How can I track down the cause of a freeze?

[InfinityX]

Occasionally my Gentoo box (Athlon XP, kernel 2.6.5-gentoo) will completely freeze, mouse, keyboard, everything. The only solution being a hard reset. It happens randomly, I can be playing a game, typing a document or just browsing.

It happened again just now, but strangely my WinXP box which my dad was using to browse could still access the internet (I've got Masquerading set up through a crossover cable), my modems lights would still be blinking.

After some earlier searches I'd heard that ACPI support in the kernel can be the source of many freezes, so I took it out and recompiled, and it still happens.

Where do I begin?

[searcher]

When your pc freezes, try to use ssh to login to the machine and see if anything still works. It might only be X freezing up on you. If this doesn't work you'll need a hard reboot to get back in, however not all is lost.

Depending on what system logger you use, you can find most stuff that happens to your pc in the /var/log directory. Look through the most recent logs for anything weird. If you're lucky you'll find out what went wrong before your pc froze. Chances are you won't be the only one experiencing this, so try searching the forum/google on any errors or warnings you got from the logs.

If this doesn't work you could increase the logging level to be more verbose in the logs; look to the manpage of your logger for more info on this.

~searcher

[squeegy]

i had a box that had heat issues and it would randomly freeze. i reapplied thermal paste and reseated the heatsink everythign was fine after that.

[InfinityX]

Well I've never used ssh but have a fair idea of what it's about. Since my other computer can get on the Internet you're probably right about the X freezing thing.

I don't really know what to look for in /var/logs:

Code: Select all

XFree86.0.log      emerge.log     lastlog   news     scrollkeeper.log  xdm.log
XFree86.0.log.old  genkernel.log  messages  ntp.log  snort
cups               kdm.log        mysql     samba    wtmp

If I could read my CPU temperature in Linux I'd probably be able to see if that was causing (I use superkaramba a lot so I'd set the monitors up for it) but my mobo isn't supported yet. The only way I can think of of seeing whether overheating is the problem would be to spend a day or two using Windows, a thought that sends shivers down my spine.

So let's say if it were just X freezing, how would I go about fixing it? I'm already on the latest version of XFree and Ati-drivers.

[searcher]

When your pc freezes get on your dad's winxp box, and use putty to log into your gentoo box. Make sure you have sshd running on your gentoo pc

Code: Select all

/etc/init.d/sshd start

and you just enter the local ip-adress of the gentoo box and select ssh in putty. If you get in, kill off the X server, which should unlock your pc.

Make sure you try ctrl+alt+backspace first to try to close the xwindows manually. If this works, you may want to check your xwindows stuff and config.
~searcher

*edit: look up the temperature in your bios when you have to do a hard reboot, the temperature should be in there somewhere.

[InfinityX]

Okay ssh is now running, and I've modified and tested my firewall and I can get in through my dads computer. But when the freeze next occurs how do I kill of X? Is it as simple as pkill X?

Ctrl-Alt-Backspace has never worked after a freeze unfortunately.

But now you've got me interested in ssh, could I use it to send files to and from my friends computers? I assume I would need a client for gentoo and they'd need a server like sshd?

*sorry for going OT with this*

[searcher]

But now you've got me interested in ssh, could I use it to send files to and from my friends computers? I assume I would need a client for gentoo and they'd need a server like sshd?

In linux i use the command line util `scp'. It can transfer files from and to a home dir on a pc running sshd. All you need is the sshd running and access, and all the file transfers should be encrypted.

When your pc freezes and you get in, just get the pid from ps and do a `kill -9 <pid>'. That should release your pc, and if it does, look into X config. When it doesn't, reboot and take a peek at the logs to find the cause.

~searcher

[InfinityX]

Well it's been several days and still no freeze, but sshd is always running since it's so useful. It's just amazing setting an emerge going whilst at college.

Never thought it would take this long for it to crash, in Windows all I have to do is open an explorer window half the time

[flindt]

I've been seing the same problem (or at least very similar).
Every now and then (maybe once / 3 weeks uptime) my keyboard dies and X does not respond to mouse clicks, however, the mouse cursor is still moving.

I've been able to shut down X and restart it remotely (ssh), and then everything works fine - right until I log out of the ssh, then the new X dies completely.

Does anyone know why this is? Is the X killed together with the terminal because of the parent / child relations between processes ?

[Vaughn]

Yes.

How do you *normally* start X?
If it's with startx, well, then just use startx. If it's with [xkg]dm, then do a /etc/init.d/xdm restart.


That aside: To avoid having X (or any other app) die on you, there are two often used solutions.
1: Prepend the command line with nohup, like this: "nohup X &". It'll stop it from dying but you won't be able to control it either. (It's X, so that doesn't matter.)
2: Use screen. Think "textmode VNC" to understand what it does.

[InfinityX]

I migrated to X.org after another lockup and sure enough 13 hours in everything froze.

I immediately went to my dad's computer to try ssh, nothing, I couldn't even ping. Ctrl-Alt-Backspace didn't work, and ordinary things the keyboard LEDs didn't work. However it did respond to the sysrq key. The normal procedure of using sysrq+r to recover from an X crash didn't work (but then that particular shortcut never has), I was able to sync and unmount before rebooting.

Where do we go from here?