Bug 9296

Summary: Intel 915GM stopped working with 7.0.99.903
Product: xorg Reporter: Gabriel C <nix.or.die>
Component: Driver/intelAssignee: Alan Hourihane <alanh>
Status: RESOLVED FIXED QA Contact:
Severity: major    
Priority: high CC: caglar
Version: 7.0.99.903 (7.1RC3)   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Bug Depends on:    
Bug Blocks: 6666    
Attachments:
Description Flags
xorg-1.1.99.901.log
none
xorg-1.1.99.903.log
none
dmesg
none
xorg.conf
none
lspci
none
log with ErrorF around the reverted stuff
none
patch to int10 none

Description Gabriel C 2006-12-09 16:12:44 UTC
Hi,

using latest RC3 release ( modular ) we got an strange bug report in Frugalware
Linux , on an IBM ThinkPad T43 notebook model 1872 the Intel card stopped
working , while was fine with RC1.

here the original email :

On the following hardware:
- IBM ThinkPad T43 notebook model 1872
- Intel 915GM VGA (lspci -vv attached)
- 1024x768 resolution

I experience the following problem: when starting X with the startkde
script, X comes up but the image is absolutely unusable. Approx. the
lower 4/5th of the screen is mangled, with a dark stripe above - looks
like when a sync value is wrong. I can use the apps with keyboard
shortcuts but I don't see what I'm doing. It is also important that
when I downgrade to 1.1.9.9.901 and then reinstall -903, on the first
start attempt, X gets locked up (no Ctrl+Alt+whatever). Before
starting X, I must start a 'sleep 60; killall X' on another tty to
regain access. Afterwards, X will start each time with the above
described unreadable display and can be killed. 1.1.99.901 works well.

Most WW and EE lines are also there with -901. Here is the difference with -903:

(EE) I810(0): vm86() syscall generated signal 8.
(EE) I810(0): vm86() syscall generated signal 8.
(EE) I810(0): unknown reason for exception
(EE) I810(0): cannot continue
(EE) I810(0): vm86() syscall generated signal 8.


Software versions:
xorg-server 1.1.99.903-4
xf86-video-i810 1.7.2-1
mesa 6.5.2-2
kernel 2.6.19-1

Attachments:
xorg.conf - note that AIGLX is disabled.
901.log - Xorg log with 1.1.99.901 as a comparison
903.log - Xorg verbose log with 1.1.99.903
lspci.LGee - lspci -vv
dmesg.LGee - dmesg

thanks
best regards
Gabor

He tested afre this all 1.6* && all 1.7.* drivers too with no luck.

I'm not really sure the problem is caused by the driver it may be a bug in
xorg-server as well.

I'll attach all logs I've got to this bug report.


Regards 

Gabriel
Comment 1 Gabriel C 2006-12-09 16:14:19 UTC
Created attachment 8043 [details]
xorg-1.1.99.901.log
Comment 2 Gabriel C 2006-12-09 16:15:02 UTC
Created attachment 8044 [details]
xorg-1.1.99.903.log
Comment 3 Gabriel C 2006-12-09 16:15:49 UTC
Created attachment 8045 [details]
dmesg
Comment 4 Gabriel C 2006-12-09 16:16:33 UTC
Created attachment 8046 [details]
xorg.conf
Comment 5 Gabriel C 2006-12-09 16:17:01 UTC
Created attachment 8047 [details]
lspci
Comment 6 Alan Hourihane 2006-12-10 13:07:37 UTC
Are your bootup scripts using vbetool to save the graphics state ??

If so, disable vbetool.
Comment 7 Gabriel C 2006-12-10 13:18:54 UTC
Hi,

no we don't use vbetool. 
Comment 8 Alan Hourihane 2006-12-10 13:35:15 UTC
Does -902 exhibit the same problem ?
Comment 9 Gabriel C 2006-12-10 13:40:49 UTC
I can't tell you about RC2 we jumped from RC1 to RC3 but if you want I could build 
RC2 xorg-server and let the user test it.
Comment 10 Alan Hourihane 2006-12-11 03:19:11 UTC
Yes, it would be useful for you to test RC2.
Comment 11 Alan Hourihane 2006-12-11 04:52:45 UTC
It might be worth reverting the patch in bug #6751 to see if that helps.

Or you could just umount /sysfs and try starting X which would have the same
effect to see if that helps.
Comment 12 Gabriel C 2006-12-11 05:02:06 UTC
Ok I'll try this.
There is the same problem but a different box :

http://lists.freedesktop.org/archives/xorg/2006-December/020405.html

( Btw I've CC'ed you on my replay to this thread I hope you don't mind )
Comment 13 Alan Hourihane 2006-12-11 07:02:20 UTC
If reverting #6751 didn't work, try reverting #6750.
Comment 14 Gabriel C 2006-12-11 12:59:36 UTC
Hi Alan,

reverting commit a33419718171a0dee0f0ffe6ceb39e6f60696416 (
http://gitweb.freedesktop.org/?p=xorg/xserver.git;a=commit;h=a33419718171a0dee0f0ffe6ceb39e6f60696416)

fixed this problem.
Comment 15 Alan Hourihane 2006-12-11 13:32:51 UTC
In that function can you add the following (with the patch reverted)...

ErrorF("Int10Current->Tag is 0x%x\n",Int10Current->Tag);
ErrorF("TAG(PciCfg1Addr) is 0x%x\n",TAG(PciCfg1Addr));

And it'll print the results out in the Xserver logfile.

Can you then upload the logfile here.
Comment 16 Gabriel C 2006-12-11 13:39:01 UTC
sure I can , just our user is off now but I think he can test this tomorrow. 
If you need it now maybe ask the guy from the Mailing List :)
Comment 17 Gabriel C 2006-12-12 12:21:19 UTC
Created attachment 8070 [details]
log with  ErrorF around the reverted stuff
Comment 18 Gabriel C 2006-12-12 12:22:16 UTC
ok here it is , let me know if you need more infos or tests.
Comment 19 Alan Hourihane 2006-12-12 12:49:07 UTC
Created attachment 8071 [details] [review]
patch to int10

Can you try this patch against the int10 file ??
Comment 20 Gabriel C 2006-12-12 13:26:10 UTC
sure I can. I'll prepare xorg-server with this patch and let the user test it.
anyway I can post the result tomorrow first. 

pardus ping :) If you get sometime could you pls test this too ? Is faster
because you own a box with this broken card and you don't need to wait for users
to test.



Comment 21 caglar 2006-12-12 13:30:10 UTC
(In reply to comment #20)
> pardus ping :) If you get sometime could you pls test this too ? Is faster
> because you own a box with this broken card and you don't need to wait for users
> to test.

I will.

Comment 22 caglar 2006-12-12 13:49:40 UTC
(In reply to comment #19)
> Can you try this patch against the int10 file ??

Same problem exists with that patch + unreverted xorg-server-1.1.99.903
Comment 23 Alan Hourihane 2006-12-12 14:34:42 UTC
Can you add an ErrorF to print the values of pvp->bus, pvp->device, pvp->func
and upload a log ?
Comment 24 Alan Hourihane 2006-12-13 04:13:01 UTC
Just committed the fix for this. Closing.

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.