Boot fails after irqbalance failure

Tumbleweed has not dumped X11 (yet). Depending on what desktop you are planning to use, X11 might still be a viable option.
Depending on the applications you need, you might still be able to calibrate your monitor in X11, then use the monitor profile even in Wayland.
But this is a long story, OT in this thread.

@OrsoBruno
I know it’s a long story
I’ve been following it on pixls.us

I use KDE plasma and only 3 months ago the project leader wrote that X11 would stay but essentially unmaintained, but trying to find out using different commands to try and find out about graphics there was only mention of Wayland. The film industry use Linux massively - I suppose they are big enough to build their own distro…

@Cyclonick colord should be there and running to import icd files for your setup?

It’s not that simple, Malcolm. Currently on Wayland only Darktable fully supports Colord. Gimp, Firefox for instance offer some support but via manual config for each single monitor, so if you switch monitor you have to reconfigure everything…
Then there is no way to calibrate a monitor with Wayland currently, so you need X11 to calibrate (generate an icd file), then import it to Wayland, then hope that your app support Colord…
All in all, better save and buy a Mac :wink:

1 Like

Colord is a pretty useless bit of tat - you’re supposed to be able to use a colorimeter with it, but there isn’t even a script for the Spyder series of instruments.

The problem is that to calibrate the program has to Access the data from the monitor and Graphics card which Wayland does not allow. Argyll contains a script that uses Colord to get the information, but I don’t know if it works. KDE are talking of doing away with Colord and writing their own. There used to Oyranos, which was much better but ceased development.

Darktable has different strategies - you can also drop a .icc into a folder in config, Rawtherapee and Krita both have similar options (Those are the 3 I use)

This is amusingly off topic !

Back on topic !
What does DRM panic mean?

@Cyclonick I’d be looking at either new non AMD hardware or reporting an upstream bug…
https://www.phoronix.com/news/AMD-DCE-DCN-DRM-Panic

A bit late for that - I’ve never had problems with AMD…
The full line is amdgpu 0000:0c:00.0: [drm] Registered 4 planes with drm panic

@Cyclonick I suspect an upstream bug report will be needed…

A new thread here ?

Otherwise where ?

@Cyclonick Upstream is https://gitlab.freedesktop.org/drm/amd

I haven’t a hope of participating there - I’m floundering here…

@Cyclonick The joys of running Tumbleweed, there is an expectation that some things won’t work and users can figure out an issue and report…

You could wait for the 6.16 kernel to drop? Or switch to use the rc version from Kernel:HEAD? I’m running it here only because of Intel ARC testing.

zypper ar -f -g -n "Kernel HEAD standard" https://download.opensuse.org/repositories/Kernel:/HEAD/standard/ repo-kernel-head
zypper -vvv dup --allow-vendor-change

I run tumbleweed because in Leap, unless I install from source, the software I use is kept 2 years behind the curve…

I don’t think (barring the patch) it’s a kernel problem.

Anyway I don’t know how to use recovery mode, but reading through the journal everything seems to set up OK, drm finds the gpu, sets up the different bits of it - it’s when amgdpu enters the picture that it seems to go wrong. (didn’t know drm was the graphics bit of the kernel until today… which shows the level I’m starting from)

I think I’ll try a new thread here - when I can work out what to say - if only to clarify what needs to be said on gitlab

Did you peruse the bugs for DRM Panic, there are lots…

Here or there (gitlab) ?

I’m looking there at the moment…

Ah, I see what you mean (looking here)

@Cyclonick Gitlab, but a search here may help.

Correct me if I’m wrong, but is irqbalance really needed? I uninstalled it from both my Intel i5 and AMD Ryzen rigs a long time ago, and they both run perfectly without it. I also uninstall it first thing in my VMs, together with nscd, cups, samba, colord, tuned, mcelog, Plymouth and dozens of other services I never need that would otherwise waste CPU time and resources.

As far as I understand, irqbalance is good for multi-socket server units, but it won’t do anything for single-socket boards.

@unix111 Sure, but that is a red herring in this case, the issue is the drm panic.

I disable lots of stuff and remove as well, but this is a new user and needs to learn some more before going down that path :wink:

IRQs go back to the very beginning - when 40Mb hard disks and 4Mb of ram was a luxury… The fact that I do remember those times should tell you that old dogs don’t learn new tricks on the basis that ā€œif it aint broke, don’t fix it !ā€

I’m intrigued by the sudden rash of problems (in the linux world) with this chipset, which has been around a while. The only thing I can think of is a change in commercial policy : that hitherto it’s been restricted to the likes of HP, Acer etc. who only make Windows computers (which linux people don’t want), but now it’s been made generally available where there’s a chance of it being used for a linux users kit.

Collecting the data for amdgpu I’m not sure if I’m looking at a kernel problem or an amdgpu one…