ipv6: Revert 'administrative down' address handling changes.
David S. Miller [Mon, 24 Jan 2011 07:27:15 +0000 (23:27 -0800)]
This reverts the following set of commits:

d1ed113f1669390da9898da3beddcc058d938587 ("ipv6: remove duplicate neigh_ifdown")
29ba5fed1bbd09c2cba890798c8f9eaab251401d ("ipv6: don't flush routes when setting loopback down")
9d82ca98f71fd686ef2f3017c5e3e6a4871b6e46 ("ipv6: fix missing in6_ifa_put in addrconf")
2de795707294972f6c34bae9de713e502c431296 ("ipv6: addrconf: don't remove address state on ifdown if the address is being kept")
8595805aafc8b077e01804c9a3668e9aa3510e89 ("IPv6: only notify protocols if address is compeletely gone")
27bdb2abcc5edb3526e25407b74bf17d1872c329 ("IPv6: keep tentative addresses in hash table")
93fa159abe50d3c55c7f83622d3f5c09b6e06f4b ("IPv6: keep route for tentative address")
8f37ada5b5f6bfb4d251a7f510f249cb855b77b3 ("IPv6: fix race between cleanup and add/delete address")
84e8b803f1e16f3a2b8b80f80a63fa2f2f8a9be6 ("IPv6: addrconf notify when address is unavailable")
dc2b99f71ef477a31020511876ab4403fb7c4420 ("IPv6: keep permanent addresses on admin down")

because the core semantic change to ipv6 address handling on ifdown
has broken some things, in particular "disable_ipv6" sysctl handling.

Stephen has made several attempts to get things back in working order,
but nothing has restored disable_ipv6 fully yet.

Reported-by: Eric W. Biederman <ebiederm@xmission.com>
Tested-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/ipv6/addrconf.c

index 24a1cf1..fd6782e 100644 (file)
@@ -2661,14 +2661,12 @@ static int addrconf_ifdown(struct net_device *dev, int how)
        struct net *net = dev_net(dev);
        struct inet6_dev *idev;
        struct inet6_ifaddr *ifa;
-       LIST_HEAD(keep_list);
-       int state;
+       int state, i;
 
        ASSERT_RTNL();
 
-       /* Flush routes if device is being removed or it is not loopback */
-       if (how || !(dev->flags & IFF_LOOPBACK))
-               rt6_ifdown(net, dev);
+       rt6_ifdown(net, dev);
+       neigh_ifdown(&nd_tbl, dev);
 
        idev = __in6_dev_get(dev);
        if (idev == NULL)
@@ -2689,6 +2687,23 @@ static int addrconf_ifdown(struct net_device *dev, int how)
 
        }
 
+       /* Step 2: clear hash table */
+       for (i = 0; i < IN6_ADDR_HSIZE; i++) {
+               struct hlist_head *h = &inet6_addr_lst[i];
+               struct hlist_node *n;
+
+               spin_lock_bh(&addrconf_hash_lock);
+       restart:
+               hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
+                       if (ifa->idev == idev) {
+                               hlist_del_init_rcu(&ifa->addr_lst);
+                               addrconf_del_timer(ifa);
+                               goto restart;
+                       }
+               }
+               spin_unlock_bh(&addrconf_hash_lock);
+       }
+
        write_lock_bh(&idev->lock);
 
        /* Step 2: clear flags for stateless addrconf */
@@ -2722,52 +2737,23 @@ static int addrconf_ifdown(struct net_device *dev, int how)
                                       struct inet6_ifaddr, if_list);
                addrconf_del_timer(ifa);
 
-               /* If just doing link down, and address is permanent
-                  and not link-local, then retain it. */
-               if (!how &&
-                   (ifa->flags&IFA_F_PERMANENT) &&
-                   !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
-                       list_move_tail(&ifa->if_list, &keep_list);
-
-                       /* If not doing DAD on this address, just keep it. */
-                       if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
-                           idev->cnf.accept_dad <= 0 ||
-                           (ifa->flags & IFA_F_NODAD))
-                               continue;
+               list_del(&ifa->if_list);
 
-                       /* If it was tentative already, no need to notify */
-                       if (ifa->flags & IFA_F_TENTATIVE)
-                               continue;
+               write_unlock_bh(&idev->lock);
 
-                       /* Flag it for later restoration when link comes up */
-                       ifa->flags |= IFA_F_TENTATIVE;
-                       ifa->state = INET6_IFADDR_STATE_DAD;
-               } else {
-                       list_del(&ifa->if_list);
-
-                       /* clear hash table */
-                       spin_lock_bh(&addrconf_hash_lock);
-                       hlist_del_init_rcu(&ifa->addr_lst);
-                       spin_unlock_bh(&addrconf_hash_lock);
-
-                       write_unlock_bh(&idev->lock);
-                       spin_lock_bh(&ifa->state_lock);
-                       state = ifa->state;
-                       ifa->state = INET6_IFADDR_STATE_DEAD;
-                       spin_unlock_bh(&ifa->state_lock);
-
-                       if (state != INET6_IFADDR_STATE_DEAD) {
-                               __ipv6_ifa_notify(RTM_DELADDR, ifa);
-                               atomic_notifier_call_chain(&inet6addr_chain,
-                                                          NETDEV_DOWN, ifa);
-                       }
+               spin_lock_bh(&ifa->state_lock);
+               state = ifa->state;
+               ifa->state = INET6_IFADDR_STATE_DEAD;
+               spin_unlock_bh(&ifa->state_lock);
 
-                       in6_ifa_put(ifa);
-                       write_lock_bh(&idev->lock);
+               if (state != INET6_IFADDR_STATE_DEAD) {
+                       __ipv6_ifa_notify(RTM_DELADDR, ifa);
+                       atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
                }
-       }
+               in6_ifa_put(ifa);
 
-       list_splice(&keep_list, &idev->addr_list);
+               write_lock_bh(&idev->lock);
+       }
 
        write_unlock_bh(&idev->lock);
 
@@ -4156,8 +4142,7 @@ static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
                addrconf_leave_solict(ifp->idev, &ifp->addr);
                dst_hold(&ifp->rt->dst);
 
-               if (ifp->state == INET6_IFADDR_STATE_DEAD &&
-                   ip6_del_rt(ifp->rt))
+               if (ip6_del_rt(ifp->rt))
                        dst_free(&ifp->rt->dst);
                break;
        }