On Mon, 2008-09-15 at 17:56 +0200, Johannes Berg wrote:
> > + if (sdata->vif.type == IEEE80211_IF_TYPE_STA) {
> > + struct ieee80211_if_sta *ifsta = &sdata->u.sta;
> > + if (erq->flags & IW_ENCODE_OPEN) {
> > + printk("wext: open\n");
> > + ifsta->auth_alg = WLAN_AUTH_OPEN;
> > + }
> > + else if (erq->flags & IW_ENCODE_RESTRICTED) {
> > + printk("wext: shared key\n");
> > + ifsta->auth_alg = WLAN_AUTH_SHARED_KEY;
> > +
>
> Come to think of it, since we already fall back from open to shared-key
> when the AP rejects, why would we want to support this at all? The
> semantics aren't well defined enough to support it _properly_, that is
> have it change the ifsta->auth_algs variable, and touching this variable
> as you're doing in these two patches is just plain wrong since it can be
> changed underneath (yes, you're removing the automatic changing, but
> that's totally wrong).
Then again, I suppose you _can_ actually implement this, but in terms of
which authentication algorithms are allowed, something like:
if (erq->flags & RESTRICTED)
ifsta->auth_algs = SHARED;
else
ifsta->auth_algs = OPEN | SHARED | LEAP;
johannes
Attachment:
signature.asc
Description: This is a digitally signed message part
