gdm fails to start

Bug #554023 reported by gmoore777 on 2010-04-02
12
This bug affects 2 people
Affects Status Importance Assigned to Milestone
xserver-xorg-video-intel (Ubuntu)
High
Unassigned

Bug Description

Binary package hint: gdm

using the LiveCD of Wednesday 3/31/2010, Thursday's or Friday's, the machine does not display anything.
The screen goes blank/black.
Similar, if machine that was working, is updated with the lastest and greatest packages. The grub menu
comes up, then monitor goes blank/black.

In the latter case, I can still ssh to the machine.
This was in the /var/log/syslog, but not sure if it has any meaning:

Apr 2 11:12:52 bee60 acpid: client 1784[0:0] has disconnected
Apr 2 11:12:52 bee60 acpid: client connected from 1958[0:0]
Apr 2 11:12:52 bee60 acpid: 1 client rule loaded
Apr 2 11:12:52 bee60 gdm-binary[875]: WARNING: GdmDisplay: display lasted 0.269835 seconds
Apr 2 11:12:52 bee60 gdm-simple-slave[2107]: WARNING: Unable to load file '/etc/gdm/custom.conf': No such file or directory
Apr 2 11:12:52 bee60 acpid: client 1958[0:0] has disconnected
Apr 2 11:12:52 bee60 acpid: client connected from 2114[0:0]
Apr 2 11:12:52 bee60 acpid: 1 client rule loaded
Apr 2 11:12:53 bee60 gdm-binary[875]: WARNING: GdmDisplay: display lasted 0.291561 seconds
Apr 2 11:12:53 bee60 gdm-binary[875]: WARNING: GdmLocalDisplayFactory: maximum number of X display failures reached: check X server log for errors
Apr 2 11:12:53 bee60 init: gdm main process (875) terminated with status 1
Apr 2 11:12:53 bee60 acpid: client 2114[0:0] has disconnected
Apr 2 11:12:53 bee60 acpid: client connected from 2137[0:0]
---
Architecture: amd64
DistroRelease: Ubuntu 10.04
DkmsStatus: Error: [Errno 2] No such file or directory
InstallationMedia: Ubuntu 10.04 "Lucid Lynx" - Alpha amd64 (20100313.1)
MachineType: Dell Inc. OptiPlex 780
Package: xserver-xorg-video-intel 2:2.9.1-3ubuntu1
PackageArchitecture: amd64
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.32-19-generic root=UUID=aac69827-1b25-489d-afc0-e8ee35fe9bc1 ro quiet splash
ProcEnviron:
 PATH=(custom, no user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.32-19.28-generic 2.6.32.10+drm33.1
Tags: lucid lucid
Uname: Linux 2.6.32-19-generic x86_64
UserGroups: adm admin audio cdrom dialout dip fax floppy galileo lpadmin plugdev tape video
dmi.bios.date: 02/13/2010
dmi.bios.vendor: Dell Inc.
dmi.bios.version: A03
dmi.board.name: 0200DY
dmi.board.vendor: Dell Inc.
dmi.board.version: A00
dmi.chassis.type: 3
dmi.chassis.vendor: Dell Inc.
dmi.modalias: dmi:bvnDellInc.:bvrA03:bd02/13/2010:svnDellInc.:pnOptiPlex780:pvr:rvnDellInc.:rn0200DY:rvrA00:cvnDellInc.:ct3:cvr:
dmi.product.name: OptiPlex 780
dmi.sys.vendor: Dell Inc.
glxinfo: Error: [Errno 2] No such file or directory
system:
 distro: Ubuntu
 codename: lucid
 architecture: x86_64
 kernel: 2.6.32-19-generic
xkbcomp:
 Error: command ['xkbcomp', ':0', '-w0', '-'] failed with exit code 1: Error: Cannot open display ":0"
                   Exiting

gmoore777 (guy-moore) wrote :

this is one thread discussing problem from more than one person.
http://ubuntuforums.org/showthread.php?p=9065231

mlindeblom (lindeblm) wrote :

I have had problems using daily build from March 29, 31, April 1 and 2.
Initial "Ubuntu" screen with 5 dots works fine then blank screen.
I have an ATI HD5750.
Previous daily builds from March 8, 15 and 22 worked.

gmoore777 (guy-moore) wrote :

LiveCD of today, 4/5/2010, still does not boot up (screen is dead) on my
Dell Optiplex 760 with " VGA compatible controller: Intel Corporation 4 Series Chipset Integrated Graphics Controller (rev 03)".

The April 5th LiveCD does work fine booting up on a Dell Optiplex GX 520 with:
"VGA compatible controller: Intel Corporation 82945G/GZ Integrated Graphics Controller (rev 02)"

gmoore777 (guy-moore) wrote :

The April 5th (32-bit) LiveCD does work fine booting up on a Dell Precision 650
with a "VGA compatible controller: nVidia Corporation NV18GL [Quadro NVS 280 SD] (rev a2)".

The April 5th (32-bit) LiveCD does work fine booting up on a Dell Inspiron Mini (netbook)
with a "VGA compatible controller:Intel Corporation System Controller Hub (SCH Poulsbo) Graphics Controller (rev 07).

Sebastien Bacher (seb128) wrote :

the issue there seems rather an xorg than a gdm one

Changed in gdm (Ubuntu):
importance: Undecided → Low
Timo Aaltonen (tjaalton) wrote :

need /var/log/Xorg.0.log from both cases (working, non-working)

gmoore777 (guy-moore) wrote :

Attached is the Xorg.0.log from the broken system.

I don't have a working system. Let me burn a LiveCD from say the Beta 1 release, load that and give you that file,....

gmoore777 (guy-moore) wrote :

Attached is Xorg.0.log from the same exact machine, but with the Beta 1 LiveCD loaded on it.

Sebastien Bacher (seb128) wrote :

the issue seems rather an xorg driver one

affects: gdm (Ubuntu) → xserver-xorg-video-intel (Ubuntu)
mlindeblom (lindeblm) wrote :

I have an ATI HD5750. My current working system failed the first reboot after yesterdays update.
The current Xorg.0.log is 0 bytes in size, thus I attached Xorg.0.log.old. Lucid 10.04 2.6.32.19 appears on grub. I was able to boot using recovery mode. Update/upgrade to not resolve issue.

As for builds from current daily build, the screen is blank/black. cntr+alt+F1 fails to respond. I do not know how to provide the /var/log/Xorg.0.log for these installation failures. The last workable daily build that I used for a fresh install was March 22. Every daily build I used from March 29 resulted in a blank/black screen a after the Ubuntu banner with 5 dots completed.

mlindeblom (lindeblm) wrote :

Since last update to working system was linux-kernel. Reinstalling fglrx resolved the problem to the previously working system.

I do not have a second system to ssh into one of my fresh installs from daily build. The installation fails before I enter any of the install information so files are in tmpfs I believe (i.e. not files or logs left after power removed). I have tried various boot line options such as vga=774 with same blank/black screen after ubuntu banner with 5 dots.

tekstr1der (tekstr1der) wrote :

I have been experiencing this periodically on a new Lenovo ThinkPad X201s with Intel graphics. After grub, the screen goes black, but the boot continues to load. Plymouth is not displayed. Following a period of disk I/O, entering the keystrokes Ctrl-Alt-Del <enter> successfully reboots the system. I believe gdm is running, but is not displayed.

If, instead of rebooting, I follow the gdm login sequence of <enter> password <enter>, there is more disk activity indicating the desktop being loaded, just not displayed.

Changed in xserver-xorg-video-intel (Ubuntu):
status: New → Incomplete
Chris Halse Rogers (raof) wrote :

We've got a couple of problems mixed up here. X bugs are generally highly hardware-specific - even from the same vendor - so it's almost always unhelpful to mix Intel, nvidia, and ati bugs as has happened here! If you've got a non-Intel system and are experiencing a bug, please file a new bug rather than add comments here. Your problem will almost certainly be different.

@gmore - Since you can SSH into your machine while it's broken, could you please boot up your broken system then SSH in and run “apport-collect 554023”? That will attach all the interesting logs we'll need for diagnosis.

mlindeblom (lindeblm) wrote :

3

This bug was re-classified as an intel bug from a generic gdm issue.

Bryce Harrington (bryce) on 2010-04-08
Changed in xserver-xorg-video-intel (Ubuntu):
status: Incomplete → New
status: New → Incomplete

apport information

tags: added: apport-collected
description: updated

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

apport information

Chris Halse Rogers (raof) wrote :

From those logs the problem is that the i915 kernel module isn't detecting any outputs connected - it doesn't think the VGA is connected, which your working Xorg.0.log shows you have connected.

If possible, it would be interesting to get the output of dmesg after booting with the “drm.debug=0x04” kernel option. That will make the intel driver much more verbose about what it's doing with respect to trying to detect what monitors are plugged in.

Secondly, since this appears to be a regression from Beta 1, it would be useful if you could try booting the intermediate kernels - 2.6.32-17 and 2.6.32-18 to see when it broke. That's only going to be possible if you've been keeping your system up to date, and haven't removed any of the previous kernels, though.

Changed in xserver-xorg-video-intel (Ubuntu):
importance: Low → High
affects: xserver-xorg-video-intel (Ubuntu) → linux (Ubuntu)
Chris Halse Rogers (raof) wrote :

@tekstr1der: Your problem is a separate bug - particularly since it's intermittent. If you could please file a separate bug, preferably with the details from a failed boot. If you can't SSH into your thinkpad, you could collect those details by logging in blindly, running a gnome-terminal, and running apport-cli. The key sequence for that would be:

Alt-F2 gnome-terminal <enter> apport-cli -f -p xorg --save blankscreen.crash <enter> (wait for disc activity to stop, then press ‘N’)

That will create a file called “blankscreen.crash” in your home directory, which you can later report to Launchpad with “apport-cli -c blankscreen.crash”.

gmoore777 (guy-moore) wrote :

On a second machine (with same problem), I just booted up 2.6.32-16 kernel, and the problem (dead screen) happened.
But having said that, that -16 kernel was installed on 3/13/2010, per /var/log/apt/history.log and I did not have the problem then.
The problem occurred about March 23rd after an update(a large one, over hundred packages).
I don't have the -17, or -18 kernel. But if I remember correctly, once the problem occurred, booting up any
older kernel didn't fix anything. And isn't the problem more in the xserver-xorg-video-intel (2.9.1-1ubuntu13, 2.9.1-1ubuntu16), package
rather than the kernel?
(xserver-xorg-video-intel did get upgraded on 3.23.2010)

(doing the drm.debug=0x04 next. As soon as I figure out how to do that.)

tekstr1der (tekstr1der) wrote :

Thanks Chris. I filed bug #559201 regarding my issue. It's not truly intermittent as I have a known STR to reproduce it reliably. Hope it gets some attention.

gmoore777 (guy-moore) wrote :

Adding an extra wrinkle to my problem(sorry). I have determined that my problem
still is a problem, but it exists only if I am using a KVM switch for the machines. If the monitor is directly
attached, the console window comes up fine, and, a LiveCD (April 5th) comes up fine.
But with the KVM switch, I have the dead screen problem.
I've tried a second KVM switch and a second monitor. Same problem.
It used to work with the KVM switch prior to March 23rd.
Would this still be the intel driver?
I've attached Xorg.0.log_noKVM (where it does work), and Xorg.0.log_withKVM where it doesn't work.

gmoore777 (guy-moore) wrote :

Xorg.0.log_withKVM is attached

gmoore777 (guy-moore) wrote :

attached is dmesg_withDebug which is /var/log/dmesg but with "linux " line in grub menu appended with "drm.debug=0x04".
This is booting the machine with the KVM switch and the dead screen.

gmoore777 (guy-moore) wrote :

just to reiterate, I am able to boot the Beta1 LiveCD , using the KVM switch.
(proving the KVM switch is good)
 I have another machine that I rebuilt out with Beta1 LiveCD and no further updates.
It boots fine on a different KVM switch, but cannot boot up using an April 5th LiveCD.
(proving again that the KVM switch is good, with a whole different set of equipment)

FYI: the KVM switches we are using, are TRENDnet, usb-type, 2-port and 4-port.
We have dozens in use with HardyHeron, and a half dozen in use with LucidLynx machines.
These are good no-frill KVM switches.
They do not have a separate power supply like Belkin switches do.

gmoore777 (guy-moore) wrote :

Problem still exists with a live CD burned today, 4/12/2010.

(this bug is critical to us, as we outfit each developer with a Windows machine, a Linux machine,
one monitor, and of course, one KVM switch.)

gmoore777 (guy-moore) wrote :

on a different machine than any of the above, but with same VGA controller, the LiveCD of 4/12/2010 boots up,
using a Belkin KVM switch. (not a TRENDnet).

One difference between the Belkin and TRENDnet KVM switches, is that Belkin has
an AC/DC adapter that provides power to the switch, the TRENDnet does not.
(does that help anyone?)

gmoore777 (guy-moore) wrote :

LiveCD of 4/16/2010 still does not work with my TRENDnet KVM switch.

Another point to make, I can boot up a machine with a monitor directly attached.
And then I can rearrange the monitor cable such that I am going through the KVM switch.
That works.

So, the problem is when "the thing" that does monitor detection, or KVM detection, is
screwing up on the KVM detection.

But the Grub Menu is displayed, so Grub has no problem throwing bits up through the
KVM switch.

gmoore777 (guy-moore) wrote :

I've updated my machine with latest Lucid updates. No change in problem: Same problem where using the TRENDnet KVM switch
causes X to not recognize the monitor. Without the KVM, everything is fine.

Looking at both Xorg.0.log files, the files are identical for the first 151 lines, then right at the line of:
       (II) intel(0): EDID for output VGA1

is where they begin to differ. When machine is hooked up to KVM switch, the next line is:
        (II) intel(0): EDID for output HDMI1

When machine is not hooked up to KVM, the next line starts hundreds of lines of Monitor resolutions and Mouse detections.

gmoore777 (guy-moore) wrote :

using the ReleaseCandidate that I burned today, 4-23-2010, booting the LiveCD exemplifies the same problem.
Black/dead screen when using KVM(TRENDnet) switch. I never see the purple screen with the 5 dots blinking.

LiveCD boots fine without KVM switch.

gmoore777 (guy-moore) on 2010-04-23
affects: linux (Ubuntu) → xserver-xorg-video-intel (Ubuntu)
Bryce Harrington (bryce) on 2010-04-24
tags: added: lucid
Chris Halse Rogers (raof) wrote :

Ok. So, what's happening here would seem to be that the KVM that you have is doing at least one of the following:
a) Not putting sufficient electrical load on the VGA port so that the intel driver isn't detecting a connected monitor there, and/or
b) Not responding to DDC requests, so the intel driver is unable to determine the acceptable display modes.

You might be able to force on the VGA connector by passing “video=VGA1:800x600” (or whatever mode you wish) to the kernel command line; you can access that from the GRUB boot menu. Just place it on the same line as “quiet splash”. I'm not sure whether this will forcibly enable the VGA output for you, but it's worth a shot.

Also, although you've said that you were going to attach “dmesg_withDebug” to the bug, but I can't seem to find it. Did you forget to attach it?

gmoore777 (guy-moore) wrote :

1A.)
I found a file dmesg_withDebug. I'll assume it was the file that I was supposed to attach a long time ago.

1B.)I'll make another one, some time today.

2.)
I just upgraded with latest LucidPackages this morning.
Rebooted. Same dead screen problem.

3.)
I just booted, successfully, off the Beta 1 LiveCD that I burned on 4/06/2010.
(that proves that it's not the KVM switch nor the monitor that are having problems, right?)
(not to mention, that earlier in my notes, I tried other KVM switches with same problem.)

4.)
I did another test, 2 weeks ago, on yet another different machine, using a Belkin switch.
These switches have their own power supply. The TRENDnet, that I'm having a problem with
do not. With the latest and greatest Lucid on that machine, with same VGA chipset, did
seem to boot successfully using that Belkin switch. But it's not one of my best tests,
since that machine belongs to someone else, including the Belkin switch, I can't
drag both over to my cube and do better divide and conquer tests. So don't jot
this note down as fact.(yet) But I wanted to mention it.

5.)
But having said all of that, even if we assume that it is a power thing,
the Beta LiveCD works fine. So it still is a software problem.
It used to work.

6.)
I will burn a HardyHeron CD, to prove that these KVM switches work fine
with older releases.

7.) I owe you one test with "video=VGA1:800x600" on linux command line via Grub menu.

gmoore777 (guy-moore) wrote :

Just added "video=VGA1:800x600" to end of linux line (without the quotes) and booted (<CTRL-X>),
No change in the outcome.

I did a diff between this Xorg.0.log and one previously without that option. It made no difference. Here is the diff:

$ diff Xorg.0.log_withKVM Xorg.0.log
7,9c7,9
< Kernel command line: BOOT_IMAGE=/boot/vmlinuz-2.6.32-21-generic root=UUID=5a4c1fd6-dd2e-4bf1-b39e-868aaba80337 ro quiet splash
< Build Date: 15 April 2010 11:05:58PM
< xorg-server 2:1.7.6-2ubuntu5 (Bryce Harrington <email address hidden>)
---
> Kernel command line: BOOT_IMAGE=/boot/vmlinuz-2.6.32-21-generic root=UUID=5a4c1fd6-dd2e-4bf1-b39e-868aaba80337 ro quiet splash video=VGA1:800x600
> Build Date: 23 April 2010 05:11:46PM
> xorg-server 2:1.7.6-2ubuntu7 (Bryce Harrington <email address hidden>)
16c16
< (==) Log file: "/var/log/Xorg.0.log", Time: Fri Apr 23 09:33:52 2010
---
> (==) Log file: "/var/log/Xorg.0.log", Time: Mon Apr 26 10:43:56 2010

gmoore777 (guy-moore) wrote :

Attached is /var/log/dmesg after having added "drm.debug=0x04" without the quotes on the linux command line via the Grub edit menu.
This file was created several minutes ago.

gmoore777 (guy-moore) wrote :

I just burned a HardyHeron 8.04.4 CD.
That LiveCD boots up fine.

gmoore777 (guy-moore) wrote :

Assuming the problem is with the package, xserver-xorg-video-intel, would it be helpful and possible
to purge this package and install various versions of this to identify the exact version where the
regression occurred?

If yes, how does one get hold of all the interim versions of xserver-xorg-video-intel
(without rebuilding them on my machine.)

Chris Halse Rogers (raof) wrote :

Odd - your dmesg+with_drm.debugAndKVM file doesn't actually have any output from the intel driver *at all*. It doesn't even appear to try to load. I'm not sure what to make of that.

The problem that you're seeing here is in the kernel - by default, the i915 kernel module is responsible for monitor detection and modesetting, which is what's going wrong here. We can go two ways: try out the modesetting code in the xserver-xorg-video-intel DDX by passing “nomodeset” to the kernel command line, or try to pin down which kernel broke modesetting for you.

You can get previous kernels from https://edge.launchpad.net/ubuntu/lucid/+source/linux

It might be worth trying both options :)

gmoore777 (guy-moore) wrote :

Thank you.

Adding "nomodeset" to the linux boot line via editing GRUB, worked 3 out of 3 boots. That is excellent.
Not using "nomodeset" failed 4 out of 4 times. (just making sure I the problem still exists)

Aside: when I initially moved the machine that I just tested from its directly attached monitor/keyboard/mouse to
my cube with the 4-way TRENDnet KVM, it actually booted up fine for the first time. Then failed on subsequent
times as expected. I may have seen that phenomena on other occasions. I felt like mentioning it may be important.

Also, when the first problem occurred, and I may have said this way above, I think, I booted up the 2 previous
kernels that were available to me on the GRUB menu and those failed as well. That's why I always thought it wasn't a kernel problem,
but I know nothing about all of this and who knows what I did so long ago.

I will grab the BETA1 kernel, if possible, since I know the BETA1 LiveCD works and see if the older kernel
amongst all the newer packages works or not.

gmoore777 (guy-moore) wrote :

If I go here, https://edge.launchpad.net/ubuntu/+source/linux/2.6.32-12.17,
and download https://edge.launchpad.net/ubuntu/+archive/primary/+files/linux_2.6.32.orig.tar.gz
this is not revision 17, this is revision 0 or 1 , right?
How do I apply patches or revs to get that souce code tar ball up to rev 17?
(sorry)
and then I would just run these 3 commands:

   make mrproper
   fakeroot make-kpkg --initrd --append-to-version=-doggy kernel-image kernel-headers

   then install whatever .deb file is lying around, via `sudo dpkg -i *.deb`
   ?

You don't need to rebuild them - click on the architecture links (for
you, amd64), which will take you here:
https://edge.launchpad.net/ubuntu/+source/linux/2.6.32-12.17/+build/1485899

From there you can download the pre-built linux-image-2.6.32-17-generic
package.

Since it works with nomodeset, could you please attach an Xorg.0.log
with nomodeset and the ModeDebug X option set. You can set ModeDebug
with the following /etc/X11/xorg.conf:
"""
Section Device
 Identifier "My card"
 Driver "intel"
 Option "ModeDebug" "true"
EndSection
"""

Once we've got that, and narrowed down the kernel revision which broke
it for you, we can (a) send this upstream and (b) hopefully identify,
and fix, the patch that broke it for you.

gmoore777 (guy-moore) wrote :

Short Story:
Kernel 2.6.32-11.15 is a kernel that causes the problem on my LucidLynx machine with all updated packages.
Kernel 2.6.32-10.14 is a kernel that boots fine on my LucidLynx machine with all updated packages.

Attached is the Xorg.0.log file from booting with the latest 2.6.32-21.32 kernel
with "nomodeset" and with ModeDebug.

SideNote: the Beta 1 LiveCD boots fine on my machine, and that release used kernel 2.6.32-16.25
(so that part doesn't make sense to me, since that kernel as of today does NOT work, but I'll let you deal with this
conflicting information. (Unless the Beta 1 LiveCD used "nomodeset" to boot up and Betat2 LiveCD did not?))

Longer story:
I did this testing today:
2.6.32-7.10 Boots fine.
2.6.32-10.14 Boots fine. (2 tests)
2.6.32-11.15 Fails but works fine with "nomodeset" (2 tests)
2.6.32-12.16 Fails but works fine with "nomodeset"
2.6.32-12.17 Fails but works fine with "nomodeset"
2.6.32-14.20 Fails but works fine with "nomodeset"
2.6.32-16.25 Fails but works fine with "nomodeset"
2.6.32-16.25 in the form of the Beta1 LiveCD boots fine. <-- does this make sense?
2.6.32-17.26 Fails but works fine with "nomodeset"
2.6.32-18.27 Fails but works fine with "nomodeset"
2.6.32-19.28 Fails but works fine with "nomodeset"
2.6.32-20.29 Fails but works fine with "nomodeset"
2.6.32-21.32 Fails but works fine with "nomodeset"

The infromation in the previous posting had a syntax error, the word Device needed quotes, as in:
   Section "Device"

and thank you again for your time on this.

Bryce Harrington (bryce) on 2010-05-21
tags: added: hardy
Launchpad Janitor (janitor) wrote :

[Expired for xserver-xorg-video-intel (Ubuntu) because there has been no activity for 60 days.]

Changed in xserver-xorg-video-intel (Ubuntu):
status: Incomplete → Expired
To post a comment you must log in.