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 | ||
d83d8461 | 418 | if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) { |
1da177e4 LT |
419 | tcp_sync_mss(sk, dst_mtu(dst)); |
420 | tcp_simple_retransmit(sk); | |
421 | } /* else let the usual retransmit timer handle it */ | |
422 | dst_release(dst); | |
423 | goto out; | |
424 | } | |
425 | ||
426 | icmpv6_err_convert(type, code, &err); | |
427 | ||
60236fdd | 428 | /* Might be for an request_sock */ |
1da177e4 | 429 | switch (sk->sk_state) { |
60236fdd | 430 | struct request_sock *req, **prev; |
1da177e4 LT |
431 | case TCP_LISTEN: |
432 | if (sock_owned_by_user(sk)) | |
433 | goto out; | |
434 | ||
8129765a ACM |
435 | req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr, |
436 | &hdr->saddr, inet6_iif(skb)); | |
1da177e4 LT |
437 | if (!req) |
438 | goto out; | |
439 | ||
440 | /* ICMPs are not backlogged, hence we cannot get | |
441 | * an established socket here. | |
442 | */ | |
547b792c | 443 | WARN_ON(req->sk != NULL); |
1da177e4 | 444 | |
2e6599cb | 445 | if (seq != tcp_rsk(req)->snt_isn) { |
de0744af | 446 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
447 | goto out; |
448 | } | |
449 | ||
463c84b9 | 450 | inet_csk_reqsk_queue_drop(sk, req, prev); |
1da177e4 LT |
451 | goto out; |
452 | ||
453 | case TCP_SYN_SENT: | |
454 | case TCP_SYN_RECV: /* Cannot happen. | |
1ab1457c | 455 | It can, it SYNs are crossed. --ANK */ |
1da177e4 | 456 | if (!sock_owned_by_user(sk)) { |
1da177e4 LT |
457 | sk->sk_err = err; |
458 | sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */ | |
459 | ||
460 | tcp_done(sk); | |
461 | } else | |
462 | sk->sk_err_soft = err; | |
463 | goto out; | |
464 | } | |
465 | ||
466 | if (!sock_owned_by_user(sk) && np->recverr) { | |
467 | sk->sk_err = err; | |
468 | sk->sk_error_report(sk); | |
469 | } else | |
470 | sk->sk_err_soft = err; | |
471 | ||
472 | out: | |
473 | bh_unlock_sock(sk); | |
474 | sock_put(sk); | |
475 | } | |
476 | ||
477 | ||
e6b4d113 WAS |
478 | static int tcp_v6_send_synack(struct sock *sk, struct request_sock *req, |
479 | struct request_values *rvp) | |
1da177e4 | 480 | { |
ca304b61 | 481 | struct inet6_request_sock *treq = inet6_rsk(req); |
1da177e4 LT |
482 | struct ipv6_pinfo *np = inet6_sk(sk); |
483 | struct sk_buff * skb; | |
484 | struct ipv6_txoptions *opt = NULL; | |
20c59de2 | 485 | struct in6_addr * final_p, final; |
4c9483b2 | 486 | struct flowi6 fl6; |
fd80eb94 | 487 | struct dst_entry *dst; |
68d0c6d3 | 488 | int err; |
1da177e4 | 489 | |
4c9483b2 DM |
490 | memset(&fl6, 0, sizeof(fl6)); |
491 | fl6.flowi6_proto = IPPROTO_TCP; | |
4e3fd7a0 AD |
492 | fl6.daddr = treq->rmt_addr; |
493 | fl6.saddr = treq->loc_addr; | |
4c9483b2 DM |
494 | fl6.flowlabel = 0; |
495 | fl6.flowi6_oif = treq->iif; | |
496 | fl6.flowi6_mark = sk->sk_mark; | |
1958b856 DM |
497 | fl6.fl6_dport = inet_rsk(req)->rmt_port; |
498 | fl6.fl6_sport = inet_rsk(req)->loc_port; | |
4c9483b2 | 499 | security_req_classify_flow(req, flowi6_to_flowi(&fl6)); |
1da177e4 | 500 | |
fd80eb94 | 501 | opt = np->opt; |
4c9483b2 | 502 | final_p = fl6_update_dst(&fl6, opt, &final); |
1da177e4 | 503 | |
4c9483b2 | 504 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false); |
68d0c6d3 DM |
505 | if (IS_ERR(dst)) { |
506 | err = PTR_ERR(dst); | |
738faca3 | 507 | dst = NULL; |
fd80eb94 | 508 | goto done; |
68d0c6d3 | 509 | } |
e6b4d113 | 510 | skb = tcp_make_synack(sk, dst, req, rvp); |
68d0c6d3 | 511 | err = -ENOMEM; |
1da177e4 | 512 | if (skb) { |
8ad50d96 | 513 | __tcp_v6_send_check(skb, &treq->loc_addr, &treq->rmt_addr); |
1da177e4 | 514 | |
4e3fd7a0 | 515 | fl6.daddr = treq->rmt_addr; |
b903d324 | 516 | err = ip6_xmit(sk, skb, &fl6, opt, np->tclass); |
b9df3cb8 | 517 | err = net_xmit_eval(err); |
1da177e4 LT |
518 | } |
519 | ||
520 | done: | |
1ab1457c | 521 | if (opt && opt != np->opt) |
1da177e4 | 522 | sock_kfree_s(sk, opt, opt->tot_len); |
78b91042 | 523 | dst_release(dst); |
1da177e4 LT |
524 | return err; |
525 | } | |
526 | ||
72659ecc OP |
527 | static int tcp_v6_rtx_synack(struct sock *sk, struct request_sock *req, |
528 | struct request_values *rvp) | |
529 | { | |
530 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS); | |
531 | return tcp_v6_send_synack(sk, req, rvp); | |
532 | } | |
533 | ||
60236fdd | 534 | static void tcp_v6_reqsk_destructor(struct request_sock *req) |
1da177e4 | 535 | { |
800d55f1 | 536 | kfree_skb(inet6_rsk(req)->pktopts); |
1da177e4 LT |
537 | } |
538 | ||
cfb6eeb4 YH |
539 | #ifdef CONFIG_TCP_MD5SIG |
540 | static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk, | |
b71d1d42 | 541 | const struct in6_addr *addr) |
cfb6eeb4 | 542 | { |
a915da9b | 543 | return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6); |
cfb6eeb4 YH |
544 | } |
545 | ||
546 | static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk, | |
547 | struct sock *addr_sk) | |
548 | { | |
549 | return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr); | |
550 | } | |
551 | ||
552 | static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk, | |
553 | struct request_sock *req) | |
554 | { | |
555 | return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr); | |
556 | } | |
557 | ||
cfb6eeb4 YH |
558 | static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval, |
559 | int optlen) | |
560 | { | |
561 | struct tcp_md5sig cmd; | |
562 | struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr; | |
cfb6eeb4 YH |
563 | |
564 | if (optlen < sizeof(cmd)) | |
565 | return -EINVAL; | |
566 | ||
567 | if (copy_from_user(&cmd, optval, sizeof(cmd))) | |
568 | return -EFAULT; | |
569 | ||
570 | if (sin6->sin6_family != AF_INET6) | |
571 | return -EINVAL; | |
572 | ||
573 | if (!cmd.tcpm_keylen) { | |
e773e4fa | 574 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) |
a915da9b ED |
575 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3], |
576 | AF_INET); | |
577 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr, | |
578 | AF_INET6); | |
cfb6eeb4 YH |
579 | } |
580 | ||
581 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
582 | return -EINVAL; | |
583 | ||
a915da9b ED |
584 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) |
585 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3], | |
586 | AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL); | |
cfb6eeb4 | 587 | |
a915da9b ED |
588 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr, |
589 | AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL); | |
cfb6eeb4 YH |
590 | } |
591 | ||
49a72dfb | 592 | static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp, |
b71d1d42 ED |
593 | const struct in6_addr *daddr, |
594 | const struct in6_addr *saddr, int nbytes) | |
cfb6eeb4 | 595 | { |
cfb6eeb4 | 596 | struct tcp6_pseudohdr *bp; |
49a72dfb | 597 | struct scatterlist sg; |
8d26d76d | 598 | |
cfb6eeb4 | 599 | bp = &hp->md5_blk.ip6; |
cfb6eeb4 | 600 | /* 1. TCP pseudo-header (RFC2460) */ |
4e3fd7a0 AD |
601 | bp->saddr = *saddr; |
602 | bp->daddr = *daddr; | |
49a72dfb | 603 | bp->protocol = cpu_to_be32(IPPROTO_TCP); |
00b1304c | 604 | bp->len = cpu_to_be32(nbytes); |
cfb6eeb4 | 605 | |
49a72dfb AL |
606 | sg_init_one(&sg, bp, sizeof(*bp)); |
607 | return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp)); | |
608 | } | |
c7da57a1 | 609 | |
49a72dfb | 610 | static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key, |
b71d1d42 | 611 | const struct in6_addr *daddr, struct in6_addr *saddr, |
318cf7aa | 612 | const struct tcphdr *th) |
49a72dfb AL |
613 | { |
614 | struct tcp_md5sig_pool *hp; | |
615 | struct hash_desc *desc; | |
616 | ||
617 | hp = tcp_get_md5sig_pool(); | |
618 | if (!hp) | |
619 | goto clear_hash_noput; | |
620 | desc = &hp->md5_desc; | |
621 | ||
622 | if (crypto_hash_init(desc)) | |
623 | goto clear_hash; | |
624 | if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2)) | |
625 | goto clear_hash; | |
626 | if (tcp_md5_hash_header(hp, th)) | |
627 | goto clear_hash; | |
628 | if (tcp_md5_hash_key(hp, key)) | |
629 | goto clear_hash; | |
630 | if (crypto_hash_final(desc, md5_hash)) | |
cfb6eeb4 | 631 | goto clear_hash; |
cfb6eeb4 | 632 | |
cfb6eeb4 | 633 | tcp_put_md5sig_pool(); |
cfb6eeb4 | 634 | return 0; |
49a72dfb | 635 | |
cfb6eeb4 YH |
636 | clear_hash: |
637 | tcp_put_md5sig_pool(); | |
638 | clear_hash_noput: | |
639 | memset(md5_hash, 0, 16); | |
49a72dfb | 640 | return 1; |
cfb6eeb4 YH |
641 | } |
642 | ||
49a72dfb | 643 | static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key, |
318cf7aa ED |
644 | const struct sock *sk, |
645 | const struct request_sock *req, | |
646 | const struct sk_buff *skb) | |
cfb6eeb4 | 647 | { |
b71d1d42 | 648 | const struct in6_addr *saddr, *daddr; |
49a72dfb AL |
649 | struct tcp_md5sig_pool *hp; |
650 | struct hash_desc *desc; | |
318cf7aa | 651 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 YH |
652 | |
653 | if (sk) { | |
654 | saddr = &inet6_sk(sk)->saddr; | |
655 | daddr = &inet6_sk(sk)->daddr; | |
49a72dfb | 656 | } else if (req) { |
cfb6eeb4 YH |
657 | saddr = &inet6_rsk(req)->loc_addr; |
658 | daddr = &inet6_rsk(req)->rmt_addr; | |
49a72dfb | 659 | } else { |
b71d1d42 | 660 | const struct ipv6hdr *ip6h = ipv6_hdr(skb); |
49a72dfb AL |
661 | saddr = &ip6h->saddr; |
662 | daddr = &ip6h->daddr; | |
cfb6eeb4 | 663 | } |
49a72dfb AL |
664 | |
665 | hp = tcp_get_md5sig_pool(); | |
666 | if (!hp) | |
667 | goto clear_hash_noput; | |
668 | desc = &hp->md5_desc; | |
669 | ||
670 | if (crypto_hash_init(desc)) | |
671 | goto clear_hash; | |
672 | ||
673 | if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len)) | |
674 | goto clear_hash; | |
675 | if (tcp_md5_hash_header(hp, th)) | |
676 | goto clear_hash; | |
677 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
678 | goto clear_hash; | |
679 | if (tcp_md5_hash_key(hp, key)) | |
680 | goto clear_hash; | |
681 | if (crypto_hash_final(desc, md5_hash)) | |
682 | goto clear_hash; | |
683 | ||
684 | tcp_put_md5sig_pool(); | |
685 | return 0; | |
686 | ||
687 | clear_hash: | |
688 | tcp_put_md5sig_pool(); | |
689 | clear_hash_noput: | |
690 | memset(md5_hash, 0, 16); | |
691 | return 1; | |
cfb6eeb4 YH |
692 | } |
693 | ||
318cf7aa | 694 | static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb) |
cfb6eeb4 | 695 | { |
cf533ea5 | 696 | const __u8 *hash_location = NULL; |
cfb6eeb4 | 697 | struct tcp_md5sig_key *hash_expected; |
b71d1d42 | 698 | const struct ipv6hdr *ip6h = ipv6_hdr(skb); |
318cf7aa | 699 | const struct tcphdr *th = tcp_hdr(skb); |
cfb6eeb4 | 700 | int genhash; |
cfb6eeb4 YH |
701 | u8 newhash[16]; |
702 | ||
703 | hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr); | |
7d5d5525 | 704 | hash_location = tcp_parse_md5sig_option(th); |
cfb6eeb4 | 705 | |
785957d3 DM |
706 | /* We've parsed the options - do we have a hash? */ |
707 | if (!hash_expected && !hash_location) | |
708 | return 0; | |
709 | ||
710 | if (hash_expected && !hash_location) { | |
711 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); | |
cfb6eeb4 YH |
712 | return 1; |
713 | } | |
714 | ||
785957d3 DM |
715 | if (!hash_expected && hash_location) { |
716 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); | |
cfb6eeb4 YH |
717 | return 1; |
718 | } | |
719 | ||
720 | /* check the signature */ | |
49a72dfb AL |
721 | genhash = tcp_v6_md5_hash_skb(newhash, |
722 | hash_expected, | |
723 | NULL, NULL, skb); | |
724 | ||
cfb6eeb4 YH |
725 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { |
726 | if (net_ratelimit()) { | |
5856b606 | 727 | printk(KERN_INFO "MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n", |
cfb6eeb4 | 728 | genhash ? "failed" : "mismatch", |
0c6ce78a HH |
729 | &ip6h->saddr, ntohs(th->source), |
730 | &ip6h->daddr, ntohs(th->dest)); | |
cfb6eeb4 YH |
731 | } |
732 | return 1; | |
733 | } | |
734 | return 0; | |
735 | } | |
736 | #endif | |
737 | ||
c6aefafb | 738 | struct request_sock_ops tcp6_request_sock_ops __read_mostly = { |
1da177e4 | 739 | .family = AF_INET6, |
2e6599cb | 740 | .obj_size = sizeof(struct tcp6_request_sock), |
72659ecc | 741 | .rtx_syn_ack = tcp_v6_rtx_synack, |
60236fdd ACM |
742 | .send_ack = tcp_v6_reqsk_send_ack, |
743 | .destructor = tcp_v6_reqsk_destructor, | |
72659ecc OP |
744 | .send_reset = tcp_v6_send_reset, |
745 | .syn_ack_timeout = tcp_syn_ack_timeout, | |
1da177e4 LT |
746 | }; |
747 | ||
cfb6eeb4 | 748 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 749 | static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = { |
cfb6eeb4 | 750 | .md5_lookup = tcp_v6_reqsk_md5_lookup, |
e3afe7b7 | 751 | .calc_md5_hash = tcp_v6_md5_hash_skb, |
cfb6eeb4 | 752 | }; |
b6332e6c | 753 | #endif |
cfb6eeb4 | 754 | |
8ad50d96 | 755 | static void __tcp_v6_send_check(struct sk_buff *skb, |
b71d1d42 | 756 | const struct in6_addr *saddr, const struct in6_addr *daddr) |
1da177e4 | 757 | { |
aa8223c7 | 758 | struct tcphdr *th = tcp_hdr(skb); |
1da177e4 | 759 | |
84fa7933 | 760 | if (skb->ip_summed == CHECKSUM_PARTIAL) { |
8ad50d96 | 761 | th->check = ~tcp_v6_check(skb->len, saddr, daddr, 0); |
663ead3b | 762 | skb->csum_start = skb_transport_header(skb) - skb->head; |
ff1dcadb | 763 | skb->csum_offset = offsetof(struct tcphdr, check); |
1da177e4 | 764 | } else { |
8ad50d96 HX |
765 | th->check = tcp_v6_check(skb->len, saddr, daddr, |
766 | csum_partial(th, th->doff << 2, | |
767 | skb->csum)); | |
1da177e4 LT |
768 | } |
769 | } | |
770 | ||
bb296246 | 771 | static void tcp_v6_send_check(struct sock *sk, struct sk_buff *skb) |
8ad50d96 HX |
772 | { |
773 | struct ipv6_pinfo *np = inet6_sk(sk); | |
774 | ||
775 | __tcp_v6_send_check(skb, &np->saddr, &np->daddr); | |
776 | } | |
777 | ||
a430a43d HX |
778 | static int tcp_v6_gso_send_check(struct sk_buff *skb) |
779 | { | |
b71d1d42 | 780 | const struct ipv6hdr *ipv6h; |
a430a43d HX |
781 | struct tcphdr *th; |
782 | ||
783 | if (!pskb_may_pull(skb, sizeof(*th))) | |
784 | return -EINVAL; | |
785 | ||
0660e03f | 786 | ipv6h = ipv6_hdr(skb); |
aa8223c7 | 787 | th = tcp_hdr(skb); |
a430a43d HX |
788 | |
789 | th->check = 0; | |
84fa7933 | 790 | skb->ip_summed = CHECKSUM_PARTIAL; |
8ad50d96 | 791 | __tcp_v6_send_check(skb, &ipv6h->saddr, &ipv6h->daddr); |
a430a43d HX |
792 | return 0; |
793 | } | |
1da177e4 | 794 | |
36990673 HX |
795 | static struct sk_buff **tcp6_gro_receive(struct sk_buff **head, |
796 | struct sk_buff *skb) | |
684f2176 | 797 | { |
b71d1d42 | 798 | const struct ipv6hdr *iph = skb_gro_network_header(skb); |
684f2176 HX |
799 | |
800 | switch (skb->ip_summed) { | |
801 | case CHECKSUM_COMPLETE: | |
86911732 | 802 | if (!tcp_v6_check(skb_gro_len(skb), &iph->saddr, &iph->daddr, |
684f2176 HX |
803 | skb->csum)) { |
804 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
805 | break; | |
806 | } | |
807 | ||
808 | /* fall through */ | |
809 | case CHECKSUM_NONE: | |
810 | NAPI_GRO_CB(skb)->flush = 1; | |
811 | return NULL; | |
812 | } | |
813 | ||
814 | return tcp_gro_receive(head, skb); | |
815 | } | |
684f2176 | 816 | |
36990673 | 817 | static int tcp6_gro_complete(struct sk_buff *skb) |
684f2176 | 818 | { |
b71d1d42 | 819 | const struct ipv6hdr *iph = ipv6_hdr(skb); |
684f2176 HX |
820 | struct tcphdr *th = tcp_hdr(skb); |
821 | ||
822 | th->check = ~tcp_v6_check(skb->len - skb_transport_offset(skb), | |
823 | &iph->saddr, &iph->daddr, 0); | |
824 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6; | |
825 | ||
826 | return tcp_gro_complete(skb); | |
827 | } | |
684f2176 | 828 | |
626e264d | 829 | static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win, |
b903d324 | 830 | u32 ts, struct tcp_md5sig_key *key, int rst, u8 tclass) |
1da177e4 | 831 | { |
cf533ea5 ED |
832 | const struct tcphdr *th = tcp_hdr(skb); |
833 | struct tcphdr *t1; | |
1da177e4 | 834 | struct sk_buff *buff; |
4c9483b2 | 835 | struct flowi6 fl6; |
adf30907 | 836 | struct net *net = dev_net(skb_dst(skb)->dev); |
e5047992 | 837 | struct sock *ctl_sk = net->ipv6.tcp_sk; |
77c676da | 838 | unsigned int tot_len = sizeof(struct tcphdr); |
adf30907 | 839 | struct dst_entry *dst; |
81ada62d | 840 | __be32 *topt; |
1da177e4 | 841 | |
626e264d IJ |
842 | if (ts) |
843 | tot_len += TCPOLEN_TSTAMP_ALIGNED; | |
cfb6eeb4 | 844 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 YH |
845 | if (key) |
846 | tot_len += TCPOLEN_MD5SIG_ALIGNED; | |
847 | #endif | |
848 | ||
cfb6eeb4 | 849 | buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len, |
1da177e4 | 850 | GFP_ATOMIC); |
1ab1457c YH |
851 | if (buff == NULL) |
852 | return; | |
1da177e4 | 853 | |
cfb6eeb4 | 854 | skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len); |
1da177e4 | 855 | |
cfb6eeb4 | 856 | t1 = (struct tcphdr *) skb_push(buff, tot_len); |
6651ffc8 | 857 | skb_reset_transport_header(buff); |
1da177e4 LT |
858 | |
859 | /* Swap the send and the receive. */ | |
860 | memset(t1, 0, sizeof(*t1)); | |
861 | t1->dest = th->source; | |
862 | t1->source = th->dest; | |
cfb6eeb4 | 863 | t1->doff = tot_len / 4; |
626e264d IJ |
864 | t1->seq = htonl(seq); |
865 | t1->ack_seq = htonl(ack); | |
866 | t1->ack = !rst || !th->ack; | |
867 | t1->rst = rst; | |
868 | t1->window = htons(win); | |
1da177e4 | 869 | |
81ada62d IJ |
870 | topt = (__be32 *)(t1 + 1); |
871 | ||
626e264d IJ |
872 | if (ts) { |
873 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | | |
874 | (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP); | |
875 | *topt++ = htonl(tcp_time_stamp); | |
876 | *topt++ = htonl(ts); | |
877 | } | |
878 | ||
cfb6eeb4 YH |
879 | #ifdef CONFIG_TCP_MD5SIG |
880 | if (key) { | |
81ada62d IJ |
881 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
882 | (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG); | |
883 | tcp_v6_md5_hash_hdr((__u8 *)topt, key, | |
78e645cb IJ |
884 | &ipv6_hdr(skb)->saddr, |
885 | &ipv6_hdr(skb)->daddr, t1); | |
cfb6eeb4 YH |
886 | } |
887 | #endif | |
888 | ||
4c9483b2 | 889 | memset(&fl6, 0, sizeof(fl6)); |
4e3fd7a0 AD |
890 | fl6.daddr = ipv6_hdr(skb)->saddr; |
891 | fl6.saddr = ipv6_hdr(skb)->daddr; | |
1da177e4 | 892 | |
e5700aff DM |
893 | buff->ip_summed = CHECKSUM_PARTIAL; |
894 | buff->csum = 0; | |
895 | ||
4c9483b2 | 896 | __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr); |
1da177e4 | 897 | |
4c9483b2 DM |
898 | fl6.flowi6_proto = IPPROTO_TCP; |
899 | fl6.flowi6_oif = inet6_iif(skb); | |
1958b856 DM |
900 | fl6.fl6_dport = t1->dest; |
901 | fl6.fl6_sport = t1->source; | |
4c9483b2 | 902 | security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); |
1da177e4 | 903 | |
c20121ae DL |
904 | /* Pass a socket to ip6_dst_lookup either it is for RST |
905 | * Underlying function will use this to retrieve the network | |
906 | * namespace | |
907 | */ | |
4c9483b2 | 908 | dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL, false); |
68d0c6d3 DM |
909 | if (!IS_ERR(dst)) { |
910 | skb_dst_set(buff, dst); | |
b903d324 | 911 | ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass); |
68d0c6d3 DM |
912 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
913 | if (rst) | |
914 | TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS); | |
915 | return; | |
1da177e4 LT |
916 | } |
917 | ||
918 | kfree_skb(buff); | |
919 | } | |
920 | ||
626e264d | 921 | static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 922 | { |
cf533ea5 | 923 | const struct tcphdr *th = tcp_hdr(skb); |
626e264d | 924 | u32 seq = 0, ack_seq = 0; |
fa3e5b4e | 925 | struct tcp_md5sig_key *key = NULL; |
1da177e4 | 926 | |
626e264d | 927 | if (th->rst) |
1da177e4 LT |
928 | return; |
929 | ||
626e264d IJ |
930 | if (!ipv6_unicast_destination(skb)) |
931 | return; | |
1da177e4 | 932 | |
cfb6eeb4 | 933 | #ifdef CONFIG_TCP_MD5SIG |
626e264d | 934 | if (sk) |
8a622e71 | 935 | key = tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->saddr); |
cfb6eeb4 YH |
936 | #endif |
937 | ||
626e264d IJ |
938 | if (th->ack) |
939 | seq = ntohl(th->ack_seq); | |
940 | else | |
941 | ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len - | |
942 | (th->doff << 2); | |
1da177e4 | 943 | |
b903d324 | 944 | tcp_v6_send_response(skb, seq, ack_seq, 0, 0, key, 1, 0); |
626e264d | 945 | } |
1da177e4 | 946 | |
626e264d | 947 | static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts, |
b903d324 | 948 | struct tcp_md5sig_key *key, u8 tclass) |
626e264d | 949 | { |
b903d324 | 950 | tcp_v6_send_response(skb, seq, ack, win, ts, key, 0, tclass); |
1da177e4 LT |
951 | } |
952 | ||
953 | static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
954 | { | |
8feaf0c0 | 955 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 956 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 957 | |
9501f972 | 958 | tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
8feaf0c0 | 959 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
b903d324 ED |
960 | tcptw->tw_ts_recent, tcp_twsk_md5_key(tcptw), |
961 | tw->tw_tclass); | |
1da177e4 | 962 | |
8feaf0c0 | 963 | inet_twsk_put(tw); |
1da177e4 LT |
964 | } |
965 | ||
6edafaaf GJ |
966 | static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
967 | struct request_sock *req) | |
1da177e4 | 968 | { |
9501f972 | 969 | tcp_v6_send_ack(skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, req->ts_recent, |
b903d324 | 970 | tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr), 0); |
1da177e4 LT |
971 | } |
972 | ||
973 | ||
974 | static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb) | |
975 | { | |
60236fdd | 976 | struct request_sock *req, **prev; |
aa8223c7 | 977 | const struct tcphdr *th = tcp_hdr(skb); |
1da177e4 LT |
978 | struct sock *nsk; |
979 | ||
980 | /* Find possible connection requests. */ | |
8129765a | 981 | req = inet6_csk_search_req(sk, &prev, th->source, |
0660e03f ACM |
982 | &ipv6_hdr(skb)->saddr, |
983 | &ipv6_hdr(skb)->daddr, inet6_iif(skb)); | |
1da177e4 LT |
984 | if (req) |
985 | return tcp_check_req(sk, skb, req, prev); | |
986 | ||
3b1e0a65 | 987 | nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo, |
d86e0dac PE |
988 | &ipv6_hdr(skb)->saddr, th->source, |
989 | &ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb)); | |
1da177e4 LT |
990 | |
991 | if (nsk) { | |
992 | if (nsk->sk_state != TCP_TIME_WAIT) { | |
993 | bh_lock_sock(nsk); | |
994 | return nsk; | |
995 | } | |
9469c7b4 | 996 | inet_twsk_put(inet_twsk(nsk)); |
1da177e4 LT |
997 | return NULL; |
998 | } | |
999 | ||
c6aefafb | 1000 | #ifdef CONFIG_SYN_COOKIES |
af9b4738 | 1001 | if (!th->syn) |
c6aefafb | 1002 | sk = cookie_v6_check(sk, skb); |
1da177e4 LT |
1003 | #endif |
1004 | return sk; | |
1005 | } | |
1006 | ||
1da177e4 LT |
1007 | /* FIXME: this is substantially similar to the ipv4 code. |
1008 | * Can some kind of merge be done? -- erics | |
1009 | */ | |
1010 | static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb) | |
1011 | { | |
4957faad | 1012 | struct tcp_extend_values tmp_ext; |
e6b4d113 | 1013 | struct tcp_options_received tmp_opt; |
cf533ea5 | 1014 | const u8 *hash_location; |
e6b4d113 | 1015 | struct request_sock *req; |
ca304b61 | 1016 | struct inet6_request_sock *treq; |
1da177e4 | 1017 | struct ipv6_pinfo *np = inet6_sk(sk); |
1da177e4 | 1018 | struct tcp_sock *tp = tcp_sk(sk); |
e6b4d113 | 1019 | __u32 isn = TCP_SKB_CB(skb)->when; |
493f377d | 1020 | struct dst_entry *dst = NULL; |
c6aefafb | 1021 | int want_cookie = 0; |
1da177e4 LT |
1022 | |
1023 | if (skb->protocol == htons(ETH_P_IP)) | |
1024 | return tcp_v4_conn_request(sk, skb); | |
1025 | ||
1026 | if (!ipv6_unicast_destination(skb)) | |
1ab1457c | 1027 | goto drop; |
1da177e4 | 1028 | |
463c84b9 | 1029 | if (inet_csk_reqsk_queue_is_full(sk) && !isn) { |
946cedcc ED |
1030 | want_cookie = tcp_syn_flood_action(sk, skb, "TCPv6"); |
1031 | if (!want_cookie) | |
1032 | goto drop; | |
1da177e4 LT |
1033 | } |
1034 | ||
463c84b9 | 1035 | if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1) |
1da177e4 LT |
1036 | goto drop; |
1037 | ||
ca304b61 | 1038 | req = inet6_reqsk_alloc(&tcp6_request_sock_ops); |
1da177e4 LT |
1039 | if (req == NULL) |
1040 | goto drop; | |
1041 | ||
cfb6eeb4 YH |
1042 | #ifdef CONFIG_TCP_MD5SIG |
1043 | tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops; | |
1044 | #endif | |
1045 | ||
1da177e4 LT |
1046 | tcp_clear_options(&tmp_opt); |
1047 | tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); | |
1048 | tmp_opt.user_mss = tp->rx_opt.user_mss; | |
bb5b7c11 | 1049 | tcp_parse_options(skb, &tmp_opt, &hash_location, 0); |
4957faad WAS |
1050 | |
1051 | if (tmp_opt.cookie_plus > 0 && | |
1052 | tmp_opt.saw_tstamp && | |
1053 | !tp->rx_opt.cookie_out_never && | |
1054 | (sysctl_tcp_cookie_size > 0 || | |
1055 | (tp->cookie_values != NULL && | |
1056 | tp->cookie_values->cookie_desired > 0))) { | |
1057 | u8 *c; | |
1058 | u32 *d; | |
1059 | u32 *mess = &tmp_ext.cookie_bakery[COOKIE_DIGEST_WORDS]; | |
1060 | int l = tmp_opt.cookie_plus - TCPOLEN_COOKIE_BASE; | |
1061 | ||
1062 | if (tcp_cookie_generator(&tmp_ext.cookie_bakery[0]) != 0) | |
1063 | goto drop_and_free; | |
1064 | ||
1065 | /* Secret recipe starts with IP addresses */ | |
0eae88f3 | 1066 | d = (__force u32 *)&ipv6_hdr(skb)->daddr.s6_addr32[0]; |
4957faad WAS |
1067 | *mess++ ^= *d++; |
1068 | *mess++ ^= *d++; | |
1069 | *mess++ ^= *d++; | |
1070 | *mess++ ^= *d++; | |
0eae88f3 | 1071 | d = (__force u32 *)&ipv6_hdr(skb)->saddr.s6_addr32[0]; |
4957faad WAS |
1072 | *mess++ ^= *d++; |
1073 | *mess++ ^= *d++; | |
1074 | *mess++ ^= *d++; | |
1075 | *mess++ ^= *d++; | |
1076 | ||
1077 | /* plus variable length Initiator Cookie */ | |
1078 | c = (u8 *)mess; | |
1079 | while (l-- > 0) | |
1080 | *c++ ^= *hash_location++; | |
1da177e4 | 1081 | |
4957faad | 1082 | want_cookie = 0; /* not our kind of cookie */ |
4957faad WAS |
1083 | tmp_ext.cookie_out_never = 0; /* false */ |
1084 | tmp_ext.cookie_plus = tmp_opt.cookie_plus; | |
1085 | } else if (!tp->rx_opt.cookie_in_always) { | |
1086 | /* redundant indications, but ensure initialization. */ | |
1087 | tmp_ext.cookie_out_never = 1; /* true */ | |
1088 | tmp_ext.cookie_plus = 0; | |
1089 | } else { | |
1090 | goto drop_and_free; | |
1091 | } | |
1092 | tmp_ext.cookie_in_always = tp->rx_opt.cookie_in_always; | |
1da177e4 | 1093 | |
4dfc2817 | 1094 | if (want_cookie && !tmp_opt.saw_tstamp) |
c6aefafb | 1095 | tcp_clear_options(&tmp_opt); |
c6aefafb | 1096 | |
1da177e4 LT |
1097 | tmp_opt.tstamp_ok = tmp_opt.saw_tstamp; |
1098 | tcp_openreq_init(req, &tmp_opt, skb); | |
1099 | ||
ca304b61 | 1100 | treq = inet6_rsk(req); |
4e3fd7a0 AD |
1101 | treq->rmt_addr = ipv6_hdr(skb)->saddr; |
1102 | treq->loc_addr = ipv6_hdr(skb)->daddr; | |
172d69e6 | 1103 | if (!want_cookie || tmp_opt.tstamp_ok) |
c6aefafb GG |
1104 | TCP_ECN_create_request(req, tcp_hdr(skb)); |
1105 | ||
4d0fe50c ED |
1106 | treq->iif = sk->sk_bound_dev_if; |
1107 | ||
1108 | /* So that link locals have meaning */ | |
1109 | if (!sk->sk_bound_dev_if && | |
1110 | ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL) | |
1111 | treq->iif = inet6_iif(skb); | |
1112 | ||
2bbdf389 | 1113 | if (!isn) { |
493f377d DM |
1114 | struct inet_peer *peer = NULL; |
1115 | ||
c6aefafb GG |
1116 | if (ipv6_opt_accepted(sk, skb) || |
1117 | np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo || | |
1118 | np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) { | |
1119 | atomic_inc(&skb->users); | |
1120 | treq->pktopts = skb; | |
1121 | } | |
493f377d DM |
1122 | |
1123 | if (want_cookie) { | |
2bbdf389 FW |
1124 | isn = cookie_v6_init_sequence(sk, skb, &req->mss); |
1125 | req->cookie_ts = tmp_opt.tstamp_ok; | |
493f377d DM |
1126 | goto have_isn; |
1127 | } | |
1128 | ||
1129 | /* VJ's idea. We save last timestamp seen | |
1130 | * from the destination in peer table, when entering | |
1131 | * state TIME-WAIT, and check against it before | |
1132 | * accepting new connection request. | |
1133 | * | |
1134 | * If "isn" is not zero, this request hit alive | |
1135 | * timewait bucket, so that all the necessary checks | |
1136 | * are made in the function processing timewait state. | |
1137 | */ | |
1138 | if (tmp_opt.saw_tstamp && | |
1139 | tcp_death_row.sysctl_tw_recycle && | |
1140 | (dst = inet6_csk_route_req(sk, req)) != NULL && | |
1141 | (peer = rt6_get_peer((struct rt6_info *)dst)) != NULL && | |
7a71ed89 | 1142 | ipv6_addr_equal((struct in6_addr *)peer->daddr.addr.a6, |
493f377d DM |
1143 | &treq->rmt_addr)) { |
1144 | inet_peer_refcheck(peer); | |
1145 | if ((u32)get_seconds() - peer->tcp_ts_stamp < TCP_PAWS_MSL && | |
1146 | (s32)(peer->tcp_ts - req->ts_recent) > | |
1147 | TCP_PAWS_WINDOW) { | |
1148 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED); | |
1149 | goto drop_and_release; | |
1150 | } | |
1151 | } | |
1152 | /* Kill the following clause, if you dislike this way. */ | |
1153 | else if (!sysctl_tcp_syncookies && | |
1154 | (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) < | |
1155 | (sysctl_max_syn_backlog >> 2)) && | |
1156 | (!peer || !peer->tcp_ts_stamp) && | |
1157 | (!dst || !dst_metric(dst, RTAX_RTT))) { | |
1158 | /* Without syncookies last quarter of | |
1159 | * backlog is filled with destinations, | |
1160 | * proven to be alive. | |
1161 | * It means that we continue to communicate | |
1162 | * to destinations, already remembered | |
1163 | * to the moment of synflood. | |
1164 | */ | |
1165 | LIMIT_NETDEBUG(KERN_DEBUG "TCP: drop open request from %pI6/%u\n", | |
1166 | &treq->rmt_addr, ntohs(tcp_hdr(skb)->source)); | |
1167 | goto drop_and_release; | |
2bbdf389 | 1168 | } |
493f377d DM |
1169 | |
1170 | isn = tcp_v6_init_sequence(skb); | |
c6aefafb | 1171 | } |
493f377d | 1172 | have_isn: |
2e6599cb | 1173 | tcp_rsk(req)->snt_isn = isn; |
9ad7c049 | 1174 | tcp_rsk(req)->snt_synack = tcp_time_stamp; |
1da177e4 | 1175 | |
4237c75c VY |
1176 | security_inet_conn_request(sk, skb, req); |
1177 | ||
4957faad WAS |
1178 | if (tcp_v6_send_synack(sk, req, |
1179 | (struct request_values *)&tmp_ext) || | |
1180 | want_cookie) | |
e6b4d113 | 1181 | goto drop_and_free; |
1da177e4 | 1182 | |
e6b4d113 WAS |
1183 | inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT); |
1184 | return 0; | |
1da177e4 | 1185 | |
493f377d DM |
1186 | drop_and_release: |
1187 | dst_release(dst); | |
e6b4d113 WAS |
1188 | drop_and_free: |
1189 | reqsk_free(req); | |
1da177e4 | 1190 | drop: |
1da177e4 LT |
1191 | return 0; /* don't send reset */ |
1192 | } | |
1193 | ||
1194 | static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb, | |
60236fdd | 1195 | struct request_sock *req, |
1da177e4 LT |
1196 | struct dst_entry *dst) |
1197 | { | |
78d15e82 | 1198 | struct inet6_request_sock *treq; |
1da177e4 LT |
1199 | struct ipv6_pinfo *newnp, *np = inet6_sk(sk); |
1200 | struct tcp6_sock *newtcp6sk; | |
1201 | struct inet_sock *newinet; | |
1202 | struct tcp_sock *newtp; | |
1203 | struct sock *newsk; | |
1204 | struct ipv6_txoptions *opt; | |
cfb6eeb4 YH |
1205 | #ifdef CONFIG_TCP_MD5SIG |
1206 | struct tcp_md5sig_key *key; | |
1207 | #endif | |
1da177e4 LT |
1208 | |
1209 | if (skb->protocol == htons(ETH_P_IP)) { | |
1210 | /* | |
1211 | * v6 mapped | |
1212 | */ | |
1213 | ||
1214 | newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst); | |
1215 | ||
1ab1457c | 1216 | if (newsk == NULL) |
1da177e4 LT |
1217 | return NULL; |
1218 | ||
1219 | newtcp6sk = (struct tcp6_sock *)newsk; | |
1220 | inet_sk(newsk)->pinet6 = &newtcp6sk->inet6; | |
1221 | ||
1222 | newinet = inet_sk(newsk); | |
1223 | newnp = inet6_sk(newsk); | |
1224 | newtp = tcp_sk(newsk); | |
1225 | ||
1226 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1227 | ||
c720c7e8 | 1228 | ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr); |
1da177e4 | 1229 | |
c720c7e8 | 1230 | ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr); |
1da177e4 | 1231 | |
4e3fd7a0 | 1232 | newnp->rcv_saddr = newnp->saddr; |
1da177e4 | 1233 | |
8292a17a | 1234 | inet_csk(newsk)->icsk_af_ops = &ipv6_mapped; |
1da177e4 | 1235 | newsk->sk_backlog_rcv = tcp_v4_do_rcv; |
cfb6eeb4 YH |
1236 | #ifdef CONFIG_TCP_MD5SIG |
1237 | newtp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
1238 | #endif | |
1239 | ||
676a1184 YZ |
1240 | newnp->ipv6_ac_list = NULL; |
1241 | newnp->ipv6_fl_list = NULL; | |
1da177e4 LT |
1242 | newnp->pktoptions = NULL; |
1243 | newnp->opt = NULL; | |
505cbfc5 | 1244 | newnp->mcast_oif = inet6_iif(skb); |
0660e03f | 1245 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; |
1da177e4 | 1246 | |
e6848976 ACM |
1247 | /* |
1248 | * No need to charge this sock to the relevant IPv6 refcnt debug socks count | |
1249 | * here, tcp_create_openreq_child now does this for us, see the comment in | |
1250 | * that function for the gory details. -acme | |
1da177e4 | 1251 | */ |
1da177e4 LT |
1252 | |
1253 | /* It is tricky place. Until this moment IPv4 tcp | |
8292a17a | 1254 | worked with IPv6 icsk.icsk_af_ops. |
1da177e4 LT |
1255 | Sync it now. |
1256 | */ | |
d83d8461 | 1257 | tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie); |
1da177e4 LT |
1258 | |
1259 | return newsk; | |
1260 | } | |
1261 | ||
78d15e82 | 1262 | treq = inet6_rsk(req); |
1da177e4 LT |
1263 | opt = np->opt; |
1264 | ||
1265 | if (sk_acceptq_is_full(sk)) | |
1266 | goto out_overflow; | |
1267 | ||
493f377d DM |
1268 | if (!dst) { |
1269 | dst = inet6_csk_route_req(sk, req); | |
1270 | if (!dst) | |
1da177e4 | 1271 | goto out; |
1ab1457c | 1272 | } |
1da177e4 LT |
1273 | |
1274 | newsk = tcp_create_openreq_child(sk, req, skb); | |
1275 | if (newsk == NULL) | |
093d2823 | 1276 | goto out_nonewsk; |
1da177e4 | 1277 | |
e6848976 ACM |
1278 | /* |
1279 | * No need to charge this sock to the relevant IPv6 refcnt debug socks | |
1280 | * count here, tcp_create_openreq_child now does this for us, see the | |
1281 | * comment in that function for the gory details. -acme | |
1282 | */ | |
1da177e4 | 1283 | |
59eed279 | 1284 | newsk->sk_gso_type = SKB_GSO_TCPV6; |
8e1ef0a9 | 1285 | __ip6_dst_store(newsk, dst, NULL, NULL); |
1da177e4 LT |
1286 | |
1287 | newtcp6sk = (struct tcp6_sock *)newsk; | |
1288 | inet_sk(newsk)->pinet6 = &newtcp6sk->inet6; | |
1289 | ||
1290 | newtp = tcp_sk(newsk); | |
1291 | newinet = inet_sk(newsk); | |
1292 | newnp = inet6_sk(newsk); | |
1293 | ||
1294 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1295 | ||
4e3fd7a0 AD |
1296 | newnp->daddr = treq->rmt_addr; |
1297 | newnp->saddr = treq->loc_addr; | |
1298 | newnp->rcv_saddr = treq->loc_addr; | |
2e6599cb | 1299 | newsk->sk_bound_dev_if = treq->iif; |
1da177e4 | 1300 | |
1ab1457c | 1301 | /* Now IPv6 options... |
1da177e4 LT |
1302 | |
1303 | First: no IPv4 options. | |
1304 | */ | |
f6d8bd05 | 1305 | newinet->inet_opt = NULL; |
676a1184 | 1306 | newnp->ipv6_ac_list = NULL; |
d35690be | 1307 | newnp->ipv6_fl_list = NULL; |
1da177e4 LT |
1308 | |
1309 | /* Clone RX bits */ | |
1310 | newnp->rxopt.all = np->rxopt.all; | |
1311 | ||
1312 | /* Clone pktoptions received with SYN */ | |
1313 | newnp->pktoptions = NULL; | |
2e6599cb ACM |
1314 | if (treq->pktopts != NULL) { |
1315 | newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC); | |
1316 | kfree_skb(treq->pktopts); | |
1317 | treq->pktopts = NULL; | |
1da177e4 LT |
1318 | if (newnp->pktoptions) |
1319 | skb_set_owner_r(newnp->pktoptions, newsk); | |
1320 | } | |
1321 | newnp->opt = NULL; | |
505cbfc5 | 1322 | newnp->mcast_oif = inet6_iif(skb); |
0660e03f | 1323 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; |
1da177e4 LT |
1324 | |
1325 | /* Clone native IPv6 options from listening socket (if any) | |
1326 | ||
1327 | Yes, keeping reference count would be much more clever, | |
1328 | but we make one more one thing there: reattach optmem | |
1329 | to newsk. | |
1330 | */ | |
1331 | if (opt) { | |
1332 | newnp->opt = ipv6_dup_options(newsk, opt); | |
1333 | if (opt != np->opt) | |
1334 | sock_kfree_s(sk, opt, opt->tot_len); | |
1335 | } | |
1336 | ||
d83d8461 | 1337 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
1da177e4 | 1338 | if (newnp->opt) |
d83d8461 ACM |
1339 | inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + |
1340 | newnp->opt->opt_flen); | |
1da177e4 | 1341 | |
5d424d5a | 1342 | tcp_mtup_init(newsk); |
1da177e4 | 1343 | tcp_sync_mss(newsk, dst_mtu(dst)); |
0dbaee3b | 1344 | newtp->advmss = dst_metric_advmss(dst); |
1da177e4 | 1345 | tcp_initialize_rcv_mss(newsk); |
9ad7c049 JC |
1346 | if (tcp_rsk(req)->snt_synack) |
1347 | tcp_valid_rtt_meas(newsk, | |
1348 | tcp_time_stamp - tcp_rsk(req)->snt_synack); | |
1349 | newtp->total_retrans = req->retrans; | |
1da177e4 | 1350 | |
c720c7e8 ED |
1351 | newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6; |
1352 | newinet->inet_rcv_saddr = LOOPBACK4_IPV6; | |
1da177e4 | 1353 | |
cfb6eeb4 YH |
1354 | #ifdef CONFIG_TCP_MD5SIG |
1355 | /* Copy over the MD5 key from the original socket */ | |
1356 | if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) { | |
1357 | /* We're using one, so create a matching key | |
1358 | * on the newsk structure. If we fail to get | |
1359 | * memory, then we end up not copying the key | |
1360 | * across. Shucks. | |
1361 | */ | |
a915da9b ED |
1362 | tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newnp->daddr, |
1363 | AF_INET6, key->key, key->keylen, GFP_ATOMIC); | |
cfb6eeb4 YH |
1364 | } |
1365 | #endif | |
1366 | ||
093d2823 BS |
1367 | if (__inet_inherit_port(sk, newsk) < 0) { |
1368 | sock_put(newsk); | |
1369 | goto out; | |
1370 | } | |
9327f705 | 1371 | __inet6_hash(newsk, NULL); |
1da177e4 LT |
1372 | |
1373 | return newsk; | |
1374 | ||
1375 | out_overflow: | |
de0744af | 1376 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 | 1377 | out_nonewsk: |
1da177e4 LT |
1378 | if (opt && opt != np->opt) |
1379 | sock_kfree_s(sk, opt, opt->tot_len); | |
1380 | dst_release(dst); | |
093d2823 BS |
1381 | out: |
1382 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); | |
1da177e4 LT |
1383 | return NULL; |
1384 | } | |
1385 | ||
b51655b9 | 1386 | static __sum16 tcp_v6_checksum_init(struct sk_buff *skb) |
1da177e4 | 1387 | { |
84fa7933 | 1388 | if (skb->ip_summed == CHECKSUM_COMPLETE) { |
684f2176 | 1389 | if (!tcp_v6_check(skb->len, &ipv6_hdr(skb)->saddr, |
0660e03f | 1390 | &ipv6_hdr(skb)->daddr, skb->csum)) { |
fb286bb2 | 1391 | skb->ip_summed = CHECKSUM_UNNECESSARY; |
1da177e4 | 1392 | return 0; |
fb286bb2 | 1393 | } |
1da177e4 | 1394 | } |
fb286bb2 | 1395 | |
684f2176 | 1396 | skb->csum = ~csum_unfold(tcp_v6_check(skb->len, |
0660e03f ACM |
1397 | &ipv6_hdr(skb)->saddr, |
1398 | &ipv6_hdr(skb)->daddr, 0)); | |
fb286bb2 | 1399 | |
1da177e4 | 1400 | if (skb->len <= 76) { |
fb286bb2 | 1401 | return __skb_checksum_complete(skb); |
1da177e4 LT |
1402 | } |
1403 | return 0; | |
1404 | } | |
1405 | ||
1406 | /* The socket must have it's spinlock held when we get | |
1407 | * here. | |
1408 | * | |
1409 | * We have a potential double-lock case here, so even when | |
1410 | * doing backlog processing we use the BH locking scheme. | |
1411 | * This is because we cannot sleep with the original spinlock | |
1412 | * held. | |
1413 | */ | |
1414 | static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1415 | { | |
1416 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1417 | struct tcp_sock *tp; | |
1418 | struct sk_buff *opt_skb = NULL; | |
1419 | ||
1420 | /* Imagine: socket is IPv6. IPv4 packet arrives, | |
1421 | goes to IPv4 receive handler and backlogged. | |
1422 | From backlog it always goes here. Kerboom... | |
1423 | Fortunately, tcp_rcv_established and rcv_established | |
1424 | handle them correctly, but it is not case with | |
1425 | tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK | |
1426 | */ | |
1427 | ||
1428 | if (skb->protocol == htons(ETH_P_IP)) | |
1429 | return tcp_v4_do_rcv(sk, skb); | |
1430 | ||
cfb6eeb4 YH |
1431 | #ifdef CONFIG_TCP_MD5SIG |
1432 | if (tcp_v6_inbound_md5_hash (sk, skb)) | |
1433 | goto discard; | |
1434 | #endif | |
1435 | ||
fda9ef5d | 1436 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1437 | goto discard; |
1438 | ||
1439 | /* | |
1440 | * socket locking is here for SMP purposes as backlog rcv | |
1441 | * is currently called with bh processing disabled. | |
1442 | */ | |
1443 | ||
1444 | /* Do Stevens' IPV6_PKTOPTIONS. | |
1445 | ||
1446 | Yes, guys, it is the only place in our code, where we | |
1447 | may make it not affecting IPv4. | |
1448 | The rest of code is protocol independent, | |
1449 | and I do not like idea to uglify IPv4. | |
1450 | ||
1451 | Actually, all the idea behind IPV6_PKTOPTIONS | |
1452 | looks not very well thought. For now we latch | |
1453 | options, received in the last packet, enqueued | |
1454 | by tcp. Feel free to propose better solution. | |
1ab1457c | 1455 | --ANK (980728) |
1da177e4 LT |
1456 | */ |
1457 | if (np->rxopt.all) | |
1458 | opt_skb = skb_clone(skb, GFP_ATOMIC); | |
1459 | ||
1460 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ | |
bdeab991 | 1461 | sock_rps_save_rxhash(sk, skb); |
aa8223c7 | 1462 | if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len)) |
1da177e4 | 1463 | goto reset; |
1da177e4 LT |
1464 | if (opt_skb) |
1465 | goto ipv6_pktoptions; | |
1466 | return 0; | |
1467 | } | |
1468 | ||
ab6a5bb6 | 1469 | if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb)) |
1da177e4 LT |
1470 | goto csum_err; |
1471 | ||
1ab1457c | 1472 | if (sk->sk_state == TCP_LISTEN) { |
1da177e4 LT |
1473 | struct sock *nsk = tcp_v6_hnd_req(sk, skb); |
1474 | if (!nsk) | |
1475 | goto discard; | |
1476 | ||
1477 | /* | |
1478 | * Queue it on the new socket if the new socket is active, | |
1479 | * otherwise we just shortcircuit this and continue with | |
1480 | * the new socket.. | |
1481 | */ | |
1ab1457c | 1482 | if(nsk != sk) { |
bdeab991 | 1483 | sock_rps_save_rxhash(nsk, skb); |
1da177e4 LT |
1484 | if (tcp_child_process(sk, nsk, skb)) |
1485 | goto reset; | |
1486 | if (opt_skb) | |
1487 | __kfree_skb(opt_skb); | |
1488 | return 0; | |
1489 | } | |
47482f13 | 1490 | } else |
bdeab991 | 1491 | sock_rps_save_rxhash(sk, skb); |
1da177e4 | 1492 | |
aa8223c7 | 1493 | if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) |
1da177e4 | 1494 | goto reset; |
1da177e4 LT |
1495 | if (opt_skb) |
1496 | goto ipv6_pktoptions; | |
1497 | return 0; | |
1498 | ||
1499 | reset: | |
cfb6eeb4 | 1500 | tcp_v6_send_reset(sk, skb); |
1da177e4 LT |
1501 | discard: |
1502 | if (opt_skb) | |
1503 | __kfree_skb(opt_skb); | |
1504 | kfree_skb(skb); | |
1505 | return 0; | |
1506 | csum_err: | |
63231bdd | 1507 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS); |
1da177e4 LT |
1508 | goto discard; |
1509 | ||
1510 | ||
1511 | ipv6_pktoptions: | |
1512 | /* Do you ask, what is it? | |
1513 | ||
1514 | 1. skb was enqueued by tcp. | |
1515 | 2. skb is added to tail of read queue, rather than out of order. | |
1516 | 3. socket is not in passive state. | |
1517 | 4. Finally, it really contains options, which user wants to receive. | |
1518 | */ | |
1519 | tp = tcp_sk(sk); | |
1520 | if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt && | |
1521 | !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) { | |
333fad53 | 1522 | if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo) |
505cbfc5 | 1523 | np->mcast_oif = inet6_iif(opt_skb); |
333fad53 | 1524 | if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) |
0660e03f | 1525 | np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit; |
1da177e4 LT |
1526 | if (ipv6_opt_accepted(sk, opt_skb)) { |
1527 | skb_set_owner_r(opt_skb, sk); | |
1528 | opt_skb = xchg(&np->pktoptions, opt_skb); | |
1529 | } else { | |
1530 | __kfree_skb(opt_skb); | |
1531 | opt_skb = xchg(&np->pktoptions, NULL); | |
1532 | } | |
1533 | } | |
1534 | ||
800d55f1 | 1535 | kfree_skb(opt_skb); |
1da177e4 LT |
1536 | return 0; |
1537 | } | |
1538 | ||
e5bbef20 | 1539 | static int tcp_v6_rcv(struct sk_buff *skb) |
1da177e4 | 1540 | { |
cf533ea5 | 1541 | const struct tcphdr *th; |
b71d1d42 | 1542 | const struct ipv6hdr *hdr; |
1da177e4 LT |
1543 | struct sock *sk; |
1544 | int ret; | |
a86b1e30 | 1545 | struct net *net = dev_net(skb->dev); |
1da177e4 LT |
1546 | |
1547 | if (skb->pkt_type != PACKET_HOST) | |
1548 | goto discard_it; | |
1549 | ||
1550 | /* | |
1551 | * Count it even if it's bad. | |
1552 | */ | |
63231bdd | 1553 | TCP_INC_STATS_BH(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1554 | |
1555 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1556 | goto discard_it; | |
1557 | ||
aa8223c7 | 1558 | th = tcp_hdr(skb); |
1da177e4 LT |
1559 | |
1560 | if (th->doff < sizeof(struct tcphdr)/4) | |
1561 | goto bad_packet; | |
1562 | if (!pskb_may_pull(skb, th->doff*4)) | |
1563 | goto discard_it; | |
1564 | ||
60476372 | 1565 | if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb)) |
1da177e4 LT |
1566 | goto bad_packet; |
1567 | ||
aa8223c7 | 1568 | th = tcp_hdr(skb); |
e802af9c | 1569 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
1570 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); |
1571 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1572 | skb->len - th->doff*4); | |
1573 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
1574 | TCP_SKB_CB(skb)->when = 0; | |
b82d1bb4 | 1575 | TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr); |
1da177e4 LT |
1576 | TCP_SKB_CB(skb)->sacked = 0; |
1577 | ||
9a1f27c4 | 1578 | sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest); |
1da177e4 LT |
1579 | if (!sk) |
1580 | goto no_tcp_socket; | |
1581 | ||
1582 | process: | |
1583 | if (sk->sk_state == TCP_TIME_WAIT) | |
1584 | goto do_time_wait; | |
1585 | ||
e802af9c SH |
1586 | if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) { |
1587 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
1588 | goto discard_and_relse; | |
1589 | } | |
1590 | ||
1da177e4 LT |
1591 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) |
1592 | goto discard_and_relse; | |
1593 | ||
fda9ef5d | 1594 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1595 | goto discard_and_relse; |
1596 | ||
1597 | skb->dev = NULL; | |
1598 | ||
293b9c42 | 1599 | bh_lock_sock_nested(sk); |
1da177e4 LT |
1600 | ret = 0; |
1601 | if (!sock_owned_by_user(sk)) { | |
1a2449a8 | 1602 | #ifdef CONFIG_NET_DMA |
1ab1457c | 1603 | struct tcp_sock *tp = tcp_sk(sk); |
b4caea8a | 1604 | if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list) |
f67b4599 | 1605 | tp->ucopy.dma_chan = dma_find_channel(DMA_MEMCPY); |
1ab1457c YH |
1606 | if (tp->ucopy.dma_chan) |
1607 | ret = tcp_v6_do_rcv(sk, skb); | |
1608 | else | |
1a2449a8 CL |
1609 | #endif |
1610 | { | |
1611 | if (!tcp_prequeue(sk, skb)) | |
1612 | ret = tcp_v6_do_rcv(sk, skb); | |
1613 | } | |
6cce09f8 | 1614 | } else if (unlikely(sk_add_backlog(sk, skb))) { |
6b03a53a | 1615 | bh_unlock_sock(sk); |
6cce09f8 | 1616 | NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP); |
6b03a53a ZY |
1617 | goto discard_and_relse; |
1618 | } | |
1da177e4 LT |
1619 | bh_unlock_sock(sk); |
1620 | ||
1621 | sock_put(sk); | |
1622 | return ret ? -1 : 0; | |
1623 | ||
1624 | no_tcp_socket: | |
1625 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1626 | goto discard_it; | |
1627 | ||
1628 | if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) { | |
1629 | bad_packet: | |
63231bdd | 1630 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
1da177e4 | 1631 | } else { |
cfb6eeb4 | 1632 | tcp_v6_send_reset(NULL, skb); |
1da177e4 LT |
1633 | } |
1634 | ||
1635 | discard_it: | |
1636 | ||
1637 | /* | |
1638 | * Discard frame | |
1639 | */ | |
1640 | ||
1641 | kfree_skb(skb); | |
1642 | return 0; | |
1643 | ||
1644 | discard_and_relse: | |
1645 | sock_put(sk); | |
1646 | goto discard_it; | |
1647 | ||
1648 | do_time_wait: | |
1649 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1650 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1651 | goto discard_it; |
1652 | } | |
1653 | ||
1654 | if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) { | |
63231bdd | 1655 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
9469c7b4 | 1656 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1657 | goto discard_it; |
1658 | } | |
1659 | ||
9469c7b4 | 1660 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 LT |
1661 | case TCP_TW_SYN: |
1662 | { | |
1663 | struct sock *sk2; | |
1664 | ||
c346dca1 | 1665 | sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo, |
0660e03f | 1666 | &ipv6_hdr(skb)->daddr, |
505cbfc5 | 1667 | ntohs(th->dest), inet6_iif(skb)); |
1da177e4 | 1668 | if (sk2 != NULL) { |
295ff7ed ACM |
1669 | struct inet_timewait_sock *tw = inet_twsk(sk); |
1670 | inet_twsk_deschedule(tw, &tcp_death_row); | |
1671 | inet_twsk_put(tw); | |
1da177e4 LT |
1672 | sk = sk2; |
1673 | goto process; | |
1674 | } | |
1675 | /* Fall through to ACK */ | |
1676 | } | |
1677 | case TCP_TW_ACK: | |
1678 | tcp_v6_timewait_ack(sk, skb); | |
1679 | break; | |
1680 | case TCP_TW_RST: | |
1681 | goto no_tcp_socket; | |
1682 | case TCP_TW_SUCCESS:; | |
1683 | } | |
1684 | goto discard_it; | |
1685 | } | |
1686 | ||
ccb7c410 DM |
1687 | static struct inet_peer *tcp_v6_get_peer(struct sock *sk, bool *release_it) |
1688 | { | |
db3949c4 DM |
1689 | struct rt6_info *rt = (struct rt6_info *) __sk_dst_get(sk); |
1690 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1691 | struct inet_peer *peer; | |
1692 | ||
1693 | if (!rt || | |
1694 | !ipv6_addr_equal(&np->daddr, &rt->rt6i_dst.addr)) { | |
1695 | peer = inet_getpeer_v6(&np->daddr, 1); | |
1696 | *release_it = true; | |
1697 | } else { | |
1698 | if (!rt->rt6i_peer) | |
1699 | rt6_bind_peer(rt, 1); | |
1700 | peer = rt->rt6i_peer; | |
457de438 | 1701 | *release_it = false; |
db3949c4 DM |
1702 | } |
1703 | ||
1704 | return peer; | |
ccb7c410 DM |
1705 | } |
1706 | ||
1707 | static void *tcp_v6_tw_get_peer(struct sock *sk) | |
1da177e4 | 1708 | { |
cf533ea5 ED |
1709 | const struct inet6_timewait_sock *tw6 = inet6_twsk(sk); |
1710 | const struct inet_timewait_sock *tw = inet_twsk(sk); | |
ccb7c410 DM |
1711 | |
1712 | if (tw->tw_family == AF_INET) | |
1713 | return tcp_v4_tw_get_peer(sk); | |
1714 | ||
db3949c4 | 1715 | return inet_getpeer_v6(&tw6->tw_v6_daddr, 1); |
1da177e4 LT |
1716 | } |
1717 | ||
ccb7c410 DM |
1718 | static struct timewait_sock_ops tcp6_timewait_sock_ops = { |
1719 | .twsk_obj_size = sizeof(struct tcp6_timewait_sock), | |
1720 | .twsk_unique = tcp_twsk_unique, | |
1721 | .twsk_destructor= tcp_twsk_destructor, | |
1722 | .twsk_getpeer = tcp_v6_tw_get_peer, | |
1723 | }; | |
1724 | ||
3b401a81 | 1725 | static const struct inet_connection_sock_af_ops ipv6_specific = { |
543d9cfe ACM |
1726 | .queue_xmit = inet6_csk_xmit, |
1727 | .send_check = tcp_v6_send_check, | |
1728 | .rebuild_header = inet6_sk_rebuild_header, | |
1729 | .conn_request = tcp_v6_conn_request, | |
1730 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
3f419d2d | 1731 | .get_peer = tcp_v6_get_peer, |
543d9cfe ACM |
1732 | .net_header_len = sizeof(struct ipv6hdr), |
1733 | .setsockopt = ipv6_setsockopt, | |
1734 | .getsockopt = ipv6_getsockopt, | |
1735 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1736 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
ab1e0a13 | 1737 | .bind_conflict = inet6_csk_bind_conflict, |
3fdadf7d | 1738 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1739 | .compat_setsockopt = compat_ipv6_setsockopt, |
1740 | .compat_getsockopt = compat_ipv6_getsockopt, | |
3fdadf7d | 1741 | #endif |
1da177e4 LT |
1742 | }; |
1743 | ||
cfb6eeb4 | 1744 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1745 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = { |
cfb6eeb4 | 1746 | .md5_lookup = tcp_v6_md5_lookup, |
49a72dfb | 1747 | .calc_md5_hash = tcp_v6_md5_hash_skb, |
cfb6eeb4 | 1748 | .md5_parse = tcp_v6_parse_md5_keys, |
cfb6eeb4 | 1749 | }; |
a928630a | 1750 | #endif |
cfb6eeb4 | 1751 | |
1da177e4 LT |
1752 | /* |
1753 | * TCP over IPv4 via INET6 API | |
1754 | */ | |
1755 | ||
3b401a81 | 1756 | static const struct inet_connection_sock_af_ops ipv6_mapped = { |
543d9cfe ACM |
1757 | .queue_xmit = ip_queue_xmit, |
1758 | .send_check = tcp_v4_send_check, | |
1759 | .rebuild_header = inet_sk_rebuild_header, | |
1760 | .conn_request = tcp_v6_conn_request, | |
1761 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
3f419d2d | 1762 | .get_peer = tcp_v4_get_peer, |
543d9cfe ACM |
1763 | .net_header_len = sizeof(struct iphdr), |
1764 | .setsockopt = ipv6_setsockopt, | |
1765 | .getsockopt = ipv6_getsockopt, | |
1766 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1767 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
ab1e0a13 | 1768 | .bind_conflict = inet6_csk_bind_conflict, |
3fdadf7d | 1769 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1770 | .compat_setsockopt = compat_ipv6_setsockopt, |
1771 | .compat_getsockopt = compat_ipv6_getsockopt, | |
3fdadf7d | 1772 | #endif |
1da177e4 LT |
1773 | }; |
1774 | ||
cfb6eeb4 | 1775 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1776 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = { |
cfb6eeb4 | 1777 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1778 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1779 | .md5_parse = tcp_v6_parse_md5_keys, |
cfb6eeb4 | 1780 | }; |
a928630a | 1781 | #endif |
cfb6eeb4 | 1782 | |
1da177e4 LT |
1783 | /* NOTE: A lot of things set to zero explicitly by call to |
1784 | * sk_alloc() so need not be done here. | |
1785 | */ | |
1786 | static int tcp_v6_init_sock(struct sock *sk) | |
1787 | { | |
6687e988 | 1788 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 LT |
1789 | struct tcp_sock *tp = tcp_sk(sk); |
1790 | ||
1791 | skb_queue_head_init(&tp->out_of_order_queue); | |
1792 | tcp_init_xmit_timers(sk); | |
1793 | tcp_prequeue_init(tp); | |
1794 | ||
6687e988 | 1795 | icsk->icsk_rto = TCP_TIMEOUT_INIT; |
1da177e4 LT |
1796 | tp->mdev = TCP_TIMEOUT_INIT; |
1797 | ||
1798 | /* So many TCP implementations out there (incorrectly) count the | |
1799 | * initial SYN frame in their delayed-ACK and congestion control | |
1800 | * algorithms that we must have the following bandaid to talk | |
1801 | * efficiently to them. -DaveM | |
1802 | */ | |
1803 | tp->snd_cwnd = 2; | |
1804 | ||
1805 | /* See draft-stevens-tcpca-spec-01 for discussion of the | |
1806 | * initialization of these values. | |
1807 | */ | |
0b6a05c1 | 1808 | tp->snd_ssthresh = TCP_INFINITE_SSTHRESH; |
1da177e4 | 1809 | tp->snd_cwnd_clamp = ~0; |
bee7ca9e | 1810 | tp->mss_cache = TCP_MSS_DEFAULT; |
1da177e4 LT |
1811 | |
1812 | tp->reordering = sysctl_tcp_reordering; | |
1813 | ||
1814 | sk->sk_state = TCP_CLOSE; | |
1815 | ||
8292a17a | 1816 | icsk->icsk_af_ops = &ipv6_specific; |
6687e988 | 1817 | icsk->icsk_ca_ops = &tcp_init_congestion_ops; |
d83d8461 | 1818 | icsk->icsk_sync_mss = tcp_sync_mss; |
1da177e4 LT |
1819 | sk->sk_write_space = sk_stream_write_space; |
1820 | sock_set_flag(sk, SOCK_USE_WRITE_QUEUE); | |
1821 | ||
cfb6eeb4 YH |
1822 | #ifdef CONFIG_TCP_MD5SIG |
1823 | tp->af_specific = &tcp_sock_ipv6_specific; | |
1824 | #endif | |
1825 | ||
435cf559 WAS |
1826 | /* TCP Cookie Transactions */ |
1827 | if (sysctl_tcp_cookie_size > 0) { | |
1828 | /* Default, cookies without s_data_payload. */ | |
1829 | tp->cookie_values = | |
1830 | kzalloc(sizeof(*tp->cookie_values), | |
1831 | sk->sk_allocation); | |
1832 | if (tp->cookie_values != NULL) | |
1833 | kref_init(&tp->cookie_values->kref); | |
1834 | } | |
1835 | /* Presumed zeroed, in order of appearance: | |
1836 | * cookie_in_always, cookie_out_never, | |
1837 | * s_data_constant, s_data_in, s_data_out | |
1838 | */ | |
1da177e4 LT |
1839 | sk->sk_sndbuf = sysctl_tcp_wmem[1]; |
1840 | sk->sk_rcvbuf = sysctl_tcp_rmem[1]; | |
1841 | ||
eb4dea58 | 1842 | local_bh_disable(); |
d1a4c0b3 | 1843 | sock_update_memcg(sk); |
180d8cd9 | 1844 | sk_sockets_allocated_inc(sk); |
eb4dea58 | 1845 | local_bh_enable(); |
1da177e4 LT |
1846 | |
1847 | return 0; | |
1848 | } | |
1849 | ||
7d06b2e0 | 1850 | static void tcp_v6_destroy_sock(struct sock *sk) |
1da177e4 | 1851 | { |
1da177e4 | 1852 | tcp_v4_destroy_sock(sk); |
7d06b2e0 | 1853 | inet6_destroy_sock(sk); |
1da177e4 LT |
1854 | } |
1855 | ||
952a10be | 1856 | #ifdef CONFIG_PROC_FS |
1da177e4 | 1857 | /* Proc filesystem TCPv6 sock list dumping. */ |
1ab1457c | 1858 | static void get_openreq6(struct seq_file *seq, |
cf533ea5 | 1859 | const struct sock *sk, struct request_sock *req, int i, int uid) |
1da177e4 | 1860 | { |
1da177e4 | 1861 | int ttd = req->expires - jiffies; |
b71d1d42 ED |
1862 | const struct in6_addr *src = &inet6_rsk(req)->loc_addr; |
1863 | const struct in6_addr *dest = &inet6_rsk(req)->rmt_addr; | |
1da177e4 LT |
1864 | |
1865 | if (ttd < 0) | |
1866 | ttd = 0; | |
1867 | ||
1da177e4 LT |
1868 | seq_printf(seq, |
1869 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 1870 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n", |
1da177e4 LT |
1871 | i, |
1872 | src->s6_addr32[0], src->s6_addr32[1], | |
1873 | src->s6_addr32[2], src->s6_addr32[3], | |
fd507037 | 1874 | ntohs(inet_rsk(req)->loc_port), |
1da177e4 LT |
1875 | dest->s6_addr32[0], dest->s6_addr32[1], |
1876 | dest->s6_addr32[2], dest->s6_addr32[3], | |
2e6599cb | 1877 | ntohs(inet_rsk(req)->rmt_port), |
1da177e4 LT |
1878 | TCP_SYN_RECV, |
1879 | 0,0, /* could print option size, but that is af dependent. */ | |
1ab1457c YH |
1880 | 1, /* timers active (only the expire timer) */ |
1881 | jiffies_to_clock_t(ttd), | |
1da177e4 LT |
1882 | req->retrans, |
1883 | uid, | |
1ab1457c | 1884 | 0, /* non standard timer */ |
1da177e4 LT |
1885 | 0, /* open_requests have no inode */ |
1886 | 0, req); | |
1887 | } | |
1888 | ||
1889 | static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i) | |
1890 | { | |
b71d1d42 | 1891 | const struct in6_addr *dest, *src; |
1da177e4 LT |
1892 | __u16 destp, srcp; |
1893 | int timer_active; | |
1894 | unsigned long timer_expires; | |
cf533ea5 ED |
1895 | const struct inet_sock *inet = inet_sk(sp); |
1896 | const struct tcp_sock *tp = tcp_sk(sp); | |
463c84b9 | 1897 | const struct inet_connection_sock *icsk = inet_csk(sp); |
cf533ea5 | 1898 | const struct ipv6_pinfo *np = inet6_sk(sp); |
1da177e4 LT |
1899 | |
1900 | dest = &np->daddr; | |
1901 | src = &np->rcv_saddr; | |
c720c7e8 ED |
1902 | destp = ntohs(inet->inet_dport); |
1903 | srcp = ntohs(inet->inet_sport); | |
463c84b9 ACM |
1904 | |
1905 | if (icsk->icsk_pending == ICSK_TIME_RETRANS) { | |
1da177e4 | 1906 | timer_active = 1; |
463c84b9 ACM |
1907 | timer_expires = icsk->icsk_timeout; |
1908 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 1909 | timer_active = 4; |
463c84b9 | 1910 | timer_expires = icsk->icsk_timeout; |
1da177e4 LT |
1911 | } else if (timer_pending(&sp->sk_timer)) { |
1912 | timer_active = 2; | |
1913 | timer_expires = sp->sk_timer.expires; | |
1914 | } else { | |
1915 | timer_active = 0; | |
1916 | timer_expires = jiffies; | |
1917 | } | |
1918 | ||
1919 | seq_printf(seq, | |
1920 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 1921 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %lu %lu %u %u %d\n", |
1da177e4 LT |
1922 | i, |
1923 | src->s6_addr32[0], src->s6_addr32[1], | |
1924 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
1925 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1926 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
1ab1457c | 1927 | sp->sk_state, |
47da8ee6 SS |
1928 | tp->write_seq-tp->snd_una, |
1929 | (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq), | |
1da177e4 LT |
1930 | timer_active, |
1931 | jiffies_to_clock_t(timer_expires - jiffies), | |
463c84b9 | 1932 | icsk->icsk_retransmits, |
1da177e4 | 1933 | sock_i_uid(sp), |
6687e988 | 1934 | icsk->icsk_probes_out, |
1da177e4 LT |
1935 | sock_i_ino(sp), |
1936 | atomic_read(&sp->sk_refcnt), sp, | |
7be87351 SH |
1937 | jiffies_to_clock_t(icsk->icsk_rto), |
1938 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
463c84b9 | 1939 | (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong, |
0b6a05c1 IJ |
1940 | tp->snd_cwnd, |
1941 | tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh | |
1da177e4 LT |
1942 | ); |
1943 | } | |
1944 | ||
1ab1457c | 1945 | static void get_timewait6_sock(struct seq_file *seq, |
8feaf0c0 | 1946 | struct inet_timewait_sock *tw, int i) |
1da177e4 | 1947 | { |
b71d1d42 | 1948 | const struct in6_addr *dest, *src; |
1da177e4 | 1949 | __u16 destp, srcp; |
cf533ea5 | 1950 | const struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw); |
1da177e4 LT |
1951 | int ttd = tw->tw_ttd - jiffies; |
1952 | ||
1953 | if (ttd < 0) | |
1954 | ttd = 0; | |
1955 | ||
0fa1a53e ACM |
1956 | dest = &tw6->tw_v6_daddr; |
1957 | src = &tw6->tw_v6_rcv_saddr; | |
1da177e4 LT |
1958 | destp = ntohs(tw->tw_dport); |
1959 | srcp = ntohs(tw->tw_sport); | |
1960 | ||
1961 | seq_printf(seq, | |
1962 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 1963 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n", |
1da177e4 LT |
1964 | i, |
1965 | src->s6_addr32[0], src->s6_addr32[1], | |
1966 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
1967 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1968 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
1969 | tw->tw_substate, 0, 0, | |
1970 | 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0, | |
1971 | atomic_read(&tw->tw_refcnt), tw); | |
1972 | } | |
1973 | ||
1da177e4 LT |
1974 | static int tcp6_seq_show(struct seq_file *seq, void *v) |
1975 | { | |
1976 | struct tcp_iter_state *st; | |
1977 | ||
1978 | if (v == SEQ_START_TOKEN) { | |
1979 | seq_puts(seq, | |
1980 | " sl " | |
1981 | "local_address " | |
1982 | "remote_address " | |
1983 | "st tx_queue rx_queue tr tm->when retrnsmt" | |
1984 | " uid timeout inode\n"); | |
1985 | goto out; | |
1986 | } | |
1987 | st = seq->private; | |
1988 | ||
1989 | switch (st->state) { | |
1990 | case TCP_SEQ_STATE_LISTENING: | |
1991 | case TCP_SEQ_STATE_ESTABLISHED: | |
1992 | get_tcp6_sock(seq, v, st->num); | |
1993 | break; | |
1994 | case TCP_SEQ_STATE_OPENREQ: | |
1995 | get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid); | |
1996 | break; | |
1997 | case TCP_SEQ_STATE_TIME_WAIT: | |
1998 | get_timewait6_sock(seq, v, st->num); | |
1999 | break; | |
2000 | } | |
2001 | out: | |
2002 | return 0; | |
2003 | } | |
2004 | ||
73cb88ec AV |
2005 | static const struct file_operations tcp6_afinfo_seq_fops = { |
2006 | .owner = THIS_MODULE, | |
2007 | .open = tcp_seq_open, | |
2008 | .read = seq_read, | |
2009 | .llseek = seq_lseek, | |
2010 | .release = seq_release_net | |
2011 | }; | |
2012 | ||
1da177e4 | 2013 | static struct tcp_seq_afinfo tcp6_seq_afinfo = { |
1da177e4 LT |
2014 | .name = "tcp6", |
2015 | .family = AF_INET6, | |
73cb88ec | 2016 | .seq_fops = &tcp6_afinfo_seq_fops, |
9427c4b3 DL |
2017 | .seq_ops = { |
2018 | .show = tcp6_seq_show, | |
2019 | }, | |
1da177e4 LT |
2020 | }; |
2021 | ||
2c8c1e72 | 2022 | int __net_init tcp6_proc_init(struct net *net) |
1da177e4 | 2023 | { |
6f8b13bc | 2024 | return tcp_proc_register(net, &tcp6_seq_afinfo); |
1da177e4 LT |
2025 | } |
2026 | ||
6f8b13bc | 2027 | void tcp6_proc_exit(struct net *net) |
1da177e4 | 2028 | { |
6f8b13bc | 2029 | tcp_proc_unregister(net, &tcp6_seq_afinfo); |
1da177e4 LT |
2030 | } |
2031 | #endif | |
2032 | ||
2033 | struct proto tcpv6_prot = { | |
2034 | .name = "TCPv6", | |
2035 | .owner = THIS_MODULE, | |
2036 | .close = tcp_close, | |
2037 | .connect = tcp_v6_connect, | |
2038 | .disconnect = tcp_disconnect, | |
463c84b9 | 2039 | .accept = inet_csk_accept, |
1da177e4 LT |
2040 | .ioctl = tcp_ioctl, |
2041 | .init = tcp_v6_init_sock, | |
2042 | .destroy = tcp_v6_destroy_sock, | |
2043 | .shutdown = tcp_shutdown, | |
2044 | .setsockopt = tcp_setsockopt, | |
2045 | .getsockopt = tcp_getsockopt, | |
1da177e4 | 2046 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
2047 | .sendmsg = tcp_sendmsg, |
2048 | .sendpage = tcp_sendpage, | |
1da177e4 LT |
2049 | .backlog_rcv = tcp_v6_do_rcv, |
2050 | .hash = tcp_v6_hash, | |
ab1e0a13 ACM |
2051 | .unhash = inet_unhash, |
2052 | .get_port = inet_csk_get_port, | |
1da177e4 LT |
2053 | .enter_memory_pressure = tcp_enter_memory_pressure, |
2054 | .sockets_allocated = &tcp_sockets_allocated, | |
2055 | .memory_allocated = &tcp_memory_allocated, | |
2056 | .memory_pressure = &tcp_memory_pressure, | |
0a5578cf | 2057 | .orphan_count = &tcp_orphan_count, |
1da177e4 LT |
2058 | .sysctl_wmem = sysctl_tcp_wmem, |
2059 | .sysctl_rmem = sysctl_tcp_rmem, | |
2060 | .max_header = MAX_TCP_HEADER, | |
2061 | .obj_size = sizeof(struct tcp6_sock), | |
3ab5aee7 | 2062 | .slab_flags = SLAB_DESTROY_BY_RCU, |
6d6ee43e | 2063 | .twsk_prot = &tcp6_timewait_sock_ops, |
60236fdd | 2064 | .rsk_prot = &tcp6_request_sock_ops, |
39d8cda7 | 2065 | .h.hashinfo = &tcp_hashinfo, |
7ba42910 | 2066 | .no_autobind = true, |
543d9cfe ACM |
2067 | #ifdef CONFIG_COMPAT |
2068 | .compat_setsockopt = compat_tcp_setsockopt, | |
2069 | .compat_getsockopt = compat_tcp_getsockopt, | |
2070 | #endif | |
d1a4c0b3 GC |
2071 | #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM |
2072 | .proto_cgroup = tcp_proto_cgroup, | |
2073 | #endif | |
1da177e4 LT |
2074 | }; |
2075 | ||
41135cc8 | 2076 | static const struct inet6_protocol tcpv6_protocol = { |
1da177e4 LT |
2077 | .handler = tcp_v6_rcv, |
2078 | .err_handler = tcp_v6_err, | |
a430a43d | 2079 | .gso_send_check = tcp_v6_gso_send_check, |
adcfc7d0 | 2080 | .gso_segment = tcp_tso_segment, |
684f2176 HX |
2081 | .gro_receive = tcp6_gro_receive, |
2082 | .gro_complete = tcp6_gro_complete, | |
1da177e4 LT |
2083 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, |
2084 | }; | |
2085 | ||
1da177e4 LT |
2086 | static struct inet_protosw tcpv6_protosw = { |
2087 | .type = SOCK_STREAM, | |
2088 | .protocol = IPPROTO_TCP, | |
2089 | .prot = &tcpv6_prot, | |
2090 | .ops = &inet6_stream_ops, | |
1da177e4 | 2091 | .no_check = 0, |
d83d8461 ACM |
2092 | .flags = INET_PROTOSW_PERMANENT | |
2093 | INET_PROTOSW_ICSK, | |
1da177e4 LT |
2094 | }; |
2095 | ||
2c8c1e72 | 2096 | static int __net_init tcpv6_net_init(struct net *net) |
93ec926b | 2097 | { |
5677242f DL |
2098 | return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6, |
2099 | SOCK_RAW, IPPROTO_TCP, net); | |
93ec926b DL |
2100 | } |
2101 | ||
2c8c1e72 | 2102 | static void __net_exit tcpv6_net_exit(struct net *net) |
93ec926b | 2103 | { |
5677242f | 2104 | inet_ctl_sock_destroy(net->ipv6.tcp_sk); |
b099ce26 EB |
2105 | } |
2106 | ||
2c8c1e72 | 2107 | static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list) |
b099ce26 EB |
2108 | { |
2109 | inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6); | |
93ec926b DL |
2110 | } |
2111 | ||
2112 | static struct pernet_operations tcpv6_net_ops = { | |
b099ce26 EB |
2113 | .init = tcpv6_net_init, |
2114 | .exit = tcpv6_net_exit, | |
2115 | .exit_batch = tcpv6_net_exit_batch, | |
93ec926b DL |
2116 | }; |
2117 | ||
7f4e4868 | 2118 | int __init tcpv6_init(void) |
1da177e4 | 2119 | { |
7f4e4868 DL |
2120 | int ret; |
2121 | ||
2122 | ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
2123 | if (ret) | |
2124 | goto out; | |
2125 | ||
1da177e4 | 2126 | /* register inet6 protocol */ |
7f4e4868 DL |
2127 | ret = inet6_register_protosw(&tcpv6_protosw); |
2128 | if (ret) | |
2129 | goto out_tcpv6_protocol; | |
2130 | ||
93ec926b | 2131 | ret = register_pernet_subsys(&tcpv6_net_ops); |
7f4e4868 DL |
2132 | if (ret) |
2133 | goto out_tcpv6_protosw; | |
2134 | out: | |
2135 | return ret; | |
ae0f7d5f | 2136 | |
7f4e4868 DL |
2137 | out_tcpv6_protocol: |
2138 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
2139 | out_tcpv6_protosw: | |
2140 | inet6_unregister_protosw(&tcpv6_protosw); | |
2141 | goto out; | |
2142 | } | |
2143 | ||
09f7709f | 2144 | void tcpv6_exit(void) |
7f4e4868 | 2145 | { |
93ec926b | 2146 | unregister_pernet_subsys(&tcpv6_net_ops); |
7f4e4868 DL |
2147 | inet6_unregister_protosw(&tcpv6_protosw); |
2148 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
1da177e4 | 2149 | } |