From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andy Furniss Date: Sat, 17 Dec 2016 15:44:25 +0000 Subject: Re: Problem to priorize SSH traffic Message-Id: <58555D59.7070809@gmail.com> List-Id: References: <1737987792.142216491.1481907014295.JavaMail.root@zimbra45-e7.priv.proxad.net> In-Reply-To: <1737987792.142216491.1481907014295.JavaMail.root@zimbra45-e7.priv.proxad.net> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: lartc@vger.kernel.org Ludovic Leroy wrote: > Andy, > > ip + 14? How do you get this result? It's IP + src/dst MAC + ethertype protocol (which you can match with tc filters using a negative offset) On a vlan interface it's it's + 18 On ppp it's just IP length. If you make a filter/class for say icmp and play with ping you can infer the overhead by looking at the byte counters. This is nothing to do with stab - it's always the case, so you need to take it into account when working out what number to use for overhead. My old DSL link was pppoa vc mux so the fixed overhead was ip + 10. When I has a PCI modem and shaped on ppp0 the overhead param of stab was 10. When I later got a stand alone modem connected by eth I had to use -4 as the overhead to allow for the fact that tc already saw ip + 14.