Client as bridge

Michael Smith msmith
Wed May 3 14:42:19 PDT 2006


I'm testing a setup where the wireless client bridges its ethernet and 
wireless interfaces. As far as I can tell, this should work if both ends 
support four-address headers. I got this setup to work:

* madwifi-ng master, svn rev r1472
* hostapd 0.4.8 configured for WPA2 with AES and preshared key
* "iwpriv ath0 wds 1" to enable 4-address headers

* madwifi-ng client, svn rev r1472
* wpa_supplicant 0.5.2 with the new -bbr0 option to select bridge device
* "iwpriv ath0 wds 1"
* "brctl setfd br0 0" to turn off forwarding delay - otherwise, if the 
  wireless card loses association and reassociates, the toggling of ath0's 
  forwarding state causes EAPOL frames to be dropped

Now I want to know if HostAP can do the same:

* hostap-driver 0.3.9, card with STA firmware 1.8.4
* just using 128-bit WEP for now

* madwifi-ng client, r1472
* 128-bit WEP
* iwpriv ath0 wds 1
* brctl setfd br0 0

When the AP receives a packet from the STA it logs:

wifi0: received WDS[4 addr] frame from unknown TA=00:02:6f:zz:xx:yy

(the MAC address of the client's wireless card)

And the AP doesn't send frames to the STA using four-address headers, 
either. Looking in hostap_80211_tx.c, in master mode four-address headers 
are never used unless the packet is going out a wlan#wds# interface.

Does HostAP just need a flag similar to "iwpriv .. wds 1" to force 
four-address headers, or is there other logic that would need to be added, 
like maybe a mapping of destination addresses to associated STA for 

Is this client-side bridging a bad idea in the first place?


More information about the Hostap mailing list