Re: [PATCH] src: proto: Add missing packet type

[Pablo Neira Ayuso <pablo@netfilter.org>]

On Tue, Mar 01, 2016 at 03:01:06PM +0100, Florian Westphal wrote:
> Shivani Bhardwaj <shivanib134@gmail.com> wrote:
> > On Tue, Mar 1, 2016 at 6:47 PM, Florian Westphal <fw@strlen.de> wrote:
> > > Shivani Bhardwaj <shivanib134@gmail.com> wrote:
> > >> Add missing packet type "invalid" for DCCP.
> > >>
> > >> Signed-off-by: Shivani Bhardwaj <shivanib134@gmail.com>
> > >> ---
> > >>  src/proto.c | 1 +
> > >>  1 file changed, 1 insertion(+)
> > >>
> > >> diff --git a/src/proto.c b/src/proto.c
> > >> index 0ed98ed..4d049f5 100644
> > >> --- a/src/proto.c
> > >> +++ b/src/proto.c
> > >> @@ -443,6 +443,7 @@ static const struct symbol_table dccp_pkttype_tbl = {
> > >>               SYMBOL("reset",         DCCP_PKT_RESET),
> > >>               SYMBOL("sync",          DCCP_PKT_SYNC),
> > >>               SYMBOL("syncack",       DCCP_PKT_SYNCACK),
> > >> +             SYMBOL("invalid",       DCCP_PKT_INVALID),
> > >
> > > I don't think this is a good idea -- when user asks to match
> > > 'invalid' then this will check type == DCCP_PKT_INVALID; however
> > > the correct way would be to ask for type '> synack' (i.e.,
> > > outside of the range of types specified).
> > 
> > OK. So, should this be removed from the enum dccp_pkt_type as well?
> 
> Yes, but we cannot do this anymore for compatibility reasons :-(
> 
> > And, if following iptables rules, one tries to get code corresponding
> > to dccp type invalid using iptables-translate utility, then the
> > corresponding nft rule should be "dccp type gt syncack". Please
> > correct me if I'm wrong here.
> 
> Good question.
> Pablo, I think -m dccp is buggy here, it should not have allowed
> INVALID match in the first place.

Right.

> Looking at kernel part it doesn't seem to special-case this
> and will just check for type == INVALID.  But such type isn't
> defined in dccp protocol, it just looks like implementation
> helper so DCCP stack can do simpler sanity tests.
>
> So I suggest to not support translation of INVALID to nft and
> to document the INVALID shortoming in iptables-extensions manpage.
> 
> Any comments?

That's fine, we can return 0 (no translation available) in that case.