[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-wlan-devel
Subject:    Re: [lwlan-devel] [PATCH] Some wireless extensions
From:       solomon () linux-wlan ! com
Date:       2004-01-06 14:47:55
Message-ID: 20040106144755.GA3658 () linux-wlan ! com
[Download RAW message or body]

[Attachment #2 (multipart/signed)]


On Tue, Jan 06, 2004 at 11:47:15AM +0100, Jan Bernhardt wrote:
> * What other modes are there? I thought I could only run in Ad-hoc (IBSS?)
>   or managed mode.

There's IBSS, BSS station and BSS access point.
 
> * What's the difference between p2 mibs and the regular ones? At least
>   for the encryption stuff, I think I have to use the p2 mibs because the
>   handlemsg function intercepts them (for host wep and internal stuff?).

Just that -- the p2 stuff is a 1:1 mapping to the prism2 firmware RIDs, 
which may or may not line up with the 802.11 MIB.  Where possible, the 
802.11 equivalents should be used.

And while the handlemsg function does intercept them, they are passed 
through to the hardware as well.
 
> What do you mean by "device-independent"? Are there plans to support other
> chipsets besides the prism2?

That was always the plan.  *grins* So much of the complexity of
linux-wlan-ng becomes irrelevant if it were to only support one chipset. 

Originally, 'linux-wlan' was written to handle the prism1/am930 chipset. 
it was a thin MAC.  'linux-wlan-ng' was written to handle the prism2
chipset, which being a thick MAC, required a completely different
internal structure.  The two driver trees were merged, and that became
the basis for our mystical internal tree, on which is what our
commercial offerings are built.  Supporting many more chipsets too, I
might add.. but thanks to many layers of NDAs, I can't talk about 'em,
much less release any of the code.

So the internal tree is radically different (and considerably more
capable!) the public trees, but still works via more or less the same
MIB interface, internally and externally.  I try to back-port all of the
bits relevant to prism2, and indeed the current wireless extension code
in linux-wlan-ng was such a backport.  

So, all that said... yes, I'd like the wireless extension code to remain 
device-independent.  Which fortunately just means picking the right 
MIBs.  *grins*  

I consider linux-wlan-ng to be a dead-end, obselete codebase.  I really 
hope I can get the new tree (as relevant to prism2/prism1) released, so 
I can shoot it in the head and be done with it.  :)  But there are many 
hurdles that need to be overcome first, mostly outside my immediate 
control.

> This was the reason I started implementing the wireless extensions. I'd
> like to see this ng drivers go more mainstream, so they integrate better
> with the current distros (in my case fedora). Though I have not tried, I
> thought rh/fedora includes support for network profiles which should be
> able to handle different authentications, wlan modes etc. I'll look into
> that, because I'm going to need this anyway.

The Fedora profile stuff amounts to a boot command line parameter, which
is still manually invoked/selected.  I just upgraded my laptop though,
and I'll be checking it out and see how it works in more detail.

> In my naive way, without having asked, I would have set the ifstate on
> driver initialization. Don't know about the onboard firmware problems.

That's my thinking too.  Except for the firmware loading problem.   
 
> Actually I'm a bit puzzled now on how to go on with the patches, so that
> you might accept them for inclusion. I think I have to learn more about
> 802.11 driver stuff. Any idea where I can find prism2 documentation these
> days? Or any other recommended documentation?

I'm under NDA, so I can't release docs.  but I'd be surprised if google 
didn't hold the answers.

Simply put, I'd like to integrate better wireless extension support.  
For the majority of linux-wlan-ng users, it's "good enough" to do what 
they need.  I'm even willing do de-emphasize the wlanctl/mib interface 
in favor of wireless extensions to achieve that end.

However, both HostAP and the Orinoco drivers support prism2 hardware and 
provide full wireless extension support, and if you don't need the full 
MIB (and/or USB) stuff, there's really little reason to use 
linux-wlan-ng. 

But that's just my opinion.  *grins*

As to your patches:

1) replace p2-specific calls with 802.11 calls.  There's usually a 1:1 
   mapping there, so it's straightforward.  
2) the SET ESSID function needs to understand the difference between 
   which mode we're in (infrastructure sta/AP/IBSS)
3) You may want to add a SET MODE hook, but that's not strictly 
   necessary for straight infrastructure support.

We're going to need to work out the kinks I mentioned earlier, but 
that's outside the realm of these patches.  

Sound good?  

I do appreciate the work you're putting into this, and I do want to see 
this stuff go in. 

 - Pizza
-- 
Solomon Peachy                        solomon@linux-wlan.com
AbsoluteValue Systems                 http://www.linux-wlan.com
715-D North Drive                     +1 (321) 259-0737  (office)
Melbourne, FL 32934                   +1 (321) 259-0286  (fax)

[Attachment #5 (application/pgp-signature)]

_______________________________________________
Linux-wlan-devel mailing list
Linux-wlan-devel@lists.linux-wlan.com
http://lists.linux-wlan.com/mailman/listinfo/linux-wlan-devel


[prev in list] [next in list] [prev in thread] [next in thread]