[SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

k3yss · 2024-05-13 10:24:08

I managed to downgrade to version 545, but it required downgrading my kernel as well. I opted to skip version 535 because it required me to install the dkms packages, which seemed unnecessary.

DFOwl · 2024-05-13 19:21:39

seth wrote:

Update, for anyone affected by this:
If you feel adventurous to try the latest nvidia driver again, the problem is increasingly likely the nvidia_uvm module.
This is mostly relevant for cuda and to use the GPU in containers, but also for https://wiki.archlinux.org/title/NVIDIA … with_NVENC and eg. utilized by gimp for HW acceleration.
If you critically rely on that, this isn't a viable solution but it would still be awesome if you can confirm the condition.
Afaict the module also comes w/ a ton of paramters, so if you've it around, posting
modinfo nvidia_uvm
might shed further light.
https://wiki.archlinux.org/title/Kernel … and_line_2
module_blacklist=nvidia_uvm
And if anyone here happens to post at https://forums.developer.nvidia.com/t/s … /284772/27 you might want to forward this.
Particularily if you could confirm it.

I am using NVIDIA LTS, but it is definitely above the 550 version of the driver. Is it still better to downgrade the driver to an older version and blacklist it from updates?

seth · 2024-05-13 21:32:55

https://archlinux.org/packages/extra/x86_64/nvidia-lts/ is currently at 550.78 and as far as we know the issue has not been fixed.
If you want to sidestep the issue (which turned out to not be limited to UVM) you can use https://aur.archlinux.org/packages/nvidia-535xx-dkms

angrynapkin · 2024-05-25 21:06:34

k3yss wrote:

I managed to downgrade to version 545, but it required downgrading my kernel as well. I opted to skip version 535 because it required me to install the dkms packages, which seemed unnecessary.

Any chance you can post the package versions of the kernel and nvidia youre using?

I have tried lts and normal, tried nvidia with and without dmks.
Nothing I try works anymore. I just updated today.
I got nvidia dkms to succeed by downgrading gcc too, but then on reboot I had a dead black screen.
I ended up ripping everything out and just doing a full system upgrade to get back to all current packages without installing nvidia. I am back in wayland, but now I got no 3d.

seth · 2024-05-25 22:24:35

https://aur.archlinux.org/packages/nvidia-535xx-dkms doesn't build for you?

angrynapkin · 2024-05-27 18:45:25

seth wrote:

https://aur.archlinux.org/packages/nvidia-535xx-dkms doesn't build for you?

I must have missed the post about this being in AUR.
Yes I got this to install fine, thank you. I have graphics again.

jongeduard · 2024-06-01 18:49:10

Just out of curiosity, has anyone here explored the possibility of making their own built from the new official Gitlab for Arch Linux?

Because I would say, checking by out the right git branch, https://gitlab.archlinux.org/archlinux/ … _type=tags, building against the current kernel might do the job.
At least 545 is a bit newer than the 535 from the AUR.

But I'm not sure about how it's exactly going to work out with the nvidia-utils dependency and if it will be a good idea.

V1del · 2024-06-02 11:21:02

You could generally try that, but 535 is an actually updated branch that receives official fixes. Picking an older random version might need to fix up much more issues for kernel or compiler compatibility. And if you don't have an actual pressing need because you require a certain feature not present in 535 picking that is quite a sustainable option.

DFOwl · 2024-06-11 19:11:28

I just updated to `nvidia-lts 1:550.90.07-1` and regrettably, I experienced a kernel panic again. Do note, though, that when I downgraded to 545, I also had to downgrade my kernel. I will try using 535 from AUR, since @V1del mentioned that it still gets updated. Hopefully this issue will get solved, but for now... gotta use the older versions.

Drivers can be real crap sometimes.

k3yss · 2024-06-11 19:21:15

kernel 6.7.6-arch1-1 and nvidia 545.29.06 works for me

DFOwl · 2024-06-11 19:25:21

k3yss wrote:

kernel 6.7.6-arch1-1 and nvidia 545.29.06 works for me

Ah, well, I guess I will try either way. If anyone gets the newest version of nvidia to work, please add it to the solution here!

Last edited by DFOwl (2024-06-11 19:26:39)

jongeduard · 2024-06-19 20:58:46

V1del wrote:

You could generally try that, but 535 is an actually updated branch that receives official fixes. Picking an older random version might need to fix up much more issues for kernel or compiler compatibility. And if you don't have an actual pressing need because you require a certain feature not present in 535 picking that is quite a sustainable option.

Good point, I get it, it's just the difference between a stable and a newer branch. I already realized that later after asking that question here. There's hardly any point in using that old update of 545.
And it looks like the 535 works fine here, so no hurries I think.

Maybe Arch Linux should just give multiple NVIDIA branch options in the official repos.
I now remember that Ubuntu does that too, and I believe that's a really great way of doing things.

Last edited by jongeduard (2024-06-19 21:02:39)

DFOwl · 2024-06-21 16:53:47

jongeduard wrote:

V1del wrote:
You could generally try that, but 535 is an actually updated branch that receives official fixes. Picking an older random version might need to fix up much more issues for kernel or compiler compatibility. And if you don't have an actual pressing need because you require a certain feature not present in 535 picking that is quite a sustainable option.
Good point, I get it, it's just the difference between a stable and a newer branch. I already realized that later after asking that question here. There's hardly any point in using that old update of 545.
And it looks like the 535 works fine here, so no hurries I think.
Maybe Arch Linux should just give multiple NVIDIA branch options in the official repos.
I now remember that Ubuntu does that too, and I believe that's a really great way of doing things.

I can confirm that the AUR 535 build of the nvidia driver works well on an MSI gaming laptop with an nvidia card and LTS kernel. So far I haven't tested out the regular kernel, but I don't think I've got a reason to.

As for nvidia branches... Arch Linux is a rolling release distro. By its very nature, it always has the newest version of packages in its database. If you want a specific branch, well, you've got the AUR! It's very handy.

V1del · 2024-06-22 11:47:06

FWIW small headsup for people on the 535 branch, if you want to use suspension with systemd 256 installed you're advised to create override files to disable user session freezing: https://gitlab.archlinux.org/archlinux/ … 4b45a60708

DFOwl · 2024-06-22 11:54:57

V1del wrote:

FWIW small headsup for people on the 535 branch, if you want to use suspension with systemd 256 installed you're advised to create override files to disable user session freezing: https://gitlab.archlinux.org/archlinux/ … 4b45a60708

What does user session freezing do? Apologies, this is the first time I've heard of this.

So we need to create the files systemd-homed-override.conf with

[Service]
Environment="SYSTEMD_HOME_LOCK_FREEZE_SESSION=false"

and systemd-suspend-override.conf with

[Service]
Environment="SYSTEMD_SLEEP_FREEZE_USER_SESSIONS=false"

if we want to disable this? Where exactly should we place these files?

Last edited by DFOwl (2024-06-22 12:11:57)

seth · 2024-06-22 14:42:28

https://bbs.archlinux.org/viewtopic.php … 8#p2179308

jongeduard · 2024-06-29 10:57:37

seth wrote:

https://bbs.archlinux.org/viewtopic.php … 8#p2179308

Thanks for sharing this. Because it's good to know about it.
But to me it looks like it's a different problem. So we basically deal with 2 different problems now?

Personally I never configured that whole PreserveVideoMemoryAllocation setting anyway, as I was also still running Xorg (Xfce destktop) mostly where it's less of a problem, and mostly dynamic graphics mode (so that my AMD CPU built in GPU is used the primarily), and occasionally only dedicated NVIDIA graphics (I can switch this setting in my BIOS). So it was generally less of an issue.

(The kernel-panics-on-shutdown-problem with 550 occurred in all my use cases, until I switched to 535 from the AUR (which still works great), regardless of this setting.)

Now I switched to KDE Plasma running on Wayland, it clearly makes sense to get this working, after reopening my laptop lid without this setting enabled all graphics break. Steps described on the Arch wiki (NVIDIA/Tips and tricks) do the job.

But now I understand systemd broke something here, and we are supposed to fix that too, but does that also apply to 535 as well?
Because, for me like it already works without that additional workaround. Am I missing something?

Last edited by jongeduard (2024-06-29 10:59:33)

jongeduard · 2024-06-29 11:05:15

DFOwl wrote:

As for nvidia branches... Arch Linux is a rolling release distro. By its very nature, it always has the newest version of packages in its database. If you want a specific branch, well, you've got the AUR! It's very handy.

I see your point, but just think about the Linux LTS, which is an older branch of the kernel and we can install it.
Arch Linux makes some exceptions sometimes.
And installing things running in kernel space (such as a driver) from the official distribution is still more secure than AUR, for which we have to trust someone else.

seth · 2024-06-29 14:52:57

But to me it looks like it's a different problem.

The thread is, but they ran into the systemd 256 issue and the specifically linked post simply answers the question "where to put that".

MartinX3 · 2024-06-30 15:44:59

And now we got the 555 drivers.
Let's see if it still randomly crashes.

Field_Master_111 · 2024-07-01 19:24:28

I'm running 555.58 - and so far it has only crashed once. That was during an -Syu update involving other packages (such as qt6, libjuice, kwin etc). Other than that I have not had any issues.

seth · 2024-07-01 20:43:42

The crash is not because nvidia gets update but because systemd re-executes (and crashes what's suggest to be because the nvidia drivers compromise the kernel memory) - ie. the most relevant package in your updates to trigger this is systemd, not nvidia.

Field_Master_111 · 2024-07-02 19:35:19

Oh, ok... understood.

So the options are to roll back / nvidia open driver or.... wait it out?

I've heard 560 will solve this but how long will that take to be released?

seth · 2024-07-02 19:58:38

https://aur.archlinux.org/packages/nvidia-535xx-dkms should not ba affected by this, nor the nvidia-open drivers (if your GPU is supported, but nb. that there're pending issues esp. around S3 support)

syscrusher · 2024-07-09 16:53:25

Field_Master_111 wrote:

I'm running 555.58 - and so far it has only crashed once. That was during an -Syu update involving other packages (such as qt6, libjuice, kwin etc). Other than that I have not had any issues.

I'm also running 555.58 and have had only one crash (which happened this morning), but it was a doozy from which it took me 3 hours to recover because tons of library files (*.so) were zapped.

The crash today involved a -Syu update in which both nvidia and systemd were affected. Based on the recent updates here, I was probably doomed from the start. The timing of the crash supports the recent assertion by seth that it was during the reload of systemd.

Arch Linux

#51 2024-05-13 10:24:08

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#52 2024-05-13 19:21:39

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#53 2024-05-13 21:32:55

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#54 2024-05-25 21:06:34

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#55 2024-05-25 22:24:35

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#56 2024-05-27 18:45:25

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#57 2024-06-01 18:49:10

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#58 2024-06-02 11:21:02

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#59 2024-06-11 19:11:28

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#60 2024-06-11 19:21:15

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#61 2024-06-11 19:25:21

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#62 2024-06-19 20:58:46

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#63 2024-06-21 16:53:47

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#64 2024-06-22 11:47:06

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#65 2024-06-22 11:54:57

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#66 2024-06-22 14:42:28

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#67 2024-06-29 10:57:37

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#68 2024-06-29 11:05:15

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#69 2024-06-29 14:52:57

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#70 2024-06-30 15:44:59

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#71 2024-07-01 19:24:28

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#72 2024-07-01 20:43:42

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#73 2024-07-02 19:35:19

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#74 2024-07-02 19:58:38

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

#75 2024-07-09 16:53:25

Re: [SOLVED] Repeated kernel problems/freezes since 6.7.6 (Nvidia 550)

Board footer