Re: RELENG_5 kernel b0rken with IPFIREWALL and without PFIL_HOOKS

From: Ruslan Ermilov <ru_at_freebsd.org>
Date: Thu, 19 Aug 2004 19:37:13 +0300
On Thu, Aug 19, 2004 at 12:04:45PM -0400, Barney Wolff wrote:
> On Thu, Aug 19, 2004 at 06:54:13PM +0300, Ruslan Ermilov wrote:
> > On Thu, Aug 19, 2004 at 11:43:34AM -0400, Barney Wolff wrote:
> > > I was inspired by the PFIL_HOOKS discussion to check my firewall rules :)
> > > There were none, other than 65535.  Apparently, /etc/rc.d/ipfw attempts
> > > to kldload ipfw, which will fail if ipfw is compiled into the kernel,
> > > and since the precmd failed, the _cmd will not be run.  When did it
> > > become mandatory to have ipfw as a module, not compiled in?  Is there
> > > some rationale for this?  It strikes me as rather dangerous, especially
> > > for firewalls, especially when default-to-accept is chosen.  Am I just
> > > confused, and missing some obvious bit of config?
> > > 
> > > Is it relevant that my /usr is on vinum, and the rules are in /usr/local/etc?
> > > 
> > net.inet.ip.fw.enable is gone, and it upsets /etc/rc.d/ipfw.
> > I asked Andre to follow up on this.
> 
> Yes, but aside from that, ipfw_precmd returns 1 if the kldload fails,
> which if I'm not confused causes ipfw_start not to be run.  At least
> that's what my system as of 8/17/04 says.
> Barney
> 
Yes sure.  Non-existing sysctl causes kldload to be attempted,
that fails (because the module already exists), and the whole
/etc/rc.d/ipfw is aborted.


Cheers,
-- 
Ruslan Ermilov
ru_at_FreeBSD.org
FreeBSD committer

Received on Thu Aug 19 2004 - 14:37:18 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:07 UTC