I have upgraded a laptop from Debian 11 (Bullseye) to Debian 12 (Bookworm)
Now I am running into trouble when trying to build DKMS modules needed by nvidia-driver (the proprietary one) on the most recent kernel version (I tried both 6.1.76-1 and 6.6.13-1~bpo12+1).
There are some compile-time errors (different errors in different attempts).
Is there any information about known problems of getting Nvidia software to interoperate with Debian 12 (Bookworm) working on up-to- date kernels?
--- Omer Zak
My experience with NVIDIA and DKMS is that with very recent Kernels the build can break, but I didn't have that for some time now (years).
Also, I get my driver from my distro repo (Fedora), which reduce the chance of these issues.
-- Rabin
On Fri, 22 Mar 2024 at 12:01, Omer Zak w1@zak.co.il wrote:
I have upgraded a laptop from Debian 11 (Bullseye) to Debian 12 (Bookworm)
Now I am running into trouble when trying to build DKMS modules needed by nvidia-driver (the proprietary one) on the most recent kernel version (I tried both 6.1.76-1 and 6.6.13-1~bpo12+1).
There are some compile-time errors (different errors in different attempts).
Is there any information about known problems of getting Nvidia software to interoperate with Debian 12 (Bookworm) working on up-to- date kernels?
--- Omer Zak
-- My Commodore 64 is suffering from slowness and insufficiency of memory; and its display device is grievously short of pixels. Can anyone help? My own blog is at https://tddpirate.zak.co.il/
My opinions, as expressed in this E-mail message, are mine alone. They do not represent the official policy of any organization with which I may be affiliated in any way. WARNING TO SPAMMERS: at https://www.zak.co.il/spamwarning.html
Linux-il mailing list -- linux-il@cs.huji.ac.il To unsubscribe send an email to linux-il-leave@cs.huji.ac.il
I am trying to recover from the bungled upgrade. I found some answers in the Internet. For example, to use kernel 6.1.0- 17 instead of 6.1.0-18 whose package version is 6.1.76-1) and to enable bookworm-updates (https://forums.developer.nvidia.com/t/debian-12-and-nvidia-driver-nvidia-lin...).
Now I am stuck with inability to uninstall xorg-video-nvidia-tesla-470 (which I think was installed in one of failed attempts to recover) due to error in its postrm script.
Before doing crazy things such as forced uninstallation or doing restore from backup which I took just before full upgrade, I'd like to know how do people deal with such problems (in Debian and maybe also in Ubuntu).
Thanks, --- Omer Zak
(The above is less than total disaster because it happens in a laptop which is not my primary work PC. I plan to upgrade my work PC only after having successfully upgraded the laptop. I also kept enough backups to have the option to perform a fresh reinstall.)
On Fri, 2024-03-22 at 15:27 +0200, Rabin Yasharzadehe wrote:
My experience with NVIDIA and DKMS is that with very recent Kernels the build can break, but I didn't have that for some time now (years).
Also, I get my driver from my distro repo (Fedora), which reduce the chance of these issues.
-- Rabin
On Fri, 22 Mar 2024 at 12:01, Omer Zak w1@zak.co.il wrote:
I have upgraded a laptop from Debian 11 (Bullseye) to Debian 12 (Bookworm)
Now I am running into trouble when trying to build DKMS modules needed by nvidia-driver (the proprietary one) on the most recent kernel version (I tried both 6.1.76-1 and 6.6.13-1~bpo12+1).
There are some compile-time errors (different errors in different attempts).
Is there any information about known problems of getting Nvidia software to interoperate with Debian 12 (Bookworm) working on up- to- date kernels?
With Fedora, I usually drop/boot into single user, and remove any package with NVIDIA in their name,
rpm -qa | grep -i nvidia | xargs dnf remove -y # rpm -qa return all locally installed packages
this brings me to a clean state, which allow me to install the nvidia driver again. I think you should be able to do the same with Debian based distro, unless you used NVIDIA installer directly.
-- Rabin
On Fri, 22 Mar 2024 at 16:09, Omer Zak w1@zak.co.il wrote:
I am trying to recover from the bungled upgrade. I found some answers in the Internet. For example, to use kernel 6.1.0- 17 instead of 6.1.0-18 whose package version is 6.1.76-1) and to enable bookworm-updates ( https://forums.developer.nvidia.com/t/debian-12-and-nvidia-driver-nvidia-lin... ).
Now I am stuck with inability to uninstall xorg-video-nvidia-tesla-470 (which I think was installed in one of failed attempts to recover) due to error in its postrm script.
Before doing crazy things such as forced uninstallation or doing restore from backup which I took just before full upgrade, I'd like to know how do people deal with such problems (in Debian and maybe also in Ubuntu).
Thanks, --- Omer Zak
(The above is less than total disaster because it happens in a laptop which is not my primary work PC. I plan to upgrade my work PC only after having successfully upgraded the laptop. I also kept enough backups to have the option to perform a fresh reinstall.)
On Fri, 2024-03-22 at 15:27 +0200, Rabin Yasharzadehe wrote:
My experience with NVIDIA and DKMS is that with very recent Kernels the build can break, but I didn't have that for some time now (years).
Also, I get my driver from my distro repo (Fedora), which reduce the chance of these issues.
-- Rabin
On Fri, 22 Mar 2024 at 12:01, Omer Zak w1@zak.co.il wrote:
I have upgraded a laptop from Debian 11 (Bullseye) to Debian 12 (Bookworm)
Now I am running into trouble when trying to build DKMS modules needed by nvidia-driver (the proprietary one) on the most recent kernel version (I tried both 6.1.76-1 and 6.6.13-1~bpo12+1).
There are some compile-time errors (different errors in different attempts).
Is there any information about known problems of getting Nvidia software to interoperate with Debian 12 (Bookworm) working on up- to- date kernels?
-- One cannot argue with a Bayesian filter. Peter Lorand Peres My own blog is at https://tddpirate.zak.co.il/
My opinions, as expressed in this E-mail message, are mine alone. They do not represent the official policy of any organization with which I may be affiliated in any way. WARNING TO SPAMMERS: at https://www.zak.co.il/spamwarning.html
Linux-il mailing list -- linux-il@cs.huji.ac.il To unsubscribe send an email to linux-il-leave@cs.huji.ac.il
On Fri, Mar 22, 2024 at 04:08:44PM +0200, Omer Zak wrote:
I am trying to recover from the bungled upgrade. I found some answers in the Internet. For example, to use kernel 6.1.0- 17 instead of 6.1.0-18 whose package version is 6.1.76-1) and to enable bookworm-updates (https://forums.developer.nvidia.com/t/debian-12-and-nvidia-driver-nvidia-lin...).
Now I am stuck with inability to uninstall xorg-video-nvidia-tesla-470 (which I think was installed in one of failed attempts to recover) due to error in its postrm script.
What is the error? Why do you need to uninstall it? You can always edit the postrm script.
The reason I was stuck with the uninstallable package was that the failure of the attempt to uninstall it seemed to block aptitude from uninstalling/upgrading other packages.
I do not remember the exact error message. I think it was something about trying to do echo <(echo) and getting a complaint that the pipeline is broken.
Eventually, I uninstalled using dpkg --remove, figuring that any unmet dependencies will be fixed later by aptitude.
With this done, I was successful in sorting out things. The only remaining problem is that packages 'dkms', 'evdi-dkms', 'libevdi0' are not properly installed. This is due to missing header file 'drm.h', which is referred to by /var/lib/dkms/evdi/1.13.1/build/evdi_drm.h.
There are also some non-fatal annoyances:
1. The Google talkplugin source (http://dl.google.com/linux/talkplugin/deb/) has now an invalid signature, and aptitude shouts about this each time I update sources. This file claims to be automatically modified, so I didn't try to edit it. The same source is perfectly OK in a Debian 11 (Bullseye) installation.
2. The terminal font is different from the one to which I am used, even though both claim to be monospace ("Monospace 12" in Debian 11 / GNOME Terminal 3.38.3; "Monospace Regular 12" in Debian 12 / GNOME Terminal 3.46.8 for GNOME 43). I was not successful in identifying the actual font being used by the terminal program in both installations.
--- Omer Zak
On Mon, 2024-03-25 at 13:58 +0200, Tzafrir Cohen wrote:
On Fri, Mar 22, 2024 at 04:08:44PM +0200, Omer Zak wrote:
I am trying to recover from the bungled upgrade. I found some answers in the Internet. For example, to use kernel 6.1.0- 17 instead of 6.1.0-18 whose package version is 6.1.76-1) and to enable bookworm-updates (https://forums.developer.nvidia.com/t/debian-12-and-nvidia-driver-nvidia-lin... ).
Now I am stuck with inability to uninstall xorg-video-nvidia-tesla- 470 (which I think was installed in one of failed attempts to recover) due to error in its postrm script.
What is the error? Why do you need to uninstall it? You can always edit the postrm script.