Merge branch 'linus' into cont_syslog
[safe/jmp/linux-2.6] / net / ipv6 / sit.c
index b7a50e9..e51e650 100644 (file)
@@ -28,6 +28,7 @@
 #include <linux/netdevice.h>
 #include <linux/if_arp.h>
 #include <linux/icmp.h>
+#include <linux/slab.h>
 #include <asm/uaccess.h>
 #include <linux/init.h>
 #include <linux/netfilter_ipv4.h>
 #define HASH_SIZE  16
 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
 
-static int ipip6_fb_tunnel_init(struct net_device *dev);
-static int ipip6_tunnel_init(struct net_device *dev);
+static void ipip6_tunnel_init(struct net_device *dev);
 static void ipip6_tunnel_setup(struct net_device *dev);
 
-static int sit_net_id;
+static int sit_net_id __read_mostly;
 struct sit_net {
        struct ip_tunnel *tunnels_r_l[HASH_SIZE];
        struct ip_tunnel *tunnels_r[HASH_SIZE];
@@ -77,30 +77,46 @@ struct sit_net {
        struct net_device *fb_tunnel_dev;
 };
 
-static DEFINE_RWLOCK(ipip6_lock);
+/*
+ * Locking : hash tables are protected by RCU and a spinlock
+ */
+static DEFINE_SPINLOCK(ipip6_lock);
 
+#define for_each_ip_tunnel_rcu(start) \
+       for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
+
+/*
+ * Must be invoked with rcu_read_lock
+ */
 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
-               __be32 remote, __be32 local)
+               struct net_device *dev, __be32 remote, __be32 local)
 {
        unsigned h0 = HASH(remote);
        unsigned h1 = HASH(local);
        struct ip_tunnel *t;
        struct sit_net *sitn = net_generic(net, sit_net_id);
 
-       for (t = sitn->tunnels_r_l[h0^h1]; t; t = t->next) {
+       for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
                if (local == t->parms.iph.saddr &&
-                   remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
+                   remote == t->parms.iph.daddr &&
+                   (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
+                   (t->dev->flags & IFF_UP))
                        return t;
        }
-       for (t = sitn->tunnels_r[h0]; t; t = t->next) {
-               if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
+       for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
+               if (remote == t->parms.iph.daddr &&
+                   (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
+                   (t->dev->flags & IFF_UP))
                        return t;
        }
-       for (t = sitn->tunnels_l[h1]; t; t = t->next) {
-               if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
+       for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
+               if (local == t->parms.iph.saddr &&
+                   (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
+                   (t->dev->flags & IFF_UP))
                        return t;
        }
-       if ((t = sitn->tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
+       t = rcu_dereference(sitn->tunnels_wc[0]);
+       if ((t != NULL) && (t->dev->flags & IFF_UP))
                return t;
        return NULL;
 }
@@ -136,9 +152,9 @@ static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 
        for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
                if (t == *tp) {
-                       write_lock_bh(&ipip6_lock);
+                       spin_lock_bh(&ipip6_lock);
                        *tp = t->next;
-                       write_unlock_bh(&ipip6_lock);
+                       spin_unlock_bh(&ipip6_lock);
                        break;
                }
        }
@@ -148,10 +164,27 @@ static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 {
        struct ip_tunnel **tp = ipip6_bucket(sitn, t);
 
+       spin_lock_bh(&ipip6_lock);
        t->next = *tp;
-       write_lock_bh(&ipip6_lock);
-       *tp = t;
-       write_unlock_bh(&ipip6_lock);
+       rcu_assign_pointer(*tp, t);
+       spin_unlock_bh(&ipip6_lock);
+}
+
+static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
+{
+#ifdef CONFIG_IPV6_SIT_6RD
+       struct ip_tunnel *t = netdev_priv(dev);
+
+       if (t->dev == sitn->fb_tunnel_dev) {
+               ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
+               t->ip6rd.relay_prefix = 0;
+               t->ip6rd.prefixlen = 16;
+               t->ip6rd.relay_prefixlen = 0;
+       } else {
+               struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
+               memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
+       }
+#endif
 }
 
 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
@@ -165,8 +198,14 @@ static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
        struct sit_net *sitn = net_generic(net, sit_net_id);
 
        for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
-               if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
-                       return t;
+               if (local == t->parms.iph.saddr &&
+                   remote == t->parms.iph.daddr &&
+                   parms->link == t->parms.link) {
+                       if (create)
+                               return NULL;
+                       else
+                               return t;
+               }
        }
        if (!create)
                goto failed;
@@ -188,8 +227,10 @@ static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
        }
 
        nt = netdev_priv(dev);
-       dev->init = ipip6_tunnel_init;
+
        nt->parms = *parms;
+       ipip6_tunnel_init(dev);
+       ipip6_tunnel_clone_6rd(dev, sitn);
 
        if (parms->i_flags & SIT_ISATAP)
                dev->priv_flags |= IFF_ISATAP;
@@ -208,15 +249,22 @@ failed:
        return NULL;
 }
 
+static DEFINE_SPINLOCK(ipip6_prl_lock);
+
+#define for_each_prl_rcu(start)                        \
+       for (prl = rcu_dereference(start);      \
+            prl;                               \
+            prl = rcu_dereference(prl->next))
+
 static struct ip_tunnel_prl_entry *
 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 {
-       struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
+       struct ip_tunnel_prl_entry *prl;
 
-       for (p = t->prl; p; p = p->next)
-               if (p->addr == addr)
+       for_each_prl_rcu(t->prl)
+               if (prl->addr == addr)
                        break;
-       return p;
+       return prl;
 
 }
 
@@ -241,7 +289,7 @@ static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
                kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
                NULL;
 
-       read_lock(&ipip6_lock);
+       rcu_read_lock();
 
        ca = t->prl_count < cmax ? t->prl_count : cmax;
 
@@ -259,8 +307,8 @@ static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
        }
 
        c = 0;
-       for (prl = t->prl; prl; prl = prl->next) {
-               if (c > cmax)
+       for_each_prl_rcu(t->prl) {
+               if (c >= cmax)
                        break;
                if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
                        continue;
@@ -271,7 +319,7 @@ static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
                        break;
        }
 out:
-       read_unlock(&ipip6_lock);
+       rcu_read_unlock();
 
        len = sizeof(*kp) * c;
        ret = 0;
@@ -292,12 +340,14 @@ ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
        if (a->addr == htonl(INADDR_ANY))
                return -EINVAL;
 
-       write_lock(&ipip6_lock);
+       spin_lock(&ipip6_prl_lock);
 
        for (p = t->prl; p; p = p->next) {
                if (p->addr == a->addr) {
-                       if (chg)
-                               goto update;
+                       if (chg) {
+                               p->flags = a->flags;
+                               goto out;
+                       }
                        err = -EEXIST;
                        goto out;
                }
@@ -315,46 +365,62 @@ ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
        }
 
        p->next = t->prl;
-       t->prl = p;
-       t->prl_count++;
-update:
        p->addr = a->addr;
        p->flags = a->flags;
+       t->prl_count++;
+       rcu_assign_pointer(t->prl, p);
 out:
-       write_unlock(&ipip6_lock);
+       spin_unlock(&ipip6_prl_lock);
        return err;
 }
 
+static void prl_entry_destroy_rcu(struct rcu_head *head)
+{
+       kfree(container_of(head, struct ip_tunnel_prl_entry, rcu_head));
+}
+
+static void prl_list_destroy_rcu(struct rcu_head *head)
+{
+       struct ip_tunnel_prl_entry *p, *n;
+
+       p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
+       do {
+               n = p->next;
+               kfree(p);
+               p = n;
+       } while (p);
+}
+
 static int
 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 {
        struct ip_tunnel_prl_entry *x, **p;
        int err = 0;
 
-       write_lock(&ipip6_lock);
+       spin_lock(&ipip6_prl_lock);
 
        if (a && a->addr != htonl(INADDR_ANY)) {
                for (p = &t->prl; *p; p = &(*p)->next) {
                        if ((*p)->addr == a->addr) {
                                x = *p;
                                *p = x->next;
-                               kfree(x);
+                               call_rcu(&x->rcu_head, prl_entry_destroy_rcu);
                                t->prl_count--;
                                goto out;
                        }
                }
                err = -ENXIO;
        } else {
-               while (t->prl) {
+               if (t->prl) {
+                       t->prl_count = 0;
                        x = t->prl;
-                       t->prl = t->prl->next;
-                       kfree(x);
-                       t->prl_count--;
+                       call_rcu(&x->rcu_head, prl_list_destroy_rcu);
+                       t->prl = NULL;
                }
        }
 out:
-       write_unlock(&ipip6_lock);
-       return 0;
+       spin_unlock(&ipip6_prl_lock);
+       return err;
 }
 
 static int
@@ -363,7 +429,7 @@ isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
        struct ip_tunnel_prl_entry *p;
        int ok = 1;
 
-       read_lock(&ipip6_lock);
+       rcu_read_lock();
        p = __ipip6_tunnel_locate_prl(t, iph->saddr);
        if (p) {
                if (p->flags & PRL_DEFAULT)
@@ -379,7 +445,7 @@ isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
                else
                        ok = 0;
        }
-       read_unlock(&ipip6_lock);
+       rcu_read_unlock();
        return ok;
 }
 
@@ -389,9 +455,9 @@ static void ipip6_tunnel_uninit(struct net_device *dev)
        struct sit_net *sitn = net_generic(net, sit_net_id);
 
        if (dev == sitn->fb_tunnel_dev) {
-               write_lock_bh(&ipip6_lock);
+               spin_lock_bh(&ipip6_lock);
                sitn->tunnels_wc[0] = NULL;
-               write_unlock_bh(&ipip6_lock);
+               spin_unlock_bh(&ipip6_lock);
                dev_put(dev);
        } else {
                ipip6_tunnel_unlink(sitn, netdev_priv(dev));
@@ -444,8 +510,11 @@ static int ipip6_err(struct sk_buff *skb, u32 info)
 
        err = -ENOENT;
 
-       read_lock(&ipip6_lock);
-       t = ipip6_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
+       rcu_read_lock();
+       t = ipip6_tunnel_lookup(dev_net(skb->dev),
+                               skb->dev,
+                               iph->daddr,
+                               iph->saddr);
        if (t == NULL || t->parms.iph.daddr == 0)
                goto out;
 
@@ -453,13 +522,13 @@ static int ipip6_err(struct sk_buff *skb, u32 info)
        if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
                goto out;
 
-       if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
+       if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
                t->err_count++;
        else
                t->err_count = 1;
        t->err_time = jiffies;
 out:
-       read_unlock(&ipip6_lock);
+       rcu_read_unlock();
        return err;
 }
 
@@ -479,9 +548,10 @@ static int ipip6_rcv(struct sk_buff *skb)
 
        iph = ip_hdr(skb);
 
-       read_lock(&ipip6_lock);
-       if ((tunnel = ipip6_tunnel_lookup(dev_net(skb->dev),
-                                       iph->saddr, iph->daddr)) != NULL) {
+       rcu_read_lock();
+       tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
+                                    iph->saddr, iph->daddr);
+       if (tunnel != NULL) {
                secpath_reset(skb);
                skb->mac_header = skb->network_header;
                skb_reset_network_header(skb);
@@ -492,40 +562,61 @@ static int ipip6_rcv(struct sk_buff *skb)
                if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
                    !isatap_chksrc(skb, iph, tunnel)) {
                        tunnel->dev->stats.rx_errors++;
-                       read_unlock(&ipip6_lock);
+                       rcu_read_unlock();
                        kfree_skb(skb);
                        return 0;
                }
-               tunnel->dev->stats.rx_packets++;
-               tunnel->dev->stats.rx_bytes += skb->len;
-               skb->dev = tunnel->dev;
-               dst_release(skb->dst);
-               skb->dst = NULL;
-               nf_reset(skb);
+
+               skb_tunnel_rx(skb, tunnel->dev);
+
                ipip6_ecn_decapsulate(iph, skb);
                netif_rx(skb);
-               read_unlock(&ipip6_lock);
+               rcu_read_unlock();
                return 0;
        }
 
        icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
-       read_unlock(&ipip6_lock);
+       rcu_read_unlock();
 out:
        kfree_skb(skb);
        return 0;
 }
 
-/* Returns the embedded IPv4 address if the IPv6 address
-   comes from 6to4 (RFC 3056) addr space */
-
-static inline __be32 try_6to4(struct in6_addr *v6dst)
+/*
+ * Returns the embedded IPv4 address if the IPv6 address
+ * comes from 6rd / 6to4 (RFC 3056) addr space.
+ */
+static inline
+__be32 try_6rd(struct in6_addr *v6dst, struct ip_tunnel *tunnel)
 {
        __be32 dst = 0;
 
+#ifdef CONFIG_IPV6_SIT_6RD
+       if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
+                             tunnel->ip6rd.prefixlen)) {
+               unsigned pbw0, pbi0;
+               int pbi1;
+               u32 d;
+
+               pbw0 = tunnel->ip6rd.prefixlen >> 5;
+               pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
+
+               d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
+                   tunnel->ip6rd.relay_prefixlen;
+
+               pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
+               if (pbi1 > 0)
+                       d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
+                            (32 - pbi1);
+
+               dst = tunnel->ip6rd.relay_prefix | htonl(d);
+       }
+#else
        if (v6dst->s6_addr16[0] == htons(0x2002)) {
                /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
                memcpy(&dst, &v6dst->s6_addr16[1], 4);
        }
+#endif
        return dst;
 }
 
@@ -534,13 +625,16 @@ static inline __be32 try_6to4(struct in6_addr *v6dst)
  *     and that skb is filled properly by that function.
  */
 
-static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
+static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
+                                    struct net_device *dev)
 {
        struct ip_tunnel *tunnel = netdev_priv(dev);
-       struct net_device_stats *stats = &tunnel->dev->stats;
+       struct net_device_stats *stats = &dev->stats;
+       struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
        struct iphdr  *tiph = &tunnel->parms.iph;
        struct ipv6hdr *iph6 = ipv6_hdr(skb);
        u8     tos = tunnel->parms.iph.tos;
+       __be16 df = tiph->frag_off;
        struct rtable *rt;                      /* Route to the other host */
        struct net_device *tdev;                        /* Device to other host */
        struct iphdr  *iph;                     /* Our new IP header */
@@ -550,11 +644,6 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        struct in6_addr *addr6;
        int addr_type;
 
-       if (tunnel->recursion++) {
-               stats->collisions++;
-               goto tx_error;
-       }
-
        if (skb->protocol != htons(ETH_P_IPV6))
                goto tx_error;
 
@@ -562,8 +651,8 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        if (dev->priv_flags & IFF_ISATAP) {
                struct neighbour *neigh = NULL;
 
-               if (skb->dst)
-                       neigh = skb->dst->neighbour;
+               if (skb_dst(skb))
+                       neigh = skb_dst(skb)->neighbour;
 
                if (neigh == NULL) {
                        if (net_ratelimit())
@@ -582,13 +671,13 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        }
 
        if (!dst)
-               dst = try_6to4(&iph6->daddr);
+               dst = try_6rd(&iph6->daddr, tunnel);
 
        if (!dst) {
                struct neighbour *neigh = NULL;
 
-               if (skb->dst)
-                       neigh = skb->dst->neighbour;
+               if (skb_dst(skb))
+                       neigh = skb_dst(skb)->neighbour;
 
                if (neigh == NULL) {
                        if (net_ratelimit())
@@ -635,29 +724,33 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
                goto tx_error;
        }
 
-       if (tiph->frag_off)
+       if (df) {
                mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
-       else
-               mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 
-       if (mtu < 68) {
-               stats->collisions++;
-               ip_rt_put(rt);
-               goto tx_error;
-       }
-       if (mtu < IPV6_MIN_MTU)
-               mtu = IPV6_MIN_MTU;
-       if (tunnel->parms.iph.daddr && skb->dst)
-               skb->dst->ops->update_pmtu(skb->dst, mtu);
+               if (mtu < 68) {
+                       stats->collisions++;
+                       ip_rt_put(rt);
+                       goto tx_error;
+               }
 
-       if (skb->len > mtu) {
-               icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
-               ip_rt_put(rt);
-               goto tx_error;
+               if (mtu < IPV6_MIN_MTU) {
+                       mtu = IPV6_MIN_MTU;
+                       df = 0;
+               }
+
+               if (tunnel->parms.iph.daddr && skb_dst(skb))
+                       skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
+
+               if (skb->len > mtu) {
+                       icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
+                       ip_rt_put(rt);
+                       goto tx_error;
+               }
        }
 
        if (tunnel->err_count > 0) {
-               if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
+               if (time_before(jiffies,
+                               tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
                        tunnel->err_count--;
                        dst_link_failure(skb);
                } else
@@ -674,10 +767,9 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
                struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
                if (!new_skb) {
                        ip_rt_put(rt);
-                       stats->tx_dropped++;
+                       txq->tx_dropped++;
                        dev_kfree_skb(skb);
-                       tunnel->recursion--;
-                       return 0;
+                       return NETDEV_TX_OK;
                }
                if (skb->sk)
                        skb_set_owner_w(new_skb, skb->sk);
@@ -691,8 +783,8 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        skb_reset_network_header(skb);
        memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
        IPCB(skb)->flags = 0;
-       dst_release(skb->dst);
-       skb->dst = &rt->u.dst;
+       skb_dst_drop(skb);
+       skb_dst_set(skb, &rt->u.dst);
 
        /*
         *      Push down and install the IPIP header.
@@ -701,11 +793,7 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        iph                     =       ip_hdr(skb);
        iph->version            =       4;
        iph->ihl                =       sizeof(struct iphdr)>>2;
-       if (mtu > IPV6_MIN_MTU)
-               iph->frag_off   =       htons(IP_DF);
-       else
-               iph->frag_off   =       0;
-
+       iph->frag_off           =       df;
        iph->protocol           =       IPPROTO_IPV6;
        iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
        iph->daddr              =       rt->rt_dst;
@@ -717,16 +805,14 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        nf_reset(skb);
 
        IPTUNNEL_XMIT();
-       tunnel->recursion--;
-       return 0;
+       return NETDEV_TX_OK;
 
 tx_error_icmp:
        dst_link_failure(skb);
 tx_error:
        stats->tx_errors++;
        dev_kfree_skb(skb);
-       tunnel->recursion--;
-       return 0;
+       return NETDEV_TX_OK;
 }
 
 static void ipip6_tunnel_bind_dev(struct net_device *dev)
@@ -774,9 +860,15 @@ ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
        struct ip_tunnel *t;
        struct net *net = dev_net(dev);
        struct sit_net *sitn = net_generic(net, sit_net_id);
+#ifdef CONFIG_IPV6_SIT_6RD
+       struct ip_tunnel_6rd ip6rd;
+#endif
 
        switch (cmd) {
        case SIOCGETTUNNEL:
+#ifdef CONFIG_IPV6_SIT_6RD
+       case SIOCGET6RD:
+#endif
                t = NULL;
                if (dev == sitn->fb_tunnel_dev) {
                        if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
@@ -787,9 +879,25 @@ ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
                }
                if (t == NULL)
                        t = netdev_priv(dev);
-               memcpy(&p, &t->parms, sizeof(p));
-               if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
-                       err = -EFAULT;
+
+               err = -EFAULT;
+               if (cmd == SIOCGETTUNNEL) {
+                       memcpy(&p, &t->parms, sizeof(p));
+                       if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
+                                        sizeof(p)))
+                               goto done;
+#ifdef CONFIG_IPV6_SIT_6RD
+               } else {
+                       ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
+                       ip6rd.relay_prefix = t->ip6rd.relay_prefix;
+                       ip6rd.prefixlen = t->ip6rd.prefixlen;
+                       ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
+                       if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
+                                        sizeof(ip6rd)))
+                               goto done;
+#endif
+               }
+               err = 0;
                break;
 
        case SIOCADDTUNNEL:
@@ -910,6 +1018,54 @@ ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
                netdev_state_change(dev);
                break;
 
+#ifdef CONFIG_IPV6_SIT_6RD
+       case SIOCADD6RD:
+       case SIOCCHG6RD:
+       case SIOCDEL6RD:
+               err = -EPERM;
+               if (!capable(CAP_NET_ADMIN))
+                       goto done;
+
+               err = -EFAULT;
+               if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
+                                  sizeof(ip6rd)))
+                       goto done;
+
+               t = netdev_priv(dev);
+
+               if (cmd != SIOCDEL6RD) {
+                       struct in6_addr prefix;
+                       __be32 relay_prefix;
+
+                       err = -EINVAL;
+                       if (ip6rd.relay_prefixlen > 32 ||
+                           ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
+                               goto done;
+
+                       ipv6_addr_prefix(&prefix, &ip6rd.prefix,
+                                        ip6rd.prefixlen);
+                       if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
+                               goto done;
+                       if (ip6rd.relay_prefixlen)
+                               relay_prefix = ip6rd.relay_prefix &
+                                              htonl(0xffffffffUL <<
+                                                    (32 - ip6rd.relay_prefixlen));
+                       else
+                               relay_prefix = 0;
+                       if (relay_prefix != ip6rd.relay_prefix)
+                               goto done;
+
+                       ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
+                       t->ip6rd.relay_prefix = relay_prefix;
+                       t->ip6rd.prefixlen = ip6rd.prefixlen;
+                       t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
+               } else
+                       ipip6_tunnel_clone_6rd(dev, sitn);
+
+               err = 0;
+               break;
+#endif
+
        default:
                err = -EINVAL;
        }
@@ -926,28 +1082,31 @@ static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
        return 0;
 }
 
+static const struct net_device_ops ipip6_netdev_ops = {
+       .ndo_uninit     = ipip6_tunnel_uninit,
+       .ndo_start_xmit = ipip6_tunnel_xmit,
+       .ndo_do_ioctl   = ipip6_tunnel_ioctl,
+       .ndo_change_mtu = ipip6_tunnel_change_mtu,
+};
+
 static void ipip6_tunnel_setup(struct net_device *dev)
 {
-       dev->uninit             = ipip6_tunnel_uninit;
+       dev->netdev_ops         = &ipip6_netdev_ops;
        dev->destructor         = free_netdev;
-       dev->hard_start_xmit    = ipip6_tunnel_xmit;
-       dev->do_ioctl           = ipip6_tunnel_ioctl;
-       dev->change_mtu         = ipip6_tunnel_change_mtu;
 
        dev->type               = ARPHRD_SIT;
        dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
        dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
        dev->flags              = IFF_NOARP;
+       dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
        dev->iflink             = 0;
        dev->addr_len           = 4;
        dev->features           |= NETIF_F_NETNS_LOCAL;
 }
 
-static int ipip6_tunnel_init(struct net_device *dev)
+static void ipip6_tunnel_init(struct net_device *dev)
 {
-       struct ip_tunnel *tunnel;
-
-       tunnel = netdev_priv(dev);
+       struct ip_tunnel *tunnel = netdev_priv(dev);
 
        tunnel->dev = dev;
        strcpy(tunnel->parms.name, dev->name);
@@ -956,11 +1115,9 @@ static int ipip6_tunnel_init(struct net_device *dev)
        memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
 
        ipip6_tunnel_bind_dev(dev);
-
-       return 0;
 }
 
-static int ipip6_fb_tunnel_init(struct net_device *dev)
+static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
 {
        struct ip_tunnel *tunnel = netdev_priv(dev);
        struct iphdr *iph = &tunnel->parms.iph;
@@ -977,7 +1134,6 @@ static int ipip6_fb_tunnel_init(struct net_device *dev)
 
        dev_hold(dev);
        sitn->tunnels_wc[0]     = tunnel;
-       return 0;
 }
 
 static struct xfrm_tunnel sit_handler = {
@@ -986,33 +1142,27 @@ static struct xfrm_tunnel sit_handler = {
        .priority       =       1,
 };
 
-static void sit_destroy_tunnels(struct sit_net *sitn)
+static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
 {
        int prio;
 
        for (prio = 1; prio < 4; prio++) {
                int h;
                for (h = 0; h < HASH_SIZE; h++) {
-                       struct ip_tunnel *t;
-                       while ((t = sitn->tunnels[prio][h]) != NULL)
-                               unregister_netdevice(t->dev);
+                       struct ip_tunnel *t = sitn->tunnels[prio][h];
+
+                       while (t != NULL) {
+                               unregister_netdevice_queue(t->dev, head);
+                               t = t->next;
+                       }
                }
        }
 }
 
-static int sit_init_net(struct net *net)
+static int __net_init sit_init_net(struct net *net)
 {
+       struct sit_net *sitn = net_generic(net, sit_net_id);
        int err;
-       struct sit_net *sitn;
-
-       err = -ENOMEM;
-       sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
-       if (sitn == NULL)
-               goto err_alloc;
-
-       err = net_assign_generic(net, sit_net_id, sitn);
-       if (err < 0)
-               goto err_assign;
 
        sitn->tunnels[0] = sitn->tunnels_wc;
        sitn->tunnels[1] = sitn->tunnels_l;
@@ -1025,47 +1175,48 @@ static int sit_init_net(struct net *net)
                err = -ENOMEM;
                goto err_alloc_dev;
        }
-
-       sitn->fb_tunnel_dev->init = ipip6_fb_tunnel_init;
        dev_net_set(sitn->fb_tunnel_dev, net);
 
+       ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
+       ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
+
        if ((err = register_netdev(sitn->fb_tunnel_dev)))
                goto err_reg_dev;
 
        return 0;
 
 err_reg_dev:
+       dev_put(sitn->fb_tunnel_dev);
        free_netdev(sitn->fb_tunnel_dev);
 err_alloc_dev:
-       /* nothing */
-err_assign:
-       kfree(sitn);
-err_alloc:
        return err;
 }
 
-static void sit_exit_net(struct net *net)
+static void __net_exit sit_exit_net(struct net *net)
 {
-       struct sit_net *sitn;
+       struct sit_net *sitn = net_generic(net, sit_net_id);
+       LIST_HEAD(list);
 
-       sitn = net_generic(net, sit_net_id);
        rtnl_lock();
-       sit_destroy_tunnels(sitn);
-       unregister_netdevice(sitn->fb_tunnel_dev);
+       sit_destroy_tunnels(sitn, &list);
+       unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
+       unregister_netdevice_many(&list);
        rtnl_unlock();
-       kfree(sitn);
 }
 
 static struct pernet_operations sit_net_ops = {
        .init = sit_init_net,
        .exit = sit_exit_net,
+       .id   = &sit_net_id,
+       .size = sizeof(struct sit_net),
 };
 
 static void __exit sit_cleanup(void)
 {
        xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 
-       unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
+       unregister_pernet_device(&sit_net_ops);
+       rcu_barrier(); /* Wait for completion of call_rcu()'s */
 }
 
 static int __init sit_init(void)
@@ -1074,15 +1225,14 @@ static int __init sit_init(void)
 
        printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
 
-       if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
+       err = register_pernet_device(&sit_net_ops);
+       if (err < 0)
+               return err;
+       err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
+       if (err < 0) {
+               unregister_pernet_device(&sit_net_ops);
                printk(KERN_INFO "sit init: Can't add protocol\n");
-               return -EAGAIN;
        }
-
-       err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
-       if (err < 0)
-               xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
-
        return err;
 }