On Sun, Jan 05, 2014 at 09:39:25PM +0100, Pablo Neira Ayuso wrote:
> Hi Patrick!
>
> On Fri, Jan 03, 2014 at 12:16:12PM +0000, Patrick McHardy wrote:
> > The following patches add support for mixed IPv4/IPv6 tables to nftables.
> >
> > The first patch fixes the chain type override logic so chain types override
> > the AF defaults instead of the other way around. The second patch adds a
> > pointer to the nf_hook_ops struct to nft_pktinfo for a later patch which
> > uses ops->pf to match the actual AF of the packet instead of the dummy
> > NFPROTO_INET value. The third patch changes the chain and hook registration
> > logic to support multiple hook registrations. The nf_tables AF modules can
> > provide a callback function to override defaults. The fourth patch finally
> > adds a new "inet" family, which basically only initializes the hook
> > functions to the IPv4 and IPv6 specific ones and registers a dummy filter
> > chain type for NFPROTO_INET. Patches 5 and 6 add support for matching on
> > the netfilter hook family and the L4 protocol number to the meta match.
> >
> > With all this in place, we can create AF-specific rules and AF-independant
> > rules that only match on the L4 protocol header and above in the inet table:
> >
> > table inet filter {
> > chain input {
> > type filter hook input priority 0;
> > }
> >
> > chain forward {
> > type filter hook forward priority 0;
> > }
> >
> > chain output {
> > type filter hook output priority 0;
> > ip protocol tcp tcp dport 1234 counter packets 2 bytes 120
> > ip6 nexthdr tcp tcp dport 1234 counter packets 2 bytes 160
> > tcp dport 1234 counter packets 4 bytes 280
> > }
> > }
> >
> > Userspace needs a bit of polishing but will most likely follow in a few
> > hours.
> >
> > Comments welcome.
>
> This patchset looks pretty good, I think there is some clashing with
> recent Arturo's work to set meta information, please see this patch.
>
> http://git.kernel.org/cgit/linux/kernel/git/pablo/nftables.git/commit/?id=e035b77ac7be430a5fef8c9c23f60b6b50ec81c5
>
> But that I can manually resolve that.
>
> We'll also need route and mangle chains for the new inet family type,
> but that's also easy to add.
Well, route is certainly fine. If you mean nat (we don't have mangle),
I'll leave that to the future since the AF-specific stuff still needs a
bit of thought.
> Let me know if I can give you a hand with the userspace patch, you can
> post a tentative RFC before polishing if you like.
Actually its already finished, I just generalized it a bit to solve a
couple of related problems wrt. protocol contexts. Final patchsets will
come tommorrow.
Cheers,
Patrick
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
