[ATI R6xx and above] X fails to start up after mesa+git update - gdm loops after logging in

Latest upgrades attached.
Could the issue be with newest mesa/libdrm ?

The correct aptitude log file.

Last week I added the xorg-edgers PPA, this installed a new package: libdrm-radeon1 and I experienced this bug.

Downgrading the packages and removing the PPA fixed it. My guess is that some changes from xorg-edgers have filtered down to vanilla Karmic and the bug has come down with it. I haven't updated my Karmic install yet, but I think it quite likely when I do I'll hit this bug again. I'll report back after updating. My GPU is an ATI HD3200.

Happened to me yesterday when I enabled the xorg-edgers PPA. Happened again when I did a clean re-install of Karmic today.

From comments, sounds likely to have broken due to the new mesa, which includes the radeon-rewrite branch for -ati.

Please run the command 'apport-collect 419126'

Architecture: amd64
DistroRelease: Ubuntu 9.10
MachineType: Hewlett-Packard HP Pavilion dv5 Notebook PC
Package: xserver-xorg-video-ati 1:6.12.99+git20090629.f39cafc5-0ubuntu6
PackageArchitecture: amd64
PccardctlIdent:

PccardctlStatus:

ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.31-6-generic root=UUID=57a680cf-a8c4-44d6-8406-f0261fe03570 ro quiet splash
ProcEnviron:
 SHELL=/bin/bash
 LANG=en_US.UTF-8
ProcVersionSignature: Ubuntu 2.6.31-6.26-generic
RelatedPackageVersions:
 xserver-xorg 1:7.4+3ubuntu5
 libgl1-mesa-glx 7.6.0~git20090817.7c422387-0ubuntu2
 libdrm2 2.4.12+git20090801.45078630-0ubuntu1
 xserver-xorg-video-intel 2:2.8.0-0ubuntu2
 xserver-xorg-video-ati 1:6.12.99+git20090629.f39cafc5-0ubuntu6
Uname: Linux 2.6.31-6-generic x86_64
UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare
XorgConf: Error: [Errno 2] No such file or directory: '/etc/X11/xorg.conf'
dmi.bios.date: 07/06/2009
dmi.bios.vendor: Hewlett-Packard
dmi.bios.version: F.35
dmi.board.asset.tag: Base Board Asset Tag
dmi.board.name: 3600
dmi.board.vendor: Quanta
dmi.board.version: 98.35
dmi.chassis.type: 10
dmi.chassis.vendor: Quanta
dmi.chassis.version: N/A
dmi.modalias: dmi:bvnHewlett-Packard:bvrF.35:bd07/06/2009:svnHewlett-Packard:pnHPPaviliondv5NotebookPC:pvrRev1:rvnQuanta:rn3600:rvr98.35:cvnQuanta:ct10:cvrN/A:
dmi.product.name: HP Pavilion dv5 Notebook PC
dmi.product.version: Rev 1
dmi.sys.vendor: Hewlett-Packard
fglrx: Not loaded
system:
 distro: Ubuntu
 architecture: x86_64kernel: 2.6.31-6-generic

Apparently was running with previous kernel (rc6), due to hangdown bug with rc7. but running kernel makes no difference.

Hi Tuomas, thanks for that. Could you also attach your gdm log files (in /var/log/gdm/)? It seems the apport collector failed to collect those, which should reveal why X isn't starting.

@Kazade, it's interesting that you saw this against xorg-edgers. For future reference, we provide xorg-edgers for testing purposes, so it is a good idea if you report bugs you find there, so we have a chance at dealing with them (or even skip doing the update) before the bugs get into the release. (I know reporting bugs can be a pain but I did go through mesa and xorg-edgers bug reports with a fine toothed comb before doing this update, so it would have definitely been of benefit in this case.)

Fwiw, I am planning on updating us to the -ati git snapshot such as the one in xorg-edgers currently. It would be helpful if you guys could confirm if this bug still occurs when running with that git snapshot. I'm guessing it will, but if you can confirm it, then I can get that bug reported upstream (and maybe try chasing down a patch, although time is short) before moving Karmic to this new -ati.

Here's few of the logs, will attach more if needed?
permissions seems kind of sneaky to me?

Bryce Harrington wrote:
"It seems the apport collector failed to collect those"

Tuomas Aavikko wrote:
"permissions seems kind of sneaky to me?"

The /var/log/gdm permissions are
drwxrwx--T root gdm

@Bryce
Were you meaning the -ati package in this ppa?

https://launchpad.net/~xorg-edgers/+archive/drivers-only

If the only steps required were:
-add above ppa to sources.list
-apt-get update
-apt-get install xserver-xorg-video-ati
-restart,
then the issue is not resolved in the newer version. If I missed any steps, please let me know.

2009/8/26 Bryce Harrington <email address hidden>

>
> Fwiw, I am planning on updating us to the -ati git snapshot such as the
> one in xorg-edgers currently. It would be helpful if you guys could
> confirm if this bug still occurs when running with that git snapshot.
> I'm guessing it will, but if you can confirm it, then I can get that bug
> reported upstream (and maybe try chasing down a patch, although time is
> short) before moving Karmic to this new -ati.
>

The one in xorg-edgers/drivers-only *ati-fc74e119 doesn't fix this.

@bryce, I'm really sorry I didn't report it sooner. I'd been fiddling with a bunch of other stuff (compiling from git etc.) so I figured it was just me. Soon after I managed to compile the 3D R600 branch and I've been using that since. It was only when I saw the comment in the Karmic forums that I realized it wasn't something I'd done.

On Wed, Aug 26, 2009 at 06:14:14PM -0000, Kiwinote wrote:
> @Bryce
> Were you meaning the -ati package in this ppa?
>
> https://launchpad.net/~xorg-edgers/+archive/drivers-only

Correct

Shouldn't this bug be with the mesa package rather than the ati package?

Using xorg-edgers, fully up-to-date, GDM looping is still an issue.
With or without nomodeset, only way to login, is to disable DRI in
xorg.conf

Using xorg-edgers PPA, GDM loop is still an issue.
With or without nomodeset, only way to login is to disable DRI in
xorg.conf.

Works for me after some iterations. I needed
https://launchpad.net/~xorg-edgers/+archive/ppa
https://launchpad.net/~xorg-edgers/+archive/radeon
sudo apt-get install drm-modules-source
sudo module-assistant -t auto-install drm-modules
sudo update-initramfs -u

Glxgears and compiz work, but there are some graphical glitches with menus, buttons and typing text. Also busy windows become solid grey boxes. I'll see if I can catch all that in screen shots and file bugs (radeon HD4670).

When I delete the /usr/lib/dri/r600_dri.so file from the stock Alpha-5 libgl1-mesa-dri package (7.6.0~git20090817.7c422387-0ubuntu3) then X will start up correctly with DRI enabled using both the Radeon and Radeon HD drivers.

AIGLX reverts to software rendering as expected.

Acer Aspire 5536 with a Radeon HD 3200 card.

2009/9/6 taavikko <email address hidden>

> Using xorg-edgers PPA, GDM loop is still an issue.
> With or without nomodeset, only way to login is to disable DRI in
> xorg.conf.
>

Sorry for double post (evolution mixup)
Was wrong, even disabling DRI in xorg.conf will result in GDM loop.
previous logs should suffice.

Tested the Xorg edgers system.

With the xserver-xorg-video-ati driver, the microcode works and fixes the drmCommandBuffer: -22 fault but I can't see what I am typing in the terminal window so there is a rendering flaw there.

With the xserver-xorg-video-radeonhd driver, everything seems fine including the rendering of typing in a terminal window.

radeonhd seems somewhat slower at rendering glxgears than radeon. Both drivers report "IRQ's not enabled, falling back to busy waits: 2 18" while running the program and the gears image always stays in front - overwriting any window you move over it.

Progress from my point of view.

Note that I was following the instructions in #109. I updated DRM as well and it is that which fixes the X server crash issue.

[QUOTE]Glxgears and compiz work, but there are some graphical glitches with menus, buttons and typing text. Also busy windows become solid grey boxes. I'll see if I can catch all that in screen shots and file bugs[/QUOTE]

[QUOTE]
radeonhd seems somewhat slower at rendering glxgears than radeon. Both drivers report "IRQ's not enabled, falling back to busy waits: 2 18" while running the program[/QUOTE]

The upstream developers already know about a lot of these issues. Graphical glitches for R600/700 3D are expected right now. See: http://www.x.org/wiki/radeonhd%3Aexperimental_3D The key word is EXPERIMENTAL ;) I'm not sure how much more the code will stabilize before the Karmic code freeze. I'm guessing that Ubuntu will configure mesa with software rendering on R600/700 if they want things to be stable, even though it's slow.

I encountered what appears to be this bug with todays daily-live disc on a hp pavilion tx2500 series tablet. could not generate log files as I couldn't get it to boot and can't install karmic on the machine right now.

Architecture: i386
DistroRelease: Ubuntu 9.10
MachineType: Hewlett-Packard HP Compaq 6830s
Package: xserver-xorg-video-ati 1:6.12.99+git20090629.f39cafc5-0ubuntu5
PackageArchitecture: i386
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.31-5-generic root=UUID=8618aa23-bc13-4875-b70e-3dcfa3cf0261 ro quiet splash
ProcEnviron:
 SHELL=/bin/bash
 LANG=sv_FI.UTF-8
ProcVersionSignature: Ubuntu 2.6.31-5.24-generic
RelatedPackageVersions:
 xserver-xorg 1:7.4+3ubuntu5
 libgl1-mesa-glx 7.5-1ubuntu1
 libdrm2 2.4.12-1ubuntu1
 xserver-xorg-video-intel 2:2.8.0-0ubuntu2
 xserver-xorg-video-ati 1:6.12.99+git20090629.f39cafc5-0ubuntu5
Uname: Linux 2.6.31-5-generic i686
UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare
XorgConf: Error: [Errno 2] No such file or directory: '/etc/X11/xorg.conf'
dmi.bios.date: 03/12/2009
dmi.bios.vendor: Hewlett-Packard
dmi.bios.version: 68PZD Ver. F.0C
dmi.board.name: 30E9
dmi.board.vendor: Hewlett-Packard
dmi.board.version: KBC Version 95.1C
dmi.chassis.asset.tag: CNU9202RW6
dmi.chassis.type: 10
dmi.chassis.vendor: Hewlett-Packard
dmi.modalias: dmi:bvnHewlett-Packard:bvr68PZDVer.F.0C:bd03/12/2009:svnHewlett-Packard:pnHPCompaq6830s:pvrF.0C:rvnHewlett-Packard:rn30E9:rvrKBCVersion95.1C:cvnHewlett-Packard:ct10:cvr:
dmi.product.name: HP Compaq 6830s
dmi.product.version: F.0C
dmi.sys.vendor: Hewlett-Packard
fglrx: Not loaded
system:
 distro: Ubuntu
 architecture: i686kernel: 2.6.31-5-generic

Architecture: i386
DistroRelease: Ubuntu 9.10
MachineType: Hewlett-Packard HP Compaq 6830s
Package: xserver-xorg-video-ati 1:6.12.99+git20090629.f39cafc5-0ubuntu5
PackageArchitecture: i386
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.31-5-generic root=UUID=8618aa23-bc13-4875-b70e-3dcfa3cf0261 ro quiet splash
ProcEnviron:
 PATH=(custom, no user)
 LANG=sv_FI.UTF-8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.31-5.24-generic
RelatedPackageVersions:
 xserver-xorg 1:7.4+3ubuntu5
 libgl1-mesa-glx 7.5-1ubuntu1
 libdrm2 2.4.12-1ubuntu1
 xserver-xorg-video-intel 2:2.8.0-0ubuntu2
 xserver-xorg-video-ati 1:6.12.99+git20090629.f39cafc5-0ubuntu5
Uname: Linux 2.6.31-5-generic i686
UserGroups:

XorgConf: Error: [Errno 2] No such file or directory: '/etc/X11/xorg.conf'
dmi.bios.date: 03/12/2009
dmi.bios.vendor: Hewlett-Packard
dmi.bios.version: 68PZD Ver. F.0C
dmi.board.name: 30E9
dmi.board.vendor: Hewlett-Packard
dmi.board.version: KBC Version 95.1C
dmi.chassis.asset.tag: CNU9202RW6
dmi.chassis.type: 10
dmi.chassis.vendor: Hewlett-Packard
dmi.modalias: dmi:bvnHewlett-Packard:bvr68PZDVer.F.0C:bd03/12/2009:svnHewlett-Packard:pnHPCompaq6830s:pvrF.0C:rvnHewlett-Packard:rn30E9:rvrKBCVersion95.1C:cvnHewlett-Packard:ct10:cvr:
dmi.product.name: HP Compaq 6830s
dmi.product.version: F.0C
dmi.sys.vendor: Hewlett-Packard
fglrx: Not loaded
system:
 distro: Ubuntu
 architecture: i686kernel: 2.6.31-5-generic

Architecture: i386
DistroRelease: Ubuntu 9.10
MachineType: Hewlett-Packard HP Compaq 6830s
Package: xserver-xorg-video-ati 1:6.12.99+git20090629.f39cafc5-0ubuntu5
PackageArchitecture: i386
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.31-5-generic root=UUID=8618aa23-bc13-4875-b70e-3dcfa3cf0261 ro quiet splash
ProcEnviron:
 SHELL=/bin/bash
 LANG=sv_FI.UTF-8
ProcVersionSignature: Ubuntu 2.6.31-5.24-generic
RelatedPackageVersions:
 xserver-xorg 1:7.4+3ubuntu5
 libgl1-mesa-glx 7.5-1ubuntu1
 libdrm2 2.4.12-1ubuntu1
 xserver-xorg-video-intel 2:2.8.0-0ubuntu2
 xserver-xorg-video-ati 1:6.12.99+git20090629.f39cafc5-0ubuntu5
Uname: Linux 2.6.31-5-generic i686
UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare
XorgConf: Error: [Errno 2] No such file or directory: '/etc/X11/xorg.conf'
dmi.bios.date: 03/12/2009
dmi.bios.vendor: Hewlett-Packard
dmi.bios.version: 68PZD Ver. F.0C
dmi.board.name: 30E9
dmi.board.vendor: Hewlett-Packard
dmi.board.version: KBC Version 95.1C
dmi.chassis.asset.tag: CNU9202RW6
dmi.chassis.type: 10
dmi.chassis.vendor: Hewlett-Packard
dmi.modalias: dmi:bvnHewlett-Packard:bvr68PZDVer.F.0C:bd03/12/2009:svnHewlett-Packard:pnHPCompaq6830s:pvrF.0C:rvnHewlett-Packard:rn30E9:rvrKBCVersion95.1C:cvnHewlett-Packard:ct10:cvr:
dmi.product.name: HP Compaq 6830s
dmi.product.version: F.0C
dmi.sys.vendor: Hewlett-Packard
fglrx: Not loaded
system:
 distro: Ubuntu
 architecture: i686kernel: 2.6.31-5-generic

Apparently some people with Nvidia cards are encountering this bug as well. there comments are at http://ubuntuforums.org/showthread.php?p=7927173

What's incomplete about the xserver-xorg-video-ati part of this bug?

some people on the forums claim to have encountered this problem while not using xserver-xorg-video-ati driver, or ati hardware for that matter. so you have to prove it is part of the problem and not just present along side the problem.

Would love to know what's incomplete about this so I can help finish it off. Still having this problem with Karmic archive today, still able to fix it by building xorg-edgers + radeon crack.

Hmmm clearing all the xorg.conf and just running apt-get xorg-driver-fglrx and rebooting for some reason seems to fix the problem for me. I'm real linux noob so sorry if this isn't helpful :)

That does only show that the problem seems to be in xserver-xorg-video-ati and people using the proprietary fglrx driver are not affected. Apparently, this does not solve the problem, since everybody with an ATI r6xx-based video card trying to boot from the Live-CD will be bound to the free radeon driver first. At the moment, those people won't be able to install karmic from the Live-CD.

Will xserver-xorg-video-ati/radeon + Mesa 7.6 branch get updated in karmic anytime soon so we can use it again?

AIUI The maintainer is on paternity leave at the moment.

2009/9/15 d2kx <email address hidden>:
> Will xserver-xorg-video-ati/radeon + Mesa 7.6 branch get updated in
> karmic anytime soon so we can use it again?
>
> --
> X fails to start up after mesa+git update - gdm loops after logging in
> https://bugs.launchpad.net/bugs/419126
> You received this bug notification because you are a direct subscriber
> of a duplicate bug.
>

--
Neil Wilson

I tried https://edge.launchpad.net/~albertomilone/+archive/x-testing on a live system (new mesa), and I still have the same problem.

When I remove /usr/lib/dri/r600_dri.so, GNOME session starts up, but with metacity. So that would be a regression compared to Jaunty, where I didn't get 3D acceleration on the r6xx, but at least compositing and Xv worked.

This bug is not only for r6xx chip sets. My HD 3200 has this bug also. As far I know the HD 3200 has the RS780 chip set.

Sorry, I misremembered. Jaunty didn't have compositing on this card, so it's not a regression.

Me to, but the RS780 in the HD 3200 uses the r600 microcode.

2009/9/16 Someone561 <email address hidden>:
> This bug is not only for r6xx chip sets. My HD 3200 has this bug also.
> As far I know the HD 3200 has the RS780 chip set.
>
> --
> [ATI R6xx] X fails to start up after mesa+git update - gdm loops after logging in
> https://bugs.launchpad.net/bugs/419126
> You received this bug notification because you are a direct subscriber
> of a duplicate bug.
>

--
Neil Wilson

@Martin:
Revision ubuntu5+ppa1 should not have r600_dri.so.

@Someone561:
As far as I know, despite its name, r600_dri.so is used for both r600 and r700.

The attached debdiff disables r600

So from the milestone, this won't be fixed until the 1st of october for the beta? That is far too long to keep such a bug around!

In #radeon irc channel we found solution for gentoo users with similar
problems. It was caused by commit a381287759b2b65e7de9fb35801c781cab016f10
(log: "drm/radeon/kms/r600: use blit for BO moves") in drm-next branch of
Airlied's kernele tree.

Reverting solves the problem until blit copy code is fixed.

On Wed, Sep 16, 2009 at 6:53 PM, Alberto Milone <
<email address hidden>> wrote:

> The attached debdiff disables r600
>
> ** Attachment added: "mesa_7.6.0~git20090817.7c422387-0ubuntu5.debdiff"
>
> http://launchpadlibrarian.net/31904627/mesa_7.6.0%7Egit20090817.7c422387-0ubuntu5.debdiff
>
> --
> [ATI R6xx and above] X fails to start up after mesa+git update - gdm loops
> after logging in
> https://bugs.launchpad.net/bugs/419126
> You received this bug notification because you are subscribed to
> xserver-xorg-video-ati in ubuntu.
>

the milestone simply indicates that they hope to have a fix in by the beta release(a development milestone release). It could be in a daily build this weekend, or prove difficult enough to fix that it misses the milestone. Time will tell.. I'm hoping for this weekend.

Fixing is straightforward. Remove the /usr/lib/dri/r600_dri.so file.

2009/9/16 executorvs <email address hidden>:
> the milestone simply indicates that they hope to have a fix in by the
> beta release(a development milestone release). It could be in a daily
> build this weekend, or prove difficult enough to fix that it misses the
> milestone. Time will tell.. I'm hoping for this weekend.
>
> --
> [ATI R6xx and above] X fails to start up after mesa+git update - gdm loops after logging in
> https://bugs.launchpad.net/bugs/419126
> You received this bug notification because you are a direct subscriber
> of a duplicate bug.
>

--
Neil Wilson

I am pretty sure that file actually has a purpose simply removing it i doubt is a sane option for "fixing" the bug. That would be a temporary solution for people until the fix is out (which i am download a daily iso to test to see if that actually works).

Actually, deleting the file will solve this problem, so that people with an r6xx or above will again be able to use karmic. Of course, this cannot be a longtime solution, since the mentioned filed has the purpose to enable DRI for r6xx based video cards (which is still in development and not quite stable as we all can see). However, since this is no regression in comparison to jaunty (were DRI did not work either on r6xx and above), I personally prefer a "working" system without DRI over a non-working system.

Just a clarification on the issue:

Removing the 3D driver (/usr/lib/dri/r600_dri.so) is a valid workaround and the debdiff that I attached in comment 132 prevents mesa from building that driver.

We are still looking into the possibility to get a working 3D driver for r600 in time for Karmic. I can't promise anything though.

On Thu, Sep 17, 2009 at 1:13 PM, Alberto Milone <
<email address hidden>> wrote:

> Just a clarification on the issue:
>
> Removing the 3D driver (/usr/lib/dri/r600_dri.so) is a valid workaround
> and the debdiff that I attached in comment 132 prevents mesa from
> building that driver.
>
> We are still looking into the possibility to get a working 3D driver for
> r600 in time for Karmic. I can't promise anything though.
>
>
You should co-operate more with upstream when you try to get still in
development code to work. It would be nice to know what kernel modules code
you are using and what extra patches etc. People will be coming to complain
to upstream anyway so fixing bugs is easier there if there is more knowledge
what you are doing.

Did you try to revert the patch that hooked in blit to kernel copy
operation? It is known to be buggy in some hw still so better not to ship
that kind of change in Ubuntu until most of problems with blit code is
fixed.

Why disable r600 dri when there is already a capable driver in the xorg-edgers PPAs? To add a datapoint (HD4670), I've been using them for the last 10 days with great success. Compiz works well and is smooth, XVideo is there with the usual 16 textured video adapters, and even some dated 3d games I tried are surprisingly playable (Supertuxcart, wine+Age of Mythology). All without a single freeze or X crash yet.

Of the glitches I described in #109 there is only bug #425303 left, The busy window bug #425334 has been fixed already. I think it would be really sad to have r600 dri disabled at this point. Are there any blockers holding up the transition from xorg-edgers to the repos?

This bug was fixed in the package mesa - 7.6.0~git20090817.7c422387-0ubuntu5

---------------
mesa (7.6.0~git20090817.7c422387-0ubuntu5) karmic; urgency=low

  * debian/rules: Disable r600: Disable r600 as 3D support is still
    experimental and causes problems with the gnome session when
    compiz is enabled (LP: #419126).

 -- Alberto Milone <email address hidden> Wed, 16 Sep 2009 17:43:59 +0200

After discussing with upstream, I have disabled the 3D driver for r600 in Mesa.

We will provide packages with the 3D driver (with the fixes from upstream) in a PPA so that you can try the experimental driver if you want.

Okay thanks for that, maybe for next ubuntu release it will be ready but as you said it's not a step backwards as we didn't have it before.

So i take it the daily cd for the 19th will have this fix right?

I tried the Sept 25 daily image and had the same experience as Chris - plain white Ubuntu logo, then text prompt, and after hitting enter:

Ubuntu karmic (development branch) ubuntu tty1

ubuntu login:

Chris wrote 10 hours ago: #20

I have tried the September 24 snapshot per Robbie Williamson's request. It does make a difference but not in the way I would like to see. The splash screen is now black with the Ubuntu logo (circle only) showing in white in the centre of the screen. After about 15 seconds the screen changes to just a binking underscore on a black background. If I hit the enter key at that point I get a console with 'Ubuntu login:' as prompt.

Can you let me know the number of the original bug report if you want to continue to receive comments there?