xfrm6: Fix xfrm6_policy.c build when SYSCTL disabled.
[safe/jmp/linux-2.6] / net / ipv6 / sit.c
index 66cf0be..d335a30 100644 (file)
@@ -6,8 +6,6 @@
  *     Pedro Roque             <roque@di.fc.ul.pt>
  *     Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
  *
- *     $Id: sit.c,v 1.53 2001/09/25 05:09:53 davem Exp $
- *
  *     This program is free software; you can redistribute it and/or
  *      modify it under the terms of the GNU General Public License
  *      as published by the Free Software Foundation; either version
@@ -17,6 +15,7 @@
  * Roger Venning <r.venning@telstra.com>:      6to4 support
  * Nate Thompson <nate@thebog.net>:            6to4 support
  * Fred Templin <fred.l.templin@boeing.com>:   isatap support
+ * Sascha Hlusiak <mail@saschahlusiak.de>:     stateless autoconf for isatap
  */
 
 #include <linux/module.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_fb_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;
 struct sit_net {
-};
+       struct ip_tunnel *tunnels_r_l[HASH_SIZE];
+       struct ip_tunnel *tunnels_r[HASH_SIZE];
+       struct ip_tunnel *tunnels_l[HASH_SIZE];
+       struct ip_tunnel *tunnels_wc[1];
+       struct ip_tunnel **tunnels[4];
 
-static struct net_device *ipip6_fb_tunnel_dev;
-
-static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
-static struct ip_tunnel *tunnels_r[HASH_SIZE];
-static struct ip_tunnel *tunnels_l[HASH_SIZE];
-static struct ip_tunnel *tunnels_wc[1];
-static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
+       struct net_device *fb_tunnel_dev;
+};
 
 static DEFINE_RWLOCK(ipip6_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 = tunnels_r_l[h0^h1]; t; t = t->next) {
+       for (t = sitn->tunnels_r_l[h0^h1]; t; t = t->next) {
                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 = tunnels_r[h0]; t; t = t->next) {
-               if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
+       for (t = sitn->tunnels_r[h0]; t; t = t->next) {
+               if (remote == t->parms.iph.daddr &&
+                   (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
+                   (t->dev->flags & IFF_UP))
                        return t;
        }
-       for (t = tunnels_l[h1]; t; t = t->next) {
-               if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
+       for (t = sitn->tunnels_l[h1]; t; t = t->next) {
+               if (local == t->parms.iph.saddr &&
+                   (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
+                   (t->dev->flags & IFF_UP))
                        return t;
        }
-       if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
+       t = sitn->tunnels_wc[0];
+       if ((t != NULL) && (t->dev->flags & IFF_UP))
                return t;
        return NULL;
 }
@@ -123,7 +129,7 @@ static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
                prio |= 1;
                h ^= HASH(local);
        }
-       return &tunnels[prio][h];
+       return &sitn->tunnels[prio][h];
 }
 
 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
@@ -167,8 +173,14 @@ 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;
@@ -182,14 +194,17 @@ static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
        if (dev == NULL)
                return NULL;
 
+       dev_net_set(dev, net);
+
        if (strchr(name, '%')) {
                if (dev_alloc_name(dev, name) < 0)
                        goto failed_free;
        }
 
        nt = netdev_priv(dev);
-       dev->init = ipip6_tunnel_init;
+
        nt->parms = *parms;
+       ipip6_tunnel_init(dev);
 
        if (parms->i_flags & SIT_ISATAP)
                dev->priv_flags |= IFF_ISATAP;
@@ -208,6 +223,44 @@ failed:
        return NULL;
 }
 
+static void ipip6_tunnel_rs_timer(unsigned long data)
+{
+       struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *) data;
+       struct inet6_dev *ifp;
+       struct inet6_ifaddr *addr;
+
+       spin_lock(&p->lock);
+       ifp = __in6_dev_get(p->tunnel->dev);
+
+       read_lock_bh(&ifp->lock);
+       for (addr = ifp->addr_list; addr; addr = addr->if_next) {
+               struct in6_addr rtr;
+
+               if (!(ipv6_addr_type(&addr->addr) & IPV6_ADDR_LINKLOCAL))
+                       continue;
+
+               /* Send RS to guessed linklocal address of router
+                *
+                * Better: send to ff02::2 encapsuled in unicast directly
+                * to router-v4 instead of guessing the v6 address.
+                *
+                * Cisco/Windows seem to not set the u/l bit correctly,
+                * so we won't guess right.
+                */
+               ipv6_addr_set(&rtr,  htonl(0xFE800000), 0, 0, 0);
+               if (!__ipv6_isatap_ifid(rtr.s6_addr + 8,
+                                       p->addr)) {
+                       ndisc_send_rs(p->tunnel->dev, &addr->addr, &rtr);
+               }
+       }
+       read_unlock_bh(&ifp->lock);
+
+       mod_timer(&p->rs_timer, jiffies + HZ * p->rs_delay);
+       spin_unlock(&p->lock);
+
+       return;
+}
+
 static struct ip_tunnel_prl_entry *
 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 {
@@ -220,15 +273,18 @@ __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 
 }
 
-static int ipip6_tunnel_get_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
+static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
+                               struct ip_tunnel_prl __user *a)
 {
-       struct ip_tunnel_prl *kp;
+       struct ip_tunnel_prl kprl, *kp;
        struct ip_tunnel_prl_entry *prl;
        unsigned int cmax, c = 0, ca, len;
        int ret = 0;
 
-       cmax = a->datalen / sizeof(*a);
-       if (cmax > 1 && a->addr != htonl(INADDR_ANY))
+       if (copy_from_user(&kprl, a, sizeof(kprl)))
+               return -EFAULT;
+       cmax = kprl.datalen / sizeof(kprl);
+       if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
                cmax = 1;
 
        /* For simple GET or for root users,
@@ -259,26 +315,26 @@ static int ipip6_tunnel_get_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
        for (prl = t->prl; prl; prl = prl->next) {
                if (c > cmax)
                        break;
-               if (a->addr != htonl(INADDR_ANY) && prl->addr != a->addr)
+               if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
                        continue;
                kp[c].addr = prl->addr;
                kp[c].flags = prl->flags;
+               kp[c].rs_delay = prl->rs_delay;
                c++;
-               if (a->addr != htonl(INADDR_ANY))
+               if (kprl.addr != htonl(INADDR_ANY))
                        break;
        }
 out:
        read_unlock(&ipip6_lock);
 
        len = sizeof(*kp) * c;
-       ret = len ? copy_to_user(a->data, kp, len) : 0;
+       ret = 0;
+       if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
+               ret = -EFAULT;
 
        kfree(kp);
-       if (ret)
-               return -EFAULT;
 
-       a->datalen = len;
-       return 0;
+       return ret;
 }
 
 static int
@@ -313,11 +369,23 @@ ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
        }
 
        p->next = t->prl;
+       p->tunnel = t;
        t->prl = p;
        t->prl_count++;
+
+       spin_lock_init(&p->lock);
+       setup_timer(&p->rs_timer, ipip6_tunnel_rs_timer, (unsigned long) p);
 update:
        p->addr = a->addr;
        p->flags = a->flags;
+       p->rs_delay = a->rs_delay;
+       if (p->rs_delay == 0)
+               p->rs_delay = IPTUNNEL_RS_DEFAULT_DELAY;
+       spin_lock(&p->lock);
+       del_timer(&p->rs_timer);
+       if (p->flags & PRL_DEFAULT)
+               mod_timer(&p->rs_timer, jiffies + 1);
+       spin_unlock(&p->lock);
 out:
        write_unlock(&ipip6_lock);
        return err;
@@ -336,6 +404,9 @@ ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
                        if ((*p)->addr == a->addr) {
                                x = *p;
                                *p = x->next;
+                               spin_lock(&x->lock);
+                               del_timer(&x->rs_timer);
+                               spin_unlock(&x->lock);
                                kfree(x);
                                t->prl_count--;
                                goto out;
@@ -346,13 +417,16 @@ ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
                while (t->prl) {
                        x = t->prl;
                        t->prl = t->prl->next;
+                       spin_lock(&x->lock);
+                       del_timer(&x->rs_timer);
+                       spin_unlock(&x->lock);
                        kfree(x);
                        t->prl_count--;
                }
        }
 out:
        write_unlock(&ipip6_lock);
-       return 0;
+       return err;
 }
 
 static int
@@ -386,9 +460,9 @@ static void ipip6_tunnel_uninit(struct net_device *dev)
        struct net *net = dev_net(dev);
        struct sit_net *sitn = net_generic(net, sit_net_id);
 
-       if (dev == ipip6_fb_tunnel_dev) {
+       if (dev == sitn->fb_tunnel_dev) {
                write_lock_bh(&ipip6_lock);
-               tunnels_wc[0] = NULL;
+               sitn->tunnels_wc[0] = NULL;
                write_unlock_bh(&ipip6_lock);
                dev_put(dev);
        } else {
@@ -401,9 +475,8 @@ static void ipip6_tunnel_uninit(struct net_device *dev)
 
 static int ipip6_err(struct sk_buff *skb, u32 info)
 {
-#ifndef I_WISH_WORLD_WERE_PERFECT
 
-/* It is not :-( All the routers (except for Linux) return only
+/* All the routers (except for Linux) return only
    8 bytes of packet payload. It means, that precise relaying of
    ICMP in the real Internet is absolutely infeasible.
  */
@@ -444,7 +517,10 @@ static int ipip6_err(struct sk_buff *skb, u32 info)
        err = -ENOENT;
 
        read_lock(&ipip6_lock);
-       t = ipip6_tunnel_lookup(&init_net, iph->daddr, iph->saddr);
+       t = ipip6_tunnel_lookup(dev_net(skb->dev),
+                               skb->dev,
+                               iph->daddr,
+                               iph->saddr);
        if (t == NULL || t->parms.iph.daddr == 0)
                goto out;
 
@@ -452,7 +528,7 @@ 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;
@@ -460,91 +536,6 @@ static int ipip6_err(struct sk_buff *skb, u32 info)
 out:
        read_unlock(&ipip6_lock);
        return err;
-#else
-       struct iphdr *iph = (struct iphdr*)dp;
-       int hlen = iph->ihl<<2;
-       struct ipv6hdr *iph6;
-       const int type = icmp_hdr(skb)->type;
-       const int code = icmp_hdr(skb)->code;
-       int rel_type = 0;
-       int rel_code = 0;
-       int rel_info = 0;
-       struct sk_buff *skb2;
-       struct rt6_info *rt6i;
-
-       if (len < hlen + sizeof(struct ipv6hdr))
-               return;
-       iph6 = (struct ipv6hdr*)(dp + hlen);
-
-       switch (type) {
-       default:
-               return;
-       case ICMP_PARAMETERPROB:
-               if (icmp_hdr(skb)->un.gateway < hlen)
-                       return;
-
-               /* So... This guy found something strange INSIDE encapsulated
-                  packet. Well, he is fool, but what can we do ?
-                */
-               rel_type = ICMPV6_PARAMPROB;
-               rel_info = icmp_hdr(skb)->un.gateway - hlen;
-               break;
-
-       case ICMP_DEST_UNREACH:
-               switch (code) {
-               case ICMP_SR_FAILED:
-               case ICMP_PORT_UNREACH:
-                       /* Impossible event. */
-                       return;
-               case ICMP_FRAG_NEEDED:
-                       /* Too complicated case ... */
-                       return;
-               default:
-                       /* All others are translated to HOST_UNREACH.
-                          rfc2003 contains "deep thoughts" about NET_UNREACH,
-                          I believe, it is just ether pollution. --ANK
-                        */
-                       rel_type = ICMPV6_DEST_UNREACH;
-                       rel_code = ICMPV6_ADDR_UNREACH;
-                       break;
-               }
-               break;
-       case ICMP_TIME_EXCEEDED:
-               if (code != ICMP_EXC_TTL)
-                       return;
-               rel_type = ICMPV6_TIME_EXCEED;
-               rel_code = ICMPV6_EXC_HOPLIMIT;
-               break;
-       }
-
-       /* Prepare fake skb to feed it to icmpv6_send */
-       skb2 = skb_clone(skb, GFP_ATOMIC);
-       if (skb2 == NULL)
-               return 0;
-       dst_release(skb2->dst);
-       skb2->dst = NULL;
-       skb_pull(skb2, skb->data - (u8*)iph6);
-       skb_reset_network_header(skb2);
-
-       /* Try to guess incoming interface */
-       rt6i = rt6_lookup(&init_net, &iph6->saddr, NULL, NULL, 0);
-       if (rt6i && rt6i->rt6i_dev) {
-               skb2->dev = rt6i->rt6i_dev;
-
-               rt6i = rt6_lookup(&init_net, &iph6->daddr, &iph6->saddr, NULL, 0);
-
-               if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
-                       struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
-                       if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
-                               rel_type = ICMPV6_DEST_UNREACH;
-                               rel_code = ICMPV6_ADDR_UNREACH;
-                       }
-                       icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
-               }
-       }
-       kfree_skb(skb2);
-       return 0;
-#endif
 }
 
 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
@@ -564,8 +555,9 @@ static int ipip6_rcv(struct sk_buff *skb)
        iph = ip_hdr(skb);
 
        read_lock(&ipip6_lock);
-       if ((tunnel = ipip6_tunnel_lookup(&init_net,
-                                       iph->saddr, iph->daddr)) != NULL) {
+       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);
@@ -575,16 +567,15 @@ static int ipip6_rcv(struct sk_buff *skb)
 
                if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
                    !isatap_chksrc(skb, iph, tunnel)) {
-                       tunnel->stat.rx_errors++;
+                       tunnel->dev->stats.rx_errors++;
                        read_unlock(&ipip6_lock);
                        kfree_skb(skb);
                        return 0;
                }
-               tunnel->stat.rx_packets++;
-               tunnel->stat.rx_bytes += skb->len;
+               tunnel->dev->stats.rx_packets++;
+               tunnel->dev->stats.rx_bytes += skb->len;
                skb->dev = tunnel->dev;
-               dst_release(skb->dst);
-               skb->dst = NULL;
+               skb_dst_drop(skb);
                nf_reset(skb);
                ipip6_ecn_decapsulate(iph, skb);
                netif_rx(skb);
@@ -593,9 +584,9 @@ static int ipip6_rcv(struct sk_buff *skb)
        }
 
        icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
-       kfree_skb(skb);
        read_unlock(&ipip6_lock);
 out:
+       kfree_skb(skb);
        return 0;
 }
 
@@ -621,7 +612,7 @@ static inline __be32 try_6to4(struct in6_addr *v6dst)
 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 {
        struct ip_tunnel *tunnel = netdev_priv(dev);
-       struct net_device_stats *stats = &tunnel->stat;
+       struct net_device_stats *stats = &tunnel->dev->stats;
        struct iphdr  *tiph = &tunnel->parms.iph;
        struct ipv6hdr *iph6 = ipv6_hdr(skb);
        u8     tos = tunnel->parms.iph.tos;
@@ -635,7 +626,7 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        int addr_type;
 
        if (tunnel->recursion++) {
-               tunnel->stat.collisions++;
+               stats->collisions++;
                goto tx_error;
        }
 
@@ -646,8 +637,8 @@ 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())
@@ -671,8 +662,8 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        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())
@@ -701,38 +692,38 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
                                                .tos = RT_TOS(tos) } },
                                    .oif = tunnel->parms.link,
                                    .proto = IPPROTO_IPV6 };
-               if (ip_route_output_key(&init_net, &rt, &fl)) {
-                       tunnel->stat.tx_carrier_errors++;
+               if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
+                       stats->tx_carrier_errors++;
                        goto tx_error_icmp;
                }
        }
        if (rt->rt_type != RTN_UNICAST) {
                ip_rt_put(rt);
-               tunnel->stat.tx_carrier_errors++;
+               stats->tx_carrier_errors++;
                goto tx_error_icmp;
        }
        tdev = rt->u.dst.dev;
 
        if (tdev == dev) {
                ip_rt_put(rt);
-               tunnel->stat.collisions++;
+               stats->collisions++;
                goto tx_error;
        }
 
        if (tiph->frag_off)
                mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
        else
-               mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
+               mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
 
        if (mtu < 68) {
-               tunnel->stat.collisions++;
+               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 (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, dev);
@@ -741,7 +732,8 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
        }
 
        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
@@ -761,7 +753,7 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
                        stats->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);
@@ -775,8 +767,8 @@ 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.
@@ -802,7 +794,7 @@ static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 
        IPTUNNEL_XMIT();
        tunnel->recursion--;
-       return 0;
+       return NETDEV_TX_OK;
 
 tx_error_icmp:
        dst_link_failure(skb);
@@ -810,7 +802,7 @@ 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)
@@ -830,7 +822,7 @@ static void ipip6_tunnel_bind_dev(struct net_device *dev)
                                    .oif = tunnel->parms.link,
                                    .proto = IPPROTO_IPV6 };
                struct rtable *rt;
-               if (!ip_route_output_key(&init_net, &rt, &fl)) {
+               if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
                        tdev = rt->u.dst.dev;
                        ip_rt_put(rt);
                }
@@ -838,7 +830,7 @@ static void ipip6_tunnel_bind_dev(struct net_device *dev)
        }
 
        if (!tdev && tunnel->parms.link)
-               tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
+               tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
 
        if (tdev) {
                dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
@@ -862,7 +854,7 @@ ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
        switch (cmd) {
        case SIOCGETTUNNEL:
                t = NULL;
-               if (dev == ipip6_fb_tunnel_dev) {
+               if (dev == sitn->fb_tunnel_dev) {
                        if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
                                err = -EFAULT;
                                break;
@@ -895,7 +887,7 @@ ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 
                t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
 
-               if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
+               if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
                        if (t != NULL) {
                                if (t->dev != dev) {
                                        err = -EEXIST;
@@ -940,7 +932,7 @@ ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
                if (!capable(CAP_NET_ADMIN))
                        goto done;
 
-               if (dev == ipip6_fb_tunnel_dev) {
+               if (dev == sitn->fb_tunnel_dev) {
                        err = -EFAULT;
                        if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
                                goto done;
@@ -948,7 +940,7 @@ ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
                        if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
                                goto done;
                        err = -EPERM;
-                       if (t == netdev_priv(ipip6_fb_tunnel_dev))
+                       if (t == netdev_priv(sitn->fb_tunnel_dev))
                                goto done;
                        dev = t->dev;
                }
@@ -957,14 +949,23 @@ ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
                break;
 
        case SIOCGETPRL:
+               err = -EINVAL;
+               if (dev == sitn->fb_tunnel_dev)
+                       goto done;
+               err = -ENOENT;
+               if (!(t = netdev_priv(dev)))
+                       goto done;
+               err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
+               break;
+
        case SIOCADDPRL:
        case SIOCDELPRL:
        case SIOCCHGPRL:
                err = -EPERM;
-               if (cmd != SIOCGETPRL && !capable(CAP_NET_ADMIN))
+               if (!capable(CAP_NET_ADMIN))
                        goto done;
                err = -EINVAL;
-               if (dev == ipip6_fb_tunnel_dev)
+               if (dev == sitn->fb_tunnel_dev)
                        goto done;
                err = -EFAULT;
                if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
@@ -974,12 +975,6 @@ ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
                        goto done;
 
                switch (cmd) {
-               case SIOCGETPRL:
-                       err = ipip6_tunnel_get_prl(t, &prl);
-                       if (!err && copy_to_user(ifr->ifr_ifru.ifru_data,
-                                                &prl, sizeof(prl)))
-                               err = -EFAULT;
-                       break;
                case SIOCDELPRL:
                        err = ipip6_tunnel_del_prl(t, &prl);
                        break;
@@ -988,8 +983,7 @@ ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
                        err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
                        break;
                }
-               if (cmd != SIOCGETPRL)
-                       netdev_state_change(dev);
+               netdev_state_change(dev);
                break;
 
        default:
@@ -1000,11 +994,6 @@ done:
        return err;
 }
 
-static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
-{
-       return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
-}
-
 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 {
        if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
@@ -1013,28 +1002,31 @@ 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->get_stats          = ipip6_tunnel_get_stats;
-       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);
@@ -1043,14 +1035,14 @@ 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 __init ipip6_fb_tunnel_init(struct net_device *dev)
+static void ipip6_fb_tunnel_init(struct net_device *dev)
 {
        struct ip_tunnel *tunnel = netdev_priv(dev);
        struct iphdr *iph = &tunnel->parms.iph;
+       struct net *net = dev_net(dev);
+       struct sit_net *sitn = net_generic(net, sit_net_id);
 
        tunnel->dev = dev;
        strcpy(tunnel->parms.name, dev->name);
@@ -1061,8 +1053,7 @@ static int __init ipip6_fb_tunnel_init(struct net_device *dev)
        iph->ttl                = 64;
 
        dev_hold(dev);
-       tunnels_wc[0]           = tunnel;
-       return 0;
+       sitn->tunnels_wc[0]     = tunnel;
 }
 
 static struct xfrm_tunnel sit_handler = {
@@ -1071,7 +1062,7 @@ static struct xfrm_tunnel sit_handler = {
        .priority       =       1,
 };
 
-static void __exit sit_destroy_tunnels(void)
+static void sit_destroy_tunnels(struct sit_net *sitn)
 {
        int prio;
 
@@ -1079,7 +1070,7 @@ static void __exit sit_destroy_tunnels(void)
                int h;
                for (h = 0; h < HASH_SIZE; h++) {
                        struct ip_tunnel *t;
-                       while ((t = tunnels[prio][h]) != NULL)
+                       while ((t = sitn->tunnels[prio][h]) != NULL)
                                unregister_netdevice(t->dev);
                }
        }
@@ -1091,7 +1082,7 @@ static int sit_init_net(struct net *net)
        struct sit_net *sitn;
 
        err = -ENOMEM;
-       sitn = kmalloc(sizeof(struct sit_net), GFP_KERNEL);
+       sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
        if (sitn == NULL)
                goto err_alloc;
 
@@ -1099,8 +1090,31 @@ static int sit_init_net(struct net *net)
        if (err < 0)
                goto err_assign;
 
+       sitn->tunnels[0] = sitn->tunnels_wc;
+       sitn->tunnels[1] = sitn->tunnels_l;
+       sitn->tunnels[2] = sitn->tunnels_r;
+       sitn->tunnels[3] = sitn->tunnels_r_l;
+
+       sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
+                                          ipip6_tunnel_setup);
+       if (!sitn->fb_tunnel_dev) {
+               err = -ENOMEM;
+               goto err_alloc_dev;
+       }
+       dev_net_set(sitn->fb_tunnel_dev, net);
+
+       ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
+
+       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:
@@ -1112,6 +1126,10 @@ static void sit_exit_net(struct net *net)
        struct sit_net *sitn;
 
        sitn = net_generic(net, sit_net_id);
+       rtnl_lock();
+       sit_destroy_tunnels(sitn);
+       unregister_netdevice(sitn->fb_tunnel_dev);
+       rtnl_unlock();
        kfree(sitn);
 }
 
@@ -1124,11 +1142,6 @@ static void __exit sit_cleanup(void)
 {
        xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 
-       rtnl_lock();
-       sit_destroy_tunnels();
-       unregister_netdevice(ipip6_fb_tunnel_dev);
-       rtnl_unlock();
-
        unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
 }
 
@@ -1143,32 +1156,11 @@ static int __init sit_init(void)
                return -EAGAIN;
        }
 
-       ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
-                                          ipip6_tunnel_setup);
-       if (!ipip6_fb_tunnel_dev) {
-               err = -ENOMEM;
-               goto err1;
-       }
-
-       ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
-
-       if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
-               goto err2;
-
        err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
        if (err < 0)
-               goto err3;
+               xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
 
- out:
        return err;
- err2:
-       free_netdev(ipip6_fb_tunnel_dev);
- err1:
-       xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
-       goto out;
-err3:
-       unregister_netdevice(ipip6_fb_tunnel_dev);
-       goto err1;
 }
 
 module_init(sit_init);