XOrg dualhead regression

Hi Jason, Thanks for your report.

Is there a particular patch to apply to fix this?

No, not afaik, I just wanted you guys to be aware of this so you didn't ship something that was completely broken.

I'm setting the severity to Major because this represents a regression, but the bug also needs more specific info.

It seems that my problem is the same. My current configuration (testing some things :) is nvidia in agp port (main card) and very old pci s3:

00:09.0 VGA compatible controller: S3 Inc. 86c775/86c785 [Trio 64V2/DX or /GX] (rev 14)
01:00.0 VGA compatible controller: nVidia Corporation NV11 [GeForce2 MX/MX 400] (rev a1)

When I boot dapper beta desktop cd, X won't load but ncurses window saying something like "X server configuration failed" appears instead. And then there is many times the prompt (ubuntu@ubuntu $) displayed. It happens with "Boot in safe graphics mode" too.

please attach Xorg.0.log from the latest Dapper Beta release if possible. Thanks.

Ok, attaching the log. A little corrections to my previous post are that in "normal" boot mode X won't start and I get only black screens on both monitors. I am not able to kill X with ctrl+alt+backspace nor switch to another console.

This log comes from booting into safe graphics mode. The dialog says about X failure and asks me if I want to display the output of X server but then the login is shown and the screen (text mode) is messed up. I am able to switch to another tty and work there.

Attaching the xorg.conf file, just in case it's needed, too.

this is my dualhead xorg.conf from dapper, which is not working in breezy

I have a matrox g450 and a SIS graphics adapter onboard. The SIS can not be switch off.
I used breezy in dual mode, with 2 TFT-Displays connected to the vga-ports. Today i reinstalled ubuntu breezy to a new partition and made a dist-upgrade to dapper. I tried then the same settings for dualhead as i had before. As the x-server starts the SIS display works correct but the tft connected to the matrox switches on but displays nothing.

better to reassign it to the right package.

Which specific configurations are affected by this problem? It does not seem that all dualhead configurations are broken in dapper.

No, not all configs are affected. It is mainly an issue when there are cards from different vendors in the same machines and they are initialized in the wrong order. It is a problem that daniels and I faced in hoary when working on multiseat.

In X7.0. it become worst but there are no fixes upstream yet.

Fabio

I can also confirm this bug as I've been wrestling with it for a few weeks while experimenting with a box that has up to 6 video cards installed. Basically, some video cards (both PCI and AGP) are not being initialized properly unless they are set as the primary video card in the BIOS. I have access to several cards of various chipsets that exhibit this behavior so if it is a driver issue then more than one driver is affected. I can also confirm that the problem does not appear to exist in Breezy Badger (Xorg 6.8.2) only in Dapper Drake testing (Xorg 7.0.0) but I'm going to test that theory some more in the next few days to be sure.

Please let me know if there is any information, no matter how detailed, I could provide to help get this nasty bug solved.

This is an Xorg 6.8.2 log from a PC with 1 AGP and 4 PCI video cards in Xinerama mode. The setup works perfectly while running Breezy.

This is an Xorg 7.0.0 log from a PC with 1 AGP and 4 PCI video cards in Xinerama mode. Although the same setup worked in Breezy, after upgrading to Dapper X freezes while attempting to initialize and the system is completely unresponsive.

I have been experiencing hard lockups every time I start X dualhead also, since upgrading to Ubuntu Dapper (worked fine with Breezy). The attached Xorg log file is what I'm able to grab before my system locks solidly (about 10-15 seconds after startx). Before the lockup, a normal X desktop is displayed on the ATI and the Ubuntu splash screen etc comes up, mouse moves etc. At the time of the lockup, the i810 display fills with vertical coloured bars, and the X desktop that had been displayed on the Mach64 blanks at the time the bars are displayed. Only C+A+D works after this point. N.B.: This is with the ATI Mach64 selected as the primary display in the BIOS.

Previously (on Breezy) I had to change my primary display (in PC BIOS) to be the i810 for dualhead to work at all; the complete opposite of the Windows install on the same machine, which only supports dualhead if the ATI Mach64 is primary display. Since upgrading to Dapper, selecting i810 as primary in BIOS no longer works, and X fails to intialise the ATI card which is simply ignored (although X works correctly as a single display setup on the i810).

My Xorg config utilises both Mach64 and i810, as below:

Section "Device"
        Identifier "ATI Rage Pro XL"
        Driver "ati"
        BusID "PCI:2:10:0"
EndSection

Section "Device"
        Identifier "i815"
        Driver "i810"
        BusID "PCI:0:2:0"
        VideoRAM 8192
EndSection

Section "Monitor"
        Identifier "LM-500"
        Option "DPMS"
        HorizSync 30-60
        VertRefresh 60-75
EndSection

Section "Monitor"
        Identifier "Philips 151AX"
        Option "DPMS"
        HorizSync 30-60
        VertRefresh 60-75
EndSection

I have been experiencing system lockups with dual-headed configs also. This is a show-stopper for me, unfortunately.
My setup uses
1 x ATI radeon QD AGP video card
1 x ATI radeon VE PCI video card
So my setup is not multi-vendor.
If I set the drivers to 'ati' or 'radeon', the system locks up as the cards initialize. If I set the drivers to 'vesa', on the primary (AGP) I get an 'unsupported video mode' box, and on the secondary, it just doesn't initialize.

/var/log/Xlog.0.log available on request.

Xorg.conf sections:

Section "Device"
  BoardName "Radeon QD"
  BusID "1:0:0"
  Driver "ati"
  Identifier "Device[0]"
# Screen 0
  VendorName "ATI"
EndSection

Section "Device"
  BoardName "Radeon VE"
  BusID "2:6:0"
  Driver "ati"
  Identifier "Device[1]"
# Screen 0
  VendorName "ATI"
EndSection

Section "ServerLayout"
  Identifier "Layout[all]"
  InputDevice "Keyboard[0]" "CoreKeyboard"
  InputDevice "Mouse[1]" "CorePointer"
  Option "Clone" "off"
  Option "Xinerama" "on"
  Screen "Screen[0]"
  Screen "Screen[1]" RightOf "Screen[0]"
EndSection

Have you tried the fglrx driver?

I think this bug is same as Bug#6751 of Xorg.
Xorg cann't find second video card.

I try to fix it use the patch of

    https://bugs.freedesktop.org/show_bug.cgi?id=6751

But it doesn't work on Dapper 6.06.

Similar to Ashley, when set up in a Dual Head configuration that worked in Breezy, I get Blue/Green/Yellow vertical bars on my second screen (really old nvidia) and Dapper locks. I reverted to a single head setup as a workaround. Below is my xorg.conf file. You can see my original configuration commented out in the ServerLayout section.

Section "Files"
 FontPath "/usr/share/X11/fonts/misc"
 FontPath "/usr/share/X11/fonts/cyrillic"
 FontPath "/usr/share/X11/fonts/100dpi/:unscaled"
 FontPath "/usr/share/X11/fonts/75dpi/:unscaled"
 FontPath "/usr/share/X11/fonts/Type1"
 FontPath "/usr/share/X11/fonts/CID"
 FontPath "/usr/share/X11/fonts/100dpi"
 FontPath "/usr/share/X11/fonts/75dpi"
        # paths to defoma fonts
 FontPath "/var/lib/defoma/x-ttcidfont-conf.d/dirs/TrueType"
 FontPath "/var/lib/defoma/x-ttcidfont-conf.d/dirs/CID"
EndSection

Section "Module"
 Load "GLcore"
 Load "i2c"
 Load "bitmap"
 Load "ddc"
 Load "dri"
 Load "extmod"
 Load "freetype"
 Load "glx"
 Load "int10"
 Load "type1"
 Load "vbe"
EndSection

Section "InputDevice"
 Identifier "Generic Keyboard"
 Driver "kbd"
 Option "CoreKeyboard"
 Option "XkbRules" "xorg"
 Option "XkbModel" "pc104"
 Option "XkbLayout" "us"
EndSection

Section "InputDevice"
 Identifier "Configured Mouse"
 Driver "mouse"
 Option "CorePointer"
 Option "Device" "/dev/input/mice"
 Option "Protocol" "ImPS/2"
 Option "Emulate3Buttons" "true"
 Option "ZAxisMapping" "4 5"
EndSection

Section "Device"
 Identifier "Intel Corporation 82845G/GL[Brookdale-G]/GE Chipset Integrated Graphics Device"
 Driver "i810"
 BusID "PCI:0:2:0"
EndSection

Section "Device"
 Identifier "Graphics Blaster [Riva TNT]"
 Driver "nv"
 BusID "PCI:1:7:0"
 Option "NvAGP" "0"
EndSection

Section "Monitor"
 Identifier "Generic Monitor"
 Option "DPMS"
 HorizSync 28-96
 VertRefresh 43-60
 Modeline "1280x800@60" 83.91 1280 1312 1624 1656 800 816 824 841
EndSection

Section "Screen"
 Identifier "Default Screen"
 Device "Intel Corporation 82845G/GL[Brookdale-G]/GE Chipset Integrated Graphics Device"
 Monitor "Generic Monitor"
 DefaultDepth 24
 SubSection "Display"
  Depth 1
  Modes "1920x1440" "1920x1200" "1856x1392" "1792x1344" "1680x1050" "1600x1200" "1440x900" "1400x1050" "1280x1024" "1280x960" "1280x854" "1280x800" "1280x768" "1200x800" "1152x864" "1152x768" "1024x768" "800x600" "640x480"
 EndSubSection
 SubSection "Display"
  Depth 4
  Modes "1920x1440" "1920x1200" "1856x1392" "1792x1344" "1680x1050" "1600x1200" "1440x900" "1400x1050" "1280x1024" "1280x960" "1280x854" "1280x800" "1280x768" "1200x800" "1152x864" "1152x768" "1024x768" "800x600" "640x480"
 EndSubSection
 SubSection "Display"
  Depth 8
  Modes "1920x1440" "1920x1200" "1856x1392" "1792x1344" "1680x1050" "1600x1200" "1440x900" "1400x1050" "1280x1024" "1280x960" "1280x854" "1280x800" "1280x768" "1200x800" "1152x864" "1152x768" "1024x768" "800x600" "640x480"
 EndSubSection
 SubSection "Display"
  Depth 15
  Modes "1920x1440" "1920x1200" "1856x1392" "1792x1344" "1680x1050" "1600x1200" "1440x900" "1400x1050" "1280x1024" "1280x960" "1280x854" "1280x800" "1280x768" "12...

meter.ten:

>I try to fix it use the patch of
> https://bugs.freedesktop.org/show_bug.cgi?id=6751
> But it doesn't work on Dapper 6.06.

Did you try it against the Xorg sources that you can download for Dapper 6.06? Or did you download the X.org CVS HEAD version from freedesktop.org? The patch will only work against HEAD - instructions on getting and installing are here http://wiki.x.org/wiki/ModularDevelopersGuide

I'm also running Dapper and am downloading HEAD now - will *try* to get time to get it built and keep everyone posted of the results.

Addendum to my last post - just checked and the patches are already in the CVS HEAD version of X.org. So, no need to patch - just follow the above install instructions and go:

To quickly satisfy the build requirements from the Dapper repos, you might want to:

sudo apt-get install build-essential libttf2 automake1.7 libtool linux-kernel-headers autotools-dev autoconf m4 cvs make dpkg-dev

After that you should have satisfied all requirements to build it as per the instructions at the above Wiki.

I can confirm that the new version of X.org in CVS no longer exhibits this behaviour. I managed to get X11 working with dualhead & Xinerama by downloading and compiling all of its modules from CVS & Git.

I have an Intel i810, which is notoriously difficult to get working in dualhead and *I* managed it, so I think that means everyone else should have a very good chance of getting it working with their own setup.

I was able to build X.org CVS in my homedir and after a lot of messing around with include paths and library paths finally managed to get everything built - frustrating, but a real learning experience!

However, I now have 2 installs of X.org on my PC and the source tree is nearly 2 GB in size so I think I'll reformat and do a server install and build X.org in the usual place with the usual prefix of /usr.

And I now know that compiling X.org, libdrm and Mesa is a *very* fiddly process (dependencies, and because bits of it are in Git & other bits in CVS).

Does anyone know if there will be another stable release of X.org before Edgy Eft is released? If not, I strongly recommend that we backport the small patch (see freedesktop bugzilla 6751) to the Dapper and/or Edgy X.org packages so that dualhead is not broken. It is really bad that dualhead is broken for so many users right now - and if it doesn't make it into Edgy we're looking at a 10 month wait for it to be fixed.

I think the patch of freedesk Bug 2597 ,can resolve the bug of dapper 6.06.
 https://bugs.freedesktop.org/show_bug.cgi?id=2597

 I download the src use "apt-get source xserver-xorg-core" , patch it as bug 2597.
Now my dualhead is ok.

 deb-src is more easy to build than CVS from X.org.

Hi Meter,

would you mind sending me a link to the exact patch you used and explaining how
you did it? I looked at the patch on Bug 2597 but it seemed to be patching
files that were in directories that don't exist in the source tree of package
xserver-xorg-core.

However, I'm pretty new to using apt-get source though so it's probably just me
:-)

Did you manage to patch your main installed version of X.org? Perhaps we
should together build a new package that people can use until Ubuntu Xorg team
gets around to releasing the fix?

--- "meter.ten" <email address hidden> wrote:

> I think the patch of freedesk Bug 2597 ,can resolve the bug of dapper 6.06.
> https://bugs.freedesktop.org/show_bug.cgi?id=2597
>
> I download the src use "apt-get source xserver-xorg-core" , patch it as bug
> 2597.
> Now my dualhead is ok.
>
> deb-src is more easy to build than CVS from X.org.
>
> --
> XOrg dualhead regression
> https://launchpad.net/bugs/36461
>

--
"We do not inherit the earth
  from our ancestors; ...-- __@
 we borrow it from our children" ....--- _\<,_
 -- Antoine de St. Exupery (_) (_)

Send instant messages to your online friends http://uk.messenger.yahoo.com

You will need the source package xserver-xorg-driver-ati in addition and split the patch into two different patches. The first diff is for xserver-xorg-driver-ati (src/radeon_driver.c) and the following diffs in the patch for xserver-xorg-core.

You can find the corresponding and recompiled packages for x86 on http://www.szczepanek.de/xinerama/

They do not work for me as I cannot extract the rom image from mit ATI X300 card as mentioned in the #2597 bug report.

Don't forget to add

Option "BiosLocation" "file:/root/test.rom"

to your xorg.conf with /root/test.rom containing your graphic card rom.

I have the same problem.

Dual head worked fine under 5.10 (with some tweaking)
AGP ATI RAGE, and VGA output of PCI radeon r100

Only one screen (AGP) works (snappy fast though, yay ubuntu)

let me know what I need to post to clarify.

This should fix:
Adjusted for X.Org 6.9/7.0 by Stefan Dirsch
https://bugzilla.novell.com/show_bug.cgi?id=171453&x=0&y=0

Can somebody please give a timeframe for the following:

1) when will X.org 7.2 (in which this X.org bug is fixed) reach Edgy?

2) will a patch be rolled out to existing Dapper users via APT or will we have to tell people to upgrade to Edgy to use their dualheads again? if so, when?

* Ashley

| Can somebody please give a timeframe for the following:
|
| 1) when will X.org 7.2 (in which this X.org bug is fixed) reach Edgy?

Given that Xorg 7.2 isn't scheduled until October/November, it
probably won't. We might be able to backport the fix, though.

| 2) will a patch be rolled out to existing Dapper users via APT or will
| we have to tell people to upgrade to Edgy to use their dualheads again?
| if so, when?

Dualheads work fine, just not this particular configuration. If it's
a small and well-contained fix, we should be able to backport it to
dapper-updates. I am not sure of the time-frame as we are all
insanely busy with Edgy at the moment and trying to get Ubuntu's and
Debian's X packages in sync. Once that happens, we might be able to
isolate the fix and apply it in Dapper.

--
Tollef Fog Heen ,''`.
UNIX is user friendly, it's just picky about who its friends are : :' :
                                                                      `. `'
                                                                        `-

--- Tollef Fog Heen <email address hidden> wrote:

> | Can somebody please give a timeframe for the following:
> |
> | 1) when will X.org 7.2 (in which this X.org bug is fixed) reach Edgy?
>
> Given that Xorg 7.2 isn't scheduled until October/November, it
> probably won't. We might be able to backport the fix, though.

OK, to my knowledge several other distros have backported this fix (Novell for
one) - this would be ideal if 7.2 won't release until Edgy+1.

> | 2) will a patch be rolled out to existing Dapper users via APT or will
> | we have to tell people to upgrade to Edgy to use their dualheads again?
> | if so, when?
>
> Dualheads work fine, just not this particular configuration. If it's
> a small and well-contained fix, we should be able to backport it to
> dapper-updates. I am not sure of the time-frame as we are all
> insanely busy with Edgy at the moment and trying to get Ubuntu's and
> Debian's X packages in sync. Once that happens, we might be able to
> isolate the fix and apply it in Dapper.

That would be greatly appreciated. It would appear that a great many people
are affected - in fact it sounds like most people who have onboard VGA and/or 2
cards from different vendors encounter this problem. ATI appears to be
especially problematic.

I already offered my services to the xserver-xorg maintainers to see if there
was some way I might be able to help but received no response.

Thanks for investigating.

Cheers
Ashley

--
"We do not inherit the earth
  from our ancestors; ...-- __@
 we borrow it from our children" ....--- _\<,_
 -- Antoine de St. Exupery (_) (_)
                                      http://backtobreath.com/

___________________________________________________________
The all-new Yahoo! Mail goes wherever you go - free your email address from your Internet provider. http://uk.docs.yahoo.com/nowyoucan.html

Yep, I have the same problem using an onboard i810 and an SiS card - I'm eagerly awaiting a fix cos I'm using Dapper at work and programming is much more convenient with dual-head.

I already look a bit stupid having recommended Ubuntu to all my colleagues and then been seen with a broken dual-head config - it'd be a great shame if I have to switch back to my previous distro in order to get both screens working :(

I have to chime in on this as well. It seems that the hardware configurations that will trigger this bug (any dual-head config using more than one greaphics card, as I understand) is not uncommon. As Dapper is the LTS release, it seems this is not a bug that should be left to bite users for the next five years!

I've been having a problem with my duel-head configuration since my upgrade to Dapper.

I'm not sure if it is related to this issue.

I was able to work around it by selecting the 2.6.12-10-386 kernel at boot.
X is broken for me with 2.6.15-25-386 and 2.6.15-26-386.

I'm hoping this will help others, let me know if I should supply more info.

Thanks.
- Roger

This is affecting me as well. Booting the breezy kernel 2.6.12-10-386 gets it working again.

My cards:
0000:01:00.0 VGA compatible controller: nVidia Corporation NV6 [Vanta/Vanta LT] (rev 15)
0000:02:09.0 VGA compatible controller: Silicon Integrated Systems [SiS] 300/305 PCI/AGP VGA Display Adapter (rev 90)

Pertinant parts of X log:
(II) Reloading /usr/lib/xorg/modules/libint10.so
Requesting insufficient memory window!: start: 0xf0000000 end: 0xf7ffffff size 0xfe010000
Requesting insufficient memory window!: start: 0xfe000000 end: 0xfe2fffff size 0xfe010000
Requesting insufficient memory window!: start: 0xf0000000 end: 0xf7ffffff size 0xfe010000
Requesting insufficient memory window!: start: 0xfe000000 end: 0xfe2fffff size 0xfe010000
Requesting insufficient memory window!: start: 0xf0000000 end: 0xf7ffffff size 0xfe010000
Requesting insufficient memory window!: start: 0xfe000000 end: 0xfe2fffff size 0xfe010000
(EE) SIS(0): Cannot read V_BIOS
(--) SIS(0): Relocated I/O registers at 0x1000
Requesting insufficient memory window!: start: 0xf0000000 end: 0xf7ffffff size 0xfe010000
Requesting insufficient memory window!: start: 0xfe000000 end: 0xfe2fffff size 0xfe010000
Requesting insufficient memory window!: start: 0xf0000000 end: 0xf7ffffff size 0xfe010000
Requesting insufficient memory window!: start: 0xfe000000 end: 0xfe2fffff size 0xfe010000
(WW) SIS(0): Could not find/read video BIOS
(EE) SIS(0): **************************************************
(EE) SIS(0): ERROR:
(EE) SIS(0): Virtual screen too big for memory; 937K needed, 508K available
(EE) SIS(0): END OF MESSAGE
(EE) SIS(0): **************************************************

This bug is hitting me too. I have i810 integrated and PCI ATI card. Heard great things about Ubuntu so it's a bummer to see this bug... i really need dual head. Once you have it, you can't go back hahaha.

Any update on a fix? Gonna have to try Suse for a while until a fix for this hits dapper. Thanks.

I applied the dpatch by Marcelo and rebuilt the xorg-server package, and what do you know, i got dual-head back. Thanks Marcelo, now i can give kubuntu a whirl. If anyone wants the debs i built, look here. Built on kubuntu dapper.

http://www.r-a-y.org/ubuntu/

A recent update of X.org fixed the problem in Edgy Eft (a.k.a Ubuntu 6.10 a.k.a. Ubuntu Unstable a.k.a. the Development release). I'd been using Edgy for a few weeks and only today tried to get it working again and to my surprise it just worked (had to set up Xinerama & whatnot of course).

Disclaimer: I am not advocating that you run out and upgrade to Edgy Eft... there are lots and lots of updates (it is synced to Debian Unstable until a 'freeze date' I believe), and so things are liable to break after updates.

To the Dapper X.org maintainers, there are 2 ways of fixing this I guess:

2. Patch source of existing xserver-xorg* packages as others have here (or even grab their package, check it's OK and use it!)

1. Push the packages from Edgy into the 6.06 repos (not sure if any dependencies on other, newer packages that are only in Edgy would be a problem though)

Actually, I spoke to soon about the problem having been fixed in Edgy. It seems I can get my dualhead working, but only by installing Ray's package (from his post directly above my last post) then attempting to restart X (which fails but allows my ATI card to initialise its BIOS), then re-upgrading to the Edgy xserver-xorg-core package. So I guess the folks looking after Debian Unstable haven't fixed this yet either...

Yes i was wondering... cause i upgraded to Edgy first to see if it was fixed there and it wasn't. So i reinstalled Dapper and built the xorg-server package with Marcelo's dpatch. Dual head has been working great for me ever since.

An update for xserver-xorg-core in Dapper came out yesterday, but i saw no mention of this bug. Fabio M. Di Nitto is listed as the maintainer... I wonder if he's following this thread haha :)

I too have been having this problem with my dual head setup here at work. I just installed Ray's binaries, and I can confirm that they work great.

Great job Ray. Thanks for building those binaries for us. Now all we need is to get the patch you used into an official update.

Im using Ray's binary and it's working great now, when this patch will be applied in xorg of dapper???

Fixed on dapper-updates release of xorg-server_1.0.2-0ubuntu10.3.

Thanks for the hard work, guys!

(just for reference, this bug is not present in the current version of xorg-server, present on Ubuntu Edgy)

I've approved Rodrigo's upload to dapper-updates.

The xserver-xorg-core update just broke X for me. I have a motherboard with the Intel 845 chipset and a Radeon 9000 Pro AGP video card. After the update, X does not detected the video card. Running Xorg -scanpci does not seem to list the card either. Let me know if I can provide more information.

yup, broke here too, no screens found - even if trying vga or vesa driver.

Dell Inspiron 9400 with ATI x1400.

This appears to be breaking a *LOT* of people's systems - see http://www.ubuntuforums.org/showthread.php?t=241286 for details.

Hey, the same for me guys: I have a MAtrox G450 and x server is failing with the new xorg core updates...

I spent A LOT of time verifyng my drivers, suspected a recent local update of mga_hal but it seems the delivery is REGRESSIVE this time ... isn't it?

Guys *please* do not spam up this bugreport with reports about the recent XOrg package snafu. This is a completely different issue which has nothing to do with the recent issue with XOrg. This bug is for dual head configurations that broke with the move to dapper.

Thanks,
Jay