Bug 39612 - radeon blocks with new style fencing
Summary: radeon blocks with new style fencing
Status: NEEDINFO
Alias: None
Product: Drivers
Classification: Unclassified
Component: Video(DRI - non Intel) (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: drivers_video-dri
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-07-19 18:45 UTC by Linas
Modified: 2014-01-06 09:20 UTC (History)
2 users (show)

See Also:
Kernel Version: 3.6.2
Subsystem:
Regression: Yes
Bisected commit-id:


Attachments
Xorg log (43.44 KB, text/plain)
2011-08-01 22:29 UTC, Linas
Details
Xorg log mieq (51.84 KB, text/plain)
2012-10-21 19:31 UTC, Linas
Details
Other Xorg log (48.64 KB, text/plain)
2012-10-21 19:34 UTC, Linas
Details

Description Linas 2011-07-19 18:45:39 UTC
I am using a Radeon HD 3600 in kernel 2.6.36 with xf86-video-ati (KMS). Since 2.6.37 when starting the X server the screen goes black and never shows the desktop.

kernel 2.6.36
startx -> Screen goes black -> The wallpaper appears -> The DE loads

kernel >= 2.6.37
startx -> Screen goes black -> You can't switch virtual terminals -> REISUB -> Downgrade kernel

Reproducible: Always

Bisecting shows that it started failing at d0f8a854c340986359a3b0a97e380c71def7a440: drm/radeon/kms/r6xx+: use new style fencing (v3)

http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.37.y.git;a=commit;h=d0f8a854c340986359a3b0a97e380c71def7a440

Downstream bug https://bugs.archlinux.org/task/23422
Comment 1 Alex Deucher 2011-07-20 04:19:06 UTC
Please attach your xorg log and dmesg output.  Does booting with radeon.no_wb=1 on the kernel command line in grub help?
Comment 2 Linas 2011-08-01 22:29:21 UTC
Created attachment 67342 [details]
Xorg log

Sorry for the delay. I'm attaching the file.
The radeon.no_wb=1 workaround does indeed work. Much appreciated.
Instead of showing only the mouse pointer (if lucky), shows the full screen again.
Comment 3 Alan 2012-08-24 15:33:20 UTC
Is this still needing workarounds or can I close it ?
Comment 4 Linas 2012-10-21 19:31:38 UTC
Created attachment 84201 [details]
Xorg log mieq

Testing with 3.6.2, radeon.no_wb=1 is still needed, although it seems a bit better.
You can get -with a little difficulty- to switch virtual terminals, so it is possible to SIGINT the xserver. Interestingly, although the first startx leads to a blank screen, a second startx runs the DE without being apparently affected. ?!
Booting with radeon.no_wb=1 the DE shows directly the first time, as it should.

In one failed instance, the Xorg log showed errors of «[mi] EQ overflowing.», «[mi] EQ overflow continuing.  100 events have been dropped.», «[mi] EQ overflow continuing.  200 events have been dropped.»... I am attaching this log.
Comment 5 Linas 2012-10-21 19:34:45 UTC
Created attachment 84211 [details]
Other Xorg log

Another failed Xorg log, this time without overflowing errors.
Comment 6 Michel Dänzer 2013-12-11 04:07:12 UTC
Is this still an issue with current kernels and userspace stack?
Comment 7 Linas 2013-12-30 19:29:47 UTC
I still have video problems, although causing a freeze at some random point*, rather than a black screen after boot.


* Or you can make it deterministic by running something like /usr/lib/xscreensaver/kumppa ;)
Comment 8 Michel Dänzer 2014-01-06 09:20:57 UTC
(In reply to Linas from comment #7)
> I still have video problems, although causing a freeze at some random
> point*, rather than a black screen after boot.

Are you sure it's still the same problem? Does no_wb=1 still work around it?

Note You need to log in before you can comment on or make changes to this bug.