With the current nouveau git drm (4a97c59) my X frequently freezes (pointer still moves, but everything else hangs), often within few minutes after starting X, sometimes much later. My kernel log shows a PFIFO_DMA_PUSHER on Ch 2, the Xorg logs a EQ overflow. I'm rather confident that commit 5837471 is causing my problems (although bisecting was very hard as the hangs occur randomly). With the commit before, I haven't had any problems yet, and reverting this commit on current HEAD seems to solve the problem, too. The revert seems also to fix s2ram, which is also broken in current HEAD for me (on resume I get a black screen with the pointer, in contrast to the other hangs the pointer still changes when moving over a text field etc.).
There's been a number of reports of random hangs etc since 2.6.36-rc1 was merged, and I agree this is probably the most likely candidate. However, I'm not certain why yet. I personally don't see any issue in normal use, but may have just encountered the problem running gtkperf with compiz enabled. I'll play around with reverting that commit now too. For the suspend issue, I've seen that too randomly, a vt-switch to console and back seems to "fix" it.
I see this issue without running any GL programs and with only swrast installed. It don't think I do anything special when I get it, it occurs in normal use for me.
Yeah, I don't think they're related. Just meant to note that I've seen regressions as a result of that commit too. I've tracked down the fix for the issue I was seeing, it's in nouveau git now. But, I don't think it'll help you still, unfortunately.
Indeed, the current HEAD doesn't work any better. After reverting 5837471, I haven't had any crashes, my system was running several hours since then, so it is really that commit causing the hangs. The s2ram issue occured nevertheless, so that problem is completely unrelated to the hangs.
I confirm the exact same issue.
This seems to be a duplicate of bug 29583, can you confirm that the patch attached to it fixes it?
This patch sort of works. My system used to hang just after few minutes of usage. With this patch I used the system for about 1/2 of hour, but then X hanged. I might be another bug. ( I also did many s2ram cycles, which sometimes garbage the screen, so I might be just lucky to have this hang) Note that in 2.6.35, I see no hangs, no matter what I do with the system. (But like I said before I might have beein just lucky. I do see occasional garbage on the screen after resume from ram) I always run compiz.
Created attachment 38258 [details] [review] nouveau_pushbuf_deadlock.patch (In reply to comment #7) > This patch sort of works. > > My system used to hang just after few minutes of usage. > With this patch I used the system for about 1/2 of hour, but then X hanged. > I might be another bug. > ( I also did many s2ram cycles, which sometimes garbage the screen, so I might > be just lucky to have this hang) > Can you reproduce it with the attached patch applied on top of the other one? If you can (and it isn't locked up hard), please provide kernel logs after the hang. > Note that in 2.6.35, I see no hangs, no matter what I do with the system. > (But like I said before I might have beein just lucky. I do see occasional > garbage on the screen after resume from ram) > > I always run compiz.
The patch from bug 29583 seems to fix the hangs for me, my system has been running for several hours without problems. I'm not running compiz or other GL programs. Should I test the new patch, too?
(In reply to comment #9) > The patch from bug 29583 seems to fix the hangs for me, my system has been > running for several hours without problems. I'm not running compiz or other GL > programs. > Thanks, I'm marking this bug as duplicate then. > Should I test the new patch, too? *** This bug has been marked as a duplicate of bug 29583 ***
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.