Installs fine, boot fails #1
Loading…
x
Reference in New Issue
Block a user
No description provided.
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
This package nuked my openmamba install.
It doesn't show any error during the install process, after installing it openmamba fails to boot.
Package version
1:470.256.02-1mamba
Legacy nvidia-470 has been updated so that the kernel module builds and a theoretical list of needed libraries and files are installed.
Unfortunately compatible hardware is not available for the packager to test and nvidia drivers are not provided with distribution friendly install scripts so that every library and component needs to be manually installed when packaged and may change from a version to another. Because of this the driver needs beta testing support to ensure it works and to keep it working with future updates.
You might want to provide information on how the boot fails in order to understand at which boot stage the problem happens and which kind of investigation and possible recover can be performed.
If instead you are looking for a system ready with nvidia-470 driver support, switching to another Linux distribution may be recommended until there is the opportunity to test and fix the nvidia-470 driver in openmamba.
I would like to provide information, but I don't know how. Please direct me on how to provide information that may be helpful.
Currently, after reboot, I see the "openmamba is starting..." message on a black screen, it disappears and appears back I believe 3 times, and then it is stuck on the screen, nothing else happens.
The first time it flashes, the (usb) keyboard turns on.
The system does restart when I press ctrl+alt+del, so the system responds to keyboard input, but I can't see anything besides that message.
I have tried to boot with the debug flag, but nothing else appears on screen.
Ok.
Does pressing Ctrl + Alt + F2 (or F3, F4, etc.) not work to open a console?
If not, you may retry after adding this to the grub kernel command line:
nvidia_drm.modeset=1 nvidia_drm.fbdev=1
I can confirm ctrl+alt+f2 (or f3...) does not change what I see on the screen.
I have added the
nvidia_drm.modeset=1 nvidia_drm.fbdev=1
kernel flags using grub's menu. Now instead of "openmamba starting..." I see "booting a command list". I believe this is because I also addeddebug
and removedquiet
. The text flashes 5 times in total. ctrl+alt+f2 (or f3...) remain unresponsive, ctrl+alt+del results in the computer rebooting.I have also tried with
nvidia-drm.modeset=1 nvidia-drm.fbdev=1
because both_
and-
came up on google search results. No difference.Assuming that nvidia module loading is causing the issue you might try to disable loading from initramfs and/or modprobe with this kernel command line options:
rd.driver.blacklist=nvidia,nvidia_drm,nvidia_modeset,nvidia_uvm modprobe.blacklist=nvidia,nvidia_drm,nvidia_modeset,nvidia_uvm
Just a subset of the above line should be needed but I put all options for completeness.
I've been busy, I'm sorry for the late reply.
I have attempted this change, with them the screen does not flash while starting openmamba, however it is still stuck in the same screen.
Is there any information that I could attempt to fetch from the os, such as logs that may contain more information on the issue? I can try to run a live cd, mount the partition and get the potentially relevant information.
You might try to boot in non-graphical mode adding a
3
to the kernel command line orinit=/bin/bash
,If you boot from a livecd and have automatic login into KDE Plasma you may try to disable it because the problem might be with Wayland. To do so, comment the
[Autologin]
block in file/etc/sddm.conf.d/kde_settings.conf
.I have no clues on where the boot stops, if before or after mounting the root filesystem, in the latter case you might find something in the log files, by the way you should mount the openmamba rootfs, bind mount /dev/, /sys, /proc, /run etc. and chroot and see the last boot journal with
journalctl
.From chroot you might also uninstall the package and attempt to reinstall and do tests later, although the nvidia modules might still be in the initramfs, which should be regenerared with dracut providing the appropriare options to match the correct kernel version. Not very easy though.
Honestly I don't know if the issue was this package any more.
I could not find useful information on the logs. I did find that openmamba received my ctrl+alt+f2 but failed there.
I have then reinstalled openmamba, after openmamba was installed I did
dnf update -y
. During this update I got a lot of linker errors, then on reboot bash would crash: whenever I opened a terminal emulator I would get"warning: /bin/bash has crashed
after a few seconds of hanging. At that point I could not even installnvidia-470
again to perform further testing.I have tested an installation from the latest livecd in VirtualBox and then run
sudo dnf update
. Yes, there are a lot of linker warnings which are all related tolibvtk*
but this is not related to any critical issue. The updates completed successully and the system rebooted to the desktop with no problems also running a console, so I cannot reproduce this problem.It may be an issue with the hardware but it is difficult to get information if bash crashes, You might check if the probem is reproducible after a reinstallation and/or check from a livecd that bash binary and anything else do not result corrupted. Quite weird problem BTW.
I have installed openmamba again, updatd with
sudo dnf update
successfully. Nothing crazy happened. No idea why the outcome is different this time. I did notice all the ld warnings disappeared, because some package was in an awkward stage?I am now installing nvidia-470.
Running scriptlet: nvidia-470-1:470.256.02-1mamba.x86_64
is taking a very long time.Before this, please note the line
find: ‘/var/lib/dkms/*/*/source’: No such file or directory
I am about to reboot, pray for me.
Update: unfortunately it didn't go well. Just like before, I see "openmamba starting..." flashing.
I think I can try this whole thing again tonight and I am looking forward to being told what to do to provide as much information as needed in order to get this package working right. (I also think that it would be nice to get nouveau to work right, because I know it can work right with this hardware)
I have uploaded the systemd logs to my google drive here. I tried to view them with
journalctl --file <file> | grep nvidia
These were the only available logs, I admittedly tried to start openmamba twice and while the first time I let it attempt to start for 20 full minutes, I have it far less time the second time, and I believe these logs are from the second time only, unfortunately.
The ldconfig warning may depend on the different packages installation ordering and when ldconfig command is triggered and are not relevant, also the error reported by dkms is about a legacy cleanup command and is not a problem.
The system log shows that
nvidia
,nvidia-drm
andnvidia-modeset
modules are loaded without apparent errors but thesddm
is failing with this error:feb 06 17:03:01 ws00 sddm[5275]: Running: /usr/bin/X -nolisten tcp -background none -seat seat0 vt2 -auth /run/sddm/xauth_mJlXnd -noreset -displayfd 16 feb 06 17:03:06 ws00 sddm[5275]: Failed to read display number from pipe
So first, it seems that the system is running but there is nothing on screen because there is not a working framebuffer device. I'm trying here a fix setting the option
fbdev=1
which we altready tried to set from the kernel command line by the way.For the
sddm
problem I don't know exaclty, but I ported from the current nvidia driver in openmamba (which works) a fix to the installation ofnvidia-drm_gbm.so
, although I think it is only related to Wayland and Arch Linux has commented it in its nvidia-470 package.These changes are in the
nvidia-470-470.256.02-2mamba
package which will be available as an update in 1-2 hours.Good news: I am able to see and log in with sddm.
Bad news: kwin uses 100% of all my cores at all times now. There is a baloo_file_extractor too taking some processor time, but nothing like kwin. This causes the computer to be unresponsive, mouse pointer movement is rendered once every 6 or 7 seconds.
How can we debug this?
The official and general information say that nvidia legacy drivers < 495 only provide a wayland support mechanism which is not supported by KDE Plasma, so you may want to login in X11 mode.
If you want next to debug the problem with the nouveau driver I suggest to open a issue in the xf86-video-nouveau project page.
You are correct. This issue is solved.