On Sat, Apr 07, 2012 at 02:12:40PM +0200, LIU Binghan wrote: > My idea is, in the beginning,' datapath' does not have any flow > table. If it found a flow that it does not know, then it would pass > the flow key and packet through netlink to 'dpif'. 'dpif' processes > the packet, and also tells the datapath through netlink to expand its > flow table about this flow. After that, packets would process by > kernel.
Yes, this is correct. > Thus, I think datapath need to learn the flow information(flow key > associate with actions) from dpif. Yes. > But, right now, as my understanding about what you say, there should > be some flowtable in the kernel(not datapath kernel module). I don't understand this distinction. What is the difference between a flow table in the kernel and a flow table in the datapath kernel module? > I asked this kind of questions is because I want to develop a new > tunnel protocol for openvswitch. Therefore, i want to know how user > space maintains the flow table, I lost in this part. Good luck! _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev