Re: Insist on cfg80211 for new drivers?

[Dan Williams <dcbw@redhat.com>]

On Thu, 2009-07-02 at 13:07 +0200, Bartlomiej Zolnierkiewicz wrote:
> Hi,
> 
> On Thursday 02 July 2009 00:18:49 Marcel Holtmann wrote:
> > Hi Greg,
> > 
> > > > > >> > That's really all I
> > > > > >> > care about, I don't want another WEXT-based driver accepted; I want all
> > > > > >> > the new ones using cfg80211.
> > > > > >>
> > > > > >> Now there is a discussion we should have had in Berlin...is it time
> > > > > >> to insist on cfg80211-based configuration for all new drivers?
> > > > > >
> > > > > > there is really nothing much to discuss on this topic. The plan is to
> > > > > > deprecate WEXT, that simple. So if the driver has no cfg80211 support,
> > > > > > then it will not be included. Period.
> > > > > >
> > > > > > Send such drivers off to staging and let them have their 6 month.  Then
> > > > > > we either remove them again or they got ported to cfg80211.
> > > > > 
> > > > > 6 months only in staging ? Is this a rule now for staging?
> > > > 
> > > > that is what Greg mentioned to me. If there is no activity for 6 month
> > > > and the driver is not getting anywhere, he going to drop it.
> > > 
> > > That is "within reason".  If a driver is still needed there, I'l
> > > probably keep it, and will take each one on a case-by-case basis.
> > 
> > what do you mean "within reason". If the driver is just sitting there
> > and no effort in making in upstream ready it is doing clearly more harm
> > than any good. And I am not talking about removing some kernel version
> > details or typedefs or coding style. Drivers with missing cfg80211 need
> 
> Well, most of those drivers need a major cleanup before porting..	
> 
> > active porting. And it is not that hard. See the orinoco one for an
> > example.
> 
> Could you please provide some more pointers here, I don't see any such
> changes in linux-next yet and I'm very interested in seeing a practical
> example of such conversion (thanks!).

http://osdir.com/ml/linux-wireless/2009-06/msg01141.html

It's not in -next, it's in wireless-testing, which gets dumped to next
periodically.  Wireless stuff happens first in wireless-testing, of
course.

> > If we see developers committed to fixing it that is a different story,
> > but a lot of drivers in staging are getting no attention all. So they
> > are fully pointless and are not doing any good for Linux. We need to
> > send the vendors a clear message that code drops of their crappy Windows
> > code are not desired.
> 
> How's about sending a clear _positive_ message for a change?
> 
> Where one can find an up to date documentation for {mac,cfg,nl}80211 (not
> just some random DocBook generated excerpts, I mean the real thing here,
> with references to kernel versions when API changes were introduced, some
> practical examples and exemplary drivers) and more importantly when one
> can find the _porting_ guide for the older stacks?

Johannes has done an much-better-than-average job of documenting
cfg80211 and nl80211, which is more than I can say for many other
subsystems.

Dan