Bss command with p2p_dev_addr= argument

[Jouni Malinen]

On Thu, May 03, 2012 at 05:52:28AM +0000, Neeraj Kumar Garg wrote:
> This is regarding the bss command recently implemented in wpa_supplicant to find out if a peer p2p device is running a bss (P2P GO interface) or not. Since I could not think of a good implementation/patch, I am requesting your opinion on below scenario. Please assume wlan0 is primary interface and p2p-wlan0-x is p2p interface.

Please notice that there has been some recent patches that bring in
alternative mechanisms for this. The p2p_connect-auto mechanism (i.e.,
replace "join" argument with "auto") uses improved mechanism to verify
whether a peer is operating as a GO in the last scan result update.

> Problems:
> 
> 1)      Bss expiration age has been configured to 180 seconds. A bss entry is removed only either after 3 minutes or if 2 directed scans for that particular bss doesn't result into a positive scan result. A p2p_find does a scan for a DIRECT- ssid.

This is by design and not really problem when used properly..
p2p_connect-auto does this automatically by verifying whether the
bss->last_update_idx was changed during the last scan run. In other
words, the information about the GO is kept for 180 seconds, but a new
scan run is used to confirm that the peer is still operating as GO at
the time of the connection request.

> 2)      If we choose to define bss expiration age to lower value say 10 seconds, then we might have issues in Group-formation stages as bss entry may expire in between of group-formation.

Yes, I don't really want to make BSS entries expire much faster.

> 3)      If we force p2p_flush to flush off the bss entries from wlan0 interface, a p2p_find followed by p2p_flush will again add an entry as GO is existing at this moment.

Similarly, I don't think we should be flushing BSS entries - that
last_update_idx is much better way of addressing this without risk for
side effects.

-- 
Jouni Malinen                                            PGP id EFC895FA