Broadcom hybrid_wl driver

[Jarod Wilson]

On Sun, 2008-10-26 at 22:21 +0000, Chris Nolan wrote:
> Jarod Wilson wrote:
> >
> > Heh, sorry 'bout that... Although to be honest, I dunno that I'd suggest
> > a kmod as a first package for *anyone*, since they consist of so much
> > black magic... :)
> >
> > However, I'd be perfectly happy if you wanted to take this on as its
> > primary maintainer, with me as a co-maintainer.
> >
> >   
> Fair point about a kmod not being an ideal place to start! I've tried to 
> read a fair bit about kmodtool and packaging kmods but clearly my 
> understanding is still rather limited. However I felt it was important 
> to start with a package that I actually use on a day-to-day basis so 
> that I'm in a good position to follow the development and 
> find/apply/create new patches.
> 
> I'd be delighted to have you as a co-maintainer - it would be great to 
> have someone to learn from and help answer any questions or problems I 
> come across!

Works for me! :)


> > Okay, so back to the Linux side and the hybrid_wl driver... I get no
> > better than ~350KB/s. I presume the driver isn't handling the
> > single-antenna very well or something, but a buddy of mine who also got
> > a Dell w/the same card said hybrid_wl performed pretty badly for him too
> > (the Dell has two antenna internally), but was having stability issues
> > with ndiswrapper. I've used ndiswrapper w/the BCM432<something> chip in
> > my AppleTV with excellent results though, so I'm off to try that next...

Used what I believe were the same bits I used on the AppleTV, and the
performance was even worse than w/hybrid_wl, and ultimately, the
connection completely died. Need to try with another (newer) driver (the
one Dell actually published for the card).


> Hmmm... I have read very varied reports on this driver - some say it is 
> fast and others report poor performance - I am guessing that it could be 
> dependent on the specific chipset but I'm not sure.

I'd lean that way too. Dunno if kernel version could be pertinent
(latest rawhide here).

> I'll try to collect 
> some info and see whether there are any patches in the wild that improve 
> performance. I am not in a position to really test the speeds since I 
> only use my wireless for broadband so the speed isn't really an issue. 
> For me the best thing about this driver is that it works flawlessly with 
> NetworkManager whereas the ndiswrapper/windows driver combo was badly 
> broken when used with WPA/WPA2 encryption and NetworkManager even though 
> it was fine when using wpa_supplicant scripts alone to connect

Huh. ndiswrapper worked just fine with WPA2 and NetworkManager for me,
both on the AppleTV and on the Mini ('tis how I connected before the
connection went belly-up).


> (FWIW I've got a 4328 chipset on x86_64).

No you don't. :)

lspci lies, someone screwed up the name for device id 4328. The 4321 and
4322 typically sit behind a bridge that has device id 4328, or something
like that, there isn't actually a BCM4328. I believe the bridge is
what's driven by the ssb driver.


-- 
Jarod Wilson
jarod at wilsonet.com