[PATCH] mac80211: Only flush workqueue when last interface was removed

Currently the ieee80211_hw->workqueue is flushed each time
an interface is being removed. However most scheduled work
is not interface specific but device specific, for example things like
periodic work for link tuners.

This patch will move the flush_workqueue() call to directly behind
the call to ops->stop() to make sure the workqueue is only flushed
when all interfaces are gone and there really shouldn't be any scheduled
work in the drivers left.

Signed-off-by: Ivo van Doorn <redacted>
---

diff --git a/net/mac80211/main.c b/net/mac80211/main.c
index 074f71a..577845e 100644
--- a/net/mac80211/main.c
+++ b/net/mac80211/main.c

@@ -533,8 +533,6 @@ static int ieee80211_stop(struct net_device *dev)
 				local->sta_hw_scanning = 0;
 		}
 
-		flush_workqueue(local->hw.workqueue);
-
 		sdata->u.sta.flags &= ~IEEE80211_STA_PRIVACY_INVOKED;
 		kfree(sdata->u.sta.extra_ie);
 		sdata->u.sta.extra_ie = NULL;

@@ -558,6 +556,8 @@ static int ieee80211_stop(struct net_device *dev)
 
 		ieee80211_led_radio(local, 0);
 
+		flush_workqueue(local->hw.workqueue);
+
 		tasklet_disable(&local->tx_pending_tasklet);
 		tasklet_disable(&local->tasklet);
 	}

On Thu, 2008-07-03 at 19:02 +0200, Ivo van Doorn wrote:

Currently the ieee80211_hw->workqueue is flushed each time
an interface is being removed. However most scheduled work
is not interface specific but device specific, for example things like
periodic work for link tuners.

This patch will move the flush_workqueue() call to directly behind
the call to ops->stop() to make sure the workqueue is only flushed
when all interfaces are gone and there really shouldn't be any scheduled
work in the drivers left.

Yeah I think that's what we always wanted.

Signed-off-by: Ivo van Doorn <redacted>

Acked-by: Johannes Berg <johannes@sipsolutions.net>