]> bbs.cooldavid.org Git - net-next-2.6.git/blobdiff - net/ipv6/addrconf.c
[IPV6] ADDRCONF: Convert addrconf_lock to RCU.
[net-next-2.6.git] / net / ipv6 / addrconf.c
index 8ea1e36bf8ebbe7b5b9f161272a66cf28c35c3bb..c09ebb7bb98a3e20a8855e1f4136b7234284d614 100644 (file)
@@ -48,6 +48,7 @@
 #include <linux/net.h>
 #include <linux/in6.h>
 #include <linux/netdevice.h>
+#include <linux/if_addr.h>
 #include <linux/if_arp.h>
 #include <linux/if_arcnet.h>
 #include <linux/if_infiniband.h>
@@ -72,6 +73,7 @@
 #include <net/addrconf.h>
 #include <net/tcp.h>
 #include <net/ip.h>
+#include <net/netlink.h>
 #include <linux/if_tunnel.h>
 #include <linux/rtnetlink.h>
 
@@ -117,9 +119,6 @@ static int ipv6_count_addresses(struct inet6_dev *idev);
 static struct inet6_ifaddr             *inet6_addr_lst[IN6_ADDR_HSIZE];
 static DEFINE_RWLOCK(addrconf_hash_lock);
 
-/* Protects inet6 devices */
-DEFINE_RWLOCK(addrconf_lock);
-
 static void addrconf_verify(unsigned long);
 
 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
@@ -144,7 +143,7 @@ static int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *de
 
 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 
-struct ipv6_devconf ipv6_devconf = {
+struct ipv6_devconf ipv6_devconf __read_mostly = {
        .forwarding             = 0,
        .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
        .mtu6                   = IPV6_MIN_MTU,
@@ -173,9 +172,10 @@ struct ipv6_devconf ipv6_devconf = {
        .accept_ra_rt_info_max_plen = 0,
 #endif
 #endif
+       .proxy_ndp              = 0,
 };
 
-static struct ipv6_devconf ipv6_devconf_dflt = {
+static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
        .forwarding             = 0,
        .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
        .mtu6                   = IPV6_MIN_MTU,
@@ -203,6 +203,7 @@ static struct ipv6_devconf ipv6_devconf_dflt = {
        .accept_ra_rt_info_max_plen = 0,
 #endif
 #endif
+       .proxy_ndp              = 0,
 };
 
 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
@@ -314,6 +315,12 @@ static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 
 /* Nobody refers to this device, we may destroy it. */
 
+static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
+{
+       struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
+       kfree(idev);
+}
+
 void in6_dev_finish_destroy(struct inet6_dev *idev)
 {
        struct net_device *dev = idev->dev;
@@ -328,7 +335,7 @@ void in6_dev_finish_destroy(struct inet6_dev *idev)
                return;
        }
        snmp6_free_dev(idev);
-       kfree(idev);
+       call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 }
 
 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
@@ -404,9 +411,8 @@ static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
        if (netif_carrier_ok(dev))
                ndev->if_flags |= IF_READY;
 
-       write_lock_bh(&addrconf_lock);
-       dev->ip6_ptr = ndev;
-       write_unlock_bh(&addrconf_lock);
+       /* protected by rtnl_lock */
+       rcu_assign_pointer(dev->ip6_ptr, ndev);
 
        ipv6_mc_init_dev(ndev);
        ndev->tstamp = jiffies;
@@ -470,7 +476,7 @@ static void addrconf_forward_change(void)
 
        read_lock(&dev_base_lock);
        for (dev=dev_base; dev; dev=dev->next) {
-               read_lock(&addrconf_lock);
+               rcu_read_lock();
                idev = __in6_dev_get(dev);
                if (idev) {
                        int changed = (!idev->cnf.forwarding) ^ (!ipv6_devconf.forwarding);
@@ -478,7 +484,7 @@ static void addrconf_forward_change(void)
                        if (changed)
                                dev_forward_change(idev);
                }
-               read_unlock(&addrconf_lock);
+               rcu_read_unlock();
        }
        read_unlock(&dev_base_lock);
 }
@@ -539,7 +545,7 @@ ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
        int hash;
        int err = 0;
 
-       read_lock_bh(&addrconf_lock);
+       rcu_read_lock_bh();
        if (idev->dead) {
                err = -ENODEV;                  /*XXX*/
                goto out2;
@@ -578,6 +584,8 @@ ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
        ifa->flags = flags | IFA_F_TENTATIVE;
        ifa->cstamp = ifa->tstamp = jiffies;
 
+       ifa->rt = rt;
+
        ifa->idev = idev;
        in6_dev_hold(idev);
        /* For caller */
@@ -603,12 +611,10 @@ ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
        }
 #endif
 
-       ifa->rt = rt;
-
        in6_ifa_hold(ifa);
        write_unlock(&idev->lock);
 out2:
-       read_unlock_bh(&addrconf_lock);
+       rcu_read_unlock_bh();
 
        if (likely(err == 0))
                atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
@@ -734,7 +740,7 @@ static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 
                if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
                        if (onlink == 0) {
-                               ip6_del_rt(rt, NULL, NULL, NULL);
+                               ip6_del_rt(rt);
                                rt = NULL;
                        } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
                                rt->rt6i_expires = expires;
@@ -911,7 +917,7 @@ int ipv6_dev_get_saddr(struct net_device *daddr_dev,
        memset(&hiscore, 0, sizeof(hiscore));
 
        read_lock(&dev_base_lock);
-       read_lock(&addrconf_lock);
+       rcu_read_lock();
 
        for (dev = dev_base; dev; dev=dev->next) {
                struct inet6_dev *idev;
@@ -1123,7 +1129,7 @@ record_it:
                }
                read_unlock_bh(&idev->lock);
        }
-       read_unlock(&addrconf_lock);
+       rcu_read_unlock();
        read_unlock(&dev_base_lock);
 
        if (!ifa_result)
@@ -1147,7 +1153,7 @@ int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr)
        struct inet6_dev *idev;
        int err = -EADDRNOTAVAIL;
 
-       read_lock(&addrconf_lock);
+       rcu_read_lock();
        if ((idev = __in6_dev_get(dev)) != NULL) {
                struct inet6_ifaddr *ifp;
 
@@ -1161,7 +1167,7 @@ int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr)
                }
                read_unlock_bh(&idev->lock);
        }
-       read_unlock(&addrconf_lock);
+       rcu_read_unlock();
        return err;
 }
 
@@ -1462,7 +1468,7 @@ static void ipv6_regen_rndid(unsigned long data)
        struct inet6_dev *idev = (struct inet6_dev *) data;
        unsigned long expires;
 
-       read_lock_bh(&addrconf_lock);
+       rcu_read_lock_bh();
        write_lock_bh(&idev->lock);
 
        if (idev->dead)
@@ -1486,7 +1492,7 @@ static void ipv6_regen_rndid(unsigned long data)
 
 out:
        write_unlock_bh(&idev->lock);
-       read_unlock_bh(&addrconf_lock);
+       rcu_read_unlock_bh();
        in6_dev_put(idev);
 }
 
@@ -1507,59 +1513,56 @@ static void
 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
                      unsigned long expires, u32 flags)
 {
-       struct in6_rtmsg rtmsg;
+       struct fib6_config cfg = {
+               .fc_table = RT6_TABLE_PREFIX,
+               .fc_metric = IP6_RT_PRIO_ADDRCONF,
+               .fc_ifindex = dev->ifindex,
+               .fc_expires = expires,
+               .fc_dst_len = plen,
+               .fc_flags = RTF_UP | flags,
+       };
 
-       memset(&rtmsg, 0, sizeof(rtmsg));
-       ipv6_addr_copy(&rtmsg.rtmsg_dst, pfx);
-       rtmsg.rtmsg_dst_len = plen;
-       rtmsg.rtmsg_metric = IP6_RT_PRIO_ADDRCONF;
-       rtmsg.rtmsg_ifindex = dev->ifindex;
-       rtmsg.rtmsg_info = expires;
-       rtmsg.rtmsg_flags = RTF_UP|flags;
-       rtmsg.rtmsg_type = RTMSG_NEWROUTE;
+       ipv6_addr_copy(&cfg.fc_dst, pfx);
 
        /* Prevent useless cloning on PtP SIT.
           This thing is done here expecting that the whole
           class of non-broadcast devices need not cloning.
         */
-       if (dev->type == ARPHRD_SIT && (dev->flags&IFF_POINTOPOINT))
-               rtmsg.rtmsg_flags |= RTF_NONEXTHOP;
+       if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
+               cfg.fc_flags |= RTF_NONEXTHOP;
 
-       ip6_route_add(&rtmsg, NULL, NULL, NULL);
+       ip6_route_add(&cfg);
 }
 
 /* Create "default" multicast route to the interface */
 
 static void addrconf_add_mroute(struct net_device *dev)
 {
-       struct in6_rtmsg rtmsg;
+       struct fib6_config cfg = {
+               .fc_table = RT6_TABLE_LOCAL,
+               .fc_metric = IP6_RT_PRIO_ADDRCONF,
+               .fc_ifindex = dev->ifindex,
+               .fc_dst_len = 8,
+               .fc_flags = RTF_UP,
+       };
+
+       ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
 
-       memset(&rtmsg, 0, sizeof(rtmsg));
-       ipv6_addr_set(&rtmsg.rtmsg_dst,
-                     htonl(0xFF000000), 0, 0, 0);
-       rtmsg.rtmsg_dst_len = 8;
-       rtmsg.rtmsg_metric = IP6_RT_PRIO_ADDRCONF;
-       rtmsg.rtmsg_ifindex = dev->ifindex;
-       rtmsg.rtmsg_flags = RTF_UP;
-       rtmsg.rtmsg_type = RTMSG_NEWROUTE;
-       ip6_route_add(&rtmsg, NULL, NULL, NULL);
+       ip6_route_add(&cfg);
 }
 
 static void sit_route_add(struct net_device *dev)
 {
-       struct in6_rtmsg rtmsg;
-
-       memset(&rtmsg, 0, sizeof(rtmsg));
-
-       rtmsg.rtmsg_type        = RTMSG_NEWROUTE;
-       rtmsg.rtmsg_metric      = IP6_RT_PRIO_ADDRCONF;
+       struct fib6_config cfg = {
+               .fc_table = RT6_TABLE_MAIN,
+               .fc_metric = IP6_RT_PRIO_ADDRCONF,
+               .fc_ifindex = dev->ifindex,
+               .fc_dst_len = 96,
+               .fc_flags = RTF_UP | RTF_NONEXTHOP,
+       };
 
        /* prefix length - 96 bits "::d.d.d.d" */
-       rtmsg.rtmsg_dst_len     = 96;
-       rtmsg.rtmsg_flags       = RTF_UP|RTF_NONEXTHOP;
-       rtmsg.rtmsg_ifindex     = dev->ifindex;
-
-       ip6_route_add(&rtmsg, NULL, NULL, NULL);
+       ip6_route_add(&cfg);
 }
 
 static void addrconf_add_lroute(struct net_device *dev)
@@ -1660,7 +1663,7 @@ void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
                if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
                        if (rt->rt6i_flags&RTF_EXPIRES) {
                                if (valid_lft == 0) {
-                                       ip6_del_rt(rt, NULL, NULL, NULL);
+                                       ip6_del_rt(rt);
                                        rt = NULL;
                                } else {
                                        rt->rt6i_expires = jiffies + rt_expires;
@@ -1887,9 +1890,6 @@ static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
        if ((dev = __dev_get_by_index(ifindex)) == NULL)
                return -ENODEV;
        
-       if (!(dev->flags&IFF_UP))
-               return -ENETDOWN;
-
        if ((idev = addrconf_add_dev(dev)) == NULL)
                return -ENOBUFS;
 
@@ -1909,11 +1909,11 @@ static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
        ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
 
        if (!IS_ERR(ifp)) {
-               spin_lock(&ifp->lock);
+               spin_lock_bh(&ifp->lock);
                ifp->valid_lft = valid_lft;
                ifp->prefered_lft = prefered_lft;
                ifp->tstamp = jiffies;
-               spin_unlock(&ifp->lock);
+               spin_unlock_bh(&ifp->lock);
 
                addrconf_dad_start(ifp, 0);
                in6_ifa_put(ifp);
@@ -2344,10 +2344,10 @@ static int addrconf_ifdown(struct net_device *dev, int how)
                   Do not dev_put!
         */
        if (how == 1) {
-               write_lock_bh(&addrconf_lock);
-               dev->ip6_ptr = NULL;
                idev->dead = 1;
-               write_unlock_bh(&addrconf_lock);
+
+               /* protected by rtnl_lock */
+               rcu_assign_pointer(dev->ip6_ptr, NULL);
 
                /* Step 1.5: remove snmp6 entry */
                snmp6_unregister_dev(idev);
@@ -2869,52 +2869,57 @@ restart:
        spin_unlock_bh(&addrconf_verify_lock);
 }
 
+static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
+{
+       struct in6_addr *pfx = NULL;
+
+       if (addr)
+               pfx = nla_data(addr);
+
+       if (local) {
+               if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
+                       pfx = NULL;
+               else
+                       pfx = nla_data(local);
+       }
+
+       return pfx;
+}
+
+static struct nla_policy ifa_ipv6_policy[IFA_MAX+1] __read_mostly = {
+       [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
+       [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
+       [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
+};
+
 static int
 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 {
-       struct rtattr **rta = arg;
-       struct ifaddrmsg *ifm = NLMSG_DATA(nlh);
+       struct ifaddrmsg *ifm;
+       struct nlattr *tb[IFA_MAX+1];
        struct in6_addr *pfx;
+       int err;
 
-       pfx = NULL;
-       if (rta[IFA_ADDRESS-1]) {
-               if (RTA_PAYLOAD(rta[IFA_ADDRESS-1]) < sizeof(*pfx))
-                       return -EINVAL;
-               pfx = RTA_DATA(rta[IFA_ADDRESS-1]);
-       }
-       if (rta[IFA_LOCAL-1]) {
-               if (RTA_PAYLOAD(rta[IFA_LOCAL-1]) < sizeof(*pfx) ||
-                   (pfx && memcmp(pfx, RTA_DATA(rta[IFA_LOCAL-1]), sizeof(*pfx))))
-                       return -EINVAL;
-               pfx = RTA_DATA(rta[IFA_LOCAL-1]);
-       }
+       err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
+       if (err < 0)
+               return err;
+
+       ifm = nlmsg_data(nlh);
+       pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
        if (pfx == NULL)
                return -EINVAL;
 
        return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
 }
 
-static int
-inet6_addr_modify(int ifindex, struct in6_addr *pfx,
-                 __u32 prefered_lft, __u32 valid_lft)
+static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 prefered_lft,
+                            u32 valid_lft)
 {
-       struct inet6_ifaddr *ifp = NULL;
-       struct net_device *dev;
        int ifa_flags = 0;
 
-       if ((dev = __dev_get_by_index(ifindex)) == NULL)
-               return -ENODEV;
-
-       if (!(dev->flags&IFF_UP))
-               return -ENETDOWN;
-
        if (!valid_lft || (prefered_lft > valid_lft))
                return -EINVAL;
 
-       ifp = ipv6_get_ifaddr(pfx, dev, 1);
-       if (ifp == NULL)
-               return -ENOENT;
-
        if (valid_lft == INFINITY_LIFE_TIME)
                ifa_flags = IFA_F_PERMANENT;
        else if (valid_lft >= 0x7FFFFFFF/HZ)
@@ -2936,7 +2941,6 @@ inet6_addr_modify(int ifindex, struct in6_addr *pfx,
        spin_unlock_bh(&ifp->lock);
        if (!(ifp->flags&IFA_F_TENTATIVE))
                ipv6_ifa_notify(0, ifp);
-       in6_ifa_put(ifp);
 
        addrconf_verify(0);
 
@@ -2946,172 +2950,185 @@ inet6_addr_modify(int ifindex, struct in6_addr *pfx,
 static int
 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 {
-       struct rtattr  **rta = arg;
-       struct ifaddrmsg *ifm = NLMSG_DATA(nlh);
+       struct ifaddrmsg *ifm;
+       struct nlattr *tb[IFA_MAX+1];
        struct in6_addr *pfx;
-       __u32 valid_lft = INFINITY_LIFE_TIME, prefered_lft = INFINITY_LIFE_TIME;
+       struct inet6_ifaddr *ifa;
+       struct net_device *dev;
+       u32 valid_lft, preferred_lft;
+       int err;
 
-       pfx = NULL;
-       if (rta[IFA_ADDRESS-1]) {
-               if (RTA_PAYLOAD(rta[IFA_ADDRESS-1]) < sizeof(*pfx))
-                       return -EINVAL;
-               pfx = RTA_DATA(rta[IFA_ADDRESS-1]);
-       }
-       if (rta[IFA_LOCAL-1]) {
-               if (RTA_PAYLOAD(rta[IFA_LOCAL-1]) < sizeof(*pfx) ||
-                   (pfx && memcmp(pfx, RTA_DATA(rta[IFA_LOCAL-1]), sizeof(*pfx))))
-                       return -EINVAL;
-               pfx = RTA_DATA(rta[IFA_LOCAL-1]);
-       }
+       err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
+       if (err < 0)
+               return err;
+
+       ifm = nlmsg_data(nlh);
+       pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
        if (pfx == NULL)
                return -EINVAL;
 
-       if (rta[IFA_CACHEINFO-1]) {
+       if (tb[IFA_CACHEINFO]) {
                struct ifa_cacheinfo *ci;
-               if (RTA_PAYLOAD(rta[IFA_CACHEINFO-1]) < sizeof(*ci))
-                       return -EINVAL;
-               ci = RTA_DATA(rta[IFA_CACHEINFO-1]);
+
+               ci = nla_data(tb[IFA_CACHEINFO]);
                valid_lft = ci->ifa_valid;
-               prefered_lft = ci->ifa_prefered;
+               preferred_lft = ci->ifa_prefered;
+       } else {
+               preferred_lft = INFINITY_LIFE_TIME;
+               valid_lft = INFINITY_LIFE_TIME;
        }
 
-       if (nlh->nlmsg_flags & NLM_F_REPLACE) {
-               int ret;
-               ret = inet6_addr_modify(ifm->ifa_index, pfx,
-                                       prefered_lft, valid_lft);
-               if (ret == 0 || !(nlh->nlmsg_flags & NLM_F_CREATE))
-                       return ret;
+       dev =  __dev_get_by_index(ifm->ifa_index);
+       if (dev == NULL)
+               return -ENODEV;
+
+       ifa = ipv6_get_ifaddr(pfx, dev, 1);
+       if (ifa == NULL) {
+               /*
+                * It would be best to check for !NLM_F_CREATE here but
+                * userspace alreay relies on not having to provide this.
+                */
+               return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
+                                     preferred_lft, valid_lft);
        }
 
-       return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
-                             prefered_lft, valid_lft);
+       if (nlh->nlmsg_flags & NLM_F_EXCL ||
+           !(nlh->nlmsg_flags & NLM_F_REPLACE))
+               err = -EEXIST;
+       else
+               err = inet6_addr_modify(ifa, preferred_lft, valid_lft);
+
+       in6_ifa_put(ifa);
+
+       return err;
+}
+
+static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
+                         u8 scope, int ifindex)
+{
+       struct ifaddrmsg *ifm;
+
+       ifm = nlmsg_data(nlh);
+       ifm->ifa_family = AF_INET6;
+       ifm->ifa_prefixlen = prefixlen;
+       ifm->ifa_flags = flags;
+       ifm->ifa_scope = scope;
+       ifm->ifa_index = ifindex;
+}
+
+static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
+                        unsigned long tstamp, u32 preferred, u32 valid)
+{
+       struct ifa_cacheinfo ci;
+
+       ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
+                       + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
+       ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
+                       + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
+       ci.ifa_prefered = preferred;
+       ci.ifa_valid = valid;
 
+       return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
 }
 
-/* Maximum length of ifa_cacheinfo attributes */
-#define INET6_IFADDR_RTA_SPACE \
-               RTA_SPACE(16) /* IFA_ADDRESS */ + \
-               RTA_SPACE(sizeof(struct ifa_cacheinfo)) /* CACHEINFO */
+static inline int rt_scope(int ifa_scope)
+{
+       if (ifa_scope & IFA_HOST)
+               return RT_SCOPE_HOST;
+       else if (ifa_scope & IFA_LINK)
+               return RT_SCOPE_LINK;
+       else if (ifa_scope & IFA_SITE)
+               return RT_SCOPE_SITE;
+       else
+               return RT_SCOPE_UNIVERSE;
+}
+
+static inline int inet6_ifaddr_msgsize(void)
+{
+       return nlmsg_total_size(sizeof(struct ifaddrmsg) +
+                               nla_total_size(16) +
+                               nla_total_size(sizeof(struct ifa_cacheinfo)) +
+                               128);
+}
 
 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
                             u32 pid, u32 seq, int event, unsigned int flags)
 {
-       struct ifaddrmsg *ifm;
        struct nlmsghdr  *nlh;
-       struct ifa_cacheinfo ci;
-       unsigned char    *b = skb->tail;
+       u32 preferred, valid;
+
+       nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
+       if (nlh == NULL)
+               return -ENOBUFS;
+
+       put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
+                     ifa->idev->dev->ifindex);
 
-       nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*ifm), flags);
-       ifm = NLMSG_DATA(nlh);
-       ifm->ifa_family = AF_INET6;
-       ifm->ifa_prefixlen = ifa->prefix_len;
-       ifm->ifa_flags = ifa->flags;
-       ifm->ifa_scope = RT_SCOPE_UNIVERSE;
-       if (ifa->scope&IFA_HOST)
-               ifm->ifa_scope = RT_SCOPE_HOST;
-       else if (ifa->scope&IFA_LINK)
-               ifm->ifa_scope = RT_SCOPE_LINK;
-       else if (ifa->scope&IFA_SITE)
-               ifm->ifa_scope = RT_SCOPE_SITE;
-       ifm->ifa_index = ifa->idev->dev->ifindex;
-       RTA_PUT(skb, IFA_ADDRESS, 16, &ifa->addr);
        if (!(ifa->flags&IFA_F_PERMANENT)) {
-               ci.ifa_prefered = ifa->prefered_lft;
-               ci.ifa_valid = ifa->valid_lft;
-               if (ci.ifa_prefered != INFINITY_LIFE_TIME) {
+               preferred = ifa->prefered_lft;
+               valid = ifa->valid_lft;
+               if (preferred != INFINITY_LIFE_TIME) {
                        long tval = (jiffies - ifa->tstamp)/HZ;
-                       ci.ifa_prefered -= tval;
-                       if (ci.ifa_valid != INFINITY_LIFE_TIME)
-                               ci.ifa_valid -= tval;
+                       preferred -= tval;
+                       if (valid != INFINITY_LIFE_TIME)
+                               valid -= tval;
                }
        } else {
-               ci.ifa_prefered = INFINITY_LIFE_TIME;
-               ci.ifa_valid = INFINITY_LIFE_TIME;
-       }
-       ci.cstamp = (__u32)(TIME_DELTA(ifa->cstamp, INITIAL_JIFFIES) / HZ * 100
-                   + TIME_DELTA(ifa->cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
-       ci.tstamp = (__u32)(TIME_DELTA(ifa->tstamp, INITIAL_JIFFIES) / HZ * 100
-                   + TIME_DELTA(ifa->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
-       RTA_PUT(skb, IFA_CACHEINFO, sizeof(ci), &ci);
-       nlh->nlmsg_len = skb->tail - b;
-       return skb->len;
+               preferred = INFINITY_LIFE_TIME;
+               valid = INFINITY_LIFE_TIME;
+       }
 
-nlmsg_failure:
-rtattr_failure:
-       skb_trim(skb, b - skb->data);
-       return -1;
+       if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
+           put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
+               return nlmsg_cancel(skb, nlh);
+
+       return nlmsg_end(skb, nlh);
 }
 
 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
                                u32 pid, u32 seq, int event, u16 flags)
 {
-       struct ifaddrmsg *ifm;
        struct nlmsghdr  *nlh;
-       struct ifa_cacheinfo ci;
-       unsigned char    *b = skb->tail;
-
-       nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*ifm), flags);
-       ifm = NLMSG_DATA(nlh);
-       ifm->ifa_family = AF_INET6;     
-       ifm->ifa_prefixlen = 128;
-       ifm->ifa_flags = IFA_F_PERMANENT;
-       ifm->ifa_scope = RT_SCOPE_UNIVERSE;
-       if (ipv6_addr_scope(&ifmca->mca_addr)&IFA_SITE)
-               ifm->ifa_scope = RT_SCOPE_SITE;
-       ifm->ifa_index = ifmca->idev->dev->ifindex;
-       RTA_PUT(skb, IFA_MULTICAST, 16, &ifmca->mca_addr);
-       ci.cstamp = (__u32)(TIME_DELTA(ifmca->mca_cstamp, INITIAL_JIFFIES) / HZ
-                   * 100 + TIME_DELTA(ifmca->mca_cstamp, INITIAL_JIFFIES) % HZ
-                   * 100 / HZ);
-       ci.tstamp = (__u32)(TIME_DELTA(ifmca->mca_tstamp, INITIAL_JIFFIES) / HZ
-                   * 100 + TIME_DELTA(ifmca->mca_tstamp, INITIAL_JIFFIES) % HZ
-                   * 100 / HZ);
-       ci.ifa_prefered = INFINITY_LIFE_TIME;
-       ci.ifa_valid = INFINITY_LIFE_TIME;
-       RTA_PUT(skb, IFA_CACHEINFO, sizeof(ci), &ci);
-       nlh->nlmsg_len = skb->tail - b;
-       return skb->len;
+       u8 scope = RT_SCOPE_UNIVERSE;
+       int ifindex = ifmca->idev->dev->ifindex;
 
-nlmsg_failure:
-rtattr_failure:
-       skb_trim(skb, b - skb->data);
-       return -1;
+       if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
+               scope = RT_SCOPE_SITE;
+
+       nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
+       if (nlh == NULL)
+               return -ENOBUFS;
+
+       put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
+       if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
+           put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
+                         INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0)
+               return nlmsg_cancel(skb, nlh);
+
+       return nlmsg_end(skb, nlh);
 }
 
 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
                                u32 pid, u32 seq, int event, unsigned int flags)
 {
-       struct ifaddrmsg *ifm;
        struct nlmsghdr  *nlh;
-       struct ifa_cacheinfo ci;
-       unsigned char    *b = skb->tail;
-
-       nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*ifm), flags);
-       ifm = NLMSG_DATA(nlh);
-       ifm->ifa_family = AF_INET6;     
-       ifm->ifa_prefixlen = 128;
-       ifm->ifa_flags = IFA_F_PERMANENT;
-       ifm->ifa_scope = RT_SCOPE_UNIVERSE;
-       if (ipv6_addr_scope(&ifaca->aca_addr)&IFA_SITE)
-               ifm->ifa_scope = RT_SCOPE_SITE;
-       ifm->ifa_index = ifaca->aca_idev->dev->ifindex;
-       RTA_PUT(skb, IFA_ANYCAST, 16, &ifaca->aca_addr);
-       ci.cstamp = (__u32)(TIME_DELTA(ifaca->aca_cstamp, INITIAL_JIFFIES) / HZ
-                   * 100 + TIME_DELTA(ifaca->aca_cstamp, INITIAL_JIFFIES) % HZ
-                   * 100 / HZ);
-       ci.tstamp = (__u32)(TIME_DELTA(ifaca->aca_tstamp, INITIAL_JIFFIES) / HZ
-                   * 100 + TIME_DELTA(ifaca->aca_tstamp, INITIAL_JIFFIES) % HZ
-                   * 100 / HZ);
-       ci.ifa_prefered = INFINITY_LIFE_TIME;
-       ci.ifa_valid = INFINITY_LIFE_TIME;
-       RTA_PUT(skb, IFA_CACHEINFO, sizeof(ci), &ci);
-       nlh->nlmsg_len = skb->tail - b;
-       return skb->len;
+       u8 scope = RT_SCOPE_UNIVERSE;
+       int ifindex = ifaca->aca_idev->dev->ifindex;
 
-nlmsg_failure:
-rtattr_failure:
-       skb_trim(skb, b - skb->data);
-       return -1;
+       if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
+               scope = RT_SCOPE_SITE;
+
+       nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
+       if (nlh == NULL)
+               return -ENOBUFS;
+
+       put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
+       if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
+           put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
+                         INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0)
+               return nlmsg_cancel(skb, nlh);
+
+       return nlmsg_end(skb, nlh);
 }
 
 enum addr_type_t
@@ -3222,79 +3239,74 @@ static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
        return inet6_dump_addr(skb, cb, type);
 }
 
-static int inet6_rtm_getaddr(struct sk_buff *in_skb,
-               struct nlmsghdr* nlh, void *arg)
+static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
+                            void *arg)
 {
-       struct rtattr **rta = arg;
-       struct ifaddrmsg *ifm = NLMSG_DATA(nlh);
+       struct ifaddrmsg *ifm;
+       struct nlattr *tb[IFA_MAX+1];
        struct in6_addr *addr = NULL;
        struct net_device *dev = NULL;
        struct inet6_ifaddr *ifa;
        struct sk_buff *skb;
-       int size = NLMSG_SPACE(sizeof(struct ifaddrmsg) + INET6_IFADDR_RTA_SPACE);
        int err;
 
-       if (rta[IFA_ADDRESS-1]) {
-               if (RTA_PAYLOAD(rta[IFA_ADDRESS-1]) < sizeof(*addr))
-                       return -EINVAL;
-               addr = RTA_DATA(rta[IFA_ADDRESS-1]);
-       }
-       if (rta[IFA_LOCAL-1]) {
-               if (RTA_PAYLOAD(rta[IFA_LOCAL-1]) < sizeof(*addr) ||
-                   (addr && memcmp(addr, RTA_DATA(rta[IFA_LOCAL-1]), sizeof(*addr))))
-                       return -EINVAL;
-               addr = RTA_DATA(rta[IFA_LOCAL-1]);
+       err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
+       if (err < 0)
+               goto errout;
+
+       addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
+       if (addr == NULL) {
+               err = -EINVAL;
+               goto errout;
        }
-       if (addr == NULL)
-               return -EINVAL;
 
+       ifm = nlmsg_data(nlh);
        if (ifm->ifa_index)
                dev = __dev_get_by_index(ifm->ifa_index);
 
-       if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL)
-               return -EADDRNOTAVAIL;
+       if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
+               err = -EADDRNOTAVAIL;
+               goto errout;
+       }
 
-       if ((skb = alloc_skb(size, GFP_KERNEL)) == NULL) {
+       if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
                err = -ENOBUFS;
-               goto out;
+               goto errout_ifa;
        }
 
-       NETLINK_CB(skb).dst_pid = NETLINK_CB(in_skb).pid;
        err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
                                nlh->nlmsg_seq, RTM_NEWADDR, 0);
        if (err < 0) {
-               err = -EMSGSIZE;
-               goto out_free;
+               kfree_skb(skb);
+               goto errout_ifa;
        }
 
-       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_ifa:
        in6_ifa_put(ifa);
+errout:
        return err;
-out_free:
-       kfree_skb(skb);
-       goto out;
 }
 
 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
 {
        struct sk_buff *skb;
-       int size = NLMSG_SPACE(sizeof(struct ifaddrmsg) + INET6_IFADDR_RTA_SPACE);
+       int err = -ENOBUFS;
 
-       skb = alloc_skb(size, GFP_ATOMIC);
-       if (!skb) {
-               netlink_set_err(rtnl, 0, RTNLGRP_IPV6_IFADDR, ENOBUFS);
-               return;
-       }
-       if (inet6_fill_ifaddr(skb, ifa, current->pid, 0, event, 0) < 0) {
+       skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
+       if (skb == NULL)
+               goto errout;
+
+       err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
+       if (err < 0) {
                kfree_skb(skb);
-               netlink_set_err(rtnl, 0, RTNLGRP_IPV6_IFADDR, EINVAL);
-               return;
+               goto errout;
        }
-       NETLINK_CB(skb).dst_group = RTNLGRP_IPV6_IFADDR;
-       netlink_broadcast(rtnl, skb, 0, RTNLGRP_IPV6_IFADDR, GFP_ATOMIC);
+
+       err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
+errout:
+       if (err < 0)
+               rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
 }
 
 static void inline ipv6_store_devconf(struct ipv6_devconf *cnf,
@@ -3329,6 +3341,7 @@ static void inline ipv6_store_devconf(struct ipv6_devconf *cnf,
        array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
 #endif
 #endif
+       array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
 }
 
 /* Maximum length of ifinfomsg attributes */
@@ -3435,20 +3448,23 @@ static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
 {
        struct sk_buff *skb;
-       int size = NLMSG_SPACE(sizeof(struct ifinfomsg) + INET6_IFINFO_RTA_SPACE);
+       int payload = sizeof(struct ifinfomsg) + INET6_IFINFO_RTA_SPACE;
+       int err = -ENOBUFS;
        
-       skb = alloc_skb(size, GFP_ATOMIC);
-       if (!skb) {
-               netlink_set_err(rtnl, 0, RTNLGRP_IPV6_IFINFO, ENOBUFS);
-               return;
-       }
-       if (inet6_fill_ifinfo(skb, idev, current->pid, 0, event, 0) < 0) {
+       skb = nlmsg_new(nlmsg_total_size(payload), GFP_ATOMIC);
+       if (skb == NULL)
+               goto errout;
+
+       err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
+       if (err < 0) {
                kfree_skb(skb);
-               netlink_set_err(rtnl, 0, RTNLGRP_IPV6_IFINFO, EINVAL);
-               return;
+               goto errout;
        }
-       NETLINK_CB(skb).dst_group = RTNLGRP_IPV6_IFINFO;
-       netlink_broadcast(rtnl, skb, 0, RTNLGRP_IPV6_IFINFO, GFP_ATOMIC);
+
+       err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
+errout:
+       if (err < 0)
+               rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
 }
 
 /* Maximum length of prefix_cacheinfo attributes */
@@ -3500,20 +3516,23 @@ static void inet6_prefix_notify(int event, struct inet6_dev *idev,
                         struct prefix_info *pinfo)
 {
        struct sk_buff *skb;
-       int size = NLMSG_SPACE(sizeof(struct prefixmsg) + INET6_PREFIX_RTA_SPACE);
+       int payload = sizeof(struct prefixmsg) + INET6_PREFIX_RTA_SPACE;
+       int err = -ENOBUFS;
 
-       skb = alloc_skb(size, GFP_ATOMIC);
-       if (!skb) {
-               netlink_set_err(rtnl, 0, RTNLGRP_IPV6_PREFIX, ENOBUFS);
-               return;
-       }
-       if (inet6_fill_prefix(skb, idev, pinfo, current->pid, 0, event, 0) < 0) {
+       skb = nlmsg_new(nlmsg_total_size(payload), GFP_ATOMIC);
+       if (skb == NULL)
+               goto errout;
+
+       err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
+       if (err < 0) {
                kfree_skb(skb);
-               netlink_set_err(rtnl, 0, RTNLGRP_IPV6_PREFIX, EINVAL);
-               return;
+               goto errout;
        }
-       NETLINK_CB(skb).dst_group = RTNLGRP_IPV6_PREFIX;
-       netlink_broadcast(rtnl, skb, 0, RTNLGRP_IPV6_PREFIX, GFP_ATOMIC);
+
+       err = rtnl_notify(skb, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
+errout:
+       if (err < 0)
+               rtnl_set_sk_err(RTNLGRP_IPV6_PREFIX, err);
 }
 
 static struct rtnetlink_link inet6_rtnetlink_table[RTM_NR_MSGTYPES] = {
@@ -3528,6 +3547,9 @@ static struct rtnetlink_link inet6_rtnetlink_table[RTM_NR_MSGTYPES] = {
        [RTM_DELROUTE - RTM_BASE] = { .doit     = inet6_rtm_delroute, },
        [RTM_GETROUTE - RTM_BASE] = { .doit     = inet6_rtm_getroute,
                                      .dumpit   = inet6_dump_fib, },
+#ifdef CONFIG_IPV6_MULTIPLE_TABLES
+       [RTM_GETRULE  - RTM_BASE] = { .dumpit   = fib6_rules_dump,   },
+#endif
 };
 
 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
@@ -3536,7 +3558,7 @@ static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
 
        switch (event) {
        case RTM_NEWADDR:
-               ip6_ins_rt(ifp->rt, NULL, NULL, NULL);
+               ip6_ins_rt(ifp->rt);
                if (ifp->idev->cnf.forwarding)
                        addrconf_join_anycast(ifp);
                break;
@@ -3545,7 +3567,7 @@ static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
                        addrconf_leave_anycast(ifp);
                addrconf_leave_solict(ifp->idev, &ifp->addr);
                dst_hold(&ifp->rt->u.dst);
-               if (ip6_del_rt(ifp->rt, NULL, NULL, NULL))
+               if (ip6_del_rt(ifp->rt))
                        dst_free(&ifp->rt->u.dst);
                break;
        }
@@ -3553,10 +3575,10 @@ static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
 
 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
 {
-       read_lock_bh(&addrconf_lock);
+       rcu_read_lock_bh();
        if (likely(ifp->idev->dead == 0))
                __ipv6_ifa_notify(event, ifp);
-       read_unlock_bh(&addrconf_lock);
+       rcu_read_unlock_bh();
 }
 
 #ifdef CONFIG_SYSCTL
@@ -3653,7 +3675,7 @@ static struct addrconf_sysctl_table
        ctl_table addrconf_conf_dir[2];
        ctl_table addrconf_proto_dir[2];
        ctl_table addrconf_root_dir[2];
-} addrconf_sysctl = {
+} addrconf_sysctl __read_mostly = {
        .sysctl_header = NULL,
        .addrconf_vars = {
                {
@@ -3842,6 +3864,14 @@ static struct addrconf_sysctl_table
                },
 #endif
 #endif
+               {
+                       .ctl_name       =       NET_IPV6_PROXY_NDP,
+                       .procname       =       "proxy_ndp",
+                       .data           =       &ipv6_devconf.proxy_ndp,
+                       .maxlen         =       sizeof(int),
+                       .mode           =       0644,
+                       .proc_handler   =       &proc_dointvec,
+               },
                {
                        .ctl_name       =       0,      /* sentinel */
                }