Aleksander Morgado <aleksan...@aleksander.es> writes: > On Tue, Jan 19, 2016 at 11:51 AM, Bjørn Mork <bj...@mork.no> wrote: >>> On Mon, Jan 18, 2016 at 10:15 PM, Glenn Washburn >>> <gwashb...@eagleeyenetworks.com> wrote: >>>> "curl -v --interface wwan0 http://www.google.com"; >>> >>> You're binding the request to the wwan interface, which I assume it's >>> because it isn't the default route. >>> >>> Try to disable the reverse path filtering, something like: >>> # for i in /proc/sys/net/ipv4/conf/*/rp_filter ; do echo 0 > $i; done >> >> Nice catch! Personally I like to leave rp filtering to my ISP, blindly >> accepting anything they let through :) >> >> But I still do not understand how those synack's could pass the filter? > > You mean because they could be seen in the tcpdump? IIRC the rp-filter > happens afterwards, although I'm not 100% sure; this is a long time > ago I faced it myself when doing multiplexing of communications over > multiple wwans.
You're right. I just tested and see the exact same behaviour. I wrongly assumed that the rp_filtering would apply somewhere around the iptables INPUT filtering, but it looks like it must be later. So you do see the synack packets in the packet dump, but tcp won't see them. So all that makes sense then. Turn off rp_filter, or set up the symmetrical routing. I usually do both myself :) Bjørn _______________________________________________ ModemManager-devel mailing list ModemManager-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/modemmanager-devel
