Re: [PATCH nf] netfilter: nft_set_rbtree: narrow down range completion to anonymous sets

[Florian Westphal <fw@strlen.de>]

Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> On Thu, Oct 08, 2026 at 12:52:35PM +0200, Pablo Neira Ayuso wrote:
> > On Thu, Oct 08, 2026 at 08:32:54AM +0200, Florian Westphal wrote:
> > > Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> > > > For anonymous sets, when adjacent ranges are found, the end element is
> > > > not added to the set to pack the set representation. Use next start
> > > > element to complete this interval.
> > > > 
> > > > Jérémy Jean says:
> > > > 
> > > > Consider two ranges in a netfilter verdict map:
> > > > * keys 10-19 accept packets, and
> > > > * keys 30-39 jump to chain 'victim'.
> > > > Now, delete the end marker at 20. The lookup array now uses element 30
> > > > as the first range's end, so an INTERVAL_END query for key 15 returns
> > > > element 30, including its jump verdict and chain pointer. This can make
> > > > KASAN reports a UAF after the GC releases reference chain for 30 that
> > > > has expired:
> > > 
> > > Dumb question: How is it possible to remove an element if the map is
> > > anonymous?
> > 
> > I suspect this is a strange combination which cannot be reached from
> > userspace.
> 
> I mean, via nft.

Even with netlink API: DELSETELEM has:

        if (nft_set_is_anonymous(set))
                return -EOPNOTSUPP;

Jérémy, can you share what your reproducer is doing to get the rbtree
map into the bad state?

Thanks!

Just to be clear: this patch is fine, but I a) wonder if there is
some other bug elsewhere, and b) if we can make a test case for
this for either nftables.git or knft so that this is covered.