[ On Fri, December 18, 1998 at 11:35:24 (+0100), Ignatios Souvatzis wrote: ]
> Subject: Re: userid partitioned swap spaces: note that running out of swap no longer crashes the machine.
>
> On Fri, Dec 18, 1998 at 12:30:18PM +0200, Jukka Marin wrote:
> > This still won't help if the unlucky process happens to be inetd or named
> > or something essential.. but the other solutions (allowing only root
> > processes to allocate swap after a certain point) would help.
> 
> Hm, if it makes sure to touch every page it needs on start-up, it will be
> safe later. ;-)

That's not really practical if you're talking about stack space, though
I suppose a robust C runtime startup could poke out on the stack a bit
to guarantee a certain amount of space would be available.  The process
could even do this on its own, but I'd rather see any such robustness
measures be programmed into the runtime, libc, etc., which might just
make it possible to link any program with a "robust" library/flag and
have it do the right things....

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>