Raspberry Pi 4 bcm2711 / Raspberry Pi 3 bcm2837 (aarch64)

sndwvs · 10-08-2022, 09:40 AM

raspberry pi 4

installation README.TXT
kernel 6.0.0

slarm64-current-aarch64-server-raspberry_pi_4-6.0.0-build-20221008.img.zst
slarm64-current-aarch64-server-raspberry_pi_4-6.0.0-build-20221008.img.zst.sha256
slarm64-current-aarch64-xfce-raspberry_pi_4-6.0.0-build-20221008.img.zst
slarm64-current-aarch64-xfce-raspberry_pi_4-6.0.0-build-20221008.img.zst.sha256

pchristy · 10-10-2022, 09:10 AM

Having a problem with kernel 6.0.0 on the Pi 400. I cannot start X or Wayland. If starting KDE, I get the gearwheel, then the screen goes blank. It does this in X11, Wayland and Full Wayland. If I select XFCE, the see the xfce desktop start to load (briefly) and then the screen goes blank. If I SSH into the Pi and run dmesg, I get the following:

Code:

[   58.712402] vc4-drm gpu: swiotlb buffer is full (sz: 1114112 bytes), total 32768 (slots), used 320 (slots)
[   62.170655] vc4-drm gpu: [drm] *ERROR* [CRTC:94:crtc-3] flip_done timed out
[   62.583147] vc4-drm gpu: swiotlb buffer is full (sz: 524288 bytes), total 32768 (slots), used 7714 (slots)

Any ideas?

Edit: Also noticed that if I reboot from SSH I get message about linuxroot not being cleanly unmounted, and it goes through the automatic correction process. This is despite being shutdown / rebooted correctly.

Cheers,

--
Pete

sndwvs · 10-10-2022, 12:10 PM

Quote:

Originally Posted by pchristy

Having a problem with kernel 6.0.0 on the Pi 400. I cannot start X or Wayland. If starting KDE, I get the gearwheel, then the screen goes blank. It does this in X11, Wayland and Full Wayland. If I select XFCE, the see the xfce desktop start to load (briefly) and then the screen goes blank. If I SSH into the Pi and run dmesg, I get the following:

Code:

[   58.712402] vc4-drm gpu: swiotlb buffer is full (sz: 1114112 bytes), total 32768 (slots), used 320 (slots)
[   62.170655] vc4-drm gpu: [drm] *ERROR* [CRTC:94:crtc-3] flip_done timed out
[   62.583147] vc4-drm gpu: swiotlb buffer is full (sz: 524288 bytes), total 32768 (slots), used 7714 (slots)

Any ideas?

Edit: Also noticed that if I reboot from SSH I get message about linuxroot not being cleanly unmounted, and it goes through the automatic correction process. This is despite being shutdown / rebooted correctly.

Cheers,

--
Pete

It looks like it's a kernel problem, maybe it's been fixed already. checked for pine64 there is no such problem.

pchristy · 10-10-2022, 01:00 PM

OK, looks like I need to wait for the next point release.

Hopefully this will warn others of the potential problem, though/

--
Pete

business_kid · 10-11-2022, 11:55 AM

Maybe do the decent thing and file a bug?

pchristy · 10-11-2022, 12:13 PM

Well, it depends where sndwvs gets his kernel source from. I don't think he uses the "pure" one from kernel.org, but a patched one from Pi. But I'm not sure - no doubt he'll be along to inform me shortly! I believe someone has been playing with Pi-patched 6rc kernels over on the "official" aarch64 port, and it was reportedly looking promising on Pis, but I haven't seen anything since. The last kernel they released completely broke my system, but this is the first time I've had a real issue with slarm64 kernels. I thought 6 was supposed to have a "proper" driver for the Pi, but it looks as if something isn't right. It could also be a configuration issue. Its still very new, so I'll see how it shakes down.

--
Pete

sndwvs · 10-11-2022, 01:28 PM

Quote:

Originally Posted by pchristy

Well, it depends where sndwvs gets his kernel source from. I don't think he uses the "pure" one from kernel.org, but a patched one from Pi. But I'm not sure - no doubt he'll be along to inform me shortly! I believe someone has been playing with Pi-patched 6rc kernels over on the "official" aarch64 port, and it was reportedly looking promising on Pis, but I haven't seen anything since. The last kernel they released completely broke my system, but this is the first time I've had a real issue with slarm64 kernels. I thought 6 was supposed to have a "proper" driver for the Pi, but it looks as if something isn't right. It could also be a configuration issue. Its still very new, so I'll see how it shakes down.

--
Pete

The kernel for raspberry pi is taken from.

This happens sometimes, even with kernels from kernel.org

business_kid · 10-11-2022, 01:32 PM

The previous post makes mine redundant. Ignore.

pchristy · 10-11-2022, 01:47 PM

Both: Not a problem for me. Sometimes you just need to be patient for things to settle!

--
Pete

pchristy · 10-12-2022, 06:57 AM

UPDATE: Found the problem, and it was a configuration issue -sort of! In my /boot/config.txt file, I had added the line

Code:

dtoverlay=rpivid-v4l2

This provided some speed increases for video processing in earlier kernels, but does not play nicely with kernel-6!

Commenting it out has got my system up and running again!

Running glxinfo provides a different output under 6:

Code:

glxinfo | grep render
direct rendering: Yes
    GLX_MESA_copy_sub_buffer, GLX_MESA_query_renderer, GLX_MESA_swap_control, 
    GLX_MESA_copy_sub_buffer, GLX_MESA_query_renderer, GLX_MESA_swap_control, 
Extended renderer info (GLX_MESA_query_renderer):
OpenGL renderer string: V3D 4.2
    GL_OES_element_index_uint, GL_OES_fbo_render_mipmap,

Although this looks significantly different from previous kernels, there is no significant improvement in video playback - I still can't play 4K videos! Again, this may need some fine tuning of the config.txt and probably the video players as well.

But at least its working again!

--
Pete

business_kid · 10-12-2022, 09:22 AM

Quote:

Originally Posted by pchristy

dtoverlay=rpivid-v4l2

What's that supposed to do? I didn't think V4L was the way things were going. It sounds like you nobbled it, but then you fixed it and it's going again. I'm happy for you, but remember rule #1:

Code:

If it works, don't fix it!

pchristy · 10-12-2022, 09:34 AM

I think it enables things like the memory-to-memory transfer when doing video encoding, etc. It worked fine in the older kernels, but looks like it has been removed in kernel 6.

Not sure why it corrupted the file systems on the SDcard, though!

--
Pete

business_kid · 10-12-2022, 01:22 PM

Quote:

Originally Posted by pchristy

Not sure why it corrupted the file systems on the SDcard, though!
Pete

Thatcan happen if the sun goes behind a cloud

.

sndwvs · 10-19-2022, 06:41 PM

at the moment when using u-boot and kernel 6.0.y the system hangs when starting xfce, for correct operation it is necessary to load the kernel by changing u-boot.bin
/boot/config.txt

Code:

#kernel=u-boot.bin
kernel=Image

sndwvs · 10-20-2022, 04:04 PM

raspberry pi 4

installation README.TXT
kernel 6.0.2

slarm64-current-aarch64-server-raspberry_pi_4-6.0.2-build-20221021.img.zst
slarm64-current-aarch64-server-raspberry_pi_4-6.0.2-build-20221021.img.zst.sha256
slarm64-current-aarch64-xfce-raspberry_pi_4-6.0.2-build-20221021.img.zst
slarm64-current-aarch64-xfce-raspberry_pi_4-6.0.2-build-20221021.img.zst.sha256

there is a problem when starting xorg for the first time and entering xfce the vc4 driver crashes:

Quote:

Oct 20 17:25:55 raspberry-pi-4 kernel: vc4-drm gpu: [drm] *ERROR* [CRTC:105:crtc-4] flip_done timed out
Oct 20 17:26:02 raspberry-pi-4 kernel: bcm2835_audio bcm2835_audio: vchi message timeout, msg=5
Oct 20 17:26:09 raspberry-pi-4 kernel: vc4-drm gpu: [drm] *ERROR* flip_done timed out
Oct 20 17:26:09 raspberry-pi-4 kernel: vc4-drm gpu: [drm] *ERROR* [CRTC:105:crtc-4] commit wait timed out
Oct 20 17:26:19 raspberry-pi-4 kernel: vc4-drm gpu: [drm] *ERROR* flip_done timed out
Oct 20 17:26:19 raspberry-pi-4 kernel: vc4-drm gpu: [drm] *ERROR* [PLANE:95

lane-4] commit wait timed out
Oct 20 17:26:29 raspberry-pi-4 kernel: vc4-drm gpu: [drm] *ERROR* flip_done timed out
Oct 20 17:26:29 raspberry-pi-4 kernel: vc4-drm gpu: [drm] *ERROR* Timed out waiting for commit

after which the monitor and usb are turned off.
after some time ~15 seconds, usb starts working (you can check by switching num lock) after that you can go to any console and restart xorg:

Code:

init 3 && init 4

after that xfce works without problems.

found a problem in the snd_bcm2835 driver.
snd_bcm2835 driver is blacklisted, to output sound through the analog output, you must manually load it after entering the desktop:

Code:

modprobe snd_bcm2835