I think I got it

[Jouni Malinen]

On Mon, Apr 05, 2004 at 01:59:18PM -0700, mike-hostap at tiedyenetworks.com wrote:

> 	So this weird problem seems to take a strange turn - the bridge 
> doesn't seem to have a complete table. For example, the station 
> '00:50:04:99:34:4e' (a server), one that the ap complains about, is on the 
> wired side and is the syslog server. If I dump the arp cache, we see:

This matches with my understanding and the part I mentioned in a
previous email about this being valid situation. I don't know why the
bridge has shorter timeout on the entry that ARP table (or
alternatively, the entry is dropped from the bridge or or not updated
based on locally generated frames). Anyway, the bridge is certainly
allowed to send unicast frames to all ports if it does not know the
correct port.

> 	You can see that 00:50:04:99:34:4e magically showed up (finially), 
> and I'm not getting "drop packet to non-associated sta" anymore. 

Thank you for tracing this. I replaced the message with a counter
(tx_drop_nonassoc) in the development branch. This should keep syslog
quite a bit cleaner on systems that trigger this behavior in the bridge
code.

-- 
Jouni Malinen                                            PGP id EFC895FA