FS#16682 - Updates on 10/8 broke X for intel graphics on Dell GX280

Attached to Project: Arch Linux
Opened by David C. Rankin (drankinatty) - Friday, 16 October 2009, 00:48 GMT
Last edited by Jan de Groot (JGC) - Wednesday, 21 October 2009, 07:05 GMT
Task Type Bug Report
Category Packages: Testing
Status Closed
Assigned To Tobias Powalowski (tpowa)
Jan de Groot (JGC)
Andreas Radke (AndyRTR)
Architecture i686
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:

Guys, I have pulled my hair out on this one. Something in the updates I did on Oct. 8 broke X for my install on my Dell GX280 box with the standard on-board intel 915 graphic controller. What happens is on boot X tries to start and then crashes. You get a glimpse of the X-cursor and then Wham, X dies. I have had no problems at all with this box running kde4 since March until the updates on 10/8. Since then, I have worked through the archwiki on intel, try both the "Early Start" and "Late Start" methods - same crash either way. I have downgraded back to libgl-7.5.1-2 intel-dri-7.5.1-2 xf86-video-intel-2.8.1-1 -- no help, same crash.

I have tried to collect as much information as possible to help get to the bottom of this. As mentioned in my posts to the list, when I ran the strace startx, I did have a sustained X session with a white xterm in the top left corner of the screen until X crashed. (of course I could do nothing with it) For starters, I have collected the following information:

strace startx:

http://www.3111skyline.com/download/Archlinux/bugs/supersff/strace-startx.txt

pacman log:

http://www.3111skyline.com/download/Archlinux/bugs/supersff/pacman.log.bz2

kdm.log

http://www.3111skyline.com/download/Archlinux/bugs/supersff/kdm.log.bz2

kdm.log - after fbdev install

http://www.3111skyline.com/download/Archlinux/bugs/supersff/kdm.log-2009-10-15.txt.bz2

Xorg.0.log

http://www.3111skyline.com/download/Archlinux/bugs/supersff/Xorg.0.log.segfault

Xorg.0.log after fbdev install

http://www.3111skyline.com/download/Archlinux/bugs/supersff/Xorg.0.log.postfbdev

What seems to be happening from Xorg.0.log is:

(==) Log file: "/var/log/Xorg.0.log", Time: Thu Oct 15 13:36:06 2009
<snip>

(II) UnloadModule: "evdev"
(II) Power Button: Close
(II) UnloadModule: "evdev"

Backtrace:
0: /usr/bin/X (xorg_backtrace+0x3b) [0x809facb]
1: /usr/bin/X (0x8048000+0x54ab5) [0x809cab5]
2: (vdso) (__kernel_rt_sigreturn+0x0) [0xb776d40c]
3: /usr/lib/xorg/modules/libshadow.so (shadowRemove+0x4e) [0xb70cfd9e]
4: /usr/lib/xorg/modules/libshadow.so (0xb70cf000+0x1234) [0xb70d0234]
5: /usr/bin/X (0x8048000+0x68cf4) [0x80b0cf4]
6: /usr/bin/X (0x8048000+0x15c2ac) [0x81a42ac]
7: /usr/bin/X (0x8048000+0x4f19c) [0x809719c]
8: /usr/bin/X (0x8048000+0x17447c) [0x81bc47c]
9: /usr/bin/X (0x8048000+0x13d3de) [0x81853de]
10: /usr/lib/xorg/modules/drivers/vesa_drv.so (0xb7740000+0x38a8) [0xb77438a8]
11: /usr/bin/X (0x8048000+0x6787c) [0x80af87c]
12: /usr/bin/X (0x8048000+0xb475d) [0x80fc75d]
13: /usr/bin/X (0x8048000+0xfa245) [0x8142245]
14: /usr/bin/X (0x8048000+0x141112) [0x8189112]
15: /usr/bin/X (0x8048000+0xa0430) [0x80e8430]
16: /usr/bin/X (0x8048000+0xc17b9) [0x81097b9]
17: /usr/bin/X (0x8048000+0x114872) [0x815c872]
18: /usr/lib/xorg/modules/extensions/libglx.so (0xb725f000+0x379e3)
[0xb72969e3]
19: /usr/bin/X (0x8048000+0x1ab48) [0x8062b48]
20: /lib/libc.so.6 (__libc_start_main+0xe6) [0xb7385a36]
21: /usr/bin/X (0x8048000+0x1a6a1) [0x80626a1]
Segmentation fault at address 0x1

Sometimes the Xorg.0.log is clean, but X still crashes. All packages are current and I have testing and community-testing repositories enabled. The intel packages are:

intel-dri-7.6-1-i686.pkg.tar.gz
xf86-video-intel-2.9.0-1-i686.pkg.tar.gz
libgl-7.6-1-i686.pkg.tar.gz

I also installed

xf86-video-fbdev-0.4.1-2-i686.pkg.tar.gz

to try and address the fbdev error. No help. One error that keeps standing out is

FATAL: Module fbcon not found.

I have the latest kernel so I guess it should be compiled in, but acts like it isn't:

kernel26-2.6.31.4-1-i686.pkg.tar.gz

Some additional files:

http://www.3111skyline.com/download/Archlinux/bugs/supersff/mkinitcpio.conf
http://www.3111skyline.com/download/Archlinux/bugs/supersff/menu.lst
http://www.3111skyline.com/download/Archlinux/bugs/supersff/rc.conf

Guys, I'm stumped! I don't mind doing extra work, so just tell me what you want to see and I'm happy to be your fingers at keyboard on this end. Thanks.
This task depends upon

Closed by  Jan de Groot (JGC)
Wednesday, 21 October 2009, 07:05 GMT
Reason for closing:  Not a bug
Additional comments about closing:  See last comment
Comment by David C. Rankin (drankinatty) - Friday, 16 October 2009, 04:13 GMT
Here were also some hints from the /var/log/errors.log file I found interesting.

errors.log:

Oct 15 15:01:48 supersff kernel: [drm:i915_wait_request] *ERROR* something (likely vbetool) disabled interrupts, re-enab
ling
Oct 15 15:01:48 supersff kernel: render error detected, EIR: 0x00000010
Oct 15 15:01:48 supersff kernel: page table error
Oct 15 15:01:48 supersff kernel: PGTBL_ER: 0x00000010
Oct 15 15:01:48 supersff kernel: [drm:i915_handle_error] *ERROR* EIR stuck: 0x00000010, masking
Oct 15 15:01:48 supersff kernel: render error detected, EIR: 0x00000010
Oct 15 15:01:48 supersff kernel: page table error
Oct 15 15:01:48 supersff kernel: PGTBL_ER: 0x00000010
Oct 15 15:02:31 supersff kdm[1639]: X server died during startup
Oct 15 15:02:31 supersff kdm[1639]: X server for display :0 can't be started, session disabled
Oct 15 15:04:26 supersff nmbd[1479]: [2009/10/15 15:04:26, 0] nmbd/nmbd.c:terminate(71)
Oct 15 15:04:26 supersff nmbd[1479]: Got SIGTERM: going down...
Oct 15 15:04:59 supersff kernel: render error detected, EIR: 0x00000010
Oct 15 15:04:59 supersff kernel: page table error
Oct 15 15:04:59 supersff kernel: PGTBL_ER: 0x00000010
Oct 15 15:04:59 supersff kernel: [drm:i915_handle_error] *ERROR* EIR stuck: 0x00000010, masking
Oct 15 15:04:59 supersff kernel: render error detected, EIR: 0x00000010
Oct 15 15:04:59 supersff kernel: page table error
Oct 15 15:04:59 supersff kernel: PGTBL_ER: 0x00000010
Oct 15 15:04:59 supersff kernel: iTCO_wdt: failed to reset NO_REBOOT flag, reboot disabled by hardware

17:46 supersff:/var/log> grep kdm errors.log
Oct 14 14:23:03 supersff kdm: :0[1496]: Abnormal termination of greeter for display :0, code 127, signal 0
Oct 14 14:45:30 supersff kdm: :0[1543]: Abnormal termination of greeter for display :0, code 127, signal 0
Oct 14 15:28:56 supersff kdm: :0[1502]: Abnormal termination of greeter for display :0, code 127, signal 0
Oct 14 16:08:18 supersff kdm: :0[1522]: Abnormal termination of greeter for display :0, code 127, signal 0
Oct 14 16:15:04 supersff kdm: :0[1575]: Abnormal termination of greeter for display :0, code 127, signal 0
Oct 15 13:21:45 supersff kdm: :0[7402]: Abnormal termination of greeter for display :0, code 127, signal 0
Oct 15 13:28:56 supersff kdm: :0[1518]: Abnormal termination of greeter for display :0, code 127, signal 0
Oct 15 13:34:22 supersff kdm: :0[1497]: Abnormal termination of greeter for display :0, code 127, signal 0
Oct 15 13:36:07 supersff kdm: :0[1541]: Abnormal termination of greeter for display :0, code 127, signal 0
Oct 15 13:45:02 supersff kdm: :0[1575]: Abnormal termination of greeter for display :0, code 127, signal 0
Oct 15 15:02:31 supersff kdm[1639]: X server died during startup
Oct 15 15:02:31 supersff kdm[1639]: X server for display :0 can't be started, session disabled
Oct 15 15:29:30 supersff kdm_greet[1628]: Cannot load /usr/share/apps/kdm/faces/.default.face: No such file or directory
Comment by Andreas Radke (AndyRTR) - Friday, 16 October 2009, 06:33 GMT
can you both please try to starx X with a different desktop, not qt/kde. I know testing intel drivers are buggy but I will make sure the reason is not kwin compositor.
Comment by David C. Rankin (drankinatty) - Saturday, 17 October 2009, 08:12 GMT
Andreas,

I have openbox and iceWM as well as twm. I will try to start those and report back tomorrow (I'm delerious presently :-)
Comment by David C. Rankin (drankinatty) - Saturday, 17 October 2009, 21:41 GMT
Well, this is strange. If I use vncserver/vncviewer, I can get twm and openbox to run fine. However, I'm can't get them to run from the console. Attempting to run them with "openbox" or "twm" gives a DISPLAY variable error. I have tried setting DISPLAY manually to :0.0 and exporting DISPLAY, but still no joy. I have tried starting openbox with ./config/openbox/autostart.sh but it didn't work (I know this problem will just take some more reading on my part)

As for anything KDE related, it is still dead-as-a-doornail. I have a new set of files from the box at:

http://www.3111skyline.com/download/Archlinux/bugs/supersff/1017/

1017-Xcrash-files.tar.bz2
daemon.log
dmesg-2009-10-17.txt
errors.log
everything.log
kdm.log
lspci-2009-10-17.txt
menu.lst
mkinitcpio.conf
pacman.log
rc.conf
Xorg.0.log
Xorg.0.log.old
Xorg.1.log

The 1017-Xcrash-files.tar.bz2 file is just a bzipped tar archive of all the files above for convenience. (~98k).

Do you have a favorite way to start an alternate X?
Comment by David C. Rankin (drankinatty) - Sunday, 18 October 2009, 04:54 GMT
Well, this is strange. If I use vncserver/vncviewer, I can get twm and openbox to run fine. However, I'm can't get them to run from the console. Attempting to run them with "openbox" or "twm" gives a DISPLAY variable error. I have tried setting DISPLAY manually to :0.0 and exporting DISPLAY, but still no joy. I have tried starting openbox with ./config/openbox/autostart.sh but it didn't work (I know this problem will just take some more reading on my part)

As for anything KDE related, it is still dead-as-a-doornail. I have a new set of files from the box at:

http://www.3111skyline.com/download/Archlinux/bugs/supersff/1017/

1017-Xcrash-files.tar.bz2
daemon.log
dmesg-2009-10-17.txt
errors.log
everything.log
kdm.log
lspci-2009-10-17.txt
menu.lst
mkinitcpio.conf
pacman.log
rc.conf
Xorg.0.log
Xorg.0.log.old
Xorg.1.log

The 1017-Xcrash-files.tar.bz2 file is just a bzipped tar archive of all the files above for convenience. (~98k).

Do you have a favorite way to start an alternate X?
Comment by David C. Rankin (drankinatty) - Sunday, 18 October 2009, 04:57 GMT
Oops, sorry for the duplicate post. My browser "resent" the data when I went back to the page in the history list.

Devs,

Here is the current status. I can get X to run twm, openbox and enlightenment just fine both from the console and via vncserver. Anything kde related is DEAD. I am next going to try and find out how the change the kernel config to enable fbcon and see if that helps. Please let me know if you have any other ideas. Thanks
Comment by David C. Rankin (drankinatty) - Sunday, 18 October 2009, 05:01 GMT
Andeas,

I set openbox to start through ~/.xinitrc - works fine. I just start enlightment from the command line with 'starte16' and it also works fine. (first time to look at enlightenment - strange, but kind of neat looking. Way way different from anything else I've used)
Comment by David C. Rankin (drankinatty) - Sunday, 18 October 2009, 09:58 GMT
Guys,

I come with hat-in-hand, in a humble fashion befitting of the fact that there is no problem with X as it relates to the intel gpu despite the turmoil caused by moving the module into KMS. In fact it seems to work quite lovely.

During the neverending troubleshooting of this problem, I encountered the fact that kde3 applications that always worked fine in openbox or twm, would no longer work and would refuse to start complaining about libqt belonging to the wrong ELF class -- an innocent enough sounding problem. Discussing the same on the list, I had thought surely this qtlib issue couldn't be the problem that is causing X to die -- could it? Master Eric Bélanger pointed out the fact that it appeared that I had somehow abused my system by installing a few x86_64 packages on a prestine i686 install. A bit more poking around showed the qt3-enhanced package was in fact the package that was causing all of the holy hell with X and a bit more poking around confirmend that /var/cache/pacman/pkg/qt3-enhanced-3.3.8-26-x86_64.pkg.tar.gz was happily residing in the cache directory. Hmmm...

......That's when the lightbulb winked on and when the feeling of SHAME begin to settle in on my consciousness. Further investigation revealed that on 10/8 when I had updated my box and the new pacman package was installed, I used the pacman.conf from an earlier update of my x86_64 Arch server as a "go-by" to quick move the new pacman.conf.pacnew into place. It would seem that there was one occasion with the community repository that an 'x86_64' did not get replaced by and 'i686', thus 'the loose nut behind the keyboard' had installed some:

04:32 supersff:~> l /var/cache/pacman/pkg/*x86_64* | wc -l
52

Yes FIFTY-TWO x86_64 packages on my fine little i686 box. (SMACK to make sure I have learned the lesson and SMACK, just for good measure)

Long story short, after reinstalling the correct qt3-enhanced package, X and kde now start without a problem on this little box. Quite amazingly with 51 x86_64 packages remaining to be removed and reinstalled as i686.

So I can certainly say I have learned a great deal more about KMS, regenerating Arch's initramfs, the handy little mkinicpio utility and a hundred more little tidbits as I have pined away at this problem. As an added benefit I even had a moment of enlightenment when I executed starte16 as a test to help diagnose the problem (moments of enlightenment are few and far between...)

And just as certainly as learning occurred, this bug can now be closedd. Because the problem was once-again not a problem with Arch software, but instead a hardware failure when the loose nut behind the keyboard went completely off on a wild architecture chase.

So thank you for all of your help and patients, I bid you a humble adieu with two final words -- CLOSE IT!

Now how in the hell do I remove and reinstall 51 other packages without thrashing the box? (one at a time ;-)
Comment by David C. Rankin (drankinatty) - Sunday, 18 October 2009, 10:09 GMT
Thankfully the remaining 51 packages are kde3 packages that will be removed with -R -d and then reinstalled -- whew:

05:06 supersff:~> for i in $(ls -1 /var/cache/pacman/pkg/*x86_64*); do pkg=${i##*/pkg/}; echo "$pkg"; done
kdemod3-kdegames-atlantik-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-base-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kasteroids-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-katomic-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kbackgammon-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kbattleship-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kblackbox-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kbounce-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kenolaba-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kfouleggs-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kgoldrunner-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kjumpingcube-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-klickety-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-klines-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kmahjongg-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kmines-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-knetwalk-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kolf-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-konquest-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kpat-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kpoker-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kreversi-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-ksame-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kshisen-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-ksirtet-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-ksmiletris-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-ksnake-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-ksokoban-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kspaceduel-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-ktron-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-ktuberling-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-kwin4-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdegames-lskat-3.5.10-4-x86_64.pkg.tar.gz
kdemod3-kdenetwork-kopete-otr-0.7-2-x86_64.pkg.tar.gz
kdemod3-kdepim-akregator-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-backup-0.2-2-x86_64.pkg.tar.gz
kdemod3-kdepim-base-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-kaddressbook-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-kalarm-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-kandy-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-karm-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-kitchensync-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-kmail-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-knode-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-knotes-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-kontact-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-korganizer-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-korn-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-kpilot-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdepim-wizards-3.5.10-1-x86_64.pkg.tar.gz
kdemod3-kdestyle-domino-0.4-3-x86_64.pkg.tar.gz
kdemod3-kdestyle-polyester-1.0.5-1-x86_64.pkg.tar.gz
Comment by David C. Rankin (drankinatty) - Tuesday, 20 October 2009, 16:31 GMT
Guys,

This bug needs to be closed, but I can't find a way to do it. So I'll leave it up to you. Close it when you can. The original problem was fixed (operator error)

Loading...