hostapd & mac80211: too many probe responses

I've just noticed that hostapd + mac80211 + rt2x00 produces way too many probe responses when setup as multi bssid AP.

The reason seems to be that mac80211 receives a broadcast probe req and drops it to the monitor interface for each vif (hence X times). And hostapd will process each received broadcast probe req on all vifs as well (hence X times again). Resulting in X*X probe responses instead of just X.

This can eat up quite some airtime as the probe reponses are sent at 1Mpbs.

IMO to fix this issue mac80211 should send a broadcast probe req only once to the monitor interface. However, I don't have a good idea how to do that right now :)

Quoted 9 lines

I've just noticed that hostapd + mac80211 + rt2x00 produces way too many probe responses when setup as multi bssid AP.

The reason seems to be that mac80211 receives a broadcast probe req and drops it to the monitor interface for each vif (hence X times). And hostapd will process each received broadcast probe req on all vifs as well (hence X times again). Resulting in X*X probe responses instead of just X.

Quoted 5 lines

This can eat up quite some airtime as the probe reponses are sent at 1Mpbs.

IMO to fix this issue mac80211 should send a broadcast probe req only once to the monitor interface. However, I don't have a good idea how to do that right now :)

Looks like we broke that then, IEEE80211_RX_CMNTR was intended to do this.

Quoted 2 lines

Looks like we broke that then, IEEE80211_RX_CMNTR was intended to do this.

Ah, I see. I'll do some more debugging in that direction first ...

Quoted 5 lines

Quoted 3 lines

Looks like we broke that then, IEEE80211_RX_CMNTR was intended to do this.

Ah, I see. I'll do some more debugging in that direction first ...

I think the 'rx->flags = 0;' thing is just wrong.