> 2013/10/23 FUJITA Tomonori <[email protected]> > >> On Wed, 23 Oct 2013 11:33:53 +0900 (JST) >> [email protected] (YAMAMOTO Takashi) wrote: >> >> >> 2013/10/21 FUJITA Tomonori <[email protected]> >> >> >> >>> On Mon, 21 Oct 2013 18:45:23 +0900 >> >>> Satoshi Kobayashi <[email protected]> wrote: >> >>> >> >>> > I was also mistaken in it and made time useless. >> >>> > Therefore I feel that it is better to change a default value. >> >>> > >> >>> > Is it like this\x85? (RFC because I do not know that this is >> excellent) >> >>> > ------------------------------ >> >>> > def __init__(self, port, max_len=ofproto_v1_3.OFPCML_NO_BUFFER, >> >>> > type_=None, len_=None): >> >>> > super(OFPActionOutput, self).__init__() >> >>> > self.port = port >> >>> > self.max_len = max_len if port == >> ofproto_v1_3.OFPP_CONTROLLER >> >>> else >> >>> > 0 >> >>> >> >>> Changing the default to non zero value is fine by me. However, using >> >>> OFPCML_NO_BUFFER has big side effects (could be harmful), the packet >> >>> is not buffered anymore? >> >>> >> >> >> >> I agree. >> >> How is it that uses OFPCML_MAX? I want to set the default size large >> >> enough. If it is a fixed number, it is nice. >> > >> > how about 128, which is the default miss_send_len? >> >> My hunch says that we would get reports "corrupted packets are sent to >> a controller" even with that. >> >> I guess that OFPCML_MAX is "safer". Let's use it for both 1.2 and 1.3 >> > > I sent the patch. But I stop that if the port is OFPP_CONTROLLER, max_len > is set to 0. > >>From the following reasons: > 1. It modifies a content without expecting when parsing a message > 2. It requires many fixes to unit test > > Probably it will be ignored by a switch if the port is not OFPP_CONTROLLER > and max_len is not 0. > (I was not able to find behavior at that time from OpenFlow switch > specification)
afaik, the meaning of max_len is defined only for port=controller. YAMAMOTO Takashi > > Thanks, > Satoshi ------------------------------------------------------------------------------ October Webinars: Code for Performance Free Intel webinars can help you accelerate application performance. Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from the latest Intel processors and coprocessors. See abstracts and register > http://pubads.g.doubleclick.net/gampad/clk?id=60135991&iu=/4140/ostg.clktrk _______________________________________________ Ryu-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ryu-devel
