Merge branch 'topic/core-cleanup' into for-linus
[safe/jmp/linux-2.6] / net / ipv6 / af_inet6.c
index 90d2f72..3f9e86b 100644 (file)
@@ -7,8 +7,6 @@
  *
  *     Adapted from linux/net/ipv4/af_inet.c
  *
- *     $Id: af_inet6.c,v 1.66 2002/02/01 22:01:04 davem Exp $
- *
  *     Fixes:
  *     piggy, Karl Knutson     :       Socket protocol table
  *     Hideaki YOSHIFUJI       :       sin6_scope_id support
@@ -38,6 +36,7 @@
 #include <linux/proc_fs.h>
 #include <linux/stat.h>
 #include <linux/init.h>
+#include <linux/slab.h>
 
 #include <linux/inet.h>
 #include <linux/netdevice.h>
@@ -52,6 +51,7 @@
 #include <net/ipip.h>
 #include <net/protocol.h>
 #include <net/inet_common.h>
+#include <net/route.h>
 #include <net/transp_v6.h>
 #include <net/ip6_route.h>
 #include <net/addrconf.h>
 
 #include <asm/uaccess.h>
 #include <asm/system.h>
+#include <linux/mroute6.h>
 
 MODULE_AUTHOR("Cast of dozens");
 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
 MODULE_LICENSE("GPL");
 
-int sysctl_ipv6_bindv6only __read_mostly;
-
 /* The inetsw6 table contains everything that inet6_create needs to
  * build a new socket.
  */
 static struct list_head inetsw6[SOCK_MAX];
 static DEFINE_SPINLOCK(inetsw6_lock);
 
+struct ipv6_params ipv6_defaults = {
+       .disable_ipv6 = 0,
+       .autoconf = 1,
+};
+
+static int disable_ipv6_mod = 0;
+
+module_param_named(disable, disable_ipv6_mod, int, 0444);
+MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
+
+module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
+MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
+
+module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
+MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
+
 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
 {
        const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
@@ -81,12 +96,12 @@ static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
        return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
 }
 
-static int inet6_create(struct net *net, struct socket *sock, int protocol)
+static int inet6_create(struct net *net, struct socket *sock, int protocol,
+                       int kern)
 {
        struct inet_sock *inet;
        struct ipv6_pinfo *np;
        struct sock *sk;
-       struct list_head *p;
        struct inet_protosw *answer;
        struct proto *answer_prot;
        unsigned char answer_flags;
@@ -94,22 +109,18 @@ static int inet6_create(struct net *net, struct socket *sock, int protocol)
        int try_loading_module = 0;
        int err;
 
-       if (net != &init_net)
-               return -EAFNOSUPPORT;
-
        if (sock->type != SOCK_RAW &&
            sock->type != SOCK_DGRAM &&
            !inet_ehash_secret)
                build_ehash_secret();
 
        /* Look for the requested type/protocol pair. */
-       answer = NULL;
 lookup_protocol:
        err = -ESOCKTNOSUPPORT;
        rcu_read_lock();
-       list_for_each_rcu(p, &inetsw6[sock->type]) {
-               answer = list_entry(p, struct inet_protosw, list);
+       list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
 
+               err = 0;
                /* Check the non-wild match. */
                if (protocol == answer->protocol) {
                        if (protocol != IPPROTO_IP)
@@ -124,10 +135,9 @@ lookup_protocol:
                                break;
                }
                err = -EPROTONOSUPPORT;
-               answer = NULL;
        }
 
-       if (!answer) {
+       if (err) {
                if (try_loading_module < 2) {
                        rcu_read_unlock();
                        /*
@@ -150,7 +160,7 @@ lookup_protocol:
        }
 
        err = -EPERM;
-       if (answer->capability > 0 && !capable(answer->capability))
+       if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
                goto out_rcu_unlock;
 
        sock->ops = answer->ops;
@@ -159,7 +169,7 @@ lookup_protocol:
        answer_flags = answer->flags;
        rcu_read_unlock();
 
-       BUG_TRAP(answer_prot->slab != NULL);
+       WARN_ON(answer_prot->slab == NULL);
 
        err = -ENOBUFS;
        sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
@@ -177,7 +187,7 @@ lookup_protocol:
        inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
 
        if (SOCK_RAW == sock->type) {
-               inet->num = protocol;
+               inet->inet_num = protocol;
                if (IPPROTO_RAW == protocol)
                        inet->hdrincl = 1;
        }
@@ -190,10 +200,10 @@ lookup_protocol:
 
        inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
        np->hop_limit   = -1;
-       np->mcast_hops  = -1;
+       np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
        np->mc_loop     = 1;
        np->pmtudisc    = IPV6_PMTUDISC_WANT;
-       np->ipv6only    = sysctl_ipv6_bindv6only;
+       np->ipv6only    = net->ipv6.sysctl.bindv6only;
 
        /* Init the ipv4 part of the socket since we can have sockets
         * using v6 API for ipv4.
@@ -220,12 +230,12 @@ lookup_protocol:
         */
        sk_refcnt_debug_inc(sk);
 
-       if (inet->num) {
+       if (inet->inet_num) {
                /* It assumes that any protocol which allows
                 * the user to assign a number at socket
                 * creation time automatically shares.
                 */
-               inet->sport = htons(inet->num);
+               inet->inet_sport = htons(inet->inet_num);
                sk->sk_prot->hash(sk);
        }
        if (sk->sk_prot->init) {
@@ -250,6 +260,7 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
        struct sock *sk = sock->sk;
        struct inet_sock *inet = inet_sk(sk);
        struct ipv6_pinfo *np = inet6_sk(sk);
+       struct net *net = sock_net(sk);
        __be32 v4addr = 0;
        unsigned short snum;
        int addr_type = 0;
@@ -272,15 +283,32 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
        lock_sock(sk);
 
        /* Check these errors (active socket, double bind). */
-       if (sk->sk_state != TCP_CLOSE || inet->num) {
+       if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
                err = -EINVAL;
                goto out;
        }
 
        /* Check if the address belongs to the host. */
        if (addr_type == IPV6_ADDR_MAPPED) {
+               int chk_addr_ret;
+
+               /* Binding to v4-mapped address on a v6-only socket
+                * makes no sense
+                */
+               if (np->ipv6only) {
+                       err = -EINVAL;
+                       goto out;
+               }
+
+               /* Reproduce AF_INET checks to make the bindings consitant */
                v4addr = addr->sin6_addr.s6_addr32[3];
-               if (inet_addr_type(v4addr) != RTN_LOCAL) {
+               chk_addr_ret = inet_addr_type(net, v4addr);
+               if (!sysctl_ip_nonlocal_bind &&
+                   !(inet->freebind || inet->transparent) &&
+                   v4addr != htonl(INADDR_ANY) &&
+                   chk_addr_ret != RTN_LOCAL &&
+                   chk_addr_ret != RTN_MULTICAST &&
+                   chk_addr_ret != RTN_BROADCAST) {
                        err = -EADDRNOTAVAIL;
                        goto out;
                }
@@ -288,6 +316,7 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
                if (addr_type != IPV6_ADDR_ANY) {
                        struct net_device *dev = NULL;
 
+                       rcu_read_lock();
                        if (addr_type & IPV6_ADDR_LINKLOCAL) {
                                if (addr_len >= sizeof(struct sockaddr_in6) &&
                                    addr->sin6_scope_id) {
@@ -300,12 +329,12 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
                                /* Binding to link-local address requires an interface */
                                if (!sk->sk_bound_dev_if) {
                                        err = -EINVAL;
-                                       goto out;
+                                       goto out_unlock;
                                }
-                               dev = dev_get_by_index(&init_net, sk->sk_bound_dev_if);
+                               dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
                                if (!dev) {
                                        err = -ENODEV;
-                                       goto out;
+                                       goto out_unlock;
                                }
                        }
 
@@ -314,20 +343,18 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
                         */
                        v4addr = LOOPBACK4_IPV6;
                        if (!(addr_type & IPV6_ADDR_MULTICAST)) {
-                               if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
-                                       if (dev)
-                                               dev_put(dev);
+                               if (!ipv6_chk_addr(net, &addr->sin6_addr,
+                                                  dev, 0)) {
                                        err = -EADDRNOTAVAIL;
-                                       goto out;
+                                       goto out_unlock;
                                }
                        }
-                       if (dev)
-                               dev_put(dev);
+                       rcu_read_unlock();
                }
        }
 
-       inet->rcv_saddr = v4addr;
-       inet->saddr = v4addr;
+       inet->inet_rcv_saddr = v4addr;
+       inet->inet_saddr = v4addr;
 
        ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
 
@@ -341,16 +368,22 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
                goto out;
        }
 
-       if (addr_type != IPV6_ADDR_ANY)
+       if (addr_type != IPV6_ADDR_ANY) {
                sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
+               if (addr_type != IPV6_ADDR_MAPPED)
+                       np->ipv6only = 1;
+       }
        if (snum)
                sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
-       inet->sport = htons(inet->num);
-       inet->dport = 0;
-       inet->daddr = 0;
+       inet->inet_sport = htons(inet->inet_num);
+       inet->inet_dport = 0;
+       inet->inet_daddr = 0;
 out:
        release_sock(sk);
        return err;
+out_unlock:
+       rcu_read_unlock();
+       goto out;
 }
 
 EXPORT_SYMBOL(inet6_bind);
@@ -373,7 +406,7 @@ int inet6_release(struct socket *sock)
 
 EXPORT_SYMBOL(inet6_release);
 
-int inet6_destroy_sock(struct sock *sk)
+void inet6_destroy_sock(struct sock *sk)
 {
        struct ipv6_pinfo *np = inet6_sk(sk);
        struct sk_buff *skb;
@@ -391,8 +424,6 @@ int inet6_destroy_sock(struct sock *sk)
 
        if ((opt = xchg(&np->opt, NULL)) != NULL)
                sock_kfree_s(sk, opt, opt->tot_len);
-
-       return 0;
 }
 
 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
@@ -413,12 +444,12 @@ int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
        sin->sin6_flowinfo = 0;
        sin->sin6_scope_id = 0;
        if (peer) {
-               if (!inet->dport)
+               if (!inet->inet_dport)
                        return -ENOTCONN;
                if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
                    peer == 1)
                        return -ENOTCONN;
-               sin->sin6_port = inet->dport;
+               sin->sin6_port = inet->inet_dport;
                ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
                if (np->sndflow)
                        sin->sin6_flowinfo = np->flow_label;
@@ -428,7 +459,7 @@ int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
                else
                        ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
 
-               sin->sin6_port = inet->sport;
+               sin->sin6_port = inet->inet_sport;
        }
        if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
                sin->sin6_scope_id = sk->sk_bound_dev_if;
@@ -441,6 +472,7 @@ EXPORT_SYMBOL(inet6_getname);
 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 {
        struct sock *sk = sock->sk;
+       struct net *net = sock_net(sk);
 
        switch(cmd)
        {
@@ -453,14 +485,14 @@ int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
        case SIOCADDRT:
        case SIOCDELRT:
 
-               return(ipv6_route_ioctl(cmd,(void __user *)arg));
+               return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
 
        case SIOCSIFADDR:
-               return addrconf_add_ifaddr((void __user *) arg);
+               return addrconf_add_ifaddr(net, (void __user *) arg);
        case SIOCDIFADDR:
-               return addrconf_del_ifaddr((void __user *) arg);
+               return addrconf_del_ifaddr(net, (void __user *) arg);
        case SIOCSIFDSTADDR:
-               return addrconf_set_dstaddr((void __user *) arg);
+               return addrconf_set_dstaddr(net, (void __user *) arg);
        default:
                if (!sk->sk_prot->ioctl)
                        return -ENOIOCTLCMD;
@@ -523,63 +555,29 @@ const struct proto_ops inet6_dgram_ops = {
 #endif
 };
 
-static struct net_proto_family inet6_family_ops = {
+static const struct net_proto_family inet6_family_ops = {
        .family = PF_INET6,
        .create = inet6_create,
        .owner  = THIS_MODULE,
 };
 
-/* Same as inet6_dgram_ops, sans udp_poll.  */
-static const struct proto_ops inet6_sockraw_ops = {
-       .family            = PF_INET6,
-       .owner             = THIS_MODULE,
-       .release           = inet6_release,
-       .bind              = inet6_bind,
-       .connect           = inet_dgram_connect,        /* ok           */
-       .socketpair        = sock_no_socketpair,        /* a do nothing */
-       .accept            = sock_no_accept,            /* a do nothing */
-       .getname           = inet6_getname,
-       .poll              = datagram_poll,             /* ok           */
-       .ioctl             = inet6_ioctl,               /* must change  */
-       .listen            = sock_no_listen,            /* ok           */
-       .shutdown          = inet_shutdown,             /* ok           */
-       .setsockopt        = sock_common_setsockopt,    /* ok           */
-       .getsockopt        = sock_common_getsockopt,    /* ok           */
-       .sendmsg           = inet_sendmsg,              /* ok           */
-       .recvmsg           = sock_common_recvmsg,       /* ok           */
-       .mmap              = sock_no_mmap,
-       .sendpage          = sock_no_sendpage,
-#ifdef CONFIG_COMPAT
-       .compat_setsockopt = compat_sock_common_setsockopt,
-       .compat_getsockopt = compat_sock_common_getsockopt,
-#endif
-};
-
-static struct inet_protosw rawv6_protosw = {
-       .type           = SOCK_RAW,
-       .protocol       = IPPROTO_IP,   /* wild card */
-       .prot           = &rawv6_prot,
-       .ops            = &inet6_sockraw_ops,
-       .capability     = CAP_NET_RAW,
-       .no_check       = UDP_CSUM_DEFAULT,
-       .flags          = INET_PROTOSW_REUSE,
-};
-
-void
-inet6_register_protosw(struct inet_protosw *p)
+int inet6_register_protosw(struct inet_protosw *p)
 {
        struct list_head *lh;
        struct inet_protosw *answer;
-       int protocol = p->protocol;
        struct list_head *last_perm;
+       int protocol = p->protocol;
+       int ret;
 
        spin_lock_bh(&inetsw6_lock);
 
+       ret = -EINVAL;
        if (p->type >= SOCK_MAX)
                goto out_illegal;
 
        /* If we are trying to override a permanent protocol, bail. */
        answer = NULL;
+       ret = -EPERM;
        last_perm = &inetsw6[p->type];
        list_for_each(lh, &inetsw6[p->type]) {
                answer = list_entry(lh, struct inet_protosw, list);
@@ -603,9 +601,10 @@ inet6_register_protosw(struct inet_protosw *p)
         * system automatically returns to the old behavior.
         */
        list_add_rcu(&p->list, last_perm);
+       ret = 0;
 out:
        spin_unlock_bh(&inetsw6_lock);
-       return;
+       return ret;
 
 out_permanent:
        printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
@@ -658,8 +657,9 @@ int inet6_sk_rebuild_header(struct sock *sk)
                ipv6_addr_copy(&fl.fl6_src, &np->saddr);
                fl.fl6_flowlabel = np->flow_label;
                fl.oif = sk->sk_bound_dev_if;
-               fl.fl_ip_dport = inet->dport;
-               fl.fl_ip_sport = inet->sport;
+               fl.mark = sk->sk_mark;
+               fl.fl_ip_dport = inet->inet_dport;
+               fl.fl_ip_sport = inet->inet_sport;
                security_sk_classify_flow(sk, &fl);
 
                if (np->opt && np->opt->srcrt) {
@@ -677,7 +677,7 @@ int inet6_sk_rebuild_header(struct sock *sk)
                if (final_p)
                        ipv6_addr_copy(&fl.fl6_dst, final_p);
 
-               if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
+               if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
                        sk->sk_err_soft = -err;
                        return err;
                }
@@ -712,62 +712,370 @@ int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
 
 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
 
-static int __init init_ipv6_mibs(void)
+static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
+{
+       const struct inet6_protocol *ops = NULL;
+
+       for (;;) {
+               struct ipv6_opt_hdr *opth;
+               int len;
+
+               if (proto != NEXTHDR_HOP) {
+                       ops = rcu_dereference(inet6_protos[proto]);
+
+                       if (unlikely(!ops))
+                               break;
+
+                       if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
+                               break;
+               }
+
+               if (unlikely(!pskb_may_pull(skb, 8)))
+                       break;
+
+               opth = (void *)skb->data;
+               len = ipv6_optlen(opth);
+
+               if (unlikely(!pskb_may_pull(skb, len)))
+                       break;
+
+               proto = opth->nexthdr;
+               __skb_pull(skb, len);
+       }
+
+       return proto;
+}
+
+static int ipv6_gso_send_check(struct sk_buff *skb)
+{
+       struct ipv6hdr *ipv6h;
+       const struct inet6_protocol *ops;
+       int err = -EINVAL;
+
+       if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
+               goto out;
+
+       ipv6h = ipv6_hdr(skb);
+       __skb_pull(skb, sizeof(*ipv6h));
+       err = -EPROTONOSUPPORT;
+
+       rcu_read_lock();
+       ops = rcu_dereference(inet6_protos[
+               ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
+
+       if (likely(ops && ops->gso_send_check)) {
+               skb_reset_transport_header(skb);
+               err = ops->gso_send_check(skb);
+       }
+       rcu_read_unlock();
+
+out:
+       return err;
+}
+
+static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
+{
+       struct sk_buff *segs = ERR_PTR(-EINVAL);
+       struct ipv6hdr *ipv6h;
+       const struct inet6_protocol *ops;
+       int proto;
+       struct frag_hdr *fptr;
+       unsigned int unfrag_ip6hlen;
+       u8 *prevhdr;
+       int offset = 0;
+
+       if (!(features & NETIF_F_V6_CSUM))
+               features &= ~NETIF_F_SG;
+
+       if (unlikely(skb_shinfo(skb)->gso_type &
+                    ~(SKB_GSO_UDP |
+                      SKB_GSO_DODGY |
+                      SKB_GSO_TCP_ECN |
+                      SKB_GSO_TCPV6 |
+                      0)))
+               goto out;
+
+       if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
+               goto out;
+
+       ipv6h = ipv6_hdr(skb);
+       __skb_pull(skb, sizeof(*ipv6h));
+       segs = ERR_PTR(-EPROTONOSUPPORT);
+
+       proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
+       rcu_read_lock();
+       ops = rcu_dereference(inet6_protos[proto]);
+       if (likely(ops && ops->gso_segment)) {
+               skb_reset_transport_header(skb);
+               segs = ops->gso_segment(skb, features);
+       }
+       rcu_read_unlock();
+
+       if (unlikely(IS_ERR(segs)))
+               goto out;
+
+       for (skb = segs; skb; skb = skb->next) {
+               ipv6h = ipv6_hdr(skb);
+               ipv6h->payload_len = htons(skb->len - skb->mac_len -
+                                          sizeof(*ipv6h));
+               if (proto == IPPROTO_UDP) {
+                       unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
+                       fptr = (struct frag_hdr *)(skb_network_header(skb) +
+                               unfrag_ip6hlen);
+                       fptr->frag_off = htons(offset);
+                       if (skb->next != NULL)
+                               fptr->frag_off |= htons(IP6_MF);
+                       offset += (ntohs(ipv6h->payload_len) -
+                                  sizeof(struct frag_hdr));
+               }
+       }
+
+out:
+       return segs;
+}
+
+struct ipv6_gro_cb {
+       struct napi_gro_cb napi;
+       int proto;
+};
+
+#define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
+
+static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
+                                        struct sk_buff *skb)
+{
+       const struct inet6_protocol *ops;
+       struct sk_buff **pp = NULL;
+       struct sk_buff *p;
+       struct ipv6hdr *iph;
+       unsigned int nlen;
+       unsigned int hlen;
+       unsigned int off;
+       int flush = 1;
+       int proto;
+       __wsum csum;
+
+       off = skb_gro_offset(skb);
+       hlen = off + sizeof(*iph);
+       iph = skb_gro_header_fast(skb, off);
+       if (skb_gro_header_hard(skb, hlen)) {
+               iph = skb_gro_header_slow(skb, hlen, off);
+               if (unlikely(!iph))
+                       goto out;
+       }
+
+       skb_gro_pull(skb, sizeof(*iph));
+       skb_set_transport_header(skb, skb_gro_offset(skb));
+
+       flush += ntohs(iph->payload_len) != skb_gro_len(skb);
+
+       rcu_read_lock();
+       proto = iph->nexthdr;
+       ops = rcu_dereference(inet6_protos[proto]);
+       if (!ops || !ops->gro_receive) {
+               __pskb_pull(skb, skb_gro_offset(skb));
+               proto = ipv6_gso_pull_exthdrs(skb, proto);
+               skb_gro_pull(skb, -skb_transport_offset(skb));
+               skb_reset_transport_header(skb);
+               __skb_push(skb, skb_gro_offset(skb));
+
+               if (!ops || !ops->gro_receive)
+                       goto out_unlock;
+
+               iph = ipv6_hdr(skb);
+       }
+
+       IPV6_GRO_CB(skb)->proto = proto;
+
+       flush--;
+       nlen = skb_network_header_len(skb);
+
+       for (p = *head; p; p = p->next) {
+               struct ipv6hdr *iph2;
+
+               if (!NAPI_GRO_CB(p)->same_flow)
+                       continue;
+
+               iph2 = ipv6_hdr(p);
+
+               /* All fields must match except length. */
+               if (nlen != skb_network_header_len(p) ||
+                   memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
+                   memcmp(&iph->nexthdr, &iph2->nexthdr,
+                          nlen - offsetof(struct ipv6hdr, nexthdr))) {
+                       NAPI_GRO_CB(p)->same_flow = 0;
+                       continue;
+               }
+
+               NAPI_GRO_CB(p)->flush |= flush;
+       }
+
+       NAPI_GRO_CB(skb)->flush |= flush;
+
+       csum = skb->csum;
+       skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
+
+       pp = ops->gro_receive(head, skb);
+
+       skb->csum = csum;
+
+out_unlock:
+       rcu_read_unlock();
+
+out:
+       NAPI_GRO_CB(skb)->flush |= flush;
+
+       return pp;
+}
+
+static int ipv6_gro_complete(struct sk_buff *skb)
+{
+       const struct inet6_protocol *ops;
+       struct ipv6hdr *iph = ipv6_hdr(skb);
+       int err = -ENOSYS;
+
+       iph->payload_len = htons(skb->len - skb_network_offset(skb) -
+                                sizeof(*iph));
+
+       rcu_read_lock();
+       ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
+       if (WARN_ON(!ops || !ops->gro_complete))
+               goto out_unlock;
+
+       err = ops->gro_complete(skb);
+
+out_unlock:
+       rcu_read_unlock();
+
+       return err;
+}
+
+static struct packet_type ipv6_packet_type __read_mostly = {
+       .type = cpu_to_be16(ETH_P_IPV6),
+       .func = ipv6_rcv,
+       .gso_send_check = ipv6_gso_send_check,
+       .gso_segment = ipv6_gso_segment,
+       .gro_receive = ipv6_gro_receive,
+       .gro_complete = ipv6_gro_complete,
+};
+
+static int __init ipv6_packet_init(void)
+{
+       dev_add_pack(&ipv6_packet_type);
+       return 0;
+}
+
+static void ipv6_packet_cleanup(void)
+{
+       dev_remove_pack(&ipv6_packet_type);
+}
+
+static int __net_init ipv6_init_mibs(struct net *net)
 {
-       if (snmp_mib_init((void **)ipv6_statistics,
+       if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
+                         sizeof (struct udp_mib)) < 0)
+               return -ENOMEM;
+       if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
+                         sizeof (struct udp_mib)) < 0)
+               goto err_udplite_mib;
+       if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
                          sizeof(struct ipstats_mib)) < 0)
                goto err_ip_mib;
-       if (snmp_mib_init((void **)icmpv6_statistics,
+       if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
                          sizeof(struct icmpv6_mib)) < 0)
                goto err_icmp_mib;
-       if (snmp_mib_init((void **)icmpv6msg_statistics,
+       if (snmp_mib_init((void __percpu **)net->mib.icmpv6msg_statistics,
                          sizeof(struct icmpv6msg_mib)) < 0)
                goto err_icmpmsg_mib;
-       if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib)) < 0)
-               goto err_udp_mib;
-       if (snmp_mib_init((void **)udplite_stats_in6,
-                         sizeof (struct udp_mib)) < 0)
-               goto err_udplite_mib;
        return 0;
 
-err_udplite_mib:
-       snmp_mib_free((void **)udp_stats_in6);
-err_udp_mib:
-       snmp_mib_free((void **)icmpv6msg_statistics);
 err_icmpmsg_mib:
-       snmp_mib_free((void **)icmpv6_statistics);
+       snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
 err_icmp_mib:
-       snmp_mib_free((void **)ipv6_statistics);
+       snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
 err_ip_mib:
+       snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
+err_udplite_mib:
+       snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
        return -ENOMEM;
+}
 
+static void ipv6_cleanup_mibs(struct net *net)
+{
+       snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
+       snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
+       snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
+       snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
+       snmp_mib_free((void __percpu **)net->mib.icmpv6msg_statistics);
 }
 
-static void cleanup_ipv6_mibs(void)
+static int __net_init inet6_net_init(struct net *net)
 {
-       snmp_mib_free((void **)ipv6_statistics);
-       snmp_mib_free((void **)icmpv6_statistics);
-       snmp_mib_free((void **)icmpv6msg_statistics);
-       snmp_mib_free((void **)udp_stats_in6);
-       snmp_mib_free((void **)udplite_stats_in6);
+       int err = 0;
+
+       net->ipv6.sysctl.bindv6only = 0;
+       net->ipv6.sysctl.icmpv6_time = 1*HZ;
+
+       err = ipv6_init_mibs(net);
+       if (err)
+               return err;
+#ifdef CONFIG_PROC_FS
+       err = udp6_proc_init(net);
+       if (err)
+               goto out;
+       err = tcp6_proc_init(net);
+       if (err)
+               goto proc_tcp6_fail;
+       err = ac6_proc_init(net);
+       if (err)
+               goto proc_ac6_fail;
+#endif
+       return err;
+
+#ifdef CONFIG_PROC_FS
+proc_ac6_fail:
+       tcp6_proc_exit(net);
+proc_tcp6_fail:
+       udp6_proc_exit(net);
+out:
+       ipv6_cleanup_mibs(net);
+       return err;
+#endif
 }
 
+static void __net_exit inet6_net_exit(struct net *net)
+{
+#ifdef CONFIG_PROC_FS
+       udp6_proc_exit(net);
+       tcp6_proc_exit(net);
+       ac6_proc_exit(net);
+#endif
+       ipv6_cleanup_mibs(net);
+}
+
+static struct pernet_operations inet6_net_ops = {
+       .init = inet6_net_init,
+       .exit = inet6_net_exit,
+};
+
 static int __init inet6_init(void)
 {
        struct sk_buff *dummy_skb;
        struct list_head *r;
-       int err;
+       int err = 0;
 
        BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
 
-#ifdef MODULE
-#if 0 /* FIXME --RR */
-       if (!mod_member_present(&__this_module, can_unload))
-         return -EINVAL;
+       /* Register the socket-side information for inet6_create.  */
+       for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
+               INIT_LIST_HEAD(r);
 
-       __this_module.can_unload = &ipv6_unload;
-#endif
-#endif
+       if (disable_ipv6_mod) {
+               printk(KERN_INFO
+                      "IPv6: Loaded, but administratively disabled, "
+                      "reboot required to enable\n");
+               goto out;
+       }
 
        err = proto_register(&tcpv6_prot, 1);
        if (err)
@@ -786,27 +1094,25 @@ static int __init inet6_init(void)
                goto out_unregister_udplite_proto;
 
 
-       /* Register the socket-side information for inet6_create.  */
-       for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
-               INIT_LIST_HEAD(r);
-
        /* We MUST register RAW sockets before we create the ICMP6,
         * IGMP6, or NDISC control sockets.
         */
-       inet6_register_protosw(&rawv6_protosw);
+       err = rawv6_init();
+       if (err)
+               goto out_unregister_raw_proto;
 
        /* Register the family here so that the init calls below will
         * be able to create sockets. (?? is this dangerous ??)
         */
        err = sock_register(&inet6_family_ops);
        if (err)
-               goto out_unregister_raw_proto;
+               goto out_sock_register_fail;
 
-       /* Initialise ipv6 mibs */
-       err = init_ipv6_mibs();
+#ifdef CONFIG_SYSCTL
+       err = ipv6_static_sysctl_register();
        if (err)
-               goto out_unregister_sock;
-
+               goto static_sysctl_fail;
+#endif
        /*
         *      ipngwg API draft makes clear that the correct semantics
         *      for TCP and UDP is to consider one TCP and UDP instance
@@ -814,16 +1120,19 @@ static int __init inet6_init(void)
         *      able to communicate via both network protocols.
         */
 
-#ifdef CONFIG_SYSCTL
-       ipv6_sysctl_register();
-#endif
-       err = icmpv6_init(&inet6_family_ops);
+       err = register_pernet_subsys(&inet6_net_ops);
+       if (err)
+               goto register_pernet_fail;
+       err = icmpv6_init();
        if (err)
                goto icmp_fail;
-       err = ndisc_init(&inet6_family_ops);
+       err = ip6_mr_init();
+       if (err)
+               goto ipmr_fail;
+       err = ndisc_init();
        if (err)
                goto ndisc_fail;
-       err = igmp6_init(&inet6_family_ops);
+       err = igmp6_init();
        if (err)
                goto igmp_fail;
        err = ipv6_netfilter_init();
@@ -834,61 +1143,85 @@ static int __init inet6_init(void)
        err = -ENOMEM;
        if (raw6_proc_init())
                goto proc_raw6_fail;
-       if (tcp6_proc_init())
-               goto proc_tcp6_fail;
-       if (udp6_proc_init())
-               goto proc_udp6_fail;
        if (udplite6_proc_init())
                goto proc_udplite6_fail;
        if (ipv6_misc_proc_init())
                goto proc_misc6_fail;
-
-       if (ac6_proc_init())
-               goto proc_anycast6_fail;
        if (if6_proc_init())
                goto proc_if6_fail;
 #endif
        err = ip6_route_init();
        if (err)
                goto ip6_route_fail;
-       ip6_flowlabel_init();
+       err = ip6_flowlabel_init();
+       if (err)
+               goto ip6_flowlabel_fail;
        err = addrconf_init();
        if (err)
                goto addrconf_fail;
 
        /* Init v6 extension headers. */
-       ipv6_rthdr_init();
-       ipv6_frag_init();
-       ipv6_nodata_init();
-       ipv6_destopt_init();
+       err = ipv6_exthdrs_init();
+       if (err)
+               goto ipv6_exthdrs_fail;
+
+       err = ipv6_frag_init();
+       if (err)
+               goto ipv6_frag_fail;
 
        /* Init v6 transport protocols. */
-       udpv6_init();
-       udplitev6_init();
-       tcpv6_init();
+       err = udpv6_init();
+       if (err)
+               goto udpv6_fail;
 
-       ipv6_packet_init();
-       err = 0;
+       err = udplitev6_init();
+       if (err)
+               goto udplitev6_fail;
+
+       err = tcpv6_init();
+       if (err)
+               goto tcpv6_fail;
+
+       err = ipv6_packet_init();
+       if (err)
+               goto ipv6_packet_fail;
+
+#ifdef CONFIG_SYSCTL
+       err = ipv6_sysctl_register();
+       if (err)
+               goto sysctl_fail;
+#endif
 out:
        return err;
 
+#ifdef CONFIG_SYSCTL
+sysctl_fail:
+       ipv6_packet_cleanup();
+#endif
+ipv6_packet_fail:
+       tcpv6_exit();
+tcpv6_fail:
+       udplitev6_exit();
+udplitev6_fail:
+       udpv6_exit();
+udpv6_fail:
+       ipv6_frag_exit();
+ipv6_frag_fail:
+       ipv6_exthdrs_exit();
+ipv6_exthdrs_fail:
+       addrconf_cleanup();
 addrconf_fail:
        ip6_flowlabel_cleanup();
+ip6_flowlabel_fail:
        ip6_route_cleanup();
 ip6_route_fail:
 #ifdef CONFIG_PROC_FS
        if6_proc_exit();
 proc_if6_fail:
-       ac6_proc_exit();
-proc_anycast6_fail:
        ipv6_misc_proc_exit();
 proc_misc6_fail:
        udplite6_proc_exit();
 proc_udplite6_fail:
-       udp6_proc_exit();
-proc_udp6_fail:
-       tcp6_proc_exit();
-proc_tcp6_fail:
        raw6_proc_exit();
 proc_raw6_fail:
 #endif
@@ -898,15 +1231,20 @@ netfilter_fail:
 igmp_fail:
        ndisc_cleanup();
 ndisc_fail:
+       ip6_mr_cleanup();
+ipmr_fail:
        icmpv6_cleanup();
 icmp_fail:
+       unregister_pernet_subsys(&inet6_net_ops);
+register_pernet_fail:
 #ifdef CONFIG_SYSCTL
-       ipv6_sysctl_unregister();
+       ipv6_static_sysctl_unregister();
+static_sysctl_fail:
 #endif
-       cleanup_ipv6_mibs();
-out_unregister_sock:
        sock_unregister(PF_INET6);
        rtnl_unregister_all(PF_INET6);
+out_sock_register_fail:
+       rawv6_exit();
 out_unregister_raw_proto:
        proto_unregister(&rawv6_prot);
 out_unregister_udplite_proto:
@@ -921,14 +1259,25 @@ module_init(inet6_init);
 
 static void __exit inet6_exit(void)
 {
+       if (disable_ipv6_mod)
+               return;
+
        /* First of all disallow new sockets creation. */
        sock_unregister(PF_INET6);
        /* Disallow any further netlink messages */
        rtnl_unregister_all(PF_INET6);
 
+#ifdef CONFIG_SYSCTL
+       ipv6_sysctl_unregister();
+#endif
+       udpv6_exit();
+       udplitev6_exit();
+       tcpv6_exit();
+
        /* Cleanup code parts. */
        ipv6_packet_cleanup();
-
+       ipv6_frag_exit();
+       ipv6_exthdrs_exit();
        addrconf_cleanup();
        ip6_flowlabel_cleanup();
        ip6_route_cleanup();
@@ -936,25 +1285,27 @@ static void __exit inet6_exit(void)
 
        /* Cleanup code parts. */
        if6_proc_exit();
-       ac6_proc_exit();
        ipv6_misc_proc_exit();
        udplite6_proc_exit();
-       udp6_proc_exit();
-       tcp6_proc_exit();
        raw6_proc_exit();
 #endif
        ipv6_netfilter_fini();
        igmp6_cleanup();
        ndisc_cleanup();
+       ip6_mr_cleanup();
        icmpv6_cleanup();
+       rawv6_exit();
+
+       unregister_pernet_subsys(&inet6_net_ops);
 #ifdef CONFIG_SYSCTL
-       ipv6_sysctl_unregister();
+       ipv6_static_sysctl_unregister();
 #endif
-       cleanup_ipv6_mibs();
        proto_unregister(&rawv6_prot);
        proto_unregister(&udplitev6_prot);
        proto_unregister(&udpv6_prot);
        proto_unregister(&tcpv6_prot);
+
+       rcu_barrier(); /* Wait for completion of call_rcu()'s */
 }
 module_exit(inet6_exit);