On Jul 20, 2013, at 12:16 PM, Daniel Borkmann <dbork...@redhat.com> wrote:
> On 07/19/2013 08:23 PM, Guy Harris wrote:
>> On Jul 3, 2013, at 3:49 AM, Daniel Borkmann <dbork...@redhat.com> wrote:
>>
>>> For pcap interoperability, introduce a common link type for netlink
>>> captures.
>>
>> What do the link-layer headers for this look like?
>
> That is struct nlmsghdr, found in include/uapi/linux/netlink.h.
I.e., what is captured are netlink messages, as described by, for example,
section 3.4 "Netlink message format" of:
http://1984.lsi.us.es/~pablo/docs/spae.pdf
or section 2.2 "Message Format" of
http://tools.ietf.org/html/rfc3549
>> Presumably making that work also involves changes to libpcap to support
> > capturing on nlmon devices (so that DLT_NETLINK is returned for them) and,
> > if you're not using the -w flag to tcpdump, changes to tcpdump to analyze
> > DLT_NETLINK packets.
>
> Right, for the device type identification, this is being exported as
> ARPHRD_NETLINK (include/uapi/linux/if_arp.h) in pf_packet's sll's
> sll_hatype member.
>
> I can have a look how libpcap handles this and send a follow-up patch
> for further inclusion next week if wished.
If the nlmon device look like regular network interfaces to which you can bind
a PF_PACKET socket, then libpcap handles this in the switch statement in
map_arphrd_to_dlt() - add a case for ARPHRD_NETLINK and map it to DLT_NETLINK.
_______________________________________________
tcpdump-workers mailing list
tcpdump-workers@lists.tcpdump.org
https://lists.sandelman.ca/mailman/listinfo/tcpdump-workers
