Mysterious Freeze

Message

iplayfast · Post by **iplayfast** » Wed Dec 08, 2004 7:20 am

I think I've narrowed it down. I was emerging openoffice and it kept stalling. Not dieing but just slowing down. After discovering I could ctrl-Z then fg to kick start it, I went to bed. The next day it was sitting at copying (one of many parts of the build). The ctrl-Z trick didn't work, and I had to reboot.

I then cleaned my Reiser4 filessystem to elimiate that possiblty (there actually were problems with it).

Then I did ebuild openoffice(whatever) compile so it would not loose the work previously done. It carried on but whenever it came to copy it would sit for a long period. I then tried doing a sync and presto it hung in the slowly degrading fashion we've come to love.

Several more experiments showed that during the "copying" part of the install, ps x would show the install with a "D" status, which the man page says is sleeping waiting for I/O. When I do sync, the Sync shows "D+" which I assume means that it's waiting on the previous process.

I'm now guessing it's a kernal bug, dealing with disk io, or a Reiser4 bug. I'm using kernel Linux version 2.6.10-rc1-mm5 (gcc version 3.4.3 (Gentoo Linux 3.4.3, ssp-3.4.3-0, pie-8.7.6.6)) #5 SMP Wed Nov 17 00:
50:37 EST 2004

Anyone with similar configurations?

BlinkEye · Post by **BlinkEye** » Wed Dec 08, 2004 8:00 am

this sure sounds like a kernel reiser4 bug. i used reiser4 for quite some time. the only kernel which worked properly were ck-sources. there's a problem now: the new ck-sources haven't included the reiser4 patch anymore, but try the nitro sources instead. i had a lock up of the system every time when using patched gentoo-dev-sources and copying or moving large files

Dolio · Post by **Dolio** » Wed Dec 08, 2004 8:17 am

It can't be reiser4 specifically. I don't have any reiser filesystems on my machine anymore (used to have reiser3, never reiser4), and they're turned off in my kernel config.

It could be disk io related, I suppose. I'm on 2.6.9-nitro2, gcc 3.3.4-r1. Single processor, no preempt. In other words, not really even close to your config, I guess.

BlinkEye · Post by **BlinkEye** » Wed Dec 08, 2004 8:59 am

i wouldn't say the problem is reiser4 for all of you. but for iplayfast i'd say it is the case

ikaro · Post by **ikaro** » Wed Dec 08, 2004 4:12 pm

i bet no one can make the box frezze if you are not using X.

iplayfast · Post by **iplayfast** » Wed Dec 08, 2004 4:35 pm

Well I went to the console prompt when my sync problem occured. That is, I had kdm waiting for me to log in, I switch to console, started the ebuild, found it dieing did the sync and it died.

Also I tried booting on an older kernel (sorry I don't recall which one) went into OpenBox and did it again, and it went all the way through. So I think, that at least for my case X isn't part of the problem.

ikaro · Post by **ikaro** » Wed Dec 08, 2004 5:53 pm

i mean, boot in INIT3 and make it crash.
dont start X.

BlinkEye · Post by **BlinkEye** » Wed Dec 08, 2004 6:16 pm

iplayfast wrote:Well I went to the console prompt when my sync problem occured. That is, I had kdm waiting for me to log in, I switch to console, started the ebuild, found it dieing did the sync and it died.

Also I tried booting on an older kernel (sorry I don't recall which one) went into OpenBox and did it again, and it went all the way through. So I think, that at least for my case X isn't part of the problem.

as said before: i assure you you have not the same problem at all. your's is simply kernel/reiser4 related. check the forums - you'll find numerous posts about freezing system while using reiser4 with the wrong kernel(-patch) and moving/copying/(un-)tar (big) files.

start a backup of your system (for example with my bash script for a stage4) and your system will freeze even without X running

NOTE: this post is dedicated to iplayfast, not to others having freezes too

ikaro · Post by **ikaro** » Wed Dec 08, 2004 6:32 pm

im not even using reiser4.
and btw, I have no problems when im _not_ using the nvidia drivers.

So in my case, the freezes are related to nvidia/kernel/xorg one of those or a mix, can't say.

I doubt very much that the freeze bug / or g.o.d bug ( grab of death ) is related alone to ReiserFS4.

I put my money in nvidia/kernel fuckup.

iplayfast · Post by **iplayfast** » Thu Dec 09, 2004 5:02 am

Looking at kernal trap, it sounds like the 2.6.9 series is very buggy. Some say to avoid it all together. 2.6.7 was much more stable. That has been my experience as well.

*edit*
I've updated my kernel to Linux version 2.6.10-rc2-mm4 and it successfully ran all night emerging about 60 packages. Maybe I've got this licked.... (famous last words)

ikaro · Post by **ikaro** » Thu Dec 09, 2004 4:29 pm

Were you using the nvidia drivers ? .. and did you left the computer running while you were sleeping or did you used the computer while it was compiling the 60 packages ?

btw, anyone knows of 1 single thread in the LKML about this issue ?

iplayfast · Post by **iplayfast** » Thu Dec 09, 2004 5:05 pm

I'm using xorg-x11 ATI drivers (9200se). I did some work and then left it. When I woke up did some more. It was still going when I left for work. (Starting another batch, lots of updates this week).

I'll have to wait and see how it pans out, but so far so good.

BlinkEye · Post by **BlinkEye** » Thu Dec 09, 2004 5:14 pm

go on ignoring me. the downside of your post is simply that you ARE misleading others. i personally wouldn't care. you had a reiser4/kernel bug not a X/nvidia bug - when do you start believing it?

iplayfast · Post by **iplayfast** » Thu Dec 09, 2004 6:08 pm

Sorry, didn't mean to ignore you. Just missed your post. I think you're right, in that my mysterious freeze isn't the same as others with nvidea problems.

How am I misleading others?!? I never claimed to have nvidea, and I originally thought my problem was agp related. I turned off agp and most of the freezes went away. I've updated my kernel and the rest seem to have disappeared. (BTW the updated kernel did have a Reiser4 fix in it as well as a whole slew of ide fixes)

BlinkEye · Post by **BlinkEye** » Thu Dec 09, 2004 6:14 pm

iplayfast wrote:Sorry, didn't mean to ignore you. Just missed your post. I think you're right, in that my mysterious freeze isn't the same as others with nvidea problems.

How am I misleading others?!? I never claimed to have nvidea, and I originally thought my problem was agp related. I turned off agp and most of the freezes went away. I've updated my kernel and the rest seem to have disappeared. (BTW the updated kernel did have a Reiser4 fix in it as well as a whole slew of ide fixes)

i'm sorry if i attacked you. i just wanted to make clear to others, that you have a different problem and that upgrading kernel wouldn't help. glad you could solve your problem though, there's nothing worse than an unstable system!

bunder · Post by **bunder** » Thu Dec 09, 2004 6:38 pm

has anyone sent in a complete bug report? if not, here's what I think we've come across:

Xfree or Xorg (verified)
Nvidia or ATI or other (verified)
GL/GLX[DRI?] dependant (verified)
2.4 or 2.6 kernel (verified)
Pre-emp on or off (verified)
AGPGART or DRM or none (verified)
udev or devfs (verified)
[IO]APIC on or off (verified)
any filesystem
flags don't seem to matter
can occur when running a GL/GLX app [including screensavers, games and even glxgears], but also under regular use
can cause hard lockups, and lockups that can be fixed with the magic sysrq

am i missing anything here?

I've turned off FW/SBA in the nvidia module, but SBA still seems to be enabled (motherboard?). I'm going to do some hard gaming tonight and see how things are going with this setup.

Dolio · Post by **Dolio** » Fri Dec 10, 2004 12:23 am

FW/SBA aren't an issue. They aren't supported on my motherboard and it still happens.

It still happens whether IOAPIC is on or off for me.

I can run GL/GLX apps just fine, and it occurs when I'm not using any GL/GLX apps at all.

korngerd · Post by **korngerd** » Fri Dec 10, 2004 12:56 am

Include me into the bunch of *annoyed* gentoo users:

xorg-x11-6.8.0-r3
gentoo-dev-sources-2.6.9-r9
nvidia-glx-1.0.6111
nvidia-kernel-1.0.6111-r3
udev-045

I have GLX enabled, and am running with the nvidia driver in X (obviously). Strange thing, though, is that I can have GLMatrix (from XScreensaver) running in the background at GDM login forever (meaning the system seems to be stable even if glmatrix is drawing the background for gdm - hence, it's using the glx extension). Once I login and start some intensive apps (firefox + xscreensaver), the system crashes. It first starts off with a "soft" crash, then ends up with a "hard" crash (although I can SSH into the machine from outside). I'm guessing that the nvidia driver has some major flaws in it...

bunder · Post by **bunder** » Fri Dec 10, 2004 1:47 am

okay, i edited it. my lockups only appear when running gl/glx apps, although i used to have them on my athlon xp system when running firefox/gtv/kmplayer.

SnEptUne · Post by **SnEptUne** » Fri Dec 10, 2004 12:32 pm

I have modified my xorg.conf to use software cursor and the freezing ceased; I can run glxgears and moves my mouse around however I want without locking up my system. It certainly doesn't solve the problem, but at least it make locking up due to mouse wheel or mouse movement very unlikely. It is like cutting toes to avoid worms until someone come up with a permanent solution.

Code: Select all

Section "Device"
    Identifier  "ATI Radeon VE (RV100)"
    Driver      "radeon"
    Option      "HWCursor"              "false"
    Option      "SWCursor"              "true"
    VideoRam    65536
EndSection

ikaro · Post by **ikaro** » Sat Dec 11, 2004 4:32 pm

I tried unpluging the mouse and use the box for a while ... well 5 minutes after it crashed.

So no mouse makes it crash too, and therefor changing from HW to SW mouse cursor is irrelevant.

joehni · Post by **joehni** » Sat Dec 11, 2004 10:11 pm

Suffer from the same problem. Locks deterministic after leaving a GLX app e.g. glxgears or matrix screen saver. Tried meanwhile to switch between agpgart and Nvidia's support, but no difference. I have also sporadic locks when I scroll a window, most of the time in the browser using the mouse, but occurs also in gvim using the cursers

kernel 2.4.26
nvidia 1.0.6111-r3

bash-2.05b# cat /proc/driver/nvidia/agp/status
Status: Enabled
Driver: NVIDIA
AGP Rate: 8x
Fast Writes: Enabled
SBA: Enabled

korngerd · Post by **korngerd** » Sun Dec 12, 2004 6:04 am

Tried disabling IO_APIC by disabling SMP in the kernel (SMP hyperthreading in Processors area). No luck

grenouille · Post by **grenouille** » Sun Dec 12, 2004 3:52 pm

Just to add myself here...

And I just noted it only happens to me when I use gnome 2.8, not when I use pekwm, no idea if it is relevant...

gspr · Post by **gspr** » Sun Dec 12, 2004 5:12 pm

Heh, seeing this ever growing flock of people, I must truly laugh at nVidia's attempt to let us all think we have faulty hardware and should go out and buy SwineForce MegaSuper720!