dlm-user: BKL pushdown
[safe/jmp/linux-2.6] / net / dccp / ipv6.c
index 627d0c3..9b1129b 100644 (file)
 #include "ipv6.h"
 #include "feat.h"
 
-/* Socket used for sending RSTs and ACKs */
-static struct socket *dccp_v6_ctl_socket;
+/* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */
 
 static struct inet_connection_sock_af_ops dccp_ipv6_mapped;
 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
 
-static int dccp_v6_get_port(struct sock *sk, unsigned short snum)
-{
-       return inet_csk_get_port(&dccp_hashinfo, sk, snum,
-                                inet6_csk_bind_conflict);
-}
-
 static void dccp_v6_hash(struct sock *sk)
 {
        if (sk->sk_state != DCCP_CLOSED) {
                if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) {
-                       dccp_hash(sk);
+                       inet_hash(sk);
                        return;
                }
                local_bh_disable();
-               __inet6_hash(&dccp_hashinfo, sk);
+               __inet6_hash(sk);
                local_bh_enable();
        }
 }
@@ -84,8 +77,8 @@ static inline __u32 secure_dccpv6_sequence_number(__be32 *saddr, __be32 *daddr,
 
 static inline __u32 dccp_v6_init_sequence(struct sk_buff *skb)
 {
-       return secure_dccpv6_sequence_number(skb->nh.ipv6h->daddr.s6_addr32,
-                                            skb->nh.ipv6h->saddr.s6_addr32,
+       return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
+                                            ipv6_hdr(skb)->saddr.s6_addr32,
                                             dccp_hdr(skb)->dccph_dport,
                                             dccp_hdr(skb)->dccph_sport     );
 
@@ -101,8 +94,9 @@ static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
        int err;
        __u64 seq;
 
-       sk = inet6_lookup(&dccp_hashinfo, &hdr->daddr, dh->dccph_dport,
-                         &hdr->saddr, dh->dccph_sport, inet6_iif(skb));
+       sk = inet6_lookup(dev_net(skb->dev), &dccp_hashinfo,
+                       &hdr->daddr, dh->dccph_dport,
+                       &hdr->saddr, dh->dccph_sport, inet6_iif(skb));
 
        if (sk == NULL) {
                ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
@@ -173,7 +167,7 @@ static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 
        icmpv6_err_convert(type, code, &err);
 
-       seq = DCCP_SKB_CB(skb)->dccpd_seq;
+       seq = dccp_hdr_seq(dh);
        /* Might be for an request_sock */
        switch (sk->sk_state) {
                struct request_sock *req, **prev;
@@ -230,8 +224,7 @@ out:
 }
 
 
-static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
-                                struct dst_entry *dst)
+static int dccp_v6_send_response(struct sock *sk, struct request_sock *req)
 {
        struct inet6_request_sock *ireq6 = inet6_rsk(req);
        struct ipv6_pinfo *np = inet6_sk(sk);
@@ -240,6 +233,7 @@ static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
        struct in6_addr *final_p = NULL, final;
        struct flowi fl;
        int err = -1;
+       struct dst_entry *dst;
 
        memset(&fl, 0, sizeof(fl));
        fl.proto = IPPROTO_DCCP;
@@ -251,39 +245,26 @@ static int dccp_v6_send_response(struct sock *sk, struct request_sock *req,
        fl.fl_ip_sport = inet_sk(sk)->sport;
        security_req_classify_flow(req, &fl);
 
-       if (dst == NULL) {
-               opt = np->opt;
-               if (opt == NULL &&
-                   np->rxopt.bits.osrcrt == 2 &&
-                   ireq6->pktopts) {
-                       struct sk_buff *pktopts = ireq6->pktopts;
-                       struct inet6_skb_parm *rxopt = IP6CB(pktopts);
-
-                       if (rxopt->srcrt)
-                               opt = ipv6_invert_rthdr(sk,
-                         (struct ipv6_rt_hdr *)(skb_network_header(pktopts) +
-                                                rxopt->srcrt));
-               }
+       opt = np->opt;
 
-               if (opt != NULL && opt->srcrt != NULL) {
-                       const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
+       if (opt != NULL && opt->srcrt != NULL) {
+               const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt;
 
-                       ipv6_addr_copy(&final, &fl.fl6_dst);
-                       ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
-                       final_p = &final;
-               }
+               ipv6_addr_copy(&final, &fl.fl6_dst);
+               ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
+               final_p = &final;
+       }
 
-               err = ip6_dst_lookup(sk, &dst, &fl);
-               if (err)
-                       goto done;
+       err = ip6_dst_lookup(sk, &dst, &fl);
+       if (err)
+               goto done;
 
-               if (final_p)
-                       ipv6_addr_copy(&fl.fl6_dst, final_p);
+       if (final_p)
+               ipv6_addr_copy(&fl.fl6_dst, final_p);
 
-               err = xfrm_lookup(&dst, &fl, sk, 0);
-               if (err < 0)
-                       goto done;
-       }
+       err = xfrm_lookup(&dst, &fl, sk, 0);
+       if (err < 0)
+               goto done;
 
        skb = dccp_make_response(sk, dst, req);
        if (skb != NULL) {
@@ -312,63 +293,40 @@ static void dccp_v6_reqsk_destructor(struct request_sock *req)
 
 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
 {
-       struct dccp_hdr *rxdh = dccp_hdr(rxskb), *dh;
-       const u32 dccp_hdr_reset_len = sizeof(struct dccp_hdr) +
-                                      sizeof(struct dccp_hdr_ext) +
-                                      sizeof(struct dccp_hdr_reset);
+       struct ipv6hdr *rxip6h;
        struct sk_buff *skb;
        struct flowi fl;
-       u64 seqno = 0;
+       struct net *net = dev_net(rxskb->dst->dev);
+       struct sock *ctl_sk = net->dccp.v6_ctl_sk;
 
-       if (rxdh->dccph_type == DCCP_PKT_RESET)
+       if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
                return;
 
        if (!ipv6_unicast_destination(rxskb))
                return;
 
-       skb = alloc_skb(dccp_v6_ctl_socket->sk->sk_prot->max_header,
-                       GFP_ATOMIC);
+       skb = dccp_ctl_make_reset(ctl_sk, rxskb);
        if (skb == NULL)
                return;
 
-       skb_reserve(skb, dccp_v6_ctl_socket->sk->sk_prot->max_header);
-
-       dh = dccp_zeroed_hdr(skb, dccp_hdr_reset_len);
-
-       /* Swap the send and the receive. */
-       dh->dccph_type  = DCCP_PKT_RESET;
-       dh->dccph_sport = rxdh->dccph_dport;
-       dh->dccph_dport = rxdh->dccph_sport;
-       dh->dccph_doff  = dccp_hdr_reset_len / 4;
-       dh->dccph_x     = 1;
-       dccp_hdr_reset(skb)->dccph_reset_code =
-                               DCCP_SKB_CB(rxskb)->dccpd_reset_code;
-
-       /* See "8.3.1. Abnormal Termination" in RFC 4340 */
-       if (DCCP_SKB_CB(rxskb)->dccpd_ack_seq != DCCP_PKT_WITHOUT_ACK_SEQ)
-               dccp_set_seqno(&seqno, DCCP_SKB_CB(rxskb)->dccpd_ack_seq + 1);
-
-       dccp_hdr_set_seq(dh, seqno);
-       dccp_hdr_set_ack(dccp_hdr_ack_bits(skb), DCCP_SKB_CB(rxskb)->dccpd_seq);
-
-       dccp_csum_outgoing(skb);
-       dh->dccph_checksum = dccp_v6_csum_finish(skb, &rxskb->nh.ipv6h->saddr,
-                                                     &rxskb->nh.ipv6h->daddr);
+       rxip6h = ipv6_hdr(rxskb);
+       dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
+                                                           &rxip6h->daddr);
 
        memset(&fl, 0, sizeof(fl));
-       ipv6_addr_copy(&fl.fl6_dst, &rxskb->nh.ipv6h->saddr);
-       ipv6_addr_copy(&fl.fl6_src, &rxskb->nh.ipv6h->daddr);
+       ipv6_addr_copy(&fl.fl6_dst, &rxip6h->saddr);
+       ipv6_addr_copy(&fl.fl6_src, &rxip6h->daddr);
 
        fl.proto = IPPROTO_DCCP;
        fl.oif = inet6_iif(rxskb);
-       fl.fl_ip_dport = dh->dccph_dport;
-       fl.fl_ip_sport = dh->dccph_sport;
+       fl.fl_ip_dport = dccp_hdr(skb)->dccph_dport;
+       fl.fl_ip_sport = dccp_hdr(skb)->dccph_sport;
        security_skb_classify_flow(rxskb, &fl);
 
        /* sk = NULL, but it is safe for now. RST socket required. */
-       if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) {
+       if (!ip6_dst_lookup(ctl_sk, &skb->dst, &fl)) {
                if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) {
-                       ip6_xmit(dccp_v6_ctl_socket->sk, skb, &fl, NULL, 0);
+                       ip6_xmit(ctl_sk, skb, &fl, NULL, 0);
                        DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
                        DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
                        return;
@@ -390,7 +348,7 @@ static struct request_sock_ops dccp6_request_sock_ops = {
 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
 {
        const struct dccp_hdr *dh = dccp_hdr(skb);
-       const struct ipv6hdr *iph = skb->nh.ipv6h;
+       const struct ipv6hdr *iph = ipv6_hdr(skb);
        struct sock *nsk;
        struct request_sock **prev;
        /* Find possible connection requests. */
@@ -402,7 +360,7 @@ static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
        if (req != NULL)
                return dccp_check_req(sk, skb, req, prev);
 
-       nsk = __inet6_lookup_established(&dccp_hashinfo,
+       nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo,
                                         &iph->saddr, dh->dccph_sport,
                                         &iph->daddr, ntohs(dh->dccph_dport),
                                         inet6_iif(skb));
@@ -426,21 +384,21 @@ static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
        struct ipv6_pinfo *np = inet6_sk(sk);
        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;
 
        if (skb->protocol == htons(ETH_P_IP))
                return dccp_v4_conn_request(sk, skb);
 
        if (!ipv6_unicast_destination(skb))
-               goto drop;
+               return 0;       /* discard, don't send a reset here */
 
        if (dccp_bad_service_code(sk, service)) {
-               reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
+               dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
                goto drop;
        }
        /*
         * There are no SYN attacks on IPv6, yet...
         */
+       dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
        if (inet_csk_reqsk_queue_is_full(sk))
                goto drop;
 
@@ -451,17 +409,18 @@ static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
        if (req == NULL)
                goto drop;
 
-       if (dccp_parse_options(sk, skb))
-               goto drop_and_free;
-
        dccp_reqsk_init(req, skb);
 
+       dreq = dccp_rsk(req);
+       if (dccp_parse_options(sk, dreq, skb))
+               goto drop_and_free;
+
        if (security_inet_conn_request(sk, skb, req))
                goto drop_and_free;
 
        ireq6 = inet6_rsk(req);
-       ipv6_addr_copy(&ireq6->rmt_addr, &skb->nh.ipv6h->saddr);
-       ipv6_addr_copy(&ireq6->loc_addr, &skb->nh.ipv6h->daddr);
+       ipv6_addr_copy(&ireq6->rmt_addr, &ipv6_hdr(skb)->saddr);
+       ipv6_addr_copy(&ireq6->loc_addr, &ipv6_hdr(skb)->daddr);
        ireq6->pktopts  = NULL;
 
        if (ipv6_opt_accepted(sk, skb) ||
@@ -485,12 +444,11 @@ static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
         *   In fact we defer setting S.GSR, S.SWL, S.SWH to
         *   dccp_create_openreq_child.
         */
-       dreq = dccp_rsk(req);
        dreq->dreq_isr     = dcb->dccpd_seq;
        dreq->dreq_iss     = dccp_v6_init_sequence(skb);
        dreq->dreq_service = service;
 
-       if (dccp_v6_send_response(sk, req, NULL))
+       if (dccp_v6_send_response(sk, req))
                goto drop_and_free;
 
        inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
@@ -500,7 +458,6 @@ drop_and_free:
        reqsk_free(req);
 drop:
        DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
-       dcb->dccpd_reset_code = reset_code;
        return -1;
 }
 
@@ -546,7 +503,7 @@ static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
                newnp->pktoptions  = NULL;
                newnp->opt         = NULL;
                newnp->mcast_oif   = inet6_iif(skb);
-               newnp->mcast_hops  = skb->nh.ipv6h->hop_limit;
+               newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
 
                /*
                 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
@@ -568,15 +525,6 @@ static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
        if (sk_acceptq_is_full(sk))
                goto out_overflow;
 
-       if (np->rxopt.bits.osrcrt == 2 && opt == NULL && ireq6->pktopts) {
-               const struct inet6_skb_parm *rxopt = IP6CB(ireq6->pktopts);
-
-               if (rxopt->srcrt)
-                       opt = ipv6_invert_rthdr(sk,
-                  (struct ipv6_rt_hdr *)(skb_network_header(ireq6->pktopts) +
-                                         rxopt->srcrt));
-       }
-
        if (dst == NULL) {
                struct in6_addr *final_p = NULL, final;
                struct flowi fl;
@@ -653,7 +601,7 @@ static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
        }
        newnp->opt        = NULL;
        newnp->mcast_oif  = inet6_iif(skb);
-       newnp->mcast_hops = skb->nh.ipv6h->hop_limit;
+       newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
 
        /*
         * Clone native IPv6 options from listening socket (if any)
@@ -676,8 +624,8 @@ static struct sock *dccp_v6_request_recv_sock(struct sock *sk,
 
        newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
 
-       __inet6_hash(&dccp_hashinfo, newsk);
-       inet_inherit_port(&dccp_hashinfo, sk, newsk);
+       __inet6_hash(newsk);
+       __inet_inherit_port(sk, newsk);
 
        return newsk;
 
@@ -813,10 +761,9 @@ discard:
        return 0;
 }
 
-static int dccp_v6_rcv(struct sk_buff **pskb)
+static int dccp_v6_rcv(struct sk_buff *skb)
 {
        const struct dccp_hdr *dh;
-       struct sk_buff *skb = *pskb;
        struct sock *sk;
        int min_cov;
 
@@ -826,15 +773,15 @@ static int dccp_v6_rcv(struct sk_buff **pskb)
                goto discard_it;
 
        /* Step 1: If header checksum is incorrect, drop packet and return. */
-       if (dccp_v6_csum_finish(skb, &skb->nh.ipv6h->saddr,
-                                    &skb->nh.ipv6h->daddr)) {
+       if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
+                                    &ipv6_hdr(skb)->daddr)) {
                DCCP_WARN("dropped packet with invalid checksum\n");
                goto discard_it;
        }
 
        dh = dccp_hdr(skb);
 
-       DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(skb);
+       DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
        DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
 
        if (dccp_packet_without_ack(skb))
@@ -844,9 +791,9 @@ static int dccp_v6_rcv(struct sk_buff **pskb)
 
        /* Step 2:
         *      Look up flow ID in table and get corresponding socket */
-       sk = __inet6_lookup(&dccp_hashinfo, &skb->nh.ipv6h->saddr,
-                           dh->dccph_sport,
-                           &skb->nh.ipv6h->daddr, ntohs(dh->dccph_dport),
+       sk = __inet6_lookup(dev_net(skb->dst->dev), &dccp_hashinfo,
+                           &ipv6_hdr(skb)->saddr, dh->dccph_sport,
+                           &ipv6_hdr(skb)->daddr, ntohs(dh->dccph_dport),
                            inet6_iif(skb));
        /*
         * Step 2:
@@ -1041,9 +988,13 @@ static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
        if (final_p)
                ipv6_addr_copy(&fl.fl6_dst, final_p);
 
-       err = xfrm_lookup(&dst, &fl, sk, 1);
-       if (err < 0)
-               goto failure;
+       err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT);
+       if (err < 0) {
+               if (err == -EREMOTE)
+                       err = ip6_dst_blackhole(sk, &dst, &fl);
+               if (err < 0)
+                       goto failure;
+       }
 
        if (saddr == NULL) {
                saddr = &fl.fl6_src;
@@ -1097,6 +1048,7 @@ static struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
        .getsockopt        = ipv6_getsockopt,
        .addr2sockaddr     = inet6_csk_addr2sockaddr,
        .sockaddr_len      = sizeof(struct sockaddr_in6),
+       .bind_conflict     = inet6_csk_bind_conflict,
 #ifdef CONFIG_COMPAT
        .compat_setsockopt = compat_ipv6_setsockopt,
        .compat_getsockopt = compat_ipv6_getsockopt,
@@ -1164,9 +1116,9 @@ static struct proto dccp_v6_prot = {
        .recvmsg           = dccp_recvmsg,
        .backlog_rcv       = dccp_v6_do_rcv,
        .hash              = dccp_v6_hash,
-       .unhash            = dccp_unhash,
+       .unhash            = inet_unhash,
        .accept            = inet_csk_accept,
-       .get_port          = dccp_v6_get_port,
+       .get_port          = inet_csk_get_port,
        .shutdown          = dccp_shutdown,
        .destroy           = dccp_v6_destroy_sock,
        .orphan_count      = &dccp_orphan_count,
@@ -1174,6 +1126,7 @@ static struct proto dccp_v6_prot = {
        .obj_size          = sizeof(struct dccp6_sock),
        .rsk_prot          = &dccp6_request_sock_ops,
        .twsk_prot         = &dccp6_timewait_sock_ops,
+       .h.hashinfo        = &dccp_hashinfo,
 #ifdef CONFIG_COMPAT
        .compat_setsockopt = compat_dccp_setsockopt,
        .compat_getsockopt = compat_dccp_getsockopt,
@@ -1220,6 +1173,25 @@ static struct inet_protosw dccp_v6_protosw = {
        .flags          = INET_PROTOSW_ICSK,
 };
 
+static int dccp_v6_init_net(struct net *net)
+{
+       int err;
+
+       err = inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
+                                  SOCK_DCCP, IPPROTO_DCCP, net);
+       return err;
+}
+
+static void dccp_v6_exit_net(struct net *net)
+{
+       inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
+}
+
+static struct pernet_operations dccp_v6_ops = {
+       .init   = dccp_v6_init_net,
+       .exit   = dccp_v6_exit_net,
+};
+
 static int __init dccp_v6_init(void)
 {
        int err = proto_register(&dccp_v6_prot, 1);
@@ -1233,13 +1205,13 @@ static int __init dccp_v6_init(void)
 
        inet6_register_protosw(&dccp_v6_protosw);
 
-       err = inet_csk_ctl_sock_create(&dccp_v6_ctl_socket, PF_INET6,
-                                      SOCK_DCCP, IPPROTO_DCCP);
+       err = register_pernet_subsys(&dccp_v6_ops);
        if (err != 0)
-               goto out_unregister_protosw;
+               goto out_destroy_ctl_sock;
 out:
        return err;
-out_unregister_protosw:
+
+out_destroy_ctl_sock:
        inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
        inet6_unregister_protosw(&dccp_v6_protosw);
 out_unregister_proto:
@@ -1249,6 +1221,7 @@ out_unregister_proto:
 
 static void __exit dccp_v6_exit(void)
 {
+       unregister_pernet_subsys(&dccp_v6_ops);
        inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
        inet6_unregister_protosw(&dccp_v6_protosw);
        proto_unregister(&dccp_v6_prot);
@@ -1262,8 +1235,8 @@ module_exit(dccp_v6_exit);
  * values directly, Also cover the case where the protocol is not specified,
  * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
  */
-MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-33-type-6");
-MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-0-type-6");
+MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
+MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");