Re: NT vulnerable to attack on CPU

jamesh (jamesh@interpath.com)
Mon, 23 Dec 1996 10:12:34 -0500

Trevor Johnson wrote:
>
> Fabio Yeon wrote:
>
> > Hum, I wouldn't go there... Last I recall, Linux had the same problem with
> > this bit of code, spawning processes exponentially until system grinds to a
> > halt...
> >
> > main() {
> > while(1) {
> > fork();
> > }
> > }
>
> I ran that under 2.1.17 and was rewarded with multiple oopses,
> interspersed with "trying to free free IRQ 3...[oops]...trying to free
> free IRQ 4" messages. Afterward, I was able to switch virtual consoles,
> but nothing else (ctrlaltdel or typing into tcsh/mingetty) worked. None
> of it got into the syslog and I didn't bother writing it down, but would
> be glad to do so if anyone asks.

This is comparing applets^H^Hs and oranges...

The NT problem is caused by running code produced somewhere else. It
could be triggered by a "Cancel" button on an applet written someone
half-way arround the world in the opposite hemisphere.

The Linux program has to be run by a user logged onto the system. Quite
a bit of difference. You can convieniently remove users that do this
sort of thing from the system. Besides, you can allocate resources via
ulimit in bash to circumvent the problem before it happens.

Even so, I ran the program on 2.0.27 with unlimited resources in bash
(in X). I was able to get a console and bring the system dowm. If ps
hadn't barfed, I could have recovered without bringing down the system.
I ran the program with process limits set at 100, and was able to kill
the parent shell and regain control of the system.

You can tell your NT friend; Nice Try - maby Next Time - Not Today...

-James