Hello! Problem: PTK (and sometimes GTK-rekeying) is broken under high network load. The high network load is produced with netperf: while true ; do netperf -t TCP_STREAM -H host netperf -t TCP_MAERTS -H host netperf -t TCP_SNDFILE -H host done During rekeying, mostly of PTK, the 4 way handshake is broken, because hostapd doesn't get the last answer from supplicant. Supplicant itself thinks, all would be fine - PTK rekeying would be ready. Goal: Near down the component, which is responsible for the broken PTK rekeying. The following tests all where done with a WPA2 / PSK / ieee80211n / 40 MHz bandwith, 2,4 GHz based connection and a PTK rekeying delta of 50 s configured on hostapd. hwcrypt was switched off in hardware for hostapd. I did the tests with different components. The result can be found in the table below. Components: Hardware for accesspoint: A1) Linksys WMP600N (rt2800pci) or A2) Atheros ar9285 (ath9k) Hardware for Supplicant: H1) WUSB600Nv2 H2) Atheros ar9285 Drivers for Supplicant D1) wpa_supplicant with wext D2) wpa_supplicant with nl80211 D3) rt3572sta without wpa_supplicant 1. test accesspoint Hardware for Driver for supplicant supplicant Rekeying -------------------------------------------------------- A1 H1 D1 broken A1 H1 D3 fine A1 H2 D1 broken A1 H2 D2 broken A2 H1 D1 broken A2 H1 D3 fine fine means: Rekeying never broke during > 1,5 h. broken means: Rekeying was broken after the first or no later then fifth incidence. Summary: Rekeying is fine as long as no wpa_supplicant / wext / or nl80211 framework is used on the supplicant. The rt3572sta-driver without wpa_supplicant works just fine - as expected. The problem could be in the wireless framework or in wpa_supplicant - that's what I cannot say. But I could see one big difference between the handling of the rekying by wpa_supplicant and rt3572sta-driver: the rt3572sta-driver stops the datastream during rekeying. This is about half a second. With wpa_supplicant, I can't see this stop. It stops after the rekeying has been broken because of deauthentication. The following reauthentication takes much longer and sometimes doesn't work at all. Could somebody please look at this problem? If you have some more questions - please ask - I'll try to answer them. Andreas -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
