On Mon, 2020-03-02 at 17:44 +0000, Cedric VONCKEN wrote:
> I continue my investigation, and I think I found an interesting point.
>
> My AP have 3 bssid, the wlan0 is set in WPA-EAP, the wlan0_1 is
> set in WPA-PSK and the latest wlan0_2 is without security. It is
> useful for my test, I only need to change the sta configuration.
>
> I added some debug in net/mac80211/tx.c
>
> When I use the BSSID set in WPA-PSK (this case works), the
> broadcast frames are sent through the netdev wlan0_1.sta1.
> When I use the BSSID set in WPA-EAP (this case doesn't work) the
> broadcast frames are sent through the netdev wlan0 (use fast_xmit
> path). But the wlan0.sta1 netdev is created in the system, added to
> the bridge, and received the frame from the bridge. But in this case
> the function ieee80211_lookup_ra_sta return ENOLINK for wlan0.sta1
> and the frame is dropped.
>
> So I think the issue is in transmit side, I don't why, but with
> WPA-EAP the mac80211 state seem different and the WDS doesn't work.
>
> Do you have any explanation or idea to locate the origin ?
Not really. The code there is just
switch (sdata->vif.type) {
case NL80211_IFTYPE_AP_VLAN:
sta = rcu_dereference(sdata->u.vlan.sta);
if (sta) {
*sta_out = sta;
return 0;
} else if (sdata->wdev.use_4addr) {
return -ENOLINK;
}
as far as I can tell, right?
So if you got into the -ENOLINK case, then it means that userspace
didn't add the right station to the 4-addr VLAN interface, which I guess
was "wlan0.sta1"? Not sure how that'd happen.
johannes
