Bug 10299

Summary: [intel] server crash with xinerama configuration
Product: xorg Reporter: Daniel Hladek <dhladek>
Component: Driver/intelAssignee: Eric Anholt <eric>
Status: RESOLVED WONTFIX QA Contact: Xorg Project Team <xorg-team>
Severity: major    
Priority: medium CC: bryce, colin, matthieu.herrb, pstehule, radek, remi
Version: 7.2 (2007.02)   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
xorg.conf with dual monitor config
none
Xorg log
none
Xorg.0.log from xorg 7.2
none
Dual monitor config with Xinerama on ubuntu Feisty
none
xinerama crash Xorg.0.log
none
Xinerama crash but different errors none

Description Daniel Hladek 2007-03-14 23:38:24 UTC
I have obtained flat panel monitor Benq fp71G+. 
My laptop: Acer tm4222 with intel945G integrated graphics, ubuntu 6.10 operating system.

I was trying to get dual monitor config working, but I have failed. Start of the x server fails with a black screen and no error messages.
I have managed to get the external screen working in single screen mode.

I have tried several configurations options of the i810 driver (MonitorLayout, DevicePresence) and screen resolutions (Mode in Screen section), but non of them worked.
Comment 1 Daniel Hladek 2007-03-14 23:39:24 UTC
Created attachment 9146 [details]
xorg.conf with dual monitor config
Comment 2 Daniel Hladek 2007-03-14 23:40:20 UTC
Created attachment 9147 [details]
Xorg log
Comment 3 Eric Anholt 2007-03-20 13:53:48 UTC
At this point, the multihead support in the current release candidate is entirely different, and we can't really support issues found with older releases.  Please re-test with current code.
Comment 4 Daniel Hladek 2007-04-20 05:52:05 UTC
The same does with 7.2 (on Ubuntu 7.04)
Comment 5 Daniel Hladek 2007-04-20 05:53:31 UTC
Created attachment 9664 [details]
Xorg.0.log from xorg 7.2
Comment 6 Daniel Hladek 2007-05-07 07:29:29 UTC
when I changed driver to xserver-xorg-video-intel (on Ubuntu Feisty and xorg 7.2), clone mode now works. But Xinerama option causes xserver crash. Xorg.0.log and xorg.conf are attached.
Comment 7 Daniel Hladek 2007-05-07 07:31:07 UTC
Created attachment 9893 [details]
Dual monitor config with Xinerama on ubuntu Feisty

Clone mode works, Xinerama crash
Comment 8 Daniel Hladek 2007-05-07 07:32:18 UTC
Created attachment 9894 [details]
xinerama crash Xorg.0.log

log with Xinerama crash
Comment 9 Rémi Cardona 2007-05-09 11:23:20 UTC
Created attachment 9906 [details]
Xinerama crash but different errors

I get this bug too but with a different log output. See attached around line 475 :

(II) intel(0): I2C bus "DVOI2C_E" initialized.
(EE) intel(0): detecting sil164
(EE) intel(0): Unable to read from DVOI2C_E Slave 112.

And later on line 1059 :
(EE) intel(0): I830 Vblank Pipe Setup Failed 0

I still have two or three different xinerama configurations available (I symlink xorg.conf using a homemade script) that used to work on xorg-server 1.1 (around last summer). None of them work anymore with xorg-server 1.3.0.0 and xf86-video-intel 2.0.0
Comment 10 Eric Anholt 2007-08-09 13:47:31 UTC
Failure with xinerama is the expected results with the current code, and we have no plans to fix it.  You can use RandR 1.2-based configuration, statically (xorg.conf(5) has information in git) or at runtime using xrandr(1), which allows for many common multihead configurations.
Comment 11 Eric Anholt 2007-08-09 15:08:45 UTC
*** Bug 10706 has been marked as a duplicate of this bug. ***
Comment 12 Eric Anholt 2007-08-09 15:30:24 UTC
*** Bug 11373 has been marked as a duplicate of this bug. ***
Comment 13 Radek Podgorny 2007-08-10 00:13:52 UTC
...no plans to fix it? Are you kidding? I'd expect a sane error message rather than a traceback...
Comment 14 Colin Guthrie 2007-08-10 00:57:55 UTC
I have to agree. It's not very nice having X crash just cos you put a special sequence of characters into a file. If the conf parser can't handle it without segfaulting then it doesn't smack of well written and considered code (regardless of how well randr 1.2 and the rest of the actual server stc. works)
Comment 15 Brice Goglin 2007-08-10 01:05:27 UTC
You should probably read "no plans to fix it" as "patch welcome" :)
Comment 16 Michael Fu 2007-08-23 18:41:42 UTC
xrandr is preferred method by now.
Comment 17 Rémi Cardona 2007-08-23 22:59:53 UTC
Will there be a way to work around the framebuffer limitation (which forces me to logically align my screens vertically when they are physically aligned horizontally) ?

Will there be a way to work around the memory limitations so that I can use the DRI if I put the Virtual 2048 2048 to "fix" the first issue ?
Comment 18 Colin Guthrie 2007-08-24 01:34:15 UTC
(In reply to comment #16)
> xrandr is preferred method by now.

We all know what the preferred method is.

I think the main issue was that incorrect syntax in xorg.conf actually CRASHES the server!!! This is not good in any application!


As of the 2048 DRI limit, this is technically fixable but involves some hard sums that no-one has stepped up to the plate to implement. Keith P has described what needs done on Xorg mailing list in the past.

You can adjust the position of your monitors but it may be that the placement you refer to is to keep them with in the 2048 bounding box?
Comment 19 Gordon Jin 2007-11-11 00:13:03 UTC
*** Bug 13163 has been marked as a duplicate of this bug. ***
Comment 20 Gordon Jin 2008-08-09 19:21:24 UTC
*** Bug 13517 has been marked as a duplicate of this bug. ***
Comment 21 Tino Keitel 2008-09-16 15:37:27 UTC
Please, can we get at least some sort of Xinerama support back? I don't care for fancy features like XV or DRI support. I just want a working dualhead setup for productivity.

Currently, RandR is only supported by ATI and Intel AFAIK. All the other drivers require Xinerama for a multihead setup. With the current RandR 1.2 vs. Xinerama conflict, there is a _very_ small subset of possible configurations that can use a multihead setup at all if one of the card is an Intel card. And as Intel is builtin, there is no chance to replace it with another card.
Comment 22 Tino Keitel 2008-09-18 04:48:42 UTC
For the records: I worked around this issue by fetching the source of the last pre-2.0 version (1.7.4), fixed it to build against the Xorg 7.3 headers, added the PCI ID of my i965GM and got working dual head using Xinerama. However, I'm afraid of changes in later Xorg versions that don't allow me to compile the old i810 driver anymore. This "solution" is also limited to hardware that is old enough to be supported by the old i810 driver.

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.