bpf and BIOCSFEEDBACK

[Mark Davies]

I'm trying to get the dynamips emulator working - specifically the pcap
based nio_gen_eth that allows the emulated network to interact with real
networks -- does this by using pcap to inject and read packets from the
host machine's ethernet interface. This works fine when talking to other
machines on the physical network but fails to communicate with the host
machine itself.

FreeBSD fixed this problem (in patches in the dynamips port) by setting
BIOCSFEEDBACK on the bpf device but if I try that on NetBSD it doesn't seem
to make a difference. From sprinkling a few printf's in bpf.c I can see
that if_input is being called but I can't tell what happens after that.

Suggestions on why this isn't working or how to track it down much
appreciated.

[I've mostly been testing this on a 5.1_RC1 system - having backported the
feedback support from current to it,
http://mail-index.netbsd.org/source-changes/2010/03/13/msg007757.html, but
i did try a current monolithic kernel at one point to confirm that it
wasn't an issue with the backport and had the same behaviour]

cheers
mark

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de

[David Young]

Post by Mark Davies
I'm trying to get the dynamips emulator working - specifically the pcap
based nio_gen_eth that allows the emulated network to interact with real
networks -- does this by using pcap to inject and read packets from the
host machine's ethernet interface. This works fine when talking to other
machines on the physical network but fails to communicate with the host
machine itself.

I believe that other emulators use tap(4), is it very hard to use that
instead of pcap(3)?

Dave

[Mark Davies]

Post by David Young
I believe that other emulators use tap(4), is it very hard to use that
instead of pcap(3)?

It has the capability to use tap though its a little more fiddly (we have
to get the students to create the tap device and bridge it -- or do that
for them) so we can go that way but given the bpf feedback stuff is there
now shouldn't it work?

cheers
mark

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de