Re: [PATCH v2 nf] netfilter: nft_set_pipapo: split gc in unlink and reclaim phase

[Stefano Brivio <sbrivio@redhat.com>]

On Tue, 10 Mar 2026 18:39:34 +0100
Florian Westphal <fw@strlen.de> wrote:

> Stefano Brivio <sbrivio@redhat.com> wrote:
> > Sorry for the late review. Just one (perhaps dumb) question:  
> 
> No problem, thanks for reviewing.
> 
> > >  	struct nft_pipapo *priv = nft_set_priv(set);
> > >  	struct net *net = read_pnet(&set->net);
> > > @@ -1697,6 +1697,8 @@ static void pipapo_gc(struct nft_set *set, struct nft_pipapo_match *m)
> > >  	if (!gc)
> > >  		return;
> > >  
> > > +	list_add(&gc->list, &priv->gc_head);  
> > 
> > ...is there a reason why we need to do this unconditionally, or could
> > we do this opportunistically if (__nft_set_elem_expired(&e->ext,
> > tstamp)) below, including the nft_trans_gc_alloc() call?  
> 
> Yes, its to make sure we run the catchall gc, which is external
> to the pipapo core datastructure.

Ah, right, it wouldn't run otherwise, I missed that. Thanks for
explaining.

> I admit we could be more clever and try to supress this
> gc container allocation, but I preferred to keep it simpler for now.

Yeah it sounds very reasonable.

-- 
Stefano