ar9170 vs otus advice?

ar9170 vs otus advice?

[Jon Loeliger]

> > mac80211 drivers:
> > 
> > agnx	-> http://git.sipsolutions.net/agnx.git
> > at76_usb	-> replacement already in kernel.org code at drivers/net/wireless/at76c50x-usb.[ch]
> > otus	-> replacement already in kernel.org code at drivers/net/wireless/ar9170/
> 
> Thanks for the summary.
> 
> Note, whenever a "real" wireless driver goes into the tree that
> obsoletes any of the staging drivers, please let me know and I'll
> instantly remove it.
> 
> Right now the otus code is staying because the ar9170 developers want it
> to.


So, I'm trying to understand the current state of the ar9170 and otus
drivers so that I may know where to direction some effort.  I'd like to
get an WN111v2 running with 11n support, and am willing to start with
either the ar9170 or otus drivers as needed.  But I've not been following
along in too much detail so far.

Advice on where I should start, what's working, what's not working,
what needs help, etc?  And should I start with one of the linville Git
trees?  Or wireless-everything?

Thanks,
jdl

Re: ar9170 vs otus advice?

[Christian Lamparter]

On Tuesday 02 June 2009 00:07:00 Jon Loeliger wrote:
> > > mac80211 drivers:
> > > 
> > > agnx	-> http://git.sipsolutions.net/agnx.git
> > > at76_usb	-> replacement already in kernel.org code at drivers/net/wireless/at76c50x-usb.[ch]
> > > otus	-> replacement already in kernel.org code at drivers/net/wireless/ar9170/
> > 
> > Thanks for the summary.
> > 
> > Note, whenever a "real" wireless driver goes into the tree that
> > obsoletes any of the staging drivers, please let me know and I'll
> > instantly remove it.
> > 
> > Right now the otus code is staying because the ar9170 developers want it
> > to.
> 
> 
> So, I'm trying to understand the current state of the ar9170 and otus
> drivers so that I may know where to direction some effort.  I'd like to
> get an WN111v2 running with 11n support, and am willing to start with
> either the ar9170 or otus drivers as needed.  But I've not been following
> along in too much detail so far.

well, otus is a dead end. However it does fully support all 11n features the
device has to offer. But you'll need a special and very old wpa_supplicant.
(see otus' driver page on wireless.kernel.org where to get it.)

ar9170 - on the other hand -  is only capable of receiving 11n frames.
( that said, enabling it makes the device unstable and therefore one can
  only _sniff_ 11n frames coming from other stations for now. )

> Advice on where I should start, what's working, what's not working,
> what needs help, etc?
a lot! first and foremost needed is the _final_ 802.11n specification,
which sadly will still take a while... (2010? I think at least...)
it's a really mess now... As other devices/firmwares are usually designed
after a different revision of the drafts which makes it complicated.
This is no joke and the original otus incorporates "workarounds" for all
11n vendors (including Atheros itself) to get it going.
( else, the performance can be worse than 802.11b )

as for the mac80211 stack... I'm still waiting for minstrel
(rate algorithm) patches for HT MCS rate indexes.

On the ar9170 driver site: there is a big patch
floating around, however its hit or miss.

> And should I start with one of the linville Git trees? Or wireless-everything?

linville's wireless-testing is where we put the most recent code.

Regards,
	Chr

Re: ar9170 vs otus advice?

[Jon Loeliger]

On Wed, 2009-06-03 at 03:37 +0200, Christian Lamparter wrote:

> > 
> > So, I'm trying to understand the current state of the ar9170 and otus
> > drivers so that I may know where to direction some effort.  I'd like to
> > get an WN111v2 running with 11n support, and am willing to start with
> > either the ar9170 or otus drivers as needed.  But I've not been following
> > along in too much detail so far.
> 
> well, otus is a dead end. However it does fully support all 11n features the
> device has to offer. But you'll need a special and very old wpa_supplicant.
> (see otus' driver page on wireless.kernel.org where to get it.)

OK, got it, thanks.

> ar9170 - on the other hand -  is only capable of receiving 11n frames.
> ( that said, enabling it makes the device unstable and therefore one can
>   only _sniff_ 11n frames coming from other stations for now. )

Ewww....  So, does the ar9170 fully handle b/g as a station at least?

> > Advice on where I should start, what's working, what's not working,
> > what needs help, etc?
> a lot! first and foremost needed is the _final_ 802.11n specification,
> which sadly will still take a while... (2010? I think at least...)
> it's a really mess now...

Heh.  But I'm not going to be able to help that much... :-)

>  As other devices/firmwares are usually designed
> after a different revision of the drafts which makes it complicated.
> This is no joke and the original otus incorporates "workarounds" for all
> 11n vendors (including Atheros itself) to get it going.
> ( else, the performance can be worse than 802.11b )

Is the in-tree Otus up-to-date with all of these workarounds?


> > And should I start with one of the linville Git trees? Or wireless-everything?
> 
> linville's wireless-testing is where we put the most recent code.

Excellent.

> Regards,
> 	Chr

Thanks,
jdl

Re: ar9170 vs otus advice?

[Christian Lamparter]

On Wednesday 03 June 2009 17:59:59 Jon Loeliger wrote:
> On Wed, 2009-06-03 at 03:37 +0200, Christian Lamparter wrote:
> > ar9170 - on the other hand -  is only capable of receiving 11n frames.
> > ( that said, enabling it makes the device unstable and therefore one can
> >   only _sniff_ 11n frames coming from other stations for now. )
> 
> Ewww....  So, does the ar9170 fully handle b/g as a station at least?
I'm not aware of any problems with/as a 802.11a/b/g station,
no one has reported any so far: so a/b/g works as _advertised_.
(at least for me and other developers)

> > > Advice on where I should start, what's working, what's not working,
> > > what needs help, etc?
> > a lot! first and foremost needed is the _final_ 802.11n specification,
> > which sadly will still take a while... (2010? I think at least...)
> > it's a really mess now...
> 
> Heh.  But I'm not going to be able to help that much... :-)
>
> >  As other devices/firmwares are usually designed
> > after a different revision of the drafts which makes it complicated.
> > This is no joke and the original otus incorporates "workarounds" for all
> > 11n vendors (including Atheros itself) to get it going.
> > ( else, the performance can be worse than 802.11b )
> 
> Is the in-tree Otus up-to-date with all of these workarounds?
I'm not really sure. the latest (and last) otus driver can be found here:
http://www.kernel.org/pub/linux/kernel/people/mcgrof/otus/

That said, I could get a decent HT20 (up to 70Mbits tx & 90 Mbit rx)
performance with the in-kernel driver... Does it work for you as well,
or are you asking, because it doesn't?

Regards,
	Chr

Re: ar9170 vs otus advice?

[Jon Loeliger]

On Wed, 2009-06-03 at 20:32 +0200, Christian Lamparter wrote:

> I'm not aware of any problems with/as a 802.11a/b/g station,
> no one has reported any so far: so a/b/g works as _advertised_.
> (at least for me and other developers)

OK, good!

> > Is the in-tree Otus up-to-date with all of these workarounds?
> I'm not really sure. the latest (and last) otus driver can be found here:
> http://www.kernel.org/pub/linux/kernel/people/mcgrof/otus/
> 
> That said, I could get a decent HT20 (up to 70Mbits tx & 90 Mbit rx)
> performance with the in-kernel driver... Does it work for you as well,
> or are you asking, because it doesn't?

I'm just working on setting expectations (mine) so far,
and getting a handle on what is presumed working so that
I may choose the correct starting point for some work I
would like to do.  I've not tested it yet.

> Regards,
> 	Chr

Thanks,
jdl