[SOLVED] Hard freeze upon KDE boot

cfgauss · Tux's lil' helper Joined: 18 May 2005 Posts: 145

During the last 48 hours my AMD64, 2.6.25-gentoo-r7, machine froze twice, the last time as soon as KDE was finished booting. The mouse and keyboard froze and I couldn't even ssh to it from another box. It acts as the gateway for other boxes at home and their Internet connection failed so my box's whole networking subsystem was down. When KDE boots, emacs 23.0.60.1, firefox 2.0.0.16, and thunderbird 2.0.0.16 boot automatically. There have been no hardware changes. This is the first time this has happened since the box was built three years ago.

I couldn't find anything in /var/log/messages which would give me a hint as to what happened. How can I debug this and, better yet, prevent it from happening again?

[EDIT] After 8 hours, memtest86 produced no errors so I finally opened the box. The BIOS reported the Northbridge fan was at 0 RPM and, sure enough, it wasn't moving. The MSI Forum reported that was enough for random crashes so I assume that's the culprit and ordered a replacement fan suggested on that forum. They claim that I can only expect about two years from any such fan. Many people order extras so they can swap them in on the spot.

Thanks for all the suggestions.[/EDIT]
[EDIT2] I installed a new Northbridge fan and it appears to be spinning. (Unfortunately it doesn't have a third wire so the RPM isn't reported by the BIOS.) But when I boot, my Gentoo Linux AMD64 froze within minutes. It's dual boot so I'm running Windows XP continuously. If it runs without a crash for three or so days, I'll assume that the hardware is OK.

Is the problem then a bad driver under Linux? If so, what's a good way of finding it so I can replace it or, perhaps, not use that piece of hardware?

I've love to get my box back again so any pointers are gratefully received.[/EDIT2]

[SOLVED] The frequency of freezing increased both under Linux and Windows. I built a new, spiffy Intel Core 2 Quad box with a new SATA hard drive. I copied over world, /home, and a bunch of config files from my old IDE hard drive, which still functions in the new box. After testing RAM, etc., in the old box, I suspect I had a failed Northbridge. Thanks for the hardware debugging pointers. [/SOLVED]

bunder · Posted: Mon Aug 18, 2008 1:06 am Post subject:

while being logged out, rm'ing /home/youruser/.kde/share/config/session/* should prevent kde from loading your apps when you log in. that won't solve the problem, but should tell you whether it's firefox/other that's causing the lockup.

some preliminary questions:
recent upgrade (kernel and software)?
ati or nvidia?
any services failing on boot?
maybe an emerge --info, lspci, dmesg?

cheers
_________________
bunder @ freenode | bunhax - tame your SMP!
gentoo shells | the remains of UFGO

Hu · Veteran Joined: 06 Mar 2007 Posts: 2595

Are you using the open source drivers for your video card or the proprietary ones from your card vendor? While gathering the information bunder wanted, please also post the output of nl /etc/X11/xorg.conf.

cfgauss · Tux's lil' helper Joined: 18 May 2005 Posts: 145

cfgauss · Tux's lil' helper Joined: 18 May 2005 Posts: 145

DirtyHairy · Posted: Wed Aug 20, 2008 3:46 am Post subject:

I wouldn't start searching for the cause in userland (à la firefox-bin), userland code that doesn't access hardware can't cause deadlocks without kernelspace bugs, and unless you also upgraded your kernel, I wouldn't suspect this to be the cause either. I more suspect the cause is related to X and the nvidia drivers. Try with the opensource nv or vesa drivers and see if the crashes disappear. Do you use any other external (not in-tree) kernel modules? What's the output of lsmod?

cfgauss · Tux's lil' helper Joined: 18 May 2005 Posts: 145

cfgauss · Tux's lil' helper Joined: 18 May 2005 Posts: 145

cfgauss · Tux's lil' helper Joined: 18 May 2005 Posts: 145

I left town for two days so I booted the box in single-user mode and just started the networking subsystem, net.eth0, net.eth1, sshd, and shorewall, so that it could act as gateway for other boxes at home. I.e. the box just had a console interface, no GUI.

It crashed while I was gone. I'm running memtest86 now to see if it reports memory errors.

I'm at a loss how to proceed from here. Do I need to run mprime to test memory? I have Windows XP on another partition. Do I run that for a while to see if it crashes? If it does, I presume there's some hardware failure but how do I determine which hardware needs to be replaced?

I don't have very much experience with debugging this kind of failure and I rely on this box to do my work. Any pointers on how to proceed will be gratefully appreciated.

Thanks.

DirtyHairy · Posted: Sun Aug 24, 2008 11:14 am Post subject:

Well, I'm no expert here either, but even if Windows run fine, it can still be a hardware error, windows has a reputation of being more tolerant with respect to hardware glitches. You could try pulling out and swapping memory to see if it helps.

cfgauss · Tux's lil' helper Joined: 18 May 2005 Posts: 145

cfgauss · Tux's lil' helper Joined: 18 May 2005 Posts: 145

I ran Windows XP for 12 hours. It then rebooted itself into Gentoo (grub's default) while I was away from the box for 30 mins. and immediately froze at the kdm login window. Since the rebooting seemed a fluke to me, I restarted XP and it ran continuously for 34 hours so I assumed that my hardware is OK and rebooted Gentoo in single-user mode with no services running.

Is it correct to assume that 34 lockup-free hours under XP "proves" the hardware's OK? I have a spare box with almost the same hardware and could swap hardware.

If so, can I conclude that it's a faulty Linux driver and, if so, how do I find it and get a replacement?

I'm at a loss how to proceed from here and would gratefully receive pointers.

Thanks.

DirtyHairy · Posted: Mon Sep 01, 2008 12:49 pm Post subject:

Well, as I said, I wouldn't take the fact that windows runs fine as a gurantee that the hardware is OK. There are also people that claim that memtest86 didn't find some hardware flaw for them, allthough I (luckily) can't share this experience.

If you are convinced that your hardware is fine, then you can start diagnosing the software side by removing everything you don't REALLY need from the kernel and then try to crash the machine. But I still would try swapping or replacing memory before that, seems like a better shot to me (and is done quicker, too).

P.S.: Did you say windows spontaneously rebooted? If so, then this IS a hint to faulty hardware...

platojones · l33t Joined: 23 Oct 2002 Posts: 862

It does sound like flakey hardware, to be honest with you. This whole thing started with a failed chipset fan (implying that the chipset (NB) overheated at some point). If you want to try something else (given that memtest86 is giving you no errors), then boot up WindowsXP (which is much more tolerant of hardware faults, until it isn't...), and get yourself a copy of Prime95 (http://www.mersenne.org/freesoft.htm) and run a 'stress test' instance for each core. I would run the large FFT test or the Blend test. This will stress Windows to Linux levels. If there is a hardware fault, it will either give you an error or could lock up your box. Either way, you know it hardware then and can start troubleshooting that.

If not, then obviously something fundamental changed in your Gentoo environment.