You are not logged in.
I -Syu'd everything today after not using my arch system for a while (should have gone through possibly system-breaking pkgs one at a time but oh well).
Today when trying to log in, SDDM comes up, and I'm still able to load into X11 but all my apps are configured to be wayland and I want to fix this issue.
In journalctl I'm getting all sorts of errors, it basically cycles through combinations of these (the first two are in red e.g. errors and the others are yellow)
kwin_wayland[897]: kwin_wayland_drm: Atomic modest commit failed! Invalid argument
kwin_wayland[897]: kwin_wayland_drm: Atomic commit failed! Invalid argument
kwin_wayland[897]: kwin_wayland_drm: Presentation failed! Invalid argument
kwin_wayland[897]: kwin_wayland_drm: Creating gbm surface failed! Function not implemented
kwin_wayland[897]: kwin_wayland_drm: Checking test buffer failed for (1)
Offline
What was updated, what's your graphics hardware, which driver are you using, full journal output? https://wiki.archlinux.org/title/List_o … n_services
Offline
I just did an upgrade of nvidia drivers and Wayland session is black screen after logging.
"The flesh knows it suffers even when the mind has forgotten."
Offline
Oops sorry, forgot to add system info.
Ryzen 5700G
NVIDIA RTX 3060 Ti
ASUS ROG STRIX B550F motherboard.
I'm on KDE Plasma using SDDM to login.
Here's the full journal http://0x0.st/HocI.txt
And basically all my packages I think. Important ones would be nvidia-utils, mesa, kwayland, mkinitcpio, linux, nvidia, nvidia-settings, plasma-wayland-session, kwayland-integration.
(I don't use mesa but I think I had a reason to still have it installed. Might try removing it)
Last edited by ffernn (2023-03-24 10:19:37)
Offline
Try installing linux-headers and installing nvidia-dkms a version lower ( https://archive.archlinux.org/packages/ … kg.tar.zst ) downgrading {lib32-}nvidia-utils to 525.89.02-2 to check whether it's a regression in the nvidia driver.
Offline
i experienced the same, had to switch back to eglstreams from GBM.
Last edited by NGStaph (2023-03-24 14:49:10)
Offline
How did you do that? Did you actively switch back to EGLStream or simply remove environment variables, which do not inherently have a relevance to whether your system/KWin is using EGLStreams or GBM?
Last edited by V1del (2023-03-24 16:23:17)
Offline
Try installing linux-headers and installing nvidia-dkms a version lower ( https://archive.archlinux.org/packages/ … kg.tar.zst ) downgrading {lib32-}nvidia-utils to 525.89.02-2 to check whether it's a regression in the nvidia driver.
System only boots to 'Reached target Graphical Interface'
http://0x0.st/HoBR.txt
Of interest would probably be "Failed to find module Nvidia-uvm'
I wasn't using dkms before so should I try to install 'nvidia' 525 instead?
[update, installing regular nvidia does the same thing ]
Last edited by ffernn (2023-03-25 00:48:07)
Offline
commented out the following environment variables which, i naively thought, are required to 'force' GBM as a backend.
GBM_BACKEND=nvidia-drm
__GLX_VENDOR_LIBRARY_NAME=nvidia
removing these lets me boot normally, however, like you mentioned... I'm apparently still using GBM.
Mar 23 23:49:39 NGSGL639SD gnome-shell[1471]: Created gbm renderer for '/dev/dri/card0'
Mar 23 23:49:39 NGSGL639SD gnome-shell[1471]: Created gbm renderer for '/dev/dri/card1'
Mar 23 23:50:00 NGSGL639SD gnome-shell[1992]: Created gbm renderer for '/dev/dri/card0'
Mar 23 23:50:00 NGSGL639SD gnome-shell[1992]: Created gbm renderer for '/dev/dri/card1'
Last edited by NGStaph (2023-03-26 10:12:14)
Offline
I'd assume it's likely the internal GBM_BACKEND= name might have changed? But it should - as mentioned - generally not be necessary either way. Can any of the affected check whether you are "unnecessarily" exporting those variables and what happens if you do not explicitly do that?
@ffernn If you intend to stick to the current kernel you need to install the linux-headers for the current kernel and the "relevantly" downgraded set of nvidia-dkms and nvidia-utils. If you are doing this correctly you should see DKMS building the nvidia kernel module for your current kernel in the post-transaction hook of pacman.
Last edited by V1del (2023-03-25 13:19:48)
Offline
I can confirm when I remove the two environment variables specified by NGStaph and upgrade from 525 to 530 it works great. If I upgrade without that, GDM and Gnome wayland session fall back to Xorg and fail.
Offline
I tried to set variables and still no way to log in to Wayland Session.
Last edited by gabriel9 (2023-03-27 18:55:07)
"The flesh knows it suffers even when the mind has forgotten."
Offline
The trick is to not set the variables.
Offline
Hi V1del,
I did not have them in the first place. Then I tried to add them, then I again removed them and still does not work
"The flesh knows it suffers even when the mind has forgotten."
Offline
In an earlier post you said it returned and sporadically disappeared again? Did you set up early module loading to make sure it's not a modesetting enablement race condition? https://wiki.archlinux.org/title/NVIDIA#Early_loading
Offline
Hi V1del,
Not sure did I explain this right.
I did not have those variables in my env.
After updating of system Wayland session was just black screen.
Tried to set:
GBM_BACKEND=nvidia-drm
__GLX_VENDOR_LIBRARY_NAME=nvidia
and do the upgrade of nvidia, still black screen.
Removed above variables from env and did reinstall of Nvidia again black screen.
This is my hook:
[Trigger]
Operation=Install
Operation=Upgrade
Operation=Remove
Type=Package
Target=nvidia
Target=linux
# Change the linux part above and in the Exec line if a different kernel is used
[Action]
Description=Update NVIDIA module in initcpio
Depends=mkinitcpio
When=PostTransaction
NeedsTargets
Exec=/bin/sh -c 'while read -r trg; do case $trg in linux) exit 0; esac; done; /usr/bin/mkinitcpio -P'
Last edited by gabriel9 (2023-03-29 13:51:05)
"The flesh knows it suffers even when the mind has forgotten."
Offline
The hook will only have a logical effect if you also added the nvidia modules to your mkinitcpio.conf as explained in the relevant subheading of the link.
Without the variables set and on an attempt to start the session what's your output for
sudo journalctl -b
Offline
Did not know that. Just added nvidia to modules array. Thank you for the hint.
As for failed session logs: https://gist.github.com/bojan2501/74ad1 … 9460eb6290
"The flesh knows it suffers even when the mind has forgotten."
Offline
Seems like a long live not smashable Mutter bug.
try add this to /etc/environment
MUTTER_DEBUG_ENABLE_ATOMIC_KMS=0
Offline
Added
MUTTER_DEBUG_ENABLE_ATOMIC_KMS=0
and it did not help.
"The flesh knows it suffers even when the mind has forgotten."
Offline
Ok so bringing this thread back because I've reinstalled arch from fresh and I'm using hyprland now, I've managed to narrow this issue down to being only when my refresh rate is set over 120 (119.878998)Hz. I remembered my monitor only supports 10bpc up to 120hz, but the default in hyprland is just to use 8bpc. Tried forcing it anyway and still no luck :<
Offline
I just forced refresh rate under 120 and managed to start Wayland session.
Current refresh rage is 119.88Hz.
Thank you ffernn.
Last edited by gabriel9 (2023-04-02 12:55:20)
"The flesh knows it suffers even when the mind has forgotten."
Offline
Hi same problem here cant get wayland working in 120Hz if I downgrade to 100Hz it's ok (plasma wayland)
AMD Ryzen 9 3900X 12-Core Processor
NVIDIA Corporation GA102 [GeForce RTX 3090] (rev a1)
Last edited by goofy03 (2023-04-03 13:01:39)
Offline
A refresh rate of 120 works for me, but maybe it's slightly under in reality. 144 does not work.
Also inside KDE plasma, if I try to change the refresh rate to 144, it does not crash, but it's not actually changed either.
Offline