Date: Fri, 5 Dec 2003 15:20:52 -0800 (PST)
To: jsun@junsun.net
Subject: Prism 2 flash success stories
From: Dave Platt <dplatt@radagast.org>

Hi.  I just succeeded in flash-upgrading two different
Prism-based cards, using the current hostap driver and
utilities.

First card: Proxim RangeLAN DS (a rebadged Zcomax XI-300,
I believe).  NIC 0x8002 version 1.0.0, primary firmware 0x15 0.3.0,
station firmware 0x1F version 0.8.3.

I upgraded the station firmware to version 1.5.6 (s1010506.hex),
and then to 1.7.1 (s1010701.hex).  Did not upgrade the primary
firmware.  The card seems to work fine with both upgraded versions,
using the hostap driver.  I'd previously been using it with the
orinoco_cs 0.13e driver, and it did _not_ work with this driver
after upgrading.  Analysis of the logs showed that the Orinoco
driver was mistakenly interpreting the new firmware as being
of the Lucent/Orinoco variety (the new firmware is "variant 1").

I modified the Orinoco driver's tests (which are acknowledged in
the code as "fundamentally broken") so that "variant 1" firmware
in the 1.xx revision range is recognized as Intersil rather
than Lucent.  With the patched Orinoco driver, the RangeLAN card
with 1.7.1 firmware is handled properly and seems to work quite
nicely.

I tested this card on my home network (40-bit WEP) using both
the firmware WEP, and host-based WEP.  In a ping-flood situation,
the firmware WEP had a significantly higher rate of lost packets
(about 23% were dropped) than the host-based WEP (about 4%), but
host WEP pegged the CPU at 100%.

Second card: D-Link DWL-650 (early rev, don't recall the subrevision
code right now).  NIC 0x8008 1.0.1, primary 0x15 0.3.0, station
0x1F 0.8.3.  I had to patch the prism2_srec tool to allow the
s101XXXX.hex firmware to be downloaded, due to the (known) mismatch
between the "supported platforms" in the firmware and the version
1.0.1 NIC in this card.

The 1.5.6 firmware worked flawlessly in this card.

The 1.7.1 had an odd behavior which I've occasionally seen before...
during times of heavy traffic, the card would occasionally report
a bogus received-signal strength.  This would cause my "wmwave"
on-screen link-indicator widget to blink out in a distracting
fashion.  I reverted back to 1.5.6 and the bogus behavior went
away, so I've left this card on 1.5.6.

Thanks for all of your work in this area - 'tis much appreciated!