Re: virtual wireless interfaces / WDS
"M. Warner Losh" <[email protected]> Sat, 19 Apr 2003 17:42:22 -0600 (MDT)
| Newsgroups | gmane.network.wireless.bsd.general |
|---|---|
| Message-ID | <[email protected]> |
In message: <[email protected]> David Young <[email protected]> writes: : > As already stated, unique MACs would need to get a consistent sta#. : : Right. I think it is important to reach some consensus on how to solve : this problem before programming begins. Here is my proposal. : : 1 Let an operator or a daemon pre-create a sta with or without : a STA's MAC pre-assigned. : : 2 Give each 802.11 device (e.g., wi) a flag that permits/forbids : creating additional stas on demand. : : 3 Preserve the (MAC, unit number) relationship for every pre-created sta : forever. : : 4 For a sta that is created on demand, only preserve the (MAC, unit : number) relationship for some user-settable time-out period. Let : two of the time-out periods be 0 and infinity. : : 5 Let a flag in the 802.11 device (e.g., wi) tell whether STA : (dis)association is indicated by adding/deleting a sta or by twiddling : a sta's media status. I don't think we can rely on staN <-> MAC mapping. It would be better to provide the MAC to staN mapping as a filter. I'm not convinced that having consistant staN is worth the hassle that it is going to cause you to get it implemented. In the end, you'd need to have these mapping on some stable storage. However, having said that, If you want to deal with all this hair, I won't object. It does mean that you have to have the sta's configured before you can start the sta daemon. This isn't that different from the limits that we have now that you have to configure the wiN interface before stations can associate. Is there an upper limit on the number of sta that can be in a given system? If not, then some ***hole user can take down your hostap by randomizing the MAC address on the packets it sends out ot associate. In a busy cafe, this can really suck :-(. You'd also want some way to say 'Don't create more than N sta's' without having to create sta0 to staN-1 to accomplish that. Warner -- *bsd wireless list, a bawug thing <http://www.bawug.org/> [un]subscribe: http://lists.bawug.org/mailman/listinfo/bsd-wireless/