netdev - Distinguishing between forwarded/bridged packets and locally generated packets

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>] [day] [month] [year] [list]

Message-ID: <1491857342293.80482@alliedtelesis.co.nz>
Date:   Mon, 10 Apr 2017 20:49:02 +0000
From:   Anthony Lineham <Anthony.Lineham@...iedtelesis.co.nz>
To:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Distinguishing between forwarded/bridged packets and locally
 generated packets

Hi,

I'm trying to find a way to distinguish between a packet that is being forwarded/bridged and one that was locally generated by the host itself. I need to know this for a particular application at tx in an ethernet device driver. Forwarded/bridged packets are processed when received, but I need to make sure locally generated packets are processed too, and avoid reprocessing the forwarded packets.

Up until now we have been using the skb_iif field in the socket buffer to do this. The logic is that forwarded/bridged packet has a non-zero value in this field, indicating the incoming interface index and a locally generated packet has 0 as there is no incoming interface. However, we've recently realised that in some situations this information is scrubbed from the skb (by skb_scrub_packet), for example when the packet is being sent over a PPP link. 

I'm looking for an alternative method to use. Currently I'm thinking about using the sk field in the skb. In a locally generated packet this would have the socket the packet was generated by. In the forwarding case it would not be set. It looks like there are some cases that don't quite conform to this, such as ARP's, but I can probably live with that.

Can anyone comment on the suitability of this method or if there is a more appropriate way to achieve this.
Any information is greatly appreciated.

Thanks,
Anthony