[BUG] Regular Freezes - NVRM: Xid (0000:02:00)

[Neurotic]

So I'm running Ubuntu 11.10, 64 bit, and I wanted to dig into why my computer freezes a few times a day at relatively random intervals.

I've messed around with the drivers that are available to Ubuntu 11.10, and am currently on 280.13. 285 was a bit more unstable I found, and 290.10 was just plain unusable it crashed on me so frequently.

Doing some digging, I could clearly see that when my system froze, I would generally see "NVRM: Xid (0000:02:00) ", or some variant thereof in my logs.

These do defintely show up from time to time in the logs, and some I think don't actually cause a crash - the system seems to recover okay.

That being said, there was definitely one at Dec 1 13:42:23, when I had the following output:

syslog:Dec  1 13:42:23 delusion kernel: [ 2347.122286] NVRM: Xid (0000:02:00): 56, CMDre 00000007 00000000 ffffffff 00000001 00000000
syslog:Dec  1 13:42:23 delusion kernel: [ 2347.122301] NVRM: Xid (0000:02:00): 56, CMDre 00000008 00000000 ffffffff 00000001 00000000
syslog:Dec  1 13:42:23 delusion kernel: [ 2347.122330] NVRM: Xid (0000:02:00): 56, CMDre 00000007 00000000 ffffffff 00000001 00000000
syslog:Dec  1 13:42:23 delusion kernel: [ 2347.122365] NVRM: Xid (0000:02:00): 56, CMDre 00000007 00000000 ffffffff 00000001 00000000
syslog:Dec  1 13:42:23 delusion kernel: [ 2347.122402] NVRM: Xid (0000:02:00): 56, CMDre 00000007 00000000 ffffffff 00000001 00000000
syslog:Dec  1 13:42:23 delusion kernel: [ 2347.122430] NVRM: Xid (0000:02:00): 56, CMDre 00000007 00000000 ffffffff 00000001 00000000
syslog:Dec  1 13:42:23 delusion kernel: [ 2347.122454] NVRM: Xid (0000:02:00): 56, CMDre 00000007 00000000 ffffffff 00000001 00000000
syslog:Dec  1 13:42:23 delusion kernel: [ 2347.122489] NVRM: Xid (0000:02:00): 56, CMDre 00000007 00000000 ffffffff 00000001 00000000
syslog:Dec  1 13:42:23 delusion kernel: [ 2347.122530] NVRM: Xid (0000:02:00): 56, CMDre 00000007 00000000 ffffffff 00000001 00000000
syslog:Dec  1 13:42:23 delusion kernel: [ 2347.122566] NVRM: Xid (0000:02:00): 56, Dec  1 13:45:01 delusion kernel: imklog 5.8.1, log source = /proc/kmsg started.

There is definitely a bug here somewhere, as I'm not the only one with this issue, and it happens on a regular basis.

Also, my Nvidia cards (2 x GTX 280M, which have the hardware SLI capabilties, but only 1 is used in Linux) where replaced earlier this week, so I know it's not the hardware.

I'm attached a grep of the NVRM issues from my syslog to this post, as well as the usual nvidia-bug-report files.

Let me know if I can provide any more details to help fix this issue. It's seriously driving me nuts, and I never had this issue when I was on Ubuntu 10.04.

[Neurotic]

Just had a complete screen turn off.

Dec  1 12:31:25 delusion vmnet-dhcpd: DHCPREQUEST for 192.168.169.128 from 00:0c:29:e0:17:38 via vmnet8
Dec  1 12:31:25 delusion vmnet-dhcpd: DHCPACK on 192.168.169.128 to 00:0c:29:e0:17:38 via vmnet8
Dec  1 12:43:17 delusion kernel: [12783.796192] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Dec  1 12:43:19 delusion kernel: [12785.796303] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Dec  1 12:44:31 delusion kernel: [12857.800220] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Dec  1 12:44:33 delusion kernel: [12859.800329] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Dec  1 12:46:00 delusion vmnet-dhcpd: DHCPREQUEST for 192.168.169.128 from 00:0c:29:e0:17:38 via vmnet8
Dec  1 12:46:00 delusion vmnet-dhcpd: DHCPACK on 192.168.169.128 to 00:0c:29:e0:17:38 via vmnet8
Dec  1 12:46:17 delusion kernel: [12963.804203] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Dec  1 12:46:19 delusion kernel: [12965.804337] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context
Dec  1 12:59:03 delusion vmnet-dhcpd: DHCPREQUEST for 192.168.169.128 from 00:0c:29:e0:17:38 via vmnet8
Dec  1 12:59:03 delusion vmnet-dhcpd: DHCPACK on 192.168.169.128 to 00:0c:29:e0:17:38 via vmnet8

[ars]

The problem is with the Kerneal 3.1 and the 290 drivers. try with Knernel 3.0-

Here the same problem:http://www.nvnews.net/vbulletin/showthread.php?t=169027

[named]

I had this with kernel 3.0.4 and nvidia 280.13. There was some discussion in this thread: http://www.nvnews.net/vbulletin/showthread.php?t=166069. Unfortunately nobody from nvidia seems to care.

[Neurotic]

Originally Posted by ars

The problem is with the Kerneal 3.1 and the 290 drivers. try with Knernel 3.0-

Here the same problem:http://www.nvnews.net/vbulletin/showthread.php?t=169027

Thanks for the tip, except I'm on 3.0.0-14-generic

I recently updated to 285.05.09 to see if that fixes the problem. It seems a bit more stable, but I am still getting crashes.

I will try 290.10, but from memory it was really unstable.

From there, I'm going to start rolling backwards, and see if I can find a stable driver.

Annoying we haven't heard from Nvidia - this is a problem that seems to be hitting *alot* of users. I'm seeing lots of bug reports across the Ubuntu bug tracker and in the forums.

[vojta]

I had this problem too, I even tried nvidia-drivers 256.63 (with xorg-server 1.9.5), they seem to not suffer with these freezes. However, when I was watching a movie with 256.53 it was not much smooth. I had to switch to nouveau for a while.

[cheechr1]

Ive found 275.09.07 to be the most stable. Any driver afterwards will crash my Ubuntu on a regular basis. I am on kernel 2.6.38-12.

[vojta]

Originally Posted by cheechr1

Ive found 275.09.07 to be the most stable. Any driver afterwards will crash my Ubuntu on a regular basis. I am on kernel 2.6.38-12.

I'm going to give it a try...

EDIT: Still freezing with this message in /var/log/messages:

[  443.046635] NVRM: Xid (0000:01:00): 8, Channel 00000001
[  577.265186] NVRM: Xid (0000:01:00): 8, Channel 00000001
[  579.261687] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context

I hope they will track this issue down and fix it soon.

[Neurotic]

Thanks for the heads up.

I've switched over to just using nouveau for now. I don't need the 3d acceleration for now, so at least my computer isn't crashing. Or at least yet

Originally Posted by cheechr1

Ive found 275.09.07 to be the most stable. Any driver afterwards will crash my Ubuntu on a regular basis. I am on kernel 2.6.38-12.

[anjal]

Originally Posted by Neurotic

Thanks for the heads up.

I've switched over to just using nouveau for now. I don't need the 3d acceleration for now, so at least my computer isn't crashing. Or at least yet

Welcome to the Club!
I have (very) old hardware on the supposedly "open source" (but obfuscated in reality) "nv" driver.
Installing a few different versions of the 96.* drivers does not make any difference.
It gives the same NVRM: Xid (0000:02:00).

I have posted recently with detailed logs, but was not expecting any answer from nVidia. Going to attach some more logs related to a few more experiments I have made.
Thought maybe someone from the community came up with something.

But, it was a long shot since similar problems have existed on different platforms for the past 5 years and many have posted over the period but doesn't look like nVidia has a clue on what's happening or, I would say they basically don't care.

The problem there is not much information on the error class either.

[Neurotic]

It's funny, up until recently, I specifically bought Nvidia cards for my laptops, as they had good support for them with their drivers on Linux.

Only on this driver update have things gone totally pear shaped. I've never had a problem before.

[dr_barnowl]

Originally Posted by cheechr1

Ive found 275.09.07 to be the most stable. Any driver afterwards will crash my Ubuntu on a regular basis. I am on kernel 2.6.38-12.

That's depressing, but it might explain why Natty was always dead stable (only had older drivers) and Oneiric has been crashing like a drunken teenager (only has driver 280 and after available). I think I was forced to upgrade my driver when I got my 560Ti because the older driver didn't support the screen mode I wanted.

I saw on another thread that they found 290 even less reliable. I've switched back to 280, in the hope that this is true - I had switched forward in the hope it would improve things. Alas, there are no packages of 275 for Oneiric.

The problems seem to persist after a reboot too ; my computer got to the point where it refused to even boot Plymouth (the Ubuntu pretty bootloader) (it would boot the loader on a LiveUSB but not any further - I don't think it could initialize the card). Windows would boot fine though...

I gave the hardware a good thrash with GIMPS mprime and Memtest86 and the CPU and RAM seem to be fine. The system isn't overclocked, conservative settings all round in the BIOS.

I've seen ; segfaults, hard lockups, and hard resets.

The problems seem to accumulate across reboots to the point where the only thing you can do is leave your computer with the power off for a while. It's almost like the problems leave residual state in your system RAM or the RAM on the card, and you have to wait for the memory cells to discharge completely to clear it away.

The frequency of problems varies - I've had 8 hours of runtime today without a problem, then BAM, hard lockup and it starts sulking and can't manage more than a few minutes of uptime.. Conversely, last Friday I struggled to get more than 5 minutes consecutive uptime, ending with the episode where it would refuse to boot Ubuntu but not Windows...

Windows merrily plays Skyrim and other games for hours on end and never locks up or hard resets. The only thing it DOES do that is odd - is that after using Ubuntu, when Windows starts up, it hard resets when Aero starts up. Then it boots again just fine and works properly. It's like the Linux driver puts the card into a bad state, then Windows clears it out (but crashes in the process...)

Honestly, until I mentioned it in an IRC channel and got a knowing "Ah, that'll be nvidia then", I thought my hardware was failing. But in some weird way that only affected Linux. Don't laugh - I've seen systems which failed the FMOV test in Memtest86 that Linux hated but Windows would run fine on.

I've dialled down the performance mode to the lowest for my card, and verified that the GPU temperature is 12 degrees lower (32, rather than 44 - neither very scary). This doesn't seem to have helped.

Since I started to write, the desktop has locked up again. All it's been doing is sitting there with a terminal and a GVim window (with my xorg.conf) open on screen. That's it. At one point, the screensaver may have kicked in. It can't even stay stable for 10 minutes (this is on the 280 driver).

Was lead here by seeing the NVRM: Xid messages in my kernel logs.

All I can do is whine about it here, it would seem.

Now I've rebooted and I'm trying to provoke it, by running youtube (it seems to especially hate flash video), it's running fine. I've got the kernel log tailed in a terminal and not a dicky bird. Maybe that's the key, like a watched pot..

I picked nvidia on the basis that I've always had good service out of their hardware with Linux, and that ATI cards (now AMD) were a bit finicky .. well, now I'm feeling a lot more affectionate towards Radeon. My (work supplied) laptop has an AMD GPU, and I've never had any difficulty using it with Ubuntu.