[RV200] ATI graphics corruption with compiz on XAA

Binary package hint: xserver-xorg-video-ati

I am seeing major graphical corruption when compiz is enabled on my T30 (ATI graphics). Some windows have no window manager decoration at all, and some windows have only window manager decoration and no contents. I will attach a photo showing an example. When compiz is disabled normality returns.

ProblemType: Bug
Architecture: i386
Date: Mon Sep 14 10:24:16 2009
DistroRelease: Ubuntu 9.10
Lsusb:
 Bus 002 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
 Bus 003 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
 Bus 001 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
MachineType: IBM 2366EG9
Package: xserver-xorg-video-ati 1:6.12.99+git20090825.fc74e119-0ubuntu1
PccardctlIdent:
 Socket 0:
   no product info available
 Socket 1:
   no product info available
PccardctlStatus:
 Socket 0:
   no card
 Socket 1:
   no card
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.31-10-generic root=UUID=a496a7f2-ff19-4c5e-8429-539a2abe365d ro quiet
ProcEnviron:
 LANG=en_GB.UTF-8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.31-10.32-generic
RelatedPackageVersions:
 xserver-xorg 1:7.4+3ubuntu5
 libgl1-mesa-glx 7.6.0~git20090817.7c422387-0ubuntu3
 libdrm2 2.4.13-1ubuntu1
 xserver-xorg-video-intel 2:2.8.1-1ubuntu1
 xserver-xorg-video-ati 1:6.12.99+git20090825.fc74e119-0ubuntu1
SourcePackage: xserver-xorg-video-ati
Uname: Linux 2.6.31-10-generic i686
dmi.bios.date: 06/11/2004
dmi.bios.vendor: IBM
dmi.bios.version: 1IET69WW (2.08 )
dmi.board.name: 2366EG9
dmi.board.vendor: IBM
dmi.board.version: Not Available
dmi.chassis.asset.tag: No Asset Information
dmi.chassis.type: 10
dmi.chassis.vendor: IBM
dmi.chassis.version: Not Available
dmi.modalias: dmi:bvnIBM:bvr1IET69WW(2.08):bd06/11/2004:svnIBM:pn2366EG9:pvrNotAvailable:rvnIBM:rn2366EG9:rvrNotAvailable:cvnIBM:ct10:cvrNotAvailable:
dmi.product.name: 2366EG9
dmi.product.version: Not Available
dmi.sys.vendor: IBM
fglrx: Not loaded
system:
 distro: Ubuntu
 architecture: i686kernel: 2.6.31-10-generic

Attaching an example of the corruption. The screen consists of the compiz appearance control middle, and an xterm bottom right.

Bryce, I have added your PPA to facilitate testing/collaboration. Also, do you recommend that I create an xorg.conf file?

See also bug 426582.

I have a Radeon M6 LY, R100. specifying EXA in my device seems to have worked around the screen corruption issues I've seen. If it works for others, can we add a quirk to the xorg.conf configuration logic that will specify EXA for affected cards?

EXA doesn't fix my issue.

lspci output:
"VGA compatible controller: ATI Technologies Inc Radeon Mobility M6 LY"

My manually created /etc/X11/xorg.conf follows:

Section "Device"
  Identifier "ATI Radeon"
  Driver "ati"
  Option "AccelMethod" "EXA"
  Option "MigrationHeuristic" "greedy"
  Option "AccelDFS" "true"
  Option "EnablePageFlip" "true"
  Option "EnableDepthMoves" "true"
EndSection

By default, there was no xorg.conf file in 9.10 Karmic Koala. I copied it over from a stable and working 9.04 Jaunty Jackalope installation.

Sadly, a regression...

Here's my Device section:

Section "Device"
       Identifier "ATI"
       Driver "radeon"
       Option "AccelMethod" "EXA"
       Option "AGPMode" "4"
# Option "AGPFastWrite" "yes"
# Option "EnablePageFlip" "on"
# Option "RenderAccel" "on"
       Option "DynamicClocks" "on"
EndSection

No, I do not use compiz.

"ATI Technologies Inc Radeon Mobility M6 LY" still broken for me in 9.10 (Karmic). No Desktop Effects possible.

Here is my xorg.conf for:
lspci output:
"VGA compatible controller: ATI Technologies Inc Radeon Mobility M6 LY"

This is now nice and snappy with window rendering and all the corruption issues have resolved. Scrolling in Firefox is also fine. If I try to enable compiz, the whole system freezes, but since I don't bother with it on this old 1 GHz celeron notebook, I don't miss it.

Section "Device"
        Identifier "ATI Technologies Inc Radeon Mobility M6 LY"
        Driver "ati"
        BusID "PCI:1:0:0"
        Option "AGPMode" "4"
        Option "AGPSize" "64" # default: 8
        Option "RingSize" "8"
        Option "BufferSize" "2"
        Option "EnablePageFlip" "True"
        Option "EnableDepthMoves" "True"
 Option "SWcursor" "off" #Faster than default (on)
 Option "AccelMethod" "EXA"
 Option "MigrationHeuristic" "greedy"
 Option "AccelDFS" "true"
 Option "BIOSHotkeys" "on"
        Option "RenderAccel" "true" # Enables hardware acceleration
        Option "DynamicClocks" "on" # Adds clock scalability / power management for the video card
EndSection

Still exists in Lucid alpha 1. For those who haven't seen it yet, bug #426582 is also related.

Lucid alpha 2 and 3 do not show corruption on what lspci identifies as "Radeon Mobility M7 LW [Radeon Mobility 7500]" (ThinkPad T30). However, checking X.0.log shows that the X server is using EXA, not XAA. I was unable to force XAA using an xorg.conf. Bug #513956 suggests that XAA still exists, however, and is being used for some cards.

Good news. Lucid Lynx Alpha 3 does not exhibit this behaviour on the Radeon Mobility M6 LY
(Neither does Alpha1 or Alpha2).

Compiz works! No corruption. Something must have been fixed between 9.10 and 10.04.

lspci output:
"VGA compatible controller: ATI Technologies Inc Radeon Mobility M6 LY"

Thanks for letting us know the issue is resolved.

On Sun, Mar 7, 2010 at 12:19 PM, Randall Ross (rrnwexec)
<email address hidden> wrote:
> Good news. Lucid Lynx Alpha 3 does not exhibit this behaviour on the Radeon Mobility M6 LY
> (Neither does Alpha1 or Alpha2).

Randall, can you confirm that you have no corruption using XAA? (Check
/var/log/Xorg.0.log for mention of either XAA or EXA.) If EXA is being
used by default now, that's a fine workaround (and may be worth
turning this into a WONTFIX), but it's not a fix.

Thanks.

Confirmed that EXA is in use. (Lucid Lynx B2)
Excerpts from /var/log/Xorg.0.log below. Full log in attachment.

excerpt from Xorg.0.log:
(II) RADEON(0): EXA: Driver will not allow EXA pixmaps in VRAM
(II) Loading sub module "exa"
(II) LoadModule: "exa"
(II) Loading /usr/lib/xorg/modules/libexa.so
(II) Module exa: vendor="X.Org Foundation"
 compiled for 1.7.6, module version = 2.5.0
 ABI class: X.Org Video Driver, version 6.0
(II) RADEON(0): Setting EXA maxPitchBytes
(II) EXA(0): Driver allocated offscreen pixmaps
(II) EXA(0): Driver registered support for the following operations:
(II) Solid
(II) Copy
(II) Composite (RENDER acceleration)
(II) UploadToScreen
(II) DownloadFromScreen
(II) RADEON(0): Acceleration enabled

Using EXA is nowhere near a proper fix. To see what I mean, open FireFox and attempt to scroll a decent sized page; the CPU will spike to 100% usage and FireFox will drop to a crawl.

A few of us from ArchLinux are having this same issue, with the card "ATI Technologies Inc Radeon Mobility M6 LY". It is *not* related to Compiz in anyway, my notebook runs a small/lightweight wm called awesome, and has never seen compiz.

For us, downgrading to xf86-video-ati-6.12.4-3 and using XAA is the only "proper" fix. See our thread about it here: http://bbs.archlinux.org/viewtopic.php?id=94960

I agree with Gary13579 that "use EXA" is not a fit for "XAA doesn't work". Upstream has decided XAA is a WONTFIX (http://bugs.freedesktop.org/show_bug.cgi?id=22055), so I suggest someone from Bug Control mark this (and bug #426582) as WONTFIX as well.

EXA performance problems should probably be another bug. There are several which *may* be relevant; those having problems may wish to look at bug #394402 and bug #405655.

I wonder if it's possible to ship both xf86-video-ati-6.12.4-3 and current 6.13.0-1, and then use probe to see if the graphics card is on a blacklist-with-xf86-video-ati-6.13.0 list? If it matches, the load 6.12.4-3 instead. Alternatively, maybe we could link Jockey into the update-alternatives framework, jockey detects the blacklisted card, and notifies the user. The user clicks a few buttons, and *bam*, Ubuntu 10.04 finally starts to deliver some LTS quality.

On the other hand, shouldn't we be able to easily inherit Debian Squeeze's fix for this, if it exists? Would rebuilding Debian's 6.13.1 package suffice?
http://packages.debian.org/squeeze/xserver-xorg-video-ati