On Tue, May 20, 2008 at 3:57 PM, Johannes Berg <johannes@xxxxxxxxxxxxxxxx> wrote: > On Tue, 2008-05-20 at 15:54 +0300, Tomas Winkler wrote: >> On Tue, May 20, 2008 at 10:56 AM, Helmut Schaa <hschaa@xxxxxxx> wrote: >> > Fix a possible NULL pointer dereference in ieee80211_compatible_rates >> > introduced in the patch "mac80211: fix association with some APs". If no bss >> > is available just use all supported rates in the association request. >> > >> > Signed-off-by: Helmut Schaa <hschaa@xxxxxxx> >> > --- >> > >> > diff --git a/net/mac80211/mlme.c b/net/mac80211/mlme.c >> > index 76ad4ed..3f7f92a 100644 >> > --- a/net/mac80211/mlme.c >> > +++ b/net/mac80211/mlme.c >> > @@ -721,7 +721,17 @@ static void ieee80211_send_assoc(struct net_device >> > *dev, >> > capab |= WLAN_CAPABILITY_PRIVACY; >> > if (bss->wmm_ie) >> > wmm = 1; >> > + >> > + /* get all rates supported by the device and the AP as >> > + * some APs don't like getting a superset of their rates >> > + * in the association request (e.g. D-Link DAP 1353 in >> > + * b-only mode) */ >> > + rates_len = ieee80211_compatible_rates(bss, sband, &rates); >> > + >> > ieee80211_rx_bss_put(dev, bss); >> > + } else { >> > + rates = ~0; >> > + rates_len = sband->n_bitrates; >> > } >> > >> > mgmt = (struct ieee80211_mgmt *) skb_put(skb, 24); >> > @@ -752,10 +762,7 @@ static void ieee80211_send_assoc(struct net_device >> > *dev, >> > *pos++ = ifsta->ssid_len; >> > memcpy(pos, ifsta->ssid, ifsta->ssid_len); >> > >> > - /* all supported rates should be added here but some APs >> > - * (e.g. D-Link DAP 1353 in b-only mode) don't like that >> > - * Therefore only add rates the AP supports */ >> > - rates_len = ieee80211_compatible_rates(bss, sband, &rates); >> > + /* add all rates which were marked to be used above */ >> > supp_rates_len = rates_len; >> > if (supp_rates_len > 8) >> > supp_rates_len = 8; >> > >> > >> >> I found one ieee80211_rx_bss_{get,put} imbalance in >> ieee80211_sta_join_ibss function >> That may cause this problem yet it doesn't look like this is the case. >> ieee80211_sta_join_ibss >> calls ieee80211_rx_bss_put on 'bss' that it receives as an argument >> Keep searching. > > Hm. Send a patch? :) > Searching for the real problem ... >> I suggest to insert at least some WARN_ON(1) for the else case. > > Disagree, not until somebody audits the code. We already know it can > happen and a WARN() won't help us track it down because it provides no > additional information (stack trace is useless) What about printk(KERN_WARN ), The else statement actually means that something wrong happened. Tomas -- 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