IPv6: fix race between cleanup and add/delete address
[linux-2.6.git] / net / netfilter / nf_conntrack_netlink.c
index d93d58d..2b2af63 100644 (file)
@@ -4,7 +4,7 @@
  * (C) 2001 by Jay Schulist <jschlst@samba.org>
  * (C) 2002-2006 by Harald Welte <laforge@gnumonks.org>
  * (C) 2003 by Patrick Mchardy <kaber@trash.net>
- * (C) 2005-2007 by Pablo Neira Ayuso <pablo@netfilter.org>
+ * (C) 2005-2008 by Pablo Neira Ayuso <pablo@netfilter.org>
  *
  * Initial connection tracking via netlink development funded and
  * generally made possible by Network Robots, Inc. (www.networkrobots.com)
@@ -18,6 +18,8 @@
 #include <linux/init.h>
 #include <linux/module.h>
 #include <linux/kernel.h>
+#include <linux/rculist.h>
+#include <linux/rculist_nulls.h>
 #include <linux/types.h>
 #include <linux/timer.h>
 #include <linux/skbuff.h>
 #include <linux/netlink.h>
 #include <linux/spinlock.h>
 #include <linux/interrupt.h>
-#include <linux/notifier.h>
 
 #include <linux/netfilter.h>
 #include <net/netlink.h>
+#include <net/sock.h>
 #include <net/netfilter/nf_conntrack.h>
 #include <net/netfilter/nf_conntrack_core.h>
 #include <net/netfilter/nf_conntrack_expect.h>
@@ -36,6 +38,8 @@
 #include <net/netfilter/nf_conntrack_l3proto.h>
 #include <net/netfilter/nf_conntrack_l4proto.h>
 #include <net/netfilter/nf_conntrack_tuple.h>
+#include <net/netfilter/nf_conntrack_acct.h>
+#include <net/netfilter/nf_conntrack_zones.h>
 #ifdef CONFIG_NF_NAT_NEEDED
 #include <net/netfilter/nf_nat_core.h>
 #include <net/netfilter/nf_nat_protocol.h>
@@ -95,7 +99,7 @@ nla_put_failure:
        return -1;
 }
 
-static inline int
+static int
 ctnetlink_dump_tuples(struct sk_buff *skb,
                      const struct nf_conntrack_tuple *tuple)
 {
@@ -103,16 +107,14 @@ ctnetlink_dump_tuples(struct sk_buff *skb,
        struct nf_conntrack_l3proto *l3proto;
        struct nf_conntrack_l4proto *l4proto;
 
-       l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
+       l3proto = __nf_ct_l3proto_find(tuple->src.l3num);
        ret = ctnetlink_dump_tuples_ip(skb, tuple, l3proto);
-       nf_ct_l3proto_put(l3proto);
 
        if (unlikely(ret < 0))
                return ret;
 
-       l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
+       l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
        ret = ctnetlink_dump_tuples_proto(skb, tuple, l4proto);
-       nf_ct_l4proto_put(l4proto);
 
        return ret;
 }
@@ -143,16 +145,15 @@ nla_put_failure:
 }
 
 static inline int
-ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct nf_conn *ct)
+ctnetlink_dump_protoinfo(struct sk_buff *skb, struct nf_conn *ct)
 {
-       struct nf_conntrack_l4proto *l4proto = nf_ct_l4proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num, ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
+       struct nf_conntrack_l4proto *l4proto;
        struct nlattr *nest_proto;
        int ret;
 
-       if (!l4proto->to_nlattr) {
-               nf_ct_l4proto_put(l4proto);
+       l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
+       if (!l4proto->to_nlattr)
                return 0;
-       }
 
        nest_proto = nla_nest_start(skb, CTA_PROTOINFO | NLA_F_NESTED);
        if (!nest_proto)
@@ -160,14 +161,11 @@ ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct nf_conn *ct)
 
        ret = l4proto->to_nlattr(skb, nest_proto, ct);
 
-       nf_ct_l4proto_put(l4proto);
-
        nla_nest_end(skb, nest_proto);
 
        return ret;
 
 nla_put_failure:
-       nf_ct_l4proto_put(l4proto);
        return -1;
 }
 
@@ -181,7 +179,6 @@ ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct nf_conn *ct)
        if (!help)
                return 0;
 
-       rcu_read_lock();
        helper = rcu_dereference(help->helper);
        if (!helper)
                goto out;
@@ -196,30 +193,32 @@ ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct nf_conn *ct)
 
        nla_nest_end(skb, nest_helper);
 out:
-       rcu_read_unlock();
        return 0;
 
 nla_put_failure:
-       rcu_read_unlock();
        return -1;
 }
 
-#ifdef CONFIG_NF_CT_ACCT
-static inline int
+static int
 ctnetlink_dump_counters(struct sk_buff *skb, const struct nf_conn *ct,
                        enum ip_conntrack_dir dir)
 {
        enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
        struct nlattr *nest_count;
+       const struct nf_conn_counter *acct;
+
+       acct = nf_conn_acct_find(ct);
+       if (!acct)
+               return 0;
 
        nest_count = nla_nest_start(skb, type | NLA_F_NESTED);
        if (!nest_count)
                goto nla_put_failure;
 
-       NLA_PUT_BE32(skb, CTA_COUNTERS32_PACKETS,
-                    htonl(ct->counters[dir].packets));
-       NLA_PUT_BE32(skb, CTA_COUNTERS32_BYTES,
-                    htonl(ct->counters[dir].bytes));
+       NLA_PUT_BE64(skb, CTA_COUNTERS_PACKETS,
+                    cpu_to_be64(acct[dir].packets));
+       NLA_PUT_BE64(skb, CTA_COUNTERS_BYTES,
+                    cpu_to_be64(acct[dir].bytes));
 
        nla_nest_end(skb, nest_count);
 
@@ -228,9 +227,6 @@ ctnetlink_dump_counters(struct sk_buff *skb, const struct nf_conn *ct,
 nla_put_failure:
        return -1;
 }
-#else
-#define ctnetlink_dump_counters(a, b, c) (0)
-#endif
 
 #ifdef CONFIG_NF_CONNTRACK_MARK
 static inline int
@@ -284,7 +280,7 @@ nla_put_failure:
 }
 
 #ifdef CONFIG_NF_NAT_NEEDED
-static inline int
+static int
 dump_nat_seq_adj(struct sk_buff *skb, const struct nf_nat_seq *natseq, int type)
 {
        struct nlattr *nest_parms;
@@ -351,42 +347,42 @@ nla_put_failure:
        return -1;
 }
 
-#define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
-
 static int
 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
-                   int event, int nowait,
-                   const struct nf_conn *ct)
+                   int event, struct nf_conn *ct)
 {
        struct nlmsghdr *nlh;
        struct nfgenmsg *nfmsg;
        struct nlattr *nest_parms;
-       unsigned char *b = skb_tail_pointer(skb);
+       unsigned int flags = pid ? NLM_F_MULTI : 0;
 
        event |= NFNL_SUBSYS_CTNETLINK << 8;
-       nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
-       nfmsg  = NLMSG_DATA(nlh);
+       nlh = nlmsg_put(skb, pid, seq, event, sizeof(*nfmsg), flags);
+       if (nlh == NULL)
+               goto nlmsg_failure;
 
-       nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
-       nfmsg->nfgen_family =
-               ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
+       nfmsg = nlmsg_data(nlh);
+       nfmsg->nfgen_family = nf_ct_l3num(ct);
        nfmsg->version      = NFNETLINK_V0;
        nfmsg->res_id       = 0;
 
        nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
        if (!nest_parms)
                goto nla_put_failure;
-       if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
+       if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
                goto nla_put_failure;
        nla_nest_end(skb, nest_parms);
 
        nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
        if (!nest_parms)
                goto nla_put_failure;
-       if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
+       if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0)
                goto nla_put_failure;
        nla_nest_end(skb, nest_parms);
 
+       if (nf_ct_zone(ct))
+               NLA_PUT_BE16(skb, CTA_ZONE, htons(nf_ct_zone(ct)));
+
        if (ctnetlink_dump_status(skb, ct) < 0 ||
            ctnetlink_dump_timeout(skb, ct) < 0 ||
            ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
@@ -401,129 +397,193 @@ ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
            ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
                goto nla_put_failure;
 
-       nlh->nlmsg_len = skb_tail_pointer(skb) - b;
+       nlmsg_end(skb, nlh);
        return skb->len;
 
 nlmsg_failure:
 nla_put_failure:
-       nlmsg_trim(skb, b);
+       nlmsg_cancel(skb, nlh);
        return -1;
 }
 
 #ifdef CONFIG_NF_CONNTRACK_EVENTS
-static int ctnetlink_conntrack_event(struct notifier_block *this,
-                                    unsigned long events, void *ptr)
+static inline size_t
+ctnetlink_proto_size(const struct nf_conn *ct)
 {
+       struct nf_conntrack_l3proto *l3proto;
+       struct nf_conntrack_l4proto *l4proto;
+       size_t len = 0;
+
+       rcu_read_lock();
+       l3proto = __nf_ct_l3proto_find(nf_ct_l3num(ct));
+       len += l3proto->nla_size;
+
+       l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
+       len += l4proto->nla_size;
+       rcu_read_unlock();
+
+       return len;
+}
+
+static inline size_t
+ctnetlink_nlmsg_size(const struct nf_conn *ct)
+{
+       return NLMSG_ALIGN(sizeof(struct nfgenmsg))
+              + 3 * nla_total_size(0) /* CTA_TUPLE_ORIG|REPL|MASTER */
+              + 3 * nla_total_size(0) /* CTA_TUPLE_IP */
+              + 3 * nla_total_size(0) /* CTA_TUPLE_PROTO */
+              + 3 * nla_total_size(sizeof(u_int8_t)) /* CTA_PROTO_NUM */
+              + nla_total_size(sizeof(u_int32_t)) /* CTA_ID */
+              + nla_total_size(sizeof(u_int32_t)) /* CTA_STATUS */
+#ifdef CONFIG_NF_CT_ACCT
+              + 2 * nla_total_size(0) /* CTA_COUNTERS_ORIG|REPL */
+              + 2 * nla_total_size(sizeof(uint64_t)) /* CTA_COUNTERS_PACKETS */
+              + 2 * nla_total_size(sizeof(uint64_t)) /* CTA_COUNTERS_BYTES */
+#endif
+              + nla_total_size(sizeof(u_int32_t)) /* CTA_TIMEOUT */
+              + nla_total_size(0) /* CTA_PROTOINFO */
+              + nla_total_size(0) /* CTA_HELP */
+              + nla_total_size(NF_CT_HELPER_NAME_LEN) /* CTA_HELP_NAME */
+#ifdef CONFIG_NF_CONNTRACK_SECMARK
+              + nla_total_size(sizeof(u_int32_t)) /* CTA_SECMARK */
+#endif
+#ifdef CONFIG_NF_NAT_NEEDED
+              + 2 * nla_total_size(0) /* CTA_NAT_SEQ_ADJ_ORIG|REPL */
+              + 6 * nla_total_size(sizeof(u_int32_t)) /* CTA_NAT_SEQ_OFFSET */
+#endif
+#ifdef CONFIG_NF_CONNTRACK_MARK
+              + nla_total_size(sizeof(u_int32_t)) /* CTA_MARK */
+#endif
+              + ctnetlink_proto_size(ct)
+              ;
+}
+
+static int
+ctnetlink_conntrack_event(unsigned int events, struct nf_ct_event *item)
+{
+       struct net *net;
        struct nlmsghdr *nlh;
        struct nfgenmsg *nfmsg;
        struct nlattr *nest_parms;
-       struct nf_conn *ct = (struct nf_conn *)ptr;
+       struct nf_conn *ct = item->ct;
        struct sk_buff *skb;
        unsigned int type;
-       sk_buff_data_t b;
        unsigned int flags = 0, group;
+       int err;
 
        /* ignore our fake conntrack entry */
        if (ct == &nf_conntrack_untracked)
-               return NOTIFY_DONE;
+               return 0;
 
-       if (events & IPCT_DESTROY) {
+       if (events & (1 << IPCT_DESTROY)) {
                type = IPCTNL_MSG_CT_DELETE;
                group = NFNLGRP_CONNTRACK_DESTROY;
-       } else  if (events & (IPCT_NEW | IPCT_RELATED)) {
+       } else  if (events & ((1 << IPCT_NEW) | (1 << IPCT_RELATED))) {
                type = IPCTNL_MSG_CT_NEW;
                flags = NLM_F_CREATE|NLM_F_EXCL;
                group = NFNLGRP_CONNTRACK_NEW;
-       } else  if (events & (IPCT_STATUS | IPCT_PROTOINFO)) {
+       } else  if (events) {
                type = IPCTNL_MSG_CT_NEW;
                group = NFNLGRP_CONNTRACK_UPDATE;
        } else
-               return NOTIFY_DONE;
-
-       if (!nfnetlink_has_listeners(group))
-               return NOTIFY_DONE;
+               return 0;
 
-       skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
-       if (!skb)
-               return NOTIFY_DONE;
+       net = nf_ct_net(ct);
+       if (!item->report && !nfnetlink_has_listeners(net, group))
+               return 0;
 
-       b = skb->tail;
+       skb = nlmsg_new(ctnetlink_nlmsg_size(ct), GFP_ATOMIC);
+       if (skb == NULL)
+               goto errout;
 
        type |= NFNL_SUBSYS_CTNETLINK << 8;
-       nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
-       nfmsg = NLMSG_DATA(nlh);
+       nlh = nlmsg_put(skb, item->pid, 0, type, sizeof(*nfmsg), flags);
+       if (nlh == NULL)
+               goto nlmsg_failure;
 
-       nlh->nlmsg_flags    = flags;
-       nfmsg->nfgen_family = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
+       nfmsg = nlmsg_data(nlh);
+       nfmsg->nfgen_family = nf_ct_l3num(ct);
        nfmsg->version  = NFNETLINK_V0;
        nfmsg->res_id   = 0;
 
+       rcu_read_lock();
        nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG | NLA_F_NESTED);
        if (!nest_parms)
                goto nla_put_failure;
-       if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
+       if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
                goto nla_put_failure;
        nla_nest_end(skb, nest_parms);
 
        nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY | NLA_F_NESTED);
        if (!nest_parms)
                goto nla_put_failure;
-       if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
+       if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0)
                goto nla_put_failure;
        nla_nest_end(skb, nest_parms);
 
-       if (events & IPCT_DESTROY) {
+       if (nf_ct_zone(ct))
+               NLA_PUT_BE16(skb, CTA_ZONE, htons(nf_ct_zone(ct)));
+
+       if (ctnetlink_dump_id(skb, ct) < 0)
+               goto nla_put_failure;
+
+       if (ctnetlink_dump_status(skb, ct) < 0)
+               goto nla_put_failure;
+
+       if (events & (1 << IPCT_DESTROY)) {
                if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
                    ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
                        goto nla_put_failure;
        } else {
-               if (ctnetlink_dump_status(skb, ct) < 0)
-                       goto nla_put_failure;
-
                if (ctnetlink_dump_timeout(skb, ct) < 0)
                        goto nla_put_failure;
 
-               if (events & IPCT_PROTOINFO
+               if (events & (1 << IPCT_PROTOINFO)
                    && ctnetlink_dump_protoinfo(skb, ct) < 0)
                        goto nla_put_failure;
 
-               if ((events & IPCT_HELPER || nfct_help(ct))
+               if ((events & (1 << IPCT_HELPER) || nfct_help(ct))
                    && ctnetlink_dump_helpinfo(skb, ct) < 0)
                        goto nla_put_failure;
 
-#ifdef CONFIG_NF_CONNTRACK_MARK
-               if ((events & IPCT_MARK || ct->mark)
-                   && ctnetlink_dump_mark(skb, ct) < 0)
-                       goto nla_put_failure;
-#endif
 #ifdef CONFIG_NF_CONNTRACK_SECMARK
-               if ((events & IPCT_SECMARK || ct->secmark)
+               if ((events & (1 << IPCT_SECMARK) || ct->secmark)
                    && ctnetlink_dump_secmark(skb, ct) < 0)
                        goto nla_put_failure;
 #endif
 
-               if (events & IPCT_COUNTER_FILLING &&
-                   (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
-                    ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0))
-                       goto nla_put_failure;
-
-               if (events & IPCT_RELATED &&
+               if (events & (1 << IPCT_RELATED) &&
                    ctnetlink_dump_master(skb, ct) < 0)
                        goto nla_put_failure;
 
-               if (events & IPCT_NATSEQADJ &&
+               if (events & (1 << IPCT_NATSEQADJ) &&
                    ctnetlink_dump_nat_seq_adj(skb, ct) < 0)
                        goto nla_put_failure;
        }
 
-       nlh->nlmsg_len = skb->tail - b;
-       nfnetlink_send(skb, 0, group, 0);
-       return NOTIFY_DONE;
+#ifdef CONFIG_NF_CONNTRACK_MARK
+       if ((events & (1 << IPCT_MARK) || ct->mark)
+           && ctnetlink_dump_mark(skb, ct) < 0)
+               goto nla_put_failure;
+#endif
+       rcu_read_unlock();
+
+       nlmsg_end(skb, nlh);
+       err = nfnetlink_send(skb, net, item->pid, group, item->report,
+                            GFP_ATOMIC);
+       if (err == -ENOBUFS || err == -EAGAIN)
+               return -ENOBUFS;
+
+       return 0;
 
-nlmsg_failure:
 nla_put_failure:
+       rcu_read_unlock();
+       nlmsg_cancel(skb, nlh);
+nlmsg_failure:
        kfree_skb(skb);
-       return NOTIFY_DONE;
+errout:
+       nfnetlink_set_err(net, 0, group, -ENOBUFS);
+       return 0;
 }
 #endif /* CONFIG_NF_CONNTRACK_EVENTS */
 
@@ -534,49 +594,54 @@ static int ctnetlink_done(struct netlink_callback *cb)
        return 0;
 }
 
-#define L3PROTO(ct) (ct)->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num
-
 static int
 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 {
+       struct net *net = sock_net(skb->sk);
        struct nf_conn *ct, *last;
        struct nf_conntrack_tuple_hash *h;
-       struct hlist_node *n;
-       struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
+       struct hlist_nulls_node *n;
+       struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
        u_int8_t l3proto = nfmsg->nfgen_family;
 
-       read_lock_bh(&nf_conntrack_lock);
+       rcu_read_lock();
        last = (struct nf_conn *)cb->args[1];
-       for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) {
+       for (; cb->args[0] < net->ct.htable_size; cb->args[0]++) {
 restart:
-               hlist_for_each_entry(h, n, &nf_conntrack_hash[cb->args[0]],
-                                    hnode) {
+               hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[cb->args[0]],
+                                        hnnode) {
                        if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL)
                                continue;
                        ct = nf_ct_tuplehash_to_ctrack(h);
+                       if (!atomic_inc_not_zero(&ct->ct_general.use))
+                               continue;
                        /* Dump entries of a given L3 protocol number.
                         * If it is not specified, ie. l3proto == 0,
                         * then dump everything. */
-                       if (l3proto && L3PROTO(ct) != l3proto)
-                               continue;
+                       if (l3proto && nf_ct_l3num(ct) != l3proto)
+                               goto releasect;
                        if (cb->args[1]) {
                                if (ct != last)
-                                       continue;
+                                       goto releasect;
                                cb->args[1] = 0;
                        }
                        if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
                                                cb->nlh->nlmsg_seq,
-                                               IPCTNL_MSG_CT_NEW,
-                                               1, ct) < 0) {
-                               nf_conntrack_get(&ct->ct_general);
+                                               IPCTNL_MSG_CT_NEW, ct) < 0) {
                                cb->args[1] = (unsigned long)ct;
                                goto out;
                        }
-#ifdef CONFIG_NF_CT_ACCT
+
                        if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
-                                               IPCTNL_MSG_CT_GET_CTRZERO)
-                               memset(&ct->counters, 0, sizeof(ct->counters));
-#endif
+                                               IPCTNL_MSG_CT_GET_CTRZERO) {
+                               struct nf_conn_counter *acct;
+
+                               acct = nf_conn_acct_find(ct);
+                               if (acct)
+                                       memset(acct, 0, sizeof(struct nf_conn_counter[IP_CT_DIR_MAX]));
+                       }
+releasect:
+               nf_ct_put(ct);
                }
                if (cb->args[1]) {
                        cb->args[1] = 0;
@@ -584,7 +649,7 @@ restart:
                }
        }
 out:
-       read_unlock_bh(&nf_conntrack_lock);
+       rcu_read_unlock();
        if (last)
                nf_ct_put(last);
 
@@ -600,7 +665,8 @@ ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
 
        nla_parse_nested(tb, CTA_IP_MAX, attr, NULL);
 
-       l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
+       rcu_read_lock();
+       l3proto = __nf_ct_l3proto_find(tuple->src.l3num);
 
        if (likely(l3proto->nlattr_to_tuple)) {
                ret = nla_validate_nested(attr, CTA_IP_MAX,
@@ -609,7 +675,7 @@ ctnetlink_parse_tuple_ip(struct nlattr *attr, struct nf_conntrack_tuple *tuple)
                        ret = l3proto->nlattr_to_tuple(tb, tuple);
        }
 
-       nf_ct_l3proto_put(l3proto);
+       rcu_read_unlock();
 
        return ret;
 }
@@ -634,7 +700,8 @@ ctnetlink_parse_tuple_proto(struct nlattr *attr,
                return -EINVAL;
        tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]);
 
-       l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
+       rcu_read_lock();
+       l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
 
        if (likely(l4proto->nlattr_to_tuple)) {
                ret = nla_validate_nested(attr, CTA_PROTO_MAX,
@@ -643,13 +710,19 @@ ctnetlink_parse_tuple_proto(struct nlattr *attr,
                        ret = l4proto->nlattr_to_tuple(tb, tuple);
        }
 
-       nf_ct_l4proto_put(l4proto);
+       rcu_read_unlock();
 
        return ret;
 }
 
-static inline int
-ctnetlink_parse_tuple(struct nlattr *cda[], struct nf_conntrack_tuple *tuple,
+static const struct nla_policy tuple_nla_policy[CTA_TUPLE_MAX+1] = {
+       [CTA_TUPLE_IP]          = { .type = NLA_NESTED },
+       [CTA_TUPLE_PROTO]       = { .type = NLA_NESTED },
+};
+
+static int
+ctnetlink_parse_tuple(const struct nlattr * const cda[],
+                     struct nf_conntrack_tuple *tuple,
                      enum ctattr_tuple type, u_int8_t l3num)
 {
        struct nlattr *tb[CTA_TUPLE_MAX+1];
@@ -657,7 +730,7 @@ ctnetlink_parse_tuple(struct nlattr *cda[], struct nf_conntrack_tuple *tuple,
 
        memset(tuple, 0, sizeof(*tuple));
 
-       nla_parse_nested(tb, CTA_TUPLE_MAX, cda[type], NULL);
+       nla_parse_nested(tb, CTA_TUPLE_MAX, cda[type], tuple_nla_policy);
 
        if (!tb[CTA_TUPLE_IP])
                return -EINVAL;
@@ -684,86 +757,31 @@ ctnetlink_parse_tuple(struct nlattr *cda[], struct nf_conntrack_tuple *tuple,
        return 0;
 }
 
-#ifdef CONFIG_NF_NAT_NEEDED
-static const struct nla_policy protonat_nla_policy[CTA_PROTONAT_MAX+1] = {
-       [CTA_PROTONAT_PORT_MIN] = { .type = NLA_U16 },
-       [CTA_PROTONAT_PORT_MAX] = { .type = NLA_U16 },
-};
-
-static int nfnetlink_parse_nat_proto(struct nlattr *attr,
-                                    const struct nf_conn *ct,
-                                    struct nf_nat_range *range)
+static int
+ctnetlink_parse_zone(const struct nlattr *attr, u16 *zone)
 {
-       struct nlattr *tb[CTA_PROTONAT_MAX+1];
-       const struct nf_nat_protocol *npt;
-       int err;
-
-       err = nla_parse_nested(tb, CTA_PROTONAT_MAX, attr, protonat_nla_policy);
-       if (err < 0)
-               return err;
-
-       npt = nf_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
-
-       if (!npt->nlattr_to_range) {
-               nf_nat_proto_put(npt);
-               return 0;
-       }
-
-       /* nlattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
-       if (npt->nlattr_to_range(tb, range) > 0)
-               range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
-
-       nf_nat_proto_put(npt);
+       if (attr)
+#ifdef CONFIG_NF_CONNTRACK_ZONES
+               *zone = ntohs(nla_get_be16(attr));
+#else
+               return -EOPNOTSUPP;
+#endif
+       else
+               *zone = 0;
 
        return 0;
 }
 
-static const struct nla_policy nat_nla_policy[CTA_NAT_MAX+1] = {
-       [CTA_NAT_MINIP]         = { .type = NLA_U32 },
-       [CTA_NAT_MAXIP]         = { .type = NLA_U32 },
+static const struct nla_policy help_nla_policy[CTA_HELP_MAX+1] = {
+       [CTA_HELP_NAME]         = { .type = NLA_NUL_STRING },
 };
 
 static inline int
-nfnetlink_parse_nat(struct nlattr *nat,
-                   const struct nf_conn *ct, struct nf_nat_range *range)
-{
-       struct nlattr *tb[CTA_NAT_MAX+1];
-       int err;
-
-       memset(range, 0, sizeof(*range));
-
-       err = nla_parse_nested(tb, CTA_NAT_MAX, nat, nat_nla_policy);
-       if (err < 0)
-               return err;
-
-       if (tb[CTA_NAT_MINIP])
-               range->min_ip = nla_get_be32(tb[CTA_NAT_MINIP]);
-
-       if (!tb[CTA_NAT_MAXIP])
-               range->max_ip = range->min_ip;
-       else
-               range->max_ip = nla_get_be32(tb[CTA_NAT_MAXIP]);
-
-       if (range->min_ip)
-               range->flags |= IP_NAT_RANGE_MAP_IPS;
-
-       if (!tb[CTA_NAT_PROTO])
-               return 0;
-
-       err = nfnetlink_parse_nat_proto(tb[CTA_NAT_PROTO], ct, range);
-       if (err < 0)
-               return err;
-
-       return 0;
-}
-#endif
-
-static inline int
-ctnetlink_parse_help(struct nlattr *attr, char **helper_name)
+ctnetlink_parse_help(const struct nlattr *attr, char **helper_name)
 {
        struct nlattr *tb[CTA_HELP_MAX+1];
 
-       nla_parse_nested(tb, CTA_HELP_MAX, attr, NULL);
+       nla_parse_nested(tb, CTA_HELP_MAX, attr, help_nla_policy);
 
        if (!tb[CTA_HELP_NAME])
                return -EINVAL;
@@ -774,23 +792,37 @@ ctnetlink_parse_help(struct nlattr *attr, char **helper_name)
 }
 
 static const struct nla_policy ct_nla_policy[CTA_MAX+1] = {
+       [CTA_TUPLE_ORIG]        = { .type = NLA_NESTED },
+       [CTA_TUPLE_REPLY]       = { .type = NLA_NESTED },
        [CTA_STATUS]            = { .type = NLA_U32 },
+       [CTA_PROTOINFO]         = { .type = NLA_NESTED },
+       [CTA_HELP]              = { .type = NLA_NESTED },
+       [CTA_NAT_SRC]           = { .type = NLA_NESTED },
        [CTA_TIMEOUT]           = { .type = NLA_U32 },
        [CTA_MARK]              = { .type = NLA_U32 },
-       [CTA_USE]               = { .type = NLA_U32 },
        [CTA_ID]                = { .type = NLA_U32 },
+       [CTA_NAT_DST]           = { .type = NLA_NESTED },
+       [CTA_TUPLE_MASTER]      = { .type = NLA_NESTED },
+       [CTA_ZONE]              = { .type = NLA_U16 },
 };
 
 static int
 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
-                       struct nlmsghdr *nlh, struct nlattr *cda[])
+                       const struct nlmsghdr *nlh,
+                       const struct nlattr * const cda[])
 {
+       struct net *net = sock_net(ctnl);
        struct nf_conntrack_tuple_hash *h;
        struct nf_conntrack_tuple tuple;
        struct nf_conn *ct;
-       struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
+       struct nfgenmsg *nfmsg = nlmsg_data(nlh);
        u_int8_t u3 = nfmsg->nfgen_family;
-       int err = 0;
+       u16 zone;
+       int err;
+
+       err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone);
+       if (err < 0)
+               return err;
 
        if (cda[CTA_TUPLE_ORIG])
                err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
@@ -798,14 +830,16 @@ ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
                err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, u3);
        else {
                /* Flush the whole table */
-               nf_conntrack_flush();
+               nf_conntrack_flush_report(net,
+                                        NETLINK_CB(skb).pid,
+                                        nlmsg_report(nlh));
                return 0;
        }
 
        if (err < 0)
                return err;
 
-       h = nf_conntrack_find_get(&tuple);
+       h = nf_conntrack_find_get(net, zone, &tuple);
        if (!h)
                return -ENOENT;
 
@@ -818,9 +852,21 @@ ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
                        return -ENOENT;
                }
        }
-       if (del_timer(&ct->timeout))
-               ct->timeout.function((unsigned long)ct);
 
+       if (nf_conntrack_event_report(IPCT_DESTROY, ct,
+                                     NETLINK_CB(skb).pid,
+                                     nlmsg_report(nlh)) < 0) {
+               nf_ct_delete_from_lists(ct);
+               /* we failed to report the event, try later */
+               nf_ct_insert_dying_list(ct);
+               nf_ct_put(ct);
+               return 0;
+       }
+
+       /* death_by_timeout would report the event again */
+       set_bit(IPS_DYING_BIT, &ct->status);
+
+       nf_ct_kill(ct);
        nf_ct_put(ct);
 
        return 0;
@@ -828,24 +874,26 @@ ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 
 static int
 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
-                       struct nlmsghdr *nlh, struct nlattr *cda[])
+                       const struct nlmsghdr *nlh,
+                       const struct nlattr * const cda[])
 {
+       struct net *net = sock_net(ctnl);
        struct nf_conntrack_tuple_hash *h;
        struct nf_conntrack_tuple tuple;
        struct nf_conn *ct;
        struct sk_buff *skb2 = NULL;
-       struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
+       struct nfgenmsg *nfmsg = nlmsg_data(nlh);
        u_int8_t u3 = nfmsg->nfgen_family;
-       int err = 0;
+       u16 zone;
+       int err;
 
-       if (nlh->nlmsg_flags & NLM_F_DUMP) {
-#ifndef CONFIG_NF_CT_ACCT
-               if (NFNL_MSG_TYPE(nlh->nlmsg_type) == IPCTNL_MSG_CT_GET_CTRZERO)
-                       return -ENOTSUPP;
-#endif
+       if (nlh->nlmsg_flags & NLM_F_DUMP)
                return netlink_dump_start(ctnl, skb, nlh, ctnetlink_dump_table,
                                          ctnetlink_done);
-       }
+
+       err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone);
+       if (err < 0)
+               return err;
 
        if (cda[CTA_TUPLE_ORIG])
                err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, u3);
@@ -857,21 +905,23 @@ ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
        if (err < 0)
                return err;
 
-       h = nf_conntrack_find_get(&tuple);
+       h = nf_conntrack_find_get(net, zone, &tuple);
        if (!h)
                return -ENOENT;
 
        ct = nf_ct_tuplehash_to_ctrack(h);
 
        err = -ENOMEM;
-       skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
-       if (!skb2) {
+       skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
+       if (skb2 == NULL) {
                nf_ct_put(ct);
                return -ENOMEM;
        }
 
+       rcu_read_lock();
        err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
-                                 IPCTNL_MSG_CT_NEW, 1, ct);
+                                 IPCTNL_MSG_CT_NEW, ct);
+       rcu_read_unlock();
        nf_ct_put(ct);
        if (err <= 0)
                goto free;
@@ -888,8 +938,41 @@ out:
        return err;
 }
 
-static inline int
-ctnetlink_change_status(struct nf_conn *ct, struct nlattr *cda[])
+#ifdef CONFIG_NF_NAT_NEEDED
+static int
+ctnetlink_parse_nat_setup(struct nf_conn *ct,
+                         enum nf_nat_manip_type manip,
+                         const struct nlattr *attr)
+{
+       typeof(nfnetlink_parse_nat_setup_hook) parse_nat_setup;
+
+       parse_nat_setup = rcu_dereference(nfnetlink_parse_nat_setup_hook);
+       if (!parse_nat_setup) {
+#ifdef CONFIG_MODULES
+               rcu_read_unlock();
+               spin_unlock_bh(&nf_conntrack_lock);
+               nfnl_unlock();
+               if (request_module("nf-nat-ipv4") < 0) {
+                       nfnl_lock();
+                       spin_lock_bh(&nf_conntrack_lock);
+                       rcu_read_lock();
+                       return -EOPNOTSUPP;
+               }
+               nfnl_lock();
+               spin_lock_bh(&nf_conntrack_lock);
+               rcu_read_lock();
+               if (nfnetlink_parse_nat_setup_hook)
+                       return -EAGAIN;
+#endif
+               return -EOPNOTSUPP;
+       }
+
+       return parse_nat_setup(ct, manip, attr);
+}
+#endif
+
+static int
+ctnetlink_change_status(struct nf_conn *ct, const struct nlattr * const cda[])
 {
        unsigned long d;
        unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS]));
@@ -897,41 +980,15 @@ ctnetlink_change_status(struct nf_conn *ct, struct nlattr *cda[])
 
        if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
                /* unchangeable */
-               return -EINVAL;
+               return -EBUSY;
 
        if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
                /* SEEN_REPLY bit can only be set */
-               return -EINVAL;
-
+               return -EBUSY;
 
        if (d & IPS_ASSURED && !(status & IPS_ASSURED))
                /* ASSURED bit can only be set */
-               return -EINVAL;
-
-       if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
-#ifndef CONFIG_NF_NAT_NEEDED
-               return -EINVAL;
-#else
-               struct nf_nat_range range;
-
-               if (cda[CTA_NAT_DST]) {
-                       if (nfnetlink_parse_nat(cda[CTA_NAT_DST], ct,
-                                               &range) < 0)
-                               return -EINVAL;
-                       if (nf_nat_initialized(ct, IP_NAT_MANIP_DST))
-                               return -EEXIST;
-                       nf_nat_setup_info(ct, &range, IP_NAT_MANIP_DST);
-               }
-               if (cda[CTA_NAT_SRC]) {
-                       if (nfnetlink_parse_nat(cda[CTA_NAT_SRC], ct,
-                                               &range) < 0)
-                               return -EINVAL;
-                       if (nf_nat_initialized(ct, IP_NAT_MANIP_SRC))
-                               return -EEXIST;
-                       nf_nat_setup_info(ct, &range, IP_NAT_MANIP_SRC);
-               }
-#endif
-       }
+               return -EBUSY;
 
        /* Be careful here, modifying NAT bits can screw up things,
         * so don't let users modify them directly if they don't pass
@@ -940,18 +997,43 @@ ctnetlink_change_status(struct nf_conn *ct, struct nlattr *cda[])
        return 0;
 }
 
+static int
+ctnetlink_change_nat(struct nf_conn *ct, const struct nlattr * const cda[])
+{
+#ifdef CONFIG_NF_NAT_NEEDED
+       int ret;
+
+       if (cda[CTA_NAT_DST]) {
+               ret = ctnetlink_parse_nat_setup(ct,
+                                               IP_NAT_MANIP_DST,
+                                               cda[CTA_NAT_DST]);
+               if (ret < 0)
+                       return ret;
+       }
+       if (cda[CTA_NAT_SRC]) {
+               ret = ctnetlink_parse_nat_setup(ct,
+                                               IP_NAT_MANIP_SRC,
+                                               cda[CTA_NAT_SRC]);
+               if (ret < 0)
+                       return ret;
+       }
+       return 0;
+#else
+       return -EOPNOTSUPP;
+#endif
+}
 
 static inline int
-ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
+ctnetlink_change_helper(struct nf_conn *ct, const struct nlattr * const cda[])
 {
        struct nf_conntrack_helper *helper;
        struct nf_conn_help *help = nfct_help(ct);
-       char *helpname;
+       char *helpname = NULL;
        int err;
 
        /* don't change helper of sibling connections */
        if (ct->master)
-               return -EINVAL;
+               return -EBUSY;
 
        err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
        if (err < 0)
@@ -967,9 +1049,25 @@ ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
                return 0;
        }
 
-       helper = __nf_conntrack_helper_find_byname(helpname);
-       if (helper == NULL)
-               return -EINVAL;
+       helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct),
+                                           nf_ct_protonum(ct));
+       if (helper == NULL) {
+#ifdef CONFIG_MODULES
+               spin_unlock_bh(&nf_conntrack_lock);
+
+               if (request_module("nfct-helper-%s", helpname) < 0) {
+                       spin_lock_bh(&nf_conntrack_lock);
+                       return -EOPNOTSUPP;
+               }
+
+               spin_lock_bh(&nf_conntrack_lock);
+               helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct),
+                                                   nf_ct_protonum(ct));
+               if (helper)
+                       return -EAGAIN;
+#endif
+               return -EOPNOTSUPP;
+       }
 
        if (help) {
                if (help->helper == helper)
@@ -979,9 +1077,8 @@ ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
                /* need to zero data of old helper */
                memset(&help->help, 0, sizeof(help->help));
        } else {
-               help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
-               if (help == NULL)
-                       return -ENOMEM;
+               /* we cannot set a helper for an existing conntrack */
+               return -EOPNOTSUPP;
        }
 
        rcu_assign_pointer(help->helper, helper);
@@ -990,7 +1087,7 @@ ctnetlink_change_helper(struct nf_conn *ct, struct nlattr *cda[])
 }
 
 static inline int
-ctnetlink_change_timeout(struct nf_conn *ct, struct nlattr *cda[])
+ctnetlink_change_timeout(struct nf_conn *ct, const struct nlattr * const cda[])
 {
        u_int32_t timeout = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
 
@@ -1003,33 +1100,44 @@ ctnetlink_change_timeout(struct nf_conn *ct, struct nlattr *cda[])
        return 0;
 }
 
+static const struct nla_policy protoinfo_policy[CTA_PROTOINFO_MAX+1] = {
+       [CTA_PROTOINFO_TCP]     = { .type = NLA_NESTED },
+       [CTA_PROTOINFO_DCCP]    = { .type = NLA_NESTED },
+       [CTA_PROTOINFO_SCTP]    = { .type = NLA_NESTED },
+};
+
 static inline int
-ctnetlink_change_protoinfo(struct nf_conn *ct, struct nlattr *cda[])
+ctnetlink_change_protoinfo(struct nf_conn *ct, const struct nlattr * const cda[])
 {
-       struct nlattr *tb[CTA_PROTOINFO_MAX+1], *attr = cda[CTA_PROTOINFO];
+       const struct nlattr *attr = cda[CTA_PROTOINFO];
+       struct nlattr *tb[CTA_PROTOINFO_MAX+1];
        struct nf_conntrack_l4proto *l4proto;
-       u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
-       u_int16_t l3num = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
        int err = 0;
 
-       nla_parse_nested(tb, CTA_PROTOINFO_MAX, attr, NULL);
-
-       l4proto = nf_ct_l4proto_find_get(l3num, npt);
+       nla_parse_nested(tb, CTA_PROTOINFO_MAX, attr, protoinfo_policy);
 
+       rcu_read_lock();
+       l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct));
        if (l4proto->from_nlattr)
                err = l4proto->from_nlattr(tb, ct);
-       nf_ct_l4proto_put(l4proto);
+       rcu_read_unlock();
 
        return err;
 }
 
 #ifdef CONFIG_NF_NAT_NEEDED
+static const struct nla_policy nat_seq_policy[CTA_NAT_SEQ_MAX+1] = {
+       [CTA_NAT_SEQ_CORRECTION_POS]    = { .type = NLA_U32 },
+       [CTA_NAT_SEQ_OFFSET_BEFORE]     = { .type = NLA_U32 },
+       [CTA_NAT_SEQ_OFFSET_AFTER]      = { .type = NLA_U32 },
+};
+
 static inline int
-change_nat_seq_adj(struct nf_nat_seq *natseq, struct nlattr *attr)
+change_nat_seq_adj(struct nf_nat_seq *natseq, const struct nlattr * const attr)
 {
        struct nlattr *cda[CTA_NAT_SEQ_MAX+1];
 
-       nla_parse_nested(cda, CTA_NAT_SEQ_MAX, attr, NULL);
+       nla_parse_nested(cda, CTA_NAT_SEQ_MAX, attr, nat_seq_policy);
 
        if (!cda[CTA_NAT_SEQ_CORRECTION_POS])
                return -EINVAL;
@@ -1053,7 +1161,8 @@ change_nat_seq_adj(struct nf_nat_seq *natseq, struct nlattr *attr)
 }
 
 static int
-ctnetlink_change_nat_seq_adj(struct nf_conn *ct, struct nlattr *cda[])
+ctnetlink_change_nat_seq_adj(struct nf_conn *ct,
+                            const struct nlattr * const cda[])
 {
        int ret = 0;
        struct nf_conn_nat *nat = nfct_nat(ct);
@@ -1084,10 +1193,15 @@ ctnetlink_change_nat_seq_adj(struct nf_conn *ct, struct nlattr *cda[])
 #endif
 
 static int
-ctnetlink_change_conntrack(struct nf_conn *ct, struct nlattr *cda[])
+ctnetlink_change_conntrack(struct nf_conn *ct,
+                          const struct nlattr * const cda[])
 {
        int err;
 
+       /* only allow NAT changes and master assignation for new conntracks */
+       if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST] || cda[CTA_TUPLE_MASTER])
+               return -EOPNOTSUPP;
+
        if (cda[CTA_HELP]) {
                err = ctnetlink_change_helper(ct, cda);
                if (err < 0)
@@ -1128,38 +1242,105 @@ ctnetlink_change_conntrack(struct nf_conn *ct, struct nlattr *cda[])
        return 0;
 }
 
-static int
-ctnetlink_create_conntrack(struct nlattr *cda[],
+static struct nf_conn *
+ctnetlink_create_conntrack(struct net *net, u16 zone,
+                          const struct nlattr * const cda[],
                           struct nf_conntrack_tuple *otuple,
                           struct nf_conntrack_tuple *rtuple,
-                          struct nf_conn *master_ct)
+                          u8 u3)
 {
        struct nf_conn *ct;
        int err = -EINVAL;
-       struct nf_conn_help *help;
        struct nf_conntrack_helper *helper;
 
-       ct = nf_conntrack_alloc(otuple, rtuple);
-       if (ct == NULL || IS_ERR(ct))
-               return -ENOMEM;
+       ct = nf_conntrack_alloc(net, zone, otuple, rtuple, GFP_ATOMIC);
+       if (IS_ERR(ct))
+               return ERR_PTR(-ENOMEM);
 
        if (!cda[CTA_TIMEOUT])
-               goto err;
+               goto err1;
        ct->timeout.expires = ntohl(nla_get_be32(cda[CTA_TIMEOUT]));
 
        ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
+
+       rcu_read_lock();
+       if (cda[CTA_HELP]) {
+               char *helpname = NULL;
+               err = ctnetlink_parse_help(cda[CTA_HELP], &helpname);
+               if (err < 0)
+                       goto err2;
+
+               helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct),
+                                                   nf_ct_protonum(ct));
+               if (helper == NULL) {
+                       rcu_read_unlock();
+#ifdef CONFIG_MODULES
+                       if (request_module("nfct-helper-%s", helpname) < 0) {
+                               err = -EOPNOTSUPP;
+                               goto err1;
+                       }
+
+                       rcu_read_lock();
+                       helper = __nf_conntrack_helper_find(helpname,
+                                                           nf_ct_l3num(ct),
+                                                           nf_ct_protonum(ct));
+                       if (helper) {
+                               err = -EAGAIN;
+                               goto err2;
+                       }
+                       rcu_read_unlock();
+#endif
+                       err = -EOPNOTSUPP;
+                       goto err1;
+               } else {
+                       struct nf_conn_help *help;
+
+                       help = nf_ct_helper_ext_add(ct, GFP_ATOMIC);
+                       if (help == NULL) {
+                               err = -ENOMEM;
+                               goto err2;
+                       }
+
+                       /* not in hash table yet so not strictly necessary */
+                       rcu_assign_pointer(help->helper, helper);
+               }
+       } else {
+               /* try an implicit helper assignation */
+               err = __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC);
+               if (err < 0)
+                       goto err2;
+       }
+
+       if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
+               err = ctnetlink_change_nat(ct, cda);
+               if (err < 0)
+                       goto err2;
+       }
+
+       nf_ct_acct_ext_add(ct, GFP_ATOMIC);
+       nf_ct_ecache_ext_add(ct, 0, 0, GFP_ATOMIC);
+       /* we must add conntrack extensions before confirmation. */
        ct->status |= IPS_CONFIRMED;
 
        if (cda[CTA_STATUS]) {
                err = ctnetlink_change_status(ct, cda);
                if (err < 0)
-                       goto err;
+                       goto err2;
        }
 
+#ifdef CONFIG_NF_NAT_NEEDED
+       if (cda[CTA_NAT_SEQ_ADJ_ORIG] || cda[CTA_NAT_SEQ_ADJ_REPLY]) {
+               err = ctnetlink_change_nat_seq_adj(ct, cda);
+               if (err < 0)
+                       goto err2;
+       }
+#endif
+
        if (cda[CTA_PROTOINFO]) {
                err = ctnetlink_change_protoinfo(ct, cda);
                if (err < 0)
-                       goto err;
+                       goto err2;
        }
 
 #if defined(CONFIG_NF_CONNTRACK_MARK)
@@ -1167,46 +1348,55 @@ ctnetlink_create_conntrack(struct nlattr *cda[],
                ct->mark = ntohl(nla_get_be32(cda[CTA_MARK]));
 #endif
 
-       helper = nf_ct_helper_find_get(rtuple);
-       if (helper) {
-               help = nf_ct_helper_ext_add(ct, GFP_KERNEL);
-               if (help == NULL) {
-                       nf_ct_helper_put(helper);
-                       err = -ENOMEM;
-                       goto err;
-               }
-               /* not in hash table yet so not strictly necessary */
-               rcu_assign_pointer(help->helper, helper);
-       }
-
        /* setup master conntrack: this is a confirmed expectation */
-       if (master_ct) {
+       if (cda[CTA_TUPLE_MASTER]) {
+               struct nf_conntrack_tuple master;
+               struct nf_conntrack_tuple_hash *master_h;
+               struct nf_conn *master_ct;
+
+               err = ctnetlink_parse_tuple(cda, &master, CTA_TUPLE_MASTER, u3);
+               if (err < 0)
+                       goto err2;
+
+               master_h = nf_conntrack_find_get(net, zone, &master);
+               if (master_h == NULL) {
+                       err = -ENOENT;
+                       goto err2;
+               }
+               master_ct = nf_ct_tuplehash_to_ctrack(master_h);
                __set_bit(IPS_EXPECTED_BIT, &ct->status);
                ct->master = master_ct;
        }
 
        add_timer(&ct->timeout);
        nf_conntrack_hash_insert(ct);
+       rcu_read_unlock();
 
-       if (helper)
-               nf_ct_helper_put(helper);
-
-       return 0;
+       return ct;
 
-err:
+err2:
+       rcu_read_unlock();
+err1:
        nf_conntrack_free(ct);
-       return err;
+       return ERR_PTR(err);
 }
 
 static int
 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
-                       struct nlmsghdr *nlh, struct nlattr *cda[])
+                       const struct nlmsghdr *nlh,
+                       const struct nlattr * const cda[])
 {
+       struct net *net = sock_net(ctnl);
        struct nf_conntrack_tuple otuple, rtuple;
        struct nf_conntrack_tuple_hash *h = NULL;
-       struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
+       struct nfgenmsg *nfmsg = nlmsg_data(nlh);
        u_int8_t u3 = nfmsg->nfgen_family;
-       int err = 0;
+       u16 zone;
+       int err;
+
+       err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone);
+       if (err < 0)
+               return err;
 
        if (cda[CTA_TUPLE_ORIG]) {
                err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG, u3);
@@ -1220,43 +1410,43 @@ ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
                        return err;
        }
 
-       write_lock_bh(&nf_conntrack_lock);
+       spin_lock_bh(&nf_conntrack_lock);
        if (cda[CTA_TUPLE_ORIG])
-               h = __nf_conntrack_find(&otuple, NULL);
+               h = __nf_conntrack_find(net, zone, &otuple);
        else if (cda[CTA_TUPLE_REPLY])
-               h = __nf_conntrack_find(&rtuple, NULL);
+               h = __nf_conntrack_find(net, zone, &rtuple);
 
        if (h == NULL) {
-               struct nf_conntrack_tuple master;
-               struct nf_conntrack_tuple_hash *master_h = NULL;
-               struct nf_conn *master_ct = NULL;
-
-               if (cda[CTA_TUPLE_MASTER]) {
-                       err = ctnetlink_parse_tuple(cda,
-                                                   &master,
-                                                   CTA_TUPLE_MASTER,
-                                                   u3);
-                       if (err < 0)
-                               return err;
-
-                       master_h = __nf_conntrack_find(&master, NULL);
-                       if (master_h == NULL) {
-                               err = -ENOENT;
+               err = -ENOENT;
+               if (nlh->nlmsg_flags & NLM_F_CREATE) {
+                       struct nf_conn *ct;
+                       enum ip_conntrack_events events;
+
+                       ct = ctnetlink_create_conntrack(net, zone, cda, &otuple,
+                                                       &rtuple, u3);
+                       if (IS_ERR(ct)) {
+                               err = PTR_ERR(ct);
                                goto out_unlock;
                        }
-                       master_ct = nf_ct_tuplehash_to_ctrack(master_h);
-                       atomic_inc(&master_ct->ct_general.use);
-               }
-
-               write_unlock_bh(&nf_conntrack_lock);
-               err = -ENOENT;
-               if (nlh->nlmsg_flags & NLM_F_CREATE)
-                       err = ctnetlink_create_conntrack(cda,
-                                                        &otuple,
-                                                        &rtuple,
-                                                        master_ct);
-               if (err < 0 && master_ct)
-                       nf_ct_put(master_ct);
+                       err = 0;
+                       nf_conntrack_get(&ct->ct_general);
+                       spin_unlock_bh(&nf_conntrack_lock);
+                       if (test_bit(IPS_EXPECTED_BIT, &ct->status))
+                               events = IPCT_RELATED;
+                       else
+                               events = IPCT_NEW;
+
+                       nf_conntrack_eventmask_report((1 << IPCT_REPLY) |
+                                                     (1 << IPCT_ASSURED) |
+                                                     (1 << IPCT_HELPER) |
+                                                     (1 << IPCT_PROTOINFO) |
+                                                     (1 << IPCT_NATSEQADJ) |
+                                                     (1 << IPCT_MARK) | events,
+                                                     ct, NETLINK_CB(skb).pid,
+                                                     nlmsg_report(nlh));
+                       nf_ct_put(ct);
+               } else
+                       spin_unlock_bh(&nf_conntrack_lock);
 
                return err;
        }
@@ -1266,22 +1456,29 @@ ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
         * so there's no need to increase the refcount */
        err = -EEXIST;
        if (!(nlh->nlmsg_flags & NLM_F_EXCL)) {
-               /* we only allow nat config for new conntracks */
-               if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST]) {
-                       err = -EINVAL;
-                       goto out_unlock;
-               }
-               /* can't link an existing conntrack to a master */
-               if (cda[CTA_TUPLE_MASTER]) {
-                       err = -EINVAL;
-                       goto out_unlock;
-               }
-               err = ctnetlink_change_conntrack(nf_ct_tuplehash_to_ctrack(h),
-                                                cda);
+               struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
+
+               err = ctnetlink_change_conntrack(ct, cda);
+               if (err == 0) {
+                       nf_conntrack_get(&ct->ct_general);
+                       spin_unlock_bh(&nf_conntrack_lock);
+                       nf_conntrack_eventmask_report((1 << IPCT_REPLY) |
+                                                     (1 << IPCT_ASSURED) |
+                                                     (1 << IPCT_HELPER) |
+                                                     (1 << IPCT_PROTOINFO) |
+                                                     (1 << IPCT_NATSEQADJ) |
+                                                     (1 << IPCT_MARK),
+                                                     ct, NETLINK_CB(skb).pid,
+                                                     nlmsg_report(nlh));
+                       nf_ct_put(ct);
+               } else
+                       spin_unlock_bh(&nf_conntrack_lock);
+
+               return err;
        }
 
 out_unlock:
-       write_unlock_bh(&nf_conntrack_lock);
+       spin_unlock_bh(&nf_conntrack_lock);
        return err;
 }
 
@@ -1321,23 +1518,22 @@ ctnetlink_exp_dump_mask(struct sk_buff *skb,
        struct nlattr *nest_parms;
 
        memset(&m, 0xFF, sizeof(m));
-       m.src.u.all = mask->src.u.all;
        memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3));
+       m.src.u.all = mask->src.u.all;
+       m.dst.protonum = tuple->dst.protonum;
 
        nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK | NLA_F_NESTED);
        if (!nest_parms)
                goto nla_put_failure;
 
-       l3proto = nf_ct_l3proto_find_get(tuple->src.l3num);
+       l3proto = __nf_ct_l3proto_find(tuple->src.l3num);
        ret = ctnetlink_dump_tuples_ip(skb, &m, l3proto);
-       nf_ct_l3proto_put(l3proto);
 
        if (unlikely(ret < 0))
                goto nla_put_failure;
 
-       l4proto = nf_ct_l4proto_find_get(tuple->src.l3num, tuple->dst.protonum);
+       l4proto = __nf_ct_l4proto_find(tuple->src.l3num, tuple->dst.protonum);
        ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto);
-       nf_ct_l4proto_put(l4proto);
        if (unlikely(ret < 0))
                goto nla_put_failure;
 
@@ -1349,11 +1545,12 @@ nla_put_failure:
        return -1;
 }
 
-static inline int
+static int
 ctnetlink_exp_dump_expect(struct sk_buff *skb,
                          const struct nf_conntrack_expect *exp)
 {
        struct nf_conn *master = exp->master;
+       struct nf_conntrack_helper *helper;
        long timeout = (exp->timeout.expires - jiffies) / HZ;
 
        if (timeout < 0)
@@ -1370,6 +1567,9 @@ ctnetlink_exp_dump_expect(struct sk_buff *skb,
 
        NLA_PUT_BE32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout));
        NLA_PUT_BE32(skb, CTA_EXPECT_ID, htonl((unsigned long)exp));
+       helper = rcu_dereference(nfct_help(master)->helper);
+       if (helper)
+               NLA_PUT_STRING(skb, CTA_EXPECT_HELP_NAME, helper->name);
 
        return 0;
 
@@ -1379,19 +1579,18 @@ nla_put_failure:
 
 static int
 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
-                   int event,
-                   int nowait,
-                   const struct nf_conntrack_expect *exp)
+                       int event, const struct nf_conntrack_expect *exp)
 {
        struct nlmsghdr *nlh;
        struct nfgenmsg *nfmsg;
-       unsigned char *b = skb_tail_pointer(skb);
+       unsigned int flags = pid ? NLM_F_MULTI : 0;
 
        event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
-       nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
-       nfmsg  = NLMSG_DATA(nlh);
+       nlh = nlmsg_put(skb, pid, seq, event, sizeof(*nfmsg), flags);
+       if (nlh == NULL)
+               goto nlmsg_failure;
 
-       nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
+       nfmsg = nlmsg_data(nlh);
        nfmsg->nfgen_family = exp->tuple.src.l3num;
        nfmsg->version      = NFNETLINK_V0;
        nfmsg->res_id       = 0;
@@ -1399,62 +1598,69 @@ ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
        if (ctnetlink_exp_dump_expect(skb, exp) < 0)
                goto nla_put_failure;
 
-       nlh->nlmsg_len = skb_tail_pointer(skb) - b;
+       nlmsg_end(skb, nlh);
        return skb->len;
 
 nlmsg_failure:
 nla_put_failure:
-       nlmsg_trim(skb, b);
+       nlmsg_cancel(skb, nlh);
        return -1;
 }
 
 #ifdef CONFIG_NF_CONNTRACK_EVENTS
-static int ctnetlink_expect_event(struct notifier_block *this,
-                                 unsigned long events, void *ptr)
+static int
+ctnetlink_expect_event(unsigned int events, struct nf_exp_event *item)
 {
+       struct nf_conntrack_expect *exp = item->exp;
+       struct net *net = nf_ct_exp_net(exp);
        struct nlmsghdr *nlh;
        struct nfgenmsg *nfmsg;
-       struct nf_conntrack_expect *exp = (struct nf_conntrack_expect *)ptr;
        struct sk_buff *skb;
        unsigned int type;
-       sk_buff_data_t b;
        int flags = 0;
 
-       if (events & IPEXP_NEW) {
+       if (events & (1 << IPEXP_NEW)) {
                type = IPCTNL_MSG_EXP_NEW;
                flags = NLM_F_CREATE|NLM_F_EXCL;
        } else
-               return NOTIFY_DONE;
-
-       if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
-               return NOTIFY_DONE;
+               return 0;
 
-       skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
-       if (!skb)
-               return NOTIFY_DONE;
+       if (!item->report &&
+           !nfnetlink_has_listeners(net, NFNLGRP_CONNTRACK_EXP_NEW))
+               return 0;
 
-       b = skb->tail;
+       skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
+       if (skb == NULL)
+               goto errout;
 
        type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
-       nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
-       nfmsg = NLMSG_DATA(nlh);
+       nlh = nlmsg_put(skb, item->pid, 0, type, sizeof(*nfmsg), flags);
+       if (nlh == NULL)
+               goto nlmsg_failure;
 
-       nlh->nlmsg_flags    = flags;
+       nfmsg = nlmsg_data(nlh);
        nfmsg->nfgen_family = exp->tuple.src.l3num;
        nfmsg->version      = NFNETLINK_V0;
        nfmsg->res_id       = 0;
 
+       rcu_read_lock();
        if (ctnetlink_exp_dump_expect(skb, exp) < 0)
                goto nla_put_failure;
+       rcu_read_unlock();
 
-       nlh->nlmsg_len = skb->tail - b;
-       nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
-       return NOTIFY_DONE;
+       nlmsg_end(skb, nlh);
+       nfnetlink_send(skb, net, item->pid, NFNLGRP_CONNTRACK_EXP_NEW,
+                      item->report, GFP_ATOMIC);
+       return 0;
 
-nlmsg_failure:
 nla_put_failure:
+       rcu_read_unlock();
+       nlmsg_cancel(skb, nlh);
+nlmsg_failure:
        kfree_skb(skb);
-       return NOTIFY_DONE;
+errout:
+       nfnetlink_set_err(net, 0, 0, -ENOBUFS);
+       return 0;
 }
 #endif
 static int ctnetlink_exp_done(struct netlink_callback *cb)
@@ -1467,16 +1673,17 @@ static int ctnetlink_exp_done(struct netlink_callback *cb)
 static int
 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 {
+       struct net *net = sock_net(skb->sk);
        struct nf_conntrack_expect *exp, *last;
-       struct nfgenmsg *nfmsg = NLMSG_DATA(cb->nlh);
+       struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh);
        struct hlist_node *n;
        u_int8_t l3proto = nfmsg->nfgen_family;
 
-       read_lock_bh(&nf_conntrack_lock);
+       rcu_read_lock();
        last = (struct nf_conntrack_expect *)cb->args[1];
        for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) {
 restart:
-               hlist_for_each_entry(exp, n, &nf_ct_expect_hash[cb->args[0]],
+               hlist_for_each_entry(exp, n, &net->ct.expect_hash[cb->args[0]],
                                     hnode) {
                        if (l3proto && exp->tuple.src.l3num != l3proto)
                                continue;
@@ -1485,11 +1692,13 @@ restart:
                                        continue;
                                cb->args[1] = 0;
                        }
-                       if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
+                       if (ctnetlink_exp_fill_info(skb,
+                                                   NETLINK_CB(cb->skb).pid,
                                                    cb->nlh->nlmsg_seq,
                                                    IPCTNL_MSG_EXP_NEW,
-                                                   1, exp) < 0) {
-                               atomic_inc(&exp->use);
+                                                   exp) < 0) {
+                               if (!atomic_inc_not_zero(&exp->use))
+                                       continue;
                                cb->args[1] = (unsigned long)exp;
                                goto out;
                        }
@@ -1500,7 +1709,7 @@ restart:
                }
        }
 out:
-       read_unlock_bh(&nf_conntrack_lock);
+       rcu_read_unlock();
        if (last)
                nf_ct_expect_put(last);
 
@@ -1508,20 +1717,27 @@ out:
 }
 
 static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = {
+       [CTA_EXPECT_MASTER]     = { .type = NLA_NESTED },
+       [CTA_EXPECT_TUPLE]      = { .type = NLA_NESTED },
+       [CTA_EXPECT_MASK]       = { .type = NLA_NESTED },
        [CTA_EXPECT_TIMEOUT]    = { .type = NLA_U32 },
        [CTA_EXPECT_ID]         = { .type = NLA_U32 },
+       [CTA_EXPECT_HELP_NAME]  = { .type = NLA_NUL_STRING },
 };
 
 static int
 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
-                    struct nlmsghdr *nlh, struct nlattr *cda[])
+                    const struct nlmsghdr *nlh,
+                    const struct nlattr * const cda[])
 {
+       struct net *net = sock_net(ctnl);
        struct nf_conntrack_tuple tuple;
        struct nf_conntrack_expect *exp;
        struct sk_buff *skb2;
-       struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
+       struct nfgenmsg *nfmsg = nlmsg_data(nlh);
        u_int8_t u3 = nfmsg->nfgen_family;
-       int err = 0;
+       u16 zone;
+       int err;
 
        if (nlh->nlmsg_flags & NLM_F_DUMP) {
                return netlink_dump_start(ctnl, skb, nlh,
@@ -1529,6 +1745,10 @@ ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
                                          ctnetlink_exp_done);
        }
 
+       err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
+       if (err < 0)
+               return err;
+
        if (cda[CTA_EXPECT_MASTER])
                err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER, u3);
        else
@@ -1537,7 +1757,7 @@ ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
        if (err < 0)
                return err;
 
-       exp = nf_ct_expect_find_get(&tuple);
+       exp = nf_ct_expect_find_get(net, zone, &tuple);
        if (!exp)
                return -ENOENT;
 
@@ -1550,13 +1770,14 @@ ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
        }
 
        err = -ENOMEM;
-       skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
-       if (!skb2)
+       skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
+       if (skb2 == NULL)
                goto out;
 
+       rcu_read_lock();
        err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
-                                     nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
-                                     1, exp);
+                                     nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW, exp);
+       rcu_read_unlock();
        if (err <= 0)
                goto free;
 
@@ -1573,25 +1794,31 @@ out:
 
 static int
 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
-                    struct nlmsghdr *nlh, struct nlattr *cda[])
+                    const struct nlmsghdr *nlh,
+                    const struct nlattr * const cda[])
 {
+       struct net *net = sock_net(ctnl);
        struct nf_conntrack_expect *exp;
        struct nf_conntrack_tuple tuple;
-       struct nf_conntrack_helper *h;
-       struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
+       struct nfgenmsg *nfmsg = nlmsg_data(nlh);
        struct hlist_node *n, *next;
        u_int8_t u3 = nfmsg->nfgen_family;
        unsigned int i;
+       u16 zone;
        int err;
 
        if (cda[CTA_EXPECT_TUPLE]) {
                /* delete a single expect by tuple */
+               err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
+               if (err < 0)
+                       return err;
+
                err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
                if (err < 0)
                        return err;
 
                /* bump usage count to 2 */
-               exp = nf_ct_expect_find_get(&tuple);
+               exp = nf_ct_expect_find_get(net, zone, &tuple);
                if (!exp)
                        return -ENOENT;
 
@@ -1613,31 +1840,26 @@ ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
                struct nf_conn_help *m_help;
 
                /* delete all expectations for this helper */
-               write_lock_bh(&nf_conntrack_lock);
-               h = __nf_conntrack_helper_find_byname(name);
-               if (!h) {
-                       write_unlock_bh(&nf_conntrack_lock);
-                       return -EINVAL;
-               }
+               spin_lock_bh(&nf_conntrack_lock);
                for (i = 0; i < nf_ct_expect_hsize; i++) {
                        hlist_for_each_entry_safe(exp, n, next,
-                                                 &nf_ct_expect_hash[i],
+                                                 &net->ct.expect_hash[i],
                                                  hnode) {
                                m_help = nfct_help(exp->master);
-                               if (m_help->helper == h
-                                   && del_timer(&exp->timeout)) {
+                               if (!strcmp(m_help->helper->name, name) &&
+                                   del_timer(&exp->timeout)) {
                                        nf_ct_unlink_expect(exp);
                                        nf_ct_expect_put(exp);
                                }
                        }
                }
-               write_unlock_bh(&nf_conntrack_lock);
+               spin_unlock_bh(&nf_conntrack_lock);
        } else {
                /* This basically means we have to flush everything*/
-               write_lock_bh(&nf_conntrack_lock);
+               spin_lock_bh(&nf_conntrack_lock);
                for (i = 0; i < nf_ct_expect_hsize; i++) {
                        hlist_for_each_entry_safe(exp, n, next,
-                                                 &nf_ct_expect_hash[i],
+                                                 &net->ct.expect_hash[i],
                                                  hnode) {
                                if (del_timer(&exp->timeout)) {
                                        nf_ct_unlink_expect(exp);
@@ -1645,19 +1867,23 @@ ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
                                }
                        }
                }
-               write_unlock_bh(&nf_conntrack_lock);
+               spin_unlock_bh(&nf_conntrack_lock);
        }
 
        return 0;
 }
 static int
-ctnetlink_change_expect(struct nf_conntrack_expect *x, struct nlattr *cda[])
+ctnetlink_change_expect(struct nf_conntrack_expect *x,
+                       const struct nlattr * const cda[])
 {
        return -EOPNOTSUPP;
 }
 
 static int
-ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3)
+ctnetlink_create_expect(struct net *net, u16 zone,
+                       const struct nlattr * const cda[],
+                       u_int8_t u3,
+                       u32 pid, int report)
 {
        struct nf_conntrack_tuple tuple, mask, master_tuple;
        struct nf_conntrack_tuple_hash *h = NULL;
@@ -1678,7 +1904,7 @@ ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3)
                return err;
 
        /* Look for master conntrack of this expectation */
-       h = nf_conntrack_find_get(&master_tuple);
+       h = nf_conntrack_find_get(net, zone, &master_tuple);
        if (!h)
                return -ENOENT;
        ct = nf_ct_tuplehash_to_ctrack(h);
@@ -1686,7 +1912,7 @@ ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3)
 
        if (!help || !help->helper) {
                /* such conntrack hasn't got any helper, abort */
-               err = -EINVAL;
+               err = -EOPNOTSUPP;
                goto out;
        }
 
@@ -1696,6 +1922,7 @@ ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3)
                goto out;
        }
 
+       exp->class = 0;
        exp->expectfn = NULL;
        exp->flags = 0;
        exp->master = ct;
@@ -1704,7 +1931,7 @@ ctnetlink_create_expect(struct nlattr *cda[], u_int8_t u3)
        memcpy(&exp->mask.src.u3, &mask.src.u3, sizeof(exp->mask.src.u3));
        exp->mask.src.u.all = mask.src.u.all;
 
-       err = nf_ct_expect_related(exp);
+       err = nf_ct_expect_related_report(exp, pid, report);
        nf_ct_expect_put(exp);
 
 out:
@@ -1714,49 +1941,60 @@ out:
 
 static int
 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
-                    struct nlmsghdr *nlh, struct nlattr *cda[])
+                    const struct nlmsghdr *nlh,
+                    const struct nlattr * const cda[])
 {
+       struct net *net = sock_net(ctnl);
        struct nf_conntrack_tuple tuple;
        struct nf_conntrack_expect *exp;
-       struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
+       struct nfgenmsg *nfmsg = nlmsg_data(nlh);
        u_int8_t u3 = nfmsg->nfgen_family;
-       int err = 0;
+       u16 zone;
+       int err;
 
        if (!cda[CTA_EXPECT_TUPLE]
            || !cda[CTA_EXPECT_MASK]
            || !cda[CTA_EXPECT_MASTER])
                return -EINVAL;
 
+       err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone);
+       if (err < 0)
+               return err;
+
        err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, u3);
        if (err < 0)
                return err;
 
-       write_lock_bh(&nf_conntrack_lock);
-       exp = __nf_ct_expect_find(&tuple);
+       spin_lock_bh(&nf_conntrack_lock);
+       exp = __nf_ct_expect_find(net, zone, &tuple);
 
        if (!exp) {
-               write_unlock_bh(&nf_conntrack_lock);
+               spin_unlock_bh(&nf_conntrack_lock);
                err = -ENOENT;
-               if (nlh->nlmsg_flags & NLM_F_CREATE)
-                       err = ctnetlink_create_expect(cda, u3);
+               if (nlh->nlmsg_flags & NLM_F_CREATE) {
+                       err = ctnetlink_create_expect(net, zone, cda,
+                                                     u3,
+                                                     NETLINK_CB(skb).pid,
+                                                     nlmsg_report(nlh));
+               }
                return err;
        }
 
        err = -EEXIST;
        if (!(nlh->nlmsg_flags & NLM_F_EXCL))
                err = ctnetlink_change_expect(exp, cda);
-       write_unlock_bh(&nf_conntrack_lock);
+       spin_unlock_bh(&nf_conntrack_lock);
 
        return err;
 }
 
 #ifdef CONFIG_NF_CONNTRACK_EVENTS
-static struct notifier_block ctnl_notifier = {
-       .notifier_call  = ctnetlink_conntrack_event,
+static struct nf_ct_event_notifier ctnl_notifier = {
+       .fcn = ctnetlink_conntrack_event,
 };
 
-static struct notifier_block ctnl_notifier_exp = {
-       .notifier_call  = ctnetlink_expect_event,
+static struct nf_exp_event_notifier ctnl_notifier_exp = {
+       .fcn = ctnetlink_expect_event,
 };
 #endif