[Legacy Radeon cards] Desktop effects/compiz: display brightness fluctuation with zoom, workspace switch or Expo

[Andreas Modinos]

Ubuntu Lucid Lynx, kernel 2.6.32.xx

Zooming in with Super+Scroll causes the screen to become very dark, practically unreadable. Adjusting Fn+Brightness doesn't fix this, although I can tell the brightness changing slightly. The same thing happens when changing workspaces (Ctrl+Alt+-> or Ctrl+Alt+<-), dragging a window from one workspace to another, or using Expo (Super+E) to view all workspaces simultaneously. Returning to normal zoom level will always restore the correct level of brightness immediately, corresponding to how it might have been adjusted during black out. I also noticed that the cursor will stay bright against the dark background.

I originally reported this bug against the kernel because with the mainline kernel there were no issues with brightness (albeit, there are some other issues). Later it was confirmed to be a bug in video drivers through the bug report at https://bugs.freedesktop.org/show_bug.cgi?id=25883

A work around for the bug is to disable the setting 'Lighting' in Compiz general settings under the display tab.

---
AlsaVersion: Advanced Linux Sound Architecture Driver Version 1.0.21.
Architecture: i386
AudioDevicesInUse:
 USER PID ACCESS COMMAND
 /dev/snd/controlC0: andreas 1416 F.... pulseaudio
CRDA: Error: [Errno 2] Tiedostoa tai hakemistoa ei ole
Card0.Amixer.info:
 Card hw:0 'I82801DBICH4'/'Intel 82801DB-ICH4 with STAC9750,51 at irq 7'
   Mixer name : 'SigmaTel STAC9750,51'
   Components : 'AC97a:83847650'
   Controls : 38
   Simple ctrls : 24
DistroRelease: Ubuntu 10.04
HibernationDevice: RESUME=UUID=30e3faba-8e9e-4616-97ab-aae747c6a3b5
InstallationMedia: Ubuntu 10.04 "Lucid Lynx" - Alpha i386 (20100224.1)
MachineType: Dell Computer Corporation Latitude D600
Package: linux (not installed)
PccardctlIdent:
 Socket 0:
   no product info available
 Socket 1:
   product info: "O2Micro", "SmartCardBus Reader", "V1.0", ""
   manfid: 0xffff, 0x0001
PccardctlStatus:
 Socket 0:
   no card
 Socket 1:
   5.0V 16-bit PC Card
   Subdevice 0 (function 0) [unbound]
ProcCmdLine: BOOT_IMAGE=/boot/vmlinuz-2.6.32-16-generic root=UUID=99ec916b-a551-4e75-85d2-60d165d6e870 ro quiet splash
ProcEnviron:
 LANG=fi_FI.utf8
 SHELL=/bin/bash
ProcVersionSignature: Ubuntu 2.6.32-16.25-generic
Regression: No
RelatedPackageVersions: linux-firmware 1.32
Reproducible: Yes
RfKill:
 0: hci0: Bluetooth
  Soft blocked: no
  Hard blocked: no
Tags: lucid needs-upstream-testing
TestedUpstream: No
Uname: Linux 2.6.32-16-generic i686
UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare
dmi.bios.date: 06/29/2005
dmi.bios.vendor: Dell Computer Corporation
dmi.bios.version: A16
dmi.board.name: 0D2125
dmi.board.vendor: Dell Computer Corporation
dmi.chassis.type: 8
dmi.chassis.vendor: Dell Computer Corporation
dmi.modalias: dmi:bvnDellComputerCorporation:bvrA16:bd06/29/2005:svnDellComputerCorporation:pnLatitudeD600:pvr:rvnDellComputerCorporation:rn0D2125:rvr:cvnDellComputerCorporation:ct8:cvr:
dmi.product.name: Latitude D600
dmi.sys.vendor: Dell Computer Corporation

[Swanson-ukfsn (swanson-ukfsn)]

Appears fixed with "radeon/r200/r300: don't clean non-emitted state" patch committed to mesa_7_7_branch and master branches.

[Swanson-ukfsn (swanson-ukfsn)]

Typical. You test it long enough to feel comfortable closing bug and problem immediately reappears.

[Pauli (paniemin)]

Does it help if you disable lightning in compiz settings? (general settings)

To me it looks like DRI2 r200 has some lightning artefacts which causes wrong output colors when some lightning settings are used. (just what compiz is using it seems)

Now it would be nice to have simplier test case on what is going wrong. :)

[Rankincj-googlemail (rankincj-googlemail)]

(In reply to comment #3)
> Does it help if you disable lightning in compiz settings? (general settings)

I have rebuilt F12's latest Mesa package with this small change from git, and will test it tonight. On the machine that I originally raised this bug on.

[Rankincj-googlemail (rankincj-googlemail)]

(In reply to comment #3)
> Does it help if you disable lightning in compiz settings? (general settings)

I am running the F12 7.7-2 mesa packages with the "radeon/r200/r300: don't clean non-emitted state" fix applied "by hand".

First impressions: this patch has (so far) prevented gnome-terminals turning completely black while rotating the compiz cube. So that's good. However, the entire desktop now looks like it's been plunged into shadow while the cube rotates instead. Sometimes, that shadow can even be pink... (After waking from a DPMS "off" state, in that particular case.)

Disabling "lighting" in the compiz settings manager makes no difference.

[Rankincj-googlemail (rankincj-googlemail)]

Created an attachment (id=32886)
Desktop im darkness while it rotates

Fortunately, the "lights" come back on when the cube settles again.

[Rankincj-googlemail (rankincj-googlemail)]

Created an attachment (id=32887)
Screenshot oddity... (Very slow shutter-speed?)

BTW, that previous screenshot was just one in a very slow series of screenshots. This is what one of the other images turned out like!

[Andreas Modinos (blueturtl)]

Running Lucid Alpha with todays updates I have some interesting visual effects:

Zooming in with Super+Scroll causes the screen to become very dark. It looks like the backlight on the display turns off, but I'm not sure. Adjusting Fn+Brightness doesn't fix this, although I can tell the brightness changing slightly. The same thing happens when changing workspaces (Ctrl+Alt+-> or Ctrl+Alt+<-), dragging a window from one workspace to another, or using Expo (Super+E) to view all workspaces simultaneously. Returning to normal zoom level will always restore the correct level of brightness, corresponding to how it might have been adjusted during black out.

I'm reporting this against the kernel and not the display driver because the bug doesn't appear with the corresponding mainline kernel release (2.6.32.9).

Attachments to follow...

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

apport information

[Andreas Modinos (blueturtl)]

Based on a tip I appended "nomodeset" to the kernel boot parameters and that fixed the brightness but introduced graphics corruption like seen in the attached screenshot.

After removing "nomodeset" and rebooting I am for the first time using the system without a sign of the bug in question (which of course makes sense seeing as I *just* decided to file a bug for it. I didn't know Murphy's law applied to open source development...

[Andreas Modinos (blueturtl)]

Spoke too soon.

Running glxgears brought the whole effect back, but I've noticed a new consistency about it:

1) While glxgears runs, zooming in or out produces a dark picture, but moving windows from workspace to another doesn't.
2) The effect lasts even after closing glxgears for zooming.
3) After moving between workspaces once, zooming works again with full brightness.

Some kind of bug in the graphics driver on either the kernel or X.org would be my guess, but I'm not an expert on those so... anyone wanting to point this bug toward a more appropriate package feel free.

[Rankincj-googlemail (rankincj-googlemail)]

The "dark-out" when rotating the compiz cube is still happening with the vanilla 2.6.33.1 kernel and Fedora 12's Mesa 7.7-4 packages.

[Andreas Modinos (blueturtl)]

Bug reappears as described originally after the system is has been powered down between reboots.

Confirming it's existence on 2.6.32-17-generic and with todays updates to X.org.

[chicken159 (services)]

Same symptoms with Radeon 9200 Pro on amd64.

[Andreas Modinos (blueturtl)]

An interesting observation: when I use the magnifying glass the cursor remains fully lit when everything else turns to dark.

I'm not sure if this is a new development or if I just failed to notice before.

[Andreas Modinos (blueturtl)]

Confirming bug with kernel 2.6.32-18-generic

Maybe this bug should be filed against xserver-xorg-video-ati package as well?

chicken159, can you post more detailed information about your hardware? Seeing as you are on a different video card and architecture you must be using a different machine from mine. The Radeon 9200 and 9000 are relatives though so I think that must be how we're connected.

[Andreas Modinos (blueturtl)]

Today there were both kernel updates and video driver updates, but the bug still occurs.

kernel 2.6.32-19-generic (2.6.32-19.28)
xserver-xorg-video-radeon (1:6.12.192-2ubuntu2)

[agd5f (agd5f)]

*** Bug 27476 has been marked as a duplicate of this bug. ***

[Brandon Cash (brandoncash)]

Confirming on two separate machines with two different Radeon video cards on Lucid.

Desktop:
uname -a: Linux chronos 2.6.32-19-generic #28-Ubuntu SMP Wed Mar 31 17:46:20 UTC 2010 i686 GNU/Linux
Card: PCI:*(0:1:0:0) 1002:5961:1043:5000 ATI Technologies Inc RV280 [Radeon 9200] rev 1
xserver-xorg-video-radeon version: 1:6.12.192-2ubuntu2

Laptop:
uname -a: Linux polaris 2.6.32-19-generic #28-Ubuntu SMP Wed Mar 31 17:46:20 UTC 2010 i686 GNU/Linux
Card: PCI:*(0:1:0:0) 1002:4c57:0e11:004a ATI Technologies Inc Radeon Mobility M7 LW [Radeon Mobility 7500] rev 0
xserver-xorg-video-radeon version: 1:6.12.192-2ubuntu2

[Andreas Modinos (blueturtl)]

I've changed the title description since this seems to apply to other older generation ATi cards as well: so far reported are 7500, 9000, and 9200 series.

[Andreas Modinos (blueturtl)]

linux-image-2.6.32-20-generic, 2.6.32-20.29
xserver-xorg-video-radeon, 1:6.13.0-1ubuntu1

Still no change for me. Is this being looked at?

I know desktop effects aren't priority but to me the desktop magnifier/zoom ability is a valuable tool.

[gator_ml (gator-ml)]

I have the same (at least very similar) problems with a Radeon RV250 (Mobility FireGL 9000) running the latest radeon driver (xserver-xorg-video-ati 1:6.13.0-1ubuntu5). The issue obviously (I tried it only shortly and then went on to 10.04) also occured with ubuntu 9.10.

Contrary to other reports, on my machine things get even worse when I turn kernel mode setting off (Desktop background and certain application windows like firefox totally black).

(See also http://ubuntuforums.org/showthread.php?p=9106012)

[agd5f (agd5f)]

*** Bug 27824 has been marked as a duplicate of this bug. ***

[Brandon Cash (brandoncash)]

For me, disabling the KMS (echo options radeon modeset=0 > /etc/modprobe.d/radeon-kms.conf) fixes the main problem, but introduces other problems (ugly boot, black borders around some windows, etc--but these are not in the scope of this bug).

In contrast to #30, I did not have the problem in 9.10--was KMS even an option there? I think it's obvious that moving some code to the kernel has introduced the problem, but we need someone qualified to make a fix. I'm afraid kernel hacking is far out of my comfort zone.

I think at this point the bug will remain in Lucid's release. Hopefully after the stress of an upcoming release has passed, someone will be able to take a look.

[Andreas Modinos (blueturtl)]

I've reported this bug upstream and against the X.org Radeon driver:
https://bugs.freedesktop.org/show_bug.cgi?id=27824

Looks like it might be a known issue.

[Andreas Modinos (blueturtl)]

Looks like I finally found the explanation...
My upstream bug was pretty quickly marked a duplicate of another bug and within that bug's comments I found this tip:
Go to Compiz settings (compizconfig-settings-manager), in there to general settings and deselect Lighting.

That fixed the issue for me. Apparently it's a bug in the display driver.

For now I am marking this bug invalid since it does not seem to be related to the kernel after all.

[Bryce Harrington (bryce)]

Workaround mentioned on one of the dupe bugs:

"""
I've fixed this issue by changing a parameter in compiz configuration.

Using ccsm, the parameter is "Lighting" in the "Display settings" tab of
"General compiz options".

If I unset this boolean parameter, transitions work fine.
"""

[agd5f (agd5f)]

(In reply to comment #11)
> Workaround mentioned on one of the dupe bugs:
>
> """
> I've fixed this issue by changing a parameter in compiz configuration.
>
> Using ccsm, the parameter is "Lighting" in the "Display settings" tab of
> "General compiz options".
>
> If I unset this boolean parameter, transitions work fine.
> """

Probably some of the lighting/material state is not emitted in 3D driver on r1xx, r2xx with kms.

[Rankincj-googlemail (rankincj-googlemail)]

(In reply to comment #11)
> Workaround mentioned on one of the dupe bugs:
>
> """
> I've fixed this issue by changing a parameter in compiz configuration.
>
> Using ccsm, the parameter is "Lighting" in the "Display settings" tab of
> "General compiz options".
>
> If I unset this boolean parameter, transitions work fine.
> """

Interesting, because unsetting that parameter makes no difference here. And I even tried rebooting the PC afterwards as well.

[imwithid (imwithid)]

I have the same problem. The contents of windows become dark and barely readable. The cursor, window top border and panels remain unaffected.

I'm running Lucid 10.04 with kernel 2.6.32-23 all updates as of an hour ago.

I have the following video hardware:

*-display
                description: VGA compatible controller
                product: Radeon RV250 [Mobility FireGL 9000]
                vendor: ATI Technologies Inc
                physical id: 0
                bus info: pci@0000:01:00.0
                version: 02
                width: 32 bits
                clock: 66MHz
                capabilities: agp agp-2.0 pm bus_master cap_list rom
                configuration: driver=radeon latency=66 mingnt=8
                resources: irq:11 memory:d8000000-dfffffff(prefetchable) ioport:

Notify me if you need any additional information.

[imwithid (imwithid)]

@ Andreas Modinos,

Your solution worked well, although it was confusing.

Specifically, you should do the following:

1. Go to compizconfig-settings-manager (if you don't have it, in terminal: sudo apt-get install compizconfig-settings-manager )
2. On the left side, go to Category > General Options > Display Settings (Tab) > Lighting (uncheck box).

I'm wondering if this could be disabled by default or if it is neccesary for new video cards.

Thanks for the fix.

[Marco Albanese (delian2)]

I've the same bug, i suppose, but with an rv200 chip (Radeon Mobility 7500).
The lighting "colour" is random. Often is light blue or light red.

I'm using KDE 4.6.4 (with desktop effects enabled) on gentoo with xf86-video-ati 6.14.2, mesa 7.10.2 and kernel linux 2.6.38-gentoo-r6 with KMS enabled.

Let me now if you need more information.

[Marco Albanese (delian2)]

Created attachment 48823
Lighting problem with amarok OSD.

It's only an example, it happens with every trasparency/lighting transition.

[Stefan Nagy (stefan-nagy)]

I'm affected by this bug on Ubuntu 10.10 (maverick). I'm seeing the same symptoms as the original reporter & the mentioned workaround solves the problem.

kernel 2.6.35-30-generic
xserver-xorg-video-ati 1:3.13.1-1ubuntu5
ATI Technologies Inc Radeon RV250 [Mobility FireGL 9000] [1002:4c66] (rev 01)

[Stefan Nagy (stefan-nagy)]

I can reproduce this bug with this graphics card (with compiz & KMS):
ATI Technologies Inc Radeon RV250 [Mobility FireGL 9000] [1002:4c66] (rev 01)

When I switch workspaces, rotate the cube or use Expo to view all workspaces the whole screen gets very dark – but the colours won't change (as described above).

I can confirm this bug for
· Ubuntu 10.10 (maverick): kernel 2.6.35-30-generic; xserver-xorg-video-ati 1:6.13.1-1ubuntu5; mesa 7.9~git20100924-0ubuntu2
· Debian Testing (wheezy): kernel 3.0.0-1-486 (3.0.0-3); xserver-xorg-video-ati 1:6.14.2-2; mesa 7.11-6

[Swanson-ukfsn (swanson-ukfsn)]

*** This bug has been marked as a duplicate of bug 22576 ***

[veldt (veldt)]

Just to confirm (unless my hardware is a corner case), this bug seems fixed in freedesktop (freedesktop.org bugzilla #22576); will it be incorporated in ubuntu, ideally 13.04?

My hardware seems close the original posters' -- "legacy radeon" -- I'm using r300 but they seem to be on r200, but the symptoms I was experiencing match (dark zoom), and the "workaround" is effective.

I have an HP Pavilion zv6130us laptop with:

01:05.0 VGA compatible controller: Advanced Micro Devices [AMD] nee ATI Radeon XPRESS 200M 5955 (PCIE) (prog-if 00 [VGA controller])

.. driven as r300 as shown in Xorg log:

[ 35.943] (II) RADEON(0): [DRI2] DRI driver: r300
[ 35.943] (II) RADEON(0): [DRI2] VDPAU driver: r300
[ 35.943] (II) RADEON(0): Front buffer size: 3072K
[ 35.943] (II) RADEON(0): VRAM usage limit set to 111369K
[ 35.943] (==) RADEON(0): Backing store disabled
[ 35.943] (II) RADEON(0): Direct rendering enabled
[ 35.943] (II) RADEON(0): Render acceleration enabled for R300/R400/R500 type cards.

kms / kernel modesetting is on.

To update, on my ubuntu 12.10 "GNOME Classic" system, the workaround in compizconfig-settings-manager (a.k.a. ccsm a.k.a (in a GNOME Classic X login session after gnome-session-fallback is installed) Applications/System Tools/Preferences/CompizConfig Settings Manager) is:

category -> general -> opengl -> lighting (uncheck)

Bug-tracking note:

This bug was marked invalid when linked to freedesktop bug 25883, presumably because that was a resolved duplicate of the resolved fixed freedesktop bug 22576. I accept that this freedesktop-assigned "xserver-xorg-driver-ati" bug is invalid. Nothing has been posted after that in the last few months. But I would point out that the "xserver-xorg-video-ati (Ubuntu)" bug is still valid. Has the fix been implemented in Ubuntu? If so, is my setup showing that it is incomplete?

Misleading link, perhaps auto-generated:

In comment #54 of this bug, it is marked as a duplicate of "bug 22576". This makes sense as a freedesktop bug at https://bugs.freedesktop.org/show_bug.cgi?id=22576, but the link is wrongly set to an unrelated launchpad bug at https://bugs.launchpad.net/bugs/22576.

I am willing to do testing.