On Mon, Aug 28, 2017 at 12:47:52PM +0000, Jan Scheurich wrote: > > -----Original Message----- > > From: Yi Yang [mailto:yi.y.y...@intel.com] > > > > +static bool > > +compose_dec_nsh_ttl_action(struct xlate_ctx *ctx) { > > + struct flow *flow = &ctx->xin->flow; > > + > > + if ((flow->packet_type == htonl(PT_NSH)) || > > + (flow->dl_type == htons(ETH_TYPE_NSH))) { > > + ctx->wc->masks.nsh.ttl = 0xff; > > + if (flow->nsh.ttl > 1) { > > + flow->nsh.ttl--; > > + return false; > > + } else { > > + execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, > > 0, > > + NULL, 0); > > + } > > + } > > + > > + /* Stop processing for current table. */ > > + xlate_report(ctx, OFT_WARN, "NSH decrement TTL exception"); > > + return true; > > I see you followed the dec_mpls_ttl implementation here in that you terminate > translation and drop the packet when trying to decrement TTL of a non-NSH > packet. This is different from the dec_nw_ttl action, which does nothing when > executed on a non-IP packet. > > @Ben: What is the reason for different approaches in OVS? > > @Yi: Why have you opted for the MPLS approach?
I don't recall why these are different. I don't think the difference is that significant in practice because, outside of OpenFlow 1.0, OVS only allows an IP TTL decrement action in a flow matches only IP packets, and similarly for MPLS TTL decrement. I wouldn't object to making the behavior uniform. _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev
