status wireless extensions API for new drivers

status wireless extensions API for new drivers

[Bjorge Dijkstra]

Hello all,

I have been working on adding support for RNDIS based USB wireless 
devices[1]. Recently I've been getting some help from another developer 
and the state of the driver has improved quite a bit. So we are thinking
about getting it into the mainline kernel.
We have used the Wireless Extensions API, but I get the impression that
things are moving towards the cfg/nl80211 API's. Is it still acceptable
for new drivers trying to get into mainline to use WEXT, or should we
switch to the new API's?

regards,
Bjorge

[1]. http:/www.jooz.net/rndis

Re: status wireless extensions API for new drivers

[John W. Linville]

On Thu, Dec 13, 2007 at 10:15:52PM +0100, Bjorge Dijkstra wrote:

> We have used the Wireless Extensions API, but I get the impression that
> things are moving towards the cfg/nl80211 API's. Is it still acceptable
> for new drivers trying to get into mainline to use WEXT, or should we
> switch to the new API's?

WEXT is fine for new drivers.  At this time cfg80211 does not offer
a suitable replacement.

John
-- 
John W. Linville
linville@tuxdriver.com

Re: status wireless extensions API for new drivers

[Bjorge Dijkstra]

John W. Linville wrote:
> 
> WEXT is fine for new drivers.  At this time cfg80211 does not offer
> a suitable replacement.

Ok, thanks. I'll stick to WEXT for now then..

> 
> John

regards,
Bjorge

Re: status wireless extensions API for new drivers

[John W. Linville]

On Thu, Dec 13, 2007 at 11:07:20PM +0100, Bjorge Dijkstra wrote:
> John W. Linville wrote:
>>
>> WEXT is fine for new drivers.  At this time cfg80211 does not offer
>> a suitable replacement.
>
> Ok, thanks. I'll stick to WEXT for now then..

BTW, how is your patchset looking?  If you want it to be in 2.6.25
then you need to post it soon.

FWIW, as long as you have just enough WEXT support to get it to work
with open networks I think that would be enough to get a patch merged.
Having experimented a bit earlier with your patchset, I suspect you
can probably support WEP and maybe more by now...

So, when will we see your patches?

Thanks,

John
-- 
John W. Linville
linville@tuxdriver.com

Re: status wireless extensions API for new drivers

[Bjorge Dijkstra]

John W. Linville wrote:
> BTW, how is your patchset looking?  If you want it to be in 2.6.25
> then you need to post it soon.
> 

It's currently in the form of a seperate self supporting package based
on an older kernel. So that probably needs a little bit of work to get
it rebased on current kernel. It's not a must for me to get it in 2.6.25
though, so if we miss it that's not a problem.

BTW, when is the approx. deadline for 2.6.25?

> FWIW, as long as you have just enough WEXT support to get it to work
> with open networks I think that would be enough to get a patch merged.
> Having experimented a bit earlier with your patchset, I suspect you
> can probably support WEP and maybe more by now...

Yes, WEP, WPA and WPA2 all should work fine. I'm using it everyday on a
WPA2 network with no issues. I don't have a wired connection here, so I
depend on it working for me :)

> 
> So, when will we see your patches?
> 

I'll try get something out during the weekend.

> Thanks,
> 
> John

regards,
Bjorge

Re: status wireless extensions API for new drivers

[Dan Williams]

On Thu, 2007-12-13 at 22:15 +0100, Bjorge Dijkstra wrote:
> Hello all,
> 
> I have been working on adding support for RNDIS based USB wireless 

Quick question about these; does the firmware on the wireless device
itself talk RNDIS directly?  i.e., is RNDIS intended to replace all the
manufacturer-specific proprietary host<->adapter protocols?  How common
are the devices these days?  I trawled through CDW this weekend and
could only find one or two adapters (out of 30 or so) that were RNDIS.
I assume it's the wave of the Windows future though.

> devices[1]. Recently I've been getting some help from another developer 
> and the state of the driver has improved quite a bit. So we are thinking
> about getting it into the mainline kernel.
> We have used the Wireless Extensions API, but I get the impression that
> things are moving towards the cfg/nl80211 API's. Is it still acceptable
> for new drivers trying to get into mainline to use WEXT, or should we
> switch to the new API's?

Well, if you want to allow your driver to be backported at all, you'll
pretty much have to support WEXT.

_But_, since your driver doesn't have any need for backwards
compatibility requirements since it's not already upstream, you might
want to be the guinea pig for cfg80211/nl80211 :)  If you decided to
only support cfg80211, you'd get WEXT support since cfg80211 provides a
backwards compat solution for WEXT.  Plus you'd be helping out a great
deal by finding and fixing bugs in cfg80211/nl80211 and proving the
framework, which would be a great help.

Dan

Re: status wireless extensions API for new drivers

[Bjorge Dijkstra]

Dan Williams wrote:
> 
> Quick question about these; does the firmware on the wireless device
> itself talk RNDIS directly?  i.e., is RNDIS intended to replace all the
> manufacturer-specific proprietary host<->adapter protocols?

Yes, the devices themselves talk RNDIS. It's basically Windows NDIS
driver API calls over a USB transport. What I have created is not much
more than a layer that translates the WEXT calls to appropriate RNDIS
messages to allow the device to be configured. The actual networking
part is handled by rndis_host and usbnet.

> How common
> are the devices these days?  I trawled through CDW this weekend and
> could only find one or two adapters (out of 30 or so) that were RNDIS.
> I assume it's the wave of the Windows future though.
> 

I know of about 5 or 6 different devices and none of them are new, but
they seem to be fairly popular. AFAIK all are based on the same
broadcom 4320 chipset.

> 
> Well, if you want to allow your driver to be backported at all, you'll
> pretty much have to support WEXT.
> 
> _But_, since your driver doesn't have any need for backwards
> compatibility requirements since it's not already upstream, you might
> want to be the guinea pig for cfg80211/nl80211 :)  If you decided to
> only support cfg80211, you'd get WEXT support since cfg80211 provides a
> backwards compat solution for WEXT.  Plus you'd be helping out a great
> deal by finding and fixing bugs in cfg80211/nl80211 and proving the
> framework, which would be a great help.

Well, regular WEXT is already more or less done, so I'd prefer to not
throw that away :)  I was just wondering about the current state of the
wireless api's.

> 
> Dan
> 

regards,
Bjorge