That "loopback" interface which allows reclassification (i.e. sending packets to a HW classifier from the core) would be an appropriate pktio, opened on a special HW device which is designed for this task. If the packets are returned to the core (this is not mandatory also), it looks like a loopback but is not , at least in the original "loopback interface" meaning. Why not using enqueue instead for reclassification?
Alex On 25 November 2014 at 15:48, Taras Kondratiuk <[email protected]> wrote: > On 11/25/2014 03:30 PM, Alexandru Badicioiu wrote: > >> What is a self-contained application? An application which uses only ODP >> calls? >> > > I mean an application that doesn't need any external equipment. > > Regarding the classification tests and loopbacks, why there's a need for >> a special kind of ODP "loopback" interface? >> Why the following scenario is not acceptable for a test - open a pktio >> (ethernet), configure it in loopback mode and transmit the test traffic >> over the pktio with odp_pktio_send(), for example. >> > > It can work this way also for test, but consider a use-case: > application processes IPsec tunneled traffic and after decryption it > wants to reclassify a packet. We agreed before that such reclassification > will be done via some loopback pktio interface. Should it be a real > physical interface in a loopback mode or it can be a special kind of > loopback? >
_______________________________________________ lng-odp mailing list [email protected] http://lists.linaro.org/mailman/listinfo/lng-odp
