Re: [libnftnl PATCH 4/6] set: Don't bypass checks in nftnl_set_set_u{32,64}()

[Phil Sutter <phil@nwl.cc>]

On Tue, Oct 15, 2019 at 06:32:39PM +0200, Pablo Neira Ayuso wrote:
> On Tue, Oct 15, 2019 at 06:11:34PM +0200, Phil Sutter wrote:
> > Hi,
> > 
> > On Tue, Oct 15, 2019 at 05:53:46PM +0200, Pablo Neira Ayuso wrote:
> > > On Tue, Oct 15, 2019 at 04:16:56PM +0200, Phil Sutter wrote:
> > > > By calling nftnl_set_set(), any data size checks are effectively
> > > > bypassed. Better call nftnl_set_set_data() directly, passing the real
> > > > size for validation.
> > > > 
> > > > Signed-off-by: Phil Sutter <phil@nwl.cc>
> > > 
> > > Acked-by: Pablo Neira Ayuso <pablo@netfilter.org>
> > > 
> > > Probably attribute((deprecated)) is better so we don't forget. Anyway,
> > > we can probably nuke this function in the next release.
> > 
> > But if we drop it, we break ABI, no? Sadly, nftables use(d) the symbol,
> > so we would break older nftables versions with the new libnftnl release.
> >
> > Should I send a v2 setting attribute((deprecated))? I think it's worth
> > doing it.
> 
> OK.

Well, given that there are more cases like this (e.g. nftnl_obj_set()),
I'll just drop the comment from existing patch and follow-up with a
separate one deprecating all unqualified setter symbols at once.

Cheers, Phil