Gnome login screen fails to start

[minifrindge]

Today I upgraded my system to the Nvidia driver 340.93 and 4.2.0-1 kernel on Debian Stretch.

While I wait for the gnome login screen to pop up I get the "Oh no something has gone wrong" message.

When I log in from another terminal and execute startx , gnome initiates and starts although settings like my resolution and other xserver settings are not loaded.

From the journalctl I believe the error is located in the following section:

Code: Select all

Oct 06 14:23:17 DEBIAN-PC org.a11y.atspi.Registry[1692]: SpiRegistry daemon is running with well-known name - org.a11y.atspi.Registry
Oct 06 14:23:17 DEBIAN-PC gnome-session[1680]: X Error of failed request:  BadValue (integer parameter out of range for operation)
Oct 06 14:23:17 DEBIAN-PC gnome-session[1680]: Major opcode of failed request:  154 (GLX)
Oct 06 14:23:17 DEBIAN-PC gnome-session[1680]: Minor opcode of failed request:  3 (X_GLXCreateContext)
Oct 06 14:23:17 DEBIAN-PC gnome-session[1680]: Value in failed request:  0x0
Oct 06 14:23:17 DEBIAN-PC gnome-session[1680]: Serial number of failed request:  25
Oct 06 14:23:17 DEBIAN-PC gnome-session[1680]: Current serial number in output stream:  26
Oct 06 14:23:17 DEBIAN-PC gnome-session[1680]: gnome-session-check-accelerated: Helper exited with code 256
Oct 06 14:23:17 DEBIAN-PC NetworkManager[898]: <info>  startup complete
Oct 06 14:23:22 DEBIAN-PC gnome-session[1680]: X Error of failed request:  BadValue (integer parameter out of range for operation)
Oct 06 14:23:22 DEBIAN-PC gnome-session[1680]: Major opcode of failed request:  154 (GLX)
Oct 06 14:23:22 DEBIAN-PC gnome-session[1680]: Minor opcode of failed request:  3 (X_GLXCreateContext)
Oct 06 14:23:22 DEBIAN-PC gnome-session[1680]: Value in failed request:  0x0
Oct 06 14:23:22 DEBIAN-PC gnome-session[1680]: Serial number of failed request:  25
Oct 06 14:23:22 DEBIAN-PC gnome-session[1680]: Current serial number in output stream:  26
Oct 06 14:23:22 DEBIAN-PC gnome-session[1680]: gnome-session-check-accelerated: Helper exited with code 256
Oct 06 14:23:22 DEBIAN-PC gnome-session[1680]: gnome-session[1680]: WARNING: software acceleration check failed: Child process exited with code 1
Oct 06 14:23:22 DEBIAN-PC gnome-session[1680]: WARNING: software acceleration check failed: Child process exited with code 1

I reinstalled the xorg-xserver and the nvidia drivers and I get the same behavior.

Can some provide some info in order to reconfigure the gnome-session to work?

[Enekk]

I know this is less than helpful, but I can confirm that this is an issue for me as well. I tried downgrading the Nvidia drivers to the older version that was available from Stable, but that didn't seem to help at all. I also tried to boot into an older Kernel coupled with the older drivers, still no success. If anyone has any idea of how to get this to work, I'd appreciate the advice.

[katox]

Hi!

I found out that the problem lies somewhere around 3D detection feature of gdm3. Light DM seems to be unaffected. For now switch the DM.

systemctl stop gdm3
apt-get install lightdm
dpkg-reconfigure lightdm

Select lightdm as default.

[Enekk]

Gah! I should have thought of that given that gdm was what was tossing errors in my logs! Thanks so much, things are now working again. I guess we should report the bug.

[katox]

I found a proper fix for GDM3 in meantime. Adding it to video group should fix the problem.

sudo adduser Debian-gdm video

[minifrindge]

For now switch the DM.

lightdm works great! Although not solving the problem lightdm seems to integrate very well with Gnome.

I found a proper fix for GDM3 in meantime. Adding it to video group should fix the problem.

sudo adduser Debian-gdm video

That fixed the issue! How did you manage to track this error? I spend a day trying to fix this and I didn't even know there was a group called video!

Also some systemd updates rolled out today that fixed this error when I tried to set again gdm as default.

$ sudo dpkg-reconfigure gdm3
Job for gdm.service invalid
invoke-rc.d: initscript gdm3, action reload failed

I tried your advice first and then upgraded so I don't know if these updates could fix something.

Code: Select all

Start-Date: 2015-10-08
Commandline: apt-get upgrade
Upgrade: gvfs-bin:amd64 (1.24.2-1, 1.26.0-2), gnome-logs:amd64 (3.16.2-1, 3.18.0-1), gir1.2-json-1.0:amd64 (1.0.4-1, 1.0.4-2), eog:amd64 (3.16.3-1, 3.18.0-1), libjson-glib-1.0-common:amd64 (1.0.4-1, 1.0.4-2), gvfs-common:amd64 (1.24.2-1, 1.26.0-2), libpam-systemd:amd64 (226-3, 226-4), fonts-lmodern:amd64 (2.004.4-5, 2.004.5-1), gvfs-daemons:amd64 (1.24.2-1, 1.26.0-2), lmodern:amd64 (2.004.4-5, 2.004.5-1), udev:amd64 (226-3, 226-4), libcap2:amd64 (2.24-11, 2.24-12), libcap2:i386 (2.24-11, 2.24-12), linux-kbuild-4.2:amd64 (4.2-1, 4.2-2), gnome-chess:amd64 (3.16.2-1, 3.18.0-1), gvfs-fuse:amd64 (1.24.2-1, 1.26.0-2), gvfs:amd64 (1.24.2-1, 1.26.0-2), libudev1:amd64 (226-3, 226-4), libudev1:i386 (226-3, 226-4), gnome-clocks:amd64 (3.16.2-1, 3.18.0-1), nodejs:amd64 (4.1.1~dfsg-2, 4.1.1~dfsg-3), nodejs-dev:amd64 (4.1.1~dfsg-2, 4.1.1~dfsg-3), gvfs-backends:amd64 (1.24.2-1, 1.26.0-2), systemd-sysv:amd64 (226-3, 226-4), gnome-music:amd64 (3.16.2-1, 3.18.0-1), libjson-glib-1.0-0:amd64 (1.0.4-1, 1.0.4-2), git-man:amd64 (2.5.3-1, 2.6.1-1), git:amd64 (2.5.3-1, 2.6.1-1), systemd:amd64 (226-3, 226-4), gvfs-libs:amd64 (1.24.2-1, 1.26.0-2), libnss-myhostname:amd64 (226-3, 226-4), libsystemd0:amd64 (226-3, 226-4), libsystemd0:i386 (226-3, 226-4), diffutils:amd64 (3.3-1+b1, 3.3-2), libgdict-common:amd64 (3.16.2-1, 3.18.0-2), libcap2-bin:amd64 (2.24-11, 2.24-12), gnome-maps:amd64 (3.16.2-1, 3.18.0.1-1)

[katox]

At first I told myself that was really an idiot updating this part of the system when there was an important work scheduled for today. Got a bit careless in few last years when everything was more or less smooth (including dist-upgrade).

The steps were like this:
- banging head against keyboard
- checking nvidia kernel module, info a glxinfo
- trying standalone X, what worked so (phew!)
- everything looked suspiciously normal, nevermind, trying to reinstall from apt cache this time with no X
- that didn't change a thing (so probably no package configuration script failure)
- running with no DM (startx), that worked as well
- trying out light-dm as a workaround (which is great, the only problem I've ever had with it was gnome keyring integration)

At that point I knew that I could go to sleep

- the next day armed with a real browser I poked google a bit
- at fedora forums I spot something about messing with gdm privileges
- trying out running everything as root seemed to work
- a bit of googling in that direction quickly took me to gdm and groups
- the first group I tried worked for me
- googling the solution took me to similar threads in other distros so I knew that should really be it

No idea why gdm3 reconfigure doesn't work. I does on my machine. You can switch dm reconfiguring lightdm too but your gdm service definition seems to be messed up at this point.

[patryk007]

Wow! @katox. Respect!
I had the same problem and it worked. Quite amazing that you tracked it down.

BTW: Has anyone reported this issue to Debian's or Gnome's BugTrack?

[mfs6174]

Thanks to @katox ! I had the same issue after the upgrade and the " user group adding " fixed it. I think that someone should submit this BUG to the Debian community ( I do not quite familiar with the Debian bug tracking system).

[GarryRicketson]

Today I upgraded my system to the Nvidia driver 340.93 and 4.2.0-1 kernel on Debian Stretch.

Stretch is still in the "testing" stages, and so are many of the packages ,

BTW: Has anyone reported this issue to Debian's or Gnome's BugTrack?

There are a lot of bugs listed , in relation to "stretch" and the testing versions, if you use a "testing" version these things are expected, if you don't know enough about keeping track of the bug reports , which are updated daily, maybe it is not a good idea to be using a unstable, testing version of Debian.
https://www.debian.org/releases/testing/
and
https://www.debian.org/devel/testing
https://bugs.debian.org/cgi-bin/pkgrepo ... t=unstable
https://bugs.debian.org

[patryk007]

There are a lot of bugs listed , in relation to "stretch" and the testing versions, if you use a "testing" version these things are expected, if you don't know enough about keeping track of the bug reports , which are updated daily, maybe it is not a good idea to be using a unstable, testing version of Debian.

I know that if I'm using Debian Sid, bugs are willing to appear. That's OK for me.

https://bugs.debian.org/cgi-bin/pkgrepo ... t=unstable

Now I can see that this issue was reported at least multiple times (see: bug report with explanation).

[shugyo]

I wrote this up at the end of September:
http://forums.debian.net/viewtopic.php?f=6&t=124950

[wgaldino]

Thanks katox! +1
Man you are my hero! hehehe... Amazing discovery!