Use sk_mark for IPv6 routing lookups
[linux-2.6.git] / net / ipv6 / inet6_connection_sock.c
1 /*
2  * INET        An implementation of the TCP/IP protocol suite for the LINUX
3  *             operating system.  INET is implemented using the  BSD Socket
4  *             interface as the means of communication with the user level.
5  *
6  *             Support for INET6 connection oriented protocols.
7  *
8  * Authors:    See the TCPv6 sources
9  *
10  *             This program is free software; you can redistribute it and/or
11  *             modify it under the terms of the GNU General Public License
12  *             as published by the Free Software Foundation; either version
13  *             2 of the License, or(at your option) any later version.
14  */
15
16 #include <linux/module.h>
17 #include <linux/in6.h>
18 #include <linux/ipv6.h>
19 #include <linux/jhash.h>
20
21 #include <net/addrconf.h>
22 #include <net/inet_connection_sock.h>
23 #include <net/inet_ecn.h>
24 #include <net/inet_hashtables.h>
25 #include <net/ip6_route.h>
26 #include <net/sock.h>
27 #include <net/inet6_connection_sock.h>
28
29 int inet6_csk_bind_conflict(const struct sock *sk,
30                             const struct inet_bind_bucket *tb)
31 {
32         const struct sock *sk2;
33         const struct hlist_node *node;
34
35         /* We must walk the whole port owner list in this case. -DaveM */
36         /*
37          * See comment in inet_csk_bind_conflict about sock lookup
38          * vs net namespaces issues.
39          */
40         sk_for_each_bound(sk2, node, &tb->owners) {
41                 if (sk != sk2 &&
42                     (!sk->sk_bound_dev_if ||
43                      !sk2->sk_bound_dev_if ||
44                      sk->sk_bound_dev_if == sk2->sk_bound_dev_if) &&
45                     (!sk->sk_reuse || !sk2->sk_reuse ||
46                      sk2->sk_state == TCP_LISTEN) &&
47                      ipv6_rcv_saddr_equal(sk, sk2))
48                         break;
49         }
50
51         return node != NULL;
52 }
53
54 EXPORT_SYMBOL_GPL(inet6_csk_bind_conflict);
55
56 /*
57  * request_sock (formerly open request) hash tables.
58  */
59 static u32 inet6_synq_hash(const struct in6_addr *raddr, const __be16 rport,
60                            const u32 rnd, const u16 synq_hsize)
61 {
62         u32 a = (__force u32)raddr->s6_addr32[0];
63         u32 b = (__force u32)raddr->s6_addr32[1];
64         u32 c = (__force u32)raddr->s6_addr32[2];
65
66         a += JHASH_GOLDEN_RATIO;
67         b += JHASH_GOLDEN_RATIO;
68         c += rnd;
69         __jhash_mix(a, b, c);
70
71         a += (__force u32)raddr->s6_addr32[3];
72         b += (__force u32)rport;
73         __jhash_mix(a, b, c);
74
75         return c & (synq_hsize - 1);
76 }
77
78 struct request_sock *inet6_csk_search_req(const struct sock *sk,
79                                           struct request_sock ***prevp,
80                                           const __be16 rport,
81                                           const struct in6_addr *raddr,
82                                           const struct in6_addr *laddr,
83                                           const int iif)
84 {
85         const struct inet_connection_sock *icsk = inet_csk(sk);
86         struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt;
87         struct request_sock *req, **prev;
88
89         for (prev = &lopt->syn_table[inet6_synq_hash(raddr, rport,
90                                                      lopt->hash_rnd,
91                                                      lopt->nr_table_entries)];
92              (req = *prev) != NULL;
93              prev = &req->dl_next) {
94                 const struct inet6_request_sock *treq = inet6_rsk(req);
95
96                 if (inet_rsk(req)->rmt_port == rport &&
97                     req->rsk_ops->family == AF_INET6 &&
98                     ipv6_addr_equal(&treq->rmt_addr, raddr) &&
99                     ipv6_addr_equal(&treq->loc_addr, laddr) &&
100                     (!treq->iif || treq->iif == iif)) {
101                         WARN_ON(req->sk != NULL);
102                         *prevp = prev;
103                         return req;
104                 }
105         }
106
107         return NULL;
108 }
109
110 EXPORT_SYMBOL_GPL(inet6_csk_search_req);
111
112 void inet6_csk_reqsk_queue_hash_add(struct sock *sk,
113                                     struct request_sock *req,
114                                     const unsigned long timeout)
115 {
116         struct inet_connection_sock *icsk = inet_csk(sk);
117         struct listen_sock *lopt = icsk->icsk_accept_queue.listen_opt;
118         const u32 h = inet6_synq_hash(&inet6_rsk(req)->rmt_addr,
119                                       inet_rsk(req)->rmt_port,
120                                       lopt->hash_rnd, lopt->nr_table_entries);
121
122         reqsk_queue_hash_req(&icsk->icsk_accept_queue, h, req, timeout);
123         inet_csk_reqsk_queue_added(sk, timeout);
124 }
125
126 EXPORT_SYMBOL_GPL(inet6_csk_reqsk_queue_hash_add);
127
128 void inet6_csk_addr2sockaddr(struct sock *sk, struct sockaddr * uaddr)
129 {
130         struct ipv6_pinfo *np = inet6_sk(sk);
131         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) uaddr;
132
133         sin6->sin6_family = AF_INET6;
134         ipv6_addr_copy(&sin6->sin6_addr, &np->daddr);
135         sin6->sin6_port = inet_sk(sk)->dport;
136         /* We do not store received flowlabel for TCP */
137         sin6->sin6_flowinfo = 0;
138         sin6->sin6_scope_id = 0;
139         if (sk->sk_bound_dev_if &&
140             ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
141                 sin6->sin6_scope_id = sk->sk_bound_dev_if;
142 }
143
144 EXPORT_SYMBOL_GPL(inet6_csk_addr2sockaddr);
145
146 static inline
147 void __inet6_csk_dst_store(struct sock *sk, struct dst_entry *dst,
148                            struct in6_addr *daddr, struct in6_addr *saddr)
149 {
150         __ip6_dst_store(sk, dst, daddr, saddr);
151
152 #ifdef CONFIG_XFRM
153         {
154                 struct rt6_info *rt = (struct rt6_info  *)dst;
155                 rt->rt6i_flow_cache_genid = atomic_read(&flow_cache_genid);
156         }
157 #endif
158 }
159
160 static inline
161 struct dst_entry *__inet6_csk_dst_check(struct sock *sk, u32 cookie)
162 {
163         struct dst_entry *dst;
164
165         dst = __sk_dst_check(sk, cookie);
166
167 #ifdef CONFIG_XFRM
168         if (dst) {
169                 struct rt6_info *rt = (struct rt6_info *)dst;
170                 if (rt->rt6i_flow_cache_genid != atomic_read(&flow_cache_genid)) {
171                         sk->sk_dst_cache = NULL;
172                         dst_release(dst);
173                         dst = NULL;
174                 }
175         }
176 #endif
177
178         return dst;
179 }
180
181 int inet6_csk_xmit(struct sk_buff *skb, int ipfragok)
182 {
183         struct sock *sk = skb->sk;
184         struct inet_sock *inet = inet_sk(sk);
185         struct ipv6_pinfo *np = inet6_sk(sk);
186         struct flowi fl;
187         struct dst_entry *dst;
188         struct in6_addr *final_p = NULL, final;
189
190         memset(&fl, 0, sizeof(fl));
191         fl.proto = sk->sk_protocol;
192         ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
193         ipv6_addr_copy(&fl.fl6_src, &np->saddr);
194         fl.fl6_flowlabel = np->flow_label;
195         IP6_ECN_flow_xmit(sk, fl.fl6_flowlabel);
196         fl.oif = sk->sk_bound_dev_if;
197         fl.mark = sk->sk_mark;
198         fl.fl_ip_sport = inet->sport;
199         fl.fl_ip_dport = inet->dport;
200         security_sk_classify_flow(sk, &fl);
201
202         if (np->opt && np->opt->srcrt) {
203                 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
204                 ipv6_addr_copy(&final, &fl.fl6_dst);
205                 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
206                 final_p = &final;
207         }
208
209         dst = __inet6_csk_dst_check(sk, np->dst_cookie);
210
211         if (dst == NULL) {
212                 int err = ip6_dst_lookup(sk, &dst, &fl);
213
214                 if (err) {
215                         sk->sk_err_soft = -err;
216                         kfree_skb(skb);
217                         return err;
218                 }
219
220                 if (final_p)
221                         ipv6_addr_copy(&fl.fl6_dst, final_p);
222
223                 if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
224                         sk->sk_route_caps = 0;
225                         kfree_skb(skb);
226                         return err;
227                 }
228
229                 __inet6_csk_dst_store(sk, dst, NULL, NULL);
230         }
231
232         skb_dst_set(skb, dst_clone(dst));
233
234         /* Restore final destination back after routing done */
235         ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
236
237         return ip6_xmit(sk, skb, &fl, np->opt, 0);
238 }
239
240 EXPORT_SYMBOL_GPL(inet6_csk_xmit);