d80211 merge plans

d80211 merge plans

[Michael Buesch]

Hi,

I am wondering about d80211 mainline merge plans.

Short:
I would like to see a mainline merge of d80211 happening
as soon as possible (next feature merge window).

Long:
I, as one of the main maintainers of the bcm43xx driver, am
maintaining two different branches of the driver at the moment.
It's getting harder and harder, but it's still ok.
But: We are so --><-- close before another branch of the project
must happen. That would mean we would have to maintain four
different branches. I am certainly not going to do this.

One might wonder why we require yet another branch of the driver.
It's simple. Broadcom effectively branched their driver, too. The
new 4.x binary driver version has _many_ differences and does not
even support older chipsets. That makes it necessary for our open
source driver to branch, too. At least for several months. We don't
know, if we are going to branch forever, or if we are going to merge
them after some months when it's stabilized, yet.

To heavily reduce maintainance burden I would like to see d80211
going mainline as soon as possible, so that we can feature-freeze
the softmac port of bcm43xx and reduce maintainance to next to zero
there. So if people want freaky features, we simply point them
to the d80211 port, which would coexist in mainline.

d80211 in mainline would clearly marked as experimental, for sure.
Notes must be added for possibly breaking usermode interfaces.

-- 
Greetings Michael.

Re: d80211 merge plans

[Jeff Garzik]

Michael Buesch wrote:
> To heavily reduce maintainance burden I would like to see d80211
> going mainline as soon as possible, so that we can feature-freeze
> the softmac port of bcm43xx and reduce maintainance to next to zero
> there. So if people want freaky features, we simply point them
> to the d80211 port, which would coexist in mainline.
> 
> d80211 in mainline would clearly marked as experimental, for sure.
> Notes must be added for possibly breaking usermode interfaces.


Does it have active maintainers?

The DeviceScape people seem to have disappeared for months, except for a 
rare appearance recently by Jouni.

Did anyone ever come up with a list of why d80211 is superior?
What are the remaining d80211 issues?

	Jeff

Re: d80211 merge plans

[David Miller]

From: Michael Buesch <mb@bu3sch.de>
Date: Wed, 9 Aug 2006 07:03:41 +0200

> I am wondering about d80211 mainline merge plans.

I think this is a non-started until the SMP problems are worked
out.  Is it still SMP challenged?

Re: d80211 merge plans

[Ivo Van Doorn]

> > I am wondering about d80211 mainline merge plans.
>
> I think this is a non-started until the SMP problems are worked
> out.  Is it still SMP challenged?

Not sure what the SMP problems actually were,
but in the test results of rt2x00 I haven't come across
a SMP related problem yet. It could be that other non-SMP problems
are still hiding the SMP related problems, but I have not
yet found real SMP problems yet.

Re: d80211 merge plans

[Christoph Hellwig]

On Wed, Aug 09, 2006 at 07:03:41AM +0200, Michael Buesch wrote:
> Hi,
> 
> I am wondering about d80211 mainline merge plans.

http://marc.theaimsgroup.com/?l=linux-netdev&m=115477844206917&w=2
has the highlevel merge requirement.  Only mentioned there as a side-note
we'll of course need a full codereview like the one Jeff is doing now.

Re: d80211 merge plans

[Dan Williams]

On Wed, 2006-08-09 at 01:40 -0400, Jeff Garzik wrote:
> Michael Buesch wrote:
> > To heavily reduce maintainance burden I would like to see d80211
> > going mainline as soon as possible, so that we can feature-freeze
> > the softmac port of bcm43xx and reduce maintainance to next to zero
> > there. So if people want freaky features, we simply point them
> > to the d80211 port, which would coexist in mainline.
> > 
> > d80211 in mainline would clearly marked as experimental, for sure.
> > Notes must be added for possibly breaking usermode interfaces.
> 
> 
> Does it have active maintainers?
> 
> The DeviceScape people seem to have disappeared for months, except for a 
> rare appearance recently by Jouni.
> 
> Did anyone ever come up with a list of why d80211 is superior?

For one, it's got a _much_ more flexible device model than the current
hostap/ipw-based stack, more like the net80211 stack for
Atheros/madwifi.  I think that's one of the main benefits.  i.e., you
can have the d80211 cards be any combination of [AP | client | wds? ] at
any given time.  Think of a repeater node, which is both an AP and
client at the same time, all with one card.  That's one of the things a
lot of people like about net80211.

Secondly, DS is also working hard on moving a lot of the higher-level
processing to user space.  Management frames and other more complex
operations will eventually be pushed up to a fused wpa_supplicant
+hostapd running in user space, which should (a) reduce the amount of
kernel driver code, (b) increase flexibility of configuration, and (c)
make any kernel API/ABI/etc much smaller since you don't have to push so
much configuration down into the kernel itself.

That's a start (please correct me if I'm wrong on any of this).

Dan

> What are the remaining d80211 issues?
> 
> 	Jeff
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: d80211 merge plans

[Mohamed Abbas]

David Miller wrote:

>From: Michael Buesch <mb@bu3sch.de>
>Date: Wed, 9 Aug 2006 07:03:41 +0200
>
>  
>
>>I am wondering about d80211 mainline merge plans.
>>    
>>
>
>I think this is a non-started until the SMP problems are worked
>out.  Is it still SMP challenged?
>-
>To unsubscribe from this list: send the line "unsubscribe netdev" in
>the body of a message to majordomo@vger.kernel.org
>More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>  
>
I been using d80211 stack for about a month I have not encounter any SMP 
issues. We are currently involving validation engineers to do more 
stress tests and will see if any SMP issues come up.
Mohamed

Re: d80211 merge plans

[Jouni Malinen]

On Wed, Aug 09, 2006 at 01:40:27AM -0400, Jeff Garzik wrote:
> Michael Buesch wrote:
> >To heavily reduce maintainance burden I would like to see d80211
> >going mainline as soon as possible

> Does it have active maintainers?
> 
> The DeviceScape people seem to have disappeared for months, except for a 
> rare appearance recently by Jouni.

Unfortunately, I have been tied up in another project that required way
too much of my time last couple of months. Getting this code into
suitable state for merging it into mainline and then making sure it
remains in working condition is my highest priority and we have
freed up more of my time for this.

We are planning on using this tree in our future products and want to be
able to remain as close to the version that is in the mainline tree as
possible, so it is indeed in our interest to make sure that this gets
maintained. We have a group of people working on this and merging rest
of our AP product to be based on this version. I'll try to be more
actively involved in getting the changes submitted to wireless-dev (and
hopefully, into mainline once the code gets there).

-- 
Jouni Malinen                                            PGP id EFC895FA

Re: d80211 merge plans

[Jeff Garzik]

Mohamed Abbas wrote:
> David Miller wrote:
>> I think this is a non-started until the SMP problems are worked
>> out.  Is it still SMP challenged?

> I been using d80211 stack for about a month I have not encounter any SMP 
> issues. We are currently involving validation engineers to do more 
> stress tests and will see if any SMP issues come up.


Well, tests are interesting, but I would rather see a real _analysis_ of 
the locking.

Locking is provable, you know...

	Jeff