Revert "bridge: only expire the mdb entry when query is received"
authorLinus Lüssing <linus.luessing@web.de>
Sat, 19 Oct 2013 22:58:57 +0000 (00:58 +0200)
committerDavid S. Miller <davem@davemloft.net>
Tue, 22 Oct 2013 18:41:02 +0000 (14:41 -0400)
While this commit was a good attempt to fix issues occuring when no
multicast querier is present, this commit still has two more issues:

1) There are cases where mdb entries do not expire even if there is a
querier present. The bridge will unnecessarily continue flooding
multicast packets on the according ports.

2) Never removing an mdb entry could be exploited for a Denial of
Service by an attacker on the local link, slowly, but steadily eating up
all memory.

Actually, this commit became obsolete with
"bridge: disable snooping if there is no querier" (b00589af3b)
which included fixes for a few more cases.

Therefore reverting the following commits (the commit stated in the
commit message plus three of its follow up fixes):

====================
Revert "bridge: update mdb expiration timer upon reports."
This reverts commit f144febd93d5ee534fdf23505ab091b2b9088edc.
Revert "bridge: do not call setup_timer() multiple times"
This reverts commit 1faabf2aab1fdaa1ace4e8c829d1b9cf7bfec2f1.
Revert "bridge: fix some kernel warning in multicast timer"
This reverts commit c7e8e8a8f7a70b343ca1e0f90a31e35ab2d16de1.
Revert "bridge: only expire the mdb entry when query is received"
This reverts commit 9f00b2e7cf241fa389733d41b615efdaa2cb0f5b.
====================

CC: Cong Wang <amwang@redhat.com>
Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Reviewed-by: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/bridge/br_mdb.c
net/bridge/br_multicast.c
net/bridge/br_private.h

index 85a09bb..b7b1914 100644 (file)
@@ -453,7 +453,7 @@ static int __br_mdb_del(struct net_bridge *br, struct br_mdb_entry *entry)
                call_rcu_bh(&p->rcu, br_multicast_free_pg);
                err = 0;
 
-               if (!mp->ports && !mp->mglist && mp->timer_armed &&
+               if (!mp->ports && !mp->mglist &&
                    netif_running(br->dev))
                        mod_timer(&mp->timer, jiffies);
                break;
index 1085f21..8b0b610 100644 (file)
@@ -272,7 +272,7 @@ static void br_multicast_del_pg(struct net_bridge *br,
                del_timer(&p->timer);
                call_rcu_bh(&p->rcu, br_multicast_free_pg);
 
-               if (!mp->ports && !mp->mglist && mp->timer_armed &&
+               if (!mp->ports && !mp->mglist &&
                    netif_running(br->dev))
                        mod_timer(&mp->timer, jiffies);
 
@@ -611,9 +611,6 @@ rehash:
                break;
 
        default:
-               /* If we have an existing entry, update it's expire timer */
-               mod_timer(&mp->timer,
-                         jiffies + br->multicast_membership_interval);
                goto out;
        }
 
@@ -623,7 +620,6 @@ rehash:
 
        mp->br = br;
        mp->addr = *group;
-
        setup_timer(&mp->timer, br_multicast_group_expired,
                    (unsigned long)mp);
 
@@ -663,6 +659,7 @@ static int br_multicast_add_group(struct net_bridge *br,
        struct net_bridge_mdb_entry *mp;
        struct net_bridge_port_group *p;
        struct net_bridge_port_group __rcu **pp;
+       unsigned long now = jiffies;
        int err;
 
        spin_lock(&br->multicast_lock);
@@ -677,18 +674,15 @@ static int br_multicast_add_group(struct net_bridge *br,
 
        if (!port) {
                mp->mglist = true;
+               mod_timer(&mp->timer, now + br->multicast_membership_interval);
                goto out;
        }
 
        for (pp = &mp->ports;
             (p = mlock_dereference(*pp, br)) != NULL;
             pp = &p->next) {
-               if (p->port == port) {
-                       /* We already have a portgroup, update the timer.  */
-                       mod_timer(&p->timer,
-                                 jiffies + br->multicast_membership_interval);
-                       goto out;
-               }
+               if (p->port == port)
+                       goto found;
                if ((unsigned long)p->port < (unsigned long)port)
                        break;
        }
@@ -699,6 +693,8 @@ static int br_multicast_add_group(struct net_bridge *br,
        rcu_assign_pointer(*pp, p);
        br_mdb_notify(br->dev, port, group, RTM_NEWMDB);
 
+found:
+       mod_timer(&p->timer, now + br->multicast_membership_interval);
 out:
        err = 0;
 
@@ -1198,9 +1194,6 @@ static int br_ip4_multicast_query(struct net_bridge *br,
        if (!mp)
                goto out;
 
-       mod_timer(&mp->timer, now + br->multicast_membership_interval);
-       mp->timer_armed = true;
-
        max_delay *= br->multicast_last_member_count;
 
        if (mp->mglist &&
@@ -1277,9 +1270,6 @@ static int br_ip6_multicast_query(struct net_bridge *br,
        if (!mp)
                goto out;
 
-       mod_timer(&mp->timer, now + br->multicast_membership_interval);
-       mp->timer_armed = true;
-
        max_delay *= br->multicast_last_member_count;
        if (mp->mglist &&
            (timer_pending(&mp->timer) ?
@@ -1365,7 +1355,7 @@ static void br_multicast_leave_group(struct net_bridge *br,
                        call_rcu_bh(&p->rcu, br_multicast_free_pg);
                        br_mdb_notify(br->dev, port, group, RTM_DELMDB);
 
-                       if (!mp->ports && !mp->mglist && mp->timer_armed &&
+                       if (!mp->ports && !mp->mglist &&
                            netif_running(br->dev))
                                mod_timer(&mp->timer, jiffies);
                }
@@ -1377,12 +1367,30 @@ static void br_multicast_leave_group(struct net_bridge *br,
                     br->multicast_last_member_interval;
 
        if (!port) {
-               if (mp->mglist && mp->timer_armed &&
+               if (mp->mglist &&
                    (timer_pending(&mp->timer) ?
                     time_after(mp->timer.expires, time) :
                     try_to_del_timer_sync(&mp->timer) >= 0)) {
                        mod_timer(&mp->timer, time);
                }
+
+               goto out;
+       }
+
+       for (p = mlock_dereference(mp->ports, br);
+            p != NULL;
+            p = mlock_dereference(p->next, br)) {
+               if (p->port != port)
+                       continue;
+
+               if (!hlist_unhashed(&p->mglist) &&
+                   (timer_pending(&p->timer) ?
+                    time_after(p->timer.expires, time) :
+                    try_to_del_timer_sync(&p->timer) >= 0)) {
+                       mod_timer(&p->timer, time);
+               }
+
+               break;
        }
 out:
        spin_unlock(&br->multicast_lock);
@@ -1805,7 +1813,6 @@ void br_multicast_stop(struct net_bridge *br)
                hlist_for_each_entry_safe(mp, n, &mdb->mhash[i],
                                          hlist[ver]) {
                        del_timer(&mp->timer);
-                       mp->timer_armed = false;
                        call_rcu_bh(&mp->rcu, br_multicast_free_group);
                }
        }
index 7ca2ae4..e14c33b 100644 (file)
@@ -126,7 +126,6 @@ struct net_bridge_mdb_entry
        struct timer_list               timer;
        struct br_ip                    addr;
        bool                            mglist;
-       bool                            timer_armed;
 };
 
 struct net_bridge_mdb_htable