]> bbs.cooldavid.org Git - net-next-2.6.git/blobdiff - net/ipv6/route.c
[NETLINK]: Do precise netlink message allocations where possible
[net-next-2.6.git] / net / ipv6 / route.c
index 73efdadb9ab89ab3ef98c63137237adaa359c447..a6472cb9054cb8193874ca7d4b03420e7e360c82 100644 (file)
@@ -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>
@@ -35,7 +37,6 @@
 #include <linux/netdevice.h>
 #include <linux/in6.h>
 #include <linux/init.h>
-#include <linux/netlink.h>
 #include <linux/if_arp.h>
 
 #ifdef         CONFIG_PROC_FS
@@ -54,6 +55,7 @@
 #include <net/dst.h>
 #include <net/xfrm.h>
 #include <net/netevent.h>
+#include <net/netlink.h>
 
 #include <asm/uaccess.h>
 
@@ -74,9 +76,6 @@
 
 #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;
@@ -140,6 +139,54 @@ struct rt6_info ip6_null_entry = {
        .rt6i_ref       = ATOMIC_INIT(1),
 };
 
+#ifdef CONFIG_IPV6_MULTIPLE_TABLES
+
+static int ip6_pkt_prohibit(struct sk_buff *skb);
+static int ip6_pkt_prohibit_out(struct sk_buff *skb);
+static int ip6_pkt_blk_hole(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_prohibit,
+                       .output         = ip6_pkt_prohibit_out,
+                       .ops            = &ip6_dst_ops,
+                       .path           = (struct dst_entry*)&ip6_prohibit_entry,
+               }
+       },
+       .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
+       .rt6i_metric    = ~(u32) 0,
+       .rt6i_ref       = ATOMIC_INIT(1),
+};
+
+struct rt6_info ip6_blk_hole_entry = {
+       .u = {
+               .dst = {
+                       .__refcnt       = ATOMIC_INIT(1),
+                       .__use          = 1,
+                       .dev            = &loopback_dev,
+                       .obsolete       = -1,
+                       .error          = -EINVAL,
+                       .metrics        = { [RTAX_HOPLIMIT - 1] = 255, },
+                       .input          = ip6_pkt_blk_hole,
+                       .output         = ip6_pkt_blk_hole,
+                       .ops            = &ip6_dst_ops,
+                       .path           = (struct dst_entry*)&ip6_blk_hole_entry,
+               }
+       },
+       .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
+       .rt6i_metric    = ~(u32) 0,
+       .rt6i_ref       = ATOMIC_INIT(1),
+};
+
+#endif
+
 /* allocate dst with ip6_dst_ops */
 static __inline__ struct rt6_info *ip6_dst_alloc(void)
 {
@@ -283,6 +330,8 @@ static int inline rt6_check_neigh(struct rt6_info *rt)
                read_lock_bh(&neigh->lock);
                if (neigh->nud_state & NUD_VALID)
                        m = 2;
+               else if (!(neigh->nud_state & NUD_FAILED))
+                       m = 1;
                read_unlock_bh(&neigh->lock);
        }
        return m;
@@ -294,15 +343,13 @@ static int rt6_score_route(struct rt6_info *rt, int oif,
        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;
 }
@@ -333,16 +380,17 @@ static struct rt6_info *rt6_select(struct rt6_info **head, int oif,
                        continue;
 
                if (m > mpri) {
-                       rt6_probe(match);
+                       if (strict & RT6_LOOKUP_F_REACHABLE)
+                               rt6_probe(match);
                        match = rt;
                        mpri = m;
-               } else {
+               } else if (strict & RT6_LOOKUP_F_REACHABLE) {
                        rt6_probe(rt);
                }
        }
 
        if (!match &&
-           (strict & RT6_SELECT_F_REACHABLE) &&
+           (strict & RT6_LOOKUP_F_REACHABLE) &&
            last && last != rt0) {
                /* no entries matched; do round-robin */
                static DEFINE_SPINLOCK(lock);
@@ -413,7 +461,7 @@ int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
        rt = rt6_get_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex);
 
        if (rt && !lifetime) {
-               ip6_del_rt(rt, NULL, NULL, NULL);
+               ip6_del_rt(rt);
                rt = NULL;
        }
 
@@ -437,17 +485,23 @@ 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(pn->subtree, 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)
@@ -459,10 +513,10 @@ 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_hold(&rt->u.dst);
        read_unlock_bh(&table->tb6_lock);
 
        rt->u.dst.lastuse = jiffies;
@@ -480,12 +534,16 @@ 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)
@@ -502,20 +560,24 @@ struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
    be destroyed.
  */
 
-int ip6_ins_rt(struct rt6_info *rt, struct nlmsghdr *nlh,
-               void *_rtattr, struct netlink_skb_parms *req)
+static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
 {
        int err;
        struct fib6_table *table;
 
        table = rt->rt6i_table;
        write_lock_bh(&table->tb6_lock);
-       err = fib6_add(&table->tb6_root, rt, nlh, _rtattr, req);
+       err = fib6_add(&table->tb6_root, rt, info);
        write_unlock_bh(&table->tb6_lock);
 
        return err;
 }
 
+int ip6_ins_rt(struct rt6_info *rt)
+{
+       return __ip6_ins_rt(rt, NULL);
+}
+
 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
                                      struct in6_addr *saddr)
 {
@@ -561,26 +623,23 @@ 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;
 }
 
-struct rt6_info *ip6_pol_route_input(struct fib6_table *table, struct flowi *fl,
-                                    int flags)
+static struct rt6_info *ip6_pol_route_input(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;
+       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);
@@ -590,7 +649,7 @@ restart_2:
 
 restart:
        rt = rt6_select(&fn->leaf, fl->iif, strict | reachable);
-       BACKTRACK();
+       BACKTRACK(&fl->fl6_src);
        if (rt == &ip6_null_entry ||
            rt->rt6i_flags & RTF_CACHE)
                goto out;
@@ -613,7 +672,7 @@ restart:
 
        dst_hold(&rt->u.dst);
        if (nrt) {
-               err = ip6_ins_rt(nrt, NULL, NULL, NULL);
+               err = ip6_ins_rt(nrt);
                if (!err)
                        goto out2;
        }
@@ -645,21 +704,22 @@ out2:
 void ip6_route_input(struct sk_buff *skb)
 {
        struct ipv6hdr *iph = skb->nh.ipv6h;
+       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);
 }
@@ -672,10 +732,9 @@ static struct rt6_info *ip6_pol_route_output(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);
@@ -685,7 +744,7 @@ restart_2:
 
 restart:
        rt = rt6_select(&fn->leaf, fl->oif, strict | reachable);
-       BACKTRACK();
+       BACKTRACK(&fl->fl6_src);
        if (rt == &ip6_null_entry ||
            rt->rt6i_flags & RTF_CACHE)
                goto out;
@@ -708,7 +767,7 @@ restart:
 
        dst_hold(&rt->u.dst);
        if (nrt) {
-               err = ip6_ins_rt(nrt, NULL, NULL, NULL);
+               err = ip6_ins_rt(nrt);
                if (!err)
                        goto out2;
        }
@@ -741,7 +800,10 @@ struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
        int flags = 0;
 
        if (rt6_need_strict(&fl->fl6_dst))
-               flags |= RT6_F_STRICT;
+               flags |= RT6_LOOKUP_F_IFACE;
+
+       if (!ipv6_addr_any(&fl->fl6_src))
+               flags |= RT6_LOOKUP_F_HAS_SADDR;
 
        return fib6_rule_lookup(fl, flags, ip6_pol_route_output);
 }
@@ -769,7 +831,7 @@ static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
 
        if (rt) {
                if (rt->rt6i_flags & RTF_CACHE)
-                       ip6_del_rt(rt, NULL, NULL, NULL);
+                       ip6_del_rt(rt);
                else
                        dst_release(dst);
        }
@@ -828,7 +890,7 @@ static inline unsigned int ipv6_advmss(unsigned int mtu)
 }
 
 static struct dst_entry *ndisc_dst_gc_list;
-DEFINE_SPINLOCK(ndisc_lock);
+static DEFINE_SPINLOCK(ndisc_lock);
 
 struct dst_entry *ndisc_dst_alloc(struct net_device *dev, 
                                  struct neighbour *neigh,
@@ -878,7 +940,7 @@ 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)
@@ -965,30 +1027,24 @@ int ipv6_get_hoplimit(struct net_device *dev)
  *
  */
 
-int ip6_route_add(struct in6_rtmsg *rtmsg, struct nlmsghdr *nlh, 
-                 void *_rtattr, struct netlink_skb_parms *req,
-                 u32 table_id)
+int ip6_route_add(struct fib6_config *cfg)
 {
        int err;
-       struct rtmsg *r;
-       struct rtattr **rta;
        struct rt6_info *rt = NULL;
        struct net_device *dev = NULL;
        struct inet6_dev *idev = NULL;
        struct fib6_table *table;
        int addr_type;
 
-       rta = (struct rtattr **) _rtattr;
-
-       if (rtmsg->rtmsg_dst_len > 128 || rtmsg->rtmsg_src_len > 128)
+       if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
                return -EINVAL;
 #ifndef CONFIG_IPV6_SUBTREES
-       if (rtmsg->rtmsg_src_len)
+       if (cfg->fc_src_len)
                return -EINVAL;
 #endif
-       if (rtmsg->rtmsg_ifindex) {
+       if (cfg->fc_ifindex) {
                err = -ENODEV;
-               dev = dev_get_by_index(rtmsg->rtmsg_ifindex);
+               dev = dev_get_by_index(cfg->fc_ifindex);
                if (!dev)
                        goto out;
                idev = in6_dev_get(dev);
@@ -996,10 +1052,10 @@ int ip6_route_add(struct in6_rtmsg *rtmsg, struct nlmsghdr *nlh,
                        goto out;
        }
 
-       if (rtmsg->rtmsg_metric == 0)
-               rtmsg->rtmsg_metric = IP6_RT_PRIO_USER;
+       if (cfg->fc_metric == 0)
+               cfg->fc_metric = IP6_RT_PRIO_USER;
 
-       table = fib6_new_table(table_id);
+       table = fib6_new_table(cfg->fc_table);
        if (table == NULL) {
                err = -ENOBUFS;
                goto out;
@@ -1013,14 +1069,13 @@ int ip6_route_add(struct in6_rtmsg *rtmsg, struct nlmsghdr *nlh,
        }
 
        rt->u.dst.obsolete = -1;
-       rt->rt6i_expires = jiffies + clock_t_to_jiffies(rtmsg->rtmsg_info);
-       if (nlh && (r = NLMSG_DATA(nlh))) {
-               rt->rt6i_protocol = r->rtm_protocol;
-       } else {
-               rt->rt6i_protocol = RTPROT_BOOT;
-       }
+       rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
 
-       addr_type = ipv6_addr_type(&rtmsg->rtmsg_dst);
+       if (cfg->fc_protocol == RTPROT_UNSPEC)
+               cfg->fc_protocol = RTPROT_BOOT;
+       rt->rt6i_protocol = cfg->fc_protocol;
+
+       addr_type = ipv6_addr_type(&cfg->fc_dst);
 
        if (addr_type & IPV6_ADDR_MULTICAST)
                rt->u.dst.input = ip6_mc_input;
@@ -1029,24 +1084,22 @@ int ip6_route_add(struct in6_rtmsg *rtmsg, struct nlmsghdr *nlh,
 
        rt->u.dst.output = ip6_output;
 
-       ipv6_addr_prefix(&rt->rt6i_dst.addr, 
-                        &rtmsg->rtmsg_dst, rtmsg->rtmsg_dst_len);
-       rt->rt6i_dst.plen = rtmsg->rtmsg_dst_len;
+       ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
+       rt->rt6i_dst.plen = cfg->fc_dst_len;
        if (rt->rt6i_dst.plen == 128)
               rt->u.dst.flags = DST_HOST;
 
 #ifdef CONFIG_IPV6_SUBTREES
-       ipv6_addr_prefix(&rt->rt6i_src.addr, 
-                        &rtmsg->rtmsg_src, rtmsg->rtmsg_src_len);
-       rt->rt6i_src.plen = rtmsg->rtmsg_src_len;
+       ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
+       rt->rt6i_src.plen = cfg->fc_src_len;
 #endif
 
-       rt->rt6i_metric = rtmsg->rtmsg_metric;
+       rt->rt6i_metric = cfg->fc_metric;
 
        /* We cannot add true routes via loopback here,
           they would result in kernel looping; promote them to reject routes
         */
-       if ((rtmsg->rtmsg_flags&RTF_REJECT) ||
+       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) {
@@ -1069,12 +1122,12 @@ int ip6_route_add(struct in6_rtmsg *rtmsg, struct nlmsghdr *nlh,
                goto install_route;
        }
 
-       if (rtmsg->rtmsg_flags & RTF_GATEWAY) {
+       if (cfg->fc_flags & RTF_GATEWAY) {
                struct in6_addr *gw_addr;
                int gwa_type;
 
-               gw_addr = &rtmsg->rtmsg_gateway;
-               ipv6_addr_copy(&rt->rt6i_gateway, &rtmsg->rtmsg_gateway);
+               gw_addr = &cfg->fc_gateway;
+               ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
                gwa_type = ipv6_addr_type(gw_addr);
 
                if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
@@ -1091,7 +1144,7 @@ int ip6_route_add(struct in6_rtmsg *rtmsg, struct nlmsghdr *nlh,
                        if (!(gwa_type&IPV6_ADDR_UNICAST))
                                goto out;
 
-                       grt = rt6_lookup(gw_addr, NULL, rtmsg->rtmsg_ifindex, 1);
+                       grt = rt6_lookup(gw_addr, NULL, cfg->fc_ifindex, 1);
 
                        err = -EHOSTUNREACH;
                        if (grt == NULL)
@@ -1123,7 +1176,7 @@ int ip6_route_add(struct in6_rtmsg *rtmsg, struct nlmsghdr *nlh,
        if (dev == NULL)
                goto out;
 
-       if (rtmsg->rtmsg_flags & (RTF_GATEWAY|RTF_NONEXTHOP)) {
+       if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
                rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
                if (IS_ERR(rt->rt6i_nexthop)) {
                        err = PTR_ERR(rt->rt6i_nexthop);
@@ -1132,24 +1185,24 @@ int ip6_route_add(struct in6_rtmsg *rtmsg, struct nlmsghdr *nlh,
                }
        }
 
-       rt->rt6i_flags = rtmsg->rtmsg_flags;
+       rt->rt6i_flags = cfg->fc_flags;
 
 install_route:
-       if (rta && rta[RTA_METRICS-1]) {
-               int attrlen = RTA_PAYLOAD(rta[RTA_METRICS-1]);
-               struct rtattr *attr = RTA_DATA(rta[RTA_METRICS-1]);
-
-               while (RTA_OK(attr, attrlen)) {
-                       unsigned flavor = attr->rta_type;
-                       if (flavor) {
-                               if (flavor > RTAX_MAX) {
+       if (cfg->fc_mx) {
+               struct nlattr *nla;
+               int remaining;
+
+               nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
+                       int type = nla->nla_type;
+
+                       if (type) {
+                               if (type > RTAX_MAX) {
                                        err = -EINVAL;
                                        goto out;
                                }
-                               rt->u.dst.metrics[flavor-1] =
-                                       *(u32 *)RTA_DATA(attr);
+
+                               rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
                        }
-                       attr = RTA_NEXT(attr, attrlen);
                }
        }
 
@@ -1162,7 +1215,7 @@ install_route:
        rt->u.dst.dev = dev;
        rt->rt6i_idev = idev;
        rt->rt6i_table = table;
-       return ip6_ins_rt(rt, nlh, _rtattr, req);
+       return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
 
 out:
        if (dev)
@@ -1170,19 +1223,22 @@ out:
        if (idev)
                in6_dev_put(idev);
        if (rt)
-               dst_free((struct dst_entry *) rt);
+               dst_free(&rt->u.dst);
        return err;
 }
 
-int ip6_del_rt(struct rt6_info *rt, struct nlmsghdr *nlh, void *_rtattr, struct netlink_skb_parms *req)
+static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
 {
        int err;
        struct fib6_table *table;
 
+       if (rt == &ip6_null_entry)
+               return -ENOENT;
+
        table = rt->rt6i_table;
        write_lock_bh(&table->tb6_lock);
 
-       err = fib6_del(rt, nlh, _rtattr, req);
+       err = fib6_del(rt, info);
        dst_release(&rt->u.dst);
 
        write_unlock_bh(&table->tb6_lock);
@@ -1190,41 +1246,43 @@ int ip6_del_rt(struct rt6_info *rt, struct nlmsghdr *nlh, void *_rtattr, struct
        return err;
 }
 
-static int ip6_route_del(struct in6_rtmsg *rtmsg, struct nlmsghdr *nlh,
-                        void *_rtattr, struct netlink_skb_parms *req,
-                        u32 table_id)
+int ip6_del_rt(struct rt6_info *rt)
+{
+       return __ip6_del_rt(rt, NULL);
+}
+
+static int ip6_route_del(struct fib6_config *cfg)
 {
        struct fib6_table *table;
        struct fib6_node *fn;
        struct rt6_info *rt;
        int err = -ESRCH;
 
-       table = fib6_get_table(table_id);
+       table = fib6_get_table(cfg->fc_table);
        if (table == NULL)
                return err;
 
        read_lock_bh(&table->tb6_lock);
 
        fn = fib6_locate(&table->tb6_root,
-                        &rtmsg->rtmsg_dst, rtmsg->rtmsg_dst_len,
-                        &rtmsg->rtmsg_src, rtmsg->rtmsg_src_len);
+                        &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) {
-                       if (rtmsg->rtmsg_ifindex &&
+                       if (cfg->fc_ifindex &&
                            (rt->rt6i_dev == NULL ||
-                            rt->rt6i_dev->ifindex != rtmsg->rtmsg_ifindex))
+                            rt->rt6i_dev->ifindex != cfg->fc_ifindex))
                                continue;
-                       if (rtmsg->rtmsg_flags&RTF_GATEWAY &&
-                           !ipv6_addr_equal(&rtmsg->rtmsg_gateway, &rt->rt6i_gateway))
+                       if (cfg->fc_flags & RTF_GATEWAY &&
+                           !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
                                continue;
-                       if (rtmsg->rtmsg_metric &&
-                           rtmsg->rtmsg_metric != rt->rt6i_metric)
+                       if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
                                continue;
                        dst_hold(&rt->u.dst);
                        read_unlock_bh(&table->tb6_lock);
 
-                       return ip6_del_rt(rt, nlh, _rtattr, req);
+                       return __ip6_del_rt(rt, &cfg->fc_nlinfo);
                }
        }
        read_unlock_bh(&table->tb6_lock);
@@ -1235,18 +1293,18 @@ static int ip6_route_del(struct in6_rtmsg *rtmsg, struct nlmsghdr *nlh,
 /*
  *     Handle redirects
  */
-void rt6_redirect(struct in6_addr *dest, 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
@@ -1260,7 +1318,7 @@ void rt6_redirect(struct in6_addr *dest, struct in6_addr *saddr,
         */
 
        read_lock_bh(&table->tb6_lock);
-       fn = fib6_lookup(&table->tb6_root, dest, NULL);
+       fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
 restart:
        for (rt = fn->leaf; rt; rt = rt->u.next) {
                /*
@@ -1275,29 +1333,63 @@ 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;
        }
 
        /*
@@ -1340,7 +1432,7 @@ restart:
        nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
        nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&nrt->u.dst));
 
-       if (ip6_ins_rt(nrt, NULL, NULL, NULL))
+       if (ip6_ins_rt(nrt))
                goto out;
 
        netevent.old = &rt->u.dst;
@@ -1348,7 +1440,7 @@ restart:
        call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
 
        if (rt->rt6i_flags&RTF_CACHE) {
-               ip6_del_rt(rt, NULL, NULL, NULL);
+               ip6_del_rt(rt);
                return;
        }
 
@@ -1430,7 +1522,7 @@ void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
                dst_set_expires(&nrt->u.dst, ip6_rt_mtu_expires);
                nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
 
-               ip6_ins_rt(nrt, NULL, NULL, NULL);
+               ip6_ins_rt(nrt);
        }
 out:
        dst_release(&rt->u.dst);
@@ -1449,6 +1541,7 @@ 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);
@@ -1507,21 +1600,23 @@ static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixle
                                           struct in6_addr *gwaddr, int ifindex,
                                           unsigned pref)
 {
-       struct in6_rtmsg rtmsg;
+       struct fib6_config cfg = {
+               .fc_table       = RT6_TABLE_INFO,
+               .fc_metric      = 1024,
+               .fc_ifindex     = ifindex,
+               .fc_dst_len     = prefixlen,
+               .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
+                                 RTF_UP | RTF_PREF(pref),
+       };
+
+       ipv6_addr_copy(&cfg.fc_dst, prefix);
+       ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
 
-       memset(&rtmsg, 0, sizeof(rtmsg));
-       rtmsg.rtmsg_type = RTMSG_NEWROUTE;
-       ipv6_addr_copy(&rtmsg.rtmsg_dst, prefix);
-       rtmsg.rtmsg_dst_len = prefixlen;
-       ipv6_addr_copy(&rtmsg.rtmsg_gateway, gwaddr);
-       rtmsg.rtmsg_metric = 1024;
-       rtmsg.rtmsg_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO | RTF_UP | RTF_PREF(pref);
        /* We should treat it as a default route if prefix length is 0. */
        if (!prefixlen)
-               rtmsg.rtmsg_flags |= RTF_DEFAULT;
-       rtmsg.rtmsg_ifindex = ifindex;
+               cfg.fc_flags |= RTF_DEFAULT;
 
-       ip6_route_add(&rtmsg, NULL, NULL, NULL, RT6_TABLE_INFO);
+       ip6_route_add(&cfg);
 
        return rt6_get_route_info(prefix, prefixlen, gwaddr, ifindex);
 }
@@ -1553,18 +1648,18 @@ struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
                                     struct net_device *dev,
                                     unsigned int pref)
 {
-       struct in6_rtmsg rtmsg;
+       struct fib6_config cfg = {
+               .fc_table       = RT6_TABLE_DFLT,
+               .fc_metric      = 1024,
+               .fc_ifindex     = dev->ifindex,
+               .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
+                                 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
+       };
 
-       memset(&rtmsg, 0, sizeof(struct in6_rtmsg));
-       rtmsg.rtmsg_type = RTMSG_NEWROUTE;
-       ipv6_addr_copy(&rtmsg.rtmsg_gateway, gwaddr);
-       rtmsg.rtmsg_metric = 1024;
-       rtmsg.rtmsg_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT | RTF_UP | RTF_EXPIRES |
-                           RTF_PREF(pref);
+       ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
 
-       rtmsg.rtmsg_ifindex = dev->ifindex;
+       ip6_route_add(&cfg);
 
-       ip6_route_add(&rtmsg, NULL, NULL, NULL, RT6_TABLE_DFLT);
        return rt6_get_dflt_router(gwaddr, dev);
 }
 
@@ -1584,15 +1679,34 @@ restart:
                if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
                        dst_hold(&rt->u.dst);
                        read_unlock_bh(&table->tb6_lock);
-                       ip6_del_rt(rt, NULL, NULL, NULL);
+                       ip6_del_rt(rt);
                        goto restart;
                }
        }
        read_unlock_bh(&table->tb6_lock);
 }
 
+static void rtmsg_to_fib6_config(struct in6_rtmsg *rtmsg,
+                                struct fib6_config *cfg)
+{
+       memset(cfg, 0, sizeof(*cfg));
+
+       cfg->fc_table = RT6_TABLE_MAIN;
+       cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
+       cfg->fc_metric = rtmsg->rtmsg_metric;
+       cfg->fc_expires = rtmsg->rtmsg_info;
+       cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
+       cfg->fc_src_len = rtmsg->rtmsg_src_len;
+       cfg->fc_flags = rtmsg->rtmsg_flags;
+
+       ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
+       ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
+       ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
+}
+
 int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
 {
+       struct fib6_config cfg;
        struct in6_rtmsg rtmsg;
        int err;
 
@@ -1605,16 +1719,16 @@ int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
                                     sizeof(struct in6_rtmsg));
                if (err)
                        return -EFAULT;
-                       
+
+               rtmsg_to_fib6_config(&rtmsg, &cfg);
+
                rtnl_lock();
                switch (cmd) {
                case SIOCADDRT:
-                       err = ip6_route_add(&rtmsg, NULL, NULL, NULL,
-                                           RT6_TABLE_MAIN);
+                       err = ip6_route_add(&cfg);
                        break;
                case SIOCDELRT:
-                       err = ip6_route_del(&rtmsg, NULL, NULL, NULL,
-                                           RT6_TABLE_MAIN);
+                       err = ip6_route_del(&cfg);
                        break;
                default:
                        err = -EINVAL;
@@ -1631,24 +1745,50 @@ 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 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(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
 
-       IP6_INC_STATS(IPSTATS_MIB_OUTNOROUTES);
-       icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_NOROUTE, 0, skb->dev);
+       IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_OUTNOROUTES);
+       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);
+}
+
 static int ip6_pkt_discard_out(struct sk_buff *skb)
 {
        skb->dev = skb->dst->dev;
        return ip6_pkt_discard(skb);
 }
 
+#ifdef CONFIG_IPV6_MULTIPLE_TABLES
+
+static int ip6_pkt_prohibit(struct sk_buff *skb)
+{
+       return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED);
+}
+
+static int ip6_pkt_prohibit_out(struct sk_buff *skb)
+{
+       skb->dev = skb->dst->dev;
+       return ip6_pkt_prohibit(skb);
+}
+
+static int ip6_pkt_blk_hole(struct sk_buff *skb)
+{
+       kfree_skb(skb);
+       return 0;
+}
+
+#endif
+
 /*
  *     Allocate a dst for local (unicast / anycast) address.
  */
@@ -1682,7 +1822,7 @@ 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);
        }
 
@@ -1765,71 +1905,120 @@ void rt6_mtu_change(struct net_device *dev, unsigned mtu)
        fib6_clean_all(rt6_mtu_change_route, 0, &arg);
 }
 
-static int inet6_rtm_to_rtmsg(struct rtmsg *r, struct rtattr **rta,
-                             struct in6_rtmsg *rtmsg)
+static struct nla_policy rtm_ipv6_policy[RTA_MAX+1] __read_mostly = {
+       [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
+       [RTA_OIF]               = { .type = NLA_U32 },
+       [RTA_IIF]               = { .type = NLA_U32 },
+       [RTA_PRIORITY]          = { .type = NLA_U32 },
+       [RTA_METRICS]           = { .type = NLA_NESTED },
+};
+
+static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
+                             struct fib6_config *cfg)
 {
-       memset(rtmsg, 0, sizeof(*rtmsg));
+       struct rtmsg *rtm;
+       struct nlattr *tb[RTA_MAX+1];
+       int err;
 
-       rtmsg->rtmsg_dst_len = r->rtm_dst_len;
-       rtmsg->rtmsg_src_len = r->rtm_src_len;
-       rtmsg->rtmsg_flags = RTF_UP;
-       if (r->rtm_type == RTN_UNREACHABLE)
-               rtmsg->rtmsg_flags |= RTF_REJECT;
+       err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
+       if (err < 0)
+               goto errout;
 
-       if (rta[RTA_GATEWAY-1]) {
-               if (rta[RTA_GATEWAY-1]->rta_len != RTA_LENGTH(16))
-                       return -EINVAL;
-               memcpy(&rtmsg->rtmsg_gateway, RTA_DATA(rta[RTA_GATEWAY-1]), 16);
-               rtmsg->rtmsg_flags |= RTF_GATEWAY;
-       }
-       if (rta[RTA_DST-1]) {
-               if (RTA_PAYLOAD(rta[RTA_DST-1]) < ((r->rtm_dst_len+7)>>3))
-                       return -EINVAL;
-               memcpy(&rtmsg->rtmsg_dst, RTA_DATA(rta[RTA_DST-1]), ((r->rtm_dst_len+7)>>3));
+       err = -EINVAL;
+       rtm = nlmsg_data(nlh);
+       memset(cfg, 0, sizeof(*cfg));
+
+       cfg->fc_table = rtm->rtm_table;
+       cfg->fc_dst_len = rtm->rtm_dst_len;
+       cfg->fc_src_len = rtm->rtm_src_len;
+       cfg->fc_flags = RTF_UP;
+       cfg->fc_protocol = rtm->rtm_protocol;
+
+       if (rtm->rtm_type == RTN_UNREACHABLE)
+               cfg->fc_flags |= RTF_REJECT;
+
+       cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
+       cfg->fc_nlinfo.nlh = nlh;
+
+       if (tb[RTA_GATEWAY]) {
+               nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
+               cfg->fc_flags |= RTF_GATEWAY;
        }
-       if (rta[RTA_SRC-1]) {
-               if (RTA_PAYLOAD(rta[RTA_SRC-1]) < ((r->rtm_src_len+7)>>3))
-                       return -EINVAL;
-               memcpy(&rtmsg->rtmsg_src, RTA_DATA(rta[RTA_SRC-1]), ((r->rtm_src_len+7)>>3));
+
+       if (tb[RTA_DST]) {
+               int plen = (rtm->rtm_dst_len + 7) >> 3;
+
+               if (nla_len(tb[RTA_DST]) < plen)
+                       goto errout;
+
+               nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
        }
-       if (rta[RTA_OIF-1]) {
-               if (rta[RTA_OIF-1]->rta_len != RTA_LENGTH(sizeof(int)))
-                       return -EINVAL;
-               memcpy(&rtmsg->rtmsg_ifindex, RTA_DATA(rta[RTA_OIF-1]), sizeof(int));
+
+       if (tb[RTA_SRC]) {
+               int plen = (rtm->rtm_src_len + 7) >> 3;
+
+               if (nla_len(tb[RTA_SRC]) < plen)
+                       goto errout;
+
+               nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
        }
-       if (rta[RTA_PRIORITY-1]) {
-               if (rta[RTA_PRIORITY-1]->rta_len != RTA_LENGTH(4))
-                       return -EINVAL;
-               memcpy(&rtmsg->rtmsg_metric, RTA_DATA(rta[RTA_PRIORITY-1]), 4);
+
+       if (tb[RTA_OIF])
+               cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
+
+       if (tb[RTA_PRIORITY])
+               cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
+
+       if (tb[RTA_METRICS]) {
+               cfg->fc_mx = nla_data(tb[RTA_METRICS]);
+               cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
        }
-       return 0;
+
+       if (tb[RTA_TABLE])
+               cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
+
+       err = 0;
+errout:
+       return err;
 }
 
 int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
 {
-       struct rtmsg *r = NLMSG_DATA(nlh);
-       struct in6_rtmsg rtmsg;
+       struct fib6_config cfg;
+       int err;
 
-       if (inet6_rtm_to_rtmsg(r, arg, &rtmsg))
-               return -EINVAL;
-       return ip6_route_del(&rtmsg, nlh, arg, &NETLINK_CB(skb), r->rtm_table);
+       err = rtm_to_fib6_config(skb, nlh, &cfg);
+       if (err < 0)
+               return err;
+
+       return ip6_route_del(&cfg);
 }
 
 int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
 {
-       struct rtmsg *r = NLMSG_DATA(nlh);
-       struct in6_rtmsg rtmsg;
+       struct fib6_config cfg;
+       int err;
 
-       if (inet6_rtm_to_rtmsg(r, arg, &rtmsg))
-               return -EINVAL;
-       return ip6_route_add(&rtmsg, nlh, arg, &NETLINK_CB(skb), r->rtm_table);
+       err = rtm_to_fib6_config(skb, nlh, &cfg);
+       if (err < 0)
+               return err;
+
+       return ip6_route_add(&cfg);
 }
 
-struct rt6_rtnl_dump_arg
+static inline size_t rt6_nlmsg_size(void)
 {
-       struct sk_buff *skb;
-       struct netlink_callback *cb;
-};
+       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 */
+              + 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,
@@ -1837,9 +2026,9 @@ static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
                         int prefix, unsigned int flags)
 {
        struct rtmsg *rtm;
-       struct nlmsghdr  *nlh;
-       unsigned char    *b = skb->tail;
+       struct nlmsghdr *nlh;
        struct rta_cacheinfo ci;
+       u32 table;
 
        if (prefix) {   /* user wants prefix routes only */
                if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
@@ -1848,17 +2037,21 @@ static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
                }
        }
 
-       nlh = NLMSG_NEW(skb, pid, seq, type, sizeof(*rtm), flags);
-       rtm = NLMSG_DATA(nlh);
+       nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
+       if (nlh == NULL)
+               return -ENOBUFS;
+
+       rtm = nlmsg_data(nlh);
        rtm->rtm_family = AF_INET6;
        rtm->rtm_dst_len = rt->rt6i_dst.plen;
        rtm->rtm_src_len = rt->rt6i_src.plen;
        rtm->rtm_tos = 0;
        if (rt->rt6i_table)
-               rtm->rtm_table = rt->rt6i_table->tb6_id;
+               table = rt->rt6i_table->tb6_id;
        else
-               rtm->rtm_table = RT6_TABLE_UNSPEC;
-       rtm->rtm_table = RT_TABLE_MAIN;
+               table = RT6_TABLE_UNSPEC;
+       rtm->rtm_table = table;
+       NLA_PUT_U32(skb, RTA_TABLE, table);
        if (rt->rt6i_flags&RTF_REJECT)
                rtm->rtm_type = RTN_UNREACHABLE;
        else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
@@ -1879,31 +2072,35 @@ static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
                rtm->rtm_flags |= RTM_F_CLONED;
 
        if (dst) {
-               RTA_PUT(skb, RTA_DST, 16, dst);
+               NLA_PUT(skb, RTA_DST, 16, dst);
                rtm->rtm_dst_len = 128;
        } else if (rtm->rtm_dst_len)
-               RTA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
+               NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
 #ifdef CONFIG_IPV6_SUBTREES
        if (src) {
-               RTA_PUT(skb, RTA_SRC, 16, src);
+               NLA_PUT(skb, RTA_SRC, 16, src);
                rtm->rtm_src_len = 128;
        } else if (rtm->rtm_src_len)
-               RTA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
+               NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
 #endif
        if (iif)
-               RTA_PUT(skb, RTA_IIF, 4, &iif);
+               NLA_PUT_U32(skb, RTA_IIF, iif);
        else if (dst) {
                struct in6_addr saddr_buf;
                if (ipv6_get_saddr(&rt->u.dst, dst, &saddr_buf) == 0)
-                       RTA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
+                       NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
        }
+
        if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
-               goto rtattr_failure;
+               goto nla_put_failure;
+
        if (rt->u.dst.neighbour)
-               RTA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
+               NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
+
        if (rt->u.dst.dev)
-               RTA_PUT(skb, RTA_OIF, sizeof(int), &rt->rt6i_dev->ifindex);
-       RTA_PUT(skb, RTA_PRIORITY, 4, &rt->rt6i_metric);
+               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);
@@ -1915,23 +2112,21 @@ static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
        ci.rta_id = 0;
        ci.rta_ts = 0;
        ci.rta_tsage = 0;
-       RTA_PUT(skb, RTA_CACHEINFO, sizeof(ci), &ci);
-       nlh->nlmsg_len = skb->tail - b;
-       return skb->len;
+       NLA_PUT(skb, RTA_CACHEINFO, sizeof(ci), &ci);
+
+       return nlmsg_end(skb, nlh);
 
-nlmsg_failure:
-rtattr_failure:
-       skb_trim(skb, b - skb->data);
-       return -1;
+nla_put_failure:
+       return nlmsg_cancel(skb, nlh);
 }
 
-static int rt6_dump_route(struct rt6_info *rt, void *p_arg)
+int rt6_dump_route(struct rt6_info *rt, void *p_arg)
 {
        struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
        int prefix;
 
-       if (arg->cb->nlh->nlmsg_len >= NLMSG_LENGTH(sizeof(struct rtmsg))) {
-               struct rtmsg *rtm = NLMSG_DATA(arg->cb->nlh);
+       if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
+               struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
                prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
        } else
                prefix = 0;
@@ -1941,219 +2136,105 @@ static int rt6_dump_route(struct rt6_info *rt, void *p_arg)
                     prefix, NLM_F_MULTI);
 }
 
-static int fib6_dump_node(struct fib6_walker_t *w)
+int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
 {
-       int res;
+       struct nlattr *tb[RTA_MAX+1];
        struct rt6_info *rt;
+       struct sk_buff *skb;
+       struct rtmsg *rtm;
+       struct flowi fl;
+       int err, iif = 0;
 
-       for (rt = w->leaf; rt; rt = rt->u.next) {
-               res = rt6_dump_route(rt, w->args);
-               if (res < 0) {
-                       /* Frame is full, suspend walking */
-                       w->leaf = rt;
-                       return 1;
-               }
-               BUG_TRAP(res!=0);
-       }
-       w->leaf = NULL;
-       return 0;
-}
-
-static void fib6_dump_end(struct netlink_callback *cb)
-{
-       struct fib6_walker_t *w = (void*)cb->args[0];
-
-       if (w) {
-               cb->args[0] = 0;
-               kfree(w);
-       }
-       cb->done = (void*)cb->args[1];
-       cb->args[1] = 0;
-}
-
-static int fib6_dump_done(struct netlink_callback *cb)
-{
-       fib6_dump_end(cb);
-       return cb->done ? cb->done(cb) : 0;
-}
-
-int inet6_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
-{
-       struct fib6_table *table;
-       struct rt6_rtnl_dump_arg arg;
-       struct fib6_walker_t *w;
-       int i, res = 0;
-
-       arg.skb = skb;
-       arg.cb = cb;
+       err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
+       if (err < 0)
+               goto errout;
 
-       /*
-        * cb->args[0] = pointer to walker structure
-        * cb->args[1] = saved cb->done() pointer
-        * cb->args[2] = current table being dumped
-        */
-
-       w = (void*)cb->args[0];
-       if (w == NULL) {
-               /* New dump:
-                * 
-                * 1. hook callback destructor.
-                */
-               cb->args[1] = (long)cb->done;
-               cb->done = fib6_dump_done;
+       err = -EINVAL;
+       memset(&fl, 0, sizeof(fl));
 
-               /*
-                * 2. allocate and initialize walker.
-                */
-               w = kzalloc(sizeof(*w), GFP_ATOMIC);
-               if (w == NULL)
-                       return -ENOMEM;
-               w->func = fib6_dump_node;
-               w->args = &arg;
-               cb->args[0] = (long)w;
-               cb->args[2] = FIB6_TABLE_MIN;
-       } else {
-               w->args = &arg;
-               i = cb->args[2];
-               if (i > FIB6_TABLE_MAX)
-                       goto end;
-
-               table = fib6_get_table(i);
-               if (table != NULL) {
-                       read_lock_bh(&table->tb6_lock);
-                       w->root = &table->tb6_root;
-                       res = fib6_walk_continue(w);
-                       read_unlock_bh(&table->tb6_lock);
-                       if (res != 0) {
-                               if (res < 0)
-                                       fib6_walker_unlink(w);
-                               goto end;
-                       }
-               }
+       if (tb[RTA_SRC]) {
+               if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
+                       goto errout;
 
-               fib6_walker_unlink(w);
-               cb->args[2] = ++i;
+               ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
        }
 
-       for (i = cb->args[2]; i <= FIB6_TABLE_MAX; i++) {
-               table = fib6_get_table(i);
-               if (table == NULL)
-                       continue;
+       if (tb[RTA_DST]) {
+               if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
+                       goto errout;
 
-               read_lock_bh(&table->tb6_lock);
-               w->root = &table->tb6_root;
-               res = fib6_walk(w);
-               read_unlock_bh(&table->tb6_lock);
-               if (res)
-                       break;
+               ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
        }
-end:
-       cb->args[2] = i;
-
-       res = res < 0 ? res : skb->len;
-       /* res < 0 is an error. (really, impossible)
-          res == 0 means that dump is complete, but skb still can contain data.
-          res > 0 dump is not complete, but frame is full.
-        */
-       /* Destroy walker, if dump of this table is complete. */
-       if (res <= 0)
-               fib6_dump_end(cb);
-       return res;
-}
-
-int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
-{
-       struct rtattr **rta = arg;
-       int iif = 0;
-       int err = -ENOBUFS;
-       struct sk_buff *skb;
-       struct flowi fl;
-       struct rt6_info *rt;
-
-       skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
-       if (skb == NULL)
-               goto out;
-
-       /* Reserve room for dummy headers, this skb can pass
-          through good chunk of routing engine.
-        */
-       skb->mac.raw = skb->data;
-       skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
 
-       memset(&fl, 0, sizeof(fl));
-       if (rta[RTA_SRC-1])
-               ipv6_addr_copy(&fl.fl6_src,
-                              (struct in6_addr*)RTA_DATA(rta[RTA_SRC-1]));
-       if (rta[RTA_DST-1])
-               ipv6_addr_copy(&fl.fl6_dst,
-                              (struct in6_addr*)RTA_DATA(rta[RTA_DST-1]));
+       if (tb[RTA_IIF])
+               iif = nla_get_u32(tb[RTA_IIF]);
 
-       if (rta[RTA_IIF-1])
-               memcpy(&iif, RTA_DATA(rta[RTA_IIF-1]), sizeof(int));
+       if (tb[RTA_OIF])
+               fl.oif = nla_get_u32(tb[RTA_OIF]);
 
        if (iif) {
                struct net_device *dev;
                dev = __dev_get_by_index(iif);
                if (!dev) {
                        err = -ENODEV;
-                       goto out_free;
+                       goto errout;
                }
        }
 
-       fl.oif = 0;
-       if (rta[RTA_OIF-1])
-               memcpy(&fl.oif, RTA_DATA(rta[RTA_OIF-1]), sizeof(int));
+       skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
+       if (skb == NULL) {
+               err = -ENOBUFS;
+               goto errout;
+       }
 
-       rt = (struct rt6_info*)ip6_route_output(NULL, &fl);
+       /* Reserve room for dummy headers, this skb can pass
+          through good chunk of routing engine.
+        */
+       skb->mac.raw = skb->data;
+       skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
 
+       rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
        skb->dst = &rt->u.dst;
 
-       NETLINK_CB(skb).dst_pid = NETLINK_CB(in_skb).pid;
-       err = rt6_fill_node(skb, rt, 
-                           &fl.fl6_dst, &fl.fl6_src,
-                           iif,
+       err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
                            RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
                            nlh->nlmsg_seq, 0, 0);
        if (err < 0) {
-               err = -EMSGSIZE;
-               goto out_free;
+               kfree_skb(skb);
+               goto errout;
        }
 
-       err = netlink_unicast(rtnl, skb, NETLINK_CB(in_skb).pid, MSG_DONTWAIT);
-       if (err > 0)
-               err = 0;
-out:
+       err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
+errout:
        return err;
-out_free:
-       kfree_skb(skb);
-       goto out;       
 }
 
-void inet6_rt_notify(int event, struct rt6_info *rt, struct nlmsghdr *nlh, 
-                       struct netlink_skb_parms *req)
+void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
 {
        struct sk_buff *skb;
-       int size = NLMSG_SPACE(sizeof(struct rtmsg)+256);
-       u32 pid = current->pid;
-       u32 seq = 0;
-
-       if (req)
-               pid = req->pid;
-       if (nlh)
-               seq = nlh->nlmsg_seq;
-       
-       skb = alloc_skb(size, gfp_any());
-       if (!skb) {
-               netlink_set_err(rtnl, 0, RTNLGRP_IPV6_ROUTE, ENOBUFS);
-               return;
-       }
-       if (rt6_fill_node(skb, rt, NULL, NULL, 0, event, pid, seq, 0, 0) < 0) {
-               kfree_skb(skb);
-               netlink_set_err(rtnl, 0, RTNLGRP_IPV6_ROUTE, EINVAL);
-               return;
+       u32 pid = 0, seq = 0;
+       struct nlmsghdr *nlh = NULL;
+       int err = -ENOBUFS;
+
+       if (info) {
+               pid = info->pid;
+               nlh = info->nlh;
+               if (nlh)
+                       seq = nlh->nlmsg_seq;
        }
-       NETLINK_CB(skb).dst_group = RTNLGRP_IPV6_ROUTE;
-       netlink_broadcast(rtnl, skb, 0, RTNLGRP_IPV6_ROUTE, 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);
+       /* failure implies BUG in rt6_nlmsg_size() */
+       BUG_ON(err < 0);
+
+       err = rtnl_notify(skb, pid, RTNLGRP_IPV6_ROUTE, nlh, gfp_any());
+errout:
+       if (err < 0)
+               rtnl_set_sk_err(RTNLGRP_IPV6_ROUTE, err);
 }
 
 /*
@@ -2176,7 +2257,6 @@ 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;
 
        if (arg->skip < arg->offset / RT6_INFO_LEN) {
                arg->skip++;
@@ -2186,38 +2266,28 @@ static int rt6_info_route(struct rt6_info *rt, void *p_arg)
        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 ",
+       arg->len += sprintf(arg->buffer + arg->len,
+                           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 ",
+       arg->len += sprintf(arg->buffer + arg->len,
+                           NIP6_SEQFMT " %02x ",
+                           NIP6(rt->rt6i_src.addr),
                            rt->rt6i_src.plen);
 #else
-       sprintf(arg->buffer + arg->len,
-               "00000000000000000000000000000000 00 ");
-       arg->len += 36;
+       arg->len += sprintf(arg->buffer + arg->len,
+                           "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;
-               }
+               arg->len += sprintf(arg->buffer + arg->len,
+                                   NIP6_SEQFMT,
+                                   NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
        } else {
-               sprintf(arg->buffer + arg->len,
-                       "00000000000000000000000000000000");
-               arg->len += 32;
+               arg->len += sprintf(arg->buffer + arg->len,
+                                   "00000000000000000000000000000000");
        }
        arg->len += sprintf(arg->buffer + arg->len,
                            " %08x %08x %08x %08x %8s\n",
@@ -2390,13 +2460,9 @@ 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, NULL);
        fib6_init();
 #ifdef         CONFIG_PROC_FS
        p = proc_net_create("ipv6_route", 0, rt6_proc_info);
@@ -2408,10 +2474,16 @@ void __init ip6_route_init(void)
 #ifdef CONFIG_XFRM
        xfrm6_init();
 #endif
+#ifdef CONFIG_IPV6_MULTIPLE_TABLES
+       fib6_rules_init();
+#endif
 }
 
 void ip6_route_cleanup(void)
 {
+#ifdef CONFIG_IPV6_MULTIPLE_TABLES
+       fib6_rules_cleanup();
+#endif
 #ifdef CONFIG_PROC_FS
        proc_net_remove("ipv6_route");
        proc_net_remove("rt6_stats");