Re: Signal Strength on Atheros?

[Dan Williams <dcbw redhat com>]

On Tue, 2006-05-02 at 17:38 -0400, Dan Williams wrote:
> On Sun, 2006-04-30 at 21:35 -0400, Darren Albers wrote:
> > This was discussed in the past, the end result was that the Atheros
> > card reports signal strength differently than other cards.  I think
> > Robert Love posted a patch for it but it was not implemented since the
> > NM devs want to avoid making special case exceptions.
> > 
> > Here are some links to previous discussions on the subject:
> > http://mail.gnome.org/archives/networkmanager-list/2005-February/msg00063.html
> > http://mail.gnome.org/archives/networkmanager-list/2006-January/msg00141.html
> > 
> > Here is the madwifi teams take on the subject:
> > http://madwifi.org/wiki/UserDocs/RSSI
> 
> Current signal strength reporting for madwifi-ng is bogus and completely
> broken WRT to the WE spec.
> 
> max_qual is set to the theoretical max value of the variable (a u8
> type), and not the real-world max of the card.
> 
> The 'qual' field of the WE quality is for link "quality", which is
> subjective, while the "level" field is for the link's actual radio
> levels.
> 
> There are two options for the driver:
> 
> 1) Actually use "quality": qual->qual must be bounded by 0 and
> max_qual->qual, and must change in a linear fashion.  It is defined as a
> relative value, and therefore is a percentage.  ipw2100 uses a signal
> quality of 0 -> 100%, which is a combination of RSSI/dBm, packet loss,
> link speed, missed beacons, etc.  That is good.
> 
> 2) Use the absolute or relative "levels" values: max_qual->level is 0
> and qual->level is a valid RSSI/dBm value, and qual->noise and/or
> max_qual->noise is valid.  OR: max_qual->level > 0, and qual->level is
> valid.
> 
> At least, that was the state when I last looked at the quality for
> madwifi-ng, which was on Jan 26 2006, and also for r1475 from early
> April.
> 
> My comments in the NM strength calculation are:
> 
> 	/* If the driver doesn't specify a complete and valid quality, we have two options:
> 	 *
> 	 * 1) dBm: driver must specify max_qual->level = 0, and have valid values for
> 	 *        qual->level and (qual->noise OR max_qual->noise)
> 	 * 2) raw RSSI: driver must specify max_qual->level > 0, and have valid values for
> 	 *        qual->level and max_qual->level

Note that with WE-19 and later, a flag was added to specify that levels
are explicitly in dBm, and therefore a max_qual->level of 0 is permitted
provided that flag is present in the max_qual->updated and qual->updated
fields.  NM doesn't honor that particular flag yet, which is a valid bug
against NM.  But I don't think madwifi-ng does this at all, and
therefore there's still a problem with madwifi.

Dan

> Just leaving it in RSSI without doing the proper encoding is a cop-out.
> It's not up to userspace to figure out WTF the quality is, it's up to
> the driver to present a sane view of it to the world, because only the
> driver knows about the variations between chipsets and radios that each
> card uses.  Manufacturers have tables of how RSSI for their chipset and
> radio map to dBm, and if the driver writers don't know that, then they
> need to pick some empirical values and fake it.  Stick your antenna
> right next to an access point, and see what the highest RSSI is.  Call
> that 100%.  Walk away until the signal drops.  Call that 0%.  If you're
> in RSSI, you don't have to do anything else; that's your 0 -> 100%
> range.  If it's dBm, you have to code it from quadratic to linear.  But
> that's pretty simple.  Madwifi kind of chickened out here.
> 
> But the patch wouldn't be all that big.  In reality, they've got bigger
> problems, like not being upstream in the kernel and not using one of the
> two softmac stacks that are now in the kernel (ipw+softmac and
> devicescape).  Madwifi is the black sheep and getting more so every
> month...
> 
> Dan
> 
> > On 4/30/06, Pat Suwalski <pat suwalski net> wrote:
> > > Hello,
> > >
> > > My ath_pci card is showing signal strength at about half of what it
> > > should be. This is consistent with how iwlist shows it, but the GNOME
> > > wireless applet has always shown it correctly.
> > >
> > > I don't know off-hand if it's always special cased, but it looks like it
> > > might have to be. Maybe when they support the wireless extensions this
> > > will be fixed.
> > >
> > > Any ideas?
> > >
> > > --Pat
> > > _______________________________________________
> > > NetworkManager-list mailing list
> > > NetworkManager-list gnome org
> > > http://mail.gnome.org/mailman/listinfo/networkmanager-list
> > >
> > _______________________________________________
> > NetworkManager-list mailing list
> > NetworkManager-list gnome org
> > http://mail.gnome.org/mailman/listinfo/networkmanager-list
> 
> _______________________________________________
> NetworkManager-list mailing list
> NetworkManager-list gnome org
> http://mail.gnome.org/mailman/listinfo/networkmanager-list