[Prism54-devel] USB advances

[Luis R. Rodriguez]

Jean-B,

On Tue, Aug 31, 2004 at 08:34:30AM +0200, Jean-Baptiste Note wrote:
> Hello,
> 
> I'm sending this mail to inform you of what's happening USB front, as
> far as i'm concerned. If anybody else is working on this, then please
> reply !

I'm glad someone is taking the initiative.

> First, from all USB devices that i could gather logs of (available on
> http://jbnote.free.fr/prism54usb ), i'm unhappy to announce that -all-
> of them use the are softmac-based. This leads me to think that there is
> no other way to manage USB prism54 devices, but i'd be happy to be
> proven wrong...

My understanding was that there were at least a few USB devices out
there with a Prism Frisbee chipset. I'm pretty sure the Linksys WUSB54G
has one. Prism Frisbee chipset-based cards work with our driver -- most
Cardbus cards are Prism Frisbee based. This is in essence why we started
prism54-usb CVS tree.

I believe though that most USB devices out there are *do not* have the
Prism Frisbee chipset and rather an Xbow/Javelin chipset (definitely
Soft MAC). To support all these we need a new driver.

> Yet, I've managed to configure the device so that 
> - you can tune its frequency
> - it receives the broadcast packets and packets directed to it (read
> beacons from the AP). You can see them for yourself by tuning the device
> to the channel of your AP. Current code only accepts 
> iwconfig ethX channel Y
> The setting wont be reported in iwconfig listings, but the change and
> received packets are printk'd, and will be available in you system logs.
> 
> The tarball can be found here
> http://jbnote.free.fr/prism54usb/prism54-usb-channel.tar.bz2
> 
> The windows logs indicate that all other actions must be taken care of
> in software (association management frames are clearly sent on the
> wire).
> 
> I now face two problems, for which your advice would be much
> appreciated:
> - for now i've done things privately, building upon what is found in
> CVS. While my code is horrible and disorganized, maybe i should put some
> of the changes i made back into CVS. 

Feyd, you have been doing most work on prism54-usb. What do you say? If
you do not reply I say we just give Jean-B CVS write access as he has
working hard on this.

> I guess that means some stability
> about code organization is required, so i badly need someone to guide me
> and tell me how to organize things... Should I submit patches here for
> review ?

OK, first thing we have to do is get a listing of all USB Devices out
there and see which ones have the Prism Frisbee. We can only work with
those with the prism54-usb code. 

If there's only a few devices out there we might as well just wait and
start work on the prism54 softmac driver. Conexant was actively working
with us to provide us specs and a new source base of the softmac driver
but I recently asked for an update and haven't received a reply. We have
also not heard from them for about 1-2 months.

In the worst case scenario they'll never get back to us and all we'd
have available is this super buggy beta release of the softmac driver to
work on:

http://www.dse.co.nz/isroot/dse/support/XH8196-linux.zip

If you're up for the challenge and don't want to wait until we hear back
from Conexant you can just start working on cleaning this driver up. The
first thing you'd have to do is get rid of all that pcmcia-cs dependency
just as we did with prism54 back last year. I'd work on this but:

1. My laptop died on me and I currently have no test box
2. I'm prioritizing finishing work on WPA right now
3. Prism54 has tons of bugs open, I should go over those first

> - i now need to interface to a 802.11 stack... I've looked at the
> available code, and the simplest and clearest i've found is in the
> ipw2100 (centrino) driver. There were talks, however, of an "official"
> 802.11 stack. Have you got any information on this ?

As per Jeff Garzik -- netdev kernel mantainer -- the kernel "plan" is to 
start work on a wireless-2.6 dev tree which will include wireless code
other drivers can use. I'm only aware of ipw2100 using some hostap code
and plans on including that code on that tree so far. I'm not sure of
the details of what code is beign re-used there. 

A softmac prism54 driver should definitely reuse a lot of code 
and we may actually end up contributing a lot to this tree if other 
softmac cards will exist.

So regardless I guess we should start poking at this XH8196-linux.zip
so we can familiarize ourselves with it. If you want you just start
working on the usb softmac driver. If we get new code its just a matter
of adapting your changes to the new source base.

	Luis

-- 
GnuPG Key fingerprint = 113F B290 C6D2 0251 4D84  A34A 6ADD 4937 E20A 525E