[PATCH v2.6.36-rc7] mac80211: cancel restart_work explicitly instead of depending on flush_scheduled_work()

[PATCH v2.6.36-rc7] mac80211: cancel restart_work explicitly instead of depending on flush_scheduled_work()

[Tejun Heo]

iee80211_hw->restart_work is the only work which uses the system
workqueue.  Instead of calling flush_scheduled_work() during
iee80211_exit(), cancel the work during unregistration.

This is to prepare for the deprecation and removal of
flush_scheduled_work().

Signed-off-by: Tejun Heo <tj@kernel.org>
---
 net/mac80211/main.c |    9 ++-------
 1 file changed, 2 insertions(+), 7 deletions(-)

Index: work/net/mac80211/main.c
===================================================================
--- work.orig/net/mac80211/main.c
+++ work/net/mac80211/main.c
@@ -606,7 +606,7 @@ int ieee80211_register_hw(struct ieee802
 		hw->queues = IEEE80211_MAX_QUEUES;

 	local->workqueue =
-		create_singlethread_workqueue(wiphy_name(local->hw.wiphy));
+		alloc_ordered_workqueue(wiphy_name(local->hw.wiphy), 0);
 	if (!local->workqueue) {
 		result = -ENOMEM;
 		goto fail_workqueue;
@@ -739,6 +739,7 @@ void ieee80211_unregister_hw(struct ieee
 	del_timer_sync(&local->work_timer);

 	cancel_work_sync(&local->reconfig_filter);
+	cancel_work_sync(&local->restart_work);

 	ieee80211_clear_tx_pending(local);
 	sta_info_stop(local);
@@ -812,12 +813,6 @@ static void __exit ieee80211_exit(void)
 	rc80211_minstrel_ht_exit();
 	rc80211_minstrel_exit();

-	/*
-	 * For key todo, it'll be empty by now but the work
-	 * might still be scheduled.
-	 */
-	flush_scheduled_work();
-
 	if (mesh_allocated)
 		ieee80211s_stop();

Re: [PATCH v2.6.36-rc7] mac80211: cancel restart_work explicitly instead of depending on flush_scheduled_work()

[Johannes Berg]

> -        create_singlethread_workqueue(wiphy_name(local->hw.wiphy));
> +        alloc_ordered_workqueue(wiphy_name(local->hw.wiphy), 0);

no description for this change?

> -    /*
> -     * For key todo, it'll be empty by now but the work
> -     * might still be scheduled.
> -     */
> -    flush_scheduled_work();

oops, todo work no longer exists...

Anyway, ACK.

Johannes
(sent on my cell phone)

Re: [PATCH v2.6.36-rc7] mac80211: cancel restart_work explicitly instead of depending on flush_scheduled_work()

[Tejun Heo]

Hello,

On 10/15/2010 06:21 PM, Johannes Berg wrote:
>> -        create_singlethread_workqueue(wiphy_name(local->hw.wiphy));
>> +        alloc_ordered_workqueue(wiphy_name(local->hw.wiphy), 0);
> 
> no description for this change?

Oops, right.  create_*workqueue() interfaces are being replaced by
alloc_*workqueue().  It's part of the workqueue interface update.

>> -    /*
>> -     * For key todo, it'll be empty by now but the work
>> -     * might still be scheduled.
>> -     */
>> -    flush_scheduled_work();
> 
> oops, todo work no longer exists...
> 
> Anyway, ACK.

Are you going to route this through wireless tree?  Shall I repost
with update patch description?

Thank you.

-- 
tejun

Re: [PATCH v2.6.36-rc7] mac80211: cancel restart_work explicitly instead of depending on flush_scheduled_work()

[Christian Lamparter]

On Friday 15 October 2010 18:39:37 Tejun Heo wrote:
> Hello,

> >> -    /*
> >> -     * For key todo, it'll be empty by now but the work
> >> -     * might still be scheduled.
> >> -     */
> >> -    flush_scheduled_work();
> > 
> > oops, todo work no longer exists...
> > 
> > Anyway, ACK.
> 
> Are you going to route this through wireless tree?  Shall I repost
> with update patch description?

wireless-testing/wireless-next has already such a patch:
"mac80211: cancel restart_work in ieee80211_unregister_hw" (2010-08-25)

<http://git.kernel.org/?p=linux/kernel/git/linville/wireless-next-2.6.git;a=commit;h=258086a48b766d12a500f98834654ffa927ca475>

John, can you please push to 2.6.36-rc8?
(I know, I am _late_, but I didn't know about any potential problem
in the mainline?)

Best regards,
	Chr

Re: [PATCH v2.6.36-rc7] mac80211: cancel restart_work explicitly instead of depending on flush_scheduled_work()

[John W. Linville]

On Fri, Oct 15, 2010 at 06:58:44PM +0200, Christian Lamparter wrote:
> On Friday 15 October 2010 18:39:37 Tejun Heo wrote:
> > Hello,
> 
> > >> -    /*
> > >> -     * For key todo, it'll be empty by now but the work
> > >> -     * might still be scheduled.
> > >> -     */
> > >> -    flush_scheduled_work();
> > > 
> > > oops, todo work no longer exists...
> > > 
> > > Anyway, ACK.
> > 
> > Are you going to route this through wireless tree?  Shall I repost
> > with update patch description?
> 
> wireless-testing/wireless-next has already such a patch:
> "mac80211: cancel restart_work in ieee80211_unregister_hw" (2010-08-25)
> 
> <http://git.kernel.org/?p=linux/kernel/git/linville/wireless-next-2.6.git;a=commit;h=258086a48b766d12a500f98834654ffa927ca475>
> 
> John, can you please push to 2.6.36-rc8?
> (I know, I am _late_, but I didn't know about any potential problem
> in the mainline?)

I don't see how either patch would merit inclusion in 2.6.36 at this point.

John
-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.

Re: [PATCH v2.6.36-rc7] mac80211: cancel restart_work explicitly instead of depending on flush_scheduled_work()

[Tejun Heo]

Hello,

On 10/15/2010 09:26 PM, John W. Linville wrote:
>> wireless-testing/wireless-next has already such a patch:
>> "mac80211: cancel restart_work in ieee80211_unregister_hw" (2010-08-25)
>>
>> <http://git.kernel.org/?p=linux/kernel/git/linville/wireless-next-2.6.git;a=commit;h=258086a48b766d12a500f98834654ffa927ca475>
>>
>> John, can you please push to 2.6.36-rc8?
>> (I know, I am _late_, but I didn't know about any potential problem
>> in the mainline?)
> 
> I don't see how either patch would merit inclusion in 2.6.36 at this point.

My patch definitely was aimed for 2.6.37-rc1 or later.  I'm not sure
whether the cancel_work_sync() is a bug fix or not but if it hasn't
actually triggered yet, I guess leaving it as-is isn't too bad at this
point.

Thanks.

-- 
tejun

Re: [PATCH v2.6.36-rc7] mac80211: cancel restart_work explicitly instead of depending on flush_scheduled_work()

[Tejun Heo]

Hello,

On 10/15/2010 10:07 PM, Christian Lamparter wrote:
> Oh, so the "[PATCH v2.6.36-rc7]" in the patch subject was just "misleading" me.

Ah, sorry about that.  I usually tag them with the version tag the
patch was generated against, not the target version.  Now that I think
about it, it could definitely be confusing.  I'll try to be clearer.

Thanks.

-- 
tejun

Re: [PATCH v2.6.36-rc7] mac80211: cancel restart_work explicitly instead of depending on flush_scheduled_work()

[Christian Lamparter]

On Friday 15 October 2010 21:33:20 Tejun Heo wrote:
> Hello,
> 
> On 10/15/2010 09:26 PM, John W. Linville wrote:
> >> wireless-testing/wireless-next has already such a patch:
> >> "mac80211: cancel restart_work in ieee80211_unregister_hw" (2010-08-25)
> >>
> >> <http://git.kernel.org/?p=linux/kernel/git/linville/wireless-next-2.6.git;a=commit;h=258086a48b766d12a500f98834654ffa927ca475>
> >>
> >> John, can you please push to 2.6.36-rc8?
> >> (I know, I am _late_, but I didn't know about any potential problem
> >> in the mainline?)
> > 
> > I don't see how either patch would merit inclusion in 2.6.36 at this point.
> 
> My patch definitely was aimed for 2.6.37-rc1 or later.
Oh, so the "[PATCH v2.6.36-rc7]" in the patch subject was just "misleading" me.
no-one could have known that there would be a 2.6.36-rc8 release.

John, please ignore my previous request of pushing the patch into 2.6.36 ;).
 
Regards,
	Chr