Bug 846505

I have the same issue. I'm using LXDE so my desktop is fine, but when I play Enemy Territory: QUAKE Wars, after several minutes there is huge rendering problem. The console log says:
"radeon: The kernel rejected CS, see dmesg for more information." many many times! (until I leave the game).
Nothing in dmesg though.
Last time, I was playing normally during 25 minutes and instantly rendering corruption appears.

This happens when I boot on the new 3.5.0-2.fc17.x86_64 kernel. No problem with the previous 3.4 series.

Software:
xorg-x11-drv-ati 6.14.4-6.20120602git930760942.fc17
libdrm 2.4.33-3.fc17
mesa-r600g (git)

Hardware:
01:00.0 VGA compatible controller [0300]: ATI Technologies Inc RV770 [Radeon HD 4850] [1002:9442]

I've seen similar problems on both KDE and GNOME. In GTK apps it were mainly font-rendering problems, GDM showing up in a weird way, although GDM was functional and the Gnome shell didn't show up.

In KDE it were large rectangles covering large parts of text in Konqueror, although resizing the window or moving seemed to make these go away again (maybe related to: https://bugzilla.redhat.com/show_bug.cgi?id=845639 ?)

On 3.4.6 all these problems disappeared.

(In reply to comment #2)
> I've seen similar problems on both KDE and GNOME. In GTK apps it were mainly
> font-rendering problems, GDM showing up in a weird way, although GDM was
> functional and the Gnome shell didn't show up.
> 
> In KDE it were large rectangles covering large parts of text in Konqueror,
> although resizing the window or moving seemed to make these go away again
> (maybe related to: https://bugzilla.redhat.com/show_bug.cgi?id=845639 ?)
> 
> On 3.4.6 all these problems disappeared.

In addition to the above: KDM (login) worked flawlessly and rendered correctly even under 3.5 kernels.

I have a HP nc8430 (ati mobility x1600) with F17 with the xorg-x11-drv-ati
driver, and also see the graphical artifacts with kernel 3.5.2 (x86_64).
There are no artifacts with kernel 3.3.4.

From /var/log/messages, maybe drm is really unhappy ... there is a timeout.

                                                                

From 3.5.2 message log:
[1.700710][drm] Initialized drm 1.1.0 20060810
...
[1.771320] [drm] Generation 2 PCI interface, using max accessible memory
[1.771326] radeon 0000:01:00.0: VRAM: 256M 0x0000000000000000 - 0x000000000FFFFFFF (256M used)
[1.771330] radeon 0000:01:00.0: GTT: 512M 0x0000000010000000 - 0x000000002FFFFFFF
[1.771356] [drm] Supports vblank timestamp caching Rev 1 (10.10.2010).
[1.771358] [drm] Driver supports precise vblank timestamp query.
[1.771422] radeon 0000:01:00.0: radeon: using MSI.
[1.771451] [drm] radeon: irq initialized.
[1.773739] [drm] Detected VRAM RAM=256M, BAR=256M
[1.773744] [drm] RAM width 128bits DDR
[1.773824] [TTM] Zone  kernel: Available graphics memory: 1025868 kiB
[1.773826] [TTM] Initializing pool allocator
[1.773833] [TTM] Initializing DMA pool allocator
[1.773868] [drm] radeon: 256M of VRAM memory ready
[1.773870] [drm] radeon: 512M of GTT memory ready.
[1.773898] [drm] GART: num cpu pages 131072, num gpu pages 131072
[1.775982] [drm] radeon: 1 quad pipes, 2 z pipes initialized.
[1.965408] Failed to wait MC idle while programming pipes. Bad things might happen.
[1.966592] [drm] PCIE GART of 512M enabled (table at 0x0000000000040000).
[1.966621] radeon 0000:01:00.0: WB enabled
....


From 3.3.4 message log:
[1.600398] [drm] Initialized drm 1.1.0 20060810
...
[1.617300] [drm] Generation 2 PCI interface, using max accessible memory
[1.617312] radeon 0000:01:00.0: VRAM: 256M 0x0000000000000000 - 0x000000000FFFFFFF (
256M used)
[1.617316] radeon 0000:01:00.0: GTT: 512M 0x0000000010000000 - 0x000000002FFFFFFF
[1.617339] [drm] Supports vblank timestamp caching Rev 1 (10.10.2010).
[1.617341] [drm] Driver supports precise vblank timestamp query.
[1.617467] radeon 0000:01:00.0: radeon: using MSI.
[1.617510] [drm] radeon: irq initialized.
[1.619911] [drm] Detected VRAM RAM=256M, BAR=256M
[1.619915] [drm] RAM width 128bits DDR
[1.621119] [TTM] Zone  kernel: Available graphics memory: 1025974 kiB.
[1.621123] [TTM] Initializing pool allocator.
[1.621129] [TTM] Initializing DMA pool allocator.
[1.621163] [drm] radeon: 256M of VRAM memory ready
[1.621165] [drm] radeon: 512M of GTT memory ready.
[1.621192] [drm] GART: num cpu pages 131072, num gpu pages 131072
[1.622309] [drm] radeon: ib pool ready.
[1.623432] [drm] radeon: 1 quad pipes, 2 z pipes initialized.
[1.624671] [drm] PCIE GART of 512M enabled (table at 0x0000000000040000).
[1.624711] radeon 0000:01:00.0: WB enabled

For reference to have all informations that might be usefull

> #:/usr/src/linux-stable ((bb63556...)|BISECTING)# git bisect bad
> bb635567291482a87e4cc46e6683419c1f365ddf is the first bad commit
> commit bb635567291482a87e4cc46e6683419c1f365ddf
> Author: Jerome Glisse <jglisse>
> Date:   Wed May 9 15:34:46 2012 +0200
>
>     drm/radeon: convert fence to uint64_t v4
>    
>     This convert fence to use uint64_t sequence number intention is
>     to use the fact that uin64_t is big enough that we don't need to
>     care about wrap around.

I have the same issues listed on a Radeon Xpress 1250. (Dell Latitude XT) and stock fedora driver.

Which application trigger the issue ? Stock gnome-shell without any modification or customization ? How do you easily trigger the issue ?

Until we can reproduce it's hard to fix.

(In reply to comment #7)
> Which application trigger the issue ? Stock gnome-shell without any
> modification or customization ? How do you easily trigger the issue ?
> 
> Until we can reproduce it's hard to fix.

For me it were mainly GTK and QT apps that triggered the issue. Framebuffer apps (e.g. links2 -G) and a terminal running in Konsole didn't show any problems. 

I'm running FC17 on a Dell E1505/Inspiron 6400 with a Mobility Radeon X1400. I ran Kernel 3.5.0-2.fc17.i686 to gather some data. I didn't succeed in making a screenshot of GDM, but I have made one of my XFCE desktop afterwards. I'll attach it in a minute. 

Furthermore I couldn't reproduce the poster in Comment 4 spoke about. I did find the following errors/warnings in /var/log/gdm/:0-greeter.log:

Window manager warning: CurrentTime used to choose focus window; focus window may not be correct.
Window manager warning: Got a request to focus the no_focus_window with a timestamp of 0.  This shouldn't happen!

(gnome-settings-daemon:1451): GLib-GObject-CRITICAL **: g_object_unref: assertion `G_IS_OBJECT (object)' failed

(gnome-settings-daemon:1451): GLib-GObject-CRITICAL **: g_object_unref: assertion `G_IS_OBJECT (object)' failed
g_dbus_connection_real_closed: Remote peer vanished with error: Underlying GIOStream returned 0 bytes on an async read (g-io-error-quark, 0). Exiting.

--------
I'll attach my /var/log/messages from that last session and the screenshot. In case you need other logs or you want me to try other things feel free to ask.

Created attachment 606677 [details]
Graphical artefacts in XFCE.

I'm using a dual monitor setup so screenshot shows both monitors. On the right you see Konsole running, on the left a gtk based terminal GUI.

Created attachment 606678 [details]
/var/log messages of 3.5.0-2.fc17.i686.PAE kernel

(In reply to comment #7)
> Which application trigger the issue ? Stock gnome-shell without any
> modification or customization ? How do you easily trigger the issue ?
> 
> Until we can reproduce it's hard to fix.

In my case, stock gnome-shell is affected (including the login screen), as well as any other graphical programs I've run.  I haven't tried installing a different desktop environment, but from the above, it sounds like the problem persists in XFCE.

Just booting the system is sufficient to trigger the issue in my case; the login screen itself is corrupted 100% of the time (in the 10-20 times that I've tried it), as is the entirety of gnome after logging in.

(In reply to comment #7)
> Which application trigger the issue ? Stock gnome-shell without any
> modification or customization ? How do you easily trigger the issue ?
> 
> Until we can reproduce it's hard to fix.

I have the same problem with my ATI Radeon Mobility X700.

Booting without Plymouth shows startup messages normally but once GDM is started, it is garbled 100% time. When logging in the whole desktop / gnome-shell is messy, going to the upper left corned for the favorities menu show the menu but all the icons are messy and I need to click like 20 times before a favorite app is being opened. Zero modification locally except for Plymouth disabled with a kernel parameter.

This is a severe regression, using kernel 3.5 is now impossible and while it's possible to stick with 3.4 on F17 I fear that upgrading to F18 can't be done before this is fixed as the whole desktop is unusable.

Does kernel patch:

http://people.freedesktop.org/~glisse/0001-drm-radeon-extra-type-safe-for-fence-
emission.patch

Helps ?

(In reply to comment #7)
> Until we can reproduce it's hard to fix.

Actually I fail to reproduce the bug on ETQW! I don't know why. It's very strange. However, I have on hypothesis:

I compiled r600g with the kernel-headers 3.4. And then upgraded to 3.5 without recompiling r600g. Can this be a source of problem?

Also can you test if booting with radeon.no_wb=1 fix the issue ?

I find that the 'Failed to wait MC idle while programming pipes. Bad things..'
line in my Comment #4 does not happen all the time, so it is likely not
relevant.

Every boot with 3.5.2 on a HP nc8430 (x1600) gives a corrupted GDM login window
as well as Gnome3 shell. I also get corruption in a single Xclient and Xserver
on another tty: xinit urvxt -display :4 -- Xorg :4 vt4. The scrolling
text in the xterm is unreadable - horizontal striations that sometimes extend
way past the xterm right boundary. But playing videos (mplayer) windowed or
full-screen is corruption free (hardware fb?).

Booting to 3.5.2 with radeon.no_wb=1 and the stock radeon driver still has
corruption in the GDM login screen. Are there help files on getting and
compiling a test radeon driver (configure/Makefile)?

(In reply to comment #15)
> Also can you test if booting with radeon.no_wb=1 fix the issue ?

No, didn't help at all with my ATI Radeon Mobility X700.

Any chances you could provide a kernel package with the patch for testing?

(In reply to comment #13)
> Does kernel patch:
> 
> http://people.freedesktop.org/~glisse/0001-drm-radeon-extra-type-safe-for-
> fence-emission.patch
> 
> Helps ?

Corruption still there with this patch, no notable change.

I have very similar issues as #11 and #12.

Stock F17, with KDE. Kernel 3.5.0-2. Mobility Radeon X700.

I'm just adding that other people including me are seeing this on NVIDIA nouveau since the 3.5 series all was fine in 3.4. So maybe not a card specific driver issue?

https://bugzilla.redhat.com/show_bug.cgi?id=845965

I've now tested kernel-PAE-3.6.0-0.rc2.git2.1.fc18.i686 from Rawhide and the problem persist.

Jerome, any suggestions, seems like this issue affects of lots of people?

kernel-3.5.4-2.fc17 has been submitted as an update for Fedora 17.
https://admin.fedoraproject.org/updates/kernel-3.5.4-2.fc17

Package kernel-3.5.4-2.fc17:
* should fix your issue,
* was pushed to the Fedora 17 testing repository,
* should be available at your local mirror within two days.
Update it with:
# su -c 'yum update --enablerepo=updates-testing kernel-3.5.4-2.fc17'
as soon as you are able to, then reboot.
Please go to the following url:
https://admin.fedoraproject.org/updates/FEDORA-2012-14952/kernel-3.5.4-2.fc17
then log in and leave karma (feedback).

Thanks a lot, 3.5.4-2.fc17 fixed the issue for me, karma added!

As I posted in bodhi (Anonymous tester) it solved the graphical problems for me as well.

kernel-3.5.4-2.fc17 has been pushed to the Fedora 17 stable repository.  If problems still persist, please make note of it in this bug report.