wpa_supplicant limitation with a large number of APs.

YenJung Chang yenjung at gmail.com
Fri Feb 25 00:12:48 EST 2005


Thanks for your reply in advance. :)

On Thu, 24 Feb 2005 20:08:44 -0800, Jouni Malinen <jkmaline at cc.hut.fi> wrote:
> On Fri, Feb 25, 2005 at 11:56:52AM +0800, YenJung Chang wrote:
> 
> > wpa_supplicant has problem to connect to a connecting available AP if
> > the AP is not in the
> > scanned result.
> > This problem would happen in the environment with much APs over than about 20.
> > The memory size of passing scanning result between wpa_supplicant and driver is
> > 4096, and the total number of APs can passed is only about 20.
> 
> That's not true with the latest wpa_supplicant version; or well, at
> least it is not true for the drivers that use wireless extensions for
> getting scan results. Yes, there is still a limit, but it is not 4k or
> 20 APs.
I said that the memory size of scanning result is 4096 is bcz the
variable "res_buf_len" in funcion wpa_driver_wext_get_scan_results in
driver_wext.c
(wpa_supplicant 0.3.8) is assigned IW_SCAN_MAX_DATA (4096).
And the max got number of AP is about 20 is the number I most saw when 
wpa_supplicant ran.

So it is possible that wpa_supplicant gets more APs?

> > Is it possible that wpa_supplicant tries to connect to target AP
> > regardless of scanning result?
> 
> Yes, if the driver supports ap_scan=2 mode, wpa_supplicant does not need
> to request scan results. Another option is to use scan_ssid=1 option to
> limit the scan to one SSID.

I tried ap_scan=2 without security. Unfortunately, it failed.
Is there any restrication on ap_scan=2 ?
If no, I will post my log of failure to the list later. :)

> --
> Jouni Malinen                                            PGP id EFC895FA


I apologize if you feel I am impolite. 
It is not my intention. :)
My English is not good enough.

Thanks,
YJ.



More information about the HostAP mailing list