Re: [ath9k-devel] Setting 11n MCS TX bitrate

Re: [ath9k-devel] Setting 11n MCS TX bitrate

[Luis R. Rodriguez]

On Wed, Feb 04, 2009 at 02:08:02PM -0800, Kishore Ramachandran wrote:
> Hi,

Hey Kishore!

> I would like to configure D-Link DWA 645 (AR5416) PCMCIA card so that it uses a fixed 11n bit-rate. Based on earlier posts, I understand that iwconfig or wireless tools will only allow me to set 11a/b/g bit-rates.

This is correct.

> There does not seem to be a "set rate" option with iw yet. Is this in the works?

Not yet, but please feel free to give a shot at hacking it up. It would require
a new cfg80211 command but keep in mind the way I believe we want this is to
use the 802.11 MLME SAP interface for this.

Right now ath9k also does not report the right MCS index through iw
though. It seems this is due to the way our RC works... anyway this
could use some love too. Sujith may be working on this, not sure,
sujith?

In the end it would be nice to see minstrel be used anyway so a good
milestone is to allow ath9k select minstrel or let it use its own RC.
By then things will start looking better I think.

> I am using ath9k (compat-wireless-2009-01-19) on Ubuntu (kernel v2.6.27-11) with iw v0.9.9.

Nice, just one tip: may want to stick to using iw from git if you are doing
development, it keeps moving on and on.

  Luis

Re: [ath9k-devel] Setting 11n MCS TX bitrate

[Luis R. Rodriguez]

On Wed, Feb 04, 2009 at 02:51:13PM -0800, Kishore Ramachandran wrote:
> "Not yet, but please feel free to give a shot at hacking it up. It would require
> a new cfg80211 command but keep in mind the way I believe we want this is to
> use the 802.11 MLME SAP interface for this."
> 
> I thought as much to start looking at the iw source but thought would check with you guys first. Would the implementation be similar to the "set channel" and/or the "set frequency" commands? Is there is another function using the 802.11 MLME SAP interface that I can use as a guideline?

Well in general I was talking about IEEE-802.11 section 10.3 stuff.
In cfg80211/nl80211 we don't want to allow things like "set essid"
as iwconfig has, instead we want to break things down a bit more,
which can really help debugging. The MLME SAP already has some
decent layout of commands/responses but after a quick glance
I do not see anything for setting a static rate. May want to review
that a bit before implementing set rate though. Just remember
post early and review early. If you have an idea of how you want
to do it let us know to get feedback.

  Luis

Re: [ath9k-devel] Setting 11n MCS TX bitrate

[Jouni Malinen]

On Wed, Feb 04, 2009 at 03:18:40PM -0800, Luis R. Rodriguez wrote:
> Well in general I was talking about IEEE-802.11 section 10.3 stuff.
> In cfg80211/nl80211 we don't want to allow things like "set essid"
> as iwconfig has, instead we want to break things down a bit more,
> which can really help debugging. The MLME SAP already has some
> decent layout of commands/responses but after a quick glance
> I do not see anything for setting a static rate. May want to review
> that a bit before implementing set rate though. Just remember
> post early and review early. If you have an idea of how you want
> to do it let us know to get feedback.

TX rate control is outside the scope of IEEE Std 802.11-2007 (see 9.6
Multirate support) and as such, there is no MLME SAP (or any other)
primitives for this type of functionality defined in the standard. The
parameters are passed into PHY through the PHY-SAP interface in TXVECTOR
(e.g., DATARATE and MCS). How the MAC sublayer decides how to set these
parameters or even things like how to control the decision process are
very much out of scope.
 
-- 
Jouni Malinen                                            PGP id EFC895FA

Re: [ath9k-devel] Setting 11n MCS TX bitrate

[Sujith]

Luis R. Rodriguez wrote:
> 
> Not yet, but please feel free to give a shot at hacking it up. It would require
> a new cfg80211 command but keep in mind the way I believe we want this is to
> use the 802.11 MLME SAP interface for this.
> 
> Right now ath9k also does not report the right MCS index through iw
> though. It seems this is due to the way our RC works... anyway this
> could use some love too. Sujith may be working on this, not sure,
> sujith?
> 
> In the end it would be nice to see minstrel be used anyway so a good
> milestone is to allow ath9k select minstrel or let it use its own RC.
> By then things will start looking better I think.
> 

Yeah, fixed rate for 11n won't work with ath9k right now.
And I'll take a look and see if we can have atleast a MCS to internal index
conversion hack - until the tables are shrunk properly.

But for now, you can do something like 'cat /debug/ath9k/*/rcstat' to see
what rates are being used.

Sujith
-- 
http://sujith-m.blogspot.com