I've never actually seen a kernel panic, but I do understand that some sort of message is printed on the screen. What I am wondering is if there is a way to have the function that displays the 'kernel panic' message, call the kernel/BIOS function that does a hard reboot of the computer?

The kernel has a function that tells the BIOS to restart, so that function should still be callable from the kernel panic function, right.

Is there an existing patch for doing this?

Do you think that would be a good idea? How would you intervene to boot from a rescue floppy, or change the command line that lilo or grub feeds to the kernel if you could not stop the cycle?

Cheers--
Charles

Once I stupidly set my initdefault to 6, which means that the first thing that the kernel does after booting is reboot. I watched my machine boot up, shutdown, boot up, shutdown... in a complete panic for several minutes before I powered off and then booted a rescue CD to begin diagnosing the problem.

I think that a kernel panic reboot would create the same sort of loop.

A loop would be caused if the kernel panics when it boots up.
I'd be interested in using such a kernel to run servers that I would like to automatically restart if there is a kernel panic once they are running, in order to increase reliability, so that they can be run for a very long time without maintenance from a technician.

I doubt these systems would have a kernel panic while they are booting up, but the solution to that problem would be to a have a special system call that must be called in order to enable this special patch (not restart, but enable restarting in the event of a kernel panic), with this new syscall only being called once the system is up and running and therefore doesn't have memory problems or other issues that can cause a kernel panic while the kernel is loading.