From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeremy Jackson Subject: Re: adding tcpdump/OAM support to usb ATM devices Date: Mon, 28 Apr 2008 11:54:42 -0400 Message-ID: <1209398082.8768.257.camel@ragnarok> References: <200804281418.m3SEImnN032542@cmf.nrl.navy.mil> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org To: chas3@users.sourceforge.net Return-path: Received: from titan.coplanar.net ([70.47.139.2]:53518 "EHLO titan.coplanar.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933996AbYD1Pyt (ORCPT ); Mon, 28 Apr 2008 11:54:49 -0400 In-Reply-To: <200804281418.m3SEImnN032542@cmf.nrl.navy.mil> Sender: netdev-owner@vger.kernel.org List-ID: On Mon, 2008-04-28 at 10:18 -0400, Chas Williams (CONTRACTOR) wrote: > In message <1209331913.8768.214.camel@ragnarok>,Jeremy Jackson writes: > >who cares about ATM? Well anyone using ADSL/2/2+ high speed internet > >access is stuck with it, and that's not a small number of people. > > i still care. good to hear! I'd love to label ATM as legacy and forget it... but we're stuck with ATM on ADSL (VDSL has packet mode with HDLC-like framing) since no hardware vendor has implemented the packet mode for ADSL2/2+ even though it's part of the spec. > >The question: what is the recommended way to push the captured cells to > >userspace? raw ATM socket, netlink, packet socket, i'm not up to speed > >here people! > > the right way is netlink. i have some patches to add more complete > oam support but they use the current ioctl interface which should be > considered deprecated. so adding new functionality using this interface > to the kernel, would likely be rejected. Can you send me some patches? It would be good for discussion and experimentation. > eventually, all the atm clients (zeppelin, atmsigd, et al) should be > converted to using netlink. > > >in 2.6.22 drivers/usb/atm/usbatm.c::usbatm_extract_one_cell() > > > >cells received for open PVC/SVC sockets are processed, however OAM cells > >are dropped, and cells to unbound VPI/VCI are dropped. > > > >This seems the likely place to route cells somewhere useful, but what > >mechanism to use? > > i guess now might be the time to try to define a netlink interface > between the userspace and the kernel. I think there are 2 cases, packet socket support for ATM, for libpcap/tcpdump etc, and control functions (presently ioctl) switched to use netlink. Generating OAM cells might belong to a control function on netlink, but I wonder about PF_RAWATM, the present setup with pcv/svc sockets is a bit wierd, you set the qos type to AAL0 to mean raw. Perhaps there should be a PF_RAWATM instead. -- Jeremy Jackson Coplanar Networks (519)489-4903 http://www.coplanar.net jerj@coplanar.net