Merge branch 'sii-m15w' into upstream
[pandora-kernel.git] / net / dccp / ipv4.c
index 656e13e..7f56f7e 100644 (file)
@@ -10,7 +10,6 @@
  *     2 of the License, or (at your option) any later version.
  */
 
-#include <linux/config.h>
 #include <linux/dccp.h>
 #include <linux/icmp.h>
 #include <linux/module.h>
 #include <linux/random.h>
 
 #include <net/icmp.h>
+#include <net/inet_common.h>
 #include <net/inet_hashtables.h>
+#include <net/inet_sock.h>
+#include <net/protocol.h>
 #include <net/sock.h>
+#include <net/timewait_sock.h>
 #include <net/tcp_states.h>
 #include <net/xfrm.h>
 
 #include "ackvec.h"
 #include "ccid.h"
 #include "dccp.h"
-
-struct inet_hashinfo __cacheline_aligned dccp_hashinfo = {
-       .lhash_lock     = RW_LOCK_UNLOCKED,
-       .lhash_users    = ATOMIC_INIT(0),
-       .lhash_wait = __WAIT_QUEUE_HEAD_INITIALIZER(dccp_hashinfo.lhash_wait),
-};
-
-EXPORT_SYMBOL_GPL(dccp_hashinfo);
-
-static int dccp_v4_get_port(struct sock *sk, const unsigned short snum)
-{
-       return inet_csk_get_port(&dccp_hashinfo, sk, snum);
-}
-
-static void dccp_v4_hash(struct sock *sk)
-{
-       inet_hash(&dccp_hashinfo, sk);
-}
-
-static void dccp_v4_unhash(struct sock *sk)
-{
-       inet_unhash(&dccp_hashinfo, sk);
-}
-
-/* called with local bh disabled */
-static int __dccp_v4_check_established(struct sock *sk, const __u16 lport,
-                                     struct inet_timewait_sock **twp)
-{
-       struct inet_sock *inet = inet_sk(sk);
-       const u32 daddr = inet->rcv_saddr;
-       const u32 saddr = inet->daddr;
-       const int dif = sk->sk_bound_dev_if;
-       INET_ADDR_COOKIE(acookie, saddr, daddr)
-       const __u32 ports = INET_COMBINED_PORTS(inet->dport, lport);
-       unsigned int hash = inet_ehashfn(daddr, lport, saddr, inet->dport);
-       struct inet_ehash_bucket *head = inet_ehash_bucket(&dccp_hashinfo, hash);
-       const struct sock *sk2;
-       const struct hlist_node *node;
-       struct inet_timewait_sock *tw;
-
-       prefetch(head->chain.first);
-       write_lock(&head->lock);
-
-       /* Check TIME-WAIT sockets first. */
-       sk_for_each(sk2, node, &(head + dccp_hashinfo.ehash_size)->chain) {
-               tw = inet_twsk(sk2);
-
-               if (INET_TW_MATCH(sk2, hash, acookie, saddr, daddr, ports, dif))
-                       goto not_unique;
-       }
-       tw = NULL;
-
-       /* And established part... */
-       sk_for_each(sk2, node, &head->chain) {
-               if (INET_MATCH(sk2, hash, acookie, saddr, daddr, ports, dif))
-                       goto not_unique;
-       }
-
-       /* Must record num and sport now. Otherwise we will see
-        * in hash table socket with a funny identity. */
-       inet->num = lport;
-       inet->sport = htons(lport);
-       sk->sk_hash = hash;
-       BUG_TRAP(sk_unhashed(sk));
-       __sk_add_node(sk, &head->chain);
-       sock_prot_inc_use(sk->sk_prot);
-       write_unlock(&head->lock);
-
-       if (twp != NULL) {
-               *twp = tw;
-               NET_INC_STATS_BH(LINUX_MIB_TIMEWAITRECYCLED);
-       } else if (tw != NULL) {
-               /* Silly. Should hash-dance instead... */
-               inet_twsk_deschedule(tw, &dccp_death_row);
-               NET_INC_STATS_BH(LINUX_MIB_TIMEWAITRECYCLED);
-
-               inet_twsk_put(tw);
-       }
-
-       return 0;
-
-not_unique:
-       write_unlock(&head->lock);
-       return -EADDRNOTAVAIL;
-}
+#include "feat.h"
 
 /*
- * Bind a port for a connect operation and hash it.
+ * This is the global socket data structure used for responding to
+ * the Out-of-the-blue (OOTB) packets. A control sock will be created
+ * for this socket at the initialization time.
  */
-static int dccp_v4_hash_connect(struct sock *sk)
-{
-       const unsigned short snum = inet_sk(sk)->num;
-       struct inet_bind_hashbucket *head;
-       struct inet_bind_bucket *tb;
-       int ret;
-
-       if (snum == 0) {
-               int low = sysctl_local_port_range[0];
-               int high = sysctl_local_port_range[1];
-               int remaining = (high - low) + 1;
-               int rover = net_random() % (high - low) + low;
-               struct hlist_node *node;
-               struct inet_timewait_sock *tw = NULL;
-
-               local_bh_disable();
-               do {
-                       head = &dccp_hashinfo.bhash[inet_bhashfn(rover,
-                                                   dccp_hashinfo.bhash_size)];
-                       spin_lock(&head->lock);
-
-                       /* Does not bother with rcv_saddr checks,
-                        * because the established check is already
-                        * unique enough.
-                        */
-                       inet_bind_bucket_for_each(tb, node, &head->chain) {
-                               if (tb->port == rover) {
-                                       BUG_TRAP(!hlist_empty(&tb->owners));
-                                       if (tb->fastreuse >= 0)
-                                               goto next_port;
-                                       if (!__dccp_v4_check_established(sk,
-                                                                        rover,
-                                                                        &tw))
-                                               goto ok;
-                                       goto next_port;
-                               }
-                       }
-
-                       tb = inet_bind_bucket_create(dccp_hashinfo.bind_bucket_cachep,
-                                                    head, rover);
-                       if (tb == NULL) {
-                               spin_unlock(&head->lock);
-                               break;
-                       }
-                       tb->fastreuse = -1;
-                       goto ok;
-
-               next_port:
-                       spin_unlock(&head->lock);
-                       if (++rover > high)
-                               rover = low;
-               } while (--remaining > 0);
-
-               local_bh_enable();
-
-               return -EADDRNOTAVAIL;
-
-ok:
-               /* All locks still held and bhs disabled */
-               inet_bind_hash(sk, tb, rover);
-               if (sk_unhashed(sk)) {
-                       inet_sk(sk)->sport = htons(rover);
-                       __inet_hash(&dccp_hashinfo, sk, 0);
-               }
-               spin_unlock(&head->lock);
-
-               if (tw != NULL) {
-                       inet_twsk_deschedule(tw, &dccp_death_row);
-                       inet_twsk_put(tw);
-               }
-
-               ret = 0;
-               goto out;
-       }
+static struct socket *dccp_v4_ctl_socket;
 
-       head = &dccp_hashinfo.bhash[inet_bhashfn(snum,
-                                                dccp_hashinfo.bhash_size)];
-       tb   = inet_csk(sk)->icsk_bind_hash;
-       spin_lock_bh(&head->lock);
-       if (sk_head(&tb->owners) == sk && sk->sk_bind_node.next == NULL) {
-               __inet_hash(&dccp_hashinfo, sk, 0);
-               spin_unlock_bh(&head->lock);
-               return 0;
-       } else {
-               spin_unlock(&head->lock);
-               /* No definite answer... Walk to established hash table */
-               ret = __dccp_v4_check_established(sk, snum, NULL);
-out:
-               local_bh_enable();
-               return ret;
-       }
+static int dccp_v4_get_port(struct sock *sk, const unsigned short snum)
+{
+       return inet_csk_get_port(&dccp_hashinfo, sk, snum,
+                                inet_csk_bind_conflict);
 }
 
-static int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr,
-                          int addr_len)
+int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 {
        struct inet_sock *inet = inet_sk(sk);
        struct dccp_sock *dp = dccp_sk(sk);
@@ -259,9 +94,9 @@ static int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr,
        inet->dport = usin->sin_port;
        inet->daddr = daddr;
 
-       dp->dccps_ext_header_len = 0;
+       inet_csk(sk)->icsk_ext_hdr_len = 0;
        if (inet->opt != NULL)
-               dp->dccps_ext_header_len = inet->opt->optlen;
+               inet_csk(sk)->icsk_ext_hdr_len = inet->opt->optlen;
        /*
         * Socket identity is still unknown (sport may be zero).
         * However we set state to DCCP_REQUESTING and not releasing socket
@@ -269,11 +104,12 @@ static int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr,
         * complete initialization after this.
         */
        dccp_set_state(sk, DCCP_REQUESTING);
-       err = dccp_v4_hash_connect(sk);
+       err = inet_hash_connect(&dccp_death_row, sk);
        if (err != 0)
                goto failure;
 
-       err = ip_route_newports(&rt, inet->sport, inet->dport, sk);
+       err = ip_route_newports(&rt, IPPROTO_DCCP, inet->sport, inet->dport,
+                               sk);
        if (err != 0)
                goto failure;
 
@@ -287,16 +123,6 @@ static int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr,
                                                            usin->sin_port);
        dccp_update_gss(sk, dp->dccps_iss);
 
-       /*
-        * SWL and AWL are initially adjusted so that they are not less than
-        * the initial Sequence Numbers received and sent, respectively:
-        *      SWL := max(GSR + 1 - floor(W/4), ISR),
-        *      AWL := max(GSS - W' + 1, ISS).
-        * These adjustments MUST be applied only at the beginning of the
-        * connection.
-        */
-       dccp_set_seqno(&dp->dccps_awl, max48(dp->dccps_awl, dp->dccps_iss));
-
        inet->id = dp->dccps_iss ^ jiffies;
 
        err = dccp_connect(sk);
@@ -316,6 +142,8 @@ failure:
        goto out;
 }
 
+EXPORT_SYMBOL_GPL(dccp_v4_connect);
+
 /*
  * This routine does path mtu discovery as defined in RFC1191.
  */
@@ -354,7 +182,7 @@ static inline void dccp_do_pmtu_discovery(struct sock *sk,
        mtu = dst_mtu(dst);
 
        if (inet->pmtudisc != IP_PMTUDISC_DONT &&
-           dp->dccps_pmtu_cookie > mtu) {
+           inet_csk(sk)->icsk_pmtu_cookie > mtu) {
                dccp_sync_mss(sk, mtu);
 
                /*
@@ -368,11 +196,12 @@ static inline void dccp_do_pmtu_discovery(struct sock *sk,
        } /* else let the usual retransmit timer handle it */
 }
 
-static void dccp_v4_ctl_send_ack(struct sk_buff *rxskb)
+static void dccp_v4_reqsk_send_ack(struct sk_buff *rxskb,
+                                  struct request_sock *req)
 {
        int err;
        struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
-       const int dccp_hdr_ack_len = sizeof(struct dccp_hdr) +
+       const u32 dccp_hdr_ack_len = sizeof(struct dccp_hdr) +
                                     sizeof(struct dccp_hdr_ext) +
                                     sizeof(struct dccp_hdr_ack_bits);
        struct sk_buff *skb;
@@ -380,12 +209,12 @@ static void dccp_v4_ctl_send_ack(struct sk_buff *rxskb)
        if (((struct rtable *)rxskb->dst)->rt_type != RTN_LOCAL)
                return;
 
-       skb = alloc_skb(MAX_DCCP_HEADER + 15, GFP_ATOMIC);
+       skb = alloc_skb(dccp_v4_ctl_socket->sk->sk_prot->max_header, GFP_ATOMIC);
        if (skb == NULL)
                return;
 
        /* Reserve space for headers. */
-       skb_reserve(skb, MAX_DCCP_HEADER);
+       skb_reserve(skb, dccp_v4_ctl_socket->sk->sk_prot->max_header);
 
        skb->dst = dst_clone(rxskb->dst);
 
@@ -404,11 +233,11 @@ static void dccp_v4_ctl_send_ack(struct sk_buff *rxskb)
        dccp_hdr_set_ack(dccp_hdr_ack_bits(skb),
                         DCCP_SKB_CB(rxskb)->dccpd_seq);
 
-       bh_lock_sock(dccp_ctl_socket->sk);
-       err = ip_build_and_send_pkt(skb, dccp_ctl_socket->sk,
+       bh_lock_sock(dccp_v4_ctl_socket->sk);
+       err = ip_build_and_send_pkt(skb, dccp_v4_ctl_socket->sk,
                                    rxskb->nh.iph->daddr,
                                    rxskb->nh.iph->saddr, NULL);
-       bh_unlock_sock(dccp_ctl_socket->sk);
+       bh_unlock_sock(dccp_v4_ctl_socket->sk);
 
        if (err == NET_XMIT_CN || err == 0) {
                DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
@@ -416,12 +245,6 @@ static void dccp_v4_ctl_send_ack(struct sk_buff *rxskb)
        }
 }
 
-static void dccp_v4_reqsk_send_ack(struct sk_buff *skb,
-                                  struct request_sock *req)
-{
-       dccp_v4_ctl_send_ack(skb);
-}
-
 static int dccp_v4_send_response(struct sock *sk, struct request_sock *req,
                                 struct dst_entry *dst)
 {
@@ -436,7 +259,10 @@ static int dccp_v4_send_response(struct sock *sk, struct request_sock *req,
        skb = dccp_make_response(sk, dst, req);
        if (skb != NULL) {
                const struct inet_request_sock *ireq = inet_rsk(req);
+               struct dccp_hdr *dh = dccp_hdr(skb);
 
+               dh->dccph_checksum = dccp_v4_checksum(skb, ireq->loc_addr,
+                                                     ireq->rmt_addr);
                memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
                err = ip_build_and_send_pkt(skb, sk, ireq->loc_addr,
                                            ireq->rmt_addr,
@@ -462,7 +288,7 @@ out:
  * check at all. A more general error queue to queue errors for later handling
  * is probably better.
  */
-void dccp_v4_err(struct sk_buff *skb, u32 info)
+static void dccp_v4_err(struct sk_buff *skb, u32 info)
 {
        const struct iphdr *iph = (struct iphdr *)skb->data;
        const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data +
@@ -606,32 +432,17 @@ out:
        sock_put(sk);
 }
 
-int dccp_v4_send_reset(struct sock *sk, enum dccp_reset_codes code)
+/* This routine computes an IPv4 DCCP checksum. */
+void dccp_v4_send_check(struct sock *sk, int len, struct sk_buff *skb)
 {
-       struct sk_buff *skb;
-       /*
-        * FIXME: what if rebuild_header fails?
-        * Should we be doing a rebuild_header here?
-        */
-       int err = inet_sk_rebuild_header(sk);
-
-       if (err != 0)
-               return err;
-
-       skb = dccp_make_reset(sk, sk->sk_dst_cache, code);
-       if (skb != NULL) {
-               const struct inet_sock *inet = inet_sk(sk);
-
-               memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
-               err = ip_build_and_send_pkt(skb, sk,
-                                           inet->saddr, inet->daddr, NULL);
-               if (err == NET_XMIT_CN)
-                       err = 0;
-       }
+       const struct inet_sock *inet = inet_sk(sk);
+       struct dccp_hdr *dh = dccp_hdr(skb);
 
-       return err;
+       dh->dccph_checksum = dccp_v4_checksum(skb, inet->saddr, inet->daddr);
 }
 
+EXPORT_SYMBOL_GPL(dccp_v4_send_check);
+
 static inline u64 dccp_v4_init_sequence(const struct sock *sk,
                                        const struct sk_buff *skb)
 {
@@ -641,28 +452,17 @@ static inline u64 dccp_v4_init_sequence(const struct sock *sk,
                                           dccp_hdr(skb)->dccph_sport);
 }
 
-static inline int dccp_bad_service_code(const struct sock *sk,
-                                       const __u32 service)
-{
-       const struct dccp_sock *dp = dccp_sk(sk);
-
-       if (dp->dccps_service == service)
-               return 0;
-       return !dccp_list_has_service(dp->dccps_service_list, service);
-}
-
 int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
 {
        struct inet_request_sock *ireq;
        struct dccp_sock dp;
        struct request_sock *req;
        struct dccp_request_sock *dreq;
-       const __u32 saddr = skb->nh.iph->saddr;
-       const __u32 daddr = skb->nh.iph->daddr;
-       const __u32 service = dccp_hdr_request(skb)->dccph_req_service;
+       const __be32 saddr = skb->nh.iph->saddr;
+       const __be32 daddr = skb->nh.iph->daddr;
+       const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
        struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
        __u8 reset_code = DCCP_RESET_CODE_TOO_BUSY;
-       struct dst_entry *dst = NULL;
 
        /* Never answer to DCCP_PKT_REQUESTs send to broadcast or multicast */
        if (((struct rtable *)skb->dst)->rt_flags &
@@ -696,16 +496,15 @@ int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
        if (req == NULL)
                goto drop;
 
-       /* FIXME: process options */
+       if (dccp_parse_options(sk, skb))
+               goto drop_and_free;
 
        dccp_openreq_init(req, &dp, skb);
 
        ireq = inet_rsk(req);
        ireq->loc_addr = daddr;
        ireq->rmt_addr = saddr;
-       /* FIXME: Merge Aristeu's option parsing code when ready */
-       req->rcv_wnd    = 100; /* Fake, option parsing will get the
-                                 right value */
+       req->rcv_wnd    = dccp_feat_default_sequence_window;
        ireq->opt       = NULL;
 
        /* 
@@ -721,23 +520,22 @@ int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
        dreq->dreq_iss     = dccp_v4_init_sequence(sk, skb);
        dreq->dreq_service = service;
 
-       if (dccp_v4_send_response(sk, req, dst))
+       if (dccp_v4_send_response(sk, req, NULL))
                goto drop_and_free;
 
        inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
        return 0;
 
 drop_and_free:
-       /*
-        * FIXME: should be reqsk_free after implementing req->rsk_ops
-        */
-       __reqsk_free(req);
+       reqsk_free(req);
 drop:
        DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
        dcb->dccpd_reset_code = reset_code;
        return -1;
 }
 
+EXPORT_SYMBOL_GPL(dccp_v4_conn_request);
+
 /*
  * The three way handshake has completed - we got a valid ACK or DATAACK -
  * now create the new socket.
@@ -792,6 +590,8 @@ exit:
        return NULL;
 }
 
+EXPORT_SYMBOL_GPL(dccp_v4_request_recv_sock);
+
 static struct sock *dccp_v4_hnd_req(struct sock *sk, struct sk_buff *skb)
 {
        const struct dccp_hdr *dh = dccp_hdr(skb);
@@ -821,8 +621,8 @@ static struct sock *dccp_v4_hnd_req(struct sock *sk, struct sk_buff *skb)
        return sk;
 }
 
-int dccp_v4_checksum(const struct sk_buff *skb, const u32 saddr,
-                    const u32 daddr)
+int dccp_v4_checksum(const struct sk_buff *skb, const __be32 saddr,
+                    const __be32 daddr)
 {
        const struct dccp_hdr* dh = dccp_hdr(skb);
        int checksum_len;
@@ -841,8 +641,10 @@ int dccp_v4_checksum(const struct sk_buff *skb, const u32 saddr,
                                 IPPROTO_DCCP, tmp);
 }
 
+EXPORT_SYMBOL_GPL(dccp_v4_checksum);
+
 static int dccp_v4_verify_checksum(struct sk_buff *skb,
-                                  const u32 saddr, const u32 daddr)
+                                  const __be32 saddr, const __be32 daddr)
 {
        struct dccp_hdr *dh = dccp_hdr(skb);
        int checksum_len;
@@ -902,16 +704,17 @@ static void dccp_v4_ctl_send_reset(struct sk_buff *rxskb)
        if (((struct rtable *)rxskb->dst)->rt_type != RTN_LOCAL)
                return;
 
-       dst = dccp_v4_route_skb(dccp_ctl_socket->sk, rxskb);
+       dst = dccp_v4_route_skb(dccp_v4_ctl_socket->sk, rxskb);
        if (dst == NULL)
                return;
 
-       skb = alloc_skb(MAX_DCCP_HEADER + 15, GFP_ATOMIC);
+       skb = alloc_skb(dccp_v4_ctl_socket->sk->sk_prot->max_header,
+                       GFP_ATOMIC);
        if (skb == NULL)
                goto out;
 
        /* Reserve space for headers. */
-       skb_reserve(skb, MAX_DCCP_HEADER);
+       skb_reserve(skb, dccp_v4_ctl_socket->sk->sk_prot->max_header);
        skb->dst = dst_clone(dst);
 
        skb->h.raw = skb_push(skb, dccp_hdr_reset_len);
@@ -939,11 +742,11 @@ static void dccp_v4_ctl_send_reset(struct sk_buff *rxskb)
        dh->dccph_checksum = dccp_v4_checksum(skb, rxskb->nh.iph->saddr,
                                              rxskb->nh.iph->daddr);
 
-       bh_lock_sock(dccp_ctl_socket->sk);
-       err = ip_build_and_send_pkt(skb, dccp_ctl_socket->sk,
+       bh_lock_sock(dccp_v4_ctl_socket->sk);
+       err = ip_build_and_send_pkt(skb, dccp_v4_ctl_socket->sk,
                                    rxskb->nh.iph->daddr,
                                    rxskb->nh.iph->saddr, NULL);
-       bh_unlock_sock(dccp_ctl_socket->sk);
+       bh_unlock_sock(dccp_v4_ctl_socket->sk);
 
        if (err == NET_XMIT_CN || err == 0) {
                DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
@@ -1011,7 +814,9 @@ discard:
        return 0;
 }
 
-static inline int dccp_invalid_packet(struct sk_buff *skb)
+EXPORT_SYMBOL_GPL(dccp_v4_do_rcv);
+
+int dccp_invalid_packet(struct sk_buff *skb)
 {
        const struct dccp_hdr *dh;
 
@@ -1065,29 +870,30 @@ static inline int dccp_invalid_packet(struct sk_buff *skb)
                return 1;
        }
 
-       /* If the header checksum is incorrect, drop packet and return */
-       if (dccp_v4_verify_checksum(skb, skb->nh.iph->saddr,
-                                   skb->nh.iph->daddr) < 0) {
-               LIMIT_NETDEBUG(KERN_WARNING "DCCP: header checksum is "
-                                           "incorrect\n");
-               return 1;
-       }
-
        return 0;
 }
 
+EXPORT_SYMBOL_GPL(dccp_invalid_packet);
+
 /* this is called when real data arrives */
-int dccp_v4_rcv(struct sk_buff *skb)
+static int dccp_v4_rcv(struct sk_buff *skb)
 {
        const struct dccp_hdr *dh;
        struct sock *sk;
-       int rc;
 
        /* Step 1: Check header basics: */
 
        if (dccp_invalid_packet(skb))
                goto discard_it;
 
+       /* If the header checksum is incorrect, drop packet and return */
+       if (dccp_v4_verify_checksum(skb, skb->nh.iph->saddr,
+                                   skb->nh.iph->daddr) < 0) {
+               LIMIT_NETDEBUG(KERN_WARNING "%s: incorrect header checksum\n",
+                              __FUNCTION__);
+               goto discard_it;
+       }
+
        dh = dccp_hdr(skb);
 
        DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(skb);
@@ -1143,28 +949,11 @@ int dccp_v4_rcv(struct sk_buff *skb)
                 goto do_time_wait;
        }
 
-       if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
-               dccp_pr_debug("xfrm4_policy_check failed\n");
+       if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
                goto discard_and_relse;
-       }
-
-        if (sk_filter(sk, skb, 0)) {
-               dccp_pr_debug("sk_filter failed\n");
-                goto discard_and_relse;
-       }
+       nf_reset(skb);
 
-       skb->dev = NULL;
-
-       bh_lock_sock(sk);
-       rc = 0;
-       if (!sock_owned_by_user(sk))
-               rc = dccp_v4_do_rcv(sk, skb);
-       else
-               sk_add_backlog(sk, skb);
-       bh_unlock_sock(sk);
-
-       sock_put(sk);
-       return rc;
+       return sk_receive_skb(sk, skb);
 
 no_dccp_socket:
        if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
@@ -1194,89 +983,35 @@ do_time_wait:
        goto no_dccp_socket;
 }
 
+static struct inet_connection_sock_af_ops dccp_ipv4_af_ops = {
+       .queue_xmit        = ip_queue_xmit,
+       .send_check        = dccp_v4_send_check,
+       .rebuild_header    = inet_sk_rebuild_header,
+       .conn_request      = dccp_v4_conn_request,
+       .syn_recv_sock     = dccp_v4_request_recv_sock,
+       .net_header_len    = sizeof(struct iphdr),
+       .setsockopt        = ip_setsockopt,
+       .getsockopt        = ip_getsockopt,
+       .addr2sockaddr     = inet_csk_addr2sockaddr,
+       .sockaddr_len      = sizeof(struct sockaddr_in),
+#ifdef CONFIG_COMPAT
+       .compat_setsockopt = compat_ip_setsockopt,
+       .compat_getsockopt = compat_ip_getsockopt,
+#endif
+};
+
 static int dccp_v4_init_sock(struct sock *sk)
 {
-       struct dccp_sock *dp = dccp_sk(sk);
-       static int dccp_ctl_socket_init = 1;
-
-       dccp_options_init(&dp->dccps_options);
-       do_gettimeofday(&dp->dccps_epoch);
+       static __u8 dccp_v4_ctl_sock_initialized;
+       int err = dccp_init_sock(sk, dccp_v4_ctl_sock_initialized);
 
-       if (dp->dccps_options.dccpo_send_ack_vector) {
-               dp->dccps_hc_rx_ackvec = dccp_ackvec_alloc(DCCP_MAX_ACKVEC_LEN,
-                                                          GFP_KERNEL);
-               if (dp->dccps_hc_rx_ackvec == NULL)
-                       return -ENOMEM;
+       if (err == 0) {
+               if (unlikely(!dccp_v4_ctl_sock_initialized))
+                       dccp_v4_ctl_sock_initialized = 1;
+               inet_csk(sk)->icsk_af_ops = &dccp_ipv4_af_ops;
        }
 
-       /*
-        * FIXME: We're hardcoding the CCID, and doing this at this point makes
-        * the listening (master) sock get CCID control blocks, which is not
-        * necessary, but for now, to not mess with the test userspace apps,
-        * lets leave it here, later the real solution is to do this in a
-        * setsockopt(CCIDs-I-want/accept). -acme
-        */
-       if (likely(!dccp_ctl_socket_init)) {
-               dp->dccps_hc_rx_ccid = ccid_init(dp->dccps_options.dccpo_rx_ccid,
-                                                sk);
-               dp->dccps_hc_tx_ccid = ccid_init(dp->dccps_options.dccpo_tx_ccid,
-                                                sk);
-               if (dp->dccps_hc_rx_ccid == NULL ||
-                   dp->dccps_hc_tx_ccid == NULL) {
-                       ccid_exit(dp->dccps_hc_rx_ccid, sk);
-                       ccid_exit(dp->dccps_hc_tx_ccid, sk);
-                       if (dp->dccps_options.dccpo_send_ack_vector) {
-                               dccp_ackvec_free(dp->dccps_hc_rx_ackvec);
-                               dp->dccps_hc_rx_ackvec = NULL;
-                       }
-                       dp->dccps_hc_rx_ccid = dp->dccps_hc_tx_ccid = NULL;
-                       return -ENOMEM;
-               }
-       } else
-               dccp_ctl_socket_init = 0;
-
-       dccp_init_xmit_timers(sk);
-       inet_csk(sk)->icsk_rto = DCCP_TIMEOUT_INIT;
-       sk->sk_state = DCCP_CLOSED;
-       sk->sk_write_space = dccp_write_space;
-       dp->dccps_mss_cache = 536;
-       dp->dccps_role = DCCP_ROLE_UNDEFINED;
-       dp->dccps_service = DCCP_SERVICE_INVALID_VALUE;
-
-       return 0;
-}
-
-static int dccp_v4_destroy_sock(struct sock *sk)
-{
-       struct dccp_sock *dp = dccp_sk(sk);
-
-       /*
-        * DCCP doesn't use sk_write_queue, just sk_send_head
-        * for retransmissions
-        */
-       if (sk->sk_send_head != NULL) {
-               kfree_skb(sk->sk_send_head);
-               sk->sk_send_head = NULL;
-       }
-
-       /* Clean up a referenced DCCP bind bucket. */
-       if (inet_csk(sk)->icsk_bind_hash != NULL)
-               inet_put_port(&dccp_hashinfo, sk);
-
-       kfree(dp->dccps_service_list);
-       dp->dccps_service_list = NULL;
-
-       ccid_hc_rx_exit(dp->dccps_hc_rx_ccid, sk);
-       ccid_hc_tx_exit(dp->dccps_hc_tx_ccid, sk);
-       if (dp->dccps_options.dccpo_send_ack_vector) {
-               dccp_ackvec_free(dp->dccps_hc_rx_ackvec);
-               dp->dccps_hc_rx_ackvec = NULL;
-       }
-       ccid_exit(dp->dccps_hc_rx_ccid, sk);
-       ccid_exit(dp->dccps_hc_tx_ccid, sk);
-       dp->dccps_hc_rx_ccid = dp->dccps_hc_tx_ccid = NULL;
-
-       return 0;
+       return err;
 }
 
 static void dccp_v4_reqsk_destructor(struct request_sock *req)
@@ -1293,7 +1028,11 @@ static struct request_sock_ops dccp_request_sock_ops = {
        .send_reset     = dccp_v4_ctl_send_reset,
 };
 
-struct proto dccp_v4_prot = {
+static struct timewait_sock_ops dccp_timewait_sock_ops = {
+       .twsk_obj_size  = sizeof(struct inet_timewait_sock),
+};
+
+static struct proto dccp_v4_prot = {
        .name                   = "DCCP",
        .owner                  = THIS_MODULE,
        .close                  = dccp_close,
@@ -1306,15 +1045,110 @@ struct proto dccp_v4_prot = {
        .sendmsg                = dccp_sendmsg,
        .recvmsg                = dccp_recvmsg,
        .backlog_rcv            = dccp_v4_do_rcv,
-       .hash                   = dccp_v4_hash,
-       .unhash                 = dccp_v4_unhash,
+       .hash                   = dccp_hash,
+       .unhash                 = dccp_unhash,
        .accept                 = inet_csk_accept,
        .get_port               = dccp_v4_get_port,
        .shutdown               = dccp_shutdown,
-       .destroy                = dccp_v4_destroy_sock,
+       .destroy                = dccp_destroy_sock,
        .orphan_count           = &dccp_orphan_count,
        .max_header             = MAX_DCCP_HEADER,
        .obj_size               = sizeof(struct dccp_sock),
        .rsk_prot               = &dccp_request_sock_ops,
-       .twsk_obj_size          = sizeof(struct inet_timewait_sock),
+       .twsk_prot              = &dccp_timewait_sock_ops,
+#ifdef CONFIG_COMPAT
+       .compat_setsockopt      = compat_dccp_setsockopt,
+       .compat_getsockopt      = compat_dccp_getsockopt,
+#endif
+};
+
+static struct net_protocol dccp_v4_protocol = {
+       .handler        = dccp_v4_rcv,
+       .err_handler    = dccp_v4_err,
+       .no_policy      = 1,
+};
+
+static const struct proto_ops inet_dccp_ops = {
+       .family            = PF_INET,
+       .owner             = THIS_MODULE,
+       .release           = inet_release,
+       .bind              = inet_bind,
+       .connect           = inet_stream_connect,
+       .socketpair        = sock_no_socketpair,
+       .accept            = inet_accept,
+       .getname           = inet_getname,
+       /* FIXME: work on tcp_poll to rename it to inet_csk_poll */
+       .poll              = dccp_poll,
+       .ioctl             = inet_ioctl,
+       /* FIXME: work on inet_listen to rename it to sock_common_listen */
+       .listen            = inet_dccp_listen,
+       .shutdown          = inet_shutdown,
+       .setsockopt        = sock_common_setsockopt,
+       .getsockopt        = sock_common_getsockopt,
+       .sendmsg           = inet_sendmsg,
+       .recvmsg           = sock_common_recvmsg,
+       .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 dccp_v4_protosw = {
+       .type           = SOCK_DCCP,
+       .protocol       = IPPROTO_DCCP,
+       .prot           = &dccp_v4_prot,
+       .ops            = &inet_dccp_ops,
+       .capability     = -1,
+       .no_check       = 0,
+       .flags          = INET_PROTOSW_ICSK,
+};
+
+static int __init dccp_v4_init(void)
+{
+       int err = proto_register(&dccp_v4_prot, 1);
+
+       if (err != 0)
+               goto out;
+
+       err = inet_add_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
+       if (err != 0)
+               goto out_proto_unregister;
+
+       inet_register_protosw(&dccp_v4_protosw);
+
+       err = inet_csk_ctl_sock_create(&dccp_v4_ctl_socket, PF_INET,
+                                      SOCK_DCCP, IPPROTO_DCCP);
+       if (err)
+               goto out_unregister_protosw;
+out:
+       return err;
+out_unregister_protosw:
+       inet_unregister_protosw(&dccp_v4_protosw);
+       inet_del_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
+out_proto_unregister:
+       proto_unregister(&dccp_v4_prot);
+       goto out;
+}
+
+static void __exit dccp_v4_exit(void)
+{
+       inet_unregister_protosw(&dccp_v4_protosw);
+       inet_del_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
+       proto_unregister(&dccp_v4_prot);
+}
+
+module_init(dccp_v4_init);
+module_exit(dccp_v4_exit);
+
+/*
+ * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
+ * values directly, Also cover the case where the protocol is not specified,
+ * i.e. net-pf-PF_INET-proto-0-type-SOCK_DCCP
+ */
+MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-33-type-6");
+MODULE_ALIAS("net-pf-" __stringify(PF_INET) "-proto-0-type-6");
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
+MODULE_DESCRIPTION("DCCP - Datagram Congestion Controlled Protocol");