nouveau 0000:01:00.0: Xorg[2356]: failed to idle channel 29

[joneall]

My system hangs every now and then, always when I am watching Netflix, be it on Chromium or Firefox. I found this in the syslog just before the hang.

Jan 22 16:56:57 jon-desktop plasmashell[12344]: [0122/165657.639060:ERROR:elf_dynamic_array_reader.h(64)] tag not found
Jan 22 16:57:12 jon-desktop kernel: [32763.140358] nouveau 0000:01:00.0: Xorg[2356]: failed to idle channel 29 [Xorg[2356]]

This site says I should install the Nvidia driver, but my tendency is not to do so, as it often brings along its own problems. Plus it's easier to stick with nouveau -- which is now rather more vieux than noveau -- like me.

Suggestions?

[ShadYoung]

Depending on the GPU, I have to disagree with the statement that the NV driver brings its own problems. It does not bring problems, it reveals them (meaning its not the driver, its the system). Nouveau is absolutely awful for Nvidia GPUs and I get rid of it as fast as I can. I have been using the NV driver for, well.. decades... without issue (I currently run dual NV GPUs with a quad monitor setup - in 20 years I have never traced a problem back to the NV driver - Linuxy things around it, yes, but the driver is always doing its job properly). It makes for a much faster and far more stable machine, and has HW accelerated encoding and decoding, CUDA, hardware monitoring and tuning, and decent Vulkan support. I did say depending on the GPU; an RTX4090 may not have mature drivers for a while, even in Windows, and GTX 700 series and below are no longer supported. Just remember to enable the i386 architecture support prior to installing it in Neon if you choose to.

[joneall]

Well, based on this wonderful review, I installed the NVIDIA driver. Now I can't boot the system at all. It doesn't even recognize my display, which circulates through all the possible connections but finds nothing. So that system is now useless.

Fortunately, I had another, older one.

Any ideas?

[claydoh]

Sooo....

Which GPU?
Which particular driver package did you install, and using which install method?

Do remember that neon is 100% identical to Ubuntu/Kubuntu. minus the Driver Manager utility (part of the Software Sources tool), which one can always install -- software-properties-qt

The commandline version ubuntu-drivers is present
So, the easiest way to get back probably is to go to recovery mode, or switch a tty if you are making it to the login screen, and run:
ubuntu-drivers install xserver-xorg-video-nouveau

This will/should switch you back
It won't hurt to run an apt autoremove afterward, just to make sure no stray nvidia packages were left behind.

Unless your GPU is unsupported, or an incompatible driver version was installed, the driver install should have set up xorg etc, though there is the chance that it failed to do that, I guess. This happens automatically when using the Driver Manager via the GUI or the CLI, Not sure if simply installing the driver packages using apt necessarily does this step.

[joneall]

Thanks, I will try that. But my first attempt did not get me to the system at all, since my display is never recognized. As I said, the screen (Samsung) doesn't can't find any connection, on hdmi or anything else.

[claydoh]

Thanks, I will try that. But my first attempt did not get me to the system at all, since my display is never recognized. As I said, the screen (Samsung) doesn't can't find any connection, on hdmi or anything else.

Not even to grub?

If you aren't making it to the grub menu, which is not using Xorg, and thus neither the Nvidia nor the nouveau drivers at all ( it uses the very basic VESA system at that level) then there is something else wrong here.
if you don't normally see the grub menu (hidden as you single-boot), you will need to hit the esc key, or possibly the shift key, at the right time just after the bios/splash screen when powering up.

[joneall]

I tried claydoh's suggestion. I managed to get a line-by-line root login and ran

ubuntu-drivers install xserver-xorg-video-nouveau.

That replied that it was already installed. But the restart gave the same problem. I get through refind and the boot starts, but at some point, it loses the screen. Or rather, the screen sees no more input. Is there an additional command to tell it to actually use nouveau.

Or is the pb that the nvidia installation just screwed up the nouveau without actually deselecting it?

[claydoh]

If the Nvidia install created an /etc/X11/xorg.conf file, you will want to delete that. I don't think it does so these days, but I have no idea.
Also run an apt autoremove if you haven't done so.

You probably need to go and manually uninstall all the Nvidia driver packages etc. The ubuntu-drivers command is supposed to help with switching drivers and do this sort of thing.

Something along the lines of the first section here: https://linuxconfig.org/how-to-unins...ellyfish-linux
This should root out anything not removed for some reason.

But people with actual, recent Nvidia experience need to chime in, lol.

[Schwarzer Kater]

[…]
You probably need to go and manually uninstall all the Nvidia driver packages etc. The ubuntu-drivers command is supposed to help with switching drivers and do this sort of thing.[…]

I would even go so far to purge the Nvidia components one by one (sudo apt purge component-names) if ubuntu-drivers does not help…

The Nouveau driver could also still be blacklisted, e.g. in /etc/default/grub or other places (that I can't remember right now) - so none of the two fitting drivers loads during boot and the screen goes black (especially if there are wrong config files).
Unfortunately I will not be home at my desktop computer until next Tuesday to look this up - perhaps another Nvidia user could be so kind?

By the way: Which Nvidia card do you use? I think you have not told us yet.

[joneall]

$ ubuntu-drivers devices
== /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0 ==
modalias : pci:v000010DEd00000FC8sv00001458sd0000368Dbc03sc00 i00
vendor : NVIDIA Corporation
model : GK107 [GeForce GT 740]
driver : nvidia-driver-418-server - distro non-free
driver : nvidia-driver-470 - distro non-free recommended
driver : nvidia-driver-390 - distro non-free
driver : nvidia-driver-470-server - distro non-free
driver : nvidia-driver-450-server - distro non-free
driver : xserver-xorg-video-nouveau - distro free builtin

I just noticed that ubuntu-drivers says the autoinstall option is deprecated. That's what I used on the other system. Maybe that's the source of my problem.

[Schwarzer Kater]

Despite the message, the "autoinstall" option should work fine - I have used it on dozens of different Ubuntu-based systems for Nvidia GPUs without any problems…