Re: Q: iw - how to scan for a specific ssid / AP mode scan

[Zefir Kurtisi <zefir.kurtisi@neratec.com>]

On 07/03/2015 08:38 PM, Johannes Berg wrote:
> On Fri, 2015-07-03 at 18:51 +0200, Zefir Kurtisi wrote:
>> Folks,
>>
>> I have difficulties using iw for a specific use case or fail to 
>> understand the documentation correctly.
>>
>> My platform is a recent OpenWRT, running ath9k.
>>
>> First use case is scanning for a given ssid in managed mode. 
>> According do iw's documentation (and the attribute description in 
>> nl80211.h), issuing
>>
>> iw dev wlan0 scan flush ssid <SSID>
>>
>> should do exactly this, but I keep receiving a full list of visible 
>> APs.
> 
> This is telling it to scan for that particular network, and that's what
> it's going to do, but it's still going to report everything that it
> heard, for example when hearing beacons while scanning.
> 
Ah, that explains it. Though, intuitively when providing an SSID to scan for, I'd
expect other networks seen not to be displayed. If this is of general interest, I
could work out some extension to iw and provide a related patch.

>> The second issue is about scanning in AP mode. Where I want to go is
>> having two
>> APs operating on arbitrary DFS channels with periodic scans to 
>> discover each
>> other. What I observe is
>> a) passive scanning: iw dev wlan0 scan flush ap-force passive
>>    => does not work - no scan results are provided
>> b) active scanning: iw dev wlan0 scan flush ap-force
>>   * finds only a subset of APs compared to a scan in managed mode
>>   * finds only APs on non-DFS channels
>>
>> Again, I might be missing some relevant documentation, but to me the 
>> observed results look rather like 'not yet implemented' than inherent
>> limitations.
>>
> 
> Not sure - but you do need to realize that the AP isn't really allowed
> to go off-channel for any period of time (like scanning) so this isn't
> really guaranteed to work well. Especially passive scanning seems like
> a really bad idea. As to why it's not actually working, I have no idea.
> 
Your confirmation that scanning in AP mode is not reliable helps a lot.

As for the 'bad idea', this is a special case: the APs don't provide a service at
all, they only need to discover each other reliably in a mobile environment. I
assumed AP background scanning would do the trick out of the box, but now will
move to a time multiplexed AP / STA operation.


> johannes

Thanks a lot,
Zefir