"Stable" branch of mac80211

"Stable" branch of mac80211

[Jiri Benc]

Hi,

I've created a new 'stable' branch in my tree
(git://git.kernel.org/pub/scm/linux/kernel/git/jbenc/mac80211.git). It
contains patches that strip mac80211 down to a stable version. The purpose
of this branch is to prepare mac80211 for vanilla inclusion.

Please note that this branch is expected to be rebased frequently and all
patches are subject to change.

Currently, it contains following patches:

      mac80211: remove 802.11n parts
      mac80211: remove WPA testing stuff
      mac80211: remove hostapd ioctls
      mac80211: remove nonstable iwpriv ioctls
      mac80211: remove iwpriv ioctls with wext equivalent
      mac80211: remove AP and WDS support


 Jiri

-- 
Jiri Benc
SUSE Labs

Re: "Stable" branch of mac80211

[Michael Wu]

[-- Attachment #1.1: Type: text/plain, Size: 679 bytes --]

On Sunday 29 April 2007 13:48, Jiri Benc wrote:
> Please note that this branch is expected to be rebased frequently and all
> patches are subject to change.
>
wireless-dev has been updated recently. One of the conflicts can be resolved 
easily. There's also an API change which conflicts with the qdisc fixes 
because tcf_destroy is no longer exported. A patch for that is attached.

Once updated to the latest wireless-dev, it should be fairly easy to pull 
together a patch for net-2.6 since wireless-dev should have all the major 
networking API changes for 2.6.22. I think mac80211 can be pushed to net-2.6 
after that. What do you think?

Thanks,
-Michael Wu

[-- Attachment #1.2: 19-fix-tcf-api-update.diff --]
[-- Type: text/x-diff, Size: 992 bytes --]

mac80211: update qdisc code to use tcf_destroy_chain

From: Michael Wu <flamingice@sourmilk.net>

tcf_destroy is no longer exported. Use tcf_destroy_chain instead.

Signed-off-by: Michael Wu <flamingice@sourmilk.net>
---

 net/mac80211/wme.c |    7 ++-----
 1 files changed, 2 insertions(+), 5 deletions(-)

diff --git a/net/mac80211/wme.c b/net/mac80211/wme.c
index 1cb6358..9874d9a 100644
--- a/net/mac80211/wme.c
+++ b/net/mac80211/wme.c
@@ -354,13 +354,10 @@ static void wme_qdiscop_destroy(struct Qdisc* qd)
 	struct ieee80211_sched_data *q = qdisc_priv(qd);
 	struct ieee80211_local *local = wdev_priv(qd->dev->ieee80211_ptr);
 	struct ieee80211_hw *hw = &local->hw;
-	struct tcf_proto *tp;
 	int queue;
 
-	while ((tp = q->filter_list) != NULL) {
-		q->filter_list = tp->next;
-		tcf_destroy(tp);
-	}
+	tcf_destroy_chain(q->filter_list);
+	q->filter_list = NULL;
 
 	for (queue=0; queue < hw->queues; queue++) {
 		skb_queue_purge(&q->requeued[queue]);

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: "Stable" branch of mac80211

[Jiri Benc]

On Sun, 29 Apr 2007 14:14:14 -0400, Michael Wu wrote:
> wireless-dev has been updated recently. One of the conflicts can be resolved 
> easily. There's also an API change which conflicts with the qdisc fixes 
> because tcf_destroy is no longer exported. A patch for that is attached.

I folded the patch into the qdisc fix patch and rebased my tree (all
branches).

I also ensured the previously misapplied radiotap patch is now applied
correctly and dropped the "move radiotap removal code to right function"
patch.

Thanks,

 Jiri

-- 
Jiri Benc
SUSE Labs

Re: "Stable" branch of mac80211

[James Ketrenos]

Jiri Benc wrote:
> Hi,
> 
> I've created a new 'stable' branch in my tree
> (git://git.kernel.org/pub/scm/linux/kernel/git/jbenc/mac80211.git). It
> contains patches that strip mac80211 down to a stable version. The purpose
> of this branch is to prepare mac80211 for vanilla inclusion.
...
> Currently, it contains following patches:
> 
>       mac80211: remove 802.11n parts
>       mac80211: remove WPA testing stuff
>       mac80211: remove hostapd ioctls
>       mac80211: remove nonstable iwpriv ioctls
>       mac80211: remove iwpriv ioctls with wext equivalent
>       mac80211: remove AP and WDS support

I'm curious what the criteria was for removing code from 'stable'  

Are there known bugs/oops?  Is it known to be broken, or ?  

For code we know we ultimately want to be redone, removed, or whatever, I can understand not wanting to merge the code.  However we ultimately want AP, WDS, and 802.11n.  At least with 802.11n, I haven't seen anything indicating the code needs to be done any differently than it is currently being done.  

I assume the entire new tree will be marked as 'experimental', so what is the advantage gained by pulling this code out now?

James

Re: "Stable" branch of mac80211

[Jiri Benc]

On Mon, 30 Apr 2007 09:52:51 -0700, James Ketrenos wrote:
> I'm curious what the criteria was for removing code from 'stable'  
> 
> Are there known bugs/oops?  Is it known to be broken, or ?  
> 
> For code we know we ultimately want to be redone, removed, or whatever,
> I can understand not wanting to merge the code.

AP and user space MLME mode don't have clarified user space interface.
We don't want to support the current mess (mgmt interface) forever. WDS
is tightly linked with AP.

802.11n code is not deeply reviewed by anyone and thus it's unclear
whether it will need to be changed (I already heard some objections) or
not.

> However we ultimately want AP, WDS, and 802.11n.

We ultimately want STA mode and - maybe - IBSS. Other things are just
nice to have. They will still stay in wireless-dev, of course.

> At least with 802.11n, I haven't seen anything indicating the code
> needs to be done any differently than it is currently being done.  

Nobody uses it. I don't see a reason why it needs to go in now. I
really want those patches to be reviewed by someone before considering
them "stable". It can be me, but I need some time to study .11n draft,
or someone else experienced with different hw designs.

> I assume the entire new tree will be marked as 'experimental', so what
> is the advantage gained by pulling this code out now?

Once the code is in vanilla, user space interfaces are set in stone.
Regarding .11n, untested and unreviewed patches shouldn't go to vanilla.

Thanks,

 Jiri

-- 
Jiri Benc
SUSE Labs

Re: "Stable" branch of mac80211

[John W. Linville]

On Mon, Apr 30, 2007 at 07:29:51PM +0200, Jiri Benc wrote:
> On Mon, 30 Apr 2007 09:52:51 -0700, James Ketrenos wrote:

> > I assume the entire new tree will be marked as 'experimental', so what
> > is the advantage gained by pulling this code out now?
> 
> Once the code is in vanilla, user space interfaces are set in stone.
> Regarding .11n, untested and unreviewed patches shouldn't go to vanilla.

ACK
-- 
John W. Linville
linville@tuxdriver.com

Re: "Stable" branch of mac80211

[Jiri Benc]

On Sun, 29 Apr 2007 19:48:55 +0200, Jiri Benc wrote:
> I've created a new 'stable' branch in my tree
> (git://git.kernel.org/pub/scm/linux/kernel/git/jbenc/mac80211.git). It
> contains patches that strip mac80211 down to a stable version. The purpose
> of this branch is to prepare mac80211 for vanilla inclusion.

And preliminary patches targeted for net-2.6 are in
http://kernel.org/pub/linux/kernel/people/jbenc/mac80211/

Please review and test (for testing purposes, a huge patch with all
mac80211 drivers currently in wireless-dev can be found at
http://kernel.org/pub/linux/kernel/people/jbenc/mac80211-test/drivers.patch).

Thanks,

 Jiri

-- 
Jiri Benc
SUSE Labs