Subject: Re: [PATCH] libertas: remove user-specified channel list
From: Dan Williams <dcbw@redhat.com>
To: Holger Schurig <hs4233@mail.mn-solutions.de>
Cc: Albert Cahalan <acahalan@gmail.com>,
	linux-wireless@vger.kernel.org, libertas-dev@lists.infradead.org,
	"John W. Linville" <linville@tuxdriver.com>
In-Reply-To: <200711290844.10501.hs4233@mail.mn-solutions.de>
References: <200711281730.21327.hs4233@mail.mn-solutions.de>
	 <787b0d920711282012k38d6d2c8g2d9ae79c89b04059@mail.gmail.com>
	 <200711290844.10501.hs4233@mail.mn-solutions.de>
Content-Type: text/plain
Date: Thu, 29 Nov 2007 14:20:36 -0500
Message-Id: <1196364036.11556.21.camel@localhost.localdomain> (sfid-20071129_192852_725486_3A94F587)
Mime-Version: 1.0
Sender: linux-wireless-owner@vger.kernel.org

On Thu, 2007-11-29 at 08:44 +0100, Holger Schurig wrote:
> On Thursday 29 November 2007 05:12:06 Albert Cahalan wrote:
> > On Nov 28, 2007 11:30 AM, Holger Schurig 
> <hs4233@mail.mn-solutions.de> wrote:
> > > Remove the ability to specify channels to scan via debugfs
> >
> > Why are you ripping out functionality? At the very least,
> > that ought to be justified in the changelog comment.
> 
> This was discussed in libertas-dev.
> 
> 
> Basically, I posted there a RFC (request-for-comment) patch about 
> a newer implementation of how scanning could be implemented. The 
> current version of scan.c is a bit complicated and not easy to 
> follow. I was rewriting that up to a point where it worked. But 
> my RFC patch had a bunch of TODO items, and I asked in 
> libertas-dev how to deal with them.
> 
> Two TODOs were about the implementation the syntax 
> of "probes=XXX" and "channel=XXX" that you can use via 
> debugfs/extgetscan.
> 
> Dan thought that we could easily rip this out, because no one is 
> known to use this feature. And I'm seconding Dan here. Both 
> debugfs-features actually just exhibits a possibility of how to 
> interact with the firmware. They don't have a real value.
> 
> Simply because you cannot set number of probes or channel-list 
> from "normal" user-space application. E.g. when libertas needs 
> to scan because of a new ESSID set via "iwlist eth1 essid BLAH" 
> it would anyway scan all channels set by the country (11, 13 or 
> 14 for b/g). It wouldn't honor your manually crafted last scan 
> from your last access to debugfs.
> 
> So extgetscan is merely a tool to play with, but nothing more.
> 
> madwifi for example also has the ability to specify "please scan 
> only on channels 1, 7, 11", but this has been implemented via 
> private ioctl's and set's this "sticky". So when you scan from 
> iwlist, implicitly from iwconfig XXX essid, from wpa_supplicant 
> etc, it would be honored. For such a feature you have a 
> real-life usage scenario, it would be much more useful than the 
> current way.

Note that WEXT supports specifying the channels you'd like to scan on in
the extscan struct of the SIOCSIWSCAN call.  Libertas should support
this.  However, I don't think iwlist supports it, but that shouldn't
stop the drivers from supporting that option.

Dan