From: Theo Buehler <tb@theobuehler.org>
Subject: Re: bgpd: start rework the rde_filter code
To: tech@openbsd.org
Date: Tue, 10 Feb 2026 20:10:59 +0100

On Tue, Feb 10, 2026 at 08:05:06PM +0100, Claudio Jeker wrote:
> On Tue, Feb 10, 2026 at 05:18:43PM +0100, Theo Buehler wrote:
> > On Tue, Feb 10, 2026 at 04:12:24PM +0100, Claudio Jeker wrote:
> > > After implementing rde_filterset which improved rde_attr_set() a fair bit
> > > it is time to tackle filter_rules.
> > > 
> > > Right now I'm just looking at the out filters (which are already per
> > > peer). So as a first step I decoupled these filters from the in filters by
> > > adding a new simplified rde_filter_out() function.
> > > Also rde_filter_match() is now just using the filter_match object and not
> > > the full rule. For that the peer.ebgp and peer.ibgp checks had to be moved
> > > up into rde_filter() which is a bit strange because of skip steps.
> > > Long term I will also redo the in filters but those will follow later and
> > > then skip steps will go away.
> > > 
> > > So this is just a big shuffle that should have no visible effect.
> > 
> > The diff reads fine.
> > 
> > ok tb
> > 
> > Since you say the skip steps will go away, I would suggest that you kill
> > RDE_FILTER_TEST_ATTRIB() as well. I would find
> > 
> > 	if (condition) {
> > 		f = next;
> > 		continue;
> > 	}
> > 
> > easier to read and I think it would even use fewer lines than now.
>  
> Like this?

Precisely. ok tb