Created attachment 66989 [details]
nouveau dmesg log

I've run a bisect on Linus' master branch, and narrowed the issue down to this one line change, which isn't much to go on. It looks like the change of struct type has caused issues where it is used elsewhere :

commit 7707b701ebfea64afa6bfb23aa318fd687892754
Author: Marcin Slusarz <marcin.slusarz@gmail.com>
Commit: Ben Skeggs <bskeggs@redhat.com>

    drm/nv40/mpeg: fix context handling

    It slipped in thanks to typeless API.

    Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
    Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

diff --git a/drivers/gpu/drm/nouveau/core/engine/mpeg/nv40.c b/drivers/gpu/drm/n
index 1241857..f7c581a 100644
--- a/drivers/gpu/drm/nouveau/core/engine/mpeg/nv40.c
+++ b/drivers/gpu/drm/nouveau/core/engine/mpeg/nv40.c
@@ -38,7 +38,7 @@ struct nv40_mpeg_priv {
 };

 struct nv40_mpeg_chan {
- struct nouveau_mpeg base;
+ struct nouveau_mpeg_chan base;
 };

Thanks for fixing... Unfortunately the bug seemed gone so I am afraid there is no way to test it now.

Alin

This also happens on my NVAC/MCP79 running Ubuntu 12.10 with the xorg-edgers PPA and Linux 3.8. Unity fails to start (only the background image and the mouse cursor appear) and, after a while, dmesg shows a couple of "failed to idle channel 0xcccc0000" messages.

Still happens here.  Both with linus/master and nouveau-drm-next.  I'm going to try reverting the above mentioned commit, although it's not obvious to me how it would be the cause...

(In reply to comment #5)
> Still happens here.  Both with linus/master and nouveau-drm-next.  I'm going
> to try reverting the above mentioned commit, although it's not obvious to me
> how it would be the cause...
As expected, reverting 7707b701ebfea64afa6bfb23aa318fd687892754 made no difference. :(

Hardware is: "NVIDIA Corporation NV35 [GeForce FX 5900] (rev a1)"

The same random X crashes with the same "failed to idle channel 0xcccc0000" messages are seen with on a NVC0 family card, NVIDIA Corporation GF119 [Quadro NVS 4200M], also on kernel 3.8.

The only way to have a stable system is to disable accel on boot.

Still present w/ fedora kernel 3.8.3-201.fc18.x86_64

Works for me since Linux 3.9-rc3. Thanks!

Last comment made me update my kernel (got it in 3.8.4) but I got it also in 3.9-rc4; trying latest nouveau patches since then, just in case.
Hardware: GF106GL (Quadro 2000) (nvc0).
I wish I could reproduce it simply.

"failed to idle channel xxx" means "the gpu appears to not have hit a sync point". This can happen for a variety of reasons, differing for different hardware. It has a special propensity for showing up on a resume where operations were going on while the gpu was suspending, but can also happen in other circumstances.

I'm going to close this bug because the issue is reportedly fixed for the original bug reporter, and every other commenter has vastly different hardware. I would advise each of you to update your software (kernel, xf86-video-nouveau, mesa) to the latest, and if problems persist file your own separate bugs, indicate what hardware you have, and what steps to reproduce, including logs. See http://nouveau.freedesktop.org/wiki/Bugs/

Guys this issue seems to be present in trusty thar now https://bugs.launchpad.net/ubuntu/+source/linux-lts-quantal/+bug/1097178. This is critical bug for tons of people. I have to restart the system 20 times a day. Unfortunately I'm having a similar issue with nvidia driver, so I'm kinda without options.

All the original reasons to have closed the bug still exist. Re-closing. Please read comment 12, in its entirety.