Re: "They have full address space protection"

[Ingo Molnar <mingo@elte.hu>]

* Philip Smolen <philip@trade-ideas.com> wrote:

> I was curious how stable TUX is when a user space module misbehaves.
> 
> I modified demo2.c to work at first, then cause a segmentation fault.  
> if (count++ == 15)
> {
>   *((char *)0) = 'X';
> }
> This locked up TUX.  I couldn't stop it with "/etc/init.d/tux stop". 
> The rest of the computer was okay, but TUX wouldn't go down until I
> rebooted.

found the bug: it's an interaction between threaded coredumps and Tux's
exit code. You can work it around by doing 'ulimit -c 0' before starting
up Tux. I've fixed it in my tree and the fix will be in -A6.

> Is this the expected behavior of TUX, or is this a bug?  Based on the
> documentation, I expected TUX to recover.

yes, Tux is very much is expected to recover in every crash scenario.

	Ingo