[Inline] > After investigating, I've figured out that the case was that, when a > host sent an ARP request (for the IP of the other host), POX received > a packet-in and then tried to put an output=FLOOD rule on the switch. > Without FlowVisor, the rule was installed OK, and all ports (except > for the packet in_port) were included in the action. But with > FlowVisor, only port 0 (local, right?) is included in the action. >
Hmm that is very strange. Could you send me your flowspace configuration? > That slice currently has acces to ANY match in ALL dpids (via > flowspace). Then I tried putting that slice as the > "default_flood_perm" in FlowVisor's general config and the arp/ping > worked fine. > > If I may complement, I was checking out the way FlowVisor handles > output actions > (https://github.com/OPENNETWORKINGLAB/flowvisor/blob/1.4-MAINT/src/org/flowvisor/message/actions/FVActionOutput.java). > I think that by putting that slice in the "default_flood_perm", it > entered one of those "shortcuts" and executed fine. > > But, according to the last part of that .java, even if the slice > doesn't have "default_flood_perms", FlowVisor should translate that > OFP_FLOOD into all allowed ports of the slice (in that switch), right? > So that my ping would work even without those flood_perms. Yes that is correct, something seems to be broken here. Could you send me a packet capture of FV receiving FLOOD actions and rewriting them incorrectly. > > My original question was because, if I really need to have those > "default flood perms" to everyone trying an output=FLOOD, why not be > able to specify more than one slice per switch? You can only specify one slice because otherwise slices would receive packets for which they are not responsible for. The whole default floodperm feature is dangerous in general anyway. > > > Sorry for the long message and thanks again! > > > Best regards, > > Victor T. > > On 26 December 2013 18:40, Ali Al-Shabibi <[email protected]> wrote: >> Hi Victor, >> >> By default FlowVisor tags LLDP packets coming from a controller so that when >> the packet comes back it can figure out which slice/controller to send it >> to. This way every controller sees the entire physical network (modulo its >> flowspace). Does this not work for you? Could you explain what you are >> trying to achieve? >> >> Cheers. >> >> -- >> Ali >> >> On 25 Dec 2013, at 20:09, Victor Torres <[email protected]> wrote: >> >>> Hello, >>> >>> I would like to know if there is a way to put multiple slices to >>> "default flood permissions" over all DPIDs, or to put multiple slices >>> as having Flood Permissions over a specific DPID. Everytime I try, it >>> overwrites it instead. >>> >>> This is critical for a testbed environment we are setting up. If there >>> is a reason for this being not implemented, I would be very glad if >>> anyone could tell me. >>> >>> >>> >>> Thanks in advance, >>> >>> Victor T. >>> _______________________________________________ >>> openflow-discuss mailing list >>> [email protected] >>> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss >> _______________________________________________ openflow-discuss mailing list [email protected] https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
