fedora core 6 1.2949 + vserver 2.2.0
[linux-2.6.git] / net / ipv6 / ip6_tunnel.c
index e815380..8d91834 100644 (file)
@@ -19,8 +19,8 @@
  *
  */
 
-#include <linux/config.h>
 #include <linux/module.h>
+#include <linux/capability.h>
 #include <linux/errno.h>
 #include <linux/types.h>
 #include <linux/sockios.h>
 
 #include <net/ip.h>
 #include <net/ipv6.h>
-#include <net/protocol.h>
 #include <net/ip6_route.h>
 #include <net/addrconf.h>
 #include <net/ip6_tunnel.h>
 #include <net/xfrm.h>
+#include <net/dsfield.h>
+#include <net/inet_ecn.h>
 
 MODULE_AUTHOR("Ville Nuorvala");
 MODULE_DESCRIPTION("IPv6-in-IPv6 tunnel");
@@ -65,7 +66,7 @@ MODULE_LICENSE("GPL");
 
 #define HASH_SIZE  32
 
-#define HASH(addr) (((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
+#define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
                     (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
                     (HASH_SIZE - 1))
 
@@ -83,7 +84,7 @@ static struct ip6_tnl *tnls_wc[1];
 static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l };
 
 /* lock for the tunnel lists */
-static rwlock_t ip6ip6_lock = RW_LOCK_UNLOCKED;
+static DEFINE_RWLOCK(ip6ip6_lock);
 
 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
 {
@@ -92,6 +93,7 @@ static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
        if (dst && dst->obsolete && 
            dst->ops->check(dst, t->dst_cookie) == NULL) {
                t->dst_cache = NULL;
+               dst_release(dst);
                return NULL;
        }
 
@@ -123,7 +125,7 @@ static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
  *   else %NULL
  **/
 
-struct ip6_tnl *
+static struct ip6_tnl *
 ip6ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
 {
        unsigned h0 = HASH(remote);
@@ -131,8 +133,8 @@ ip6ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
        struct ip6_tnl *t;
 
        for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) {
-               if (!ipv6_addr_cmp(local, &t->parms.laddr) &&
-                   !ipv6_addr_cmp(remote, &t->parms.raddr) &&
+               if (ipv6_addr_equal(local, &t->parms.laddr) &&
+                   ipv6_addr_equal(remote, &t->parms.raddr) &&
                    (t->dev->flags & IFF_UP))
                        return t;
        }
@@ -178,10 +180,10 @@ ip6ip6_tnl_link(struct ip6_tnl *t)
 {
        struct ip6_tnl **tp = ip6ip6_bucket(&t->parms);
 
-       write_lock_bh(&ip6ip6_lock);
        t->next = *tp;
-       write_unlock_bh(&ip6ip6_lock);
+       write_lock_bh(&ip6ip6_lock);
        *tp = t;
+       write_unlock_bh(&ip6ip6_lock);
 }
 
 /**
@@ -213,11 +215,10 @@ ip6ip6_tnl_unlink(struct ip6_tnl *t)
  *   Create tunnel matching given parameters.
  * 
  * Return: 
- *   0 on success
+ *   created tunnel or NULL
  **/
 
-static int
-ip6_tnl_create(struct ip6_tnl_parm *p, struct ip6_tnl **pt)
+static struct ip6_tnl *ip6_tnl_create(struct ip6_tnl_parm *p)
 {
        struct net_device *dev;
        struct ip6_tnl *t;
@@ -234,25 +235,25 @@ ip6_tnl_create(struct ip6_tnl_parm *p, struct ip6_tnl **pt)
                                break;
                }
                if (i == IP6_TNL_MAX) 
-                       return -ENOBUFS;
+                       goto failed;
        }
        dev = alloc_netdev(sizeof (*t), name, ip6ip6_tnl_dev_setup);
        if (dev == NULL)
-               return -ENOMEM;
+               goto failed;
 
-       t = dev->priv;
+       t = netdev_priv(dev);
        dev->init = ip6ip6_tnl_dev_init;
        t->parms = *p;
 
        if ((err = register_netdevice(dev)) < 0) {
                free_netdev(dev);
-               return err;
+               goto failed;
        }
        dev_hold(dev);
-
        ip6ip6_tnl_link(t);
-       *pt = t;
-       return 0;
+       return t;
+failed:
+       return NULL;
 }
 
 /**
@@ -266,32 +267,23 @@ ip6_tnl_create(struct ip6_tnl_parm *p, struct ip6_tnl **pt)
  *   tunnel device is created and registered for use.
  *
  * Return:
- *   0 if tunnel located or created,
- *   -EINVAL if parameters incorrect,
- *   -ENODEV if no matching tunnel available
+ *   matching tunnel or NULL
  **/
 
-static int
-ip6ip6_tnl_locate(struct ip6_tnl_parm *p, struct ip6_tnl **pt, int create)
+static struct ip6_tnl *ip6ip6_tnl_locate(struct ip6_tnl_parm *p, int create)
 {
        struct in6_addr *remote = &p->raddr;
        struct in6_addr *local = &p->laddr;
        struct ip6_tnl *t;
 
-       if (p->proto != IPPROTO_IPV6)
-               return -EINVAL;
-
        for (t = *ip6ip6_bucket(p); t; t = t->next) {
-               if (!ipv6_addr_cmp(local, &t->parms.laddr) &&
-                   !ipv6_addr_cmp(remote, &t->parms.raddr)) {
-                       *pt = t;
-                       return (create ? -EEXIST : 0);
-               }
+               if (ipv6_addr_equal(local, &t->parms.laddr) &&
+                   ipv6_addr_equal(remote, &t->parms.raddr))
+                       return t;
        }
        if (!create)
-               return -ENODEV;
-       
-       return ip6_tnl_create(p, pt);
+               return NULL;
+       return ip6_tnl_create(p);
 }
 
 /**
@@ -305,7 +297,7 @@ ip6ip6_tnl_locate(struct ip6_tnl_parm *p, struct ip6_tnl **pt, int create)
 static void
 ip6ip6_tnl_dev_uninit(struct net_device *dev)
 {
-       struct ip6_tnl *t = dev->priv;
+       struct ip6_tnl *t = netdev_priv(dev);
 
        if (dev == ip6ip6_fb_tnl_dev) {
                write_lock_bh(&ip6ip6_lock);
@@ -387,8 +379,9 @@ parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
  *   to the specifications in RFC 2473.
  **/
 
-void ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
-                  int type, int code, int offset, __u32 info)
+static int
+ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
+          int type, int code, int offset, __be32 info)
 {
        struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
        struct ip6_tnl *t;
@@ -397,6 +390,7 @@ void ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
        int rel_code = ICMPV6_ADDR_UNREACH;
        __u32 rel_info = 0;
        __u16 len;
+       int err = -ENOENT;
 
        /* If the packet doesn't contain the original IPv6 header we are 
           in trouble since we might need the source address for further 
@@ -406,6 +400,8 @@ void ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
        if ((t = ip6ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
                goto out;
 
+       err = 0;
+
        switch (type) {
                __u32 teli;
                struct ipv6_tlv_tnl_enc_lim *tel;
@@ -428,12 +424,9 @@ void ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
                }
                break;
        case ICMPV6_PARAMPROB:
-               /* ignore if parameter problem not caused by a tunnel
-                  encapsulation limit sub-option */
-               if (code != ICMPV6_HDR_FIELD) {
-                       break;
-               }
-               teli = parse_tlv_tnl_enc_lim(skb, skb->data);
+               teli = 0;
+               if (code == ICMPV6_HDR_FIELD)
+                       teli = parse_tlv_tnl_enc_lim(skb, skb->data);
 
                if (teli && teli == ntohl(info) - 2) {
                        tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
@@ -445,6 +438,10 @@ void ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
                                               "tunnel!\n", t->parms.name);
                                rel_msg = 1;
                        }
+               } else if (net_ratelimit()) {
+                       printk(KERN_WARNING
+                              "%s: Recipient unable to parse tunneled "
+                              "packet!\n ", t->parms.name);
                }
                break;
        case ICMPV6_PKT_TOOBIG:
@@ -453,7 +450,7 @@ void ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
                        mtu = IPV6_MIN_MTU;
                t->dev->mtu = mtu;
 
-               if ((len = sizeof (*ipv6h) + ipv6h->payload_len) > mtu) {
+               if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
                        rel_type = ICMPV6_PKT_TOOBIG;
                        rel_code = 0;
                        rel_info = mtu;
@@ -464,6 +461,7 @@ void ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
        if (rel_msg &&  pskb_may_pull(skb, offset + sizeof (*ipv6h))) {
                struct rt6_info *rt;
                struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
+
                if (!skb2)
                        goto out;
 
@@ -487,6 +485,37 @@ void ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
        }
 out:
        read_unlock(&ip6ip6_lock);
+       return err;
+}
+
+static inline void ip6ip6_ecn_decapsulate(struct ipv6hdr *outer_iph,
+                                         struct sk_buff *skb)
+{
+       struct ipv6hdr *inner_iph = skb->nh.ipv6h;
+
+       if (INET_ECN_is_ce(ipv6_get_dsfield(outer_iph)))
+               IP6_ECN_set_ce(inner_iph);
+}
+static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
+{
+       struct ip6_tnl_parm *p = &t->parms;
+       int ret = 0;
+
+       if (p->flags & IP6_TNL_F_CAP_RCV) {
+               struct net_device *ldev = NULL;
+
+               if (p->link)
+                       ldev = dev_get_by_index(p->link);
+
+               if ((ipv6_addr_is_multicast(&p->laddr) ||
+                    likely(ipv6_chk_addr(&p->laddr, ldev, 0))) &&
+                   likely(!ipv6_chk_addr(&p->raddr, NULL, 0)))
+                       ret = 1;
+
+               if (ldev)
+                       dev_put(ldev);
+       }
+       return ret;
 }
 
 /**
@@ -496,21 +525,23 @@ out:
  * Return: 0
  **/
 
-int ip6ip6_rcv(struct sk_buff **pskb, unsigned int *nhoffp)
+static int 
+ip6ip6_rcv(struct sk_buff *skb)
 {
-       struct sk_buff *skb = *pskb;
        struct ipv6hdr *ipv6h;
        struct ip6_tnl *t;
 
-       if (!pskb_may_pull(skb, sizeof (*ipv6h)))
-               goto discard;
-
        ipv6h = skb->nh.ipv6h;
 
        read_lock(&ip6ip6_lock);
 
        if ((t = ip6ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
-               if (!(t->parms.flags & IP6_TNL_F_CAP_RCV)) {
+               if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
+                       read_unlock(&ip6ip6_lock);
+                       goto discard;
+               }
+
+               if (!ip6_tnl_rcv_ctl(t)) {
                        t->stat.rx_dropped++;
                        read_unlock(&ip6ip6_lock);
                        goto discard;
@@ -524,6 +555,10 @@ int ip6ip6_rcv(struct sk_buff **pskb, unsigned int *nhoffp)
                skb->dev = t->dev;
                dst_release(skb->dst);
                skb->dst = NULL;
+               nf_reset(skb);
+               if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
+                       ipv6_copy_dscp(ipv6h, skb->nh.ipv6h);
+               ip6ip6_ecn_decapsulate(ipv6h, skb);
                t->stat.rx_packets++;
                t->stat.rx_bytes += skb->len;
                netif_rx(skb);
@@ -531,38 +566,30 @@ int ip6ip6_rcv(struct sk_buff **pskb, unsigned int *nhoffp)
                return 0;
        }
        read_unlock(&ip6ip6_lock);
-       icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
+       return 1;
+
 discard:
        kfree_skb(skb);
        return 0;
 }
 
-static inline struct ipv6_txoptions *create_tel(__u8 encap_limit)
-{
-       struct ipv6_tlv_tnl_enc_lim *tel;
-       struct ipv6_txoptions *opt;
-       __u8 *raw;
-
-       int opt_len = sizeof(*opt) + 8;
-
-       if (!(opt = kmalloc(opt_len, GFP_ATOMIC))) {
-               return NULL;
-       }
-       memset(opt, 0, opt_len);
-       opt->tot_len = opt_len;
-       opt->dst0opt = (struct ipv6_opt_hdr *) (opt + 1);
-       opt->opt_nflen = 8;
+struct ipv6_tel_txoption {
+       struct ipv6_txoptions ops;
+       __u8 dst_opt[8];
+};
 
-       tel = (struct ipv6_tlv_tnl_enc_lim *) (opt->dst0opt + 1);
-       tel->type = IPV6_TLV_TNL_ENCAP_LIMIT;
-       tel->length = 1;
-       tel->encap_limit = encap_limit;
+static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
+{
+       memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 
-       raw = (__u8 *) opt->dst0opt;
-       raw[5] = IPV6_TLV_PADN;
-       raw[6] = 1;
+       opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
+       opt->dst_opt[3] = 1;
+       opt->dst_opt[4] = encap_limit;
+       opt->dst_opt[5] = IPV6_TLV_PADN;
+       opt->dst_opt[6] = 1;
 
-       return opt;
+       opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
+       opt->ops.opt_nflen = 8;
 }
 
 /**
@@ -582,9 +609,37 @@ static inline struct ipv6_txoptions *create_tel(__u8 encap_limit)
 static inline int
 ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
 {
-       return !ipv6_addr_cmp(&t->parms.raddr, &hdr->saddr);
+       return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 }
 
+static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
+{
+       struct ip6_tnl_parm *p = &t->parms;
+       int ret = 0;
+
+       if (p->flags & IP6_TNL_F_CAP_XMIT) {
+               struct net_device *ldev = NULL;
+
+               if (p->link)
+                       ldev = dev_get_by_index(p->link);
+
+               if (unlikely(!ipv6_chk_addr(&p->laddr, ldev, 0)))
+                       printk(KERN_WARNING
+                              "%s xmit: Local address not yet configured!\n",
+                              p->name);
+               else if (!ipv6_addr_is_multicast(&p->raddr) &&
+                        unlikely(ipv6_chk_addr(&p->raddr, NULL, 0)))
+                       printk(KERN_WARNING
+                              "%s xmit: Routing loop! "
+                              "Remote address found on this node!\n",
+                              p->name);
+               else
+                       ret = 1;
+               if (ldev)
+                       dev_put(ldev);
+       }
+       return ret;
+}
 /**
  * ip6ip6_tnl_xmit - encapsulate packet and send 
  *   @skb: the outgoing socket buffer
@@ -598,13 +653,14 @@ ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
  *   0
  **/
 
-int ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
+static int 
+ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 {
-       struct ip6_tnl *t = (struct ip6_tnl *) dev->priv;
+       struct ip6_tnl *t = netdev_priv(dev);
        struct net_device_stats *stats = &t->stat;
        struct ipv6hdr *ipv6h = skb->nh.ipv6h;
-       struct ipv6_txoptions *opt = NULL;
        int encap_limit = -1;
+       struct ipv6_tel_txoption opt;
        __u16 offset;
        struct flowi fl;
        struct dst_entry *dst;
@@ -614,16 +670,16 @@ int ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
        u8 proto;
        int err;
        int pkt_len;
+       int dsfield;
 
        if (t->recursion++) {
                stats->collisions++;
                goto tx_err;
        }
        if (skb->protocol != htons(ETH_P_IPV6) ||
-           !(t->parms.flags & IP6_TNL_F_CAP_XMIT) ||
-           ip6ip6_tnl_addr_conflict(t, ipv6h)) {
+           !ip6_tnl_xmit_ctl(t) || ip6ip6_tnl_addr_conflict(t, ipv6h))
                goto tx_err;
-       }
+
        if ((offset = parse_tlv_tnl_enc_lim(skb, skb->nh.raw)) > 0) {
                struct ipv6_tlv_tnl_enc_lim *tel;
                tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->nh.raw[offset];
@@ -633,27 +689,26 @@ int ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
                        goto tx_err;
                }
                encap_limit = tel->encap_limit - 1;
-       } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
+       } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
                encap_limit = t->parms.encap_limit;
-       }
+
        memcpy(&fl, &t->fl, sizeof (fl));
        proto = fl.proto;
 
+       dsfield = ipv6_get_dsfield(ipv6h);
        if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
-               fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_TCLASS_MASK);
+               fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
        if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
-               fl.fl6_flowlabel |= (*(__u32 *) ipv6h & IPV6_FLOWLABEL_MASK);
-
-       if (encap_limit >= 0 && (opt = create_tel(encap_limit)) == NULL)
-               goto tx_err;
+               fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
 
        if ((dst = ip6_tnl_dst_check(t)) != NULL)
                dst_hold(dst);
-       else
+       else {
                dst = ip6_route_output(NULL, &fl);
 
-       if (dst->error || xfrm_lookup(&dst, &fl, NULL, 0) < 0)
-               goto tx_err_link_failure;
+               if (dst->error || xfrm_lookup(&dst, &fl, NULL, 0) < 0)
+                       goto tx_err_link_failure;
+       }
 
        tdev = dst->dev;
 
@@ -665,14 +720,14 @@ int ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
                               t->parms.name);
                goto tx_err_dst_release;
        }
-       mtu = dst_pmtu(dst) - sizeof (*ipv6h);
-       if (opt) {
+       mtu = dst_mtu(dst) - sizeof (*ipv6h);
+       if (encap_limit >= 0) {
                max_headroom += 8;
                mtu -= 8;
        }
        if (mtu < IPV6_MIN_MTU)
                mtu = IPV6_MIN_MTU;
-       if (skb->dst && mtu < dst_pmtu(skb->dst)) {
+       if (skb->dst && mtu < dst_mtu(skb->dst)) {
                struct rt6_info *rt = (struct rt6_info *) skb->dst;
                rt->rt6i_flags |= RTF_MODIFIED;
                rt->u.dst.metrics[RTAX_MTU-1] = mtu;
@@ -704,29 +759,26 @@ int ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 
        skb->h.raw = skb->nh.raw;
 
-       if (opt)
-               ipv6_push_nfrag_opts(skb, opt, &proto, NULL);
-
+       if (encap_limit >= 0) {
+               init_tel_txopt(&opt, encap_limit);
+               ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
+       }
        skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr));
        ipv6h = skb->nh.ipv6h;
-       *(u32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000);
+       *(__be32*)ipv6h = fl.fl6_flowlabel | htonl(0x60000000);
+       dsfield = INET_ECN_encapsulate(0, dsfield);
+       ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
        ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
        ipv6h->hop_limit = t->parms.hop_limit;
        ipv6h->nexthdr = proto;
        ipv6_addr_copy(&ipv6h->saddr, &fl.fl6_src);
        ipv6_addr_copy(&ipv6h->daddr, &fl.fl6_dst);
-#ifdef CONFIG_NETFILTER
-       nf_conntrack_put(skb->nfct);
-       skb->nfct = NULL;
-#ifdef CONFIG_NETFILTER_DEBUG
-       skb->nf_debug = 0;
-#endif
-#endif
+       nf_reset(skb);
        pkt_len = skb->len;
        err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, 
                      skb->dst->dev, dst_output);
 
-       if (err == NET_XMIT_SUCCESS || err == NET_XMIT_CN) {
+       if (net_xmit_eval(err) == 0) {
                stats->tx_bytes += pkt_len;
                stats->tx_packets++;
        } else {
@@ -734,10 +786,6 @@ int ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
                stats->tx_aborted_errors++;
        }
        ip6_tnl_dst_store(t, dst);
-
-       if (opt)
-               kfree(opt);
-
        t->recursion--;
        return 0;
 tx_err_link_failure:
@@ -745,8 +793,6 @@ tx_err_link_failure:
        dst_link_failure(skb);
 tx_err_dst_release:
        dst_release(dst);
-       if (opt)
-               kfree(opt);
 tx_err:
        stats->tx_errors++;
        stats->tx_dropped++;
@@ -758,39 +804,19 @@ tx_err:
 static void ip6_tnl_set_cap(struct ip6_tnl *t)
 {
        struct ip6_tnl_parm *p = &t->parms;
-       struct in6_addr *laddr = &p->laddr;
-       struct in6_addr *raddr = &p->raddr;
-       int ltype = ipv6_addr_type(laddr);
-       int rtype = ipv6_addr_type(raddr);
+       int ltype = ipv6_addr_type(&p->laddr);
+       int rtype = ipv6_addr_type(&p->raddr);
 
        p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
 
-       if (ltype != IPV6_ADDR_ANY && rtype != IPV6_ADDR_ANY &&
-           ((ltype|rtype) &
-            (IPV6_ADDR_UNICAST|
-             IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL|
-             IPV6_ADDR_MAPPED|IPV6_ADDR_RESERVED)) == IPV6_ADDR_UNICAST) {
-               struct net_device *ldev = NULL;
-               int l_ok = 1;
-               int r_ok = 1;
-
-               if (p->link)
-                       ldev = dev_get_by_index(p->link);
-               
-               if (ltype&IPV6_ADDR_UNICAST && !ipv6_chk_addr(laddr, ldev, 0))
-                       l_ok = 0;
-               
-               if (rtype&IPV6_ADDR_UNICAST && ipv6_chk_addr(raddr, NULL, 0))
-                       r_ok = 0;
-               
-               if (l_ok && r_ok) {
-                       if (ltype&IPV6_ADDR_UNICAST)
-                               p->flags |= IP6_TNL_F_CAP_XMIT;
-                       if (rtype&IPV6_ADDR_UNICAST)
-                               p->flags |= IP6_TNL_F_CAP_RCV;
-               }
-               if (ldev)
-                       dev_put(ldev);
+       if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
+           rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
+           !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
+           (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
+               if (ltype&IPV6_ADDR_UNICAST)
+                       p->flags |= IP6_TNL_F_CAP_XMIT;
+               if (rtype&IPV6_ADDR_UNICAST)
+                       p->flags |= IP6_TNL_F_CAP_RCV;
        }
 }
 
@@ -824,8 +850,11 @@ static void ip6ip6_tnl_link_config(struct ip6_tnl *t)
        dev->iflink = p->link;
 
        if (p->flags & IP6_TNL_F_CAP_XMIT) {
+               int strict = (ipv6_addr_type(&p->raddr) &
+                             (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
+
                struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
-                                                p->link, 0);
+                                                p->link, strict);
 
                if (rt == NULL)
                        return;
@@ -862,6 +891,8 @@ ip6ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
        t->parms.hop_limit = p->hop_limit;
        t->parms.encap_limit = p->encap_limit;
        t->parms.flowinfo = p->flowinfo;
+       t->parms.link = p->link;
+       ip6_tnl_dst_reset(t);
        ip6ip6_tnl_link_config(t);
        return 0;
 }
@@ -898,26 +929,20 @@ static int
 ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 {
        int err = 0;
-       int create;
        struct ip6_tnl_parm p;
        struct ip6_tnl *t = NULL;
 
        switch (cmd) {
        case SIOCGETTUNNEL:
                if (dev == ip6ip6_fb_tnl_dev) {
-                       if (copy_from_user(&p,
-                                          ifr->ifr_ifru.ifru_data,
-                                          sizeof (p))) {
+                       if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
                                err = -EFAULT;
                                break;
                        }
-                       if ((err = ip6ip6_tnl_locate(&p, &t, 0)) == -ENODEV)
-                               t = (struct ip6_tnl *) dev->priv;
-                       else if (err)
-                               break;
-               } else
-                       t = (struct ip6_tnl *) dev->priv;
-
+                       t = ip6ip6_tnl_locate(&p, 0);
+               }
+               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;
@@ -926,35 +951,36 @@ ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
        case SIOCADDTUNNEL:
        case SIOCCHGTUNNEL:
                err = -EPERM;
-               create = (cmd == SIOCADDTUNNEL);
                if (!capable(CAP_NET_ADMIN))
                        break;
-               if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
-                       err = -EFAULT;
+               err = -EFAULT;
+               if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
                        break;
-               }
-               if (!create && dev != ip6ip6_fb_tnl_dev) {
-                       t = (struct ip6_tnl *) dev->priv;
-               }
-               if (!t && (err = ip6ip6_tnl_locate(&p, &t, create))) {
+               err = -EINVAL;
+               if (p.proto != IPPROTO_IPV6)
                        break;
-               }
-               if (cmd == SIOCCHGTUNNEL) {
-                       if (t->dev != dev) {
-                               err = -EEXIST;
-                               break;
-                       }
+               t = ip6ip6_tnl_locate(&p, cmd == SIOCADDTUNNEL);
+               if (dev != ip6ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
+                       if (t != NULL) {
+                               if (t->dev != dev) {
+                                       err = -EEXIST;
+                                       break;
+                               }
+                       } else
+                               t = netdev_priv(dev);
+
                        ip6ip6_tnl_unlink(t);
                        err = ip6ip6_tnl_change(t, &p);
                        ip6ip6_tnl_link(t);
                        netdev_state_change(dev);
                }
-               if (copy_to_user(ifr->ifr_ifru.ifru_data,
-                                &t->parms, sizeof (p))) {
-                       err = -EFAULT;
-               } else {
+               if (t) {
                        err = 0;
-               }
+                       if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
+                               err = -EFAULT;
+
+               } else
+                       err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
                break;
        case SIOCDELTUNNEL:
                err = -EPERM;
@@ -962,22 +988,18 @@ ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
                        break;
 
                if (dev == ip6ip6_fb_tnl_dev) {
-                       if (copy_from_user(&p, ifr->ifr_ifru.ifru_data,
-                                          sizeof (p))) {
-                               err = -EFAULT;
+                       err = -EFAULT;
+                       if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
                                break;
-                       }
-                       err = ip6ip6_tnl_locate(&p, &t, 0);
-                       if (err)
+                       err = -ENOENT;
+                       if ((t = ip6ip6_tnl_locate(&p, 0)) == NULL)
                                break;
-                       if (t == ip6ip6_fb_tnl_dev->priv) {
-                               err = -EPERM;
+                       err = -EPERM;
+                       if (t->dev == ip6ip6_fb_tnl_dev)
                                break;
-                       }
-               } else {
-                       t = (struct ip6_tnl *) dev->priv;
+                       dev = t->dev;
                }
-               err = unregister_netdevice(t->dev);
+               err = unregister_netdevice(dev);
                break;
        default:
                err = -EINVAL;
@@ -995,7 +1017,7 @@ ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 static struct net_device_stats *
 ip6ip6_tnl_get_stats(struct net_device *dev)
 {
-       return &(((struct ip6_tnl *) dev->priv)->stat);
+       return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
 }
 
 /**
@@ -1052,7 +1074,7 @@ static void ip6ip6_tnl_dev_setup(struct net_device *dev)
 static inline void
 ip6ip6_tnl_dev_init_gen(struct net_device *dev)
 {
-       struct ip6_tnl *t = (struct ip6_tnl *) dev->priv;
+       struct ip6_tnl *t = netdev_priv(dev);
        t->fl.proto = IPPROTO_IPV6;
        t->dev = dev;
        strcpy(t->parms.name, dev->name);
@@ -1066,7 +1088,7 @@ ip6ip6_tnl_dev_init_gen(struct net_device *dev)
 static int
 ip6ip6_tnl_dev_init(struct net_device *dev)
 {
-       struct ip6_tnl *t = (struct ip6_tnl *) dev->priv;
+       struct ip6_tnl *t = netdev_priv(dev);
        ip6ip6_tnl_dev_init_gen(dev);
        ip6ip6_tnl_link_config(t);
        return 0;
@@ -1079,19 +1101,20 @@ ip6ip6_tnl_dev_init(struct net_device *dev)
  * Return: 0
  **/
 
-int ip6ip6_fb_tnl_dev_init(struct net_device *dev)
+static int 
+ip6ip6_fb_tnl_dev_init(struct net_device *dev)
 {
-       struct ip6_tnl *t = dev->priv;
+       struct ip6_tnl *t = netdev_priv(dev);
        ip6ip6_tnl_dev_init_gen(dev);
        dev_hold(dev);
        tnls_wc[0] = t;
        return 0;
 }
 
-static struct inet6_protocol ip6ip6_protocol = {
-       .handler = ip6ip6_rcv,
-       .err_handler = ip6ip6_err,
-       .flags = INET6_PROTO_FINAL
+static struct xfrm6_tunnel ip6ip6_handler = {
+       .handler        = ip6ip6_rcv,
+       .err_handler    = ip6ip6_err,
+       .priority       =       1,
 };
 
 /**
@@ -1104,9 +1127,9 @@ static int __init ip6_tunnel_init(void)
 {
        int  err;
 
-       if ((err = inet6_add_protocol(&ip6ip6_protocol, IPPROTO_IPV6)) < 0) {
-               printk(KERN_ERR "Failed to register IPv6 protocol\n");
-               return err;
+       if (xfrm6_tunnel_register(&ip6ip6_handler)) {
+               printk(KERN_ERR "ip6ip6 init: can't register tunnel\n");
+               return -EAGAIN;
        }
        ip6ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
                                         ip6ip6_tnl_dev_setup);
@@ -1123,18 +1146,36 @@ static int __init ip6_tunnel_init(void)
        }
        return 0;
 fail:
-       inet6_del_protocol(&ip6ip6_protocol, IPPROTO_IPV6);
+       xfrm6_tunnel_deregister(&ip6ip6_handler);
        return err;
 }
 
+static void __exit ip6ip6_destroy_tunnels(void)
+{
+       int h;
+       struct ip6_tnl *t;
+
+       for (h = 0; h < HASH_SIZE; h++) {
+               while ((t = tnls_r_l[h]) != NULL)
+                       unregister_netdevice(t->dev);
+       }
+
+       t = tnls_wc[0];
+       unregister_netdevice(t->dev);
+}
+
 /**
  * ip6_tunnel_cleanup - free resources and unregister protocol
  **/
 
 static void __exit ip6_tunnel_cleanup(void)
 {
-       unregister_netdev(ip6ip6_fb_tnl_dev);
-       inet6_del_protocol(&ip6ip6_protocol, IPPROTO_IPV6);
+       if (xfrm6_tunnel_deregister(&ip6ip6_handler))
+               printk(KERN_INFO "ip6ip6 close: can't deregister tunnel\n");
+
+       rtnl_lock();
+       ip6ip6_destroy_tunnels();
+       rtnl_unlock();
 }
 
 module_init(ip6_tunnel_init);