On Mon, Sep 11, 2017 at 08:56:30AM +0000, Chandran, Sugesh wrote:
>
>
> Regards
> _Sugesh
>
>
> > -----Original Message-----
> > From: Yuanhan Liu [mailto:y...@fridaylinux.org]
> > Sent: Monday, September 11, 2017 7:36 AM
> > To: Chandran, Sugesh <sugesh.chand...@intel.com>
> > Cc: d...@openvswitch.org
> > Subject: Re: [ovs-dev] [PATCH v2 2/8] dpif-netdev: retrieve flow directly
> > from the flow mark
> >
> > On Sun, Sep 10, 2017 at 04:15:42PM +0000, Chandran, Sugesh wrote:
> > > > atomic_read_relaxed(&pmd->dp->emc_insert_min, &cur_min);
> > > >
> > > > DP_PACKET_BATCH_REFILL_FOR_EACH (i, size, packet, packets_) {
> > > > struct dp_netdev_flow *flow;
> > > > + uint32_t flow_mark;
> > > >
> > > > if (OVS_UNLIKELY(dp_packet_size(packet) < ETH_HEADER_LEN)) {
> > > > dp_packet_delete(packet); @@ -4972,6 +4974,16 @@
> > > > emc_processing(struct dp_netdev_pmd_thread *pmd,
> > > > continue;
> > > > }
> > > >
> > > > + if (dp_packet_has_flow_mark(packet, &flow_mark)) {
> > > > + flow = dp_netdev_pmd_find_flow_by_mark(pmd, flow_mark);
> > > > + if (flow) {
> > > [Sugesh] If the NIC/hardware can match on tcp_flags then this parsing can
> > be avoided?
> > > Is that true?
> >
> > Maybe. If so, we could get better performance, as I have showed in v1.
> [Sugesh] Then another probing feature to see if this parsing needs to be
> done/not.?
I think so. I'd leave it as it is (only sw implementation) and will check
the possibility after this patchset has been merged.
--yliu
_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
