nvidia driver sometime does not load at boot

Restart GDM (sudo /etc/init.d/gdm restart) cause X/NVidia to load properly. I maybe a timing issue.

Still a problem in beta 1, probably somewhat worst.

Are you able to reproduce the problem if you uninstall (with --purge) plymouth?

Nvidia loads fine after removing plymouth.

The problem is still happening at random, but not at frequently since I uninstalled plymouth.

Plymouth isn't relevant here, all you're doing there is changing boot timings, etc.

This is still a driver issue

Subscribing upstream to this bug report

Please subscribe Daniel Dadap to NVIDIA bugs instead of me. He's our support engineer now.

I should also mention that there's no indication in the log file that the NVIDIA driver is related. I don't think any of the usual suspects, like crashes (SEGV or BUS), unresolved symbols (ABRT), or divides by zero (FPE), can be blamed here. I would guess that some external program is sending SIGQUIT to the server.

I haven't seen this happen during any of my recent testing with Lucid. If it is a timing issue, it might be easiest to pursue a repro on the same hardware. I'll try to get access to one of these notebook systems (Dell Studio XPS 1340) for testing.

In the meantime, if anybody else is experiencing this on other systems, it would be helpful to know about those systems as well.

Still appears to be a problem with a fresh install of Ubuntu Lucid RC. Seems only to be a problem with the GeForce 9400M G, my GeForce 9800GT card on my desktop does not have this issue.

Having the same problem with Lucid RC and a GeForce 7100 / nForce 630i onboard. After switching back to nouveau it is gone.

No change trying the 195.36.24 driver in X-Swat PPA.

This sounds like it might the same as this:

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=563624

Could the affected people please try increasing the GDM timeout (GdmXserverTimeout in /usr/share/gdm/defaults.conf) to see if it alleviates the issue?

@daniel Ubuntu does not have that file, or it is somewhere else.

Hmm, I'll fire up my Ubuntu Lucid system and see if the setting for GDM timeout has moved somewhere else. It might be in another file, or in gconf, or who knows where...

Hi Mike, you're right; I can't seem to figure out where the GDM configuration went...

Try creating a simple custom.conf file at /usr/share/gdm/custom.conf containing the following two lines:

[daemon]
GdmXserverTimeout=60

If this doesn't work, it will tell us that either:

1) Your problem is not the GDM timeout issue
2) The version of GDM distributed with Lucid ignores custom.conf
3) Both of the above

So far go good with the /usr/share/gdm/custom.con edit, but this problem can be pretty random and not happen for a week or two at a time.

I'm having the same problem with an HP laptop with an Nvidia card. It only happens with the 2.6.32- kernel. I added the custom.conf file to both /usr/share/gdm and /etc/gdm but no joy. Running '/etc/init.d/gdm restart' fixes it every time.

While problem is less frequent, it still appears to be a problem. I'm not sure if usr/share/gdm/custom.con edit work or if the problem randomness just did not happen for awhile.

I'm having this problem too and none of the solutions above to increase the gdm timeouts are working for me. In my case, about 80% of the time I get the console login upon bootup. So much more serious. My brute force work around was to add a little script that gets run at boot up after a small delay to start gdm. If it's already started, the script doesn't affect anything, but otherwise, at least GDM gets loaded:

1. sudo nano /etc/crontab

add this line to the end:

@reboot root /root/gdmstart

2. sudo nano /root/gdmstart

with this contents:

#!/bin/bash
sleep 6
/usr/sbin/service gdm start

3. sudo chmod u+x /root/gdmstart

Hi just wanted to post my config as I suffer from the same problem about 50% of boots

Lucid-Lynx on
Elitegroup
Geforce6100PM-M2 Motherboard

Seballa

New beta drivers for NVidia:
http://www.nvnews.net/vbulletin/showthread.php?p=2255561

Someone had tested them?

I have the problem with a G210.

@vpablo No change with the new beta driver.

Ok, thanks. We must restart gdm every time.

I had installed 256 drivers from x-swat PPA and in two times I have booted I had no problems:

https://launchpad.net/~ubuntu-x-swat/+archive/x-updates

Please test.

And update kernel to 2.6.32-22.36 (if it will be the solution for this bug).

I had same problem I think, look at https://bugs.launchpad.net/ubuntu/+bug/573455.
But I'm not sure if it's duplicate bug but it's at least similar, I didn't test drivers from x-swapp PPA.

Four boots with new drivers and no problem. Please update and test.

No change.

After many days without problem today returns the bug.

I don't know.

Again. After yesterday's update of xorg two boots without problem. Now I have an update for GDM, I will make it and check to reboot.

I just wanted to update from Jaunty to the latest and greatest LTS but i heavily regret having done that.. fortunately i clonezill'ed the system before the clean install.
But how such a problem slipped into an LTS?
I experience the same problem with any version of the nvidia drivers:
With a clean install ubuntu10.04-64:
  - tried ubuntu-supplied nvidia proprietary drivers, the problem arise;
  - tried x-swat repos, same problem;
  - tried nvidia-supplied .run packages, same problem with both 195.xx/256.xx;
It's much worse now, after a reboot it loose configuration settings for some applications, this time gedit loose its entire preferences.. i just don't know what's happening..

I forgot to mention that "sudo service gdm restart" DOES start gdm as it should.

So far it seems i managed to have the gdm start correctly by adding a "GdmXserverTimeout" of 60 seconds in the "[daemon]" section.
Also, it seems Ubuntu changed the way (custom) gdm settings are being looked up since while i was keeping trying play with the timeout in "/usr/share/gdm/defaults.conf" and other paths, i just got it running at the first try when modifying the configuration file at "/etc/gdm/custom.conf", so it seems this is the configuration file actually being used, hope this bit of info will help you out.

I came into this bug report from my personal report #583660, which seems to be quite similar, or maybe a duplicate.

Today I tried to add "GdmXserverTimeout=60" into the "[daemon]" section of /etc/gdm/custom.conf. From that moment on, gdm won't start anymore nor at boot nor at command line (sudo start gdm). Then I needed to remove that configuration line to get gdm work again (only from the command line).

I found into my log files the same error as above You reported:
(EE) Jun 22 14:43:47 NVIDIA(0): Failed to initialize the NVIDIA kernel module. Please see the
(EE) Jun 22 14:43:47 NVIDIA(0): system's kernel log for additional error messages and
(EE) Jun 22 14:43:47 NVIDIA(0): consult the NVIDIA README for details.
(EE) NVIDIA(0): *** Aborting ***
(EE) Screen(s) found, but none have a usable configuration

But when I run GDM from command line, no error is issued.

Today, the automatic check of disks performed a chkdsk routine at the boot. After this event, GDM started automatically as one should expect. I hope this will help You.

(Should I mark my bug #583660 as duplicate of this bug #532436 ???)

Ciao
CM

@Cesare, this is quite interesting.
I've got your very same error message if i *omit* the GdmXserverTimeout=60 setting instead, now that i'm trying i can't even boot anymore if omitting it, then i'll leave it as is for now: i could bet i'll need to revert that gdm.conf line in two weeks.

Hi there. Since yesterday, after an automatic upgrade, the problem onto my PC is vanished. GDM starts ad boot without problems. (I'm closing my bug report #583660, which was linked to the present one as "duplicate".)

I don't know why the bug vanished. I can only say that yesterday APT-GET downloaded the following packages:
libparted0debian1 (2.2-5ubuntu5, 2.2-5ubuntu5.1), empathy-common (2.30.1-0ubuntu1, 2.30.1.1-0ubuntu1), language-selector-common (0.5.7, 0.5.8), empathy (2.30.1-0ubuntu1, 2.30.1.1-0ubuntu1), parted (2.2-5ubuntu5, 2.2-5ubuntu5.1), language-selector (0.5.7, 0.5.8), gdm (2.30.0-0ubuntu5, 2.30.2.is.2.30.0-0ubuntu2), nautilus-sendto-empathy (2.30.1-0ubuntu1, 2.30.1.1-0ubuntu1)

Then I suppose that new version of gdm fixed the problem.

Many thanks.
Ciao and good luck!
CM

Thank you for reporting that Cesare!
This evening i'm going to try the updates: oddly the update manager never prompted me for updates although is setup to do so ;-\
Anyway, just yesterday the problem started to arise after a week of working as expected, so i hope this update will finally bring this long-awaited fix.
I'll let you know if that worked for me too!

Ciao da Trento (Italia)!
MB

No, no. Several times I had booted right but then the bug has raised again. In fact, this morning the problem was in my PC.

So was your problem related to gdm in some ways? Or was it laying in your PC (hardware)?

Yesterday updates didn't help: now with gdm 2.30.2.is.2.30.0-0ubuntu2 (curious naming btw) i'm experiencing the same problem at every boot, no matter if i disconnect/connect usb drives, reboot or shutdown; furthermore, i'm now unable to access to a console prompt since everything hang after asking to "exit to console" when the X server gives the configuration error and CTRL+ALT+F<1-7> only show just a blinking cursor and nothing else.
The only way i can boot to a working desktop is to CTRL+ALT+RSYS+B and hope the best since i can no longer issue a "(sudo) service gdm start": also, most of the times this lead to a desktop where the gnome panels are quite screwed up, gnome sensors applet hangs, shortcuts aren't there, date/time prefs are gone, etc.. but everything is back at normal if "sudo killall gnome-panel".
Someone with other experiences?

Manuel: I'm not computer expert. However I think that You are describing two different bugs. The first one is related somehow to gdm and its (missing) startup. The other one is related to gnome-panel. I suppose the latter should refers to bug #439448.

Please, no flames on me if I'm wrong ... this was just a hint.

Ciao
CM

Thank you for the head up Cesare, now i'm sure i'm also affected by that very same problem too: the thing is curious because everytime i was manually restarting the gdm, the panels never get messy and everything would load as normal.

It looks like as the boot scripts at startup are loaded in parallel and there is some sort of syncing problem: i don't know why gnome panels and applets get screwed up at a so late booting stage, but i feel some of the components get loaded too early without syncing with others from which they could depend upon.
I don't know if and what configuration services are at stake, but i have both panels and/or applets screwed up whenever the problem occurs, while applications (gedit/vi/chrome..) always load fine, as far as i can tell.

Today's updates kernel to 2.6.32-24 and NVidia drivers to 256.35-0ubuntu2~xup from PPA ubuntu X team:
https://launchpad.net/~ubuntu-x-swat/+archive/x-updates

One reboot and no problems. Hope it works.

I saw that yesterday evening, i updated and forgot to reboot and test ;) This afternoon i'm going to report if that has worked or not.

Sorry to bother You again ... but after the GDM upgrade on June 28th, everything worked well at bootstrap. However, since yesterday, a new problem happened.

After hardware power-up, boot goes on to the "splash" screen with the small white-red dots. After 3/4 seconds the dots become all red, and the everything is freezed: no key shortcut is accepted (CTRL+ALT+F1,2,3,4 ... ALT+F1,2,3,4 ... CTRL+ALT+BCKSPC) but only CTRL+ALT+CANC is accepted and it works. At every second boot, everything runs well!!!

I mean, after hardware power-up the boot freezes at the "splash" screen; the soft reboot (CTRL+ALT+CANC) leads to a good and complete bootstrap. If I turn power off, then the problem starts again.

I extensively searched for warning or error messages about NVIDIA or GDM or X into /var/log/* ... but I can't find anything relevant, but only this:
   /var/log/gdm/:0.log.1 output --->
   X: /tmp/.X11-unix has suspicious mode (not 1777) or is not a directory, aborting.

However, as long as I'm not an expert, I could missed to detect something relevant. Please, let me know if is there some specific key or string to search for in /var/log/* or something else. See attached files. The last wrong event (bad hard-boot failed) is dated locally Jul 9 - 07:54, while the last good soft reboot is dated locally July 9 - 07:55. I hope this will help.

Ciao
CM

Hi Cesare, i also updated my system and since then, no problems so far: but i had your very same problem before the update, but used CTRL+ALT+SysReq+B since CTRL+ALT+CANC wasn't working at all. Lately, it seems things are a little better, but i know this problem presents itself in various different ways, and with no way to estimate any future occurence.

Ok, here i'm again: no problems for some days, today the problem reoccurred, so still no luck.

Following my comment #49, since a few days on everything now seems to be ok again. I had no more problem. Reading again the logfiles the error messge "X: /tmp/.X11-unix has suspicious mode (not 1777) or is not a directory, aborting" is present on file dated July 10th (last time). Therefore, from July 11th the system is ok.

Thanks.
CM

Hi.

I upgraded to 2.6.32-24-generic via update manager. The machine then hung exactly as described in comment #49. I suspect the module build failed as I couldn't find it in the 2.6.32-24 modules when I booted using an older kernel.

I then upgraded my NVidia drivers to 256.35-0ubuntu2~xup from PPA ubuntu X team as suggested in comment #47 (thanks VPablo!). I noticed it built the module for the running kernel and then the new 2.6.32-24 kernel. I rebooted and everything was fine.

Since a couple of days the problem at comment #49 appeared again. I then installed the PPA driver and everything now seems correct.

Please, note that a new splash screen (with nvidia logo) flashes before gdm starts up.

Thanks You all
Ciao
CM

Update to #53. I spoke too soon. Next time I rebooted it hung in exactly the same way again. This is a real show stopper. My system just will not boot with the latest official 2.6.32-24-generic kernel.

:(

Comment to issue #55: my system seems to be well fixed by the PPA at issue #47. Here details:

Linux 2.6.32-24-generic #38-Ubuntu SMP Mon Jul 5 09:20:59 UTC 2010 x86_64 GNU/Linux
xserver-xorg-video-ati (6.13.0-1ubuntu5, 6.13.1-1ubuntu1~xup3~lucid)
nvidia-current (195.36.24-0ubuntu1~10.04, 256.35-0ubuntu2~xup2)
nvidia-current-modaliases (195.36.24-0ubuntu1~10.04, 256.35-0ubuntu2~xup2)
nvidia-glx-185 (195.36.24-0ubuntu1~10.04, 256.35-0ubuntu2~xup2)
gdm (2.30.2.is.2.30.0-0ubuntu2, 2.30.2.is.2.30.0-0ubuntu3)
fglrx-modaliases (8.723.1-0ubuntu4, 8.741-0ubuntu0sarvatt2~lucid)
nvidia-96-modaliases (96.43.17-0ubuntu1, 96.43.18-0ubuntu1)
xserver-xorg-video-radeon (6.13.0-1ubuntu5, 6.13.1-1ubuntu1~xup3~lucid)
nvidia-173-modaliases (173.14.22-0ubuntu11, 173.14.27-0ubuntu1)
nvidia-185-modaliases (195.36.24-0ubuntu1~10.04, 256.35-0ubuntu2~xup2)
nvidia-settings (195.36.08-0ubuntu2, 256.35-0ubuntu1)

All these packages together have fixed the problem onto my system.

Ciao
CM

I am still having this problem with the latest nvidia drivers.

At nearly every boot nvidia fails to load and gdm tries to run a failsafe xsession.

Unfortunately, there's nothing of importance in either the xorg or kernel logs. The kernel log has no mention of any failure at all and xorg's log just says nvidia failed to load and I should contact support and read my kernel log....

If I exit to terminal, i can start gdm with either "sudo gdm" or "sudo service gdm start" and everything runs normally.

Could this be a mistake in user/group permissions?

Perhaps gdm is being started by a system account that has somehow lost permission to load nvidia.

@quequotion - I would say this is more likely a race condition between the loading of the nvidia kernel module and the starting of X / gdm - basically I think what is happening is that the module is loaded but X / gdm starts up before it has finished initialising the hardware and hence the X nvidia driver fails to load correctly. At least this is what I think is probably happening...

I'm not computer expert, and perhaps my following comment could be a stupid matter ...

Recently I was affected by both bugs: bug #439448 and bug bug #575410 - both are related to gnome-panel.

Both bugs were temporary fixed (workaround) onto my system just adding a 3 seconds delay before gnome-panel starts after gdm login.

If there is a race conditions for the nvidia driver, it could be worth trying to add a 3-5 seconds delay to X / gdm start up.

If someone could tell me how to try this X startup delay, then I will be happy to help You: I will downgrade to previous drivers (removing the PPA versions) and I will try the delayed X startup.

Ciao
CM

Sorry to bother You with another new question ... After the PPA installation (https://launchpad.net/~ubuntu-x-swat/+archive/x-updates) everything is running well.

However, today, the command "sudo apt-get update ; sudo apt-get upgrade" showed this output:
I seguenti pacchetti sono stati mantenuti alla versione attuale:
  xserver-xorg-video-intel
I seguenti pacchetti saranno aggiornati:
  fglrx-modaliases libmysqlclient16 libsmbios2 mysql-common
  nvidia-185-modaliases nvidia-current nvidia-current-modaliases
  nvidia-glx-185 nvidia-settings ubuntu-system-service
10 aggiornati, 0 installati, 0 da rimuovere e 1 non aggiornati.
È necessario scaricare 45,2MB di archivi.

It seems that new versions of nvidia drivers are available. I didn't reply "yes" to this proposed upgrade. Should I reply "yes" ? Are these new drivers patched (bug fixed) ?

Thank You in advance.
Ciao
CM

Today, after yesterday's updates, gdm start works well, but we know that this happens always, in a week the bug will appear again.

I'm experiencing this on 10.10 Alpha 3 using the drivers from the repo's.

Take a look at this thread in the forums:

http://ubuntuforums.org/showthread.php?t=1501156

one temporary workaround i found was to comment out a part of /etc/init/gdm.conf

after a week of working very well, the problem came back and now ubuntu goes directly to text login.

I think this is part of a larger issue in init or gdm and not an issue with nvidia.

This looks like a duplicate of bug #459639 marking as such to clear things up

Alex,

I'm not sure it's a duplicate. I'm not sure if it's the same behaviour for all other affected, but in my case this particular bug even prevents the failsafe dialog from loading, and I'm left with either a black screen or a screen with the Ubuntu logo.

Using nVidia driver 260.19.04 with Lucid. Same issues as above. My current fix is to restart gdm every time from terminal. Still experimenting/ trying to find permanent fix.

I haven't seen this issue since upgrading to 2.6.32-25 - X now starts perfectly every time. My symptoms are described under bug number 616197.