[Dapper] Firefox hard locks when RenderAccel is enabled

Relevant "lspci -v" output:

0000:01:00.0 VGA compatible controller: nVidia Corporation NV10DDR [GeForce 256 DDR] (rev 10) (prog-if 00 [VGA])
        Subsystem: Guillemot Corporation 3D Prophet DDR-DVI
        Flags: bus master, 66MHz, medium devsel, latency 248, IRQ 10
        Memory at dc000000 (32-bit, non-prefetchable) [size=16M]
        Memory at d0000000 (32-bit, prefetchable) [size=128M]
        Expansion ROM at dd000000 [disabled] [size=64K]
        Capabilities: <available only to root>

And "uname -r"
2.6.15-9-386

Thanks for your bug report. Which version of Ubuntu do you use?

Sorry. Dapper. Dist-upgraded from Flight 2.

I just realized that there doesn't seem to be any nvidia-legacy support for 2.6.15 kernels. I'm a bit confused as to how I was able to use the nvidia driver anyways...

I'm experiencing this bug with the latest Dapper as of Apr 23, 2006. However, it is not Firefox specific, it just occurs during times of high load, which happen to be when Firefox or a java app like Azureus are running. I'm cutting and pasting my descriptions from this forum thread: http://www.ubuntuforums.org/showthread.php?t=159483

For the last 3 months I have experienced unexplained lockups while using many applications but especially while using Firefox and Azureus. What happens exactly is the GUI will become totally unresponsive, but music and movies will typically continue to play. I can also move the mouse. But any amount of pressing CTRL ALT DEL or attempting to click the close button on the in-focus application (typically Firefox) will have no effect. I have been using Dapper since at least Flight-3 and have experienced this with increasing frequency since then. This week the rate of lockups has peaked to about 2 per day. At first install, the rate was about once every 2-3 weeks, maybe longer.

I am using an nvidia card and w/ the 'nvidia' drivers and RenderAccel enabled. I am NOT using XGL.

This is the last item present in my syslog before the machine restarts:
Apr 21 14:28:39 localhost kernel: [4296241.244000] NVRM: Xid (0001:00): 6, PE0000 04d0 00000000 0000fbdc dfff0000 00ffffff
Apr 21 14:28:39 localhost kernel: [4296241.258000] NVRM: Xid (0001:00): 28, L0 -> L0

I backed up my bookmarks, deleted by ~/.mozilla folder, started firefox, restored by bookmarks, and reinstalled only the extensions I use and now things were much smoother for a few days, presumably because Firefox was less load on the system. This lasted for about 2 days and then the crashes came back.

A user pointed out "A long shot. I've been having some similar lockups until I noticed that my swap partition hat "disapeared". Restored swap and no lockups since."

I replied "I checked and my swap partition is functioning fine. I am not using an underpowered machine by any means, I have one gig of ram and one gig of swap, on a 2ghz processor. There's no good reason for this."

I attempted to use the nvidia driver with RenderAccel turned to false. I experienced a lockup as I described while using Firefox within 5 minutes of logging in. Switching back to the nv driver and no lockups yet.

sudo lspci -v
0000:01:00.0 VGA compatible controller: nVidia Corporation NV43 [GeForce 6600/GeForce 6600 GT] (rev a2) (prog-if 00 [VGA])
        Subsystem: XFX Pine Group Inc.: Unknown device 2119
        Flags: bus master, 66MHz, medium devsel, latency 248, IRQ 185
        Memory at fd000000 (32-bit, non-prefetchable) [size=16M]
        Memory at b0000000 (32-bit, prefetchable) [size=256M]
        Memory at fc000000 (32-bit, non-prefetchable) [size=16M]
        Expansion ROM at feae0000 [disabled] [size=128K]
        Capabilities: [60] Power Management version 2
        Capabilities: [44] AGP version 3.0

uname -a
Linux ubuntu 2.6.15-20-k7 #1 SMP PREEMPT Tue Apr 4 19:26:13 UTC 2006 i686 GNU/Linux

"Whether this is Firefox 1.5 or the nvidia-legacy driver is beyond me..."

I have this issue also in the 64bit release of 6.06 BETA. It completly locks the system when I try to type somethign in the firefox address bar.

I have an ATI X800PRO, which implies that this problem is not only a nvidia problem.

I believed this to be a problem related to my arcane nvidia card, so I bought a new one, but it still shows the same symptoms.

The new card is a Winfast A340 with an nVidia FX 5200 chipset (rev. a1).

I can typically reproduce this by alt-tabbing with a mozilla window open somewhere.

As dguido points out, the music keeps on playing and I can move the mouse.

But my sure way to reproduce is to alt-tab with mozilla open.

After spending a few days on the 'nv' driver and seeing a few updates go by I switched back to the 'nvidia' driver today.

I stressed tested my box quite a bit since I switched back, with heavy use of Firefox, Azureus, and a number of other processes that frequently brought my CPU and memory use up to 100%. No crashes.

The issue appears to be fixed for me.

Just dist-upgraded. I've found out that the problem is even worse than I first expected.

Firstly; it has nothing to do with RenderAccel. I just set RenderAccel to "False" in my xorg.conf, and I can still reproduce.

Which leads me to my next point. I can reproduce this without involving firefox at all. Just opening a few Nautilus windows and alt-tabbing a few times, freezes my box as described previously.

Maybe out problems are infact different after all dguido. Perhaps I should open a bug report on my specific issue...

The Nvidia drivers has worked perfectly in every other distro I've been running on this box, so it has to be something in the Ubuntu stack.

Another note:
If I run a failsafe session and just run "metacity" and "nautilus" from a terminal I cannot reproduce...

I had one strange lock-up yesterday, but it didn't seem to be related to my Nvidia issue. Though, other than that I have experienced no crashes since Sunday night (about 2 1/2 days).

I'll declare I'm free of this problem if I still haven't crashed by Friday (3 more days).

(I also found that my problem had nothing to do with RenderAccel, after testing it on and off. I also found that this was not directly tied to Firefox, but conditions of high load/memory use.)

I crashed today while in screensaver. syslog lists:

May 3 12:55:14 localhost kernel: [4380226.485000] NVRM: Xid (0001:00): 6, PE0000 0490 01e30231 0000fcfc fffbe304 02fb00cd

Looks like it's not gone, just slightly abated.

When my display crashes I can still log in to the box over SSH and do a 'sudo shutdown -r now' to cleanly reboot the box. Keyboard is unresponsive though, so CTRL ALT DEL doesn't work.

There are a couple of threads in the forums about the bug(s?). One such thread is at http://www.ubuntuforums.org/showthread.php?t=193283 .
Here's the summary:

Problem occurs with the nv driver and the official nvidia drivers. There is also an ATI user having the same issues, so it doesn't look like a driver issue. It's not a Firefox issue as the exact same binaries on a Mandriva system (and same hardware) do not produce the problem. Someone else reported the same thing running both Dapper and FC5. FC5 doesn't have the problem, Dapper does. The problem is also reported with Thunderbird, VMware, and even Opera. A few other programs were mentioned too.

I have swap working, so it's not a swap problem (although not having swap working can be a problem!). On thing I have done and I've not seen reported is I've kept top running in a terminal. I notice that the amount of memory xorg uses slowly climbs and never comes back down, even if I close Firefox. It climbs every time Firefox does a "page refresh" (on it's own, BTW).

I left Firefox open all night with only a simple static page. At the start, xorg was using a little less than 2% of my memory. By morning, it was up to 39% (as reported by top)! No flash or Java on the page either and only a few terminals running in the background (one with top).

I'm thinking it is a bug in xorg, maybe a memory leak. I had this problem in Breezy as well as Dapper (what I'm running now).

I have this issue too. I can confirm it is a Xorg issue, since if I am running Xgl, I have NO problems at all, but as soon as I switch to Xorg I get the lockups.

I have this problem on my two computers, one being a PIII 600Mhz with a Geforce2 GTS (nvidia-legacy) and the other a P4 2Ghz with Geforce2 MX400 (nvidia-glx). Running with the nv driver I had not experienced this problem yet.

It seems to occur at random, not related to any specific app. I had seen this occur mainly on Firefox and Nautilus tho.
The freeze seems to lock the keyboard but not the mouse inputs, it also locks the whole desktop, but pressing the shutdown button on the case causes the system to cleanly shutdown. I have also seen some instances where it completely locks the whole system, tho more rarely.

Some relevant nVidia (nVNews) forum threads:
http://www.nvnews.net/vbulletin/showthread.php?t=47502
http://www.nvnews.net/vbulletin/showthread.php?t=49117

For me, it really appears to be related to X somehow...

RenderAccel in the nvidia-legacy drivers was always experimental which is why it defaulted to off. From ftp://download.nvidia.com/XFree86/Linux-x86/1.0-7174/README.txt :
"Enable or disable hardware acceleration of the RENDER extension. THIS OPTION IS EXPERIMENTAL. ENABLE IT AT YOUR OWN RISK. There is no correctness test suite for the RENDER extension so NVIDIA can not verify that RENDER acceleration works correctly. Default: hardware acceleration of the RENDER extension is disabled." So if you are getting crashes enabling it the legacy drivers I'm afraid that's to be expected. NVIDIA fixed RenderAccel in later releases of their drivers.

If NVRM: Xid is being printed to the console or is turning up in /var/log/messges you definitely have an issue with the binary drivers. NVRM: Xid messages are output by the binary driver basically saying "something unexpected happened" but only NVIDIA can really make sense of them (they can indicate genuine driver problems as well as troublesome hardware).

This bug has become confused. It is not clear whether everyone is really seeing the same issue. I would say that if you are not using a legacy card then you are probably better off not posting to this bug but instead opening an issue of your own.

Mikkel:
Is the issue still happening under Feisty? Are any NVRM: Xid message being printed in dmesg/ /var/log/messges ?

Setting to needsinfo pending reply from Mikkel.

We need confirmation that this issue still exists in Feisty. Can someone please test it with a more recent version of Ubuntu?

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