[PATCH 03/11] wpa_s: use sched_scan in driver init

[Luciano Coelho]

Hi Alex,

On Wed, 2011-10-05 at 12:38 +0200, Bondar, Alexander wrote: 
> > diff --git a/wpa_supplicant/sme.c b/wpa_supplicant/sme.c index dbf385e..4cb5062 100644
> > --- a/wpa_supplicant/sme.c
> > +++ b/wpa_supplicant/sme.c
> > @@ -237,6 +237,7 @@ void sme_authenticate(struct wpa_supplicant *wpa_s,
> >  	}
> >  #endif /* CONFIG_P2P */
> >  
> > +	wpa_supplicant_cancel_sched_scan(wpa_s);
> >  	wpa_supplicant_cancel_scan(wpa_s);
> 
> Don't you need to wait for EVENT_SCHED_SCAN_STOPPED before proceeding with authentication flow? There may be
> race condition when authentication request is set to the driver while sched_scan is still in progress.

I don't see a problem with that.  Nothing prevents the driver and the
firmware from supporting authentication while a sched_scan is running
(in theory).  If it cannot be done, the lower driver/hardware can always
send a sched_scan stopped event before continuing with authentication
(like the wl12xx driver does by checking the mac80211 idle flag).

And furthermore, here it's the same thing as with normal scan.  We don't
want to keep scanning after we found what we want, so we call
cancel_scan().  But we don't really care if the scan keeps running
during authentication.

-- 
Cheers,
Luca.