[RAW]: Introduce raw_hashinfo structure
[safe/jmp/linux-2.6] / net / ipv6 / route.c
index a9b08a2..d7ec4c9 100644 (file)
@@ -3,7 +3,7 @@
  *     FIB front-end.
  *
  *     Authors:
- *     Pedro Roque             <roque@di.fc.ul.pt>     
+ *     Pedro Roque             <roque@di.fc.ul.pt>
  *
  *     $Id: route.c,v 1.56 2001/10/31 21:55:55 davem Exp $
  *
@@ -22,6 +22,8 @@
  *             routers in REACHABLE, STALE, DELAY or PROBE states).
  *             - always select the same router if it is (probably)
  *             reachable.  otherwise, round-robin the list.
+ *     Ville Nuorvala
+ *             Fixed routing subtrees.
  */
 
 #include <linux/capability.h>
 #include <linux/in6.h>
 #include <linux/init.h>
 #include <linux/if_arp.h>
-
-#ifdef         CONFIG_PROC_FS
 #include <linux/proc_fs.h>
 #include <linux/seq_file.h>
-#endif
-
+#include <net/net_namespace.h>
 #include <net/snmp.h>
 #include <net/ipv6.h>
 #include <net/ip6_fib.h>
@@ -74,9 +73,6 @@
 
 #define CLONE_OFFLINK_ROUTE 0
 
-#define RT6_SELECT_F_IFACE     0x1
-#define RT6_SELECT_F_REACHABLE 0x2
-
 static int ip6_rt_max_size = 4096;
 static int ip6_rt_gc_min_interval = HZ / 2;
 static int ip6_rt_gc_timeout = 60*HZ;
@@ -117,6 +113,20 @@ static struct dst_ops ip6_dst_ops = {
        .negative_advice        =       ip6_negative_advice,
        .link_failure           =       ip6_link_failure,
        .update_pmtu            =       ip6_rt_update_pmtu,
+       .local_out              =       ip6_local_out,
+       .entry_size             =       sizeof(struct rt6_info),
+};
+
+static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
+{
+}
+
+static struct dst_ops ip6_dst_blackhole_ops = {
+       .family                 =       AF_INET6,
+       .protocol               =       __constant_htons(ETH_P_IPV6),
+       .destroy                =       ip6_dst_destroy,
+       .check                  =       ip6_dst_check,
+       .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
        .entry_size             =       sizeof(struct rt6_info),
 };
 
@@ -125,7 +135,6 @@ struct rt6_info ip6_null_entry = {
                .dst = {
                        .__refcnt       = ATOMIC_INIT(1),
                        .__use          = 1,
-                       .dev            = &loopback_dev,
                        .obsolete       = -1,
                        .error          = -ENETUNREACH,
                        .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
@@ -142,17 +151,19 @@ struct rt6_info ip6_null_entry = {
 
 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 
+static int ip6_pkt_prohibit(struct sk_buff *skb);
+static int ip6_pkt_prohibit_out(struct sk_buff *skb);
+
 struct rt6_info ip6_prohibit_entry = {
        .u = {
                .dst = {
                        .__refcnt       = ATOMIC_INIT(1),
                        .__use          = 1,
-                       .dev            = &loopback_dev,
                        .obsolete       = -1,
                        .error          = -EACCES,
                        .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
-                       .input          = ip6_pkt_discard,
-                       .output         = ip6_pkt_discard_out,
+                       .input          = ip6_pkt_prohibit,
+                       .output         = ip6_pkt_prohibit_out,
                        .ops            = &ip6_dst_ops,
                        .path           = (struct dst_entry*)&ip6_prohibit_entry,
                }
@@ -167,12 +178,11 @@ struct rt6_info ip6_blk_hole_entry = {
                .dst = {
                        .__refcnt       = ATOMIC_INIT(1),
                        .__use          = 1,
-                       .dev            = &loopback_dev,
                        .obsolete       = -1,
                        .error          = -EINVAL,
                        .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
-                       .input          = ip6_pkt_discard,
-                       .output         = ip6_pkt_discard_out,
+                       .input          = dst_discard,
+                       .output         = dst_discard,
                        .ops            = &ip6_dst_ops,
                        .path           = (struct dst_entry*)&ip6_blk_hole_entry,
                }
@@ -198,7 +208,7 @@ static void ip6_dst_destroy(struct dst_entry *dst)
        if (idev != NULL) {
                rt->rt6i_idev = NULL;
                in6_dev_put(idev);
-       }       
+       }
 }
 
 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
@@ -207,8 +217,8 @@ static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
        struct rt6_info *rt = (struct rt6_info *)dst;
        struct inet6_dev *idev = rt->rt6i_idev;
 
-       if (dev != &loopback_dev && idev != NULL && idev->dev == dev) {
-               struct inet6_dev *loopback_idev = in6_dev_get(&loopback_dev);
+       if (dev != init_net.loopback_dev && idev != NULL && idev->dev == dev) {
+               struct inet6_dev *loopback_idev = in6_dev_get(init_net.loopback_dev);
                if (loopback_idev != NULL) {
                        rt->rt6i_idev = loopback_idev;
                        in6_dev_put(idev);
@@ -240,7 +250,7 @@ static __inline__ struct rt6_info *rt6_device_match(struct rt6_info *rt,
        struct rt6_info *sprt;
 
        if (oif) {
-               for (sprt = rt; sprt; sprt = sprt->u.next) {
+               for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
                        struct net_device *dev = sprt->rt6i_dev;
                        if (dev->ifindex == oif)
                                return sprt;
@@ -249,7 +259,7 @@ static __inline__ struct rt6_info *rt6_device_match(struct rt6_info *rt,
                                    sprt->rt6i_idev->dev->ifindex != oif) {
                                        if (strict && oif)
                                                continue;
-                                       if (local && (!oif || 
+                                       if (local && (!oif ||
                                                      local->rt6i_idev->dev->ifindex == oif))
                                                continue;
                                }
@@ -305,7 +315,7 @@ static inline void rt6_probe(struct rt6_info *rt)
 /*
  * Default Router Selection (RFC 2461 6.3.6)
  */
-static int inline rt6_check_dev(struct rt6_info *rt, int oif)
+static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 {
        struct net_device *dev = rt->rt6i_dev;
        if (!oif || dev->ifindex == oif)
@@ -316,10 +326,10 @@ static int inline rt6_check_dev(struct rt6_info *rt, int oif)
        return 0;
 }
 
-static int inline rt6_check_neigh(struct rt6_info *rt)
+static inline int rt6_check_neigh(struct rt6_info *rt)
 {
        struct neighbour *neigh = rt->rt6i_nexthop;
-       int m = 0;
+       int m;
        if (rt->rt6i_flags & RTF_NONEXTHOP ||
            !(rt->rt6i_flags & RTF_GATEWAY))
                m = 1;
@@ -327,8 +337,15 @@ static int inline rt6_check_neigh(struct rt6_info *rt)
                read_lock_bh(&neigh->lock);
                if (neigh->nud_state & NUD_VALID)
                        m = 2;
+#ifdef CONFIG_IPV6_ROUTER_PREF
+               else if (neigh->nud_state & NUD_FAILED)
+                       m = 0;
+#endif
+               else
+                       m = 1;
                read_unlock_bh(&neigh->lock);
-       }
+       } else
+               m = 0;
        return m;
 }
 
@@ -336,69 +353,89 @@ static int rt6_score_route(struct rt6_info *rt, int oif,
                           int strict)
 {
        int m, n;
-               
+
        m = rt6_check_dev(rt, oif);
-       if (!m && (strict & RT6_SELECT_F_IFACE))
+       if (!m && (strict & RT6_LOOKUP_F_IFACE))
                return -1;
 #ifdef CONFIG_IPV6_ROUTER_PREF
        m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 #endif
        n = rt6_check_neigh(rt);
-       if (n > 1)
-               m |= 16;
-       else if (!n && strict & RT6_SELECT_F_REACHABLE)
+       if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
                return -1;
        return m;
 }
 
-static struct rt6_info *rt6_select(struct rt6_info **head, int oif,
-                                  int strict)
+static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
+                                  int *mpri, struct rt6_info *match)
 {
-       struct rt6_info *match = NULL, *last = NULL;
-       struct rt6_info *rt, *rt0 = *head;
-       u32 metric;
+       int m;
+
+       if (rt6_check_expired(rt))
+               goto out;
+
+       m = rt6_score_route(rt, oif, strict);
+       if (m < 0)
+               goto out;
+
+       if (m > *mpri) {
+               if (strict & RT6_LOOKUP_F_REACHABLE)
+                       rt6_probe(match);
+               *mpri = m;
+               match = rt;
+       } else if (strict & RT6_LOOKUP_F_REACHABLE) {
+               rt6_probe(rt);
+       }
+
+out:
+       return match;
+}
+
+static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
+                                    struct rt6_info *rr_head,
+                                    u32 metric, int oif, int strict)
+{
+       struct rt6_info *rt, *match;
        int mpri = -1;
 
-       RT6_TRACE("%s(head=%p(*head=%p), oif=%d)\n",
-                 __FUNCTION__, head, head ? *head : NULL, oif);
+       match = NULL;
+       for (rt = rr_head; rt && rt->rt6i_metric == metric;
+            rt = rt->u.dst.rt6_next)
+               match = find_match(rt, oif, strict, &mpri, match);
+       for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
+            rt = rt->u.dst.rt6_next)
+               match = find_match(rt, oif, strict, &mpri, match);
 
-       for (rt = rt0, metric = rt0->rt6i_metric;
-            rt && rt->rt6i_metric == metric && (!last || rt != rt0);
-            rt = rt->u.next) {
-               int m;
+       return match;
+}
 
-               if (rt6_check_expired(rt))
-                       continue;
+static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
+{
+       struct rt6_info *match, *rt0;
 
-               last = rt;
+       RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
+                 __FUNCTION__, fn->leaf, oif);
 
-               m = rt6_score_route(rt, oif, strict);
-               if (m < 0)
-                       continue;
+       rt0 = fn->rr_ptr;
+       if (!rt0)
+               fn->rr_ptr = rt0 = fn->leaf;
 
-               if (m > mpri) {
-                       rt6_probe(match);
-                       match = rt;
-                       mpri = m;
-               } else {
-                       rt6_probe(rt);
-               }
-       }
+       match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
 
        if (!match &&
-           (strict & RT6_SELECT_F_REACHABLE) &&
-           last && last != rt0) {
+           (strict & RT6_LOOKUP_F_REACHABLE)) {
+               struct rt6_info *next = rt0->u.dst.rt6_next;
+
                /* no entries matched; do round-robin */
-               static DEFINE_SPINLOCK(lock);
-               spin_lock(&lock);
-               *head = rt0->u.next;
-               rt0->u.next = last->u.next;
-               last->u.next = rt0;
-               spin_unlock(&lock);
+               if (!next || next->rt6i_metric != rt0->rt6i_metric)
+                       next = fn->leaf;
+
+               if (next != rt0)
+                       fn->rr_ptr = next;
        }
 
-       RT6_TRACE("%s() => %p, score=%d\n",
-                 __FUNCTION__, match, mpri);
+       RT6_TRACE("%s() => %p\n",
+                 __FUNCTION__, match);
 
        return (match ? match : &ip6_null_entry);
 }
@@ -436,7 +473,7 @@ int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
        if (pref == ICMPV6_ROUTER_PREF_INVALID)
                pref = ICMPV6_ROUTER_PREF_MEDIUM;
 
-       lifetime = htonl(rinfo->lifetime);
+       lifetime = ntohl(rinfo->lifetime);
        if (lifetime == 0xffffffff) {
                /* infinity */
        } else if (lifetime > 0x7fffffff/HZ) {
@@ -481,17 +518,23 @@ int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 }
 #endif
 
-#define BACKTRACK() \
-if (rt == &ip6_null_entry && flags & RT6_F_STRICT) { \
-       while ((fn = fn->parent) != NULL) { \
-               if (fn->fn_flags & RTN_TL_ROOT) { \
-                       dst_hold(&rt->u.dst); \
-                       goto out; \
+#define BACKTRACK(saddr) \
+do { \
+       if (rt == &ip6_null_entry) { \
+               struct fib6_node *pn; \
+               while (1) { \
+                       if (fn->fn_flags & RTN_TL_ROOT) \
+                               goto out; \
+                       pn = fn->parent; \
+                       if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
+                               fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
+                       else \
+                               fn = pn; \
+                       if (fn->fn_flags & RTN_RTINFO) \
+                               goto restart; \
                } \
-               if (fn->fn_flags & RTN_RTINFO) \
-                       goto restart; \
        } \
-}
+} while(0)
 
 static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
                                             struct flowi *fl, int flags)
@@ -503,15 +546,11 @@ static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
        fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 restart:
        rt = fn->leaf;
-       rt = rt6_device_match(rt, fl->oif, flags & RT6_F_STRICT);
-       BACKTRACK();
-       dst_hold(&rt->u.dst);
+       rt = rt6_device_match(rt, fl->oif, flags);
+       BACKTRACK(&fl->fl6_src);
 out:
+       dst_use(&rt->u.dst, jiffies);
        read_unlock_bh(&table->tb6_lock);
-
-       rt->u.dst.lastuse = jiffies;
-       rt->u.dst.__use++;
-
        return rt;
 
 }
@@ -524,12 +563,16 @@ struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
                .nl_u = {
                        .ip6_u = {
                                .daddr = *daddr,
-                               /* TODO: saddr */
                        },
                },
        };
        struct dst_entry *dst;
-       int flags = strict ? RT6_F_STRICT : 0;
+       int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
+
+       if (saddr) {
+               memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
+               flags |= RT6_LOOKUP_F_HAS_SADDR;
+       }
 
        dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_lookup);
        if (dst->error == 0)
@@ -540,6 +583,8 @@ struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
        return NULL;
 }
 
+EXPORT_SYMBOL(rt6_lookup);
+
 /* ip6_ins_rt is called with FREE table->tb6_lock.
    It takes new route entry, the addition fails by any reason the
    route is freed. In any case, if caller does not hold it, it may
@@ -609,15 +654,13 @@ static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *d
                ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
                rt->rt6i_dst.plen = 128;
                rt->rt6i_flags |= RTF_CACHE;
-               if (rt->rt6i_flags & RTF_REJECT)
-                       rt->u.dst.error = ort->u.dst.error;
                rt->u.dst.flags |= DST_HOST;
                rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
        }
        return rt;
 }
 
-static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
+static struct rt6_info *ip6_pol_route(struct fib6_table *table, int oif,
                                            struct flowi *fl, int flags)
 {
        struct fib6_node *fn;
@@ -625,10 +668,9 @@ static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
        int strict = 0;
        int attempts = 3;
        int err;
-       int reachable = RT6_SELECT_F_REACHABLE;
+       int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
 
-       if (flags & RT6_F_STRICT)
-               strict = RT6_SELECT_F_IFACE;
+       strict |= flags & RT6_LOOKUP_F_IFACE;
 
 relookup:
        read_lock_bh(&table->tb6_lock);
@@ -637,8 +679,8 @@ restart_2:
        fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 
 restart:
-       rt = rt6_select(&fn->leaf, fl->iif, strict | reachable);
-       BACKTRACK();
+       rt = rt6_select(fn, oif, strict | reachable);
+       BACKTRACK(&fl->fl6_src);
        if (rt == &ip6_null_entry ||
            rt->rt6i_flags & RTF_CACHE)
                goto out;
@@ -690,24 +732,31 @@ out2:
        return rt;
 }
 
+static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
+                                           struct flowi *fl, int flags)
+{
+       return ip6_pol_route(table, fl->iif, fl, flags);
+}
+
 void ip6_route_input(struct sk_buff *skb)
 {
-       struct ipv6hdr *iph = skb->nh.ipv6h;
+       struct ipv6hdr *iph = ipv6_hdr(skb);
+       int flags = RT6_LOOKUP_F_HAS_SADDR;
        struct flowi fl = {
                .iif = skb->dev->ifindex,
                .nl_u = {
                        .ip6_u = {
                                .daddr = iph->daddr,
                                .saddr = iph->saddr,
-                               .flowlabel = (* (u32 *) iph)&IPV6_FLOWINFO_MASK,
+                               .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
                        },
                },
+               .mark = skb->mark,
                .proto = iph->nexthdr,
        };
-       int flags = 0;
 
        if (rt6_need_strict(&iph->daddr))
-               flags |= RT6_F_STRICT;
+               flags |= RT6_LOOKUP_F_IFACE;
 
        skb->dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_input);
 }
@@ -715,85 +764,65 @@ void ip6_route_input(struct sk_buff *skb)
 static struct rt6_info *ip6_pol_route_output(struct fib6_table *table,
                                             struct flowi *fl, int flags)
 {
-       struct fib6_node *fn;
-       struct rt6_info *rt, *nrt;
-       int strict = 0;
-       int attempts = 3;
-       int err;
-       int reachable = RT6_SELECT_F_REACHABLE;
-
-       if (flags & RT6_F_STRICT)
-               strict = RT6_SELECT_F_IFACE;
+       return ip6_pol_route(table, fl->oif, fl, flags);
+}
 
-relookup:
-       read_lock_bh(&table->tb6_lock);
+struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
+{
+       int flags = 0;
 
-restart_2:
-       fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
+       if (rt6_need_strict(&fl->fl6_dst))
+               flags |= RT6_LOOKUP_F_IFACE;
 
-restart:
-       rt = rt6_select(&fn->leaf, fl->oif, strict | reachable);
-       BACKTRACK();
-       if (rt == &ip6_null_entry ||
-           rt->rt6i_flags & RTF_CACHE)
-               goto out;
+       if (!ipv6_addr_any(&fl->fl6_src))
+               flags |= RT6_LOOKUP_F_HAS_SADDR;
 
-       dst_hold(&rt->u.dst);
-       read_unlock_bh(&table->tb6_lock);
+       return fib6_rule_lookup(fl, flags, ip6_pol_route_output);
+}
 
-       if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
-               nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
-       else {
-#if CLONE_OFFLINK_ROUTE
-               nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
-#else
-               goto out2;
-#endif
-       }
+EXPORT_SYMBOL(ip6_route_output);
 
-       dst_release(&rt->u.dst);
-       rt = nrt ? : &ip6_null_entry;
+int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
+{
+       struct rt6_info *ort = (struct rt6_info *) *dstp;
+       struct rt6_info *rt = (struct rt6_info *)
+               dst_alloc(&ip6_dst_blackhole_ops);
+       struct dst_entry *new = NULL;
 
-       dst_hold(&rt->u.dst);
-       if (nrt) {
-               err = ip6_ins_rt(nrt);
-               if (!err)
-                       goto out2;
-       }
+       if (rt) {
+               new = &rt->u.dst;
 
-       if (--attempts <= 0)
-               goto out2;
+               atomic_set(&new->__refcnt, 1);
+               new->__use = 1;
+               new->input = dst_discard;
+               new->output = dst_discard;
 
-       /*
-        * Race condition! In the gap, when table->tb6_lock was
-        * released someone could insert this route.  Relookup.
-        */
-       dst_release(&rt->u.dst);
-       goto relookup;
+               memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
+               new->dev = ort->u.dst.dev;
+               if (new->dev)
+                       dev_hold(new->dev);
+               rt->rt6i_idev = ort->rt6i_idev;
+               if (rt->rt6i_idev)
+                       in6_dev_hold(rt->rt6i_idev);
+               rt->rt6i_expires = 0;
 
-out:
-       if (reachable) {
-               reachable = 0;
-               goto restart_2;
-       }
-       dst_hold(&rt->u.dst);
-       read_unlock_bh(&table->tb6_lock);
-out2:
-       rt->u.dst.lastuse = jiffies;
-       rt->u.dst.__use++;
-       return rt;
-}
+               ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
+               rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
+               rt->rt6i_metric = 0;
 
-struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
-{
-       int flags = 0;
+               memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
+#ifdef CONFIG_IPV6_SUBTREES
+               memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
+#endif
 
-       if (rt6_need_strict(&fl->fl6_dst))
-               flags |= RT6_F_STRICT;
+               dst_free(new);
+       }
 
-       return fib6_rule_lookup(fl, flags, ip6_pol_route_output);
+       dst_release(*dstp);
+       *dstp = new;
+       return (new ? 0 : -ENOMEM);
 }
-
+EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
 
 /*
  *     Destination cache support functions
@@ -865,9 +894,9 @@ static inline unsigned int ipv6_advmss(unsigned int mtu)
                mtu = ip6_rt_min_advmss;
 
        /*
-        * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and 
-        * corresponding MSS is IPV6_MAXPLEN - tcp_header_size. 
-        * IPV6_MAXPLEN is also valid and means: "any MSS, 
+        * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
+        * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
+        * IPV6_MAXPLEN is also valid and means: "any MSS,
         * rely only on pmtu discovery"
         */
        if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
@@ -878,7 +907,7 @@ static inline unsigned int ipv6_advmss(unsigned int mtu)
 static struct dst_entry *ndisc_dst_gc_list;
 static DEFINE_SPINLOCK(ndisc_lock);
 
-struct dst_entry *ndisc_dst_alloc(struct net_device *dev, 
+struct dst_entry *ndisc_dst_alloc(struct net_device *dev,
                                  struct neighbour *neigh,
                                  struct in6_addr *addr,
                                  int (*output)(struct sk_buff *))
@@ -911,8 +940,8 @@ struct dst_entry *ndisc_dst_alloc(struct net_device *dev,
        rt->u.dst.output  = output;
 
 #if 0  /* there's no chance to use these for ndisc */
-       rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST 
-                               ? DST_HOST 
+       rt->u.dst.flags   = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
+                               ? DST_HOST
                                : 0;
        ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
        rt->rt6i_dst.plen = 128;
@@ -926,7 +955,7 @@ struct dst_entry *ndisc_dst_alloc(struct net_device *dev,
        fib6_force_start_gc();
 
 out:
-       return (struct dst_entry *)rt;
+       return &rt->u.dst;
 }
 
 int ndisc_dst_gc(int *more)
@@ -935,7 +964,7 @@ int ndisc_dst_gc(int *more)
        int freed;
 
        next = NULL;
-       freed = 0;
+       freed = 0;
 
        spin_lock_bh(&ndisc_lock);
        pprev = &ndisc_dst_gc_list;
@@ -1030,7 +1059,7 @@ int ip6_route_add(struct fib6_config *cfg)
 #endif
        if (cfg->fc_ifindex) {
                err = -ENODEV;
-               dev = dev_get_by_index(cfg->fc_ifindex);
+               dev = dev_get_by_index(&init_net, cfg->fc_ifindex);
                if (!dev)
                        goto out;
                idev = in6_dev_get(dev);
@@ -1088,12 +1117,12 @@ int ip6_route_add(struct fib6_config *cfg)
        if ((cfg->fc_flags & RTF_REJECT) ||
            (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
                /* hold loopback dev/idev if we haven't done so. */
-               if (dev != &loopback_dev) {
+               if (dev != init_net.loopback_dev) {
                        if (dev) {
                                dev_put(dev);
                                in6_dev_put(idev);
                        }
-                       dev = &loopback_dev;
+                       dev = init_net.loopback_dev;
                        dev_hold(dev);
                        idev = in6_dev_get(dev);
                        if (!idev) {
@@ -1179,7 +1208,7 @@ install_route:
                int remaining;
 
                nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
-                       int type = nla->nla_type;
+                       int type = nla_type(nla);
 
                        if (type) {
                                if (type > RTAX_MAX) {
@@ -1209,7 +1238,7 @@ out:
        if (idev)
                in6_dev_put(idev);
        if (rt)
-               dst_free((struct dst_entry *) rt);
+               dst_free(&rt->u.dst);
        return err;
 }
 
@@ -1253,9 +1282,9 @@ static int ip6_route_del(struct fib6_config *cfg)
        fn = fib6_locate(&table->tb6_root,
                         &cfg->fc_dst, cfg->fc_dst_len,
                         &cfg->fc_src, cfg->fc_src_len);
-       
+
        if (fn) {
-               for (rt = fn->leaf; rt; rt = rt->u.next) {
+               for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
                        if (cfg->fc_ifindex &&
                            (rt->rt6i_dev == NULL ||
                             rt->rt6i_dev->ifindex != cfg->fc_ifindex))
@@ -1279,19 +1308,18 @@ static int ip6_route_del(struct fib6_config *cfg)
 /*
  *     Handle redirects
  */
-void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
-                 struct in6_addr *saddr,
-                 struct neighbour *neigh, u8 *lladdr, int on_link)
+struct ip6rd_flowi {
+       struct flowi fl;
+       struct in6_addr gateway;
+};
+
+static struct rt6_info *__ip6_route_redirect(struct fib6_table *table,
+                                            struct flowi *fl,
+                                            int flags)
 {
-       struct rt6_info *rt, *nrt = NULL;
+       struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
+       struct rt6_info *rt;
        struct fib6_node *fn;
-       struct fib6_table *table;
-       struct netevent_redirect netevent;
-
-       /* TODO: Very lazy, might need to check all tables */
-       table = fib6_get_table(RT6_TABLE_MAIN);
-       if (table == NULL)
-               return;
 
        /*
         * Get the "current" route for this destination and
@@ -1305,9 +1333,9 @@ void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
         */
 
        read_lock_bh(&table->tb6_lock);
-       fn = fib6_lookup(&table->tb6_root, dest, src);
+       fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 restart:
-       for (rt = fn->leaf; rt; rt = rt->u.next) {
+       for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
                /*
                 * Current route is on-link; redirect is always invalid.
                 *
@@ -1320,36 +1348,70 @@ restart:
                        continue;
                if (!(rt->rt6i_flags & RTF_GATEWAY))
                        continue;
-               if (neigh->dev != rt->rt6i_dev)
+               if (fl->oif != rt->rt6i_dev->ifindex)
                        continue;
-               if (!ipv6_addr_equal(saddr, &rt->rt6i_gateway))
+               if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
                        continue;
                break;
        }
-       if (rt)
-               dst_hold(&rt->u.dst);
-       else if (rt6_need_strict(dest)) {
-               while ((fn = fn->parent) != NULL) {
-                       if (fn->fn_flags & RTN_ROOT)
-                               break;
-                       if (fn->fn_flags & RTN_RTINFO)
-                               goto restart;
-               }
-       }
+
+       if (!rt)
+               rt = &ip6_null_entry;
+       BACKTRACK(&fl->fl6_src);
+out:
+       dst_hold(&rt->u.dst);
+
        read_unlock_bh(&table->tb6_lock);
 
-       if (!rt) {
+       return rt;
+};
+
+static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
+                                          struct in6_addr *src,
+                                          struct in6_addr *gateway,
+                                          struct net_device *dev)
+{
+       int flags = RT6_LOOKUP_F_HAS_SADDR;
+       struct ip6rd_flowi rdfl = {
+               .fl = {
+                       .oif = dev->ifindex,
+                       .nl_u = {
+                               .ip6_u = {
+                                       .daddr = *dest,
+                                       .saddr = *src,
+                               },
+                       },
+               },
+               .gateway = *gateway,
+       };
+
+       if (rt6_need_strict(dest))
+               flags |= RT6_LOOKUP_F_IFACE;
+
+       return (struct rt6_info *)fib6_rule_lookup((struct flowi *)&rdfl, flags, __ip6_route_redirect);
+}
+
+void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
+                 struct in6_addr *saddr,
+                 struct neighbour *neigh, u8 *lladdr, int on_link)
+{
+       struct rt6_info *rt, *nrt = NULL;
+       struct netevent_redirect netevent;
+
+       rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
+
+       if (rt == &ip6_null_entry) {
                if (net_ratelimit())
                        printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
                               "for redirect target\n");
-               return;
+               goto out;
        }
 
        /*
         *      We have finally decided to accept it.
         */
 
-       neigh_update(neigh, lladdr, NUD_STALE, 
+       neigh_update(neigh, lladdr, NUD_STALE,
                     NEIGH_UPDATE_F_WEAK_OVERRIDE|
                     NEIGH_UPDATE_F_OVERRIDE|
                     (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
@@ -1398,7 +1460,7 @@ restart:
        }
 
 out:
-        dst_release(&rt->u.dst);
+       dst_release(&rt->u.dst);
        return;
 }
 
@@ -1422,7 +1484,7 @@ void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
 
        if (pmtu < IPV6_MIN_MTU) {
                /*
-                * According to RFC2460, PMTU is set to the IPv6 Minimum Link 
+                * According to RFC2460, PMTU is set to the IPv6 Minimum Link
                 * MTU (1280) and a fragment header should always be included
                 * after a node receiving Too Big message reporting PMTU is
                 * less than the IPv6 Minimum Link MTU.
@@ -1494,6 +1556,7 @@ static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
                rt->u.dst.output = ort->u.dst.output;
 
                memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
+               rt->u.dst.error = ort->u.dst.error;
                rt->u.dst.dev = ort->u.dst.dev;
                if (rt->u.dst.dev)
                        dev_hold(rt->u.dst.dev);
@@ -1533,7 +1596,7 @@ static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixle
        if (!fn)
                goto out;
 
-       for (rt = fn->leaf; rt; rt = rt->u.next) {
+       for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
                if (rt->rt6i_dev->ifindex != ifindex)
                        continue;
                if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
@@ -1575,7 +1638,7 @@ static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixle
 #endif
 
 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
-{      
+{
        struct rt6_info *rt;
        struct fib6_table *table;
 
@@ -1584,7 +1647,7 @@ struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *d
                return NULL;
 
        write_lock_bh(&table->tb6_lock);
-       for (rt = table->tb6_root.leaf; rt; rt=rt->u.next) {
+       for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
                if (dev == rt->rt6i_dev &&
                    ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
                    ipv6_addr_equal(&rt->rt6i_gateway, addr))
@@ -1627,7 +1690,7 @@ void rt6_purge_dflt_routers(void)
 
 restart:
        read_lock_bh(&table->tb6_lock);
-       for (rt = table->tb6_root.leaf; rt; rt = rt->u.next) {
+       for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
                if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
                        dst_hold(&rt->u.dst);
                        read_unlock_bh(&table->tb6_lock);
@@ -1688,7 +1751,7 @@ int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
                rtnl_unlock();
 
                return err;
-       };
+       }
 
        return -EINVAL;
 }
@@ -1697,24 +1760,53 @@ int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
  *     Drop the packet on the floor
  */
 
-static int ip6_pkt_discard(struct sk_buff *skb)
+static inline int ip6_pkt_drop(struct sk_buff *skb, int code,
+                              int ipstats_mib_noroutes)
 {
-       int type = ipv6_addr_type(&skb->nh.ipv6h->daddr);
-       if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED)
-               IP6_INC_STATS(IPSTATS_MIB_INADDRERRORS);
-
-       IP6_INC_STATS(IPSTATS_MIB_OUTNOROUTES);
-       icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_NOROUTE, 0, skb->dev);
+       int type;
+       switch (ipstats_mib_noroutes) {
+       case IPSTATS_MIB_INNOROUTES:
+               type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
+               if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
+                       IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
+                       break;
+               }
+               /* FALLTHROUGH */
+       case IPSTATS_MIB_OUTNOROUTES:
+               IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
+               break;
+       }
+       icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
        kfree_skb(skb);
        return 0;
 }
 
+static int ip6_pkt_discard(struct sk_buff *skb)
+{
+       return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
+}
+
 static int ip6_pkt_discard_out(struct sk_buff *skb)
 {
        skb->dev = skb->dst->dev;
-       return ip6_pkt_discard(skb);
+       return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
 }
 
+#ifdef CONFIG_IPV6_MULTIPLE_TABLES
+
+static int ip6_pkt_prohibit(struct sk_buff *skb)
+{
+       return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
+}
+
+static int ip6_pkt_prohibit_out(struct sk_buff *skb)
+{
+       skb->dev = skb->dst->dev;
+       return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
+}
+
+#endif
+
 /*
  *     Allocate a dst for local (unicast / anycast) address.
  */
@@ -1728,13 +1820,13 @@ struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
        if (rt == NULL)
                return ERR_PTR(-ENOMEM);
 
-       dev_hold(&loopback_dev);
+       dev_hold(init_net.loopback_dev);
        in6_dev_hold(idev);
 
        rt->u.dst.flags = DST_HOST;
        rt->u.dst.input = ip6_input;
        rt->u.dst.output = ip6_output;
-       rt->rt6i_dev = &loopback_dev;
+       rt->rt6i_dev = init_net.loopback_dev;
        rt->rt6i_idev = idev;
        rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
        rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
@@ -1748,7 +1840,7 @@ struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
                rt->rt6i_flags |= RTF_LOCAL;
        rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
        if (rt->rt6i_nexthop == NULL) {
-               dst_free((struct dst_entry *) rt);
+               dst_free(&rt->u.dst);
                return ERR_PTR(-ENOMEM);
        }
 
@@ -1813,11 +1905,12 @@ static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
         */
        if (rt->rt6i_dev == arg->dev &&
            !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
-            (dst_mtu(&rt->u.dst) > arg->mtu ||
-             (dst_mtu(&rt->u.dst) < arg->mtu &&
-             dst_mtu(&rt->u.dst) == idev->cnf.mtu6)))
+           (dst_mtu(&rt->u.dst) > arg->mtu ||
+            (dst_mtu(&rt->u.dst) < arg->mtu &&
+             dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
                rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
-       rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(arg->mtu);
+               rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(arg->mtu);
+       }
        return 0;
 }
 
@@ -1831,8 +1924,8 @@ void rt6_mtu_change(struct net_device *dev, unsigned mtu)
        fib6_clean_all(rt6_mtu_change_route, 0, &arg);
 }
 
-static struct nla_policy rtm_ipv6_policy[RTA_MAX+1] __read_mostly = {
-       [RTA_GATEWAY]           = { .minlen = sizeof(struct in6_addr) },
+static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
+       [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
        [RTA_OIF]               = { .type = NLA_U32 },
        [RTA_IIF]               = { .type = NLA_U32 },
        [RTA_PRIORITY]          = { .type = NLA_U32 },
@@ -1908,11 +2001,15 @@ errout:
        return err;
 }
 
-int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
+static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
 {
+       struct net *net = skb->sk->sk_net;
        struct fib6_config cfg;
        int err;
 
+       if (net != &init_net)
+               return -EINVAL;
+
        err = rtm_to_fib6_config(skb, nlh, &cfg);
        if (err < 0)
                return err;
@@ -1920,11 +2017,15 @@ int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
        return ip6_route_del(&cfg);
 }
 
-int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
+static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
 {
+       struct net *net = skb->sk->sk_net;
        struct fib6_config cfg;
        int err;
 
+       if (net != &init_net)
+               return -EINVAL;
+
        err = rtm_to_fib6_config(skb, nlh, &cfg);
        if (err < 0)
                return err;
@@ -1932,6 +2033,21 @@ int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
        return ip6_route_add(&cfg);
 }
 
+static inline size_t rt6_nlmsg_size(void)
+{
+       return NLMSG_ALIGN(sizeof(struct rtmsg))
+              + nla_total_size(16) /* RTA_SRC */
+              + nla_total_size(16) /* RTA_DST */
+              + nla_total_size(16) /* RTA_GATEWAY */
+              + nla_total_size(16) /* RTA_PREFSRC */
+              + nla_total_size(4) /* RTA_TABLE */
+              + nla_total_size(4) /* RTA_IIF */
+              + nla_total_size(4) /* RTA_OIF */
+              + nla_total_size(4) /* RTA_PRIORITY */
+              + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
+              + nla_total_size(sizeof(struct rta_cacheinfo));
+}
+
 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
                         struct in6_addr *dst, struct in6_addr *src,
                         int iif, int type, u32 pid, u32 seq,
@@ -1939,7 +2055,7 @@ static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
 {
        struct rtmsg *rtm;
        struct nlmsghdr *nlh;
-       struct rta_cacheinfo ci;
+       long expires;
        u32 table;
 
        if (prefix) {   /* user wants prefix routes only */
@@ -1951,7 +2067,7 @@ static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
 
        nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
        if (nlh == NULL)
-               return -ENOBUFS;
+               return -EMSGSIZE;
 
        rtm = nlmsg_data(nlh);
        rtm->rtm_family = AF_INET6;
@@ -1985,13 +2101,13 @@ static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
 
        if (dst) {
                NLA_PUT(skb, RTA_DST, 16, dst);
-               rtm->rtm_dst_len = 128;
+               rtm->rtm_dst_len = 128;
        } else if (rtm->rtm_dst_len)
                NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
 #ifdef CONFIG_IPV6_SUBTREES
        if (src) {
                NLA_PUT(skb, RTA_SRC, 16, src);
-               rtm->rtm_src_len = 128;
+               rtm->rtm_src_len = 128;
        } else if (rtm->rtm_src_len)
                NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
 #endif
@@ -2013,23 +2129,17 @@ static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
                NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
 
        NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
-       ci.rta_lastuse = jiffies_to_clock_t(jiffies - rt->u.dst.lastuse);
-       if (rt->rt6i_expires)
-               ci.rta_expires = jiffies_to_clock_t(rt->rt6i_expires - jiffies);
-       else
-               ci.rta_expires = 0;
-       ci.rta_used = rt->u.dst.__use;
-       ci.rta_clntref = atomic_read(&rt->u.dst.__refcnt);
-       ci.rta_error = rt->u.dst.error;
-       ci.rta_id = 0;
-       ci.rta_ts = 0;
-       ci.rta_tsage = 0;
-       NLA_PUT(skb, RTA_CACHEINFO, sizeof(ci), &ci);
+
+       expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
+       if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
+                              expires, rt->u.dst.error) < 0)
+               goto nla_put_failure;
 
        return nlmsg_end(skb, nlh);
 
 nla_put_failure:
-       return nlmsg_cancel(skb, nlh);
+       nlmsg_cancel(skb, nlh);
+       return -EMSGSIZE;
 }
 
 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
@@ -2048,8 +2158,9 @@ int rt6_dump_route(struct rt6_info *rt, void *p_arg)
                     prefix, NLM_F_MULTI);
 }
 
-int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
+static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
 {
+       struct net *net = in_skb->sk->sk_net;
        struct nlattr *tb[RTA_MAX+1];
        struct rt6_info *rt;
        struct sk_buff *skb;
@@ -2057,6 +2168,9 @@ int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
        struct flowi fl;
        int err, iif = 0;
 
+       if (net != &init_net)
+               return -EINVAL;
+
        err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
        if (err < 0)
                goto errout;
@@ -2086,7 +2200,7 @@ int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
 
        if (iif) {
                struct net_device *dev;
-               dev = __dev_get_by_index(iif);
+               dev = __dev_get_by_index(&init_net, iif);
                if (!dev) {
                        err = -ENODEV;
                        goto errout;
@@ -2102,7 +2216,7 @@ int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
        /* Reserve room for dummy headers, this skb can pass
           through good chunk of routing engine.
         */
-       skb->mac.raw = skb->data;
+       skb_reset_mac_header(skb);
        skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
 
        rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
@@ -2116,7 +2230,7 @@ int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
                goto errout;
        }
 
-       err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
+       err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
 errout:
        return err;
 }
@@ -2126,7 +2240,6 @@ void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
        struct sk_buff *skb;
        u32 pid = 0, seq = 0;
        struct nlmsghdr *nlh = NULL;
-       int payload = sizeof(struct rtmsg) + 256;
        int err = -ENOBUFS;
 
        if (info) {
@@ -2136,20 +2249,21 @@ void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
                        seq = nlh->nlmsg_seq;
        }
 
-       skb = nlmsg_new(nlmsg_total_size(payload), gfp_any());
+       skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
        if (skb == NULL)
                goto errout;
 
        err = rt6_fill_node(skb, rt, NULL, NULL, 0, event, pid, seq, 0, 0);
        if (err < 0) {
+               /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
+               WARN_ON(err == -EMSGSIZE);
                kfree_skb(skb);
                goto errout;
        }
-
-       err = rtnl_notify(skb, pid, RTNLGRP_IPV6_ROUTE, nlh, gfp_any());
+       err = rtnl_notify(skb, &init_net, pid, RTNLGRP_IPV6_ROUTE, nlh, gfp_any());
 errout:
        if (err < 0)
-               rtnl_set_sk_err(RTNLGRP_IPV6_ROUTE, err);
+               rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_ROUTE, err);
 }
 
 /*
@@ -2171,82 +2285,50 @@ struct rt6_proc_arg
 
 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
 {
-       struct rt6_proc_arg *arg = (struct rt6_proc_arg *) p_arg;
-       int i;
+       struct seq_file *m = p_arg;
 
-       if (arg->skip < arg->offset / RT6_INFO_LEN) {
-               arg->skip++;
-               return 0;
-       }
-
-       if (arg->len >= arg->length)
-               return 0;
-
-       for (i=0; i<16; i++) {
-               sprintf(arg->buffer + arg->len, "%02x",
-                       rt->rt6i_dst.addr.s6_addr[i]);
-               arg->len += 2;
-       }
-       arg->len += sprintf(arg->buffer + arg->len, " %02x ",
-                           rt->rt6i_dst.plen);
+       seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
+                  rt->rt6i_dst.plen);
 
 #ifdef CONFIG_IPV6_SUBTREES
-       for (i=0; i<16; i++) {
-               sprintf(arg->buffer + arg->len, "%02x",
-                       rt->rt6i_src.addr.s6_addr[i]);
-               arg->len += 2;
-       }
-       arg->len += sprintf(arg->buffer + arg->len, " %02x ",
-                           rt->rt6i_src.plen);
+       seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
+                  rt->rt6i_src.plen);
 #else
-       sprintf(arg->buffer + arg->len,
-               "00000000000000000000000000000000 00 ");
-       arg->len += 36;
+       seq_puts(m, "00000000000000000000000000000000 00 ");
 #endif
 
        if (rt->rt6i_nexthop) {
-               for (i=0; i<16; i++) {
-                       sprintf(arg->buffer + arg->len, "%02x",
-                               rt->rt6i_nexthop->primary_key[i]);
-                       arg->len += 2;
-               }
+               seq_printf(m, NIP6_SEQFMT,
+                          NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
        } else {
-               sprintf(arg->buffer + arg->len,
-                       "00000000000000000000000000000000");
-               arg->len += 32;
+               seq_puts(m, "00000000000000000000000000000000");
        }
-       arg->len += sprintf(arg->buffer + arg->len,
-                           " %08x %08x %08x %08x %8s\n",
-                           rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
-                           rt->u.dst.__use, rt->rt6i_flags, 
-                           rt->rt6i_dev ? rt->rt6i_dev->name : "");
+       seq_printf(m, " %08x %08x %08x %08x %8s\n",
+                  rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
+                  rt->u.dst.__use, rt->rt6i_flags,
+                  rt->rt6i_dev ? rt->rt6i_dev->name : "");
        return 0;
 }
 
-static int rt6_proc_info(char *buffer, char **start, off_t offset, int length)
+static int ipv6_route_show(struct seq_file *m, void *v)
 {
-       struct rt6_proc_arg arg = {
-               .buffer = buffer,
-               .offset = offset,
-               .length = length,
-       };
-
-       fib6_clean_all(rt6_info_route, 0, &arg);
-
-       *start = buffer;
-       if (offset)
-               *start += offset % RT6_INFO_LEN;
-
-       arg.len -= offset % RT6_INFO_LEN;
-
-       if (arg.len > length)
-               arg.len = length;
-       if (arg.len < 0)
-               arg.len = 0;
+       fib6_clean_all(rt6_info_route, 0, m);
+       return 0;
+}
 
-       return arg.len;
+static int ipv6_route_open(struct inode *inode, struct file *file)
+{
+       return single_open(file, ipv6_route_show, NULL);
 }
 
+static const struct file_operations ipv6_route_proc_fops = {
+       .owner          = THIS_MODULE,
+       .open           = ipv6_route_open,
+       .read           = seq_read,
+       .llseek         = seq_lseek,
+       .release        = single_release,
+};
+
 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
 {
        seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
@@ -2264,7 +2346,7 @@ static int rt6_stats_seq_open(struct inode *inode, struct file *file)
        return single_open(file, rt6_stats_seq_show, NULL);
 }
 
-static struct file_operations rt6_stats_seq_fops = {
+static const struct file_operations rt6_stats_seq_fops = {
        .owner   = THIS_MODULE,
        .open    = rt6_stats_seq_open,
        .read    = seq_read,
@@ -2290,91 +2372,90 @@ int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
 }
 
 ctl_table ipv6_route_table[] = {
-        {
-               .ctl_name       =       NET_IPV6_ROUTE_FLUSH, 
+       {
                .procname       =       "flush",
-               .data           =       &flush_delay,
+               .data           =       &flush_delay,
                .maxlen         =       sizeof(int),
                .mode           =       0200,
-               .proc_handler   =       &ipv6_sysctl_rtcache_flush
+               .proc_handler   =       &ipv6_sysctl_rtcache_flush
        },
        {
                .ctl_name       =       NET_IPV6_ROUTE_GC_THRESH,
                .procname       =       "gc_thresh",
-               .data           =       &ip6_dst_ops.gc_thresh,
+               .data           =       &ip6_dst_ops.gc_thresh,
                .maxlen         =       sizeof(int),
                .mode           =       0644,
-               .proc_handler   =       &proc_dointvec,
+               .proc_handler   =       &proc_dointvec,
        },
        {
                .ctl_name       =       NET_IPV6_ROUTE_MAX_SIZE,
                .procname       =       "max_size",
-               .data           =       &ip6_rt_max_size,
+               .data           =       &ip6_rt_max_size,
                .maxlen         =       sizeof(int),
                .mode           =       0644,
-               .proc_handler   =       &proc_dointvec,
+               .proc_handler   =       &proc_dointvec,
        },
        {
                .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL,
                .procname       =       "gc_min_interval",
-               .data           =       &ip6_rt_gc_min_interval,
+               .data           =       &ip6_rt_gc_min_interval,
                .maxlen         =       sizeof(int),
                .mode           =       0644,
-               .proc_handler   =       &proc_dointvec_jiffies,
+               .proc_handler   =       &proc_dointvec_jiffies,
                .strategy       =       &sysctl_jiffies,
        },
        {
                .ctl_name       =       NET_IPV6_ROUTE_GC_TIMEOUT,
                .procname       =       "gc_timeout",
-               .data           =       &ip6_rt_gc_timeout,
+               .data           =       &ip6_rt_gc_timeout,
                .maxlen         =       sizeof(int),
                .mode           =       0644,
-               .proc_handler   =       &proc_dointvec_jiffies,
+               .proc_handler   =       &proc_dointvec_jiffies,
                .strategy       =       &sysctl_jiffies,
        },
        {
                .ctl_name       =       NET_IPV6_ROUTE_GC_INTERVAL,
                .procname       =       "gc_interval",
-               .data           =       &ip6_rt_gc_interval,
+               .data           =       &ip6_rt_gc_interval,
                .maxlen         =       sizeof(int),
                .mode           =       0644,
-               .proc_handler   =       &proc_dointvec_jiffies,
+               .proc_handler   =       &proc_dointvec_jiffies,
                .strategy       =       &sysctl_jiffies,
        },
        {
                .ctl_name       =       NET_IPV6_ROUTE_GC_ELASTICITY,
                .procname       =       "gc_elasticity",
-               .data           =       &ip6_rt_gc_elasticity,
+               .data           =       &ip6_rt_gc_elasticity,
                .maxlen         =       sizeof(int),
                .mode           =       0644,
-               .proc_handler   =       &proc_dointvec_jiffies,
+               .proc_handler   =       &proc_dointvec_jiffies,
                .strategy       =       &sysctl_jiffies,
        },
        {
                .ctl_name       =       NET_IPV6_ROUTE_MTU_EXPIRES,
                .procname       =       "mtu_expires",
-               .data           =       &ip6_rt_mtu_expires,
+               .data           =       &ip6_rt_mtu_expires,
                .maxlen         =       sizeof(int),
                .mode           =       0644,
-               .proc_handler   =       &proc_dointvec_jiffies,
+               .proc_handler   =       &proc_dointvec_jiffies,
                .strategy       =       &sysctl_jiffies,
        },
        {
                .ctl_name       =       NET_IPV6_ROUTE_MIN_ADVMSS,
                .procname       =       "min_adv_mss",
-               .data           =       &ip6_rt_min_advmss,
+               .data           =       &ip6_rt_min_advmss,
                .maxlen         =       sizeof(int),
                .mode           =       0644,
-               .proc_handler   =       &proc_dointvec_jiffies,
+               .proc_handler   =       &proc_dointvec_jiffies,
                .strategy       =       &sysctl_jiffies,
        },
        {
                .ctl_name       =       NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
                .procname       =       "gc_min_interval_ms",
-               .data           =       &ip6_rt_gc_min_interval,
+               .data           =       &ip6_rt_gc_min_interval,
                .maxlen         =       sizeof(int),
                .mode           =       0644,
-               .proc_handler   =       &proc_dointvec_ms_jiffies,
+               .proc_handler   =       &proc_dointvec_ms_jiffies,
                .strategy       =       &sysctl_ms_jiffies,
        },
        { .ctl_name = 0 }
@@ -2384,29 +2465,24 @@ ctl_table ipv6_route_table[] = {
 
 void __init ip6_route_init(void)
 {
-       struct proc_dir_entry *p;
-
-       ip6_dst_ops.kmem_cachep = kmem_cache_create("ip6_dst_cache",
-                                                    sizeof(struct rt6_info),
-                                                    0, SLAB_HWCACHE_ALIGN,
-                                                    NULL, NULL);
-       if (!ip6_dst_ops.kmem_cachep)
-               panic("cannot create ip6_dst_cache");
+       ip6_dst_ops.kmem_cachep =
+               kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
+                                 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
+       ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops.kmem_cachep;
 
        fib6_init();
-#ifdef         CONFIG_PROC_FS
-       p = proc_net_create("ipv6_route", 0, rt6_proc_info);
-       if (p)
-               p->owner = THIS_MODULE;
-
-       proc_net_fops_create("rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
-#endif
+       proc_net_fops_create(&init_net, "ipv6_route", 0, &ipv6_route_proc_fops);
+       proc_net_fops_create(&init_net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
 #ifdef CONFIG_XFRM
        xfrm6_init();
 #endif
 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
        fib6_rules_init();
 #endif
+
+       __rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL);
+       __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL);
+       __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL);
 }
 
 void ip6_route_cleanup(void)
@@ -2415,8 +2491,8 @@ void ip6_route_cleanup(void)
        fib6_rules_cleanup();
 #endif
 #ifdef CONFIG_PROC_FS
-       proc_net_remove("ipv6_route");
-       proc_net_remove("rt6_stats");
+       proc_net_remove(&init_net, "ipv6_route");
+       proc_net_remove(&init_net, "rt6_stats");
 #endif
 #ifdef CONFIG_XFRM
        xfrm6_fini();