wpa_supplicant seems to prefer 2.4GHz over 5GHz

[Jouni Malinen]

On Thu, Feb 16, 2017 at 08:10:44AM +0000, Peer, Ilan wrote:
> According to the wpa_supplicant AP scoring, it would indeed seem that the 2.4GHz AP is proffered since it has better SNR (although the SNR for both is good. Generally, wpa_s considers any SNR above 30 as good).
> 
> a0:55:4f:66:d5:18 freq=2447 qual=0 noise=-89~ level=-49 snr=40* flags=0xb age=0 est=65000
> a0:55:4f:66:d5:10 freq=5240 qual=0 noise=-92~ level=-67 snr=25 flags=0xb age=0 est=390001
> 
> perhaps the scoring can be tweaked to more aggressively favor 5.2 in case of close good SNR. 

SNR of 25 is not above the current GREAT_SNR (30) definition. Because
of that, this happens to be just one dB too large a difference to allow
estimated throughput to be considered since abs(snr_b - snr_a) == 5
in this case (that snr=40 gets dropped to GREAT_SNR = 30) and < 5 would
have resulted in using the throughput estimate (of which the 5 GHz AP
here has significantly higher due to possibility of using VHT80).

Taken into account the throughput estimates (which are much newer than
the GREAT_SNR definition) are actually assuming the highest rate to be
available with SNR of 24, it would sound reasonable to drop the
GREAT_SNR value to something like 25. The 30 is already documented as
being "a conservative value"..

If you (John) want to test the impact of such a change, modify the
following line in wpa_supplicant/scan.c to use value 25 instead of 30:

#define GREAT_SNR 30

and rebuild wpa_supplicant.

-- 
Jouni Malinen                                            PGP id EFC895FA