setjmp/lonjmp

Joerg Sonnenberger joerg at britannica.bec.de
Fri Feb 4 05:09:15 PST 2005


On Fri, Feb 04, 2005 at 02:41:45PM +0200, Yury Tarasievich wrote:
> A word from "the masses": this is completely justified -- developer's -- 
> point of view. On the other hand: cleaning of the code per se gives 
> plenty of opportunities to break what's working. Like, reducing vinum 
> functionality (or breaking it), or having ipfw removed. Doesn't look good.

I had a quite impossible panic from vinum a while ago which most likely
relates to longjmp handling. But I'm not sure and the current code is
exactly for the interwindling of code pathes via longjmp hard to follow.

Concerning the removal of ipfw1, it won't happen soon. Once ipfw2
fully works, it can die. Another requirement or ipfw(2) is to use
the normal firewall API, since it currently hooks into way too
much places directly. This means an improvement in usability too,
because ipfw would be fully dynamically loadable.

Joerg





More information about the Kernel mailing list