You are not logged in.
FWIW for context for why we're so sure of that, the crashing threads prompted me to try and get the systemd packager to reconsider an unconditional load of udevadm trigger, which we originally assumed to be the cause (... which can lead to similar effects if you're unlucky) but as we've found in the discussion it's presumably the systemd daemon reload (that somewhat "has" to happen on systemd updates) which triggers the condition: https://gitlab.archlinux.org/archlinux/ … ote_176919
Offline
535xx no longer builds against 6.10, fyi. There's a comment on the AUR thread about applying a patch but the patch doesn't fully work.
I'm thinking of trying out 560 open once it moves to stable, but if that has problems I'm stuck on 535, because it seems like Nvidia has decided they'll never fix this bug with their proprietary drivers.
Last edited by mesaprotector (2024-07-25 17:23:11)
Offline
I'm stuck too... Currently running 555.
Downgraded to 535dkms but lost my second monitor and a bunch of settings - reverted back to 555 (timeshift backup)
I'm too nervous to run a -Syu due to kernel panic (which happened on the last one)....
Tried to move to nvidia-open but there is a outdated dependency with an older package that I cannot update due to conflict with nvidia package itself, (lib32-nvidia-utils) I feel backed into a corner lol! I have read a lot about this damn thing and there is a chance the 'bug' is due to other factors than nvidia drivers - perhaps even the kernel itself?
All the while my system is getting more and more out of sync package wise...
Last edited by Field_Master_111 (2024-07-25 17:39:50)
Offline
535xx no longer builds against 6.10, fyi. There's a comment on the AUR thread about applying a patch but the patch doesn't fully work.
I'm thinking of trying out 560 open once it moves to stable, but if that has problems I'm stuck on 535, because it seems like Nvidia has decided they'll never fix this bug with their proprietary drivers.
You should try using linux-lts. I've been using it for years now with no issue. I recommend.
Offline
With 555 the issue still occurs.
Also still sometimes the org crashes on system updates.
These issues doesn't happen if I remove the nvidia-dkms driver and only use the intel iGPU.
(And I only use the iGPU and use the nvidia with prime-run for games, but the issue happens in run cycles without using prime-run)
I can't use the "open" drivers since my gpu lacks the hardware support to run the user space closed source stuff firmware.
Last edited by MartinX3 (2024-07-26 09:20:26)
Offline
On my nvidia desktop (1030) I am using 555 with linux, and I resorted to rebooting with nouveau in text mode just for the update, then rebooting with nvidia afterwards. So far no crashes.
Offline
6.10.1-zen, nvidia-dkms 555,intel/nvidia optimus laptop, GTX 1660 ti, sometimes experiencing kernel panics upon shutdown.
What I noticed — kernel panic might occur only, if I boot my laptop connected to a power supply aka AC mode, then, unplug the power supply while the laptop is on, thus switching to a battery mode. Doesn't matter if I switch back to AC mode — unplugging the power supply once gives me 50 % chance of panicking upon shutdown. Also, I can't remember if I ever had a kernel panic, if I booted the laptop in a battery mode, without plugging it before shutting down.
Looks like nvidia gpu doesn't like power changes, and the kernel doesn't like the gpu.
Last edited by thoushaltnot (2024-07-28 19:44:01)
Closing all the applications (especially the browser) and then wait for 5-10 mins before shutting down is working for me.
I notice that when I shutdown right after closing the browser (or when the browser/or other application is still open) will always result in this panic issue
Offline
The Nvidia guy on their Linux forum says they might have fixed the bug with version 550.107.02.
I'm not sure what exactly that means for 555+ seeing as 550 is no longer the live version of the driver, but worth trying.
Last edited by mesaprotector (2024-07-31 18:45:26)
Offline
This workaround fixed the problem for me (Dell G16 7630, with i9-13900HX and NVIDIA RTX 4060):
systemctl stop lightdm # or stop your login manager / graphical session
pacman -S thermald
systemctl enable --now thermald
pacman -U https://archive.archlinux.org/repos/2023/10/31/extra/os/x86_64/nvidia-535.113.01-8-x86_64.pkg.tar.zst https://archive.archlinux.org/repos/2023/10/31/core/os/x86_64/linux-6.5.9.arch2-1-x86_64.pkg.tar.zst https://archive.archlinux.org/repos/2023/10/31/core/os/x86_64/linux-headers-6.5.9.arch2-1-x86_64.pkg.tar.zst https://archive.archlinux.org/repos/2023/10/31/extra/os/x86_64/nvidia-settings-535.113.01-1-x86_64.pkg.tar.zst https://archive.archlinux.org/repos/2023/10/31/extra/os/x86_64/nvidia-utils-535.113.01-2-x86_64.pkg.tar.zst https://archive.archlinux.org/repos/2023/10/31/multilib/os/x86_64/lib32-nvidia-utils-535.113.01-1-x86_64.pkg.tar.zst https://archive.archlinux.org/repos/2023/10/31/extra/os/x86_64/opencl-nvidia-535.113.01-2-x86_64.pkg.tar.zst https://archive.archlinux.org/repos/2023/10/31/extra/os/x86_64/libxnvctrl-535.113.01-1-x86_64.pkg.tar.zst
reboot
To avoid accidental upgrade in the short term, I also added this to /etc/pacman.conf :
IgnorePkg = nvidia linux linux-headers nvidia-settings nvidia-utils lib32-nvidia-utils opencl-nvidia libxnvctrl
Last edited by bouloumag (2024-08-01 01:47:52)
Offline
The canonical suggestion would be to use the 535xx dkms module in the AUR instead of locking the kernel to a dated version.
Offline
The Nvidia guy on their Linux forum says they might have fixed the bug with version 550.107.02.
I'm not sure what exactly that means for 555+ seeing as 550 is no longer the live version of the driver, but worth trying.
There are no new 555 drivers.
https://www.nvidia.com/download/driverR … 410/en-us/
That's funny because 550 is the production branch and 555 the feature branch.
I would expecting to add fixes to test in the 555 first.
Anyway I think I could finally workaround the kernel panics and upgrade crashes.
But that only works if you got a second GPU.
Using
nvidia-smi
I saw that the
xorg
process was running on my nvidia instead of my iGPU.
So I fixed my configuration issue and now no process runs on my GPU except I use
prime-run
.
So apps/games executed with
prime-run
may randomly crash but at least my whole system is stable now.
Last edited by MartinX3 (2024-08-01 11:20:23)
Offline
There are no new 555 drivers.
There's a direct link in the nvidia forum, https://us.download.nvidia.com/XFree86/ … 107.02.run
Offline
There are no new 555 drivers.
There's a direct link in the nvidia forum, https://us.download.nvidia.com/XFree86/ … 107.02.run
That's the 550 driver.
Offline
Which doesn't matter if they provided an updated version of 550 which might have the bug fix that's not yet present in the 555 release. Nvidia tends to have multiple active branches, so if they fixed something in a subrelease for 550 that might still be worth testing (though chances are they'll push out an update containing the same fix to other versions - but are potentially waiting for confirmation from users)
Offline
Any chance this fix will be provided to 555 and 560? Strange how an issue to newer drivers don't get the patch.
Kinda stuck on 555 since 550 doesn't really work for me.
Offline
nvidia-open 560 seems already fixed this problem for me, maybe you can have a try
Offline
It was never a problem with these fake open drivers.
Last edited by MartinX3 (2024-08-12 15:57:32)
Offline
Has anyone still been experiencing issues with 560?
Just curious, because until this day I am still on nvidia-535xx from AUR (better safe than sorry), but if the issues are actually resolved, I will be happy to switch back to the newer one of course.
Offline
Everything is normal. I am using 560 drivers for one month. I didn't have any issues.
Offline
Still broken for me on 560.35.03
Offline
What exactly? Over the course of this thread a somewhat clear cause and effect was found (updates involving a reload of the systemd daemon) if you have any issues outside of that you're most likely not running into the same issue.
And as said, nvidia-open assuming you have a Turing+ GPU was never affected.
Going ahead and close this old thread, based on the reports that it should be fixed, please open new topics with new and actually current details about whatever it is you're facing.
Offline