Preserving memory allocation with Nvidia and suspend

[napolitan]

Hello, everyone.

I’m trying to fix the issue where the Wayland session comes back from suspend with a broken screen when using the Nvidia proprietary drivers. For that I came up with the article on the wiki that seems to point towards a solution but with no success.

I did create a file on /etc/modprobe.d/nvidia-power-management.conf with

options nvidia NVreg_PreserveVideoMemoryAllocations=1 NVreg_TemporaryFilePath=/var/tmp

as pointed out by the article, and enabled both the nvidia-suspend and nvidia-resume services, but still no dice.

I guess I didn’t understand how am I supposed to deal with this temporary file, because from what I’ve seen from other posts, this should fix it.

I am using all the latest drivers and kernels, and I’m on a desktop with a Nvidia RTX 2070 card, so no hybrid graphics.

What I’m doing wrong here?

[jake99]

Hi,

It looks like you are doing nothing wrong

How much video memory do you have on your graphics card? Do you have enough free space on the partition where the /var/tmp is located? What filesystem are you using there?

Can you show the output of this

glxinfo -B | grep -i memory

and this command

df -hT

Have you checked if your kernel module is loaded with correct params

sudo systool -m nvidia -av

Can you provide some more details from logs, for example

journalctl -b -0 --grep=nvidia --no-hostname

[napolitan]

Hello!

So it turns out I don’t have enough space inside of /var/tmp for that, huh. I was counting with the whole root filesystem as available space

The output of

glxinfo -B | grep -i memory

is

Memory info (GL_NVX_gpu_memory_info):
    Dedicated video memory: 8192 MB
    Total available memory: 8192 MB
    Currently available dedicated video memory: 7603 MB

df -hT

Filesystem     Type      Size  Used Avail Use% Mounted on
devtmpfs       devtmpfs  4.0M     0  4.0M   0% /dev
tmpfs          tmpfs     7.8G  252M  7.6G   4% /dev/shm
tmpfs          tmpfs     3.2G  1.9M  3.2G   1% /run
/dev/nvme0n1p7 xfs       103G   24G   80G  23% /
tmpfs          tmpfs     7.8G  9.3M  7.8G   1% /tmp
tmpfs          tmpfs     1.6G  120K  1.6G   1% /run/user/1000

For the Nvidia kernel modules I had to enable the usual

nvidia nvidia_modeset nvidia_uvm nvidia_drm

inside of /etc/dracut.conf to even have the Wayland session available on gdm, however df -hT only shows the nvidia kernel module enabled… I have also enabled nvidia-drm.modeset=1and NVreg_PreserveVideoMemoryAllocations=1 inside of /etc/kernel/cmdline but I don’t remember if I was supposed to do that. I’m used to edit mkinitcipio.conf and actually just used dracut and systemd-boot for the first time this week for my recent install. There might be some problem here as well.

[napolitan]

Just for further reference if someone ever stumble upon this with the same issue, I fixed it.

Besides the steps above I was missing both additional steps:

• Increase the size of /var/tmp.

• Include the kernel modules properly.

In my case, I’m just using dracut and systemd-boot as defaults, so according to this and this article I created a file under /etc/dracut.conf.d/ named nvidia.conf with the options

force_drivers+=" nvidia nvidia_modeset nvidia_uvm nvidia_drm "

then regenerated the initrd.

I haven’t tested this with mkinitcpio but I guess if the /var/tmp partition was big enough it would be enough to just pass the Nvidia modules above as usual.

Thank you @jake99 for the tips and help!

[ricklinux]

http://us.download.nvidia.com/XFree86/Linux-x86_64/470.63.01/README/powermanagement.html#Backgroundc73a7