NVidia module causes linux kernel crash?

[lachlan]

I am getting occasional kernel crashes (once or twice a day!) - not fun at all. It happens randomly. It doesn't seem to matter if I am stressing the CPU, using sound, or anything.

I see in /var/log/messages that it says this:

Feb 1 14:17:03 localhost kernel: ------------[ cut here ]------------
Feb 1 14:17:03 localhost kernel: kernel BUG at vmscan.c:545!
Feb 1 14:17:03 localhost kernel: invalid operand: 0000
Feb 1 14:17:03 localhost kernel: snd-seq-midi snd-emu10k1-synth snd-emux-synth
snd-seq-midi-emul snd-seq-virmidi snd-seq-oss snd-seq-midi-event snd-seq snd-pcm
-oss snd-mixer-oss nvidia snd-em
Feb 1 14:17:03 localhost kernel: CPU: 0
Feb 1 14:17:03 localhost kernel: EIP: 0060:[<c013985c>] Tainted: P
Feb 1 14:17:03 localhost kernel: EFLAGS: 00010206
etc...

I can see this from dmesg:

0: nvidia: loading NVIDIA Linux x86 NVIDIA Kernel Module 1.0-5336 Wed Jan 14 18:29:26 PST 2004
Linux agpgart interface v0.99 (c) Jeff Hartmann
agpgart: Maximum main memory to use for agp memory: 941M
agpgart: unsupported bridge
agpgart: no supported devices found.
0: NVRM: AGPGART: unable to retrieve symbol table

My system info is as follows:

I am running kernel version: 2.4.20-8
With a "Fancy Force" (MX440SE 64MB AGP Video Card)
ASUS A7N8X-X motherboard (which has lame onboard sound that I am not using (disabled at the bios level) - I am using SB Live! instead)
AMD 2600
1GB of RAM

[lachlan]

Forgot to attach my /var/log/XFree86.0.log for the above problem

[lachlan]

This thread is related, I believe:

https://bugzilla.redhat.com/bugzilla...g.cgi?id=73733

[whig]

Instead of using the nvidia driver try the nv or vesa one for a few hours. You could do this while you go about your usual 2D work. If the crashes keep coming then it's not a nvidia related.

[zander]

In any case, it would be helpful if you provided the complete trace and more information on the kernel you're using (I'm assuming it's a RedHat kernel?).

[lachlan]

It is a fresh install of Redhat 9.0 (3 months old), and I haven't upgraded or modified the kernel in any way since then. So, as I said, the kernel version is 2.4.20-8.

I was originally blaming the probems on KDE, but after switching to a thinner window manger (icewm), the crashes kept coming, so I investigated further.

Here is the complete trace of that crash:

Feb 1 14:17:03 localhost kernel: ------------[ cut here ]------------
Feb 1 14:17:03 localhost kernel: kernel BUG at vmscan.c:545!
Feb 1 14:17:03 localhost kernel: invalid operand: 0000
Feb 1 14:17:03 localhost kernel: snd-seq-midi snd-emu10k1-synth snd-emux-synth
snd-seq-midi-emul snd-seq-virmidi snd-seq-oss snd-seq-midi-event snd-seq snd-pcm
-oss snd-mixer-oss nvidia snd-em
Feb 1 14:17:03 localhost kernel: CPU: 0
Feb 1 14:17:03 localhost kernel: EIP: 0060:[<c013985c>] Tainted: P
Feb 1 14:17:03 localhost kernel: EFLAGS: 00010206
Feb 1 14:17:03 localhost kernel:
Feb 1 14:17:03 localhost kernel: EIP is at refill_inactive_zone [kernel] 0x430
(2.4.20-8)
Feb 1 14:17:03 localhost kernel: eax: 010c0048 ebx: c1b7ff84 ecx: c0303900
edx: 00000000
Feb 1 14:17:03 localhost kernel: esi: 0000e4dc edi: c1b7ff68 ebp: 0000e4dd
esp: c34b1f9c
Feb 1 14:17:03 localhost kernel: ds: 0068 es: 0068 ss: 0068
Feb 1 14:17:03 localhost kernel: Process kswapd (pid: 5, stackpage=c34b1000)
Feb 1 14:17:03 localhost kernel: Stack: c34b1fbc c013a716 c03039f4 c0303a74 000
00033 fffffffe 00000006 0000000d
Feb 1 14:17:03 localhost kernel: 00000a9f 00000001 c34b0000 c0303900 c34
b0305 00000000 c013a45c c0303900
Feb 1 14:17:03 localhost kernel: 00000006 00000033 c013a368 00000000 000
00000 c0107389 00000000 00000000
Feb 1 14:17:03 localhost kernel: Call Trace: [<c013a716>] wakeup_memwaiters [
kernel] 0xde (0xc34b1fa0))
Feb 1 14:17:03 localhost kernel: [<c013a45c>] kswapd [kernel] 0xf4 (0xc34b1fd4)
)
Feb 1 14:17:03 localhost kernel: [<c013a368>] kswapd [kernel] 0x0 (0xc34b1fe4))
Feb 1 14:17:03 localhost kernel: [<c0107389>] kernel_thread_helper [kernel] 0x5
(0xc34b1ff0))
Feb 1 14:17:03 localhost kernel:
Feb 1 14:17:03 localhost kernel:
Feb 1 14:17:03 localhost kernel: Code: 0f 0b 21 02 41 61 25 c0 e9 67 fe ff ff 8
d b4 26 00 00 00 00

...thanks for your help.
-Lachlan

[lachlan]

Well, I switched back to "nv". So far so good. Normally it takes between 1-8 hours for a kernel crash. I have been going strong for a good 15 minutes now. I'll report back after the first crash or 2 days of running without one.

[lachlan]

Well, I have been running with "nv" instead of "nvidia" for about 24 hours now, and I haven't had a kernel crash yet. It is too bad that I will probably have to give up on "nvidia" because "nv" seems slower.

[whig]

Was it crashing using the latest nVidia drivers? Yes, it is probably the first thing you covered but I'm just checking.

For what it's worth I was sometimes getting freezes closing down X. This hasn't happened at all since I put on 2.4.23+. I note your kernel is older than that. Time for a kernel upgrade?

[lachlan]

Yes, I was using the latest nvidia driver (1.0-5336). Maybe a kernel upgrade would help, I don't know.

[lanzz]

i'm experiencing regular crashing too, two or three crashes EVERY day, kinda feels like windows. crashes always happen while something is scolling, so i assume it is a 2d-acceleration issue. crashes come in many flavors - kernel panic (complete freeze with blinking leds), whole X suddenly dying and dropping back to console (i don't run X from init, i run it by hand with startx), various applications randomly crashing with either segmentation fault or illegal instruction (so i guess the nvidia module ****s up the kernelspace pretty badly). doing "ls -lR /" on a disk with about 15gigs used is NEVER able to finish if you watch it - segfaults somewhere on the way. running it in a terminal and keeping it out of visibility (to avoid actual on-screen scrolling) finishes without any problems whatsoever, every time. overall stability deteriorates with time, with applications crashing more and more often as scrolling happens (i try to keep down the scrolling these days, jumping pages at a time, but that can't be done all the time and feels kinda stupid). cannot switch to the opensource nv driver as i could not get it to run properly - all that is visible after starting X is a whole screen filled with regularly spaced vertical blue lines.

card is gf4 mx440 agp8x (though i keep agp disabled, in naive hopes of increased stability)
kernel 2.4.21-ac1
xfree86 4.3.0
athlon xp 2400+, 512mb ram

[lachlan]

Ok, 48 hours now without a crash using nv instead of nvidia, I think it is clear:

The [1.0-5336] nvidia driver corrupts the [2.4.20-8] linux kernel.