Re: No inter-AP Roaming?

[Dan Williams <dcbw redhat com>]

On Tue, 2007-12-04 at 22:34 +0100, dragoran wrote:
> On Dec 4, 2007 10:08 PM, Derek Atkins <warlord mit edu> wrote:
> > Quoting Dan Williams <dcbw redhat com>:
> >
> > > On Tue, 2007-12-04 at 15:31 -0500, Derek Atkins wrote:
> > >> I'm sitting at the IETF, connected to the 802.11 network..
> > >> But as I wander from room to room, NM doesn't roam.  It wont
> > >> jump from one AP to another as I move around the conference
> > >> facility.  I'll get to my new room and find that I no longer
> > >> have any network and then have to wait for a full connection
> > >> sequence.
> > >>
> > >> What gives?  I certainly don't have this problem if I just
> > >> let the driver/hardware roam on its own.
> > >
> > > Well, NM doesn't control roaming at all, wpa_supplicant and the driver
> > > do.  Sometimes wpa_supplicant will push the BSSID down to the driver
> > > when connecting (not entirely sure when that is) and if the driver
> > > actually follows WEXT (which lots don't) that would cause the driver to
> > > lock to that single BSSID and not switch.  That bit is highly driver
> > > dependent.  It looks like wpa_supplicant will only do this when it has
> > > previously timed out authentication with the station but tries again
> > > (like it should).  It then passes the BSSID down to the association code
> > > which tries to lock the card to that BSSID specifically.
> >
> > Interesting.
> >
> > I'm using the iwl3945 driver.  I DO see some authentication timeouts,
> > which is really weird.. Sometimes to the point where it wont even
> > successfully authenticate to the card at all and I have to reset.
> > Very frustrating.
> 
> that might be because the card is doing a scan at that time.. there is
> a patch posted upstream (linux-wireless) that should fix this by
> delaying scans while authenticate

Geez, it allowed scans during (re)association?  That's so broken it's
almost funny...

Dan