Bug 1333

Summary: Error switching between graphical and text logins when Direct Rendering is enabled
Product: xorg Reporter: Mike <lsomike>
Component: Driver/intelAssignee: Xorg Project Team <xorg-team>
Status: RESOLVED FIXED QA Contact:
Severity: normal    
Priority: high CC: barryn, mej
Version: git   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Xorg.0.log
none
dmesg log
none
Xorg.0.log.old
none
situation bad refresh,may be DRI related?
none
Logs for: 16 Bit - DRI = No
none
Logs for: 16 Bit - DRI = Yes
none
Logs for: 24 Bit
none
Further Test Descriptions for various graphic modes
none
At 1024x768 things are better for performance
none
Confirming DRI connection.
none
Mode switching causes refresh problem.
none
This setting of the xorg.conf file causes a refresh problem.
none
This configuration rocks!
none
These are the performance related information items for the 1280 x 1024 mode
none
All the known tools - gears etc
none
xorg.conf for rolled back version
none
xorg.0.log for rolled back xorg
none
This file is from an 845 graphics controller.
none
comparative xorg.conf for 845 Graphics controller
none
This is performance for comparison of 845 GC
none
DMESG output from running system. none

Description Mike 2004-09-11 00:03:17 UTC
========= Problem ===============

Logging into init level 5 graphical login screen, 
open up a terminal session,
press:

 <CTRL><ALT><F1>

init level 3 login screen is displayed

press:

 <CTRL><ALT><F7>

Instead of being taken back to the terminal session I left, I am 
presented with a new init level 5 graphical login screen.  This is 
recorded in /var/log/messages:

 
gpm[2841]: *** info [mice.c(1766)]:

gpm[2841]: imps2: Auto-detected intellimouse PS/2

kernel: [drm:i810_wait_ring] *ERROR* space: 65520
wanted 65528

kernel: [drm:i810_wait_ring] *ERROR* lockup

gconfd (root-3715): Received signal 15, shutting
down cleanly

gdm[3365]: gdm_slave_xioerror_handler: Fatal X error
- Restarting :0

gdm(pam_unix)[3365]: session closed for user root
gconfd (root-3715): Exiting


This is a Fedora Core 3 Test 1 OS on a Sony VAIO Laptop.
Comment 1 Mike 2004-09-11 00:04:26 UTC
Created attachment 849 [details]
Xorg.0.log
Comment 2 Mike 2004-09-11 00:05:27 UTC
Created attachment 850 [details]
dmesg log
Comment 3 Mike 2004-09-11 00:06:32 UTC
Created attachment 851 [details]
Xorg.0.log.old
Comment 4 Mike 2004-09-11 00:11:24 UTC
The Fedora packages for xorg-x11* and kernel are:

xorg-x11-6.7.99.903-6
kernel-2.6.8-1.541
Comment 5 Jim Cornette 2004-09-11 07:23:21 UTC
I cannot use this driver at all for it loses refreshing functionality. This was
when booting up in runlevel 5 and then logging into X as a regular user.
This is most likely from a similar cause that Mike is experiencing w/ DRI enabled.

I'll litter this report with my Xorg.0.log file that can be compared with the
terminal to GUI screen F7 or F8
Comment 6 Jim Cornette 2004-09-11 07:26:20 UTC
Created attachment 852 [details]
situation bad refresh,may be DRI related?

The same xorg-x11 version and similar kernel used. Both Intel 815 cards.
Comment 7 Mike 2004-09-12 13:19:37 UTC
Created attachment 857 [details]
Logs for: 16 Bit - DRI = No
Comment 8 Mike 2004-09-12 13:20:32 UTC
Created attachment 858 [details]
Logs for: 16 Bit - DRI = Yes
Comment 9 Mike 2004-09-12 13:21:24 UTC
Created attachment 859 [details]
Logs for: 24 Bit
Comment 10 Mike 2004-09-12 13:24:36 UTC
Created attachment 860 [details]
Further Test Descriptions for various graphic modes

Descripton of attachments in comments 7, 8, and 9 with test results for various
graphic configurations.
Comment 11 Mike 2004-09-12 13:33:59 UTC
Comment on attachment 860 [details]
Further Test Descriptions for various graphic modes

>
>Tested three different graphics configurations:
>(all are at 1024x768)
>
>16 Bit, DRI = No.
>16 Bit, DRI = Yes.
>24 Bit.
>
>*****************
>16 Bit, DRI = No.
>*****************
>
>Open up a terminal session
>
>The sequence:
><CTRL><ALT><F1>
>
>Took me to the text login screen.
>
><CTRL><ALT><F7>
>
>Brought me back to the terminal session I left.
>
>During the terminal session I ran glxgears and an odd thing happened when
>I exited the program. Notice the last line here:
>
>-------------------------------------
>[root@linmaster log]# glxgears
>771 frames in 5.0 seconds = 154.200 FPS
>600 frames in 5.0 seconds = 120.000 FPS
>720 frames in 5.0 seconds = 144.000 FPS
>600 frames in 5.0 seconds = 120.000 FPS
>600 frames in 5.0 seconds = 120.000 FPS
>720 frames in 5.0 seconds = 144.000 FPS
>600 frames in 5.0 seconds = 120.000 FPS
>600 frames in 5.0 seconds = 120.000 FPS
>600 frames in 5.0 seconds = 120.000 FPS
>X connection to :0.0 broken (explicit kill or server shutdown).
>---------------------------------------
>
>The attachment in comment # 7
>contains; xorg.conf, Xorg.0.log, and dmesg
>
>
>******************
>16 Bit, DRI = Yes.
>******************
>
>This is my normal setup mode.  I have booted this computer 100's
>of times into this mode without problem until today (after the
>24 Bit test below)
>
>Open up a terminal session
>
>The sequence:
><CTRL><ALT><F1>
>
>Took me to the text login screen.
>
><CTRL><ALT><F7>
>
>Caused the computer to lock up without re-displaying the graphical
>login screen.  I had to reboot.
>
>The attachment in comment # 8
>contains; xorg.conf, Xorg.0.log, and dmesg
>
>
>******
>24 Bit
>******
>
>This was a disaster.  All the nastiness described in
>https://freedesktop.org/bugzilla/show_bug.cgi?id=1232
>occurred;i.e. the slowness and refresh problems.
>
>The attachment in comment # 9
>contains; xorg.conf, Xorg.0.log, and dmesg
>
>What's also puzzling, is that even the 16Bit, DRI=Yes mode
>above was a problem after running this configuration
>(slowness and refresh problems).  I finally had
>to do a cold boot in order to restore things to working condition.
>(Mounting the root partition on another OS (FC1) and copying the 16 Bit,
>DRI=Yes Xorg.conf file into place and then reboot into Fedora Core 3
>
>Also of note is the fact that I am certain that I've run 24 Bit
>graphics on an earlier version of Xorg-X11 software without problems.  If
>it will be of use to troubleshoot this problem I'll be happy to roll back
>to earlier versions of Xorg-X11* software.
>
Comment 12 Jim Cornette 2004-09-12 18:53:04 UTC
Created attachment 865 [details]
At 1024x768 things are better for performance

Though I prefer 1280 x 1024 mode, the 1024 x 768 mode enables DRI. Refer to
attached file for details regarding glxinfo, glxgears and xdpyinfo. To help
with confirming 815 behavior similar to the ctl-alt-Fn to terminal crash, I
wanted to get DRI activated.
After attachment submitted, I'll try the ctl-alt-Fn to a mingetty terminal. I
am running in runlevel 3.
Comment 13 Jim Cornette 2004-09-12 18:59:13 UTC
Created attachment 866 [details]
Confirming DRI connection. 

This happened to me with DRI enabled and at 1024 x 768 resolution. Refer to
Xorg.0.log attached
Comment 14 Jim Cornette 2004-09-12 19:28:38 UTC
Created attachment 868 [details]
Mode switching causes refresh problem.

The history of this log covers switching resolutions via system-config-display
to 1280 x 1024 after DRI testing at 1024 x 768. A later attachment regarding
this tools reconfiguration of the Xorg.conf file will be posted along with
information with another xorg.conf file that was manually edited for 1280 x
1024. The config file created with the tool bombed and caused the refresh
problem indicated.
Reverting back and rebooting was the only solution to get back to the working
server condition.
Comment 15 Jim Cornette 2004-09-12 19:34:37 UTC
Created attachment 869 [details]
This setting of the xorg.conf file causes a refresh problem.

This file was transitional from testing. It was generated from
system-conf-display and is very bad. I used the tool to select default depth of
24 and 1280 x 1024 resolution.
Comment 16 Jim Cornette 2004-09-12 19:37:19 UTC
Created attachment 870 [details]
This configuration rocks!

This file is from initial setup of the system and was set to 24 depth
originally. This is an ideal setup and exhibits no problems. The server for the
815 card looks great.
Comment 17 Jim Cornette 2004-09-12 19:38:55 UTC
Created attachment 871 [details]
These are the performance related information items for the 1280 x 1024 mode

This should complete my testing
Comment 18 Jim Cornette 2004-09-12 19:42:32 UTC
Created attachment 872 [details]
All the known tools - gears etc
Comment 19 Mike 2004-09-13 09:18:43 UTC
I rolled the version of Xorg back to the Fedora package xorg-x11-6.7.99.903-5
and the problems I noted with the 24 bit configuration above in "Further Test
Descriptions for various graphic modes" (Comment #10) have disappeared although
it didn't fix the problem switching from graphic to text mode logins.

I'll attach the xorg.conf and xorg.0.log ...

Mike Klinke
Comment 20 Mike 2004-09-13 09:20:37 UTC
Created attachment 885 [details]
xorg.conf for rolled back version
Comment 21 Mike 2004-09-13 09:21:34 UTC
Created attachment 886 [details]
xorg.0.log for rolled back xorg
Comment 22 Jim Cornette 2004-09-13 15:02:15 UTC
Created attachment 888 [details]
This file is from an 845 graphics controller. 

Test is to see if DRI is causing lockups for all versions of Graphics
controllers that use the i810 driver. The test was basically starting X from
runlevel 3, then ctl-alt-Fn to a terminal. The crash did not happen. DRI was
active and resolution was 1280 x 1024. Other attachments that will be included
are DMESG, glxinfo and the xorg.conf from the Dell with the 845 card.
Comment 23 Jim Cornette 2004-09-13 15:04:55 UTC
Created attachment 889 [details]
comparative xorg.conf for 845 Graphics controller
Comment 24 Jim Cornette 2004-09-13 15:06:10 UTC
Created attachment 890 [details]
This is performance for comparison of 845 GC
Comment 25 Jim Cornette 2004-09-13 15:10:22 UTC
Created attachment 891 [details]
DMESG output from running system.

Now that the problem is comparatively isolated to the 815 chipset and the 845
card is working fine, I hope the corrections can be made to the driver or split
the driver into 810/815 for one driver type and 830/845 for a secondary card w/
later chipsets following suit.
Comment 26 Egbert Eich 2005-03-03 06:37:36 UTC
It is very likely that these problems are gone in 6.8.2.
Please test.
Comment 27 Mike 2005-03-29 20:56:07 UTC
Today the 6.8.2 version was released to Fedora Core 3 as:

xorg-x11-6.8.2-1.FC3.13

and as, has been suggested, this problem doesn't exist any longer, at least on
my laptop.  Swithing back and forth works fine.


Regards, Mike Klinke

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.