From mboxrd@z Thu Jan 1 00:00:00 1970 From: Atis Elsts Subject: Re: [PATCH] Add sk_mark route lookup support for IPv4 listening sockets, and for IPv4 multicast forwarding Date: Wed, 7 Oct 2009 15:59:56 +0300 Message-ID: <200910071559.56526.atis@mikrotik.com> References: <200910051646.34770.atis@mikrotik.com> <20091007.031957.197632672.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, panther@balabit.hu, eric.dumazet@gmail.com, brian.haley@hp.com, zenczykowski@gmail.com To: David Miller Return-path: Received: from bute.mt.lv ([159.148.172.196]:37089 "EHLO bute.mt.lv" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933271AbZJGNAe (ORCPT ); Wed, 7 Oct 2009 09:00:34 -0400 In-Reply-To: <20091007.031957.197632672.davem@davemloft.net> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-ID: On Wednesday 07 October 2009 13:19:57 David Miller wrote: > From: Atis Elsts > Date: Mon, 5 Oct 2009 16:46:34 +0300 > > > @@ -1238,6 +1238,7 @@ static void ipmr_queue_xmit(struct sk_buff *skb, struct mfc_cache *c, int vifi) > > > > if (vif->flags&VIFF_TUNNEL) { > > struct flowi fl = { .oif = vif->link, > > + .mark = skb->mark, > > .nl_u = { .ip4_u = > > { .daddr = vif->remote, > > .saddr = vif->local, > > I'm not so sure if this is right. > > I understand what you're trying to do, inherit the socket's > mark when it goes over a multicast tunnel. > > But I'm not so sure that's what we want to do, semantically. > > Could you split out these skb->mark cases into a seperate > patch? The parts that only use sk->mark are fine and I > would like to apply a patch from you which just does that > while we discuss the skb->mark case. > Here is the sk_mark part. As for the ipmr.c code, I agree with your comment. Using mark from skb probably is wrong in case of tunnel interface (i.e. in the "if (vif->flags&VIFF_TUNNEL)" part of the patch), my mistake. I still think that the "else" part is correct, though, because using mark from skb there mirrors behaviour for unicast forwarding routing lookup in ip_route_input_slow(). The same applies to IPv6 code in ip6mr_forward2(). Add support for route lookup using sk_mark on IPv4 listening sockets. Signed-off-by: Atis Elsts --- net/ipv4/inet_connection_sock.c | 1 + net/ipv4/syncookies.c | 3 ++- 2 files changed, 3 insertions(+), 1 deletion(-) diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c index 4351ca2..9139e8f 100644 --- a/net/ipv4/inet_connection_sock.c +++ b/net/ipv4/inet_connection_sock.c @@ -358,6 +358,7 @@ struct dst_entry *inet_csk_route_req(struct sock *sk, const struct inet_request_sock *ireq = inet_rsk(req); struct ip_options *opt = inet_rsk(req)->opt; struct flowi fl = { .oif = sk->sk_bound_dev_if, + .mark = sk->sk_mark, .nl_u = { .ip4_u = { .daddr = ((opt && opt->srr) ? opt->faddr : diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c index a6e0e07..5ec678a 100644 --- a/net/ipv4/syncookies.c +++ b/net/ipv4/syncookies.c @@ -333,7 +333,8 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb, * no easy way to do this. */ { - struct flowi fl = { .nl_u = { .ip4_u = + struct flowi fl = { .mark = sk->sk_mark, + .nl_u = { .ip4_u = { .daddr = ((opt && opt->srr) ? opt->faddr : ireq->rmt_addr),