netdev - Re: [PATCH net] netfilter: set skb transport_header before calling sctp_compute

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 12 Mar 2019 10:48:46 +0100
From:   Pablo Neira Ayuso <pablo@...filter.org>
To:     Xin Long <lucien.xin@...il.com>
Cc:     Neil Horman <nhorman@...driver.com>,
        Florian Westphal <fw@...len.de>,
        network dev <netdev@...r.kernel.org>,
        netfilter-devel@...r.kernel.org,
        Marcelo Ricardo Leitner <marcelo.leitner@...il.com>
Subject: Re: [PATCH net] netfilter: set skb transport_header before calling
 sctp_compute_cksum

On Tue, Mar 12, 2019 at 04:39:46PM +0800, Xin Long wrote:
> On Mon, Mar 11, 2019 at 7:08 PM Neil Horman <nhorman@...driver.com> wrote:
> >
> > On Sat, Mar 09, 2019 at 10:24:34AM +0100, Florian Westphal wrote:
> > > Xin Long <lucien.xin@...il.com> wrote:
> > > >   https://marc.info/?l=linux-netdev&m=155109395226858&w=2
> > > > But from sctp side, Neil preferred sctp_hdr().
> > > >
> > > > We need to either add skb_set_transport_header() in sctp_s/dnat_handler()
> > > > and sctp_manip_pkt(), or bring that patch back?
> > > >
> > > > Now it seems not good to set skb->transport_header in netfilter code.
> > >
> > > I think its fine, but I wonder why we need to do it.
> > >
> > > Since 21d1196a35f5686c4323e42a62fdb4b23b0ab4a3 ipv4 input path sets
> > > transport header before netfilter.  The only problem is that linear
> > > access is illegal without may_pull checks, but in this case the
> > > make_writable call takes care of this already.
> > >
> > Yes, this.  It seems to me we should be setting the transport header prior to
> > ever getting into the netfilter code, which does imply that we need the may_pull
> > check to linearize enough of the packet to do so, just like tcp and udp do.
> >
> > > So, why was this patch needed?
> 
> The issue was reported when going to nf_conntrack by br_netfilter's
> bridge-nf-call-iptables, which could be:
> 
> br_prerouting->inet_prerouting->
> br_forward->inet_forward->
> br_postrouting->inet_postrouting

Can you fix this from br_netfilter then? ie. set the transport header
before prerouting to emulate the IP stack behaviour.