x server with fglrx driver doesn't start anymore

[Philippe Lefevre]

I didn't know what was the fglrx package, so when I bought my new video card, I first installed Catalyst 10.8 with Lucid Linx and kernel 2.6.32.23 ... it worked but was a bit buggy (bad 2D display)
When the catalyst 10.9 was out, I installed it. It worked too and it fixed the bugs.
Then I did a kernel upgrade at 2.6.32-24. Catalyst 10.9 was still working.
Later I noticed that the Catalyst control panel was still displaying "Catalyst 10.8". I then tried to reinstall it to see if it fixed this.
I was unable to do it due to the bug #642518.
Then I applied the patch given by Luc Bruninx (post #14) and rebuild the package. There was no build errors and the patched package was successfully installed but the xserver didn't start anymore. (my own posts #19 to 22)

In the boot process, the framebuffer display is ok (with the ubuntu banner and the moving 5 dots), but when it shift to the xserver it hangs.
Neither any CtrAltFx key or CtrlAltBackspace key are working. The only way is to electricaly stop the box.
I tried many things and in particular, to remove/purge all old fglrx prevously installed following the doc at https://wiki.ubuntu.com/X/Troubleshooting/FglrxInteferesWithRadeonDriver.
I finally uninstall all and reinstalled the official new package with the new patch 2:8.723.1-0Ubuntu5. I still have the same trouble.

So something has probably been removed, or modified when I reinstalled the package (via synaptic).

I'm running :
  Description: Ubuntu 10.04.1 LTS
  Release: 10.04
  Kernel: Linux phlsys 2.6.32-24-generic #43-Ubuntu SMP Thu Sep 16 14:58:24 UTC 2010 x86_64 GNU/Linux
  Video card: ati HD5670
  Packages:

fglrx:
    Installé : 2:8.723.1-0ubuntu5
    Candidat : 2:8.723.1-0ubuntu5
   Table de version :
   *** 2:8.723.1-0ubuntu5 0
        500 ftp://ftp.free.fr/mirrors/ftp.ubuntu.com/ubuntu/ lucid-updates/restricted Packages
        500 http://archive.ubuntu.com/ubuntu/ lucid-updates/restricted Packages
        100 /var/lib/dpkg/status
     2:8.723.1-0ubuntu3 0
        500 ftp://ftp.free.fr/mirrors/ftp.ubuntu.com/ubuntu/ lucid/restricted Packages
        500 http://archive.ubuntu.com/ubuntu/ lucid/restricted Packages

fglrx-amdcccle:
  Installé : 2:8.723.1-0ubuntu5
  Candidat : 2:8.723.1-0ubuntu5
 Table de version :
 *** 2:8.723.1-0ubuntu5 0
        500 ftp://ftp.free.fr/mirrors/ftp.ubuntu.com/ubuntu/ lucid-updates/restricted Packages
        500 http://archive.ubuntu.com/ubuntu/ lucid-updates/restricted Packages
        100 /var/lib/dpkg/status
     2:8.723.1-0ubuntu3 0
        500 ftp://ftp.free.fr/mirrors/ftp.ubuntu.com/ubuntu/ lucid/restricted Packages
        500 http://archive.ubuntu.com/ubuntu/ lucid/restricted Packages

[Philippe Lefevre (ph-l)]

Sorry, I forgot to add the logs

[Philippe Lefevre (ph-l)]

Here some infos that might help too

dmesg | grep fglrx
[ 11.792142] fglrx: module license 'Proprietary. (C) 2002 - ATI Technologies, Starnberg, GERMANY' taints kernel.
[ 11.852413] [fglrx] Maximum main memory to use for locked dma buffers: 2872 MBytes.
[ 11.852488] [fglrx] vendor: 1002 device: 68d8 count: 1
[ 11.852853] [fglrx] ioport: bar 4, base 0xbc00, size: 0x100
[ 11.853688] [fglrx] Kernel PAT support is enabled
[ 11.853729] [fglrx] module loaded - fglrx 8.72.11 [Apr 8 2010] with 1 minors

ldconfig -p | grep GL
 libQtOpenGL.so.4 (libc6,x86-64) => /usr/lib/libQtOpenGL.so.4
 libQtOpenGL.so.4 (libc6) => /usr/lib32/libQtOpenGL.so.4
 libQtOpenGL.so (libc6) => /usr/lib32/libQtOpenGL.so
 libGLU.so.1 (libc6,x86-64) => /usr/lib/libGLU.so.1
 libGLU.so.1 (libc6) => /usr/lib32/libGLU.so.1
 libGLU.so (libc6) => /usr/lib32/libGLU.so
 libGLEW.so.1.5 (libc6,x86-64) => /usr/lib/libGLEW.so.1.5
 libGL.so.1 (libc6,x86-64) => /usr/lib/fglrx/libGL.so.1
 libGL.so.1 (libc6,x86-64) => /usr/lib/libGL.so.1
 libGL.so.1 (libc6) => /usr/lib32/fglrx/libGL.so.1
 libGL.so.1 (libc6) => /usr/lib32/libGL.so.1
 libGL.so (libc6,x86-64) => /usr/lib/fglrx/libGL.so
 libGL.so (libc6) => /usr/lib32/libGL.so

[Philippe Lefevre (ph-l)]

I did this test today.
I boot my machine with "normal" mode.
From an other machine I've been able to do a ssh connection to it.
With a ps -ef I can see that plymouth is still running and x is started
  root 437 1 0 17:18 ? 00:00:01 /sbin/plymouthd --mode=boot --attach-to-session --pid-file=/dev/.initramfs/plymouth.pid
  root 1991 1 0 17:20 ? 00:00:00 gdm-binary
  root 2461 1991 0 17:20 ? 00:00:00 /usr/lib/gdm/gdm-simple-slave --display-id /org/gnome/DisplayManager/Display1
  root 2470 2461 99 17:20 tty7 00:21:08 /usr/bin/X :0 -nr -verbose -auth /var/run/gdm/auth-for-gdm-FGXrKO/database -nolisten tcp vt7

I can see that the fglrx module is in place
  cat syslog | grep fglrx
  Sep 25 17:20:49 localhost kernel: [ 12.002294] fglrx: module license 'Proprietary. (C) 2002 - ATI Technologies, Starnberg, GERMANY' taints kernel.
  Sep 25 17:20:49 localhost kernel: [ 12.062691] [fglrx] Maximum main memory to use for locked dma buffers: 2872 MBytes.
  Sep 25 17:20:49 localhost kernel: [ 12.062763] [fglrx] vendor: 1002 device: 68d8 count: 1
  Sep 25 17:20:49 localhost kernel: [ 12.063116] [fglrx] ioport: bar 4, base 0xbc00, size: 0x100
  Sep 25 17:20:49 localhost kernel: [ 12.063848] [fglrx] Kernel PAT support is enabled
  Sep 25 17:20:49 localhost kernel: [ 12.063877] [fglrx] module loaded - fglrx 8.72.11 [Apr 8 2010] with 1 minors

  lsmod|grep fglrx
  fglrx 2353486 3

If I try to stop gdm with a service stop, gdm keep alive (even a kill -9 can't stop it)

Hope these details could help

Please, see the current xorg log file attached

[Philippe Lefevre (ph-l)]

I uninstalled and reinstalled all X11 xorg packages. It din't fix anything.
Today I deinstalled the xinit package and fglrx then reinstalled xinit and the ATI package.
To try to debug, I launched xinit via strace and I was able to see that my user files .Xauthority .Xauthority-c .Xauthority-l were owned by root and not overwritable by my user processes. So I deleted them. (Probably, I created them when I did some tests)
Then I reboot my box to get the traces but this time, it worked !

I still don't know what was the problem. I have some doubts that the xinit package was the cause.
The ATI package install/uninstall is may be somewhere a bit different from the fglrx one. May be the Xauthxx files stuck X.
I dunno.
So now my problem is fixed, and I'm absolutly sure of ... nothing :(