mac80211: handle sched_scan_stopped vs. hw restart race
authorEliad Peller <eliad@wizery.com>
Tue, 5 Jan 2016 14:28:06 +0000 (16:28 +0200)
committerJohannes Berg <johannes.berg@intel.com>
Thu, 14 Jan 2016 10:10:20 +0000 (11:10 +0100)
On hw restart, mac80211 might try to reconfigure already
stopped sched scan, if ieee80211_sched_scan_stopped_work()
wasn't scheduled yet.

This in turn will keep the device driver with scheduled scan
configured, while both mac80211 and cfg80211 will clear
their sched scan state once the work is scheduled.

Fix it by ignoring ieee80211_sched_scan_stopped() calls
while in hw restart, and flush the work before starting
the reconfiguration.

Signed-off-by: Eliad Peller <eliadx.peller@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
net/mac80211/main.c
net/mac80211/scan.c

index ed4c8e66b44a9edd9188634996f641dbb2200ab5..8190bf27ebfff8b0a594b7c029130dc9b1b99957 100644 (file)
@@ -248,6 +248,7 @@ static void ieee80211_restart_work(struct work_struct *work)
 
        /* wait for scan work complete */
        flush_workqueue(local->workqueue);
+       flush_work(&local->sched_scan_stopped_work);
 
        WARN(test_bit(SCAN_HW_SCANNING, &local->scanning),
             "%s called with hardware scan in progress\n", __func__);
index a413e52f7691418116d928f684a0fc27b308c2d6..8eb68ef42e8c9c5a951b11e9774d82b347fc76af 100644 (file)
@@ -1213,6 +1213,14 @@ void ieee80211_sched_scan_stopped(struct ieee80211_hw *hw)
 
        trace_api_sched_scan_stopped(local);
 
+       /*
+        * this shouldn't really happen, so for simplicity
+        * simply ignore it, and let mac80211 reconfigure
+        * the sched scan later on.
+        */
+       if (local->in_reconfig)
+               return;
+
        schedule_work(&local->sched_scan_stopped_work);
 }
 EXPORT_SYMBOL(ieee80211_sched_scan_stopped);