SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

timminybo · 2025-12-18 21:52:52

Last night, I performed a full system upgrade which included going from linux 6.17.9.arch1-1 to 6.18.1.arch1-2 and from nvidia 580.105.08-4 to nvidia 580.119.02-2 and powered down the machine.

This morning, I logged in and executed startx. It failed with message: nvidia failed to acquire modesetting permission.
I rebooted and executed startx which failed with a different message: nvidia xf86enableio failed to enable i/o ports 0000-03ff operation not permitted.

A quick fix was to boot using the lts kernel which worked. startx worked just fine with that kernel.

Both kernels have the same kernel options/parameters:

options rd.luks.name=...=root rd.luks.name=...=home rd.luks.name=...=data ipv6.disable=1 nvidia_drm.modeset=1 nvidia_drm.fbdev=0 NVreg_PreserveVideoMemoryAllocations=1

I read a couple of other recent forum posts with different error messages that indicated that the two nvidia parameters above are no longer working. Perhaps that is responsible for my errors as well.

Last edited by timminybo (2025-12-23 19:11:45)

gerasc · 2025-12-18 23:08:55

I also upgraded NVIDIA and Linux, and when I did, I got the same error: "failed to acquire modesetting permission." Then I added "nvidia_drm.fbdev=1" to the kernel parameters and was able to get in with startx. It worked fine at first, but later it started having a ton of flickering and tearing, so I ended up reverting the upgrade for nvidia, nvidia-utils, and linux. Currenlty using (Linux 6.17.9-arch1-1, nvidia: 580.105.08-4, nvidia-utils 580.105.08-5) no flickering or tearing

Last edited by gerasc (2025-12-18 23:09:29)

timminybo · 2025-12-21 17:01:38

Attempted another system upgrade a couple of hours ago. Opted not to install nvidia-open since my nvidia Quadro P1000 (GP107GL = NV130 = Pascal) is not supported by it. Had dependency cycle with nvidia-utils I think. Removed packages nvidia, nvidia-utils, nvidia-lts, nvidia-settings, lib32-nvidia-utils with intention of reinstalling them after the upgrade. Upgraded. Now nvidia package doesn't exist anymore. Searched around and found this page: https://archlinux.org/news/nvidia-590-d … l-modules/
which says
NVIDIA 590 driver drops Pascal and lower support.
Uninstall the official nvidia, nvidia-lts, or nvidia-dkms packages.
Install nvidia-580xx-dkms from the AUR

I installed nvidia-580xx-dkms from AUR. Now when I run the linux kernel (not linux-lts) I can now startx.

But ... picom doesnt start ... backend failed to initialize

In picom config, I changed the backend from egl to glx. restarted i3 wm.
picom runs but when I type, nothing appears in the terminal. If I restart i3 in place, what I typed is now showing but I still can't type.

Switched to xrender backend. Picom runs. I can type as normal in the terminal. Performance is poor though. Polybar blanks out whenever switching workspaces, opening a second terminal window makes the first terminal window disappear.

I tried booting into the linux-lts kernel and got an error: vesa: refusing to run, framebuffer or dri device present. Oh. Right. I removed nvidia-lts earlier.

Maybe I should see what happens if I install the nvidia-open driver. I don't want to make more of a mess though. I'm not sure what else to try. Any suggestions?

timminybo · 2025-12-21 19:26:56

Some new discoveries. My second monitor is not detected. When I run xrandr, I don't get monitor ids like DP-1 or DP-3 which I had before the most recent upgrade. For identifier, xrandr just says 0x37 instead of DP-3.

plrigaux · 2025-12-21 21:05:28

Hello

I have an Optimus card in my Laptop, meaning that I have a NVIDIA GPU with a Intel GPU. I did an pacman update yesterday and I shutdown. This morning my HDMI monitor is not detected as well. I checked the monitor is working and if I boot from my broken Fedoda partition I see log on my second Monitor, it works.

I will try the revert and update if my HDMI monitor detection is fixed.

EDIT

I did the revert, but my laptop was not able to detect my second monitor

nvidia-open 580.119.02-2
nvidia-open-lts 1:580.119.02-3
nvidia-utils 580.119.02-1
linux 6.17.9.arch1-1

Above here my config

Last edited by plrigaux (2025-12-22 00:14:19)

timminybo · 2025-12-21 22:11:32

I'm beginning to suspect that the Nvidia card is not being used.

glxinfo | grep "OpenGl renderer" shows this output:

string: llvmpip (LLVM 21.1.6, 256 bits)

nvidia-smi command shows this output:

NVIDIA-SMA has failed because it couldn't communicate with the NVIDIA driver. Make sure the latest NVIDIA driver is installed and running.

I installed the nvidia-580xx-dkms from the AUR. When I click on nvidia-580xx-dkms in the AUR package list, the link takes me to
https://aur.archlinux.org/packages/nvidia-580xx-dkms which has the git clone link https://aur.archlinux.org/nvidia-580xx-utils.git
I was expecting a dkms page not a utils page but a https://aur.archlinux.org/nvidia-580xx-dkms page doesn't exist. I don't know if that is relevant at all.

I believe there is Intel onboard graphics on the motherboard. Perhaps that is being used instead of the NVIDIA card.

I found this message from journalctl -b: systemd-modules-load[856] Failed to find module 'nvidia_uvm'
and this one "xdg-desktop-portal-gtk: cannot open display

dmseg: Load kernel module drm was skipped because of an unmet condition check (ConditionKernelModuleLoaded=!drm)

timminybo · 2025-12-21 22:46:26

In the "Latest Comments" of this page https://aur.archlinux.org/packages/nvidia-580xx-dkms I saw mention of installing linux-headers package being necessary. So I installed that which also installed a package pahole.

After rebooting, I cannot startx now. But I do have console output on both monitors.

Xorg log has this:

503.699] (**) OutputClass "nvidia" ModulePath extended to "/usr/lib/nvidia/xorg,/usr/lib/xorg/modules,/usr/lib/xorg/modules"
[ 503.701] (II) Applying OutputClass "nvidia" to /dev/dri/card1
[ 503.701] loading driver: nvidia
[ 503.701] (==) Matched nvidia as autoconfigured driver 2
[ 503.702] (II) LoadModule: "nvidia"
[ 503.702] (II) Loading /usr/lib/xorg/modules/drivers/nvidia_drv.so
[ 503.702] (II) Module nvidia: vendor="NVIDIA Corporation"
[ 504.227] (II) Applying OutputClass "nvidia" options to /dev/dri/card1
[ 504.227] (II) Loading sub module "glxserver_nvidia"
[ 504.227] (II) LoadModule: "glxserver_nvidia"
[ 504.227] (II) Loading /usr/lib/nvidia/xorg/libglxserver_nvidia.so
[ 504.238] (II) Module glxserver_nvidia: vendor="NVIDIA Corporation"
[ 504.240] (II) UnloadModule: "nvidia"
[ 504.240] (II) UnloadSubModule: "glxserver_nvidia"
[ 504.240] (II) Unloading glxserver_nvidia

journalctl -b and dmesg have these messages:

17.974952] nvidia: loading out-of-tree module taints kernel.
[ 17.974962] nvidia: module license 'NVIDIA' taints kernel.
[ 17.974966] nvidia: module verification failed: signature and/or required key missing - tainting kernel
[ 17.974967] nvidia: module license taints kernel.
[ 18.497494] nvidia-nvlink: Nvlink Core is being initialized, major device number 235
[ 18.508677] nvidia 0000:01:00.0: vgaarb: VGA decodes changed: olddecodes=io+mem,decodes=none:owns=io+mem
[ 18.785502] nvidia_uvm: module uses symbols nvUvmInterfaceDisableAccessCntr from proprietary module nvidia, inheriting taint.
[ 18.786857] nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 580.119.02 Mon Dec 8 07:37:54 UTC 2025
[ 18.957910] [drm] [nvidia-drm] [GPU ID 0x00000100] Loading driver
[ 19.684037] [drm] Initialized nvidia-drm 0.0.0 for 0000:01:00.0 on minor 1

If I am interpreting all of these messages correctly, it seems like the nvidia card is now being used but the X11 server isn't starting. I suppose that is progress.

timminybo · 2025-12-21 23:25:05

with the linux-headers package installed and logged in to console, nvidia-smi shows this:

+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|================================================
| No running processes found |
+-----------------------------------------------------------------------------------------+

which indicates to me that the nvidia card and driver are at least communicating with each other.

startx gives console message: xf86EnableIO: failed to enable I/O ports 0000-03ff Operation not permitted). Fatal server error. no screens found.

This was my original problem but maybe it is some type of generic message.

I'm stuck for now.

seth · 2025-12-21 23:43:20

Please post your Xorg log, you can cat it into " | curl -F 'file=@-' 0x0.st"

timminybo · 2025-12-21 23:55:03

Thank you for your reply. Please see latest Xorg log here: https://0x0.st/Psda.txt

seth · 2025-12-22 00:12:16

"nvidia_drm.modeset=1" is default, "nvidia_drm.fbdev=0" isn't a good idea because the driver cannot take over from the simpledrm device (which you're trying to use and fail because of the incompatible nvidia libraries) and "NVreg_PreserveVideoMemoryAllocations=1" should be "nvidia.NVreg_PreserveVideoMemoryAllocations=1" but is default anyway.

Remove all three and every single xf86-video-* package you've installed.
Then report back w/ an updated X11 log

timminybo · 2025-12-22 00:33:56

I removed the 3 kernel parameters. I removed xf86-video-nouveau and xf86-video-vesa packages. The latest xorg log after rebooting and startx is here: https://0x0.st/Psdv.txt

Note: at console after login, there is a strip of black along bottom and right side of main monitor.

X11 session experience seems otherwise normal so far.

Last edited by timminybo (2025-12-23 07:11:42)

timminybo · 2025-12-22 01:59:24

I've been running my normal X11 programs for the past hour or so and the behavior seems fine with the nvidia-580xx-dkms package from the AUR and the 6.18.2-arch2-1 kernel with the changes that Seth suggested.

The console at login looks different: there is a black strip along the right side and bottom of the screen and the font size is smaller (maybe the font is different too, I don't know). But I don't spend much time in a tty console or virtual console. I'll leave this open for a day or two before marking it SOLVED in case I find out what is causing these new minor issues and can post the solution.

Seth, I'm grateful for your help. You are the arch linux whisperer. Thanks for the rescue.

timminybo · 2025-12-22 04:50:32

My best guess about the console issues is the following:

I have a two monitor setup. My primary monitor (on the right) is physically larger and has a higher resolution than the monitor on the left.
I think the screen size of the left monitor is being used for both monitors at boot. That would explain the dead real estate (black border only the right side and bottom of the larger monitor on the right) and the smaller text size.

Once X11 starts, my xinitrc file has xrandr commands that setup the monitors how I want.

After coming up with this monitor settings swapping at boot idea, I now had some keywords to search with and found this post:

bbs.archlinux.org/viewtopic.php?id=278369

It sounds like my issue. And there's Seth again. He's everywhere.

Tomorrow I'll re-read that post and start fresh again.

(Maybe I can just physically swap the video cables at the card ports and edit my xinitrc and i3 config files accordingly.)

JontomXire · 2025-12-22 08:31:00

Have you tried the AUR packages? They partly solved my problem, but I am getting wierd behaviour, and when I run a game through Lutris, the window never appeared.

nvidia-580xx-dkms
nvidia-580xx-utils
lib32-nvidia-580xx-utils
nvidia-580xx-settings

seth · 2025-12-22 09:17:21

The console at login looks different: there is a black strip along the right side and bottom of the screen and the font size is smaller

You might have switched from the VGA to the framebuffer console or are using a bad modeline (EDID problems?)

You can force modes https://www.kernel.org/doc/Documentation/fb/modedb.rst and in doubt want to use a bigger console font, https://wiki.archlinux.org/title/Linux_console#Fonts
What's the output of

xrandr --verbose

?

Cpt_Liam · 2025-12-22 14:03:46

TLDR: I can confirm that the linux-headers package seems to be a missing dependency of nvidia-580xx-dkms.

I just went through the nvidia upgrade process with an older graphics card (GTX 1070) which falls into the version range covered by nvidia-580xx-dkms (a helpful source for verification are the links under https://wiki.archlinux.org/title/NVIDIA). During installation there were some errors about missing headers/modules (something "kernel headers missing for nvidia", I don't remember the details, sorry). After rebooting, X didn't work. There was a login screen but in very low resolution, clearly not using the nvidia driver.

The fix was to install linux-headers (and reinstalling nvidia-580xx-dkms for good measure to trigger repackaging the initrd image). Works now.

WorMzy · 2025-12-22 14:31:39

Cpt_Liam wrote:

TLDR: I can confirm that the linux-headers package seems to be a missing dependency of nvidia-580xx-dkms.

It's not. The dkms package optdepends on the various -header packages for the kernels that Arch ships: https://gitlab.archlinux.org/archlinux/ … ILD#L25-28

You should install the ones that make sense for you.

https://wiki.archlinux.org/title/Dynami … stallation

Mod note: Moving to AUR Issues.

Scimmia · 2025-12-22 14:31:55

Cpt_Liam wrote:

TLDR: I can confirm that the linux-headers package seems to be a missing dependency of nvidia-580xx-dkms.

It's not. You need the header package for whatever kernel you use. The optional deps were printed when you installed the dkms package, and the dkms wiki page covers this.

vvv2542 · 2025-12-22 23:08:43

timminybo wrote:

Last night, I performed a full system upgrade which included going from linux 6.17.9.arch1-1 to 6.18.1.arch1-2 and from nvidia 580.105.08-4 to nvidia 580.119.02-2 and powered down the machine.
Both kernels have the same kernel options/parameters:
options rd.luks.name=...=root rd.luks.name=...=home rd.luks.name=...=data ipv6.disable=1 nvidia_drm.modeset=1 nvidia_drm.fbdev=0 NVreg_PreserveVideoMemoryAllocations=1

The same problem after upgrade from 580.105.08 to nvidia 580.119.02. And I also use nvidia_drm.fbdev=0 in kernel parameters. Looks like new nvidia driver tries to initialize built-in Intel GPU instead of nvidia card and failing. The problem disappeared after I added the explicit BusID of a card to the xorg configuration:

Section "Device"
    Identifier  "Card0"
    BusID       "PCI:1:0:0"
EndSection

vvv2542 · 2025-12-22 23:20:28

seth wrote:

"nvidia_drm.modeset=1" is default, "nvidia_drm.fbdev=0" isn't a good idea because the driver cannot take over from the simpledrm device (which you're trying to use and fail because of the incompatible nvidia libraries)

nvidia_drm.fbdev=1 (default value) fixes the issue, but brings other problems for me: after loading nvidia_drm module resolution of frame buffer changed from 3840x2160 to 1920x1080. So, only left upper quarter of screen is used to display text console. Also my default terminus console font 32x16 resets to default 16x8.

Last edited by vvv2542 (2025-12-22 23:21:08)

timminybo · 2025-12-23 00:19:04

Since my last post, I created a systemd service attempting to use fbset at boot to set the monitor resolution. My service is failing though so I have to read up on systemd to figure out what I'm doing wrong. I kind of winged it last night so its no surprise its not working.

JontomXire,

Yes. I have those AUR packages installed. My X11 session has been working fine since I installed them and followed Seth's advice (reference post #11)

Seth,

I don't know how I would "have switched from the VGA to the framebuffer console or are using a bad modeline (EDID problems?)". Frankly, that's over my head even after quickly reading through the link you provided. I'll need to re-read it carefully and do some research if I hope to understand it.

As far as the console font, I do set that in /etc/vconsole.conf but the text size got smaller than it has been when the display borders appeared.

Please find the xrandr --verbose output here: https://0x0.st/PzKE.txt

seth · 2025-12-23 01:00:28

    DTD 1:  1680x1050   59.883253 Hz  16:10    64.674 kHz    119.000000 MHz (430 mm x 270 mm)
      Monitor ranges (GTF): 56-75 Hz V, 30-83 kHz H, max dotclock 140 MHz
    Display Product Name: 'DELL E207WFP'

    DTD 1:  1920x1080   60.000000 Hz  16:9     67.500 kHz    148.500000 MHz (521 mm x 293 mm)
                 Hfront   88 Hsync  44 Hback  148 Hpol P
                 Vfront    4 Vsync   5 Vback   36 Vpol P
    DTD 2:  1280x720    60.000000 Hz  16:9     45.000 kHz     74.250000 MHz (521 mm x 293 mm)
                 Hfront  110 Hsync  40 Hback  220 Hpol P
                 Vfront    5 Vsync   5 Vback   20 Vpol P
      Monitor ranges (GTF): 50-75 Hz V, 30-81 kHz H, max dotclock 170 MHz
    Display Product Name: 'S24D300'

What's the output of

fbset --info

Looks like new nvidia driver tries to initialize built-in Intel GPU instead of nvidia card and failing.

It will more look like the simpledrm device is in the way.

after loading nvidia_drm module resolution of frame buffer changed from 3840x2160 to 1920x1080. So, only left upper quarter of screen is used

tries to initialize built-in Intel GPU

I suspect you'll also have two monitors with different resolutions?
You can enforce resolutions, https://www.kernel.org/doc/Documentation/fb/modedb.rst and also control which GPU controls which terminal, see https://raw.githubusercontent.com/torva … /fbcon.rst - "fbcon=map"

This will most likely hinge on what gets loaded when though, notably whether IGP or nvidia drivers are in the initramfs (the simplydumb device can flip the order of the cards by blocking the first slot and having it surrendered to the first GPU when the second one comes along)
In doubt please post your complete system journal for the boot:

sudo journalctl -b | curl -F 'file=@-' 0x0.st

(this uploads the entire boot, but up to the multi-user.target is gonna suffice) and also your xorg log so we can see what's going on.

timminybo · 2025-12-23 01:19:54

fbset --info output is as follows:

mode "1680x1050"
geometry 1680 1050 1920 1080 32
timings 0 0 0 0 0 0 0
rgba 8/16,8/8,8/0,0/0
endmode

Frame buffer device information:
Name : nvidia-drmdrmfb
Address : (nil)
Size : 8294400
Type : PACKED PIXELS
Visual : TRUECOLOR
XPanStep : 1
YPanStep : 1
YWrapStep : 0
LineLength : 7680
Accelerator : No

timminybo · 2025-12-23 06:20:07

I managed to fix the systemd service I wrote. It gets rid of the black borders along bottom and right side of the larger (primary) monitor's display at boot (console) by explicitly forcing a specific monitor resolution. It cuts off some text on the smaller (left) monitor sometimes but I don't care about that.

The text is still the wrong size. But since it worked once, I created another systemd service to set the console font. That one works too.

In case someone finds it useful, these are my systemd service files.

# /etc/systemd/system/monitor-settings-at-boot.service
# 2025.12.22
# https://bbs.archlinux.org/viewtopic.php?id=278369&p=2
# https://bbs.archlinux.org/viewtopic.php?id=311099

[Unit]
Description=monitor settings at boot
After=multi-user.target

[Service]
Type=simple
User=root
ExecStart=/usr/bin/fbset -xres 1920 -yres 1080

# /etc/systemd/system/console-font-at-boot.service
# 2025.12.22
# https://bbs.archlinux.org/viewtopic.php?id=278369&p=2
# https://bbs.archlinux.org/viewtopic.php?id=311099

[Unit]
Description=console font settings at boot
After=multi-user.target

[Service]
Type=simple
User=
ExecStart=setfont ter-u32n

After creating the service files, I executed these commands:

sudo systemctl daemon-reload
sudo systemctl start monitor-settings-at-boot.service
sudo systemctl start console-font-at-boot.service
sudo ln -s /etc/systemd/system/monitor-settings-at-boot.service /etc/systemd/system/multi-user.target.wants/
sudo ln -s /etc/systemd/system/console-font-at-boot.service /etc/systemd/system/multi-user.target.wants/

Now when I boot up, the display is normal.

I was having trouble even executing fbset commands at the console command line at first. I was using sudo fbset ... but either got an error or the help screen. I had to switch the user to root (sudo su root) first and then run the command.

That was the problem with the monitor systemd service file too. I had used my user login initially. Once I switched the user to root in the service file, it worked.

I tried adding the setfont command to the monitor service file but it didn't work. Maybe I don't know the correct syntax. So I just created a separate service file for it and left the user blank (means any user).

So that is a work-around that does what I want. I don't know what is causing these minor console issues. The effect seems to be the smaller monitor's resolution is being used for the larger monitor. This wasn't the case before the nvidia/kernel changes.

Oh, and in case you're as clueless as me and go searching the internet and the Arch wiki to find out how to use code tags (put text in a box) in an Arch Linux Forum post and don't find anything, just click on the "BBCode: on" link in blue just below the reply box you are typing in. There are other ones there too: "url tag", "img tag", and. "smilies".
Clicking any of them takes you here: https://bbs.archlinux.org/help.php#bbcode which explains how to use all the pretty formatting.

Last edited by timminybo (2025-12-23 06:21:40)

Arch Linux

#1 2025-12-18 21:52:52

SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#2 2025-12-18 23:08:55

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#3 2025-12-21 17:01:38

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#4 2025-12-21 19:26:56

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#5 2025-12-21 21:05:28

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#6 2025-12-21 22:11:32

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#7 2025-12-21 22:46:26

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#8 2025-12-21 23:25:05

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#9 2025-12-21 23:43:20

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#10 2025-12-21 23:55:03

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#11 2025-12-22 00:12:16

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#12 2025-12-22 00:33:56

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#13 2025-12-22 01:59:24

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#14 2025-12-22 04:50:32

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#15 2025-12-22 08:31:00

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#16 2025-12-22 09:17:21

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#17 2025-12-22 14:03:46

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#18 2025-12-22 14:31:39

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#19 2025-12-22 14:31:55

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#20 2025-12-22 23:08:43

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#21 2025-12-22 23:20:28

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#22 2025-12-23 00:19:04

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#23 2025-12-23 01:00:28

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#24 2025-12-23 01:19:54

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

#25 2025-12-23 06:20:07

Re: SOLVED nvidia errors using X11 after uprading kernel & nvidia driver

Board footer