FS#48772 - [gnome-shell] After a few minutes the screen goes black
Attached to Project:
Arch Linux
Opened by Dario Ostuni (dariost) - Thursday, 31 March 2016, 14:53 GMT
Last edited by Jan de Groot (JGC) - Tuesday, 12 April 2016, 19:21 GMT
Opened by Dario Ostuni (dariost) - Thursday, 31 March 2016, 14:53 GMT
Last edited by Jan de Groot (JGC) - Tuesday, 12 April 2016, 19:21 GMT
|
Details
Description:
After a few minutes I've logged in gnome-shell the screen goes black and I can't do anything. Additional info: gnome-shell is version 3.20.0-1 The only 2 installed extensions are: gnome-shell-extension-freon and gnome-shell-extension-gravatar In dmesg it says "[ 330.712504] traps: gnome-shell[702] trap int3 ip:7fa24bd967eb sp:7ffd156555c0 error:0" Steps to reproduce: Log in gnome-shell and wait some minutes |
This task depends upon
Just a guess, of course :)
Also I should mention that "echo $DISPLAY" gives ":1"
Xorg.1.log (41.4 KiB)
In the attachments there are /boot/grub/grub.cfg and /etc/default/grub
grub (1.5 KiB)
I get the same errors in journalctl than ademidov.
"Description:
After a few minutes I've logged in gnome-shell the screen goes black and I can't do anything."
I can switch to TTY2 when the screen went black. I only have a garbaged wallpaper, after alt+f2 and 'r' (restarting the shell), everything is well again.
Journalctl: http://pastebin.com/NTnmwnGZ
Try disabling wayland for gdm: https://wiki.archlinux.org/index.php/GDM#Use_Xorg_backend
╰─$ cat /etc/gdm/custom.conf
# GDM configuration storage
[daemon]
# Uncoment the line below to force the login screen to use Xorg
WaylandEnable=false
And, yes I have nvidia drivers.
Indeed. With nouveau, the problem does not occur, so it seems to be related. Replaced nvidia with xf86-video-nouveau and the system runs stable for now. I'm using an GTX 760.
- was this a problem with GNOME 3.18 also (same cogl version)?
- is this still a problem when using nvidia drivers from extra?
Looks like cogl_framebuffer_allocate fails with nvidia drivers at some point, causing a fatal error leading to SIGABRT.
Also, can anyone recompile cogl with debug symbols and post complete backtraces?
I am not sure, but I think - no. There was no this problem with Gnome 3.18. I had fresh install of system with Gnome 3.18 and worked with it only 4-5 hours, but there was not this problem.
>> is this still a problem when using nvidia drivers from extra?
I have installed nvidia drivers from extra right now, and yes - there problem is still here
1: No.
2: Yes, I've tested it some minutes ago. Same result. (nvidia-361.28-4)
3: Maybe I can make a backtrace later, I will be busy at work for now.
@ademidov: Are you sure, you're using nvidia from extra and not from testing?
╭─demas@localhost ~
╰─$ pacman -Ss nvidia
extra/libvdpau 1.1.1-2 [установлен]
Nvidia VDPAU library
extra/nvidia 361.28-4 [установлен]
NVIDIA drivers for linux
extra/nvidia-libgl 361.28-5 [установлен]
NVIDIA drivers libraries symlinks
extra/nvidia-utils 361.28-5 [установлен]
NVIDIA drivers utilities
Now I don't understand which file do I need to run with `# gdb /path/to/file`
backtrace
extra/libvdpau 1.1.1-2
extra/nvidia 361.28-4
extra/nvidia-libgl 361.28-5
extra/nvidia-utils 361.28-5
multilib/lib32-nvidia-libgl 361.28-4
multilib/lib32-nvidia-utils 361.28-4
extra/gnome-desktop 1:3.20.0-1
extra/gdm 3.20.0-1
journalctl: http://pastebin.com/raw/TXNqF2A8
Packages:
(crash happens on both linux and linux-zen)
core/linux 4.4.5-1
extra/linux-zen 4.4.5-1
extra/libvdpau 1.1.1-2
extra/nvidia-dkms 361.28-4
extra/nvidia-libgl 361.28-5
extra/nvidia-utils 361.28-5
multilib/lib32-nvidia-libgl 361.28-4
multilib/lib32-nvidia-utils 361.28-4
extra/gnome-desktop 1:3.20.0-1
extra/gdm 3.20.0-1
extra/libgdm 3.20.0-1
Logs:
gdm.service status: http://pastebin.com/ACcA5Kiw
( ⌁ ) dmesg | grep gnome
[ 396.473419] traps: gnome-shell[1101] trap int3 ip:7f2faa2117eb sp:7ffc6340ea00 error:0
Xorg.1.log (36.2 KiB)
http://comments.gmane.org/gmane.comp.gnome.gdm.general/2710
My logs:
http://m-reimer.de/tmp/gdm-crash-bug/journal
http://m-reimer.de/tmp/gdm-crash-bug/coredump
Can someone tell me if there are any disadvantages if I use an alternative login manager? Is it still possible to lock the screen or switch the user?
Patch that should fix this issue (at least the backtraces) is in that directory too.
Cogl allocates framebuffer objects lazy on the first drawing operation. The check on offscreen != COGL_INVALID_HANDLE in gnome-shell will never fail, as NULL/COGL_INVALID_HANDLE is never returned from the function that creates the offscreen object.
Edit: More info,
linux: 4.4.5-1-ARCH
vga: nvidia 361.28-4
gnome: 3.20.1
3.20.0-2 -> Bug still present.
3.20.0-3 -> Bug fixed. ;)
Thanks and sorry for my reopen.