Bug 19537 - [GM965] X locks up machine completely on running startx sometimes.
Summary: [GM965] X locks up machine completely on running startx sometimes.
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: 7.3 (2007.09)
Hardware: x86 (IA32) Linux (All)
: high major
Assignee: Wang Zhenyu
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-01-13 09:56 UTC by W Unruh
Modified: 2009-01-19 19:24 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
dmesg of a good bootup (Cannot get one from a frozen bootup) (34.46 KB, text/plain)
2009-01-13 09:56 UTC, W Unruh
no flags Details
Xorg.0.log from crash when kde starts up from startx (26.55 KB, text/plain)
2009-01-13 09:57 UTC, W Unruh
no flags Details
xorg.conf file from machine that crashes. (2.19 KB, text/plain)
2009-01-13 09:59 UTC, W Unruh
no flags Details
xrandf --verbose output from machine that crashes ( but from a non-crash session) (4.15 KB, text/plain)
2009-01-13 10:00 UTC, W Unruh
no flags Details
glxinfo output from machine that crashes (non-crash session) (6.53 KB, text/plain)
2009-01-13 10:01 UTC, W Unruh
no flags Details
Ouput of startx on a non-crash startup (1.19 KB, application/octet-stream)
2009-01-13 10:08 UTC, W Unruh
no flags Details
Xorg.0.log file from a non-crash startup (32.95 KB, text/plain)
2009-01-13 10:13 UTC, W Unruh
no flags Details
crash Xorg.0.log (26.49 KB, text/plain)
2009-01-13 22:26 UTC, W Unruh
no flags Details
Saved output from startx from the crash (972 bytes, text/plain)
2009-01-13 22:28 UTC, W Unruh
no flags Details

Description W Unruh 2009-01-13 09:56:15 UTC
Created attachment 21943 [details]
dmesg of a good bootup (Cannot get one from a frozen bootup)

Intel Mobile GM965/GL960 Integrated Graphics controller, (rev 03) (from lspci)
2.6.27.7-1mnb SMP kernel from Mandriva 2009.0
libdrm 2.4.3, mesa 7.3, xf86-video-intel-2.5.99.2 
substituted for the  standard 2009.0 versions. 
Panasonic CF-WT Toughbook notebook computer.

I boot up in run level 3 (Console boot) and start X by running startx, with KDE the desktop.
After running startx, the screeen flashes various colours, mainly blue, and then a Mandriva splash screen comes up as the start of the kde boot up. On a regular bootup, the various text "progress messages" come up, but on the bad ones, after that splash screen and before any progress messages, the system freezes solid, so only the power switch rescues me. Machine does not respond even to ping from outsice ( and it does before the crash)
Crash happens maybe 1 in 3 times of running startx.
Comment 1 W Unruh 2009-01-13 09:57:52 UTC
Created attachment 21944 [details]
Xorg.0.log from crash when kde starts up from startx
Comment 2 W Unruh 2009-01-13 09:59:18 UTC
Created attachment 21945 [details]
xorg.conf file from machine that crashes.
Comment 3 W Unruh 2009-01-13 10:00:29 UTC
Created attachment 21946 [details]
xrandf --verbose output from machine that crashes ( but from a non-crash session)
Comment 4 W Unruh 2009-01-13 10:01:50 UTC
Created attachment 21947 [details]
glxinfo output from machine that crashes (non-crash session)
Comment 5 W Unruh 2009-01-13 10:08:41 UTC
Created attachment 21948 [details]
Ouput of startx on a non-crash startup
Comment 6 W Unruh 2009-01-13 10:13:05 UTC
Created attachment 21949 [details]
Xorg.0.log file from a non-crash startup

Of course now that I want it to crash it has refused to do so for the last 6 attempts.
Comment 7 W Unruh 2009-01-13 10:21:46 UTC
Just to clarify. The machine starts up, displays a blue screen with a cursor arrow and a flashing "busy" circle (I think this is Mandriva specific) erases that and comes up with a dark almost black splash screen with mandriva in lower right corner. In a good bootup, text (starting kded) is then displayed in lower left of screen, but in crash, everything freezes at this point.  
Comment 8 W Unruh 2009-01-13 22:26:31 UTC
Created attachment 21970 [details]
crash Xorg.0.log

I finally after about 10 tries got it to crash again.
This is the Xorg.0.log file after that crash
Comment 9 W Unruh 2009-01-13 22:28:15 UTC
Created attachment 21971 [details]
Saved output from startx from the crash

This is the output of startx (startx 2>/tmp/X) which was saved during the crash.
It is probable that the full file was not saved before the crash occurred.
Comment 10 Wang Zhenyu 2009-01-13 23:34:17 UTC
Could you try current xf86-video-intel-2.6-branch in git? That contains a critical fix which might be relate to your failure. And what's your kernel version? Have you tried newer xserver?
Comment 11 W Unruh 2009-01-14 21:13:11 UTC
OK, tried the git version-- of the intel driver-- I had to back out the lvds patch since the DS_VENDOR variable is not defined in my system. I assume it is part of the newer x11-server (I have 1.4.2) I cannot install the newer x11-server 1.5.99 version since it seems to have lots of dependencies that I cannot clear up right now. 

The git version seems to be working so far-- have tried running startx about 20 times and so far no crash. Mind you since  the crashes are sporadic, it is no proof, but I will keep trying.

I do get 
(EE) AIGLX error: dlsym for __driCreateNewScreen_20050727 failed
(/usr/lib/dri/i965_dri.so: undefined symbol: __driCreateNewScreen_20050727)
(EE) AIGLX: reverting to software rendering

Is this driCreateNewScreen_20050727 a part of the newer x server?


 
Comment 12 Wang Zhenyu 2009-01-14 21:15:12 UTC
It's in mesa, you may try intel-2008-q4 branch in mesa too.
Comment 13 W Unruh 2009-01-15 14:57:40 UTC
Is that intel-2008-q4 a complete replacement for mesa_7.3 or does one also need the regular mesa as well somehow?
Comment 14 Wang Zhenyu 2009-01-19 19:24:37 UTC
intel-2008-q4 is standalone Intel Q4 release, which makes sure all Intel driver fixes are in. Not sure about the diff between 7.3.

The lockup should be fixed reliably in master and 2.6.0 now, so I close this. Feel free to reopen if you still met that problem.

The fix commit is
commit db9f5915ce812144ffd9d2aa42e8ba856129c35e
Author: Ma Ling <ling.ma@intel.com>
Date:   Wed Jan 14 14:46:52 2009 +0800

    Disable VGA plane reliably
    
    This fixes #17235, VGA random hang on recent G45/43 board.
    From spec, SR01 bit 5 should be set before VGA plane disable through
    control register, otherwise we might get random crash and lockups.



Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.