From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <4C5997DD.8080200@free.fr> Date: Wed, 04 Aug 2010 18:39:57 +0200 From: =?ISO-8859-1?Q?Nicolas_de_Peslo=FCan?= MIME-Version: 1.0 References: <4C5953C8.2040708@free.fr> <51AC42EFFBC60FE7E630D78B@Ximines.local> In-Reply-To: Content-Type: text/plain; charset="iso-8859-1"; format="flowed" Content-Transfer-Encoding: quoted-printable Subject: Re: [Bridge] ebtables PREROUTING -drop List-Id: Linux Ethernet Bridging List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Jan Engelhardt Cc: Netfilter mailing list , bridge@lists.linux-foundation.org, Alex Bligh Le 04/08/2010 16:32, Jan Engelhardt a =E9crit : > > On Wednesday 2010-08-04 16:25, Alex Bligh wrote: >> >>>>> Did you read http://ebtables.sourceforge.net/br_fw_ia/br_fw_ia.html a= nd >>>>> http://ebtables.sourceforge.net/br_fw_ia/PacketFlow.png ? >>>> >>>> A useful improvement to those would be documenting where libpcap >>>> (which does both input and, less well known, output) samples/injects >>>> packets. I /think/ sampling is right on the left and injection right >>>> on the right. >>> >>> pcap grabbing and injection is completely outside any of the graphs >>> currently floating around. >> >> If by 'outside' you mean 'to the extreme left or extreme right' >> that was my conclusion. But the absence of any documentation means >> this makes debugging with tcpdump (for instance) harder >> because you don't know where you are sampling. > > Well perhaps not extreme. If you inject into a tunnel, it may well > walk through Xtables after all - but then of course, only in its > encapsulated form. > >> I'm not 100% sure it is completely outside though. For instance, >> if you do tcdump on a bridge device (as opposed to the corresponding >> physical participant interface), isn't that after ingress ebtales >> processing, but before egress? IE is in the graph somewhere. > > Huh, all once investigated already. See > http://jengelh.medozas.de/images/nf-packet-flow.png for where > in/egress happen to be. :) Nice work! May be just missing other netif_receive_skb() magic, like bonding for examp= le. Nicolas.