Patch "wifi: mac80211: lock wiphy in IP address notifier" has been added to the 6.6-stable tree

[Sasha Levin <sashal@xxxxxxxxxx>]

This is a note to let you know that I've just added the patch titled

    wifi: mac80211: lock wiphy in IP address notifier

to the 6.6-stable tree which can be found at:
    http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
     wifi-mac80211-lock-wiphy-in-ip-address-notifier.patch
and it can be found in the queue-6.6 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree,
please let <stable@xxxxxxxxxxxxxxx> know about it.



commit 2adfa2f3a6f4d4d8981bf119cfe2681271712b6c
Author: Johannes Berg <johannes.berg@xxxxxxxxx>
Date:   Mon Aug 28 13:59:41 2023 +0200

    wifi: mac80211: lock wiphy in IP address notifier
    
    [ Upstream commit 730538edc8e0eb14b02708f65100a0deaf43e6cd ]
    
    Lock the wiphy in the IP address notifier as another
    place that should have it locked before calling into
    the driver. This needs a bit of attention since the
    notifier can be called while the wiphy is already
    locked, when we remove an interface. Handle this by
    not running the notifier in this case, and instead
    calling out to the driver directly.
    
    Signed-off-by: Johannes Berg <johannes.berg@xxxxxxxxx>
    Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>

diff --git a/net/mac80211/iface.c b/net/mac80211/iface.c
index 52b048807feae..a7c39e895b1e5 100644
--- a/net/mac80211/iface.c
+++ b/net/mac80211/iface.c
@@ -2315,6 +2315,20 @@ void ieee80211_remove_interfaces(struct ieee80211_local *local)
 	list_for_each_entry_safe(sdata, tmp, &unreg_list, list) {
 		bool netdev = sdata->dev;
 
+		/*
+		 * Remove IP addresses explicitly, since the notifier will
+		 * skip the callbacks if wdev->registered is false, since
+		 * we can't acquire the wiphy_lock() again there if already
+		 * inside this locked section.
+		 */
+		sdata_lock(sdata);
+		sdata->vif.cfg.arp_addr_cnt = 0;
+		if (sdata->vif.type == NL80211_IFTYPE_STATION &&
+		    sdata->u.mgd.associated)
+			ieee80211_vif_cfg_change_notify(sdata,
+							BSS_CHANGED_ARP_FILTER);
+		sdata_unlock(sdata);
+
 		list_del(&sdata->list);
 		cfg80211_unregister_wdev(&sdata->wdev);
 
diff --git a/net/mac80211/main.c b/net/mac80211/main.c
index e0c701c5e5f93..066424e62ff09 100644
--- a/net/mac80211/main.c
+++ b/net/mac80211/main.c
@@ -443,7 +443,7 @@ static int ieee80211_ifa_changed(struct notifier_block *nb,
 	if (!wdev)
 		return NOTIFY_DONE;
 
-	if (wdev->wiphy != local->hw.wiphy)
+	if (wdev->wiphy != local->hw.wiphy || !wdev->registered)
 		return NOTIFY_DONE;
 
 	sdata = IEEE80211_DEV_TO_SUB_IF(ndev);
@@ -458,6 +458,25 @@ static int ieee80211_ifa_changed(struct notifier_block *nb,
 		return NOTIFY_DONE;
 
 	ifmgd = &sdata->u.mgd;
+
+	/*
+	 * The nested here is needed to convince lockdep that this is
+	 * all OK. Yes, we lock the wiphy mutex here while we already
+	 * hold the notifier rwsem, that's the normal case. And yes,
+	 * we also acquire the notifier rwsem again when unregistering
+	 * a netdev while we already hold the wiphy mutex, so it does
+	 * look like a typical ABBA deadlock.
+	 *
+	 * However, both of these things happen with the RTNL held
+	 * already. Therefore, they can't actually happen, since the
+	 * lock orders really are ABC and ACB, which is fine due to
+	 * the RTNL (A).
+	 *
+	 * We still need to prevent recursion, which is accomplished
+	 * by the !wdev->registered check above.
+	 */
+	mutex_lock_nested(&local->hw.wiphy->mtx, 1);
+	__acquire(&local->hw.wiphy->mtx);
 	sdata_lock(sdata);
 
 	/* Copy the addresses to the vif config list */
@@ -476,6 +495,7 @@ static int ieee80211_ifa_changed(struct notifier_block *nb,
 		ieee80211_vif_cfg_change_notify(sdata, BSS_CHANGED_ARP_FILTER);
 
 	sdata_unlock(sdata);
+	wiphy_unlock(local->hw.wiphy);
 
 	return NOTIFY_OK;
 }