Subject: Re: [PATCH 1/5] Move standard wireless defintions out of mac80211
From: Dan Williams <dcbw@redhat.com>
To: "Luis R. Rodriguez" <mcgrof@gmail.com>
Cc: Johannes Berg <johannes@sipsolutions.net>,
	John Linville <linville@tuxdriver.com>,
	linux-wireless@vger.kernel.org,
	Michael Wu <flamingice@sourmilk.net>,
	Daniel Drake <dsd@gentoo.org>,
	Larry Finger <Larry.Finger@lwfinger.net>
In-Reply-To: <43e72e890709241049y63a88d6aoa6456d2f1465ee04@mail.gmail.com>
References: <20070921210014.GE31768@pogo>
	 <1190409462.18521.160.camel@johannes.berg>
	 <43e72e890709211433hf6669f1y97d24663a97b21ff@mail.gmail.com>
	 <20070924094959.GY10493@thot.informatik.uni-kl.de>
	 <43e72e890709241049y63a88d6aoa6456d2f1465ee04@mail.gmail.com>
Content-Type: text/plain
Date: Wed, 26 Sep 2007 11:01:02 -0400
Message-Id: <1190818863.20955.4.camel@xo-3E-67-34.localdomain>
Mime-Version: 1.0
Sender: linux-wireless-owner@vger.kernel.org

On Mon, 2007-09-24 at 13:49 -0400, Luis R. Rodriguez wrote:
> On 9/24/07, Joerg Mayer <jmayer@loplof.de> wrote:
> > On Fri, Sep 21, 2007 at 05:33:42PM -0400, Luis R. Rodriguez wrote:
> > > > > + * ieee80211_channel - internal structure definiton for an IEEE-802.11 channel
> > > > > + *
> > > > > + * This defines an ieee80211_channel. The IEEE-802.11 regulatory domain agent
> > > > > + * is in charge of filling most of these fields out. The low-level driver
> > > > > + * is expected to fill in, if needed, the val field. Note that val is already
> > > > > + * set by the regulatory agent to the same channel as in chan.
> > > >
> > > > Shouldn't you also set chan and freq in the driver?
> > [...]
> > > Initially I wanted to move mac80211 to use a linked list of channels
> > > instead of an array for each mode. As we discussed it over (you, me
> > > and a few others) we determined it wasn't best to do this (I have a
> > > patch that allows this just in case we later change our mind). But --
> > > previous to doing this I had a wrapper on ieee80211_channel as I
> > > wanted the central regdomain to have a linked list of reg domain
> > > channels. If we don't want a wrapper we have to add this to the
> > > ieee8021_channel struct. Otherwise I need to define my own channel
> > > struct which wraps around ieee80211_channel.
> > >
> > > > > +struct ieee80211_channel {
> > > > > +     /* XXX change to u8 */
> > > > > +     short chan;
> > > >
> > > > Not sure, is a u8 always enough? I thought 802.11a had huge channel
> > > > numbers?
> > >
> > > Not that huge :-) Highest I've heard is 220. u8 should suffice.
> >
> > What's the idea of working with channels? Channels are non-uniq numbers.
> > Internally everything should work with center frequency and not with
> > channels. If necessary, information should be converted from channels
> > by the driver when reporting it into the stack. That way there also won't
> > be any problems if there ever is a channel 417.
> 
> I agree. I'm OK with just having center of freq. Right now WE supports
> channel or freqs and the value is passed directly to the driver (for
> FullMAC) or mac80211 (for SoftMAC). For nl80211 we can just support
> channel passing onto the kernel, if the user supplies a channel we'll
> do the conversion to freq in userspace. On step at a time.

Definitely need to use freq and not channel.  Channels are not unique
(they just _happen_ to be so right now).  Of course remember that we
can't just pass channel really, we need to pass the tuple of (channel,
band) or maybe even (channel, [A|B|G|N]) or something.

Dan