with this kernels. No problem downgraded to 5.4.79-v7l+ #1373

Lutz

[ShorTie]

But, But, but, I thought the use of rpi-update was a no no ??

[DougieLawson]

But, But, but, I thought the use of rpi-update was a no no ??

Not if you're joining in with the beta testing of the 5.10.2 kernel. You can't install that for testing without using sudo rpi-update (and having a good backup in case your system won't boot).

[itsmedoofer]

Hi,

I've been trying to get audio working over the new KMS driver but cant seem to crack it, I think it's because of my set up rather than the software though..

I have a DVI monitor that I connect to the Pi with the standard adaptor, to get the sound I have a HDMI audio extractor that I plug the desktop speakers in to.

For this to work with the FKMS driverI have to have the following in my config.txt, but it works well

Code: Select all

hdmi_drive=2

I think I have followed things correctly , dtoverlay=vc4-kms-v3d has been set, dtparam=audio=on has been removed and the edits to /usr/share/alsa/cards/vc4-hdmi.conf have been made.

When I look at the available sounds devices I only see a "Dummy" device.

My thinking is that the monitor is not being detected as a sound capable device and so nothing is being registered ?

I have some errors in DMESG...

Code: Select all

[    4.996529] bcm2835-codec bcm2835-codec: Device registered as /dev/video12
[    4.996606] bcm2835-codec bcm2835-codec: Loaded V4L2 isp
[    5.215201] [drm] Initialized v3d 1.0.0 20180419 for fec00000.v3d on minor 0
[    5.260732] cfg80211: Loading compiled-in X.509 certificates for regulatory database
[    5.330538] debugfs: Directory 'fef00700.hdmi' with parent 'vc4-hdmi-0' already present!
[    5.333779] vc4_hdmi fef00700.hdmi: ASoC: no DMI vendor name!
[    5.334690] vc4-drm gpu: bound fef00700.hdmi (ops vc4_hdmi_ops [vc4])
[    5.338724] cfg80211: Loaded X.509 cert 'sforshee: 00b28ddf47aef9cea7'
[    5.338967] debugfs: Directory 'fef05700.hdmi' with parent 'vc4-hdmi-1' already present!
[    5.341033] vc4_hdmi fef05700.hdmi: ASoC: no DMI vendor name!
[    5.344356] vc4-drm gpu: bound fef05700.hdmi (ops vc4_hdmi_ops [vc4])
[    5.344674] vc4-drm gpu: bound fe400000.hvs (ops vc4_hvs_ops [vc4])
[    5.344996] vc4-drm gpu: bound fe004000.txp (ops vc4_txp_ops [vc4])
[    5.345287] vc4-drm gpu: bound fe206000.pixelvalve (ops vc4_crtc_ops [vc4])
[    5.345659] vc4-drm gpu: bound fe207000.pixelvalve (ops vc4_crtc_ops [vc4])
[    5.347727] vc4-drm gpu: bound fe20a000.pixelvalve (ops vc4_crtc_ops [vc4])
[    5.348162] vc4-drm gpu: bound fe216000.pixelvalve (ops vc4_crtc_ops [vc4])
[    5.350990] vc4-drm gpu: bound fec12000.pixelvalve (ops vc4_crtc_ops [vc4])

And

Code: Select all

[    9.504470] brcmfmac: brcmf_cfg80211_set_power_mgmt: power save enabled
[    9.989908] broken atomic modeset userspace detected, disabling atomic
[   11.577424] fuse: init (API version 7.32)
[   12.811322] brcmfmac: brcmf_cfg80211_set_power_mgmt: power save enabled
[   13.178260] vc4_hdmi fef00700.hdmi: ASoC: error at snd_soc_dai_startup on fef00700.hdmi: -19
[   13.179079] vc4_hdmi fef00700.hdmi: ASoC: error at snd_soc_dai_startup on fef00700.hdmi: -19
[   13.179937] vc4_hdmi fef00700.hdmi: ASoC: error at snd_soc_dai_startup on fef00700.hdmi: -19
[   13.200052] vc4_hdmi fef00700.hdmi: ASoC: error at snd_soc_dai_startup on fef00700.hdmi: -19
[   13.200556] vc4_hdmi fef00700.hdmi: ASoC: error at snd_soc_dai_startup on fef00700.hdmi: -19
[   13.273452] vc4_hdmi fef05700.hdmi: ASoC: error at snd_soc_dai_startup on fef05700.hdmi: -19
[   13.283219] vc4_hdmi fef05700.hdmi: ASoC: error at snd_soc_dai_startup on fef05700.hdmi: -19
[   13.283927] vc4_hdmi fef05700.hdmi: ASoC: error at snd_soc_dai_startup on fef05700.hdmi: -19
[   13.293563] vc4_hdmi fef05700.hdmi: ASoC: error at snd_soc_dai_startup on fef05700.hdmi: -19
[   13.293979] vc4_hdmi fef05700.hdmi: ASoC: error at snd_soc_dai_startup on fef05700.hdmi: -19
[   14.585700] Bluetooth: Core ver 2.22
[   14.585803] NET: Registered protocol family 31
[   14.585809] Bluetooth: HCI device and connection manager initialized
[   14.585825] Bluetooth: HCI socket layer initialized

Any ideas on how to fix this ?

I have tried this on a clean PiOS64 updated and then updated the kernel

Code: Select all

Linux raspberrypi64 5.10.2-v8+ #1385 SMP PREEMPT Wed Dec 23 12:00:51 GMT 2020 aarch64 GNU/Linux

[busywait]

Hi,

I've been trying to get audio working over the new KMS driver but cant seem to crack it, I think it's because of my set up rather than the software though..

I have a DVI monitor that I connect to the Pi with the standard adaptor, to get the sound I have a HDMI audio extractor that I plug the desktop speakers in to.

For this to work with the FKMS driverI have to have the following in my config.txt, but it works well

Code: Select all

hdmi_drive=2

I think I have followed things correctly , dtoverlay=vc4-kms-v3d has been set, dtparam=audio=on has been removed and the edits to /usr/share/alsa/cards/vc4-hdmi.conf have been made.

When I look at the available sounds devices I only see a "Dummy" device.

Code: Select all

[    9.504470] brcmfmac: brcmf_cfg80211_set_power_mgmt: power save enabled
[    9.989908] broken atomic modeset userspace detected, disabling atomic
[   11.577424] fuse: init (API version 7.32)
[   12.811322] brcmfmac: brcmf_cfg80211_set_power_mgmt: power save enabled
[   13.178260] vc4_hdmi fef00700.hdmi: ASoC: error at snd_soc_dai_startup on fef00700.hdmi: -19
[   13.179079] vc4_hdmi fef00700.hdmi: ASoC: error at snd_soc_dai_startup on fef00700.hdmi: -19
[   13.179937] vc4_hdmi fef00700.hdmi: ASoC: error at snd_soc_dai_startup on fef00700.hdmi: -19
[   13.200052] vc4_hdmi fef00700.hdmi: ASoC: error at snd_soc_dai_startup on fef00700.hdmi: -19
[   13.200556] vc4_hdmi fef00700.hdmi: ASoC: error at snd_soc_dai_startup on fef00700.hdmi: -19
[   13.273452] vc4_hdmi fef05700.hdmi: ASoC: error at snd_soc_dai_startup on fef05700.hdmi: -19
[   13.283219] vc4_hdmi fef05700.hdmi: ASoC: error at snd_soc_dai_startup on fef05700.hdmi: -19
[   13.283927] vc4_hdmi fef05700.hdmi: ASoC: error at snd_soc_dai_startup on fef05700.hdmi: -19
[   13.293563] vc4_hdmi fef05700.hdmi: ASoC: error at snd_soc_dai_startup on fef05700.hdmi: -19
[   13.293979] vc4_hdmi fef05700.hdmi: ASoC: error at snd_soc_dai_startup on fef05700.hdmi: -19
[   14.585700] Bluetooth: Core ver 2.22
[   14.585803] NET: Registered protocol family 31
[   14.585809] Bluetooth: HCI device and connection manager initialized
[   14.585825] Bluetooth: HCI socket layer initialized

Any ideas on how to fix this ?

I have tried this on a clean PiOS64 updated and then updated the kernel

I think that there is still an unresolved problem with PulseAudio+the new ARM-side audio. (PulseAudio is part of the latest updates that you have installed).

I'd do these 2 things:
- follow the instructions from the December 2020 update blog https://www.raspberrypi.org/blog/new-ra ... mber-2020/ to ensure that you're switched over to PA in the desktop
- look back in this thread to Dec 16, and 17 when Dom suggested workarounds for the problems loading vc4_hdmi audio

[itsmedoofer]

Hi,

Thanks for the reply, appreciated.

So I have followed the thread, made sure things were working before switching to the KMS driver, made the switch and followed the possible fixes, still not go...

Having read the thread again I'm less certain it's my setup causing the problem, especially after reading your posts, will park the idea of using it for the moment and keep an eye open here...

[Bluestang]

Hi,

Thanks for the reply, appreciated.

So I have followed the thread, made sure things were working before switching to the KMS driver, made the switch and followed the possible fixes, still not go...

Having read the thread again I'm less certain it's my setup causing the problem, especially after reading your posts, will park the idea of using it for the moment and keep an eye open here...

viewtopic.php?p=1781377#p1781377

Did you follow these instructions as well?

[itsmedoofer]

Hi,

Thanks for the reply.

Yes followed that as well no different.

Regards.

[Bluestang]

Hi,

Thanks for the reply.

Yes followed that as well no different.

Regards.

Can you post the contents of your config.txt and cmdline.txt in your boot folder?

Also are you sure that vc4-hdmi.conf was changed? IIRC you need to edit it with root access...

[HiassofT]

I have a DVI monitor that I connect to the Pi with the standard adaptor, to get the sound I have a HDMI audio extractor that I plug the desktop speakers in to.
...
My thinking is that the monitor is not being detected as a sound capable device and so nothing is being registered ?

To get audio with KMS drivers it's required that audio capabilities are signalled in the EDID (CEA audio and speaker allocation blocks). My guess is the HDMI audio extractor passes the edid of the DVI monitor on without altering it (AVRs usually modify the EDID and add the supported audio modes) and therefore you don't get audio from the driver.

To fix this you can manually override the edid:

Either use the edid file of a monitor with identical video support that also supports audio or grab the edid of your monitor (cat /sys/class/drm/card0-HDMI-A-1/edid > my-edid.bin) and add CEA extension blocks with an EDID editor (eg Deltacast E-EDID under windows / wine should do the job).

Then store the edid file in eg /lib/firmware/edid/my-edid.bin and add drm.edid_firmware=edid/my-edid.bin to cmdline.txt to use it.

so long,

Hias

[andrum99]

Hi,

I've been trying to get audio working over the new KMS driver but cant seem to crack it, I think it's because of my set up rather than the software though..

I have a DVI monitor that I connect to the Pi with the standard adaptor, to get the sound I have a HDMI audio extractor that I plug the desktop speakers in to.

For this to work with the FKMS driverI have to have the following in my config.txt, but it works well

Code: Select all

hdmi_drive=2

Probably hdmi_drive has no effect with the KMS driver, due it to doing its own mode setting.

Some HDMI audio extractors have a switch on the side to force a specific audio mode. If yours has one, that's probably the easiest way to get sound working again. For example, the one I use has 3 settings, PASS (passthrough, i.e. just pass what the display says straight through to the host), 2CH (force 2 channel stereo), 5.1CH (force 5.1 channel surround sound).

[noee]

Very latest rpi-update here (12/27) has broken something with either wifi or network in general. I updated and immediately JRiver MediaCenter and Chromium could not stream music from SomaFM. Didn't try video or another music sources. I say wifi, because this Pi4 4GB (32b) is only using wifi.

That said, all other network stuff seemed okay (file sharing, ssh, krdc, vnc).

Went back to 5.4.79 (Nov 23) and all is well with music streaming from both JRiver and Chromium.

EDIT: I should say, I was running both fake and real KMS with the old and new VC4....conf, respectively on 5.10.2, no joy with either.

Code: Select all

pi@arya:~ $ uname -a
Linux arya 5.4.79-v7l+ #1373 SMP Mon Nov 23 13:27:40 GMT 2020 armv7l GNU/Linux

pi@arya:~ $ sudo rpi-eeprom-update 
BCM2711 detected
Dedicated VL805 EEPROM detected
BOOTLOADER: up-to-date
CURRENT: Fri 11 Dec 2020 11:15:17 AM UTC (1607685317)
 LATEST: Fri 11 Dec 2020 11:15:17 AM UTC (1607685317)
 FW DIR: /lib/firmware/raspberrypi/bootloader/beta
VL805: up-to-date
CURRENT: 000138a1
 LATEST: 000138a1

[carldani]

I had a sound card problem similar to https://github.com/raspberrypi/linux/issues/3962 , so I tried my luck with rpi-update.

# rpi-update
*** Raspberry Pi firmware updater by Hexxeh, enhanced by AndrewS and Dom
*** Performing self-update
*** Relaunching after update
*** Raspberry Pi firmware updater by Hexxeh, enhanced by AndrewS and Dom
*** We're running for the first time
*** Backing up files (this will take a few minutes)
*** Backing up firmware
*** Backing up modules 5.4.79-v7l+
#############################################################
WARNING: This update bumps to rpi-5.10.y linux tree
See: viewtopic.php?f=29&t=288234
'rpi-update' should only be used if there is a specific
reason to do so - for example, a request by a Raspberry Pi
engineer or if you want to help the testing effort
and are comfortable with restoring if there are regressions.

DO NOT use 'rpi-update' as part of a regular update process.

##############################################################
Would you like to proceed? (y/N)
*** Downloading specific firmware revision (this will take a few minutes)
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
100 168 100 168 0 0 3169 0 --:--:-- --:--:-- --:--:-- 3169
100 119M 100 119M 0 0 293k 0 0:06:57 0:06:57 --:--:-- 472k
*** Updating firmware
*** Updating kernel modules
*** depmod 5.10.3-v7l+
*** depmod 5.10.3-v7+
*** depmod 5.10.3+
*** depmod 5.10.3-v8+
*** Updating VideoCore libraries
*** Using HardFP libraries
*** Updating SDK
*** Running ldconfig
*** Storing current firmware revision
*** Deleting downloaded files
*** Syncing changes to disk
*** If no errors appeared, your firmware was successfully updated to da59cb1161dc7c75727ec5c7636f632c52170961
*** A reboot is needed to activate the new firmware

The 5.10 kernel does boot on my Raspberry Pi 4 4GB, but IPsec doesn't work anymore. I tracked it down to dozens of processes waiting/hanging, all the same:
/sbin/modprobe -q -- crypto-cbc(aes)

crypto-cbc(aes) is an alias of aes-arm-bs. To replicate the hang, just run
# modprobe aes-arm-bs

That command will hang. I admit I didn't wait more than 20 minutes for it to complete, but I expect module loading to be done in milliseconds, not minutes.

# uname -a
Linux pi4 5.10.3-v7l+ #1386 SMP Sun Dec 27 14:05:40 GMT 2020 armv7l GNU/Linux
# vcgencmd version
Dec 15 2020 18:29:44
Copyright (c) 2012 Broadcom
version 4b4aff21f72c5b9ba39d83c7b0f8fa910a6ef99b (clean) (release) (start_x)

A downgrade to Linux 5.4.79-v7l+ is a workaround for now, but I'd rather have a working aes-arm-bs module.

[dom]

Very latest rpi-update here (12/27) has broken something with either wifi or network in general. I updated and immediately JRiver MediaCenter and Chromium could not stream music from SomaFM. Didn't try video or another music sources. I say wifi, because this Pi4 4GB (32b) is only using wifi.

It's hard to imagine a wifi issue that affect SomaFM and not other web sites. More information would be useful. e.g.
does youtube work? Other radio streaming sites? Does it work if you connect ethernet cable?

[itsmedoofer]

Probably hdmi_drive has no effect with the KMS driver, due it to doing its own mode setting.

Some HDMI audio extractors have a switch on the side to force a specific audio mode. If yours has one, that's probably the easiest way to get sound working again. For example, the one I use has 3 settings, PASS (passthrough, i.e. just pass what the display says straight through to the host), 2CH (force 2 channel stereo), 5.1CH (force 5.1 channel surround sound).

Thanks for the reply, that was my thinking that it was redundant with the KMS driver.

Mine has a switch for 2.0 & 5.1 that didn't make any difference when testing with the FKMS driver, I removed hdmi_drive just to see if it was picked up.

Thanks again.