[NETFILTER]: Introduce NF_INET_ hook values
[safe/jmp/linux-2.6] / net / ipv6 / ip6_output.c
index 1900c62..d54da61 100644 (file)
@@ -29,7 +29,7 @@
  */
 
 #include <linux/errno.h>
-#include <linux/types.h>
+#include <linux/kernel.h>
 #include <linux/string.h>
 #include <linux/socket.h>
 #include <linux/net.h>
@@ -70,7 +70,32 @@ static __inline__ void ipv6_select_ident(struct sk_buff *skb, struct frag_hdr *f
        spin_unlock_bh(&ip6_id_lock);
 }
 
-static inline int ip6_output_finish(struct sk_buff *skb)
+int __ip6_local_out(struct sk_buff *skb)
+{
+       int len;
+
+       len = skb->len - sizeof(struct ipv6hdr);
+       if (len > IPV6_MAXPLEN)
+               len = 0;
+       ipv6_hdr(skb)->payload_len = htons(len);
+
+       return nf_hook(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, skb->dst->dev,
+                      dst_output);
+}
+
+int ip6_local_out(struct sk_buff *skb)
+{
+       int err;
+
+       err = __ip6_local_out(skb);
+       if (likely(err == 1))
+               err = dst_output(skb);
+
+       return err;
+}
+EXPORT_SYMBOL_GPL(ip6_local_out);
+
+static int ip6_output_finish(struct sk_buff *skb)
 {
        struct dst_entry *dst = skb->dst;
 
@@ -120,8 +145,8 @@ static int ip6_output2(struct sk_buff *skb)
                           is not supported in any case.
                         */
                        if (newskb)
-                               NF_HOOK(PF_INET6, NF_IP6_POST_ROUTING, newskb, NULL,
-                                       newskb->dev,
+                               NF_HOOK(PF_INET6, NF_INET_POST_ROUTING, newskb,
+                                       NULL, newskb->dev,
                                        ip6_dev_loopback_xmit);
 
                        if (ipv6_hdr(skb)->hop_limit == 0) {
@@ -134,12 +159,21 @@ static int ip6_output2(struct sk_buff *skb)
                IP6_INC_STATS(idev, IPSTATS_MIB_OUTMCASTPKTS);
        }
 
-       return NF_HOOK(PF_INET6, NF_IP6_POST_ROUTING, skb,NULL, skb->dev,ip6_output_finish);
+       return NF_HOOK(PF_INET6, NF_INET_POST_ROUTING, skb, NULL, skb->dev,
+                      ip6_output_finish);
+}
+
+static inline int ip6_skb_dst_mtu(struct sk_buff *skb)
+{
+       struct ipv6_pinfo *np = skb->sk ? inet6_sk(skb->sk) : NULL;
+
+       return (np && np->pmtudisc == IPV6_PMTUDISC_PROBE) ?
+              skb->dst->dev->mtu : dst_mtu(skb->dst);
 }
 
 int ip6_output(struct sk_buff *skb)
 {
-       if ((skb->len > dst_mtu(skb->dst) && !skb_is_gso(skb)) ||
+       if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
                                dst_allfrag(skb->dst))
                return ip6_fragment(skb, ip6_output2);
        else
@@ -163,7 +197,7 @@ int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
        u32 mtu;
 
        if (opt) {
-               int head_room;
+               unsigned int head_room;
 
                /* First: exthdrs may take lots of space (~8K for now)
                   MAX_HEADER is not enough.
@@ -228,7 +262,7 @@ int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
        if ((skb->len <= mtu) || ipfragok || skb_is_gso(skb)) {
                IP6_INC_STATS(ip6_dst_idev(skb->dst),
                              IPSTATS_MIB_OUTREQUESTS);
-               return NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, dst->dev,
+               return NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, dst->dev,
                                dst_output);
        }
 
@@ -263,7 +297,8 @@ int ip6_nd_hdr(struct sock *sk, struct sk_buff *skb, struct net_device *dev,
 
        totlen = len + sizeof(struct ipv6hdr);
 
-       skb->nh.raw = skb_put(skb, sizeof(struct ipv6hdr));
+       skb_reset_network_header(skb);
+       skb_put(skb, sizeof(struct ipv6hdr));
        hdr = ipv6_hdr(skb);
 
        *(__be32*)hdr = htonl(0x60000000);
@@ -377,7 +412,7 @@ int ip6_forward(struct sk_buff *skb)
                goto drop;
        }
 
-       skb->ip_summed = CHECKSUM_NONE;
+       skb_forward_csum(skb);
 
        /*
         *      We DO NOT make any processing on
@@ -432,8 +467,10 @@ int ip6_forward(struct sk_buff *skb)
 
        /* IPv6 specs say nothing about it, but it is clear that we cannot
           send redirects to source routed frames.
+          We don't send redirects to frames decapsulated from IPsec.
         */
-       if (skb->dev == dst->dev && dst->neighbour && opt->srcrt == 0) {
+       if (skb->dev == dst->dev && dst->neighbour && opt->srcrt == 0 &&
+           !skb->sp) {
                struct in6_addr *target = NULL;
                struct rt6_info *rt;
                struct neighbour *n = dst->neighbour;
@@ -454,10 +491,17 @@ int ip6_forward(struct sk_buff *skb)
                 */
                if (xrlim_allow(dst, 1*HZ))
                        ndisc_send_redirect(skb, n, target);
-       } else if (ipv6_addr_type(&hdr->saddr)&(IPV6_ADDR_MULTICAST|IPV6_ADDR_LOOPBACK
-                                               |IPV6_ADDR_LINKLOCAL)) {
+       } else {
+               int addrtype = ipv6_addr_type(&hdr->saddr);
+
                /* This check is security critical. */
-               goto error;
+               if (addrtype & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LOOPBACK))
+                       goto error;
+               if (addrtype & IPV6_ADDR_LINKLOCAL) {
+                       icmpv6_send(skb, ICMPV6_DEST_UNREACH,
+                               ICMPV6_NOT_NEIGHBOUR, 0, skb->dev);
+                       goto error;
+               }
        }
 
        if (skb->len > dst_mtu(dst)) {
@@ -482,7 +526,8 @@ int ip6_forward(struct sk_buff *skb)
        hdr->hop_limit--;
 
        IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_OUTFORWDATAGRAMS);
-       return NF_HOOK(PF_INET6,NF_IP6_FORWARD, skb, skb->dev, dst->dev, ip6_forward_finish);
+       return NF_HOOK(PF_INET6, NF_INET_FORWARD, skb, skb->dev, dst->dev,
+                      ip6_forward_finish);
 
 error:
        IP6_INC_STATS_BH(ip6_dst_idev(dst), IPSTATS_MIB_INADDRERRORS);
@@ -504,22 +549,10 @@ static void ip6_copy_metadata(struct sk_buff *to, struct sk_buff *from)
 #ifdef CONFIG_NET_SCHED
        to->tc_index = from->tc_index;
 #endif
-#ifdef CONFIG_NETFILTER
-       /* Connection association is same as pre-frag packet */
-       nf_conntrack_put(to->nfct);
-       to->nfct = from->nfct;
-       nf_conntrack_get(to->nfct);
-       to->nfctinfo = from->nfctinfo;
-#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
-       nf_conntrack_put_reasm(to->nfct_reasm);
-       to->nfct_reasm = from->nfct_reasm;
-       nf_conntrack_get_reasm(to->nfct_reasm);
-#endif
-#ifdef CONFIG_BRIDGE_NETFILTER
-       nf_bridge_put(to->nf_bridge);
-       to->nf_bridge = from->nf_bridge;
-       nf_bridge_get(to->nf_bridge);
-#endif
+       nf_copy(to, from);
+#if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
+    defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
+       to->nf_trace = from->nf_trace;
 #endif
        skb_copy_secmark(to, from);
 }
@@ -529,7 +562,7 @@ int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr)
        u16 offset = sizeof(struct ipv6hdr);
        struct ipv6_opt_hdr *exthdr =
                                (struct ipv6_opt_hdr *)(ipv6_hdr(skb) + 1);
-       unsigned int packet_len = skb->tail - skb_network_header(skb);
+       unsigned int packet_len = skb->tail - skb->network_header;
        int found_rhdr = 0;
        *nexthdr = &ipv6_hdr(skb)->nexthdr;
 
@@ -543,7 +576,7 @@ int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr)
                        found_rhdr = 1;
                        break;
                case NEXTHDR_DEST:
-#ifdef CONFIG_IPV6_MIP6
+#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
                        if (ipv6_find_tlv(skb, offset, IPV6_TLV_HAO) >= 0)
                                break;
 #endif
@@ -581,7 +614,20 @@ static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
        hlen = ip6_find_1stfragopt(skb, &prevhdr);
        nexthdr = *prevhdr;
 
-       mtu = dst_mtu(&rt->u.dst);
+       mtu = ip6_skb_dst_mtu(skb);
+
+       /* We must not fragment if the socket is set to force MTU discovery
+        * or if the skb it not generated by a local socket.  (This last
+        * check should be redundant, but it's free.)
+        */
+       if (!np || np->pmtudisc >= IPV6_PMTUDISC_DO) {
+               skb->dev = skb->dst->dev;
+               icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, skb->dev);
+               IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_FRAGFAILS);
+               kfree_skb(skb);
+               return -EMSGSIZE;
+       }
+
        if (np && np->frag_size < mtu) {
                if (np->frag_size)
                        mtu = np->frag_size;
@@ -654,7 +700,7 @@ static int ip6_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
                         * before previous one went down. */
                        if (frag) {
                                frag->ip_summed = CHECKSUM_NONE;
-                               frag->h.raw = frag->data;
+                               skb_reset_transport_header(frag);
                                fh = (struct frag_hdr*)__skb_push(frag, sizeof(struct frag_hdr));
                                __skb_push(frag, hlen);
                                skb_reset_network_header(frag);
@@ -747,8 +793,9 @@ slow_path:
                skb_reserve(frag, LL_RESERVED_SPACE(rt->u.dst.dev));
                skb_put(frag, len + hlen + sizeof(struct frag_hdr));
                skb_reset_network_header(frag);
-               fh = (struct frag_hdr*)(frag->data + hlen);
-               frag->h.raw = frag->data + hlen + sizeof(struct frag_hdr);
+               fh = (struct frag_hdr *)(skb_network_header(frag) + hlen);
+               frag->transport_header = (frag->network_header + hlen +
+                                         sizeof(struct frag_hdr));
 
                /*
                 *      Charge the memory for the fragment to any owner
@@ -760,7 +807,7 @@ slow_path:
                /*
                 *      Copy the packet header into the new buffer.
                 */
-               memcpy(skb_network_header(frag), skb->data, hlen);
+               skb_copy_from_linear_data(skb, skb_network_header(frag), hlen);
 
                /*
                 *      Build fragment header.
@@ -776,7 +823,7 @@ slow_path:
                /*
                 *      Copy a block of the IP datagram.
                 */
-               if (skb_copy_bits(skb, ptr, frag->h.raw, len))
+               if (skb_copy_bits(skb, ptr, skb_transport_header(frag), len))
                        BUG();
                left -= len;
 
@@ -913,6 +960,8 @@ static int ip6_dst_lookup_tail(struct sock *sk,
        return 0;
 
 out_err_release:
+       if (err == -ENETUNREACH)
+               IP6_INC_STATS_BH(NULL, IPSTATS_MIB_OUTNOROUTES);
        dst_release(*dst);
        *dst = NULL;
        return err;
@@ -991,7 +1040,7 @@ static inline int ip6_ufo_append_data(struct sock *sk,
                skb_reset_network_header(skb);
 
                /* initialize protocol header pointer */
-               skb->h.raw = skb->data + fragheaderlen;
+               skb->transport_header = skb->network_header + fragheaderlen;
 
                skb->ip_summed = CHECKSUM_PARTIAL;
                skb->csum = 0;
@@ -1064,7 +1113,8 @@ int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
                inet->cork.fl = *fl;
                np->cork.hop_limit = hlimit;
                np->cork.tclass = tclass;
-               mtu = dst_mtu(rt->u.dst.path);
+               mtu = np->pmtudisc == IPV6_PMTUDISC_PROBE ?
+                     rt->u.dst.dev->mtu : dst_mtu(rt->u.dst.path);
                if (np->frag_size < mtu) {
                        if (np->frag_size)
                                mtu = np->frag_size;
@@ -1075,7 +1125,8 @@ int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
                inet->cork.length = 0;
                sk->sk_sndmsg_page = NULL;
                sk->sk_sndmsg_off = 0;
-               exthdrlen = rt->u.dst.header_len + (opt ? opt->opt_flen : 0);
+               exthdrlen = rt->u.dst.header_len + (opt ? opt->opt_flen : 0) -
+                           rt->nfheader_len;
                length += exthdrlen;
                transhdrlen += exthdrlen;
        } else {
@@ -1090,7 +1141,8 @@ int ip6_append_data(struct sock *sk, int getfrag(void *from, char *to,
 
        hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
 
-       fragheaderlen = sizeof(struct ipv6hdr) + rt->u.dst.nfheader_len + (opt ? opt->opt_nflen : 0);
+       fragheaderlen = sizeof(struct ipv6hdr) + rt->nfheader_len +
+                       (opt ? opt->opt_nflen : 0);
        maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen - sizeof(struct frag_hdr);
 
        if (mtu <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN) {
@@ -1213,8 +1265,8 @@ alloc_new_skb:
                        data = skb_put(skb, fraglen);
                        skb_set_network_header(skb, exthdrlen);
                        data += fragheaderlen;
-                       skb->h.raw = skb->nh.raw + fragheaderlen;
-
+                       skb->transport_header = (skb->network_header +
+                                                fragheaderlen);
                        if (fraggap) {
                                skb->csum = skb_copy_and_csum_bits(
                                        skb_prev, maxfraglen,
@@ -1293,8 +1345,6 @@ alloc_new_skb:
 
                                skb_fill_page_desc(skb, i, page, 0, 0);
                                frag = &skb_shinfo(skb)->frags[i];
-                               skb->truesize += PAGE_SIZE;
-                               atomic_add(PAGE_SIZE, &sk->sk_wmem_alloc);
                        } else {
                                err = -EMSGSIZE;
                                goto error;
@@ -1307,6 +1357,8 @@ alloc_new_skb:
                        frag->size += copy;
                        skb->len += copy;
                        skb->data_len += copy;
+                       skb->truesize += copy;
+                       atomic_add(copy, &sk->sk_wmem_alloc);
                }
                offset += copy;
                length -= copy;
@@ -1318,6 +1370,19 @@ error:
        return err;
 }
 
+static void ip6_cork_release(struct inet_sock *inet, struct ipv6_pinfo *np)
+{
+       inet->cork.flags &= ~IPCORK_OPT;
+       kfree(np->cork.opt);
+       np->cork.opt = NULL;
+       if (np->cork.rt) {
+               dst_release(&np->cork.rt->u.dst);
+               np->cork.rt = NULL;
+               inet->cork.flags &= ~IPCORK_ALLFRAG;
+       }
+       memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
+}
+
 int ip6_push_pending_frames(struct sock *sk)
 {
        struct sk_buff *skb, *tmp_skb;
@@ -1340,7 +1405,7 @@ int ip6_push_pending_frames(struct sock *sk)
        if (skb->data < skb_network_header(skb))
                __skb_pull(skb, skb_network_offset(skb));
        while ((tmp_skb = __skb_dequeue(&sk->sk_write_queue)) != NULL) {
-               __skb_pull(tmp_skb, skb->h.raw - skb->nh.raw);
+               __skb_pull(tmp_skb, skb_network_header_len(skb));
                *tail_skb = tmp_skb;
                tail_skb = &(tmp_skb->next);
                skb->len += tmp_skb->len;
@@ -1352,7 +1417,7 @@ int ip6_push_pending_frames(struct sock *sk)
        }
 
        ipv6_addr_copy(final_dst, &fl->fl6_dst);
-       __skb_pull(skb, skb->h.raw - skb->nh.raw);
+       __skb_pull(skb, skb_network_header_len(skb));
        if (opt && opt->opt_flen)
                ipv6_push_frag_opts(skb, opt, &proto);
        if (opt && opt->opt_nflen)
@@ -1365,10 +1430,6 @@ int ip6_push_pending_frames(struct sock *sk)
        *(__be32*)hdr = fl->fl6_flowlabel |
                     htonl(0x60000000 | ((int)np->cork.tclass << 20));
 
-       if (skb->len <= sizeof(struct ipv6hdr) + IPV6_MAXPLEN)
-               hdr->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
-       else
-               hdr->payload_len = 0;
        hdr->hop_limit = np->cork.hop_limit;
        hdr->nexthdr = proto;
        ipv6_addr_copy(&hdr->saddr, &fl->fl6_src);
@@ -1378,7 +1439,14 @@ int ip6_push_pending_frames(struct sock *sk)
 
        skb->dst = dst_clone(&rt->u.dst);
        IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
-       err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, skb->dst->dev, dst_output);
+       if (proto == IPPROTO_ICMPV6) {
+               struct inet6_dev *idev = ip6_dst_idev(skb->dst);
+
+               ICMP6MSGOUT_INC_STATS_BH(idev, icmp6_hdr(skb)->icmp6_type);
+               ICMP6_INC_STATS_BH(idev, ICMP6_MIB_OUTMSGS);
+       }
+
+       err = ip6_local_out(skb);
        if (err) {
                if (err > 0)
                        err = np->recverr ? net_xmit_errno(err) : 0;
@@ -1387,15 +1455,7 @@ int ip6_push_pending_frames(struct sock *sk)
        }
 
 out:
-       inet->cork.flags &= ~IPCORK_OPT;
-       kfree(np->cork.opt);
-       np->cork.opt = NULL;
-       if (np->cork.rt) {
-               dst_release(&np->cork.rt->u.dst);
-               np->cork.rt = NULL;
-               inet->cork.flags &= ~IPCORK_ALLFRAG;
-       }
-       memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
+       ip6_cork_release(inet, np);
        return err;
 error:
        goto out;
@@ -1403,24 +1463,14 @@ error:
 
 void ip6_flush_pending_frames(struct sock *sk)
 {
-       struct inet_sock *inet = inet_sk(sk);
-       struct ipv6_pinfo *np = inet6_sk(sk);
        struct sk_buff *skb;
 
        while ((skb = __skb_dequeue_tail(&sk->sk_write_queue)) != NULL) {
-               IP6_INC_STATS(ip6_dst_idev(skb->dst),
-                             IPSTATS_MIB_OUTDISCARDS);
+               if (skb->dst)
+                       IP6_INC_STATS(ip6_dst_idev(skb->dst),
+                                     IPSTATS_MIB_OUTDISCARDS);
                kfree_skb(skb);
        }
 
-       inet->cork.flags &= ~IPCORK_OPT;
-
-       kfree(np->cork.opt);
-       np->cork.opt = NULL;
-       if (np->cork.rt) {
-               dst_release(&np->cork.rt->u.dst);
-               np->cork.rt = NULL;
-               inet->cork.flags &= ~IPCORK_ALLFRAG;
-       }
-       memset(&inet->cork.fl, 0, sizeof(inet->cork.fl));
+       ip6_cork_release(inet_sk(sk), inet6_sk(sk));
 }