Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * TCP over IPv6 | |
1ab1457c | 3 | * Linux INET6 implementation |
1da177e4 LT |
4 | * |
5 | * Authors: | |
1ab1457c | 6 | * Pedro Roque <roque@di.fc.ul.pt> |
1da177e4 | 7 | * |
1ab1457c | 8 | * Based on: |
1da177e4 LT |
9 | * linux/net/ipv4/tcp.c |
10 | * linux/net/ipv4/tcp_input.c | |
11 | * linux/net/ipv4/tcp_output.c | |
12 | * | |
13 | * Fixes: | |
14 | * Hideaki YOSHIFUJI : sin6_scope_id support | |
15 | * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which | |
16 | * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind | |
17 | * a single port at the same time. | |
18 | * YOSHIFUJI Hideaki @USAGI: convert /proc/net/tcp6 to seq_file. | |
19 | * | |
20 | * This program is free software; you can redistribute it and/or | |
21 | * modify it under the terms of the GNU General Public License | |
22 | * as published by the Free Software Foundation; either version | |
23 | * 2 of the License, or (at your option) any later version. | |
24 | */ | |
25 | ||
eb4dea58 | 26 | #include <linux/bottom_half.h> |
1da177e4 | 27 | #include <linux/module.h> |
1da177e4 LT |
28 | #include <linux/errno.h> |
29 | #include <linux/types.h> | |
30 | #include <linux/socket.h> | |
31 | #include <linux/sockios.h> | |
32 | #include <linux/net.h> | |
33 | #include <linux/jiffies.h> | |
34 | #include <linux/in.h> | |
35 | #include <linux/in6.h> | |
36 | #include <linux/netdevice.h> | |
37 | #include <linux/init.h> | |
38 | #include <linux/jhash.h> | |
39 | #include <linux/ipsec.h> | |
40 | #include <linux/times.h> | |
5a0e3ad6 | 41 | #include <linux/slab.h> |
1da177e4 LT |
42 | |
43 | #include <linux/ipv6.h> | |
44 | #include <linux/icmpv6.h> | |
45 | #include <linux/random.h> | |
46 | ||
47 | #include <net/tcp.h> | |
48 | #include <net/ndisc.h> | |
5324a040 | 49 | #include <net/inet6_hashtables.h> |
8129765a | 50 | #include <net/inet6_connection_sock.h> |
1da177e4 LT |
51 | #include <net/ipv6.h> |
52 | #include <net/transp_v6.h> | |
53 | #include <net/addrconf.h> | |
54 | #include <net/ip6_route.h> | |
55 | #include <net/ip6_checksum.h> | |
56 | #include <net/inet_ecn.h> | |
57 | #include <net/protocol.h> | |
58 | #include <net/xfrm.h> | |
1da177e4 LT |
59 | #include <net/snmp.h> |
60 | #include <net/dsfield.h> | |
6d6ee43e | 61 | #include <net/timewait_sock.h> |
18134bed | 62 | #include <net/netdma.h> |
3d58b5fa | 63 | #include <net/inet_common.h> |
6e5714ea | 64 | #include <net/secure_seq.h> |
d1a4c0b3 | 65 | #include <net/tcp_memcontrol.h> |
1da177e4 LT |
66 | |
67 | #include <asm/uaccess.h> | |
68 | ||
69 | #include <linux/proc_fs.h> | |
70 | #include <linux/seq_file.h> | |
71 | ||
cfb6eeb4 YH |
72 | #include <linux/crypto.h> |
73 | #include <linux/scatterlist.h> | |
74 | ||
cfb6eeb4 | 75 | static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb); |
6edafaaf GJ |
76 | static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
77 | struct request_sock *req); | |
1da177e4 LT |
78 | |
79 | static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb); | |
8ad50d96 | 80 | static void __tcp_v6_send_check(struct sk_buff *skb, |
b71d1d42 ED |
81 | const struct in6_addr *saddr, |
82 | const struct in6_addr *daddr); | |
1da177e4 | 83 | |
3b401a81 SH |
84 | static const struct inet_connection_sock_af_ops ipv6_mapped; |
85 | static const struct inet_connection_sock_af_ops ipv6_specific; | |
a928630a | 86 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de SH |
87 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific; |
88 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific; | |
9501f972 YH |
89 | #else |
90 | static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk, | |
b71d1d42 | 91 | const struct in6_addr *addr) |
9501f972 YH |
92 | { |
93 | return NULL; | |
94 | } | |
a928630a | 95 | #endif |
1da177e4 | 96 | |
1da177e4 LT |
97 | static void tcp_v6_hash(struct sock *sk) |
98 | { | |
99 | if (sk->sk_state != TCP_CLOSE) { | |
8292a17a | 100 | if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) { |
1da177e4 LT |
101 | tcp_prot.hash(sk); |
102 | return; | |
103 | } | |
104 | local_bh_disable(); | |
9327f705 | 105 | __inet6_hash(sk, NULL); |
1da177e4 LT |
106 | local_bh_enable(); |
107 | } | |
108 | } | |
109 | ||
684f2176 | 110 | static __inline__ __sum16 tcp_v6_check(int len, |
b71d1d42 ED |
111 | const struct in6_addr *saddr, |
112 | const struct in6_addr *daddr, | |
868c86bc | 113 | __wsum base) |
1da177e4 LT |
114 | { |
115 | return csum_ipv6_magic(saddr, daddr, len, IPPROTO_TCP, base); | |
116 | } | |
117 | ||
cf533ea5 | 118 | static __u32 tcp_v6_init_sequence(const struct sk_buff *skb) |
1da177e4 | 119 | { |
0660e03f ACM |
120 | return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32, |
121 | ipv6_hdr(skb)->saddr.s6_addr32, | |
aa8223c7 ACM |
122 | tcp_hdr(skb)->dest, |
123 | tcp_hdr(skb)->source); | |
1da177e4 LT |
124 | } |
125 | ||
1ab1457c | 126 | static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, |
1da177e4 LT |
127 | int addr_len) |
128 | { | |
129 | struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr; | |
1ab1457c | 130 | struct inet_sock *inet = inet_sk(sk); |
d83d8461 | 131 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 LT |
132 | struct ipv6_pinfo *np = inet6_sk(sk); |
133 | struct tcp_sock *tp = tcp_sk(sk); | |
20c59de2 | 134 | struct in6_addr *saddr = NULL, *final_p, final; |
493f377d | 135 | struct rt6_info *rt; |
4c9483b2 | 136 | struct flowi6 fl6; |
1da177e4 LT |
137 | struct dst_entry *dst; |
138 | int addr_type; | |
139 | int err; | |
140 | ||
1ab1457c | 141 | if (addr_len < SIN6_LEN_RFC2133) |
1da177e4 LT |
142 | return -EINVAL; |
143 | ||
1ab1457c | 144 | if (usin->sin6_family != AF_INET6) |
a02cec21 | 145 | return -EAFNOSUPPORT; |
1da177e4 | 146 | |
4c9483b2 | 147 | memset(&fl6, 0, sizeof(fl6)); |
1da177e4 LT |
148 | |
149 | if (np->sndflow) { | |
4c9483b2 DM |
150 | fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK; |
151 | IP6_ECN_flow_init(fl6.flowlabel); | |
152 | if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { | |
1da177e4 | 153 | struct ip6_flowlabel *flowlabel; |
4c9483b2 | 154 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); |
1da177e4 LT |
155 | if (flowlabel == NULL) |
156 | return -EINVAL; | |
4e3fd7a0 | 157 | usin->sin6_addr = flowlabel->dst; |
1da177e4 LT |
158 | fl6_sock_release(flowlabel); |
159 | } | |
160 | } | |
161 | ||
162 | /* | |
1ab1457c YH |
163 | * connect() to INADDR_ANY means loopback (BSD'ism). |
164 | */ | |
165 | ||
166 | if(ipv6_addr_any(&usin->sin6_addr)) | |
167 | usin->sin6_addr.s6_addr[15] = 0x1; | |
1da177e4 LT |
168 | |
169 | addr_type = ipv6_addr_type(&usin->sin6_addr); | |
170 | ||
171 | if(addr_type & IPV6_ADDR_MULTICAST) | |
172 | return -ENETUNREACH; | |
173 | ||
174 | if (addr_type&IPV6_ADDR_LINKLOCAL) { | |
175 | if (addr_len >= sizeof(struct sockaddr_in6) && | |
176 | usin->sin6_scope_id) { | |
177 | /* If interface is set while binding, indices | |
178 | * must coincide. | |
179 | */ | |
180 | if (sk->sk_bound_dev_if && | |
181 | sk->sk_bound_dev_if != usin->sin6_scope_id) | |
182 | return -EINVAL; | |
183 | ||
184 | sk->sk_bound_dev_if = usin->sin6_scope_id; | |
185 | } | |
186 | ||
187 | /* Connect to link-local address requires an interface */ | |
188 | if (!sk->sk_bound_dev_if) | |
189 | return -EINVAL; | |
190 | } | |
191 | ||
192 | if (tp->rx_opt.ts_recent_stamp && | |
193 | !ipv6_addr_equal(&np->daddr, &usin->sin6_addr)) { | |
194 | tp->rx_opt.ts_recent = 0; | |
195 | tp->rx_opt.ts_recent_stamp = 0; | |
196 | tp->write_seq = 0; | |
197 | } | |
198 | ||
4e3fd7a0 | 199 | np->daddr = usin->sin6_addr; |
4c9483b2 | 200 | np->flow_label = fl6.flowlabel; |
1da177e4 LT |
201 | |
202 | /* | |
203 | * TCP over IPv4 | |
204 | */ | |
205 | ||
206 | if (addr_type == IPV6_ADDR_MAPPED) { | |
d83d8461 | 207 | u32 exthdrlen = icsk->icsk_ext_hdr_len; |
1da177e4 LT |
208 | struct sockaddr_in sin; |
209 | ||
210 | SOCK_DEBUG(sk, "connect: ipv4 mapped\n"); | |
211 | ||
212 | if (__ipv6_only_sock(sk)) | |
213 | return -ENETUNREACH; | |
214 | ||
215 | sin.sin_family = AF_INET; | |
216 | sin.sin_port = usin->sin6_port; | |
217 | sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3]; | |
218 | ||
d83d8461 | 219 | icsk->icsk_af_ops = &ipv6_mapped; |
1da177e4 | 220 | sk->sk_backlog_rcv = tcp_v4_do_rcv; |
cfb6eeb4 YH |
221 | #ifdef CONFIG_TCP_MD5SIG |
222 | tp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
223 | #endif | |
1da177e4 LT |
224 | |
225 | err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin)); | |
226 | ||
227 | if (err) { | |
d83d8461 ACM |
228 | icsk->icsk_ext_hdr_len = exthdrlen; |
229 | icsk->icsk_af_ops = &ipv6_specific; | |
1da177e4 | 230 | sk->sk_backlog_rcv = tcp_v6_do_rcv; |
cfb6eeb4 YH |
231 | #ifdef CONFIG_TCP_MD5SIG |
232 | tp->af_specific = &tcp_sock_ipv6_specific; | |
233 | #endif | |
1da177e4 LT |
234 | goto failure; |
235 | } else { | |
c720c7e8 ED |
236 | ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr); |
237 | ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, | |
238 | &np->rcv_saddr); | |
1da177e4 LT |
239 | } |
240 | ||
241 | return err; | |
242 | } | |
243 | ||
244 | if (!ipv6_addr_any(&np->rcv_saddr)) | |
245 | saddr = &np->rcv_saddr; | |
246 | ||
4c9483b2 | 247 | fl6.flowi6_proto = IPPROTO_TCP; |
4e3fd7a0 AD |
248 | fl6.daddr = np->daddr; |
249 | fl6.saddr = saddr ? *saddr : np->saddr; | |
4c9483b2 DM |
250 | fl6.flowi6_oif = sk->sk_bound_dev_if; |
251 | fl6.flowi6_mark = sk->sk_mark; | |
1958b856 DM |
252 | fl6.fl6_dport = usin->sin6_port; |
253 | fl6.fl6_sport = inet->inet_sport; | |
1da177e4 | 254 | |
4c9483b2 | 255 | final_p = fl6_update_dst(&fl6, np->opt, &final); |
1da177e4 | 256 | |
4c9483b2 | 257 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
beb8d13b | 258 | |
4c9483b2 | 259 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true); |
68d0c6d3 DM |
260 | if (IS_ERR(dst)) { |
261 | err = PTR_ERR(dst); | |
1da177e4 | 262 | goto failure; |
14e50e57 | 263 | } |
1da177e4 LT |
264 | |
265 | if (saddr == NULL) { | |
4c9483b2 | 266 | saddr = &fl6.saddr; |
4e3fd7a0 | 267 | np->rcv_saddr = *saddr; |
1da177e4 LT |
268 | } |
269 | ||
270 | /* set the source address */ | |
4e3fd7a0 | 271 | np->saddr = *saddr; |
c720c7e8 | 272 | inet->inet_rcv_saddr = LOOPBACK4_IPV6; |
1da177e4 | 273 | |
f83ef8c0 | 274 | sk->sk_gso_type = SKB_GSO_TCPV6; |
8e1ef0a9 | 275 | __ip6_dst_store(sk, dst, NULL, NULL); |
1da177e4 | 276 | |
493f377d DM |
277 | rt = (struct rt6_info *) dst; |
278 | if (tcp_death_row.sysctl_tw_recycle && | |
279 | !tp->rx_opt.ts_recent_stamp && | |
280 | ipv6_addr_equal(&rt->rt6i_dst.addr, &np->daddr)) { | |
281 | struct inet_peer *peer = rt6_get_peer(rt); | |
282 | /* | |
283 | * VJ's idea. We save last timestamp seen from | |
284 | * the destination in peer table, when entering state | |
285 | * TIME-WAIT * and initialize rx_opt.ts_recent from it, | |
286 | * when trying new connection. | |
287 | */ | |
288 | if (peer) { | |
289 | inet_peer_refcheck(peer); | |
290 | if ((u32)get_seconds() - peer->tcp_ts_stamp <= TCP_PAWS_MSL) { | |
291 | tp->rx_opt.ts_recent_stamp = peer->tcp_ts_stamp; | |
292 | tp->rx_opt.ts_recent = peer->tcp_ts; | |
293 | } | |
294 | } | |
295 | } | |
296 | ||
d83d8461 | 297 | icsk->icsk_ext_hdr_len = 0; |
1da177e4 | 298 | if (np->opt) |
d83d8461 ACM |
299 | icsk->icsk_ext_hdr_len = (np->opt->opt_flen + |
300 | np->opt->opt_nflen); | |
1da177e4 LT |
301 | |
302 | tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); | |
303 | ||
c720c7e8 | 304 | inet->inet_dport = usin->sin6_port; |
1da177e4 LT |
305 | |
306 | tcp_set_state(sk, TCP_SYN_SENT); | |
d8313f5c | 307 | err = inet6_hash_connect(&tcp_death_row, sk); |
1da177e4 LT |
308 | if (err) |
309 | goto late_failure; | |
310 | ||
311 | if (!tp->write_seq) | |
312 | tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32, | |
313 | np->daddr.s6_addr32, | |
c720c7e8 ED |
314 | inet->inet_sport, |
315 | inet->inet_dport); | |
1da177e4 LT |
316 | |
317 | err = tcp_connect(sk); | |
318 | if (err) | |
319 | goto late_failure; | |
320 | ||
321 | return 0; | |
322 | ||
323 | late_failure: | |
324 | tcp_set_state(sk, TCP_CLOSE); | |
325 | __sk_dst_reset(sk); | |
326 | failure: | |
c720c7e8 | 327 | inet->inet_dport = 0; |
1da177e4 LT |
328 | sk->sk_route_caps = 0; |
329 | return err; | |
330 | } | |
331 | ||
332 | static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, | |
d5fdd6ba | 333 | u8 type, u8 code, int offset, __be32 info) |
1da177e4 | 334 | { |
b71d1d42 | 335 | const struct ipv6hdr *hdr = (const struct ipv6hdr*)skb->data; |
505cbfc5 | 336 | const struct tcphdr *th = (struct tcphdr *)(skb->data+offset); |
1da177e4 LT |
337 | struct ipv6_pinfo *np; |
338 | struct sock *sk; | |
339 | int err; | |
1ab1457c | 340 | struct tcp_sock *tp; |
1da177e4 | 341 | __u32 seq; |
ca12a1a4 | 342 | struct net *net = dev_net(skb->dev); |
1da177e4 | 343 | |
ca12a1a4 | 344 | sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr, |
d86e0dac | 345 | th->dest, &hdr->saddr, th->source, skb->dev->ifindex); |
1da177e4 LT |
346 | |
347 | if (sk == NULL) { | |
e41b5368 DL |
348 | ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev), |
349 | ICMP6_MIB_INERRORS); | |
1da177e4 LT |
350 | return; |
351 | } | |
352 | ||
353 | if (sk->sk_state == TCP_TIME_WAIT) { | |
9469c7b4 | 354 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
355 | return; |
356 | } | |
357 | ||
358 | bh_lock_sock(sk); | |
359 | if (sock_owned_by_user(sk)) | |
de0744af | 360 | NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS); |
1da177e4 LT |
361 | |
362 | if (sk->sk_state == TCP_CLOSE) | |
363 | goto out; | |
364 | ||
e802af9c SH |
365 | if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) { |
366 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
367 | goto out; | |
368 | } | |
369 | ||
1da177e4 | 370 | tp = tcp_sk(sk); |
1ab1457c | 371 | seq = ntohl(th->seq); |
1da177e4 LT |
372 | if (sk->sk_state != TCP_LISTEN && |
373 | !between(seq, tp->snd_una, tp->snd_nxt)) { | |
de0744af | 374 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
375 | goto out; |
376 | } | |
377 | ||
378 | np = inet6_sk(sk); | |
379 | ||
380 | if (type == ICMPV6_PKT_TOOBIG) { | |
68d0c6d3 | 381 | struct dst_entry *dst; |
1da177e4 LT |
382 | |
383 | if (sock_owned_by_user(sk)) | |
384 | goto out; | |
385 | if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) | |
386 | goto out; | |
387 | ||
388 | /* icmp should have updated the destination cache entry */ | |
389 | dst = __sk_dst_check(sk, np->dst_cookie); | |
390 | ||
391 | if (dst == NULL) { | |
392 | struct inet_sock *inet = inet_sk(sk); | |
4c9483b2 | 393 | struct flowi6 fl6; |
1da177e4 LT |
394 | |
395 | /* BUGGG_FUTURE: Again, it is not clear how | |
396 | to handle rthdr case. Ignore this complexity | |
397 | for now. | |
398 | */ | |
4c9483b2 DM |
399 | memset(&fl6, 0, sizeof(fl6)); |
400 | fl6.flowi6_proto = IPPROTO_TCP; | |
4e3fd7a0 AD |
401 | fl6.daddr = np->daddr; |
402 | fl6.saddr = np->saddr; | |
4c9483b2 DM |
403 | fl6.flowi6_oif = sk->sk_bound_dev_if; |
404 | fl6.flowi6_mark = sk->sk_mark; | |
1958b856 DM |
405 | fl6.fl6_dport = inet->inet_dport; |
406 | fl6.fl6_sport = inet->inet_sport; | |
4c9483b2 DM |
407 | security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); |
408 | ||
409 | dst = ip6_dst_lookup_flow(sk, &fl6, NULL, false); | |
68d0c6d3 DM |
410 | if (IS_ERR(dst)) { |
411 | sk->sk_err_soft = -PTR_ERR(dst); | |
1da177e4 LT |
412 | goto out; |
413 | } | |
414 | ||
415 | } else | |
416 | dst_hold(dst); | |
417 | ||
81aded24 DM |
418 | dst->ops->update_pmtu(dst, ntohl(info)); |
419 | ||
d83d8461 | 420 | if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) { |
1da177e4 LT |
421 | tcp_sync_mss(sk, dst_mtu(dst)); |
422 | tcp_simple_retransmit(sk); | |
423 | } /* else let the usual retransmit timer handle it */ | |
424 | dst_release(dst); | |
425 | goto out; | |
426 | } | |
427 | ||
428 | icmpv6_err_convert(type, code, &err); | |
429 | ||
60236fdd | 430 | /* Might be for an request_sock */ |
1da177e4 | 431 | switch (sk->sk_state) { |
60236fdd | 432 | struct request_sock *req, **prev; |
1da177e4 LT |
433 | case TCP_LISTEN: |
434 | if (sock_owned_by_user(sk)) | |
435 | goto out; | |
436 | ||
8129765a ACM |
437 | req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr, |
438 | &hdr->saddr, inet6_iif(skb)); | |
1da177e4 LT |
439 | if (!req) |
440 | goto out; | |
441 | ||
442 | /* ICMPs are not backlogged, hence we cannot get | |
443 | * an established socket here. | |
444 | */ | |
547b792c | 445 | WARN_ON(req->sk != NULL); |
1da177e4 | 446 | |
2e6599cb | 447 | if (seq != tcp_rsk(req)->snt_isn) { |
de0744af | 448 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
449 | goto out; |
450 | } | |
451 | ||
463c84b9 | 452 | inet_csk_reqsk_queue_drop(sk, req, prev); |
1da177e4 LT |
453 | goto out; |
454 | ||
455 | case TCP_SYN_SENT: | |
456 | case TCP_SYN_RECV: /* Cannot happen. | |
1ab1457c | 457 | It can, it SYNs are crossed. --ANK */ |
1da177e4 | 458 | if (!sock_owned_by_user(sk)) { |
1da177e4 LT |
459 | sk->sk_err = err; |
460 | sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */ | |
461 | ||
462 | tcp_done(sk); | |
463 | } else | |
464 | sk->sk_err_soft = err; | |
465 | goto out; | |
466 | } | |
467 | ||
468 | if (!sock_owned_by_user(sk) && np->recverr) { | |
469 | sk->sk_err = err; | |
470 | sk->sk_error_report(sk); | |
471 | } else | |
472 | sk->sk_err_soft = err; | |
473 | ||
474 | out: | |
475 | bh_unlock_sock(sk); | |
476 | sock_put(sk); | |
477 | } | |
478 | ||
479 | ||
3840a06e NC |
480 | static int tcp_v6_send_synack(struct sock *sk, |
481 | struct request_sock *req, | |
fff32699 ED |
482 | struct request_values *rvp, |
483 | u16 queue_mapping) | |
1da177e4 | 484 | { |
ca304b61 | 485 | struct inet6_request_sock *treq = inet6_rsk(req); |
1da177e4 LT |
486 | struct ipv6_pinfo *np = inet6_sk(sk); |
487 | struct sk_buff * skb; | |
9494218f | 488 | struct ipv6_txoptions *opt = np->opt; |
4c9483b2 | 489 | struct flowi6 fl6; |
fd80eb94 | 490 | struct dst_entry *dst; |
9494218f | 491 | int err = -ENOMEM; |
1da177e4 | 492 | |
9494218f NC |
493 | dst = inet6_csk_route_req(sk, &fl6, req); |
494 | if (!dst) | |
fd80eb94 | 495 | goto done; |
9494218f | 496 | |
e6b4d113 | 497 | skb = tcp_make_synack(sk, dst, req, rvp); |
9494218f | 498 | |
1da177e4 | 499 | if (skb) { |
8ad50d96 | 500 | __tcp_v6_send_check(skb, &treq->loc_addr, &treq->rmt_addr); |
1da177e4 | 501 | |
4e3fd7a0 | 502 | fl6.daddr = treq->rmt_addr; |
fff32699 | 503 | skb_set_queue_mapping(skb, queue_mapping); |
b903d324 | 504 | err = ip6_xmit(sk, skb, &fl6, opt, np->tclass); |
b9df3cb8 | 505 | err = net_xmit_eval(err); |
1da177e4 LT |
506 | } |
507 | ||
508 | done: | |
1ab1457c | 509 | if (opt && opt != np->opt) |
1da177e4 LT |
510 | sock_kfree_s(sk, opt, opt->tot_len); |
511 | return err; | |
512 | } | |
513 | ||
72659ecc OP |
514 | static int tcp_v6_rtx_synack(struct sock *sk, struct request_sock *req, |
515 | struct request_values *rvp) | |
516 | { | |
517 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS); | |
fff32699 | 518 | return tcp_v6_send_synack(sk, req, rvp, 0); |
72659ecc OP |
519 | } |
520 | ||
60236fdd | 521 | static void tcp_v6_reqsk_destructor(struct request_sock *req) |
1da177e4 | 522 | { |
800d55f1 | 523 | kfree_skb(inet6_rsk(req)->pktopts); |
1da177e4 LT |
524 | } |
525 | ||
cfb6eeb4 YH |
526 | #ifdef CONFIG_TCP_MD5SIG |
527 | static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk, | |
b71d1d42 | 528 | const struct in6_addr *addr) |
cfb6eeb4 | 529 | { |
a915da9b | 530 | return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6); |
cfb6eeb4 YH |
531 | } |
532 | ||
533 | static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk, | |
534 | struct sock *addr_sk) | |
535 | { | |
536 | return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr); | |
537 | } | |
538 | ||
539 | static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk, | |
540 | struct request_sock *req) | |
541 | { | |
542 | return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr); | |
543 | } | |
544 | ||
cfb6eeb4 YH |
545 | static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval, |
546 | int optlen) | |
547 | { | |
548 | struct tcp_md5sig cmd; | |
549 | struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr; | |
cfb6eeb4 YH |
550 | |
551 | if (optlen < sizeof(cmd)) | |
552 | return -EINVAL; | |
553 | ||
554 | if (copy_from_user(&cmd, optval, sizeof(cmd))) | |
555 | return -EFAULT; | |
556 | ||
557 | if (sin6->sin6_family != AF_INET6) | |
558 | return -EINVAL; | |
559 | ||
560 | if (!cmd.tcpm_keylen) { | |
e773e4fa | 561 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) |
a915da9b ED |
562 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3], |
563 | AF_INET); | |
564 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr, | |
565 | AF_INET6); | |
cfb6eeb4 YH |
566 | } |
567 | ||
568 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
569 | return -EINVAL; | |
570 | ||
a915da9b ED |
571 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) |
572 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3], | |
573 | AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL); | |
cfb6eeb4 | 574 | |
a915da9b ED |
575 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr, |
576 | AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL); | |
cfb6eeb4 YH |
577 | } |
578 | ||
49a72dfb | 579 | static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp, |
b71d1d42 ED |
580 | const struct in6_addr *daddr, |
581 | const struct in6_addr *saddr, int nbytes) | |
cfb6eeb4 | 582 | { |
cfb6eeb4 | 583 | struct tcp6_pseudohdr *bp; |
49a72dfb | 584 | struct scatterlist sg; |
8d26d76d | 585 | |
cfb6eeb4 | 586 | bp = &hp->md5_blk.ip6; |
cfb6eeb4 | 587 | /* 1. TCP pseudo-header (RFC2460) */ |
4e3fd7a0 AD |
588 | bp->saddr = *saddr; |
589 | bp->daddr = *daddr; | |
49a72dfb | 590 | bp->protocol = cpu_to_be32(IPPROTO_TCP); |
00b1304c | 591 | bp->len = cpu_to_be32(nbytes); |
cfb6eeb4 | 592 | |
49a72dfb AL |
593 | sg_init_one(&sg, bp, sizeof(*bp)); |
594 | return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp)); | |
595 | } | |
c7da57a1 | 596 | |
49a72dfb | 597 | static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key, |
b71d1d42 | 598 | const struct in6_addr *daddr, struct in6_addr *saddr, |
318cf7aa | 599 | const struct tcphdr *th) |
49a72dfb AL |
600 | { |
601 | struct tcp_md5sig_pool *hp; | |
602 | struct hash_desc *desc; | |
603 | ||
604 | hp = tcp_get_md5sig_pool(); | |
605 | if (!hp) | |
606 | goto clear_hash_noput; | |
607 | desc = &hp->md5_desc; | |
608 | ||
609 | if (crypto_hash_init(desc)) | |
610 | goto clear_hash; | |
611 | if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2)) | |
612 | goto clear_hash; | |
613 | if (tcp_md5_hash_header(hp, th)) | |
614 | goto clear_hash; | |
615 | if (tcp_md5_hash_key(hp, key)) | |
616 | goto clear_hash; | |
617 | if (crypto_hash_final(desc, md5_hash)) | |
cfb6eeb4 | 618 | goto clear_hash; |
cfb6eeb4 | 619 | |
cfb6eeb4 | 620 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 621 | return 0; |
49a72dfb | 622 | |
cfb6eeb4 YH |
623 | clear_hash: |
624 | tcp_put_md5sig_pool(); | |
625 | clear_hash_noput: | |
626 | memset(md5_hash, 0, 16); | |
49a72dfb | 627 | return 1; |
cfb6eeb4 YH |
628 | } |
629 | ||
49a72dfb | 630 | static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key, |
318cf7aa ED |
631 | const struct sock *sk, |
632 | const struct request_sock *req, | |
633 | const struct sk_buff *skb) | |
cfb6eeb4 | 634 | { |
b71d1d42 | 635 | const struct in6_addr *saddr, *daddr; |
49a72dfb AL |
636 | struct tcp_md5sig_pool *hp; |
637 | struct hash_desc *desc; | |
318cf7aa | 638 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
639 | |
640 | if (sk) { | |
641 | saddr = &inet6_sk(sk)->saddr; | |
642 | daddr = &inet6_sk(sk)->daddr; | |
49a72dfb | 643 | } else if (req) { |
cfb6eeb4 YH |
644 | saddr = &inet6_rsk(req)->loc_addr; |
645 | daddr = &inet6_rsk(req)->rmt_addr; | |
49a72dfb | 646 | } else { |
b71d1d42 | 647 | const struct ipv6hdr *ip6h = ipv6_hdr(skb); |
49a72dfb AL |
648 | saddr = &ip6h->saddr; |
649 | daddr = &ip6h->daddr; | |
cfb6eeb4 | 650 | } |
49a72dfb AL |
651 | |
652 | hp = tcp_get_md5sig_pool(); | |
653 | if (!hp) | |
654 | goto clear_hash_noput; | |
655 | desc = &hp->md5_desc; | |
656 | ||
657 | if (crypto_hash_init(desc)) | |
658 | goto clear_hash; | |
659 | ||
660 | if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len)) | |
661 | goto clear_hash; | |
662 | if (tcp_md5_hash_header(hp, th)) | |
663 | goto clear_hash; | |
664 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
665 | goto clear_hash; | |
666 | if (tcp_md5_hash_key(hp, key)) | |
667 | goto clear_hash; | |
668 | if (crypto_hash_final(desc, md5_hash)) | |
669 | goto clear_hash; | |
670 | ||
671 | tcp_put_md5sig_pool(); | |
672 | return 0; | |
673 | ||
674 | clear_hash: | |
675 | tcp_put_md5sig_pool(); | |
676 | clear_hash_noput: | |
677 | memset(md5_hash, 0, 16); | |
678 | return 1; | |
cfb6eeb4 YH |
679 | } |
680 | ||
318cf7aa | 681 | static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb) |
cfb6eeb4 | 682 | { |
cf533ea5 | 683 | const __u8 *hash_location = NULL; |
cfb6eeb4 | 684 | struct tcp_md5sig_key *hash_expected; |
b71d1d42 | 685 | const struct ipv6hdr *ip6h = ipv6_hdr(skb); |
318cf7aa | 686 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 687 | int genhash; |
cfb6eeb4 YH |
688 | u8 newhash[16]; |
689 | ||
690 | hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr); | |
7d5d5525 | 691 | hash_location = tcp_parse_md5sig_option(th); |
cfb6eeb4 | 692 | |
785957d3 DM |
693 | /* We've parsed the options - do we have a hash? */ |
694 | if (!hash_expected && !hash_location) | |
695 | return 0; | |
696 | ||
697 | if (hash_expected && !hash_location) { | |
698 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); | |
cfb6eeb4 YH |
699 | return 1; |
700 | } | |
701 | ||
785957d3 DM |
702 | if (!hash_expected && hash_location) { |
703 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); | |
cfb6eeb4 YH |
704 | return 1; |
705 | } | |
706 | ||
707 | /* check the signature */ | |
49a72dfb AL |
708 | genhash = tcp_v6_md5_hash_skb(newhash, |
709 | hash_expected, | |
710 | NULL, NULL, skb); | |
711 | ||
cfb6eeb4 | 712 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { |
e87cc472 JP |
713 | net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n", |
714 | genhash ? "failed" : "mismatch", | |
715 | &ip6h->saddr, ntohs(th->source), | |
716 | &ip6h->daddr, ntohs(th->dest)); | |
cfb6eeb4 YH |
717 | return 1; |
718 | } | |
719 | return 0; | |
720 | } | |
721 | #endif | |
722 | ||
c6aefafb | 723 | struct request_sock_ops tcp6_request_sock_ops __read_mostly = { |
1da177e4 | 724 | .family = AF_INET6, |
2e6599cb | 725 | .obj_size = sizeof(struct tcp6_request_sock), |
72659ecc | 726 | .rtx_syn_ack = tcp_v6_rtx_synack, |
60236fdd ACM |
727 | .send_ack = tcp_v6_reqsk_send_ack, |
728 | .destructor = tcp_v6_reqsk_destructor, | |
72659ecc OP |
729 | .send_reset = tcp_v6_send_reset, |
730 | .syn_ack_timeout = tcp_syn_ack_timeout, | |
1da177e4 LT |
731 | }; |
732 | ||
cfb6eeb4 | 733 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 734 | static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = { |
cfb6eeb4 | 735 | .md5_lookup = tcp_v6_reqsk_md5_lookup, |
e3afe7b7 | 736 | .calc_md5_hash = tcp_v6_md5_hash_skb, |
cfb6eeb4 | 737 | }; |
b6332e6c | 738 | #endif |
cfb6eeb4 | 739 | |
8ad50d96 | 740 | static void __tcp_v6_send_check(struct sk_buff *skb, |
b71d1d42 | 741 | const struct in6_addr *saddr, const struct in6_addr *daddr) |
1da177e4 | 742 | { |
aa8223c7 | 743 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 744 | |
84fa7933 | 745 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
8ad50d96 | 746 | th->check = ~tcp_v6_check(skb->len, saddr, daddr, 0); |
663ead3b | 747 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 748 | skb->csum_offset = offsetof(struct tcphdr, check); |
1da177e4 | 749 | } else { |
8ad50d96 HX |
750 | th->check = tcp_v6_check(skb->len, saddr, daddr, |
751 | csum_partial(th, th->doff << 2, | |
752 | skb->csum)); | |
1da177e4 LT |
753 | } |
754 | } | |
755 | ||
bb296246 | 756 | static void tcp_v6_send_check(struct sock *sk, struct sk_buff *skb) |
8ad50d96 HX |
757 | { |
758 | struct ipv6_pinfo *np = inet6_sk(sk); | |
759 | ||
760 | __tcp_v6_send_check(skb, &np->saddr, &np->daddr); | |
761 | } | |
762 | ||
a430a43d HX |
763 | static int tcp_v6_gso_send_check(struct sk_buff *skb) |
764 | { | |
b71d1d42 | 765 | const struct ipv6hdr *ipv6h; |
a430a43d HX |
766 | struct tcphdr *th; |
767 | ||
768 | if (!pskb_may_pull(skb, sizeof(*th))) | |
769 | return -EINVAL; | |
770 | ||
0660e03f | 771 | ipv6h = ipv6_hdr(skb); |
aa8223c7 | 772 | th = tcp_hdr(skb); |
a430a43d HX |
773 | |
774 | th->check = 0; | |
84fa7933 | 775 | skb->ip_summed = CHECKSUM_PARTIAL; |
8ad50d96 | 776 | __tcp_v6_send_check(skb, &ipv6h->saddr, &ipv6h->daddr); |
a430a43d HX |
777 | return 0; |
778 | } | |
1da177e4 | 779 | |
36990673 HX |
780 | static struct sk_buff **tcp6_gro_receive(struct sk_buff **head, |
781 | struct sk_buff *skb) | |
684f2176 | 782 | { |
b71d1d42 | 783 | const struct ipv6hdr *iph = skb_gro_network_header(skb); |
684f2176 HX |
784 | |
785 | switch (skb->ip_summed) { | |
786 | case CHECKSUM_COMPLETE: | |
86911732 | 787 | if (!tcp_v6_check(skb_gro_len(skb), &iph->saddr, &iph->daddr, |
684f2176 HX |
788 | skb->csum)) { |
789 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
790 | break; | |
791 | } | |
792 | ||
793 | /* fall through */ | |
794 | case CHECKSUM_NONE: | |
795 | NAPI_GRO_CB(skb)->flush = 1; | |
796 | return NULL; | |
797 | } | |
798 | ||
799 | return tcp_gro_receive(head, skb); | |
800 | } | |
684f2176 | 801 | |
36990673 | 802 | static int tcp6_gro_complete(struct sk_buff *skb) |
684f2176 | 803 | { |
b71d1d42 | 804 | const struct ipv6hdr *iph = ipv6_hdr(skb); |
684f2176 HX |
805 | struct tcphdr *th = tcp_hdr(skb); |
806 | ||
807 | th->check = ~tcp_v6_check(skb->len - skb_transport_offset(skb), | |
808 | &iph->saddr, &iph->daddr, 0); | |
809 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6; | |
810 | ||
811 | return tcp_gro_complete(skb); | |
812 | } | |
684f2176 | 813 | |
626e264d | 814 | static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win, |
b903d324 | 815 | u32 ts, struct tcp_md5sig_key *key, int rst, u8 tclass) |
1da177e4 | 816 | { |
cf533ea5 ED |
817 | const struct tcphdr *th = tcp_hdr(skb); |
818 | struct tcphdr *t1; | |
1da177e4 | 819 | struct sk_buff *buff; |
4c9483b2 | 820 | struct flowi6 fl6; |
adf30907 | 821 | struct net *net = dev_net(skb_dst(skb)->dev); |
e5047992 | 822 | struct sock *ctl_sk = net->ipv6.tcp_sk; |
77c676da | 823 | unsigned int tot_len = sizeof(struct tcphdr); |
adf30907 | 824 | struct dst_entry *dst; |
81ada62d | 825 | __be32 *topt; |
1da177e4 | 826 | |
626e264d IJ |
827 | if (ts) |
828 | tot_len += TCPOLEN_TSTAMP_ALIGNED; | |
cfb6eeb4 | 829 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 YH |
830 | if (key) |
831 | tot_len += TCPOLEN_MD5SIG_ALIGNED; | |
832 | #endif | |
833 | ||
cfb6eeb4 | 834 | buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len, |
1da177e4 | 835 | GFP_ATOMIC); |
1ab1457c YH |
836 | if (buff == NULL) |
837 | return; | |
1da177e4 | 838 | |
cfb6eeb4 | 839 | skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len); |
1da177e4 | 840 | |
cfb6eeb4 | 841 | t1 = (struct tcphdr *) skb_push(buff, tot_len); |
6651ffc8 | 842 | skb_reset_transport_header(buff); |
1da177e4 LT |
843 | |
844 | /* Swap the send and the receive. */ | |
845 | memset(t1, 0, sizeof(*t1)); | |
846 | t1->dest = th->source; | |
847 | t1->source = th->dest; | |
cfb6eeb4 | 848 | t1->doff = tot_len / 4; |
626e264d IJ |
849 | t1->seq = htonl(seq); |
850 | t1->ack_seq = htonl(ack); | |
851 | t1->ack = !rst || !th->ack; | |
852 | t1->rst = rst; | |
853 | t1->window = htons(win); | |
1da177e4 | 854 | |
81ada62d IJ |
855 | topt = (__be32 *)(t1 + 1); |
856 | ||
626e264d IJ |
857 | if (ts) { |
858 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | | |
859 | (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP); | |
860 | *topt++ = htonl(tcp_time_stamp); | |
861 | *topt++ = htonl(ts); | |
862 | } | |
863 | ||
cfb6eeb4 YH |
864 | #ifdef CONFIG_TCP_MD5SIG |
865 | if (key) { | |
81ada62d IJ |
866 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
867 | (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG); | |
868 | tcp_v6_md5_hash_hdr((__u8 *)topt, key, | |
78e645cb IJ |
869 | &ipv6_hdr(skb)->saddr, |
870 | &ipv6_hdr(skb)->daddr, t1); | |
cfb6eeb4 YH |
871 | } |
872 | #endif | |
873 | ||
4c9483b2 | 874 | memset(&fl6, 0, sizeof(fl6)); |
4e3fd7a0 AD |
875 | fl6.daddr = ipv6_hdr(skb)->saddr; |
876 | fl6.saddr = ipv6_hdr(skb)->daddr; | |
1da177e4 | 877 | |
e5700aff DM |
878 | buff->ip_summed = CHECKSUM_PARTIAL; |
879 | buff->csum = 0; | |
880 | ||
4c9483b2 | 881 | __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr); |
1da177e4 | 882 | |
4c9483b2 DM |
883 | fl6.flowi6_proto = IPPROTO_TCP; |
884 | fl6.flowi6_oif = inet6_iif(skb); | |
1958b856 DM |
885 | fl6.fl6_dport = t1->dest; |
886 | fl6.fl6_sport = t1->source; | |
4c9483b2 | 887 | security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); |
1da177e4 | 888 | |
c20121ae DL |
889 | /* Pass a socket to ip6_dst_lookup either it is for RST |
890 | * Underlying function will use this to retrieve the network | |
891 | * namespace | |
892 | */ | |
4c9483b2 | 893 | dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL, false); |
68d0c6d3 DM |
894 | if (!IS_ERR(dst)) { |
895 | skb_dst_set(buff, dst); | |
b903d324 | 896 | ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass); |
68d0c6d3 DM |
897 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
898 | if (rst) | |
899 | TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS); | |
900 | return; | |
1da177e4 LT |
901 | } |
902 | ||
903 | kfree_skb(buff); | |
904 | } | |
905 | ||
626e264d | 906 | static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 907 | { |
cf533ea5 | 908 | const struct tcphdr *th = tcp_hdr(skb); |
626e264d | 909 | u32 seq = 0, ack_seq = 0; |
fa3e5b4e | 910 | struct tcp_md5sig_key *key = NULL; |
658ddaaf SL |
911 | #ifdef CONFIG_TCP_MD5SIG |
912 | const __u8 *hash_location = NULL; | |
913 | struct ipv6hdr *ipv6h = ipv6_hdr(skb); | |
914 | unsigned char newhash[16]; | |
915 | int genhash; | |
916 | struct sock *sk1 = NULL; | |
917 | #endif | |
1da177e4 | 918 | |
626e264d | 919 | if (th->rst) |
1da177e4 LT |
920 | return; |
921 | ||
626e264d IJ |
922 | if (!ipv6_unicast_destination(skb)) |
923 | return; | |
1da177e4 | 924 | |
cfb6eeb4 | 925 | #ifdef CONFIG_TCP_MD5SIG |
658ddaaf SL |
926 | hash_location = tcp_parse_md5sig_option(th); |
927 | if (!sk && hash_location) { | |
928 | /* | |
929 | * active side is lost. Try to find listening socket through | |
930 | * source port, and then find md5 key through listening socket. | |
931 | * we are not loose security here: | |
932 | * Incoming packet is checked with md5 hash with finding key, | |
933 | * no RST generated if md5 hash doesn't match. | |
934 | */ | |
935 | sk1 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev), | |
936 | &tcp_hashinfo, &ipv6h->daddr, | |
937 | ntohs(th->source), inet6_iif(skb)); | |
938 | if (!sk1) | |
939 | return; | |
940 | ||
941 | rcu_read_lock(); | |
942 | key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr); | |
943 | if (!key) | |
944 | goto release_sk1; | |
945 | ||
946 | genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, NULL, skb); | |
947 | if (genhash || memcmp(hash_location, newhash, 16) != 0) | |
948 | goto release_sk1; | |
949 | } else { | |
950 | key = sk ? tcp_v6_md5_do_lookup(sk, &ipv6h->saddr) : NULL; | |
951 | } | |
cfb6eeb4 YH |
952 | #endif |
953 | ||
626e264d IJ |
954 | if (th->ack) |
955 | seq = ntohl(th->ack_seq); | |
956 | else | |
957 | ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len - | |
958 | (th->doff << 2); | |
1da177e4 | 959 | |
b903d324 | 960 | tcp_v6_send_response(skb, seq, ack_seq, 0, 0, key, 1, 0); |
658ddaaf SL |
961 | |
962 | #ifdef CONFIG_TCP_MD5SIG | |
963 | release_sk1: | |
964 | if (sk1) { | |
965 | rcu_read_unlock(); | |
966 | sock_put(sk1); | |
967 | } | |
968 | #endif | |
626e264d | 969 | } |
1da177e4 | 970 | |
626e264d | 971 | static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts, |
b903d324 | 972 | struct tcp_md5sig_key *key, u8 tclass) |
626e264d | 973 | { |
b903d324 | 974 | tcp_v6_send_response(skb, seq, ack, win, ts, key, 0, tclass); |
1da177e4 LT |
975 | } |
976 | ||
977 | static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
978 | { | |
8feaf0c0 | 979 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 980 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 981 | |
9501f972 | 982 | tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
8feaf0c0 | 983 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
b903d324 ED |
984 | tcptw->tw_ts_recent, tcp_twsk_md5_key(tcptw), |
985 | tw->tw_tclass); | |
1da177e4 | 986 | |
8feaf0c0 | 987 | inet_twsk_put(tw); |
1da177e4 LT |
988 | } |
989 | ||
6edafaaf GJ |
990 | static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
991 | struct request_sock *req) | |
1da177e4 | 992 | { |
9501f972 | 993 | tcp_v6_send_ack(skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, req->ts_recent, |
b903d324 | 994 | tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr), 0); |
1da177e4 LT |
995 | } |
996 | ||
997 | ||
998 | static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb) | |
999 | { | |
60236fdd | 1000 | struct request_sock *req, **prev; |
aa8223c7 | 1001 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 LT |
1002 | struct sock *nsk; |
1003 | ||
1004 | /* Find possible connection requests. */ | |
8129765a | 1005 | req = inet6_csk_search_req(sk, &prev, th->source, |
0660e03f ACM |
1006 | &ipv6_hdr(skb)->saddr, |
1007 | &ipv6_hdr(skb)->daddr, inet6_iif(skb)); | |
1da177e4 LT |
1008 | if (req) |
1009 | return tcp_check_req(sk, skb, req, prev); | |
1010 | ||
3b1e0a65 | 1011 | nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo, |
d86e0dac PE |
1012 | &ipv6_hdr(skb)->saddr, th->source, |
1013 | &ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb)); | |
1da177e4 LT |
1014 | |
1015 | if (nsk) { | |
1016 | if (nsk->sk_state != TCP_TIME_WAIT) { | |
1017 | bh_lock_sock(nsk); | |
1018 | return nsk; | |
1019 | } | |
9469c7b4 | 1020 | inet_twsk_put(inet_twsk(nsk)); |
1da177e4 LT |
1021 | return NULL; |
1022 | } | |
1023 | ||
c6aefafb | 1024 | #ifdef CONFIG_SYN_COOKIES |
af9b4738 | 1025 | if (!th->syn) |
c6aefafb | 1026 | sk = cookie_v6_check(sk, skb); |
1da177e4 LT |
1027 | #endif |
1028 | return sk; | |
1029 | } | |
1030 | ||
1da177e4 LT |
1031 | /* FIXME: this is substantially similar to the ipv4 code. |
1032 | * Can some kind of merge be done? -- erics | |
1033 | */ | |
1034 | static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb) | |
1035 | { | |
4957faad | 1036 | struct tcp_extend_values tmp_ext; |
e6b4d113 | 1037 | struct tcp_options_received tmp_opt; |
cf533ea5 | 1038 | const u8 *hash_location; |
e6b4d113 | 1039 | struct request_sock *req; |
ca304b61 | 1040 | struct inet6_request_sock *treq; |
1da177e4 | 1041 | struct ipv6_pinfo *np = inet6_sk(sk); |
1da177e4 | 1042 | struct tcp_sock *tp = tcp_sk(sk); |
e6b4d113 | 1043 | __u32 isn = TCP_SKB_CB(skb)->when; |
493f377d | 1044 | struct dst_entry *dst = NULL; |
3840a06e | 1045 | struct flowi6 fl6; |
a2a385d6 | 1046 | bool want_cookie = false; |
1da177e4 LT |
1047 | |
1048 | if (skb->protocol == htons(ETH_P_IP)) | |
1049 | return tcp_v4_conn_request(sk, skb); | |
1050 | ||
1051 | if (!ipv6_unicast_destination(skb)) | |
1ab1457c | 1052 | goto drop; |
1da177e4 | 1053 | |
463c84b9 | 1054 | if (inet_csk_reqsk_queue_is_full(sk) && !isn) { |
946cedcc ED |
1055 | want_cookie = tcp_syn_flood_action(sk, skb, "TCPv6"); |
1056 | if (!want_cookie) | |
1057 | goto drop; | |
1da177e4 LT |
1058 | } |
1059 | ||
463c84b9 | 1060 | if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1) |
1da177e4 LT |
1061 | goto drop; |
1062 | ||
ca304b61 | 1063 | req = inet6_reqsk_alloc(&tcp6_request_sock_ops); |
1da177e4 LT |
1064 | if (req == NULL) |
1065 | goto drop; | |
1066 | ||
cfb6eeb4 YH |
1067 | #ifdef CONFIG_TCP_MD5SIG |
1068 | tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops; | |
1069 | #endif | |
1070 | ||
1da177e4 LT |
1071 | tcp_clear_options(&tmp_opt); |
1072 | tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); | |
1073 | tmp_opt.user_mss = tp->rx_opt.user_mss; | |
bb5b7c11 | 1074 | tcp_parse_options(skb, &tmp_opt, &hash_location, 0); |
4957faad WAS |
1075 | |
1076 | if (tmp_opt.cookie_plus > 0 && | |
1077 | tmp_opt.saw_tstamp && | |
1078 | !tp->rx_opt.cookie_out_never && | |
1079 | (sysctl_tcp_cookie_size > 0 || | |
1080 | (tp->cookie_values != NULL && | |
1081 | tp->cookie_values->cookie_desired > 0))) { | |
1082 | u8 *c; | |
1083 | u32 *d; | |
1084 | u32 *mess = &tmp_ext.cookie_bakery[COOKIE_DIGEST_WORDS]; | |
1085 | int l = tmp_opt.cookie_plus - TCPOLEN_COOKIE_BASE; | |
1086 | ||
1087 | if (tcp_cookie_generator(&tmp_ext.cookie_bakery[0]) != 0) | |
1088 | goto drop_and_free; | |
1089 | ||
1090 | /* Secret recipe starts with IP addresses */ | |
0eae88f3 | 1091 | d = (__force u32 *)&ipv6_hdr(skb)->daddr.s6_addr32[0]; |
4957faad WAS |
1092 | *mess++ ^= *d++; |
1093 | *mess++ ^= *d++; | |
1094 | *mess++ ^= *d++; | |
1095 | *mess++ ^= *d++; | |
0eae88f3 | 1096 | d = (__force u32 *)&ipv6_hdr(skb)->saddr.s6_addr32[0]; |
4957faad WAS |
1097 | *mess++ ^= *d++; |
1098 | *mess++ ^= *d++; | |
1099 | *mess++ ^= *d++; | |
1100 | *mess++ ^= *d++; | |
1101 | ||
1102 | /* plus variable length Initiator Cookie */ | |
1103 | c = (u8 *)mess; | |
1104 | while (l-- > 0) | |
1105 | *c++ ^= *hash_location++; | |
1da177e4 | 1106 | |
a2a385d6 | 1107 | want_cookie = false; /* not our kind of cookie */ |
4957faad WAS |
1108 | tmp_ext.cookie_out_never = 0; /* false */ |
1109 | tmp_ext.cookie_plus = tmp_opt.cookie_plus; | |
1110 | } else if (!tp->rx_opt.cookie_in_always) { | |
1111 | /* redundant indications, but ensure initialization. */ | |
1112 | tmp_ext.cookie_out_never = 1; /* true */ | |
1113 | tmp_ext.cookie_plus = 0; | |
1114 | } else { | |
1115 | goto drop_and_free; | |
1116 | } | |
1117 | tmp_ext.cookie_in_always = tp->rx_opt.cookie_in_always; | |
1da177e4 | 1118 | |
4dfc2817 | 1119 | if (want_cookie && !tmp_opt.saw_tstamp) |
c6aefafb | 1120 | tcp_clear_options(&tmp_opt); |
c6aefafb | 1121 | |
1da177e4 LT |
1122 | tmp_opt.tstamp_ok = tmp_opt.saw_tstamp; |
1123 | tcp_openreq_init(req, &tmp_opt, skb); | |
1124 | ||
ca304b61 | 1125 | treq = inet6_rsk(req); |
4e3fd7a0 AD |
1126 | treq->rmt_addr = ipv6_hdr(skb)->saddr; |
1127 | treq->loc_addr = ipv6_hdr(skb)->daddr; | |
172d69e6 | 1128 | if (!want_cookie || tmp_opt.tstamp_ok) |
bd14b1b2 | 1129 | TCP_ECN_create_request(req, skb); |
c6aefafb | 1130 | |
4d0fe50c ED |
1131 | treq->iif = sk->sk_bound_dev_if; |
1132 | ||
1133 | /* So that link locals have meaning */ | |
1134 | if (!sk->sk_bound_dev_if && | |
1135 | ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL) | |
1136 | treq->iif = inet6_iif(skb); | |
1137 | ||
2bbdf389 | 1138 | if (!isn) { |
493f377d DM |
1139 | struct inet_peer *peer = NULL; |
1140 | ||
c6aefafb GG |
1141 | if (ipv6_opt_accepted(sk, skb) || |
1142 | np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo || | |
1143 | np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) { | |
1144 | atomic_inc(&skb->users); | |
1145 | treq->pktopts = skb; | |
1146 | } | |
493f377d DM |
1147 | |
1148 | if (want_cookie) { | |
2bbdf389 FW |
1149 | isn = cookie_v6_init_sequence(sk, skb, &req->mss); |
1150 | req->cookie_ts = tmp_opt.tstamp_ok; | |
493f377d DM |
1151 | goto have_isn; |
1152 | } | |
1153 | ||
1154 | /* VJ's idea. We save last timestamp seen | |
1155 | * from the destination in peer table, when entering | |
1156 | * state TIME-WAIT, and check against it before | |
1157 | * accepting new connection request. | |
1158 | * | |
1159 | * If "isn" is not zero, this request hit alive | |
1160 | * timewait bucket, so that all the necessary checks | |
1161 | * are made in the function processing timewait state. | |
1162 | */ | |
1163 | if (tmp_opt.saw_tstamp && | |
1164 | tcp_death_row.sysctl_tw_recycle && | |
3840a06e | 1165 | (dst = inet6_csk_route_req(sk, &fl6, req)) != NULL && |
493f377d | 1166 | (peer = rt6_get_peer((struct rt6_info *)dst)) != NULL && |
7a71ed89 | 1167 | ipv6_addr_equal((struct in6_addr *)peer->daddr.addr.a6, |
493f377d DM |
1168 | &treq->rmt_addr)) { |
1169 | inet_peer_refcheck(peer); | |
1170 | if ((u32)get_seconds() - peer->tcp_ts_stamp < TCP_PAWS_MSL && | |
1171 | (s32)(peer->tcp_ts - req->ts_recent) > | |
1172 | TCP_PAWS_WINDOW) { | |
1173 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED); | |
1174 | goto drop_and_release; | |
1175 | } | |
1176 | } | |
1177 | /* Kill the following clause, if you dislike this way. */ | |
1178 | else if (!sysctl_tcp_syncookies && | |
1179 | (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) < | |
1180 | (sysctl_max_syn_backlog >> 2)) && | |
1181 | (!peer || !peer->tcp_ts_stamp) && | |
1182 | (!dst || !dst_metric(dst, RTAX_RTT))) { | |
1183 | /* Without syncookies last quarter of | |
1184 | * backlog is filled with destinations, | |
1185 | * proven to be alive. | |
1186 | * It means that we continue to communicate | |
1187 | * to destinations, already remembered | |
1188 | * to the moment of synflood. | |
1189 | */ | |
1190 | LIMIT_NETDEBUG(KERN_DEBUG "TCP: drop open request from %pI6/%u\n", | |
1191 | &treq->rmt_addr, ntohs(tcp_hdr(skb)->source)); | |
1192 | goto drop_and_release; | |
2bbdf389 | 1193 | } |
493f377d DM |
1194 | |
1195 | isn = tcp_v6_init_sequence(skb); | |
c6aefafb | 1196 | } |
493f377d | 1197 | have_isn: |
2e6599cb | 1198 | tcp_rsk(req)->snt_isn = isn; |
9ad7c049 | 1199 | tcp_rsk(req)->snt_synack = tcp_time_stamp; |
1da177e4 | 1200 | |
437c5b53 NC |
1201 | if (security_inet_conn_request(sk, skb, req)) |
1202 | goto drop_and_release; | |
4237c75c | 1203 | |
4957faad | 1204 | if (tcp_v6_send_synack(sk, req, |
fff32699 ED |
1205 | (struct request_values *)&tmp_ext, |
1206 | skb_get_queue_mapping(skb)) || | |
4957faad | 1207 | want_cookie) |
e6b4d113 | 1208 | goto drop_and_free; |
1da177e4 | 1209 | |
e6b4d113 WAS |
1210 | inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT); |
1211 | return 0; | |
1da177e4 | 1212 | |
493f377d DM |
1213 | drop_and_release: |
1214 | dst_release(dst); | |
e6b4d113 WAS |
1215 | drop_and_free: |
1216 | reqsk_free(req); | |
1da177e4 | 1217 | drop: |
1da177e4 LT |
1218 | return 0; /* don't send reset */ |
1219 | } | |
1220 | ||
1221 | static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb, | |
60236fdd | 1222 | struct request_sock *req, |
1da177e4 LT |
1223 | struct dst_entry *dst) |
1224 | { | |
78d15e82 | 1225 | struct inet6_request_sock *treq; |
1da177e4 LT |
1226 | struct ipv6_pinfo *newnp, *np = inet6_sk(sk); |
1227 | struct tcp6_sock *newtcp6sk; | |
1228 | struct inet_sock *newinet; | |
1229 | struct tcp_sock *newtp; | |
1230 | struct sock *newsk; | |
1231 | struct ipv6_txoptions *opt; | |
cfb6eeb4 YH |
1232 | #ifdef CONFIG_TCP_MD5SIG |
1233 | struct tcp_md5sig_key *key; | |
1234 | #endif | |
3840a06e | 1235 | struct flowi6 fl6; |
1da177e4 LT |
1236 | |
1237 | if (skb->protocol == htons(ETH_P_IP)) { | |
1238 | /* | |
1239 | * v6 mapped | |
1240 | */ | |
1241 | ||
1242 | newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst); | |
1243 | ||
1ab1457c | 1244 | if (newsk == NULL) |
1da177e4 LT |
1245 | return NULL; |
1246 | ||
1247 | newtcp6sk = (struct tcp6_sock *)newsk; | |
1248 | inet_sk(newsk)->pinet6 = &newtcp6sk->inet6; | |
1249 | ||
1250 | newinet = inet_sk(newsk); | |
1251 | newnp = inet6_sk(newsk); | |
1252 | newtp = tcp_sk(newsk); | |
1253 | ||
1254 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1255 | ||
c720c7e8 | 1256 | ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr); |
1da177e4 | 1257 | |
c720c7e8 | 1258 | ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr); |
1da177e4 | 1259 | |
4e3fd7a0 | 1260 | newnp->rcv_saddr = newnp->saddr; |
1da177e4 | 1261 | |
8292a17a | 1262 | inet_csk(newsk)->icsk_af_ops = &ipv6_mapped; |
1da177e4 | 1263 | newsk->sk_backlog_rcv = tcp_v4_do_rcv; |
cfb6eeb4 YH |
1264 | #ifdef CONFIG_TCP_MD5SIG |
1265 | newtp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
1266 | #endif | |
1267 | ||
676a1184 YZ |
1268 | newnp->ipv6_ac_list = NULL; |
1269 | newnp->ipv6_fl_list = NULL; | |
1da177e4 LT |
1270 | newnp->pktoptions = NULL; |
1271 | newnp->opt = NULL; | |
505cbfc5 | 1272 | newnp->mcast_oif = inet6_iif(skb); |
0660e03f | 1273 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; |
4c507d28 | 1274 | newnp->rcv_tclass = ipv6_tclass(ipv6_hdr(skb)); |
1da177e4 | 1275 | |
e6848976 ACM |
1276 | /* |
1277 | * No need to charge this sock to the relevant IPv6 refcnt debug socks count | |
1278 | * here, tcp_create_openreq_child now does this for us, see the comment in | |
1279 | * that function for the gory details. -acme | |
1da177e4 | 1280 | */ |
1da177e4 LT |
1281 | |
1282 | /* It is tricky place. Until this moment IPv4 tcp | |
8292a17a | 1283 | worked with IPv6 icsk.icsk_af_ops. |
1da177e4 LT |
1284 | Sync it now. |
1285 | */ | |
d83d8461 | 1286 | tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie); |
1da177e4 LT |
1287 | |
1288 | return newsk; | |
1289 | } | |
1290 | ||
78d15e82 | 1291 | treq = inet6_rsk(req); |
1da177e4 LT |
1292 | opt = np->opt; |
1293 | ||
1294 | if (sk_acceptq_is_full(sk)) | |
1295 | goto out_overflow; | |
1296 | ||
493f377d | 1297 | if (!dst) { |
3840a06e | 1298 | dst = inet6_csk_route_req(sk, &fl6, req); |
493f377d | 1299 | if (!dst) |
1da177e4 | 1300 | goto out; |
1ab1457c | 1301 | } |
1da177e4 LT |
1302 | |
1303 | newsk = tcp_create_openreq_child(sk, req, skb); | |
1304 | if (newsk == NULL) | |
093d2823 | 1305 | goto out_nonewsk; |
1da177e4 | 1306 | |
e6848976 ACM |
1307 | /* |
1308 | * No need to charge this sock to the relevant IPv6 refcnt debug socks | |
1309 | * count here, tcp_create_openreq_child now does this for us, see the | |
1310 | * comment in that function for the gory details. -acme | |
1311 | */ | |
1da177e4 | 1312 | |
59eed279 | 1313 | newsk->sk_gso_type = SKB_GSO_TCPV6; |
8e1ef0a9 | 1314 | __ip6_dst_store(newsk, dst, NULL, NULL); |
1da177e4 LT |
1315 | |
1316 | newtcp6sk = (struct tcp6_sock *)newsk; | |
1317 | inet_sk(newsk)->pinet6 = &newtcp6sk->inet6; | |
1318 | ||
1319 | newtp = tcp_sk(newsk); | |
1320 | newinet = inet_sk(newsk); | |
1321 | newnp = inet6_sk(newsk); | |
1322 | ||
1323 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1324 | ||
4e3fd7a0 AD |
1325 | newnp->daddr = treq->rmt_addr; |
1326 | newnp->saddr = treq->loc_addr; | |
1327 | newnp->rcv_saddr = treq->loc_addr; | |
2e6599cb | 1328 | newsk->sk_bound_dev_if = treq->iif; |
1da177e4 | 1329 | |
1ab1457c | 1330 | /* Now IPv6 options... |
1da177e4 LT |
1331 | |
1332 | First: no IPv4 options. | |
1333 | */ | |
f6d8bd05 | 1334 | newinet->inet_opt = NULL; |
676a1184 | 1335 | newnp->ipv6_ac_list = NULL; |
d35690be | 1336 | newnp->ipv6_fl_list = NULL; |
1da177e4 LT |
1337 | |
1338 | /* Clone RX bits */ | |
1339 | newnp->rxopt.all = np->rxopt.all; | |
1340 | ||
1341 | /* Clone pktoptions received with SYN */ | |
1342 | newnp->pktoptions = NULL; | |
2e6599cb ACM |
1343 | if (treq->pktopts != NULL) { |
1344 | newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC); | |
ab185d7b | 1345 | consume_skb(treq->pktopts); |
2e6599cb | 1346 | treq->pktopts = NULL; |
1da177e4 LT |
1347 | if (newnp->pktoptions) |
1348 | skb_set_owner_r(newnp->pktoptions, newsk); | |
1349 | } | |
1350 | newnp->opt = NULL; | |
505cbfc5 | 1351 | newnp->mcast_oif = inet6_iif(skb); |
0660e03f | 1352 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; |
4c507d28 | 1353 | newnp->rcv_tclass = ipv6_tclass(ipv6_hdr(skb)); |
1da177e4 LT |
1354 | |
1355 | /* Clone native IPv6 options from listening socket (if any) | |
1356 | ||
1357 | Yes, keeping reference count would be much more clever, | |
1358 | but we make one more one thing there: reattach optmem | |
1359 | to newsk. | |
1360 | */ | |
1361 | if (opt) { | |
1362 | newnp->opt = ipv6_dup_options(newsk, opt); | |
1363 | if (opt != np->opt) | |
1364 | sock_kfree_s(sk, opt, opt->tot_len); | |
1365 | } | |
1366 | ||
d83d8461 | 1367 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
1da177e4 | 1368 | if (newnp->opt) |
d83d8461 ACM |
1369 | inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + |
1370 | newnp->opt->opt_flen); | |
1da177e4 | 1371 | |
5d424d5a | 1372 | tcp_mtup_init(newsk); |
1da177e4 | 1373 | tcp_sync_mss(newsk, dst_mtu(dst)); |
0dbaee3b | 1374 | newtp->advmss = dst_metric_advmss(dst); |
d135c522 NC |
1375 | if (tcp_sk(sk)->rx_opt.user_mss && |
1376 | tcp_sk(sk)->rx_opt.user_mss < newtp->advmss) | |
1377 | newtp->advmss = tcp_sk(sk)->rx_opt.user_mss; | |
1378 | ||
1da177e4 | 1379 | tcp_initialize_rcv_mss(newsk); |
9ad7c049 JC |
1380 | if (tcp_rsk(req)->snt_synack) |
1381 | tcp_valid_rtt_meas(newsk, | |
1382 | tcp_time_stamp - tcp_rsk(req)->snt_synack); | |
1383 | newtp->total_retrans = req->retrans; | |
1da177e4 | 1384 | |
c720c7e8 ED |
1385 | newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6; |
1386 | newinet->inet_rcv_saddr = LOOPBACK4_IPV6; | |
1da177e4 | 1387 | |
cfb6eeb4 YH |
1388 | #ifdef CONFIG_TCP_MD5SIG |
1389 | /* Copy over the MD5 key from the original socket */ | |
1390 | if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) { | |
1391 | /* We're using one, so create a matching key | |
1392 | * on the newsk structure. If we fail to get | |
1393 | * memory, then we end up not copying the key | |
1394 | * across. Shucks. | |
1395 | */ | |
a915da9b ED |
1396 | tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newnp->daddr, |
1397 | AF_INET6, key->key, key->keylen, GFP_ATOMIC); | |
cfb6eeb4 YH |
1398 | } |
1399 | #endif | |
1400 | ||
093d2823 BS |
1401 | if (__inet_inherit_port(sk, newsk) < 0) { |
1402 | sock_put(newsk); | |
1403 | goto out; | |
1404 | } | |
9327f705 | 1405 | __inet6_hash(newsk, NULL); |
1da177e4 LT |
1406 | |
1407 | return newsk; | |
1408 | ||
1409 | out_overflow: | |
de0744af | 1410 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 | 1411 | out_nonewsk: |
1da177e4 LT |
1412 | if (opt && opt != np->opt) |
1413 | sock_kfree_s(sk, opt, opt->tot_len); | |
1414 | dst_release(dst); | |
093d2823 BS |
1415 | out: |
1416 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); | |
1da177e4 LT |
1417 | return NULL; |
1418 | } | |
1419 | ||
b51655b9 | 1420 | static __sum16 tcp_v6_checksum_init(struct sk_buff *skb) |
1da177e4 | 1421 | { |
84fa7933 | 1422 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
684f2176 | 1423 | if (!tcp_v6_check(skb->len, &ipv6_hdr(skb)->saddr, |
0660e03f | 1424 | &ipv6_hdr(skb)->daddr, skb->csum)) { |
fb286bb2 | 1425 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
1da177e4 | 1426 | return 0; |
fb286bb2 | 1427 | } |
1da177e4 | 1428 | } |
fb286bb2 | 1429 | |
684f2176 | 1430 | skb->csum = ~csum_unfold(tcp_v6_check(skb->len, |
0660e03f ACM |
1431 | &ipv6_hdr(skb)->saddr, |
1432 | &ipv6_hdr(skb)->daddr, 0)); | |
fb286bb2 | 1433 | |
1da177e4 | 1434 | if (skb->len <= 76) { |
fb286bb2 | 1435 | return __skb_checksum_complete(skb); |
1da177e4 LT |
1436 | } |
1437 | return 0; | |
1438 | } | |
1439 | ||
1440 | /* The socket must have it's spinlock held when we get | |
1441 | * here. | |
1442 | * | |
1443 | * We have a potential double-lock case here, so even when | |
1444 | * doing backlog processing we use the BH locking scheme. | |
1445 | * This is because we cannot sleep with the original spinlock | |
1446 | * held. | |
1447 | */ | |
1448 | static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1449 | { | |
1450 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1451 | struct tcp_sock *tp; | |
1452 | struct sk_buff *opt_skb = NULL; | |
1453 | ||
1454 | /* Imagine: socket is IPv6. IPv4 packet arrives, | |
1455 | goes to IPv4 receive handler and backlogged. | |
1456 | From backlog it always goes here. Kerboom... | |
1457 | Fortunately, tcp_rcv_established and rcv_established | |
1458 | handle them correctly, but it is not case with | |
1459 | tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK | |
1460 | */ | |
1461 | ||
1462 | if (skb->protocol == htons(ETH_P_IP)) | |
1463 | return tcp_v4_do_rcv(sk, skb); | |
1464 | ||
cfb6eeb4 YH |
1465 | #ifdef CONFIG_TCP_MD5SIG |
1466 | if (tcp_v6_inbound_md5_hash (sk, skb)) | |
1467 | goto discard; | |
1468 | #endif | |
1469 | ||
fda9ef5d | 1470 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1471 | goto discard; |
1472 | ||
1473 | /* | |
1474 | * socket locking is here for SMP purposes as backlog rcv | |
1475 | * is currently called with bh processing disabled. | |
1476 | */ | |
1477 | ||
1478 | /* Do Stevens' IPV6_PKTOPTIONS. | |
1479 | ||
1480 | Yes, guys, it is the only place in our code, where we | |
1481 | may make it not affecting IPv4. | |
1482 | The rest of code is protocol independent, | |
1483 | and I do not like idea to uglify IPv4. | |
1484 | ||
1485 | Actually, all the idea behind IPV6_PKTOPTIONS | |
1486 | looks not very well thought. For now we latch | |
1487 | options, received in the last packet, enqueued | |
1488 | by tcp. Feel free to propose better solution. | |
1ab1457c | 1489 | --ANK (980728) |
1da177e4 LT |
1490 | */ |
1491 | if (np->rxopt.all) | |
1492 | opt_skb = skb_clone(skb, GFP_ATOMIC); | |
1493 | ||
1494 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ | |
bdeab991 | 1495 | sock_rps_save_rxhash(sk, skb); |
aa8223c7 | 1496 | if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len)) |
1da177e4 | 1497 | goto reset; |
1da177e4 LT |
1498 | if (opt_skb) |
1499 | goto ipv6_pktoptions; | |
1500 | return 0; | |
1501 | } | |
1502 | ||
ab6a5bb6 | 1503 | if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb)) |
1da177e4 LT |
1504 | goto csum_err; |
1505 | ||
1ab1457c | 1506 | if (sk->sk_state == TCP_LISTEN) { |
1da177e4 LT |
1507 | struct sock *nsk = tcp_v6_hnd_req(sk, skb); |
1508 | if (!nsk) | |
1509 | goto discard; | |
1510 | ||
1511 | /* | |
1512 | * Queue it on the new socket if the new socket is active, | |
1513 | * otherwise we just shortcircuit this and continue with | |
1514 | * the new socket.. | |
1515 | */ | |
1ab1457c | 1516 | if(nsk != sk) { |
bdeab991 | 1517 | sock_rps_save_rxhash(nsk, skb); |
1da177e4 LT |
1518 | if (tcp_child_process(sk, nsk, skb)) |
1519 | goto reset; | |
1520 | if (opt_skb) | |
1521 | __kfree_skb(opt_skb); | |
1522 | return 0; | |
1523 | } | |
47482f13 | 1524 | } else |
bdeab991 | 1525 | sock_rps_save_rxhash(sk, skb); |
1da177e4 | 1526 | |
aa8223c7 | 1527 | if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) |
1da177e4 | 1528 | goto reset; |
1da177e4 LT |
1529 | if (opt_skb) |
1530 | goto ipv6_pktoptions; | |
1531 | return 0; | |
1532 | ||
1533 | reset: | |
cfb6eeb4 | 1534 | tcp_v6_send_reset(sk, skb); |
1da177e4 LT |
1535 | discard: |
1536 | if (opt_skb) | |
1537 | __kfree_skb(opt_skb); | |
1538 | kfree_skb(skb); | |
1539 | return 0; | |
1540 | csum_err: | |
63231bdd | 1541 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS); |
1da177e4 LT |
1542 | goto discard; |
1543 | ||
1544 | ||
1545 | ipv6_pktoptions: | |
1546 | /* Do you ask, what is it? | |
1547 | ||
1548 | 1. skb was enqueued by tcp. | |
1549 | 2. skb is added to tail of read queue, rather than out of order. | |
1550 | 3. socket is not in passive state. | |
1551 | 4. Finally, it really contains options, which user wants to receive. | |
1552 | */ | |
1553 | tp = tcp_sk(sk); | |
1554 | if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt && | |
1555 | !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) { | |
333fad53 | 1556 | if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo) |
505cbfc5 | 1557 | np->mcast_oif = inet6_iif(opt_skb); |
333fad53 | 1558 | if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) |
0660e03f | 1559 | np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit; |
4c507d28 JB |
1560 | if (np->rxopt.bits.rxtclass) |
1561 | np->rcv_tclass = ipv6_tclass(ipv6_hdr(skb)); | |
1da177e4 LT |
1562 | if (ipv6_opt_accepted(sk, opt_skb)) { |
1563 | skb_set_owner_r(opt_skb, sk); | |
1564 | opt_skb = xchg(&np->pktoptions, opt_skb); | |
1565 | } else { | |
1566 | __kfree_skb(opt_skb); | |
1567 | opt_skb = xchg(&np->pktoptions, NULL); | |
1568 | } | |
1569 | } | |
1570 | ||
800d55f1 | 1571 | kfree_skb(opt_skb); |
1da177e4 LT |
1572 | return 0; |
1573 | } | |
1574 | ||
e5bbef20 | 1575 | static int tcp_v6_rcv(struct sk_buff *skb) |
1da177e4 | 1576 | { |
cf533ea5 | 1577 | const struct tcphdr *th; |
b71d1d42 | 1578 | const struct ipv6hdr *hdr; |
1da177e4 LT |
1579 | struct sock *sk; |
1580 | int ret; | |
a86b1e30 | 1581 | struct net *net = dev_net(skb->dev); |
1da177e4 LT |
1582 | |
1583 | if (skb->pkt_type != PACKET_HOST) | |
1584 | goto discard_it; | |
1585 | ||
1586 | /* | |
1587 | * Count it even if it's bad. | |
1588 | */ | |
63231bdd | 1589 | TCP_INC_STATS_BH(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1590 | |
1591 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1592 | goto discard_it; | |
1593 | ||
aa8223c7 | 1594 | th = tcp_hdr(skb); |
1da177e4 LT |
1595 | |
1596 | if (th->doff < sizeof(struct tcphdr)/4) | |
1597 | goto bad_packet; | |
1598 | if (!pskb_may_pull(skb, th->doff*4)) | |
1599 | goto discard_it; | |
1600 | ||
60476372 | 1601 | if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb)) |
1da177e4 LT |
1602 | goto bad_packet; |
1603 | ||
aa8223c7 | 1604 | th = tcp_hdr(skb); |
e802af9c | 1605 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
1606 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); |
1607 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1608 | skb->len - th->doff*4); | |
1609 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
1610 | TCP_SKB_CB(skb)->when = 0; | |
b82d1bb4 | 1611 | TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr); |
1da177e4 LT |
1612 | TCP_SKB_CB(skb)->sacked = 0; |
1613 | ||
9a1f27c4 | 1614 | sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest); |
1da177e4 LT |
1615 | if (!sk) |
1616 | goto no_tcp_socket; | |
1617 | ||
1618 | process: | |
1619 | if (sk->sk_state == TCP_TIME_WAIT) | |
1620 | goto do_time_wait; | |
1621 | ||
e802af9c SH |
1622 | if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) { |
1623 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
1624 | goto discard_and_relse; | |
1625 | } | |
1626 | ||
1da177e4 LT |
1627 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) |
1628 | goto discard_and_relse; | |
1629 | ||
fda9ef5d | 1630 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1631 | goto discard_and_relse; |
1632 | ||
1633 | skb->dev = NULL; | |
1634 | ||
293b9c42 | 1635 | bh_lock_sock_nested(sk); |
1da177e4 LT |
1636 | ret = 0; |
1637 | if (!sock_owned_by_user(sk)) { | |
1a2449a8 | 1638 | #ifdef CONFIG_NET_DMA |
1ab1457c | 1639 | struct tcp_sock *tp = tcp_sk(sk); |
b4caea8a | 1640 | if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list) |
a2bd1140 | 1641 | tp->ucopy.dma_chan = net_dma_find_channel(); |
1ab1457c YH |
1642 | if (tp->ucopy.dma_chan) |
1643 | ret = tcp_v6_do_rcv(sk, skb); | |
1644 | else | |
1a2449a8 CL |
1645 | #endif |
1646 | { | |
1647 | if (!tcp_prequeue(sk, skb)) | |
1648 | ret = tcp_v6_do_rcv(sk, skb); | |
1649 | } | |
da882c1f ED |
1650 | } else if (unlikely(sk_add_backlog(sk, skb, |
1651 | sk->sk_rcvbuf + sk->sk_sndbuf))) { | |
6b03a53a | 1652 | bh_unlock_sock(sk); |
6cce09f8 | 1653 | NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP); |
6b03a53a ZY |
1654 | goto discard_and_relse; |
1655 | } | |
1da177e4 LT |
1656 | bh_unlock_sock(sk); |
1657 | ||
1658 | sock_put(sk); | |
1659 | return ret ? -1 : 0; | |
1660 | ||
1661 | no_tcp_socket: | |
1662 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1663 | goto discard_it; | |
1664 | ||
1665 | if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) { | |
1666 | bad_packet: | |
63231bdd | 1667 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
1da177e4 | 1668 | } else { |
cfb6eeb4 | 1669 | tcp_v6_send_reset(NULL, skb); |
1da177e4 LT |
1670 | } |
1671 | ||
1672 | discard_it: | |
1673 | ||
1674 | /* | |
1675 | * Discard frame | |
1676 | */ | |
1677 | ||
1678 | kfree_skb(skb); | |
1679 | return 0; | |
1680 | ||
1681 | discard_and_relse: | |
1682 | sock_put(sk); | |
1683 | goto discard_it; | |
1684 | ||
1685 | do_time_wait: | |
1686 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1687 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1688 | goto discard_it; |
1689 | } | |
1690 | ||
1691 | if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) { | |
63231bdd | 1692 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
9469c7b4 | 1693 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1694 | goto discard_it; |
1695 | } | |
1696 | ||
9469c7b4 | 1697 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 LT |
1698 | case TCP_TW_SYN: |
1699 | { | |
1700 | struct sock *sk2; | |
1701 | ||
c346dca1 | 1702 | sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo, |
0660e03f | 1703 | &ipv6_hdr(skb)->daddr, |
505cbfc5 | 1704 | ntohs(th->dest), inet6_iif(skb)); |
1da177e4 | 1705 | if (sk2 != NULL) { |
295ff7ed ACM |
1706 | struct inet_timewait_sock *tw = inet_twsk(sk); |
1707 | inet_twsk_deschedule(tw, &tcp_death_row); | |
1708 | inet_twsk_put(tw); | |
1da177e4 LT |
1709 | sk = sk2; |
1710 | goto process; | |
1711 | } | |
1712 | /* Fall through to ACK */ | |
1713 | } | |
1714 | case TCP_TW_ACK: | |
1715 | tcp_v6_timewait_ack(sk, skb); | |
1716 | break; | |
1717 | case TCP_TW_RST: | |
1718 | goto no_tcp_socket; | |
1719 | case TCP_TW_SUCCESS:; | |
1720 | } | |
1721 | goto discard_it; | |
1722 | } | |
1723 | ||
4670fd81 | 1724 | static struct inet_peer *tcp_v6_get_peer(struct sock *sk) |
ccb7c410 | 1725 | { |
db3949c4 DM |
1726 | struct rt6_info *rt = (struct rt6_info *) __sk_dst_get(sk); |
1727 | struct ipv6_pinfo *np = inet6_sk(sk); | |
db3949c4 | 1728 | |
4670fd81 DM |
1729 | /* If we don't have a valid cached route, or we're doing IP |
1730 | * options which make the IPv6 header destination address | |
1731 | * different from our peer's, do not bother with this. | |
1732 | */ | |
1733 | if (!rt || !ipv6_addr_equal(&np->daddr, &rt->rt6i_dst.addr)) | |
1734 | return NULL; | |
1735 | return rt6_get_peer_create(rt); | |
ccb7c410 DM |
1736 | } |
1737 | ||
ccb7c410 DM |
1738 | static struct timewait_sock_ops tcp6_timewait_sock_ops = { |
1739 | .twsk_obj_size = sizeof(struct tcp6_timewait_sock), | |
1740 | .twsk_unique = tcp_twsk_unique, | |
1741 | .twsk_destructor= tcp_twsk_destructor, | |
ccb7c410 DM |
1742 | }; |
1743 | ||
3b401a81 | 1744 | static const struct inet_connection_sock_af_ops ipv6_specific = { |
543d9cfe ACM |
1745 | .queue_xmit = inet6_csk_xmit, |
1746 | .send_check = tcp_v6_send_check, | |
1747 | .rebuild_header = inet6_sk_rebuild_header, | |
1748 | .conn_request = tcp_v6_conn_request, | |
1749 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
3f419d2d | 1750 | .get_peer = tcp_v6_get_peer, |
543d9cfe | 1751 | .net_header_len = sizeof(struct ipv6hdr), |
67469601 | 1752 | .net_frag_header_len = sizeof(struct frag_hdr), |
543d9cfe ACM |
1753 | .setsockopt = ipv6_setsockopt, |
1754 | .getsockopt = ipv6_getsockopt, | |
1755 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1756 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
ab1e0a13 | 1757 | .bind_conflict = inet6_csk_bind_conflict, |
3fdadf7d | 1758 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1759 | .compat_setsockopt = compat_ipv6_setsockopt, |
1760 | .compat_getsockopt = compat_ipv6_getsockopt, | |
3fdadf7d | 1761 | #endif |
1da177e4 LT |
1762 | }; |
1763 | ||
cfb6eeb4 | 1764 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1765 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = { |
cfb6eeb4 | 1766 | .md5_lookup = tcp_v6_md5_lookup, |
49a72dfb | 1767 | .calc_md5_hash = tcp_v6_md5_hash_skb, |
cfb6eeb4 | 1768 | .md5_parse = tcp_v6_parse_md5_keys, |
cfb6eeb4 | 1769 | }; |
a928630a | 1770 | #endif |
cfb6eeb4 | 1771 | |
1da177e4 LT |
1772 | /* |
1773 | * TCP over IPv4 via INET6 API | |
1774 | */ | |
1775 | ||
3b401a81 | 1776 | static const struct inet_connection_sock_af_ops ipv6_mapped = { |
543d9cfe ACM |
1777 | .queue_xmit = ip_queue_xmit, |
1778 | .send_check = tcp_v4_send_check, | |
1779 | .rebuild_header = inet_sk_rebuild_header, | |
1780 | .conn_request = tcp_v6_conn_request, | |
1781 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
3f419d2d | 1782 | .get_peer = tcp_v4_get_peer, |
543d9cfe ACM |
1783 | .net_header_len = sizeof(struct iphdr), |
1784 | .setsockopt = ipv6_setsockopt, | |
1785 | .getsockopt = ipv6_getsockopt, | |
1786 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1787 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
ab1e0a13 | 1788 | .bind_conflict = inet6_csk_bind_conflict, |
3fdadf7d | 1789 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1790 | .compat_setsockopt = compat_ipv6_setsockopt, |
1791 | .compat_getsockopt = compat_ipv6_getsockopt, | |
3fdadf7d | 1792 | #endif |
1da177e4 LT |
1793 | }; |
1794 | ||
cfb6eeb4 | 1795 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1796 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = { |
cfb6eeb4 | 1797 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1798 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1799 | .md5_parse = tcp_v6_parse_md5_keys, |
cfb6eeb4 | 1800 | }; |
a928630a | 1801 | #endif |
cfb6eeb4 | 1802 | |
1da177e4 LT |
1803 | /* NOTE: A lot of things set to zero explicitly by call to |
1804 | * sk_alloc() so need not be done here. | |
1805 | */ | |
1806 | static int tcp_v6_init_sock(struct sock *sk) | |
1807 | { | |
6687e988 | 1808 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 | 1809 | |
900f65d3 | 1810 | tcp_init_sock(sk); |
1da177e4 | 1811 | |
8292a17a | 1812 | icsk->icsk_af_ops = &ipv6_specific; |
1da177e4 | 1813 | |
cfb6eeb4 | 1814 | #ifdef CONFIG_TCP_MD5SIG |
ac807fa8 | 1815 | tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific; |
cfb6eeb4 YH |
1816 | #endif |
1817 | ||
1da177e4 LT |
1818 | return 0; |
1819 | } | |
1820 | ||
7d06b2e0 | 1821 | static void tcp_v6_destroy_sock(struct sock *sk) |
1da177e4 | 1822 | { |
1da177e4 | 1823 | tcp_v4_destroy_sock(sk); |
7d06b2e0 | 1824 | inet6_destroy_sock(sk); |
1da177e4 LT |
1825 | } |
1826 | ||
952a10be | 1827 | #ifdef CONFIG_PROC_FS |
1da177e4 | 1828 | /* Proc filesystem TCPv6 sock list dumping. */ |
1ab1457c | 1829 | static void get_openreq6(struct seq_file *seq, |
cf533ea5 | 1830 | const struct sock *sk, struct request_sock *req, int i, int uid) |
1da177e4 | 1831 | { |
1da177e4 | 1832 | int ttd = req->expires - jiffies; |
b71d1d42 ED |
1833 | const struct in6_addr *src = &inet6_rsk(req)->loc_addr; |
1834 | const struct in6_addr *dest = &inet6_rsk(req)->rmt_addr; | |
1da177e4 LT |
1835 | |
1836 | if (ttd < 0) | |
1837 | ttd = 0; | |
1838 | ||
1da177e4 LT |
1839 | seq_printf(seq, |
1840 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 1841 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n", |
1da177e4 LT |
1842 | i, |
1843 | src->s6_addr32[0], src->s6_addr32[1], | |
1844 | src->s6_addr32[2], src->s6_addr32[3], | |
fd507037 | 1845 | ntohs(inet_rsk(req)->loc_port), |
1da177e4 LT |
1846 | dest->s6_addr32[0], dest->s6_addr32[1], |
1847 | dest->s6_addr32[2], dest->s6_addr32[3], | |
2e6599cb | 1848 | ntohs(inet_rsk(req)->rmt_port), |
1da177e4 LT |
1849 | TCP_SYN_RECV, |
1850 | 0,0, /* could print option size, but that is af dependent. */ | |
1ab1457c YH |
1851 | 1, /* timers active (only the expire timer) */ |
1852 | jiffies_to_clock_t(ttd), | |
1da177e4 LT |
1853 | req->retrans, |
1854 | uid, | |
1ab1457c | 1855 | 0, /* non standard timer */ |
1da177e4 LT |
1856 | 0, /* open_requests have no inode */ |
1857 | 0, req); | |
1858 | } | |
1859 | ||
1860 | static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i) | |
1861 | { | |
b71d1d42 | 1862 | const struct in6_addr *dest, *src; |
1da177e4 LT |
1863 | __u16 destp, srcp; |
1864 | int timer_active; | |
1865 | unsigned long timer_expires; | |
cf533ea5 ED |
1866 | const struct inet_sock *inet = inet_sk(sp); |
1867 | const struct tcp_sock *tp = tcp_sk(sp); | |
463c84b9 | 1868 | const struct inet_connection_sock *icsk = inet_csk(sp); |
cf533ea5 | 1869 | const struct ipv6_pinfo *np = inet6_sk(sp); |
1da177e4 LT |
1870 | |
1871 | dest = &np->daddr; | |
1872 | src = &np->rcv_saddr; | |
c720c7e8 ED |
1873 | destp = ntohs(inet->inet_dport); |
1874 | srcp = ntohs(inet->inet_sport); | |
463c84b9 ACM |
1875 | |
1876 | if (icsk->icsk_pending == ICSK_TIME_RETRANS) { | |
1da177e4 | 1877 | timer_active = 1; |
463c84b9 ACM |
1878 | timer_expires = icsk->icsk_timeout; |
1879 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 1880 | timer_active = 4; |
463c84b9 | 1881 | timer_expires = icsk->icsk_timeout; |
1da177e4 LT |
1882 | } else if (timer_pending(&sp->sk_timer)) { |
1883 | timer_active = 2; | |
1884 | timer_expires = sp->sk_timer.expires; | |
1885 | } else { | |
1886 | timer_active = 0; | |
1887 | timer_expires = jiffies; | |
1888 | } | |
1889 | ||
1890 | seq_printf(seq, | |
1891 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 1892 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %lu %lu %u %u %d\n", |
1da177e4 LT |
1893 | i, |
1894 | src->s6_addr32[0], src->s6_addr32[1], | |
1895 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
1896 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1897 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
1ab1457c | 1898 | sp->sk_state, |
47da8ee6 SS |
1899 | tp->write_seq-tp->snd_una, |
1900 | (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq), | |
1da177e4 LT |
1901 | timer_active, |
1902 | jiffies_to_clock_t(timer_expires - jiffies), | |
463c84b9 | 1903 | icsk->icsk_retransmits, |
1da177e4 | 1904 | sock_i_uid(sp), |
6687e988 | 1905 | icsk->icsk_probes_out, |
1da177e4 LT |
1906 | sock_i_ino(sp), |
1907 | atomic_read(&sp->sk_refcnt), sp, | |
7be87351 SH |
1908 | jiffies_to_clock_t(icsk->icsk_rto), |
1909 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
463c84b9 | 1910 | (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong, |
0b6a05c1 IJ |
1911 | tp->snd_cwnd, |
1912 | tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh | |
1da177e4 LT |
1913 | ); |
1914 | } | |
1915 | ||
1ab1457c | 1916 | static void get_timewait6_sock(struct seq_file *seq, |
8feaf0c0 | 1917 | struct inet_timewait_sock *tw, int i) |
1da177e4 | 1918 | { |
b71d1d42 | 1919 | const struct in6_addr *dest, *src; |
1da177e4 | 1920 | __u16 destp, srcp; |
cf533ea5 | 1921 | const struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw); |
1da177e4 LT |
1922 | int ttd = tw->tw_ttd - jiffies; |
1923 | ||
1924 | if (ttd < 0) | |
1925 | ttd = 0; | |
1926 | ||
0fa1a53e ACM |
1927 | dest = &tw6->tw_v6_daddr; |
1928 | src = &tw6->tw_v6_rcv_saddr; | |
1da177e4 LT |
1929 | destp = ntohs(tw->tw_dport); |
1930 | srcp = ntohs(tw->tw_sport); | |
1931 | ||
1932 | seq_printf(seq, | |
1933 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 1934 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n", |
1da177e4 LT |
1935 | i, |
1936 | src->s6_addr32[0], src->s6_addr32[1], | |
1937 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
1938 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1939 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
1940 | tw->tw_substate, 0, 0, | |
1941 | 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0, | |
1942 | atomic_read(&tw->tw_refcnt), tw); | |
1943 | } | |
1944 | ||
1da177e4 LT |
1945 | static int tcp6_seq_show(struct seq_file *seq, void *v) |
1946 | { | |
1947 | struct tcp_iter_state *st; | |
1948 | ||
1949 | if (v == SEQ_START_TOKEN) { | |
1950 | seq_puts(seq, | |
1951 | " sl " | |
1952 | "local_address " | |
1953 | "remote_address " | |
1954 | "st tx_queue rx_queue tr tm->when retrnsmt" | |
1955 | " uid timeout inode\n"); | |
1956 | goto out; | |
1957 | } | |
1958 | st = seq->private; | |
1959 | ||
1960 | switch (st->state) { | |
1961 | case TCP_SEQ_STATE_LISTENING: | |
1962 | case TCP_SEQ_STATE_ESTABLISHED: | |
1963 | get_tcp6_sock(seq, v, st->num); | |
1964 | break; | |
1965 | case TCP_SEQ_STATE_OPENREQ: | |
1966 | get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid); | |
1967 | break; | |
1968 | case TCP_SEQ_STATE_TIME_WAIT: | |
1969 | get_timewait6_sock(seq, v, st->num); | |
1970 | break; | |
1971 | } | |
1972 | out: | |
1973 | return 0; | |
1974 | } | |
1975 | ||
73cb88ec AV |
1976 | static const struct file_operations tcp6_afinfo_seq_fops = { |
1977 | .owner = THIS_MODULE, | |
1978 | .open = tcp_seq_open, | |
1979 | .read = seq_read, | |
1980 | .llseek = seq_lseek, | |
1981 | .release = seq_release_net | |
1982 | }; | |
1983 | ||
1da177e4 | 1984 | static struct tcp_seq_afinfo tcp6_seq_afinfo = { |
1da177e4 LT |
1985 | .name = "tcp6", |
1986 | .family = AF_INET6, | |
73cb88ec | 1987 | .seq_fops = &tcp6_afinfo_seq_fops, |
9427c4b3 DL |
1988 | .seq_ops = { |
1989 | .show = tcp6_seq_show, | |
1990 | }, | |
1da177e4 LT |
1991 | }; |
1992 | ||
2c8c1e72 | 1993 | int __net_init tcp6_proc_init(struct net *net) |
1da177e4 | 1994 | { |
6f8b13bc | 1995 | return tcp_proc_register(net, &tcp6_seq_afinfo); |
1da177e4 LT |
1996 | } |
1997 | ||
6f8b13bc | 1998 | void tcp6_proc_exit(struct net *net) |
1da177e4 | 1999 | { |
6f8b13bc | 2000 | tcp_proc_unregister(net, &tcp6_seq_afinfo); |
1da177e4 LT |
2001 | } |
2002 | #endif | |
2003 | ||
2004 | struct proto tcpv6_prot = { | |
2005 | .name = "TCPv6", | |
2006 | .owner = THIS_MODULE, | |
2007 | .close = tcp_close, | |
2008 | .connect = tcp_v6_connect, | |
2009 | .disconnect = tcp_disconnect, | |
463c84b9 | 2010 | .accept = inet_csk_accept, |
1da177e4 LT |
2011 | .ioctl = tcp_ioctl, |
2012 | .init = tcp_v6_init_sock, | |
2013 | .destroy = tcp_v6_destroy_sock, | |
2014 | .shutdown = tcp_shutdown, | |
2015 | .setsockopt = tcp_setsockopt, | |
2016 | .getsockopt = tcp_getsockopt, | |
1da177e4 | 2017 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
2018 | .sendmsg = tcp_sendmsg, |
2019 | .sendpage = tcp_sendpage, | |
1da177e4 LT |
2020 | .backlog_rcv = tcp_v6_do_rcv, |
2021 | .hash = tcp_v6_hash, | |
ab1e0a13 ACM |
2022 | .unhash = inet_unhash, |
2023 | .get_port = inet_csk_get_port, | |
1da177e4 LT |
2024 | .enter_memory_pressure = tcp_enter_memory_pressure, |
2025 | .sockets_allocated = &tcp_sockets_allocated, | |
2026 | .memory_allocated = &tcp_memory_allocated, | |
2027 | .memory_pressure = &tcp_memory_pressure, | |
0a5578cf | 2028 | .orphan_count = &tcp_orphan_count, |
1da177e4 LT |
2029 | .sysctl_wmem = sysctl_tcp_wmem, |
2030 | .sysctl_rmem = sysctl_tcp_rmem, | |
2031 | .max_header = MAX_TCP_HEADER, | |
2032 | .obj_size = sizeof(struct tcp6_sock), | |
3ab5aee7 | 2033 | .slab_flags = SLAB_DESTROY_BY_RCU, |
6d6ee43e | 2034 | .twsk_prot = &tcp6_timewait_sock_ops, |
60236fdd | 2035 | .rsk_prot = &tcp6_request_sock_ops, |
39d8cda7 | 2036 | .h.hashinfo = &tcp_hashinfo, |
7ba42910 | 2037 | .no_autobind = true, |
543d9cfe ACM |
2038 | #ifdef CONFIG_COMPAT |
2039 | .compat_setsockopt = compat_tcp_setsockopt, | |
2040 | .compat_getsockopt = compat_tcp_getsockopt, | |
2041 | #endif | |
d1a4c0b3 GC |
2042 | #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM |
2043 | .proto_cgroup = tcp_proto_cgroup, | |
2044 | #endif | |
1da177e4 LT |
2045 | }; |
2046 | ||
41135cc8 | 2047 | static const struct inet6_protocol tcpv6_protocol = { |
1da177e4 LT |
2048 | .handler = tcp_v6_rcv, |
2049 | .err_handler = tcp_v6_err, | |
a430a43d | 2050 | .gso_send_check = tcp_v6_gso_send_check, |
adcfc7d0 | 2051 | .gso_segment = tcp_tso_segment, |
684f2176 HX |
2052 | .gro_receive = tcp6_gro_receive, |
2053 | .gro_complete = tcp6_gro_complete, | |
1da177e4 LT |
2054 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, |
2055 | }; | |
2056 | ||
1da177e4 LT |
2057 | static struct inet_protosw tcpv6_protosw = { |
2058 | .type = SOCK_STREAM, | |
2059 | .protocol = IPPROTO_TCP, | |
2060 | .prot = &tcpv6_prot, | |
2061 | .ops = &inet6_stream_ops, | |
1da177e4 | 2062 | .no_check = 0, |
d83d8461 ACM |
2063 | .flags = INET_PROTOSW_PERMANENT | |
2064 | INET_PROTOSW_ICSK, | |
1da177e4 LT |
2065 | }; |
2066 | ||
2c8c1e72 | 2067 | static int __net_init tcpv6_net_init(struct net *net) |
93ec926b | 2068 | { |
5677242f DL |
2069 | return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6, |
2070 | SOCK_RAW, IPPROTO_TCP, net); | |
93ec926b DL |
2071 | } |
2072 | ||
2c8c1e72 | 2073 | static void __net_exit tcpv6_net_exit(struct net *net) |
93ec926b | 2074 | { |
5677242f | 2075 | inet_ctl_sock_destroy(net->ipv6.tcp_sk); |
b099ce26 EB |
2076 | } |
2077 | ||
2c8c1e72 | 2078 | static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list) |
b099ce26 EB |
2079 | { |
2080 | inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6); | |
93ec926b DL |
2081 | } |
2082 | ||
2083 | static struct pernet_operations tcpv6_net_ops = { | |
b099ce26 EB |
2084 | .init = tcpv6_net_init, |
2085 | .exit = tcpv6_net_exit, | |
2086 | .exit_batch = tcpv6_net_exit_batch, | |
93ec926b DL |
2087 | }; |
2088 | ||
7f4e4868 | 2089 | int __init tcpv6_init(void) |
1da177e4 | 2090 | { |
7f4e4868 DL |
2091 | int ret; |
2092 | ||
2093 | ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
2094 | if (ret) | |
2095 | goto out; | |
2096 | ||
1da177e4 | 2097 | /* register inet6 protocol */ |
7f4e4868 DL |
2098 | ret = inet6_register_protosw(&tcpv6_protosw); |
2099 | if (ret) | |
2100 | goto out_tcpv6_protocol; | |
2101 | ||
93ec926b | 2102 | ret = register_pernet_subsys(&tcpv6_net_ops); |
7f4e4868 DL |
2103 | if (ret) |
2104 | goto out_tcpv6_protosw; | |
2105 | out: | |
2106 | return ret; | |
ae0f7d5f | 2107 | |
7f4e4868 DL |
2108 | out_tcpv6_protocol: |
2109 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
2110 | out_tcpv6_protosw: | |
2111 | inet6_unregister_protosw(&tcpv6_protosw); | |
2112 | goto out; | |
2113 | } | |
2114 | ||
09f7709f | 2115 | void tcpv6_exit(void) |
7f4e4868 | 2116 | { |
93ec926b | 2117 | unregister_pernet_subsys(&tcpv6_net_ops); |
7f4e4868 DL |
2118 | inet6_unregister_protosw(&tcpv6_protosw); |
2119 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
1da177e4 | 2120 | } |