CFG80211 support in wpa_supplicant
dcbw at redhat.com
Mon Jan 18 15:19:57 EST 2010
On Fri, 2010-01-15 at 23:39 +0200, Maxim Levitsky wrote:
> On Fri, 2010-01-15 at 11:54 -0800, Dan Williams wrote:
> > On Fri, 2010-01-15 at 11:25 +0100, Holger Schurig wrote:
> > > > Also NM doesn't show proper signal levels. I also created more
> > > or less a hack to make it work:
> > >
> > > > + if (max_qual->updated & IW_QUAL_DBM)
> > >
> > >
> > > That looks like NM's signal level is still based on WEXT,
> > > IW_QUAL_xxx looks like a wext identifier.
> > >
> > > On my device I use a self-written app that get's the signal/noise
> > > via nl80211 (like the "iw" tool), and then it works pretty nice.
> > Yeah, NM's signal quality code hasn't been touched in quite a while.
> > I'd appreciate suggestions for fixing it up... Given that mac80211
> > drivers will now only return level, and I believe only in dBm, and often
> > without a noise floor, what should the calculation look like?
> I don't know yet an exact formula, but I know solid facts about quality
> versus dBm:
> * anything above -80 dBm is plain perfect.
> * -80 to -85 is more or less ok, but expect trouble.
> * -86 to -89 is very low, unstable connection, almost unusable
> * -90 is the signal at which it pretty much disappears.
> I also strongly dislike the signal bars that NM has now. Sure they are
> pretty, but they give much less information about signal level that old
> progress bar indicators.
> But if you add the dBm near the indicator, it would be just perfect.
The tooltip on mouseover gives you the actual quality % of your current
connection (as reported by the driver or calculated by NM); the applet
doesn't try to be a wardriving tool but maybe we could push dBm into the
The problem is that since the drivers are so inconsistent WRT to signal
strength reporting, and since a difference of 5% in the signal bars is
actually statistically irrelevant because of the signal strength
reporting, it's pretty pointless to have the progress bars. I'd argue
that even a 10% resolution in the menu is too much resolution because of
this. The current bars give you a 20% resolution, so maybe we split the
difference and 15% (six signal bars) is the right amount to make most
people happy. But anything more than that and we're in wifi-ricer-land
because a 5% difference simply isn't large enough to make an intelligent
decision about, for the computer or the user.
More information about the HostAP