signal quality, prism firmware, WPA

Paul Fox pgf at brightstareng.com
Mon Dec 12 14:00:17 EST 2005


hi -- i have a very repeatable problem i'm hoping someone here
can shed light on.

we're using PCMCIA cards based on the (now Intersil) Prism 2.5
chipset.  the cards are the Engenius (or Senao) NL-2511CD Plus,
like this:
    http://www.netgate.com/product_info.php?cPath=26_39&products_id=43

as shipped, the cards we've gotten have firmware version 1.4.9,
which works just fine with WEP and the hostap drivers (version
0.3.9).  (we're not using hostapd -- we're just a station.)

however, we need WPA.  so i've upgraded some of our cards to
1.8.4.  (i've also tried intermediate versions, like 1.7.1, with
similar results.)  device identification currently looks like this:
    wifi0: NIC: id=0x800c v1.0.0
    wifi0: PRI: id=0x15 v1.1.1
    wifi0: STA: id=0x1f v1.8.4

when using WEP, everything still works fine, just as it did with 1.4.9.

however, when using WPA, the signal quality, strength, and noise
values all report bogus results.  signal quality is consistently
reported as 0, strength and noise report as a constant non-zero
value.  (i've verified this at a low level with driver printks, so
specific tool versions can be eliminated i think.)

there are two RID values for fetching these parameters -- one is
"raw" (HFA384X_RID_COMMSQUALITY, or 0xFD43) and the other is
newer, with results in dbm (HFA384X_RID_DBMCOMMSQUALITY, or
0xFD51).  normally with newer firmware the latter RID command
is used.  i modified the driver to force the older api, and
the results are the same -- quality is consistently 0, strength
and noise are unvarying non-zero numbers.

i have (legitimate) access to the prism driver programmers
manual, and from what i can tell, the hostap driver is using the
API correctly.  (and, indeed, it seems to work just fine when WPA
isn't involved.)

testing the 1.8.4 firmware in a windows laptop shows that, as
expected, quality and strength values are somehow available,
since the prism driver and status utility report the values
without complaining.

has anyone else seen this behavior?  i've found references to
similar problems in the list archives, but the messages were
almost two years old...

many thanks in advance.

(so far the only oddness i've noticed is what i've detailed here. 
are there other stats i should be checking in particular to see
if they're also not reported correctly?)

paul
=---------------------
 paul fox, pgf at brightstareng.com



More information about the HostAP mailing list