Hi, Sam On Tue, Mar 26, 2013 at 1:54 AM, Sam Leffler <[email protected]> wrote: > It appears there's no ordering to the list of known mesh gates. I'm > assuming this is because noone uses mesh gates except to fallback for > unresolved paths?
As far as I know, currently, we don't have mechanism to choose the best mesh gates. If one mesh STA located outside the MBSS is reachable via two mesh gates, only the the latest mesh gate updated based on last received frame from that mesh STA will be used. Assuming STP to be turned on to avoid looping. In ieee80211_rx_h_mesh_fwding, ..... if (!ether_addr_equal(mppath->mpp, mpp_addr)) memcpy(mppath->mpp, mpp_addr, ETH_ALEN); ..... For mesh STA located inside the MBSS, only the latest path will be used which is updated with specific interval. > I'm adding an anycast mechanism for finding the "best mesh gate" (a la what > OLPC did/does). It seems mesh gates are installed in the mesh_paths hash > table so all I need is to map my fixed anycast address to a known mesh gate > and everything else will just work. The only decision I see is whether to > register a hash-table entry for the anycast address that points to the > current "best mesh gate" or to intercept the address and select the best > mesh gate on every use. The former seems better since it also lets me see > what the current choice is with iw. > > OTOH after writing the above it seems like I can even avoid doing this if > all my mesh gates publish a path to the anycast address. Then the normal > path discovery mechanism would find it because only mesh gates would answer > for the anycast address. Perhaps, can make use of RANN element to publish this. > Comments? Others? --- Chun-Yeow _______________________________________________ Devel mailing list [email protected] http://lists.open80211s.org/cgi-bin/mailman/listinfo/devel
