[nouveau] 12.04 Unity 3d hard freezes

[Brendan Brigham]

This is a fresh install of Precise, there is constant hard freezes in Unity 3d. It doesn't matter if it's a video or text editor, heavy graphical use doesn't dictate when it freezes. I have replicated it anywhere from watching a video to opening an email or even editing one of my conkyrc files. I have been running in Unity 2d for a couple weeks without one such problem. I am running dual monitors and can have videos going on both screens without a hitch. Edit a file in 3d and I have to hard reset. I have been following a couple other bug reports and have tried different fixes that they've been experimenting with. I am now running in kernel 3.4.0-030400 with no success, I have also tried the xorg backports they've suggested with no positive results.
Upon install this was originally kernel 3.2.0.24, which was the start of freezes.

My system specs are :
Desktop with wired LAN
Ubuntu 12.04LTS : kernel 3.4.0-030400
Sandy Bridge Core i3
MSI H55M-E33 mobo
4gigs ddr3 ram
Asus ENGT 240 graphics card using <s>Nouveau</s> drivers (Nvidia drivers now working) into dual 19" monitors

I can not SSH into the machine, no second computer access. If there's some other information I can provide just let me know, I'll try to get it.

ProblemType: Bug
DistroRelease: Ubuntu 12.04
Package: unity 5.12-0ubuntu1.1
Uname: Linux 3.4.0-030400-generic i686
ApportVersion: 2.0.1-0ubuntu8
Architecture: i386
CompizPlugins: [core,composite,opengl,compiztoolbox,decor,imgpng,vpswitch,mousepoll,regex,wall,grid,snap,move,resize,place,gnomecompat,unitymtgrabhandles,animation,session,fade,expo,workarounds,scale,ezoom,unityshell]
Date: Wed Jun 13 20:09:23 2012
InstallationMedia: Ubuntu 12.04 LTS "Precise Pangolin" - Release i386 (20120423)
ProcEnviron:
 PATH=(custom, no user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: unity
UpgradeStatus: No upgrade log present (probably fresh install)

[Daniel van Vugt (vanvugt)]

Brendan,

The NVIDIA driver does support multiple monitors. You can configure them in "nvidia-settings". Please try that, even briefly, so we can identify if this is a problem with nouveau.

I recall having several system lockups myself when I was testing nouveau last week.

[Brendan Brigham (brigham)]

Hi Daniel, I've actually tried several builds of the Nvidia drivers up to and including the newest releases. I never had any problems with them in Ubuntu11.04 but in 12.04 one monitor goes white and the cursor on that screen is an x. I have seen several other complaints of the Nvidia drives doing the same with 12.04.
I will try them again tonight to be sure there wasn't any more recent changes, I do prefer Nvidia so I can monitor the temps. I will let you know shortly for sure.

[Daniel van Vugt (vanvugt)]

Brendan,

You can find out the GPU temperature with nouveau too. Just install package "lm-senors", configure it with "sensors-detect" and then run "sensors" to view the temp.

[Brendan Brigham (brigham)]

Hi Daniel, well how bout that, they must have updated their drivers since last week. They work now o.O
And oddly enough, so far it hasn't froze in 3d, I usually able to force a freeze within 10 minutes. I will keep pushing it to see if I can force it again. If not maybe this is indeed a Nouveau issue ?! I know a Lot of people who will be happy about that. Well, other than those not using Nvidia cards lol

[Brendan Brigham (brigham)]

Good evening Daniel, I've pushed my system pretty hard running videos on both monitors while recording video from my hd webcam and every other thing I could think of and have had not one glitch since the new Nvidia drivers.
Definitely a Nouveau issue.

[markb (mark-blakeney)]

I'm using nouveau from xorg-edgers ppa on 12.04 and I think I am having this same bug just recently. My Unity 3D session will hang anywhere from 1 minute to a few hours after I boot. Sometimes the entire UI locks up, other times the UI locks up but my mouse cursor will still move (but can't select anything). I can not select an alternate console via ctrl+alt+f1 etc. Note I mainly use gnome shell and I have never seen this problem there, only in Unity 3D which I occasionally try out to see where it is at. This bug essentially stops me from using Unity atm. It could be related to kernel 3.5.0-2 which xorg-edgers seems to have pulled in to my system 3 days ago?

[markb (mark-blakeney)]

Further to my comment #7 above, I noticed an xorg-core update today so I tried Unity for a while but it locked up within 30 mins. System died over about 20 secs, i.e. my mouse became sluggish then after about 10 secs it stopped moving completely. Music player stopped a few secs after that and all UI was frozen. I found I could log in remotely via ssh from my laptop and found skype (v4) at the top of the run queue. Killed it (using default kill TERM) but system remained frozen. Found a heap of the following messages in dmesg/kern.log:

[ 1413.350588] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/2 Mthd 0x08d0 Data 0x00000000
[ 1413.350606] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/2 Mthd 0x08d4 Data 0x0000029f
[ 1413.350624] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/2 Mthd 0x08d8 Data 0x00000000
[ 1413.350641] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/2 Mthd 0x08dc Data 0x0000042e
[ 1413.350660] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/2 Mthd 0x0110 Data 0x00000000
[ 1413.350679] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/2 Mthd 0x088c Data 0x00000000
[ 1413.350697] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/2 Mthd 0x08b0 Data 0x000002a0
[ 1413.350715] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/2 Mthd 0x08b4 Data 0x00000487
[ 1413.350733] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/2 Mthd 0x08b8 Data 0x00000001

and then

[ 2638.295874] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2638.513969] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2638.731928] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2638.950519] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2639.168399] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2639.386458] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2639.604410] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2639.822783] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2640.047224] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2640.265387] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2640.483451] [drm] nouveau 0000:01:00.0: nv50cal_space: -16
[ 2640.701437] [drm] nouveau 0000:01:00.0: nv50cal_space: -16

Another odd thing which I also noticed the previous time this happened - I planned to grab the Xorg.0.log.old file after I rebooted but found that the nouveau driver seems to die (or get killed) and then restarted during my system startup, i.e. Xorg.0.log.old contains:

[ 32.513] (II) NOUVEAU(0): NVLeaveVT is called.
[ 32.520] (II) NOUVEAU(0): Closed GPU channel 2
[ 32.540] ddxSigGiveUp: Closing log
[ 32.540] Server terminated successfully (0). Closing log file.

and then the system comes up ok logging in Xorg.0.log. So I lose the real "previous" log file. Even though this happens, there is no abnormal screen "glitch" noticeable to me as the system comes up. Hope all this makes sense and I hope posting this here is appropriate even though I am using xorg-edgers ppa? Looks like an xorg/nouveau bug - but may be something brought on by skype 4?

I should add that, like the OP, I am using dual monitors: 2 by 1900x1200.

[Brendan Brigham (brigham)]

Mark, can you use the latest Nvidia drivers ? For the longest time Nvidia drivers were problem after problem but they have since fixed them and now work flawlessly, I haven't had a freeze since installing their latest.

[markb (mark-blakeney)]

@brigham, I have used nvidia drivers with my system (8600 GT) for a few years without problems but since ubuntu 12.04 (i.e. nvidia v295.33, v295.40, v295.49, v295.53, v302.07, v302.11, v302.17?), I suffer an xorg segmentation fault within a hour or two of booting when using nvidia driver (and that is tried with all these versions with and without xorg-edgers ppa enabled). See https://bugs.launchpad.net/ubuntu/+source/nvidia-graphics-drivers/+bug/973096. Hence that is why I am now running nouveau. Note I have run nouveau with Unity 3D fine since 12.04 release except only recently when these nouveau lockups started (could be around the release of skype 4, or that new xorg-edgers ppa kernel?). Also, note I mainly use Gnome Shell (all day every day) and have never seen this freeze or any other problems. Gnome Shell is a 3D desktop also so seems Compiz + nouveau related?

[Brendan Brigham (brigham)]

Ok, I am using 302.17 now with no issues, looks like the rabbit hole is getting deeper :s

[markb (mark-blakeney)]

Well this just happened again to me with same PFIFO_CACHE_ERROR nouveau errors in dmesg. Although this time it happened on gnome-shell, which is my normal desktop where I have never seen this before. I have only had this happen on those rare times I try Unity. I had updated earlier today and noticed my kernel went from linux-image-generic-pae = 3.5.0.2.2 -> 3.5.0.3.3 and xserver-xorg-video-nouveau = 1.0.1+git20120628.0d2b414a-0ubuntu0sarvatt~precise -> 1.0.1+git20120702.0d2b414a-0ubuntu0sarvatt~precise. At least we now know it is not specific to Compiz/Unity. Looks like a kernel and/or nouveau issue. Skype was running and I had music playing (via guayadeque) which I think has been true every other time this has happened.

[markb (mark-blakeney)]

So I decided to use Unity for a while. Locked up twice within 10 mins, second time without skype 4 running so not skype related. Unity/Compiz brings on this problem MUCH more than gnome shell/mutter.

[markb (mark-blakeney)]

Well I have determined that for me this problem is caused by linux kernel 3.5.0-3 (pulled in via linux-image-generic) which was installed on my system via xorg-edgers ppa. I removed this and went back to the generic precise kernel which is currently 3.2.0-26 and my lockups have gone. I am still using xorg-edgers ppa, just back-graded that kernel version.

[markb (mark-blakeney)]

I don't know if anybody is following my story here but I noticed an xorg-edgers kernel update from 3.5.0.3.3 to 3.5.0.4.4 today so I thought I would give it a test running Unity. Note, as my previous comment I can run kernel 3.2 forever without issue. So I upgraded and rebooted to 3.5.0.4.4 and I still see PFIFO_CACHE_ERROR nouveau errors but this time my system does not lock up. I don't see any UI problems. Here are the dmesg logs:

[ 6392.463172] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/0 Mthd 0x0060 Data 0x8000000f
[ 6610.102638] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/0 Mthd 0x0060 Data 0x8000000f
[ 6625.011826] [drm] nouveau 0000:01:00.0: PFIFO_CACHE_ERROR - Ch 4/0 Mthd 0x0060 Data 0x8000000f

So kernel 3.5.0.4.4 has improved things, but there still seems a problem there.

[Oibaf (oibaf)]

Is this still an issue?

[Brendan Brigham (brigham)]

Not any longer for me. I've fought with it too long without any true fix and have since abandoned Ubuntu 12.02 for Linux Mint 13 with zero problems.

[Launchpad Janitor (janitor)]

[Expired for xserver-xorg-video-nouveau (Ubuntu) because there has been no activity for 60 days.]