Re: Making sense of cfg80211_scan_request.n_ssids

[Luciano Coelho <coelho@ti.com>]

On Mon, 2011-08-08 at 10:37 +0200, Johannes Berg wrote: 
> On Tue, 2011-08-02 at 18:10 -0400, Pavel Roskin wrote:
> 
> > > Both of those commands result in libertas being called with
> > > cfg80211_scan_request with n_ssids=1 and the "test" SSID present in
> > > cfg80211_scan_request.ssids[0], as expected.
> > >
> > >
> > >    iwlist eth0 scan
> > >    iw dev eth0 scan
> > >
> > > Both of those commands result in libertas being called with
> > > cfg80211_scan_request with n_ssids=1 and a zero-length SSID present in
> > > cfg80211_scan_request.ssids[0].
> > > Is this a bug? I would expect it to be passed n_ssids=0 in this case.
> > 
> > I believe n_ssids is the number of probe requests to be sent per 
> > channel.  It scanning for all SSIDs, one probe request is sent.  It is 
> > only for passive scanning that we don't need any probes.
> 
> Yeah. "iw dev eth0 scan passive" will cause n_ssids to be 0.

I think the question was slightly different.  As I understand, Daniel
wants to know why the n_ssids is 1 even if no SSID is passed from user
space.

A zero-length SSID is a broadcast SSID.  So probe requests *will* be
sent, but with the broadcast SSID.  If you explicitly ask for a passive
scan (as Johannes mentioned), then n_ssids will be 0 which means that no
probe requests should be sent at all.

The n_ssids field is the number of *different* probe_reqs that will be
sent per channel.  Some drivers will send more than one probe_req with
the same SSID per channel (eg. wl12xx sends 2 probe_reqs per SSID on
each channel).


-- 
Cheers,
Luca.