wpa_supplicant_select_bss error?

[Dan Williams]

On Sun, 2013-12-15 at 21:53 +1100, a.brooks2 at marathon-targets.com wrote:
> Hi,
> 
> I've been trying to track down why a client seems to connect to the 
> (significantly) weaker of two access-points with identical SSIDs, and 
> refuses to roam to the stronger one subsequently.
> 
> I think there might be an error in wpa_supplicant_select_bss:

Scan results are sorted when they arrive from the driver in
wpa_supplicant_get_scan_results() using the wpa_scan_result_compar()
function, a huge component of which is signal strength.  Thus, when
wpa_supplicant_select_bss() returns the first matching AP, it should
already be the best AP because the AP list has been sorted.

So I would concentrate your debugging on that function, and find out why
it decides your "better" AP is less usable than your "worse" AP.

Dan

> static struct wpa_bss *
> wpa_supplicant_select_bss(struct wpa_supplicant *wpa_s,
>                struct wpa_ssid *group,
>                struct wpa_ssid **selected_ssid)
> {
>      unsigned int i;
> 
>      wpa_dbg(wpa_s, MSG_DEBUG, "Selecting BSS from priority group %d",
>          group->priority);
> 
>      for (i = 0; i < wpa_s->last_scan_res_used; i++) {
>          struct wpa_bss *bss = wpa_s->last_scan_res[i];
>          *selected_ssid = wpa_scan_res_match(wpa_s, i, bss, group);
>          if (!*selected_ssid)
>              continue;
>          wpa_dbg(wpa_s, MSG_DEBUG, "   selected BSS " MACSTR
>              " ssid='%s'",
>              MAC2STR(bss->bssid),
>              wpa_ssid_txt(bss->ssid, bss->ssid_len));
>          return bss;
>      }
> 
>      return NULL;
> }
> 
> 
> The loop returns on the first BSS it finds with the appropriate ssid.  
> So the results depend on the order of the scan results.
> In my case I have a client which happens to list the weaker BSS first, 
> so it does the wrong thing.
> 
> I think this loop should find the best BSS instead.  I made this change 
> and now my client happily connects to the correct BSS, and roams 
> appropriately.
> 
> I can submit a patch but thought I'd ask the question first (this code 
> seems to have been around for a while so I'm surprised no-one has found 
> it if it is in fact a bug), let me know.
> 
> 
> 
> Cheers,
> 
> Alex
> _______________________________________________
> HostAP mailing list
> HostAP at lists.shmoo.com
> http://lists.shmoo.com/mailman/listinfo/hostap