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> |
4aa956d8 | 42 | #include <linux/uaccess.h> |
1da177e4 LT |
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> |
076bb0c8 | 66 | #include <net/busy_poll.h> |
1da177e4 | 67 | |
1da177e4 LT |
68 | #include <linux/proc_fs.h> |
69 | #include <linux/seq_file.h> | |
70 | ||
cfb6eeb4 YH |
71 | #include <linux/crypto.h> |
72 | #include <linux/scatterlist.h> | |
73 | ||
cfb6eeb4 | 74 | static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb); |
6edafaaf GJ |
75 | static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
76 | struct request_sock *req); | |
1da177e4 LT |
77 | |
78 | static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb); | |
1da177e4 | 79 | |
3b401a81 SH |
80 | static const struct inet_connection_sock_af_ops ipv6_mapped; |
81 | static const struct inet_connection_sock_af_ops ipv6_specific; | |
a928630a | 82 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de SH |
83 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific; |
84 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific; | |
9501f972 YH |
85 | #else |
86 | static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk, | |
b71d1d42 | 87 | const struct in6_addr *addr) |
9501f972 YH |
88 | { |
89 | return NULL; | |
90 | } | |
a928630a | 91 | #endif |
1da177e4 | 92 | |
fae6ef87 NC |
93 | static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) |
94 | { | |
95 | struct dst_entry *dst = skb_dst(skb); | |
96 | const struct rt6_info *rt = (const struct rt6_info *)dst; | |
97 | ||
98 | dst_hold(dst); | |
99 | sk->sk_rx_dst = dst; | |
100 | inet_sk(sk)->rx_dst_ifindex = skb->skb_iif; | |
101 | if (rt->rt6i_node) | |
102 | inet6_sk(sk)->rx_dst_cookie = rt->rt6i_node->fn_sernum; | |
103 | } | |
104 | ||
1da177e4 LT |
105 | static void tcp_v6_hash(struct sock *sk) |
106 | { | |
107 | if (sk->sk_state != TCP_CLOSE) { | |
8292a17a | 108 | if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) { |
1da177e4 LT |
109 | tcp_prot.hash(sk); |
110 | return; | |
111 | } | |
112 | local_bh_disable(); | |
9327f705 | 113 | __inet6_hash(sk, NULL); |
1da177e4 LT |
114 | local_bh_enable(); |
115 | } | |
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; | |
1da177e4 LT |
157 | fl6_sock_release(flowlabel); |
158 | } | |
159 | } | |
160 | ||
161 | /* | |
1ab1457c YH |
162 | * connect() to INADDR_ANY means loopback (BSD'ism). |
163 | */ | |
164 | ||
4c99aa40 | 165 | if (ipv6_addr_any(&usin->sin6_addr)) |
1ab1457c | 166 | usin->sin6_addr.s6_addr[15] = 0x1; |
1da177e4 LT |
167 | |
168 | addr_type = ipv6_addr_type(&usin->sin6_addr); | |
169 | ||
4c99aa40 | 170 | if (addr_type & IPV6_ADDR_MULTICAST) |
1da177e4 LT |
171 | return -ENETUNREACH; |
172 | ||
173 | if (addr_type&IPV6_ADDR_LINKLOCAL) { | |
174 | if (addr_len >= sizeof(struct sockaddr_in6) && | |
175 | usin->sin6_scope_id) { | |
176 | /* If interface is set while binding, indices | |
177 | * must coincide. | |
178 | */ | |
179 | if (sk->sk_bound_dev_if && | |
180 | sk->sk_bound_dev_if != usin->sin6_scope_id) | |
181 | return -EINVAL; | |
182 | ||
183 | sk->sk_bound_dev_if = usin->sin6_scope_id; | |
184 | } | |
185 | ||
186 | /* Connect to link-local address requires an interface */ | |
187 | if (!sk->sk_bound_dev_if) | |
188 | return -EINVAL; | |
189 | } | |
190 | ||
191 | if (tp->rx_opt.ts_recent_stamp && | |
efe4208f | 192 | !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) { |
1da177e4 LT |
193 | tp->rx_opt.ts_recent = 0; |
194 | tp->rx_opt.ts_recent_stamp = 0; | |
195 | tp->write_seq = 0; | |
196 | } | |
197 | ||
efe4208f | 198 | sk->sk_v6_daddr = usin->sin6_addr; |
4c9483b2 | 199 | np->flow_label = fl6.flowlabel; |
1da177e4 LT |
200 | |
201 | /* | |
202 | * TCP over IPv4 | |
203 | */ | |
204 | ||
205 | if (addr_type == IPV6_ADDR_MAPPED) { | |
d83d8461 | 206 | u32 exthdrlen = icsk->icsk_ext_hdr_len; |
1da177e4 LT |
207 | struct sockaddr_in sin; |
208 | ||
209 | SOCK_DEBUG(sk, "connect: ipv4 mapped\n"); | |
210 | ||
211 | if (__ipv6_only_sock(sk)) | |
212 | return -ENETUNREACH; | |
213 | ||
214 | sin.sin_family = AF_INET; | |
215 | sin.sin_port = usin->sin6_port; | |
216 | sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3]; | |
217 | ||
d83d8461 | 218 | icsk->icsk_af_ops = &ipv6_mapped; |
1da177e4 | 219 | sk->sk_backlog_rcv = tcp_v4_do_rcv; |
cfb6eeb4 YH |
220 | #ifdef CONFIG_TCP_MD5SIG |
221 | tp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
222 | #endif | |
1da177e4 LT |
223 | |
224 | err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin)); | |
225 | ||
226 | if (err) { | |
d83d8461 ACM |
227 | icsk->icsk_ext_hdr_len = exthdrlen; |
228 | icsk->icsk_af_ops = &ipv6_specific; | |
1da177e4 | 229 | sk->sk_backlog_rcv = tcp_v6_do_rcv; |
cfb6eeb4 YH |
230 | #ifdef CONFIG_TCP_MD5SIG |
231 | tp->af_specific = &tcp_sock_ipv6_specific; | |
232 | #endif | |
1da177e4 LT |
233 | goto failure; |
234 | } else { | |
c720c7e8 ED |
235 | ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr); |
236 | ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, | |
efe4208f | 237 | &sk->sk_v6_rcv_saddr); |
1da177e4 LT |
238 | } |
239 | ||
240 | return err; | |
241 | } | |
242 | ||
efe4208f ED |
243 | if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) |
244 | saddr = &sk->sk_v6_rcv_saddr; | |
1da177e4 | 245 | |
4c9483b2 | 246 | fl6.flowi6_proto = IPPROTO_TCP; |
efe4208f | 247 | fl6.daddr = sk->sk_v6_daddr; |
4e3fd7a0 | 248 | fl6.saddr = saddr ? *saddr : np->saddr; |
4c9483b2 DM |
249 | fl6.flowi6_oif = sk->sk_bound_dev_if; |
250 | fl6.flowi6_mark = sk->sk_mark; | |
1958b856 DM |
251 | fl6.fl6_dport = usin->sin6_port; |
252 | fl6.fl6_sport = inet->inet_sport; | |
1da177e4 | 253 | |
4c9483b2 | 254 | final_p = fl6_update_dst(&fl6, np->opt, &final); |
1da177e4 | 255 | |
4c9483b2 | 256 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
beb8d13b | 257 | |
0e0d44ab | 258 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p); |
68d0c6d3 DM |
259 | if (IS_ERR(dst)) { |
260 | err = PTR_ERR(dst); | |
1da177e4 | 261 | goto failure; |
14e50e57 | 262 | } |
1da177e4 LT |
263 | |
264 | if (saddr == NULL) { | |
4c9483b2 | 265 | saddr = &fl6.saddr; |
efe4208f | 266 | sk->sk_v6_rcv_saddr = *saddr; |
1da177e4 LT |
267 | } |
268 | ||
269 | /* set the source address */ | |
4e3fd7a0 | 270 | np->saddr = *saddr; |
c720c7e8 | 271 | inet->inet_rcv_saddr = LOOPBACK4_IPV6; |
1da177e4 | 272 | |
f83ef8c0 | 273 | sk->sk_gso_type = SKB_GSO_TCPV6; |
8e1ef0a9 | 274 | __ip6_dst_store(sk, dst, NULL, NULL); |
1da177e4 | 275 | |
493f377d DM |
276 | rt = (struct rt6_info *) dst; |
277 | if (tcp_death_row.sysctl_tw_recycle && | |
278 | !tp->rx_opt.ts_recent_stamp && | |
efe4208f | 279 | ipv6_addr_equal(&rt->rt6i_dst.addr, &sk->sk_v6_daddr)) |
81166dd6 | 280 | tcp_fetch_timewait_stamp(sk, dst); |
493f377d | 281 | |
d83d8461 | 282 | icsk->icsk_ext_hdr_len = 0; |
1da177e4 | 283 | if (np->opt) |
d83d8461 ACM |
284 | icsk->icsk_ext_hdr_len = (np->opt->opt_flen + |
285 | np->opt->opt_nflen); | |
1da177e4 LT |
286 | |
287 | tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); | |
288 | ||
c720c7e8 | 289 | inet->inet_dport = usin->sin6_port; |
1da177e4 LT |
290 | |
291 | tcp_set_state(sk, TCP_SYN_SENT); | |
d8313f5c | 292 | err = inet6_hash_connect(&tcp_death_row, sk); |
1da177e4 LT |
293 | if (err) |
294 | goto late_failure; | |
295 | ||
2b916477 | 296 | if (!tp->write_seq && likely(!tp->repair)) |
1da177e4 | 297 | tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32, |
efe4208f | 298 | sk->sk_v6_daddr.s6_addr32, |
c720c7e8 ED |
299 | inet->inet_sport, |
300 | inet->inet_dport); | |
1da177e4 LT |
301 | |
302 | err = tcp_connect(sk); | |
303 | if (err) | |
304 | goto late_failure; | |
305 | ||
306 | return 0; | |
307 | ||
308 | late_failure: | |
309 | tcp_set_state(sk, TCP_CLOSE); | |
310 | __sk_dst_reset(sk); | |
311 | failure: | |
c720c7e8 | 312 | inet->inet_dport = 0; |
1da177e4 LT |
313 | sk->sk_route_caps = 0; |
314 | return err; | |
315 | } | |
316 | ||
563d34d0 ED |
317 | static void tcp_v6_mtu_reduced(struct sock *sk) |
318 | { | |
319 | struct dst_entry *dst; | |
320 | ||
321 | if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) | |
322 | return; | |
323 | ||
324 | dst = inet6_csk_update_pmtu(sk, tcp_sk(sk)->mtu_info); | |
325 | if (!dst) | |
326 | return; | |
327 | ||
328 | if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) { | |
329 | tcp_sync_mss(sk, dst_mtu(dst)); | |
330 | tcp_simple_retransmit(sk); | |
331 | } | |
332 | } | |
333 | ||
1da177e4 | 334 | static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, |
d5fdd6ba | 335 | u8 type, u8 code, int offset, __be32 info) |
1da177e4 | 336 | { |
4c99aa40 | 337 | const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; |
505cbfc5 | 338 | const struct tcphdr *th = (struct tcphdr *)(skb->data+offset); |
1da177e4 LT |
339 | struct ipv6_pinfo *np; |
340 | struct sock *sk; | |
341 | int err; | |
1ab1457c | 342 | struct tcp_sock *tp; |
0a672f74 YC |
343 | struct request_sock *fastopen; |
344 | __u32 seq, snd_una; | |
ca12a1a4 | 345 | struct net *net = dev_net(skb->dev); |
1da177e4 | 346 | |
ca12a1a4 | 347 | sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr, |
d86e0dac | 348 | th->dest, &hdr->saddr, th->source, skb->dev->ifindex); |
1da177e4 LT |
349 | |
350 | if (sk == NULL) { | |
e41b5368 DL |
351 | ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev), |
352 | ICMP6_MIB_INERRORS); | |
1da177e4 LT |
353 | return; |
354 | } | |
355 | ||
356 | if (sk->sk_state == TCP_TIME_WAIT) { | |
9469c7b4 | 357 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
358 | return; |
359 | } | |
360 | ||
361 | bh_lock_sock(sk); | |
563d34d0 | 362 | if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG) |
de0744af | 363 | NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS); |
1da177e4 LT |
364 | |
365 | if (sk->sk_state == TCP_CLOSE) | |
366 | goto out; | |
367 | ||
e802af9c SH |
368 | if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) { |
369 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
370 | goto out; | |
371 | } | |
372 | ||
1da177e4 | 373 | tp = tcp_sk(sk); |
1ab1457c | 374 | seq = ntohl(th->seq); |
0a672f74 YC |
375 | /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */ |
376 | fastopen = tp->fastopen_rsk; | |
377 | snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una; | |
1da177e4 | 378 | if (sk->sk_state != TCP_LISTEN && |
0a672f74 | 379 | !between(seq, snd_una, tp->snd_nxt)) { |
de0744af | 380 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
381 | goto out; |
382 | } | |
383 | ||
384 | np = inet6_sk(sk); | |
385 | ||
ec18d9a2 DM |
386 | if (type == NDISC_REDIRECT) { |
387 | struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie); | |
388 | ||
1ed5c48f | 389 | if (dst) |
6700c270 | 390 | dst->ops->redirect(dst, sk, skb); |
50a75a89 | 391 | goto out; |
ec18d9a2 DM |
392 | } |
393 | ||
1da177e4 | 394 | if (type == ICMPV6_PKT_TOOBIG) { |
0d4f0608 ED |
395 | /* We are not interested in TCP_LISTEN and open_requests |
396 | * (SYN-ACKs send out by Linux are always <576bytes so | |
397 | * they should go through unfragmented). | |
398 | */ | |
399 | if (sk->sk_state == TCP_LISTEN) | |
400 | goto out; | |
401 | ||
93b36cf3 HFS |
402 | if (!ip6_sk_accept_pmtu(sk)) |
403 | goto out; | |
404 | ||
563d34d0 ED |
405 | tp->mtu_info = ntohl(info); |
406 | if (!sock_owned_by_user(sk)) | |
407 | tcp_v6_mtu_reduced(sk); | |
d013ef2a JA |
408 | else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, |
409 | &tp->tsq_flags)) | |
410 | sock_hold(sk); | |
1da177e4 LT |
411 | goto out; |
412 | } | |
413 | ||
414 | icmpv6_err_convert(type, code, &err); | |
415 | ||
60236fdd | 416 | /* Might be for an request_sock */ |
1da177e4 | 417 | switch (sk->sk_state) { |
60236fdd | 418 | struct request_sock *req, **prev; |
1da177e4 LT |
419 | case TCP_LISTEN: |
420 | if (sock_owned_by_user(sk)) | |
421 | goto out; | |
422 | ||
8129765a ACM |
423 | req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr, |
424 | &hdr->saddr, inet6_iif(skb)); | |
1da177e4 LT |
425 | if (!req) |
426 | goto out; | |
427 | ||
428 | /* ICMPs are not backlogged, hence we cannot get | |
429 | * an established socket here. | |
430 | */ | |
547b792c | 431 | WARN_ON(req->sk != NULL); |
1da177e4 | 432 | |
2e6599cb | 433 | if (seq != tcp_rsk(req)->snt_isn) { |
de0744af | 434 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); |
1da177e4 LT |
435 | goto out; |
436 | } | |
437 | ||
463c84b9 | 438 | inet_csk_reqsk_queue_drop(sk, req, prev); |
5f1e942c | 439 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 LT |
440 | goto out; |
441 | ||
442 | case TCP_SYN_SENT: | |
0a672f74 YC |
443 | case TCP_SYN_RECV: |
444 | /* Only in fast or simultaneous open. If a fast open socket is | |
445 | * is already accepted it is treated as a connected one below. | |
446 | */ | |
447 | if (fastopen && fastopen->sk == NULL) | |
448 | break; | |
449 | ||
1da177e4 | 450 | if (!sock_owned_by_user(sk)) { |
1da177e4 LT |
451 | sk->sk_err = err; |
452 | sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */ | |
453 | ||
454 | tcp_done(sk); | |
455 | } else | |
456 | sk->sk_err_soft = err; | |
457 | goto out; | |
458 | } | |
459 | ||
460 | if (!sock_owned_by_user(sk) && np->recverr) { | |
461 | sk->sk_err = err; | |
462 | sk->sk_error_report(sk); | |
463 | } else | |
464 | sk->sk_err_soft = err; | |
465 | ||
466 | out: | |
467 | bh_unlock_sock(sk); | |
468 | sock_put(sk); | |
469 | } | |
470 | ||
471 | ||
9f10d3f6 NC |
472 | static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst, |
473 | struct flowi6 *fl6, | |
3840a06e | 474 | struct request_sock *req, |
3a19ce0e DL |
475 | u16 queue_mapping, |
476 | struct tcp_fastopen_cookie *foc) | |
1da177e4 | 477 | { |
634fb979 | 478 | struct inet_request_sock *ireq = inet_rsk(req); |
1da177e4 | 479 | struct ipv6_pinfo *np = inet6_sk(sk); |
4c99aa40 | 480 | struct sk_buff *skb; |
9494218f | 481 | int err = -ENOMEM; |
1da177e4 | 482 | |
9f10d3f6 NC |
483 | /* First, grab a route. */ |
484 | if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL) | |
fd80eb94 | 485 | goto done; |
9494218f | 486 | |
3a19ce0e | 487 | skb = tcp_make_synack(sk, dst, req, foc); |
9494218f | 488 | |
1da177e4 | 489 | if (skb) { |
634fb979 ED |
490 | __tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr, |
491 | &ireq->ir_v6_rmt_addr); | |
1da177e4 | 492 | |
634fb979 | 493 | fl6->daddr = ireq->ir_v6_rmt_addr; |
df3687ff FF |
494 | if (np->repflow && (ireq->pktopts != NULL)) |
495 | fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts)); | |
496 | ||
fff32699 | 497 | skb_set_queue_mapping(skb, queue_mapping); |
43264e0b | 498 | err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass); |
b9df3cb8 | 499 | err = net_xmit_eval(err); |
aa27fc50 OP |
500 | if (!tcp_rsk(req)->snt_synack && !err) |
501 | tcp_rsk(req)->snt_synack = tcp_time_stamp; | |
1da177e4 LT |
502 | } |
503 | ||
504 | done: | |
1da177e4 LT |
505 | return err; |
506 | } | |
507 | ||
1a2c6181 | 508 | static int tcp_v6_rtx_synack(struct sock *sk, struct request_sock *req) |
72659ecc | 509 | { |
9f10d3f6 | 510 | struct flowi6 fl6; |
e6c022a4 | 511 | int res; |
9f10d3f6 | 512 | |
3a19ce0e | 513 | res = tcp_v6_send_synack(sk, NULL, &fl6, req, 0, NULL); |
f19c29e3 | 514 | if (!res) { |
e6c022a4 | 515 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_RETRANSSEGS); |
f19c29e3 YC |
516 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPSYNRETRANS); |
517 | } | |
e6c022a4 | 518 | return res; |
72659ecc OP |
519 | } |
520 | ||
60236fdd | 521 | static void tcp_v6_reqsk_destructor(struct request_sock *req) |
1da177e4 | 522 | { |
634fb979 | 523 | kfree_skb(inet_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 | { | |
efe4208f | 536 | return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr); |
cfb6eeb4 YH |
537 | } |
538 | ||
539 | static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk, | |
540 | struct request_sock *req) | |
541 | { | |
634fb979 | 542 | return tcp_v6_md5_do_lookup(sk, &inet_rsk(req)->ir_v6_rmt_addr); |
cfb6eeb4 YH |
543 | } |
544 | ||
4aa956d8 WY |
545 | static int tcp_v6_parse_md5_keys(struct sock *sk, char __user *optval, |
546 | int optlen) | |
cfb6eeb4 YH |
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; | |
efe4208f | 642 | daddr = &sk->sk_v6_daddr; |
49a72dfb | 643 | } else if (req) { |
634fb979 ED |
644 | saddr = &inet_rsk(req)->ir_v6_loc_addr; |
645 | daddr = &inet_rsk(req)->ir_v6_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 | ||
16bea70a OP |
723 | static void tcp_v6_init_req(struct request_sock *req, struct sock *sk, |
724 | struct sk_buff *skb) | |
725 | { | |
726 | struct inet_request_sock *ireq = inet_rsk(req); | |
727 | struct ipv6_pinfo *np = inet6_sk(sk); | |
728 | ||
729 | ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr; | |
730 | ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr; | |
731 | ||
732 | ireq->ir_iif = sk->sk_bound_dev_if; | |
733 | ||
734 | /* So that link locals have meaning */ | |
735 | if (!sk->sk_bound_dev_if && | |
736 | ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL) | |
737 | ireq->ir_iif = inet6_iif(skb); | |
738 | ||
739 | if (!TCP_SKB_CB(skb)->when && | |
740 | (ipv6_opt_accepted(sk, skb) || np->rxopt.bits.rxinfo || | |
741 | np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim || | |
742 | np->rxopt.bits.rxohlim || np->repflow)) { | |
743 | atomic_inc(&skb->users); | |
744 | ireq->pktopts = skb; | |
745 | } | |
746 | } | |
747 | ||
c6aefafb | 748 | struct request_sock_ops tcp6_request_sock_ops __read_mostly = { |
1da177e4 | 749 | .family = AF_INET6, |
2e6599cb | 750 | .obj_size = sizeof(struct tcp6_request_sock), |
72659ecc | 751 | .rtx_syn_ack = tcp_v6_rtx_synack, |
60236fdd ACM |
752 | .send_ack = tcp_v6_reqsk_send_ack, |
753 | .destructor = tcp_v6_reqsk_destructor, | |
72659ecc | 754 | .send_reset = tcp_v6_send_reset, |
4aa956d8 | 755 | .syn_ack_timeout = tcp_syn_ack_timeout, |
1da177e4 LT |
756 | }; |
757 | ||
b2e4b3de | 758 | static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = { |
16bea70a | 759 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 | 760 | .md5_lookup = tcp_v6_reqsk_md5_lookup, |
e3afe7b7 | 761 | .calc_md5_hash = tcp_v6_md5_hash_skb, |
b6332e6c | 762 | #endif |
16bea70a | 763 | .init_req = tcp_v6_init_req, |
fb7b37a7 OP |
764 | #ifdef CONFIG_SYN_COOKIES |
765 | .cookie_init_seq = cookie_v6_init_sequence, | |
766 | #endif | |
16bea70a | 767 | }; |
cfb6eeb4 | 768 | |
626e264d | 769 | static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win, |
9c76a114 | 770 | u32 tsval, u32 tsecr, int oif, |
1d13a96c FF |
771 | struct tcp_md5sig_key *key, int rst, u8 tclass, |
772 | u32 label) | |
1da177e4 | 773 | { |
cf533ea5 ED |
774 | const struct tcphdr *th = tcp_hdr(skb); |
775 | struct tcphdr *t1; | |
1da177e4 | 776 | struct sk_buff *buff; |
4c9483b2 | 777 | struct flowi6 fl6; |
adf30907 | 778 | struct net *net = dev_net(skb_dst(skb)->dev); |
e5047992 | 779 | struct sock *ctl_sk = net->ipv6.tcp_sk; |
77c676da | 780 | unsigned int tot_len = sizeof(struct tcphdr); |
adf30907 | 781 | struct dst_entry *dst; |
81ada62d | 782 | __be32 *topt; |
1da177e4 | 783 | |
ee684b6f | 784 | if (tsecr) |
626e264d | 785 | tot_len += TCPOLEN_TSTAMP_ALIGNED; |
cfb6eeb4 | 786 | #ifdef CONFIG_TCP_MD5SIG |
cfb6eeb4 YH |
787 | if (key) |
788 | tot_len += TCPOLEN_MD5SIG_ALIGNED; | |
789 | #endif | |
790 | ||
cfb6eeb4 | 791 | buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len, |
1da177e4 | 792 | GFP_ATOMIC); |
1ab1457c YH |
793 | if (buff == NULL) |
794 | return; | |
1da177e4 | 795 | |
cfb6eeb4 | 796 | skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len); |
1da177e4 | 797 | |
cfb6eeb4 | 798 | t1 = (struct tcphdr *) skb_push(buff, tot_len); |
6651ffc8 | 799 | skb_reset_transport_header(buff); |
1da177e4 LT |
800 | |
801 | /* Swap the send and the receive. */ | |
802 | memset(t1, 0, sizeof(*t1)); | |
803 | t1->dest = th->source; | |
804 | t1->source = th->dest; | |
cfb6eeb4 | 805 | t1->doff = tot_len / 4; |
626e264d IJ |
806 | t1->seq = htonl(seq); |
807 | t1->ack_seq = htonl(ack); | |
808 | t1->ack = !rst || !th->ack; | |
809 | t1->rst = rst; | |
810 | t1->window = htons(win); | |
1da177e4 | 811 | |
81ada62d IJ |
812 | topt = (__be32 *)(t1 + 1); |
813 | ||
ee684b6f | 814 | if (tsecr) { |
626e264d IJ |
815 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
816 | (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP); | |
ee684b6f AV |
817 | *topt++ = htonl(tsval); |
818 | *topt++ = htonl(tsecr); | |
626e264d IJ |
819 | } |
820 | ||
cfb6eeb4 YH |
821 | #ifdef CONFIG_TCP_MD5SIG |
822 | if (key) { | |
81ada62d IJ |
823 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | |
824 | (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG); | |
825 | tcp_v6_md5_hash_hdr((__u8 *)topt, key, | |
78e645cb IJ |
826 | &ipv6_hdr(skb)->saddr, |
827 | &ipv6_hdr(skb)->daddr, t1); | |
cfb6eeb4 YH |
828 | } |
829 | #endif | |
830 | ||
4c9483b2 | 831 | memset(&fl6, 0, sizeof(fl6)); |
4e3fd7a0 AD |
832 | fl6.daddr = ipv6_hdr(skb)->saddr; |
833 | fl6.saddr = ipv6_hdr(skb)->daddr; | |
1d13a96c | 834 | fl6.flowlabel = label; |
1da177e4 | 835 | |
e5700aff DM |
836 | buff->ip_summed = CHECKSUM_PARTIAL; |
837 | buff->csum = 0; | |
838 | ||
4c9483b2 | 839 | __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr); |
1da177e4 | 840 | |
4c9483b2 | 841 | fl6.flowi6_proto = IPPROTO_TCP; |
a36dbdb2 | 842 | if (rt6_need_strict(&fl6.daddr) && !oif) |
4c675258 | 843 | fl6.flowi6_oif = inet6_iif(skb); |
9c76a114 WY |
844 | else |
845 | fl6.flowi6_oif = oif; | |
e110861f | 846 | fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark); |
1958b856 DM |
847 | fl6.fl6_dport = t1->dest; |
848 | fl6.fl6_sport = t1->source; | |
4c9483b2 | 849 | security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); |
1da177e4 | 850 | |
c20121ae DL |
851 | /* Pass a socket to ip6_dst_lookup either it is for RST |
852 | * Underlying function will use this to retrieve the network | |
853 | * namespace | |
854 | */ | |
0e0d44ab | 855 | dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL); |
68d0c6d3 DM |
856 | if (!IS_ERR(dst)) { |
857 | skb_dst_set(buff, dst); | |
b903d324 | 858 | ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass); |
68d0c6d3 DM |
859 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); |
860 | if (rst) | |
861 | TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS); | |
862 | return; | |
1da177e4 LT |
863 | } |
864 | ||
865 | kfree_skb(buff); | |
866 | } | |
867 | ||
626e264d | 868 | static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 869 | { |
cf533ea5 | 870 | const struct tcphdr *th = tcp_hdr(skb); |
626e264d | 871 | u32 seq = 0, ack_seq = 0; |
fa3e5b4e | 872 | struct tcp_md5sig_key *key = NULL; |
658ddaaf SL |
873 | #ifdef CONFIG_TCP_MD5SIG |
874 | const __u8 *hash_location = NULL; | |
875 | struct ipv6hdr *ipv6h = ipv6_hdr(skb); | |
876 | unsigned char newhash[16]; | |
877 | int genhash; | |
878 | struct sock *sk1 = NULL; | |
879 | #endif | |
9c76a114 | 880 | int oif; |
1da177e4 | 881 | |
626e264d | 882 | if (th->rst) |
1da177e4 LT |
883 | return; |
884 | ||
626e264d IJ |
885 | if (!ipv6_unicast_destination(skb)) |
886 | return; | |
1da177e4 | 887 | |
cfb6eeb4 | 888 | #ifdef CONFIG_TCP_MD5SIG |
658ddaaf SL |
889 | hash_location = tcp_parse_md5sig_option(th); |
890 | if (!sk && hash_location) { | |
891 | /* | |
892 | * active side is lost. Try to find listening socket through | |
893 | * source port, and then find md5 key through listening socket. | |
894 | * we are not loose security here: | |
895 | * Incoming packet is checked with md5 hash with finding key, | |
896 | * no RST generated if md5 hash doesn't match. | |
897 | */ | |
898 | sk1 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev), | |
5ba24953 TH |
899 | &tcp_hashinfo, &ipv6h->saddr, |
900 | th->source, &ipv6h->daddr, | |
658ddaaf SL |
901 | ntohs(th->source), inet6_iif(skb)); |
902 | if (!sk1) | |
903 | return; | |
904 | ||
905 | rcu_read_lock(); | |
906 | key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr); | |
907 | if (!key) | |
908 | goto release_sk1; | |
909 | ||
910 | genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, NULL, skb); | |
911 | if (genhash || memcmp(hash_location, newhash, 16) != 0) | |
912 | goto release_sk1; | |
913 | } else { | |
914 | key = sk ? tcp_v6_md5_do_lookup(sk, &ipv6h->saddr) : NULL; | |
915 | } | |
cfb6eeb4 YH |
916 | #endif |
917 | ||
626e264d IJ |
918 | if (th->ack) |
919 | seq = ntohl(th->ack_seq); | |
920 | else | |
921 | ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len - | |
922 | (th->doff << 2); | |
1da177e4 | 923 | |
9c76a114 WY |
924 | oif = sk ? sk->sk_bound_dev_if : 0; |
925 | tcp_v6_send_response(skb, seq, ack_seq, 0, 0, 0, oif, key, 1, 0, 0); | |
658ddaaf SL |
926 | |
927 | #ifdef CONFIG_TCP_MD5SIG | |
928 | release_sk1: | |
929 | if (sk1) { | |
930 | rcu_read_unlock(); | |
931 | sock_put(sk1); | |
932 | } | |
933 | #endif | |
626e264d | 934 | } |
1da177e4 | 935 | |
ee684b6f | 936 | static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, |
9c76a114 | 937 | u32 win, u32 tsval, u32 tsecr, int oif, |
1d13a96c FF |
938 | struct tcp_md5sig_key *key, u8 tclass, |
939 | u32 label) | |
626e264d | 940 | { |
9c76a114 | 941 | tcp_v6_send_response(skb, seq, ack, win, tsval, tsecr, oif, key, 0, tclass, |
1d13a96c | 942 | label); |
1da177e4 LT |
943 | } |
944 | ||
945 | static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
946 | { | |
8feaf0c0 | 947 | struct inet_timewait_sock *tw = inet_twsk(sk); |
cfb6eeb4 | 948 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); |
1da177e4 | 949 | |
9501f972 | 950 | tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, |
8feaf0c0 | 951 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, |
ee684b6f | 952 | tcp_time_stamp + tcptw->tw_ts_offset, |
9c76a114 | 953 | tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw), |
1d13a96c | 954 | tw->tw_tclass, (tw->tw_flowlabel << 12)); |
1da177e4 | 955 | |
8feaf0c0 | 956 | inet_twsk_put(tw); |
1da177e4 LT |
957 | } |
958 | ||
6edafaaf GJ |
959 | static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, |
960 | struct request_sock *req) | |
1da177e4 | 961 | { |
3a19ce0e DL |
962 | /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV |
963 | * sk->sk_state == TCP_SYN_RECV -> for Fast Open. | |
964 | */ | |
965 | tcp_v6_send_ack(skb, (sk->sk_state == TCP_LISTEN) ? | |
966 | tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt, | |
967 | tcp_rsk(req)->rcv_nxt, | |
9c76a114 | 968 | req->rcv_wnd, tcp_time_stamp, req->ts_recent, sk->sk_bound_dev_if, |
1d13a96c FF |
969 | tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr), |
970 | 0, 0); | |
1da177e4 LT |
971 | } |
972 | ||
973 | ||
4c99aa40 | 974 | static struct sock *tcp_v6_hnd_req(struct sock *sk, struct sk_buff *skb) |
1da177e4 | 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 | 984 | if (req) |
8336886f | 985 | return tcp_check_req(sk, skb, req, prev, false); |
1da177e4 | 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 | { | |
e6b4d113 WAS |
1012 | struct tcp_options_received tmp_opt; |
1013 | struct request_sock *req; | |
634fb979 | 1014 | struct inet_request_sock *ireq; |
1da177e4 | 1015 | struct tcp_sock *tp = tcp_sk(sk); |
e6b4d113 | 1016 | __u32 isn = TCP_SKB_CB(skb)->when; |
493f377d | 1017 | struct dst_entry *dst = NULL; |
3a19ce0e DL |
1018 | struct tcp_fastopen_cookie foc = { .len = -1 }; |
1019 | bool want_cookie = false, fastopen; | |
3840a06e | 1020 | struct flowi6 fl6; |
16bea70a | 1021 | const struct tcp_request_sock_ops *af_ops; |
3a19ce0e | 1022 | int err; |
1da177e4 LT |
1023 | |
1024 | if (skb->protocol == htons(ETH_P_IP)) | |
1025 | return tcp_v4_conn_request(sk, skb); | |
1026 | ||
1027 | if (!ipv6_unicast_destination(skb)) | |
1ab1457c | 1028 | goto drop; |
1da177e4 | 1029 | |
5ad37d5d HFS |
1030 | if ((sysctl_tcp_syncookies == 2 || |
1031 | inet_csk_reqsk_queue_is_full(sk)) && !isn) { | |
946cedcc ED |
1032 | want_cookie = tcp_syn_flood_action(sk, skb, "TCPv6"); |
1033 | if (!want_cookie) | |
1034 | goto drop; | |
1da177e4 LT |
1035 | } |
1036 | ||
5f1e942c VS |
1037 | if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1) { |
1038 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); | |
1da177e4 | 1039 | goto drop; |
5f1e942c | 1040 | } |
1da177e4 | 1041 | |
476eab82 | 1042 | req = inet_reqsk_alloc(&tcp6_request_sock_ops); |
1da177e4 LT |
1043 | if (req == NULL) |
1044 | goto drop; | |
1045 | ||
16bea70a | 1046 | af_ops = tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops; |
cfb6eeb4 | 1047 | |
1da177e4 LT |
1048 | tcp_clear_options(&tmp_opt); |
1049 | tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); | |
1050 | tmp_opt.user_mss = tp->rx_opt.user_mss; | |
3a19ce0e | 1051 | tcp_parse_options(skb, &tmp_opt, 0, want_cookie ? NULL : &foc); |
1da177e4 | 1052 | |
4dfc2817 | 1053 | if (want_cookie && !tmp_opt.saw_tstamp) |
c6aefafb | 1054 | tcp_clear_options(&tmp_opt); |
c6aefafb | 1055 | |
1da177e4 | 1056 | tmp_opt.tstamp_ok = tmp_opt.saw_tstamp; |
e0f802fb | 1057 | tcp_openreq_init(req, &tmp_opt, skb, sk); |
1da177e4 | 1058 | |
634fb979 | 1059 | ireq = inet_rsk(req); |
16bea70a OP |
1060 | af_ops->init_req(req, sk, skb); |
1061 | ||
172d69e6 | 1062 | if (!want_cookie || tmp_opt.tstamp_ok) |
5d134f1c | 1063 | TCP_ECN_create_request(req, skb, sock_net(sk)); |
c6aefafb | 1064 | |
2bbdf389 | 1065 | if (!isn) { |
493f377d | 1066 | if (want_cookie) { |
fb7b37a7 | 1067 | isn = cookie_init_sequence(af_ops, sk, skb, &req->mss); |
2bbdf389 | 1068 | req->cookie_ts = tmp_opt.tstamp_ok; |
493f377d DM |
1069 | goto have_isn; |
1070 | } | |
1071 | ||
1072 | /* VJ's idea. We save last timestamp seen | |
1073 | * from the destination in peer table, when entering | |
1074 | * state TIME-WAIT, and check against it before | |
1075 | * accepting new connection request. | |
1076 | * | |
1077 | * If "isn" is not zero, this request hit alive | |
1078 | * timewait bucket, so that all the necessary checks | |
1079 | * are made in the function processing timewait state. | |
1080 | */ | |
1081 | if (tmp_opt.saw_tstamp && | |
1082 | tcp_death_row.sysctl_tw_recycle && | |
81166dd6 DM |
1083 | (dst = inet6_csk_route_req(sk, &fl6, req)) != NULL) { |
1084 | if (!tcp_peer_is_proven(req, dst, true)) { | |
493f377d DM |
1085 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_PAWSPASSIVEREJECTED); |
1086 | goto drop_and_release; | |
1087 | } | |
1088 | } | |
1089 | /* Kill the following clause, if you dislike this way. */ | |
1090 | else if (!sysctl_tcp_syncookies && | |
1091 | (sysctl_max_syn_backlog - inet_csk_reqsk_queue_len(sk) < | |
1092 | (sysctl_max_syn_backlog >> 2)) && | |
81166dd6 | 1093 | !tcp_peer_is_proven(req, dst, false)) { |
493f377d DM |
1094 | /* Without syncookies last quarter of |
1095 | * backlog is filled with destinations, | |
1096 | * proven to be alive. | |
1097 | * It means that we continue to communicate | |
1098 | * to destinations, already remembered | |
1099 | * to the moment of synflood. | |
1100 | */ | |
1101 | LIMIT_NETDEBUG(KERN_DEBUG "TCP: drop open request from %pI6/%u\n", | |
634fb979 | 1102 | &ireq->ir_v6_rmt_addr, ntohs(tcp_hdr(skb)->source)); |
493f377d | 1103 | goto drop_and_release; |
2bbdf389 | 1104 | } |
493f377d DM |
1105 | |
1106 | isn = tcp_v6_init_sequence(skb); | |
c6aefafb | 1107 | } |
493f377d | 1108 | have_isn: |
1da177e4 | 1109 | |
437c5b53 NC |
1110 | if (security_inet_conn_request(sk, skb, req)) |
1111 | goto drop_and_release; | |
4237c75c | 1112 | |
3a19ce0e | 1113 | if (!dst && (dst = inet6_csk_route_req(sk, &fl6, req)) == NULL) |
e6b4d113 | 1114 | goto drop_and_free; |
1da177e4 | 1115 | |
3a19ce0e | 1116 | tcp_rsk(req)->snt_isn = isn; |
3a19ce0e DL |
1117 | tcp_openreq_init_rwin(req, sk, dst); |
1118 | fastopen = !want_cookie && | |
1119 | tcp_try_fastopen(sk, skb, req, &foc, dst); | |
1120 | err = tcp_v6_send_synack(sk, dst, &fl6, req, | |
1121 | skb_get_queue_mapping(skb), &foc); | |
1122 | if (!fastopen) { | |
1123 | if (err || want_cookie) | |
1124 | goto drop_and_free; | |
1125 | ||
1126 | tcp_rsk(req)->listener = NULL; | |
1127 | inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT); | |
1128 | } | |
e6b4d113 | 1129 | return 0; |
1da177e4 | 1130 | |
493f377d DM |
1131 | drop_and_release: |
1132 | dst_release(dst); | |
e6b4d113 WAS |
1133 | drop_and_free: |
1134 | reqsk_free(req); | |
1da177e4 | 1135 | drop: |
5f1e942c | 1136 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); |
1da177e4 LT |
1137 | return 0; /* don't send reset */ |
1138 | } | |
1139 | ||
4c99aa40 WC |
1140 | static struct sock *tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb, |
1141 | struct request_sock *req, | |
1142 | struct dst_entry *dst) | |
1da177e4 | 1143 | { |
634fb979 | 1144 | struct inet_request_sock *ireq; |
1da177e4 LT |
1145 | struct ipv6_pinfo *newnp, *np = inet6_sk(sk); |
1146 | struct tcp6_sock *newtcp6sk; | |
1147 | struct inet_sock *newinet; | |
1148 | struct tcp_sock *newtp; | |
1149 | struct sock *newsk; | |
cfb6eeb4 YH |
1150 | #ifdef CONFIG_TCP_MD5SIG |
1151 | struct tcp_md5sig_key *key; | |
1152 | #endif | |
3840a06e | 1153 | struct flowi6 fl6; |
1da177e4 LT |
1154 | |
1155 | if (skb->protocol == htons(ETH_P_IP)) { | |
1156 | /* | |
1157 | * v6 mapped | |
1158 | */ | |
1159 | ||
1160 | newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst); | |
1161 | ||
1ab1457c | 1162 | if (newsk == NULL) |
1da177e4 LT |
1163 | return NULL; |
1164 | ||
1165 | newtcp6sk = (struct tcp6_sock *)newsk; | |
1166 | inet_sk(newsk)->pinet6 = &newtcp6sk->inet6; | |
1167 | ||
1168 | newinet = inet_sk(newsk); | |
1169 | newnp = inet6_sk(newsk); | |
1170 | newtp = tcp_sk(newsk); | |
1171 | ||
1172 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1173 | ||
efe4208f | 1174 | ipv6_addr_set_v4mapped(newinet->inet_daddr, &newsk->sk_v6_daddr); |
1da177e4 | 1175 | |
c720c7e8 | 1176 | ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr); |
1da177e4 | 1177 | |
efe4208f | 1178 | newsk->sk_v6_rcv_saddr = newnp->saddr; |
1da177e4 | 1179 | |
8292a17a | 1180 | inet_csk(newsk)->icsk_af_ops = &ipv6_mapped; |
1da177e4 | 1181 | newsk->sk_backlog_rcv = tcp_v4_do_rcv; |
cfb6eeb4 YH |
1182 | #ifdef CONFIG_TCP_MD5SIG |
1183 | newtp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
1184 | #endif | |
1185 | ||
676a1184 YZ |
1186 | newnp->ipv6_ac_list = NULL; |
1187 | newnp->ipv6_fl_list = NULL; | |
1da177e4 LT |
1188 | newnp->pktoptions = NULL; |
1189 | newnp->opt = NULL; | |
505cbfc5 | 1190 | newnp->mcast_oif = inet6_iif(skb); |
0660e03f | 1191 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; |
1397ed35 | 1192 | newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb)); |
df3687ff FF |
1193 | if (np->repflow) |
1194 | newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb)); | |
1da177e4 | 1195 | |
e6848976 ACM |
1196 | /* |
1197 | * No need to charge this sock to the relevant IPv6 refcnt debug socks count | |
1198 | * here, tcp_create_openreq_child now does this for us, see the comment in | |
1199 | * that function for the gory details. -acme | |
1da177e4 | 1200 | */ |
1da177e4 LT |
1201 | |
1202 | /* It is tricky place. Until this moment IPv4 tcp | |
8292a17a | 1203 | worked with IPv6 icsk.icsk_af_ops. |
1da177e4 LT |
1204 | Sync it now. |
1205 | */ | |
d83d8461 | 1206 | tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie); |
1da177e4 LT |
1207 | |
1208 | return newsk; | |
1209 | } | |
1210 | ||
634fb979 | 1211 | ireq = inet_rsk(req); |
1da177e4 LT |
1212 | |
1213 | if (sk_acceptq_is_full(sk)) | |
1214 | goto out_overflow; | |
1215 | ||
493f377d | 1216 | if (!dst) { |
3840a06e | 1217 | dst = inet6_csk_route_req(sk, &fl6, req); |
493f377d | 1218 | if (!dst) |
1da177e4 | 1219 | goto out; |
1ab1457c | 1220 | } |
1da177e4 LT |
1221 | |
1222 | newsk = tcp_create_openreq_child(sk, req, skb); | |
1223 | if (newsk == NULL) | |
093d2823 | 1224 | goto out_nonewsk; |
1da177e4 | 1225 | |
e6848976 ACM |
1226 | /* |
1227 | * No need to charge this sock to the relevant IPv6 refcnt debug socks | |
1228 | * count here, tcp_create_openreq_child now does this for us, see the | |
1229 | * comment in that function for the gory details. -acme | |
1230 | */ | |
1da177e4 | 1231 | |
59eed279 | 1232 | newsk->sk_gso_type = SKB_GSO_TCPV6; |
8e1ef0a9 | 1233 | __ip6_dst_store(newsk, dst, NULL, NULL); |
fae6ef87 | 1234 | inet6_sk_rx_dst_set(newsk, skb); |
1da177e4 LT |
1235 | |
1236 | newtcp6sk = (struct tcp6_sock *)newsk; | |
1237 | inet_sk(newsk)->pinet6 = &newtcp6sk->inet6; | |
1238 | ||
1239 | newtp = tcp_sk(newsk); | |
1240 | newinet = inet_sk(newsk); | |
1241 | newnp = inet6_sk(newsk); | |
1242 | ||
1243 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1244 | ||
634fb979 ED |
1245 | newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr; |
1246 | newnp->saddr = ireq->ir_v6_loc_addr; | |
1247 | newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr; | |
1248 | newsk->sk_bound_dev_if = ireq->ir_iif; | |
1da177e4 | 1249 | |
1ab1457c | 1250 | /* Now IPv6 options... |
1da177e4 LT |
1251 | |
1252 | First: no IPv4 options. | |
1253 | */ | |
f6d8bd05 | 1254 | newinet->inet_opt = NULL; |
676a1184 | 1255 | newnp->ipv6_ac_list = NULL; |
d35690be | 1256 | newnp->ipv6_fl_list = NULL; |
1da177e4 LT |
1257 | |
1258 | /* Clone RX bits */ | |
1259 | newnp->rxopt.all = np->rxopt.all; | |
1260 | ||
1261 | /* Clone pktoptions received with SYN */ | |
1262 | newnp->pktoptions = NULL; | |
634fb979 ED |
1263 | if (ireq->pktopts != NULL) { |
1264 | newnp->pktoptions = skb_clone(ireq->pktopts, | |
99a1dec7 | 1265 | sk_gfp_atomic(sk, GFP_ATOMIC)); |
634fb979 ED |
1266 | consume_skb(ireq->pktopts); |
1267 | ireq->pktopts = NULL; | |
1da177e4 LT |
1268 | if (newnp->pktoptions) |
1269 | skb_set_owner_r(newnp->pktoptions, newsk); | |
1270 | } | |
1271 | newnp->opt = NULL; | |
505cbfc5 | 1272 | newnp->mcast_oif = inet6_iif(skb); |
0660e03f | 1273 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; |
1397ed35 | 1274 | newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb)); |
df3687ff FF |
1275 | if (np->repflow) |
1276 | newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb)); | |
1da177e4 LT |
1277 | |
1278 | /* Clone native IPv6 options from listening socket (if any) | |
1279 | ||
1280 | Yes, keeping reference count would be much more clever, | |
1281 | but we make one more one thing there: reattach optmem | |
1282 | to newsk. | |
1283 | */ | |
43264e0b RL |
1284 | if (np->opt) |
1285 | newnp->opt = ipv6_dup_options(newsk, np->opt); | |
1da177e4 | 1286 | |
d83d8461 | 1287 | inet_csk(newsk)->icsk_ext_hdr_len = 0; |
1da177e4 | 1288 | if (newnp->opt) |
d83d8461 ACM |
1289 | inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + |
1290 | newnp->opt->opt_flen); | |
1da177e4 LT |
1291 | |
1292 | tcp_sync_mss(newsk, dst_mtu(dst)); | |
0dbaee3b | 1293 | newtp->advmss = dst_metric_advmss(dst); |
d135c522 NC |
1294 | if (tcp_sk(sk)->rx_opt.user_mss && |
1295 | tcp_sk(sk)->rx_opt.user_mss < newtp->advmss) | |
1296 | newtp->advmss = tcp_sk(sk)->rx_opt.user_mss; | |
1297 | ||
1da177e4 LT |
1298 | tcp_initialize_rcv_mss(newsk); |
1299 | ||
c720c7e8 ED |
1300 | newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6; |
1301 | newinet->inet_rcv_saddr = LOOPBACK4_IPV6; | |
1da177e4 | 1302 | |
cfb6eeb4 YH |
1303 | #ifdef CONFIG_TCP_MD5SIG |
1304 | /* Copy over the MD5 key from the original socket */ | |
4aa956d8 WY |
1305 | key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr); |
1306 | if (key != NULL) { | |
cfb6eeb4 YH |
1307 | /* We're using one, so create a matching key |
1308 | * on the newsk structure. If we fail to get | |
1309 | * memory, then we end up not copying the key | |
1310 | * across. Shucks. | |
1311 | */ | |
efe4208f | 1312 | tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr, |
99a1dec7 MG |
1313 | AF_INET6, key->key, key->keylen, |
1314 | sk_gfp_atomic(sk, GFP_ATOMIC)); | |
cfb6eeb4 YH |
1315 | } |
1316 | #endif | |
1317 | ||
093d2823 | 1318 | if (__inet_inherit_port(sk, newsk) < 0) { |
e337e24d CP |
1319 | inet_csk_prepare_forced_close(newsk); |
1320 | tcp_done(newsk); | |
093d2823 BS |
1321 | goto out; |
1322 | } | |
9327f705 | 1323 | __inet6_hash(newsk, NULL); |
1da177e4 LT |
1324 | |
1325 | return newsk; | |
1326 | ||
1327 | out_overflow: | |
de0744af | 1328 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); |
093d2823 | 1329 | out_nonewsk: |
1da177e4 | 1330 | dst_release(dst); |
093d2823 BS |
1331 | out: |
1332 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); | |
1da177e4 LT |
1333 | return NULL; |
1334 | } | |
1335 | ||
1da177e4 LT |
1336 | /* The socket must have it's spinlock held when we get |
1337 | * here. | |
1338 | * | |
1339 | * We have a potential double-lock case here, so even when | |
1340 | * doing backlog processing we use the BH locking scheme. | |
1341 | * This is because we cannot sleep with the original spinlock | |
1342 | * held. | |
1343 | */ | |
1344 | static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1345 | { | |
1346 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1347 | struct tcp_sock *tp; | |
1348 | struct sk_buff *opt_skb = NULL; | |
1349 | ||
1350 | /* Imagine: socket is IPv6. IPv4 packet arrives, | |
1351 | goes to IPv4 receive handler and backlogged. | |
1352 | From backlog it always goes here. Kerboom... | |
1353 | Fortunately, tcp_rcv_established and rcv_established | |
1354 | handle them correctly, but it is not case with | |
1355 | tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK | |
1356 | */ | |
1357 | ||
1358 | if (skb->protocol == htons(ETH_P_IP)) | |
1359 | return tcp_v4_do_rcv(sk, skb); | |
1360 | ||
cfb6eeb4 | 1361 | #ifdef CONFIG_TCP_MD5SIG |
4aa956d8 | 1362 | if (tcp_v6_inbound_md5_hash(sk, skb)) |
cfb6eeb4 YH |
1363 | goto discard; |
1364 | #endif | |
1365 | ||
fda9ef5d | 1366 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1367 | goto discard; |
1368 | ||
1369 | /* | |
1370 | * socket locking is here for SMP purposes as backlog rcv | |
1371 | * is currently called with bh processing disabled. | |
1372 | */ | |
1373 | ||
1374 | /* Do Stevens' IPV6_PKTOPTIONS. | |
1375 | ||
1376 | Yes, guys, it is the only place in our code, where we | |
1377 | may make it not affecting IPv4. | |
1378 | The rest of code is protocol independent, | |
1379 | and I do not like idea to uglify IPv4. | |
1380 | ||
1381 | Actually, all the idea behind IPV6_PKTOPTIONS | |
1382 | looks not very well thought. For now we latch | |
1383 | options, received in the last packet, enqueued | |
1384 | by tcp. Feel free to propose better solution. | |
1ab1457c | 1385 | --ANK (980728) |
1da177e4 LT |
1386 | */ |
1387 | if (np->rxopt.all) | |
99a1dec7 | 1388 | opt_skb = skb_clone(skb, sk_gfp_atomic(sk, GFP_ATOMIC)); |
1da177e4 LT |
1389 | |
1390 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ | |
5d299f3d ED |
1391 | struct dst_entry *dst = sk->sk_rx_dst; |
1392 | ||
bdeab991 | 1393 | sock_rps_save_rxhash(sk, skb); |
5d299f3d ED |
1394 | if (dst) { |
1395 | if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif || | |
1396 | dst->ops->check(dst, np->rx_dst_cookie) == NULL) { | |
1397 | dst_release(dst); | |
1398 | sk->sk_rx_dst = NULL; | |
1399 | } | |
1400 | } | |
1401 | ||
c995ae22 | 1402 | tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len); |
1da177e4 LT |
1403 | if (opt_skb) |
1404 | goto ipv6_pktoptions; | |
1405 | return 0; | |
1406 | } | |
1407 | ||
ab6a5bb6 | 1408 | if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb)) |
1da177e4 LT |
1409 | goto csum_err; |
1410 | ||
1ab1457c | 1411 | if (sk->sk_state == TCP_LISTEN) { |
1da177e4 LT |
1412 | struct sock *nsk = tcp_v6_hnd_req(sk, skb); |
1413 | if (!nsk) | |
1414 | goto discard; | |
1415 | ||
1416 | /* | |
1417 | * Queue it on the new socket if the new socket is active, | |
1418 | * otherwise we just shortcircuit this and continue with | |
1419 | * the new socket.. | |
1420 | */ | |
4c99aa40 | 1421 | if (nsk != sk) { |
bdeab991 | 1422 | sock_rps_save_rxhash(nsk, skb); |
1da177e4 LT |
1423 | if (tcp_child_process(sk, nsk, skb)) |
1424 | goto reset; | |
1425 | if (opt_skb) | |
1426 | __kfree_skb(opt_skb); | |
1427 | return 0; | |
1428 | } | |
47482f13 | 1429 | } else |
bdeab991 | 1430 | sock_rps_save_rxhash(sk, skb); |
1da177e4 | 1431 | |
aa8223c7 | 1432 | if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) |
1da177e4 | 1433 | goto reset; |
1da177e4 LT |
1434 | if (opt_skb) |
1435 | goto ipv6_pktoptions; | |
1436 | return 0; | |
1437 | ||
1438 | reset: | |
cfb6eeb4 | 1439 | tcp_v6_send_reset(sk, skb); |
1da177e4 LT |
1440 | discard: |
1441 | if (opt_skb) | |
1442 | __kfree_skb(opt_skb); | |
1443 | kfree_skb(skb); | |
1444 | return 0; | |
1445 | csum_err: | |
6a5dc9e5 | 1446 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS); |
63231bdd | 1447 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS); |
1da177e4 LT |
1448 | goto discard; |
1449 | ||
1450 | ||
1451 | ipv6_pktoptions: | |
1452 | /* Do you ask, what is it? | |
1453 | ||
1454 | 1. skb was enqueued by tcp. | |
1455 | 2. skb is added to tail of read queue, rather than out of order. | |
1456 | 3. socket is not in passive state. | |
1457 | 4. Finally, it really contains options, which user wants to receive. | |
1458 | */ | |
1459 | tp = tcp_sk(sk); | |
1460 | if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt && | |
1461 | !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) { | |
333fad53 | 1462 | if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo) |
505cbfc5 | 1463 | np->mcast_oif = inet6_iif(opt_skb); |
333fad53 | 1464 | if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) |
0660e03f | 1465 | np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit; |
82e9f105 | 1466 | if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass) |
1397ed35 | 1467 | np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb)); |
df3687ff FF |
1468 | if (np->repflow) |
1469 | np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb)); | |
1da177e4 LT |
1470 | if (ipv6_opt_accepted(sk, opt_skb)) { |
1471 | skb_set_owner_r(opt_skb, sk); | |
1472 | opt_skb = xchg(&np->pktoptions, opt_skb); | |
1473 | } else { | |
1474 | __kfree_skb(opt_skb); | |
1475 | opt_skb = xchg(&np->pktoptions, NULL); | |
1476 | } | |
1477 | } | |
1478 | ||
800d55f1 | 1479 | kfree_skb(opt_skb); |
1da177e4 LT |
1480 | return 0; |
1481 | } | |
1482 | ||
e5bbef20 | 1483 | static int tcp_v6_rcv(struct sk_buff *skb) |
1da177e4 | 1484 | { |
cf533ea5 | 1485 | const struct tcphdr *th; |
b71d1d42 | 1486 | const struct ipv6hdr *hdr; |
1da177e4 LT |
1487 | struct sock *sk; |
1488 | int ret; | |
a86b1e30 | 1489 | struct net *net = dev_net(skb->dev); |
1da177e4 LT |
1490 | |
1491 | if (skb->pkt_type != PACKET_HOST) | |
1492 | goto discard_it; | |
1493 | ||
1494 | /* | |
1495 | * Count it even if it's bad. | |
1496 | */ | |
63231bdd | 1497 | TCP_INC_STATS_BH(net, TCP_MIB_INSEGS); |
1da177e4 LT |
1498 | |
1499 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1500 | goto discard_it; | |
1501 | ||
aa8223c7 | 1502 | th = tcp_hdr(skb); |
1da177e4 LT |
1503 | |
1504 | if (th->doff < sizeof(struct tcphdr)/4) | |
1505 | goto bad_packet; | |
1506 | if (!pskb_may_pull(skb, th->doff*4)) | |
1507 | goto discard_it; | |
1508 | ||
e4f45b7f | 1509 | if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo)) |
6a5dc9e5 | 1510 | goto csum_error; |
1da177e4 | 1511 | |
aa8223c7 | 1512 | th = tcp_hdr(skb); |
e802af9c | 1513 | hdr = ipv6_hdr(skb); |
1da177e4 LT |
1514 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); |
1515 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1516 | skb->len - th->doff*4); | |
1517 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
1518 | TCP_SKB_CB(skb)->when = 0; | |
b82d1bb4 | 1519 | TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr); |
1da177e4 LT |
1520 | TCP_SKB_CB(skb)->sacked = 0; |
1521 | ||
9a1f27c4 | 1522 | sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest); |
1da177e4 LT |
1523 | if (!sk) |
1524 | goto no_tcp_socket; | |
1525 | ||
1526 | process: | |
1527 | if (sk->sk_state == TCP_TIME_WAIT) | |
1528 | goto do_time_wait; | |
1529 | ||
e802af9c SH |
1530 | if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) { |
1531 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
1532 | goto discard_and_relse; | |
1533 | } | |
1534 | ||
1da177e4 LT |
1535 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) |
1536 | goto discard_and_relse; | |
1537 | ||
fda9ef5d | 1538 | if (sk_filter(sk, skb)) |
1da177e4 LT |
1539 | goto discard_and_relse; |
1540 | ||
8b80cda5 | 1541 | sk_mark_napi_id(sk, skb); |
1da177e4 LT |
1542 | skb->dev = NULL; |
1543 | ||
293b9c42 | 1544 | bh_lock_sock_nested(sk); |
1da177e4 LT |
1545 | ret = 0; |
1546 | if (!sock_owned_by_user(sk)) { | |
1a2449a8 | 1547 | #ifdef CONFIG_NET_DMA |
1ab1457c | 1548 | struct tcp_sock *tp = tcp_sk(sk); |
b4caea8a | 1549 | if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list) |
a2bd1140 | 1550 | tp->ucopy.dma_chan = net_dma_find_channel(); |
1ab1457c YH |
1551 | if (tp->ucopy.dma_chan) |
1552 | ret = tcp_v6_do_rcv(sk, skb); | |
1553 | else | |
1a2449a8 CL |
1554 | #endif |
1555 | { | |
1556 | if (!tcp_prequeue(sk, skb)) | |
1557 | ret = tcp_v6_do_rcv(sk, skb); | |
1558 | } | |
da882c1f ED |
1559 | } else if (unlikely(sk_add_backlog(sk, skb, |
1560 | sk->sk_rcvbuf + sk->sk_sndbuf))) { | |
6b03a53a | 1561 | bh_unlock_sock(sk); |
6cce09f8 | 1562 | NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP); |
6b03a53a ZY |
1563 | goto discard_and_relse; |
1564 | } | |
1da177e4 LT |
1565 | bh_unlock_sock(sk); |
1566 | ||
1567 | sock_put(sk); | |
1568 | return ret ? -1 : 0; | |
1569 | ||
1570 | no_tcp_socket: | |
1571 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1572 | goto discard_it; | |
1573 | ||
1574 | if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) { | |
6a5dc9e5 ED |
1575 | csum_error: |
1576 | TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS); | |
1da177e4 | 1577 | bad_packet: |
63231bdd | 1578 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); |
1da177e4 | 1579 | } else { |
cfb6eeb4 | 1580 | tcp_v6_send_reset(NULL, skb); |
1da177e4 LT |
1581 | } |
1582 | ||
1583 | discard_it: | |
1da177e4 LT |
1584 | kfree_skb(skb); |
1585 | return 0; | |
1586 | ||
1587 | discard_and_relse: | |
1588 | sock_put(sk); | |
1589 | goto discard_it; | |
1590 | ||
1591 | do_time_wait: | |
1592 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
9469c7b4 | 1593 | inet_twsk_put(inet_twsk(sk)); |
1da177e4 LT |
1594 | goto discard_it; |
1595 | } | |
1596 | ||
6a5dc9e5 | 1597 | if (skb->len < (th->doff<<2)) { |
9469c7b4 | 1598 | inet_twsk_put(inet_twsk(sk)); |
6a5dc9e5 ED |
1599 | goto bad_packet; |
1600 | } | |
1601 | if (tcp_checksum_complete(skb)) { | |
1602 | inet_twsk_put(inet_twsk(sk)); | |
1603 | goto csum_error; | |
1da177e4 LT |
1604 | } |
1605 | ||
9469c7b4 | 1606 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { |
1da177e4 LT |
1607 | case TCP_TW_SYN: |
1608 | { | |
1609 | struct sock *sk2; | |
1610 | ||
c346dca1 | 1611 | sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo, |
5ba24953 | 1612 | &ipv6_hdr(skb)->saddr, th->source, |
0660e03f | 1613 | &ipv6_hdr(skb)->daddr, |
505cbfc5 | 1614 | ntohs(th->dest), inet6_iif(skb)); |
1da177e4 | 1615 | if (sk2 != NULL) { |
295ff7ed ACM |
1616 | struct inet_timewait_sock *tw = inet_twsk(sk); |
1617 | inet_twsk_deschedule(tw, &tcp_death_row); | |
1618 | inet_twsk_put(tw); | |
1da177e4 LT |
1619 | sk = sk2; |
1620 | goto process; | |
1621 | } | |
1622 | /* Fall through to ACK */ | |
1623 | } | |
1624 | case TCP_TW_ACK: | |
1625 | tcp_v6_timewait_ack(sk, skb); | |
1626 | break; | |
1627 | case TCP_TW_RST: | |
1628 | goto no_tcp_socket; | |
4aa956d8 WY |
1629 | case TCP_TW_SUCCESS: |
1630 | ; | |
1da177e4 LT |
1631 | } |
1632 | goto discard_it; | |
1633 | } | |
1634 | ||
c7109986 ED |
1635 | static void tcp_v6_early_demux(struct sk_buff *skb) |
1636 | { | |
1637 | const struct ipv6hdr *hdr; | |
1638 | const struct tcphdr *th; | |
1639 | struct sock *sk; | |
1640 | ||
1641 | if (skb->pkt_type != PACKET_HOST) | |
1642 | return; | |
1643 | ||
1644 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr))) | |
1645 | return; | |
1646 | ||
1647 | hdr = ipv6_hdr(skb); | |
1648 | th = tcp_hdr(skb); | |
1649 | ||
1650 | if (th->doff < sizeof(struct tcphdr) / 4) | |
1651 | return; | |
1652 | ||
1653 | sk = __inet6_lookup_established(dev_net(skb->dev), &tcp_hashinfo, | |
1654 | &hdr->saddr, th->source, | |
1655 | &hdr->daddr, ntohs(th->dest), | |
1656 | inet6_iif(skb)); | |
1657 | if (sk) { | |
1658 | skb->sk = sk; | |
1659 | skb->destructor = sock_edemux; | |
1660 | if (sk->sk_state != TCP_TIME_WAIT) { | |
1661 | struct dst_entry *dst = sk->sk_rx_dst; | |
f3f12135 | 1662 | |
c7109986 | 1663 | if (dst) |
5d299f3d | 1664 | dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie); |
c7109986 | 1665 | if (dst && |
f3f12135 | 1666 | inet_sk(sk)->rx_dst_ifindex == skb->skb_iif) |
c7109986 ED |
1667 | skb_dst_set_noref(skb, dst); |
1668 | } | |
1669 | } | |
1670 | } | |
1671 | ||
ccb7c410 DM |
1672 | static struct timewait_sock_ops tcp6_timewait_sock_ops = { |
1673 | .twsk_obj_size = sizeof(struct tcp6_timewait_sock), | |
1674 | .twsk_unique = tcp_twsk_unique, | |
4aa956d8 | 1675 | .twsk_destructor = tcp_twsk_destructor, |
ccb7c410 DM |
1676 | }; |
1677 | ||
3b401a81 | 1678 | static const struct inet_connection_sock_af_ops ipv6_specific = { |
543d9cfe ACM |
1679 | .queue_xmit = inet6_csk_xmit, |
1680 | .send_check = tcp_v6_send_check, | |
1681 | .rebuild_header = inet6_sk_rebuild_header, | |
5d299f3d | 1682 | .sk_rx_dst_set = inet6_sk_rx_dst_set, |
543d9cfe ACM |
1683 | .conn_request = tcp_v6_conn_request, |
1684 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
543d9cfe | 1685 | .net_header_len = sizeof(struct ipv6hdr), |
67469601 | 1686 | .net_frag_header_len = sizeof(struct frag_hdr), |
543d9cfe ACM |
1687 | .setsockopt = ipv6_setsockopt, |
1688 | .getsockopt = ipv6_getsockopt, | |
1689 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1690 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
ab1e0a13 | 1691 | .bind_conflict = inet6_csk_bind_conflict, |
3fdadf7d | 1692 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1693 | .compat_setsockopt = compat_ipv6_setsockopt, |
1694 | .compat_getsockopt = compat_ipv6_getsockopt, | |
3fdadf7d | 1695 | #endif |
1da177e4 LT |
1696 | }; |
1697 | ||
cfb6eeb4 | 1698 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1699 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = { |
cfb6eeb4 | 1700 | .md5_lookup = tcp_v6_md5_lookup, |
49a72dfb | 1701 | .calc_md5_hash = tcp_v6_md5_hash_skb, |
cfb6eeb4 | 1702 | .md5_parse = tcp_v6_parse_md5_keys, |
cfb6eeb4 | 1703 | }; |
a928630a | 1704 | #endif |
cfb6eeb4 | 1705 | |
1da177e4 LT |
1706 | /* |
1707 | * TCP over IPv4 via INET6 API | |
1708 | */ | |
3b401a81 | 1709 | static const struct inet_connection_sock_af_ops ipv6_mapped = { |
543d9cfe ACM |
1710 | .queue_xmit = ip_queue_xmit, |
1711 | .send_check = tcp_v4_send_check, | |
1712 | .rebuild_header = inet_sk_rebuild_header, | |
63d02d15 | 1713 | .sk_rx_dst_set = inet_sk_rx_dst_set, |
543d9cfe ACM |
1714 | .conn_request = tcp_v6_conn_request, |
1715 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
543d9cfe ACM |
1716 | .net_header_len = sizeof(struct iphdr), |
1717 | .setsockopt = ipv6_setsockopt, | |
1718 | .getsockopt = ipv6_getsockopt, | |
1719 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1720 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
ab1e0a13 | 1721 | .bind_conflict = inet6_csk_bind_conflict, |
3fdadf7d | 1722 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
1723 | .compat_setsockopt = compat_ipv6_setsockopt, |
1724 | .compat_getsockopt = compat_ipv6_getsockopt, | |
3fdadf7d | 1725 | #endif |
1da177e4 LT |
1726 | }; |
1727 | ||
cfb6eeb4 | 1728 | #ifdef CONFIG_TCP_MD5SIG |
b2e4b3de | 1729 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = { |
cfb6eeb4 | 1730 | .md5_lookup = tcp_v4_md5_lookup, |
49a72dfb | 1731 | .calc_md5_hash = tcp_v4_md5_hash_skb, |
cfb6eeb4 | 1732 | .md5_parse = tcp_v6_parse_md5_keys, |
cfb6eeb4 | 1733 | }; |
a928630a | 1734 | #endif |
cfb6eeb4 | 1735 | |
1da177e4 LT |
1736 | /* NOTE: A lot of things set to zero explicitly by call to |
1737 | * sk_alloc() so need not be done here. | |
1738 | */ | |
1739 | static int tcp_v6_init_sock(struct sock *sk) | |
1740 | { | |
6687e988 | 1741 | struct inet_connection_sock *icsk = inet_csk(sk); |
1da177e4 | 1742 | |
900f65d3 | 1743 | tcp_init_sock(sk); |
1da177e4 | 1744 | |
8292a17a | 1745 | icsk->icsk_af_ops = &ipv6_specific; |
1da177e4 | 1746 | |
cfb6eeb4 | 1747 | #ifdef CONFIG_TCP_MD5SIG |
ac807fa8 | 1748 | tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific; |
cfb6eeb4 YH |
1749 | #endif |
1750 | ||
1da177e4 LT |
1751 | return 0; |
1752 | } | |
1753 | ||
7d06b2e0 | 1754 | static void tcp_v6_destroy_sock(struct sock *sk) |
1da177e4 | 1755 | { |
1da177e4 | 1756 | tcp_v4_destroy_sock(sk); |
7d06b2e0 | 1757 | inet6_destroy_sock(sk); |
1da177e4 LT |
1758 | } |
1759 | ||
952a10be | 1760 | #ifdef CONFIG_PROC_FS |
1da177e4 | 1761 | /* Proc filesystem TCPv6 sock list dumping. */ |
1ab1457c | 1762 | static void get_openreq6(struct seq_file *seq, |
a7cb5a49 | 1763 | const struct sock *sk, struct request_sock *req, int i, kuid_t uid) |
1da177e4 | 1764 | { |
1da177e4 | 1765 | int ttd = req->expires - jiffies; |
634fb979 ED |
1766 | const struct in6_addr *src = &inet_rsk(req)->ir_v6_loc_addr; |
1767 | const struct in6_addr *dest = &inet_rsk(req)->ir_v6_rmt_addr; | |
1da177e4 LT |
1768 | |
1769 | if (ttd < 0) | |
1770 | ttd = 0; | |
1771 | ||
1da177e4 LT |
1772 | seq_printf(seq, |
1773 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
d14c5ab6 | 1774 | "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n", |
1da177e4 LT |
1775 | i, |
1776 | src->s6_addr32[0], src->s6_addr32[1], | |
1777 | src->s6_addr32[2], src->s6_addr32[3], | |
b44084c2 | 1778 | inet_rsk(req)->ir_num, |
1da177e4 LT |
1779 | dest->s6_addr32[0], dest->s6_addr32[1], |
1780 | dest->s6_addr32[2], dest->s6_addr32[3], | |
634fb979 | 1781 | ntohs(inet_rsk(req)->ir_rmt_port), |
1da177e4 | 1782 | TCP_SYN_RECV, |
4c99aa40 | 1783 | 0, 0, /* could print option size, but that is af dependent. */ |
1ab1457c YH |
1784 | 1, /* timers active (only the expire timer) */ |
1785 | jiffies_to_clock_t(ttd), | |
e6c022a4 | 1786 | req->num_timeout, |
a7cb5a49 | 1787 | from_kuid_munged(seq_user_ns(seq), uid), |
1ab1457c | 1788 | 0, /* non standard timer */ |
1da177e4 LT |
1789 | 0, /* open_requests have no inode */ |
1790 | 0, req); | |
1791 | } | |
1792 | ||
1793 | static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i) | |
1794 | { | |
b71d1d42 | 1795 | const struct in6_addr *dest, *src; |
1da177e4 LT |
1796 | __u16 destp, srcp; |
1797 | int timer_active; | |
1798 | unsigned long timer_expires; | |
cf533ea5 ED |
1799 | const struct inet_sock *inet = inet_sk(sp); |
1800 | const struct tcp_sock *tp = tcp_sk(sp); | |
463c84b9 | 1801 | const struct inet_connection_sock *icsk = inet_csk(sp); |
0a672f74 | 1802 | struct fastopen_queue *fastopenq = icsk->icsk_accept_queue.fastopenq; |
1da177e4 | 1803 | |
efe4208f ED |
1804 | dest = &sp->sk_v6_daddr; |
1805 | src = &sp->sk_v6_rcv_saddr; | |
c720c7e8 ED |
1806 | destp = ntohs(inet->inet_dport); |
1807 | srcp = ntohs(inet->inet_sport); | |
463c84b9 ACM |
1808 | |
1809 | if (icsk->icsk_pending == ICSK_TIME_RETRANS) { | |
1da177e4 | 1810 | timer_active = 1; |
463c84b9 ACM |
1811 | timer_expires = icsk->icsk_timeout; |
1812 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1da177e4 | 1813 | timer_active = 4; |
463c84b9 | 1814 | timer_expires = icsk->icsk_timeout; |
1da177e4 LT |
1815 | } else if (timer_pending(&sp->sk_timer)) { |
1816 | timer_active = 2; | |
1817 | timer_expires = sp->sk_timer.expires; | |
1818 | } else { | |
1819 | timer_active = 0; | |
1820 | timer_expires = jiffies; | |
1821 | } | |
1822 | ||
1823 | seq_printf(seq, | |
1824 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
d14c5ab6 | 1825 | "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n", |
1da177e4 LT |
1826 | i, |
1827 | src->s6_addr32[0], src->s6_addr32[1], | |
1828 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
1829 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1830 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
1ab1457c | 1831 | sp->sk_state, |
47da8ee6 SS |
1832 | tp->write_seq-tp->snd_una, |
1833 | (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq), | |
1da177e4 | 1834 | timer_active, |
a399a805 | 1835 | jiffies_delta_to_clock_t(timer_expires - jiffies), |
463c84b9 | 1836 | icsk->icsk_retransmits, |
a7cb5a49 | 1837 | from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), |
6687e988 | 1838 | icsk->icsk_probes_out, |
1da177e4 LT |
1839 | sock_i_ino(sp), |
1840 | atomic_read(&sp->sk_refcnt), sp, | |
7be87351 SH |
1841 | jiffies_to_clock_t(icsk->icsk_rto), |
1842 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
4c99aa40 | 1843 | (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong, |
0b6a05c1 | 1844 | tp->snd_cwnd, |
0a672f74 YC |
1845 | sp->sk_state == TCP_LISTEN ? |
1846 | (fastopenq ? fastopenq->max_qlen : 0) : | |
1847 | (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh) | |
1da177e4 LT |
1848 | ); |
1849 | } | |
1850 | ||
1ab1457c | 1851 | static void get_timewait6_sock(struct seq_file *seq, |
8feaf0c0 | 1852 | struct inet_timewait_sock *tw, int i) |
1da177e4 | 1853 | { |
b71d1d42 | 1854 | const struct in6_addr *dest, *src; |
1da177e4 | 1855 | __u16 destp, srcp; |
96f817fe | 1856 | s32 delta = tw->tw_ttd - inet_tw_time_stamp(); |
1da177e4 | 1857 | |
efe4208f ED |
1858 | dest = &tw->tw_v6_daddr; |
1859 | src = &tw->tw_v6_rcv_saddr; | |
1da177e4 LT |
1860 | destp = ntohs(tw->tw_dport); |
1861 | srcp = ntohs(tw->tw_sport); | |
1862 | ||
1863 | seq_printf(seq, | |
1864 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
71338aa7 | 1865 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n", |
1da177e4 LT |
1866 | i, |
1867 | src->s6_addr32[0], src->s6_addr32[1], | |
1868 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
1869 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1870 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
1871 | tw->tw_substate, 0, 0, | |
a399a805 | 1872 | 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0, |
1da177e4 LT |
1873 | atomic_read(&tw->tw_refcnt), tw); |
1874 | } | |
1875 | ||
1da177e4 LT |
1876 | static int tcp6_seq_show(struct seq_file *seq, void *v) |
1877 | { | |
1878 | struct tcp_iter_state *st; | |
05dbc7b5 | 1879 | struct sock *sk = v; |
1da177e4 LT |
1880 | |
1881 | if (v == SEQ_START_TOKEN) { | |
1882 | seq_puts(seq, | |
1883 | " sl " | |
1884 | "local_address " | |
1885 | "remote_address " | |
1886 | "st tx_queue rx_queue tr tm->when retrnsmt" | |
1887 | " uid timeout inode\n"); | |
1888 | goto out; | |
1889 | } | |
1890 | st = seq->private; | |
1891 | ||
1892 | switch (st->state) { | |
1893 | case TCP_SEQ_STATE_LISTENING: | |
1894 | case TCP_SEQ_STATE_ESTABLISHED: | |
05dbc7b5 ED |
1895 | if (sk->sk_state == TCP_TIME_WAIT) |
1896 | get_timewait6_sock(seq, v, st->num); | |
1897 | else | |
1898 | get_tcp6_sock(seq, v, st->num); | |
1da177e4 LT |
1899 | break; |
1900 | case TCP_SEQ_STATE_OPENREQ: | |
1901 | get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid); | |
1902 | break; | |
1da177e4 LT |
1903 | } |
1904 | out: | |
1905 | return 0; | |
1906 | } | |
1907 | ||
73cb88ec AV |
1908 | static const struct file_operations tcp6_afinfo_seq_fops = { |
1909 | .owner = THIS_MODULE, | |
1910 | .open = tcp_seq_open, | |
1911 | .read = seq_read, | |
1912 | .llseek = seq_lseek, | |
1913 | .release = seq_release_net | |
1914 | }; | |
1915 | ||
1da177e4 | 1916 | static struct tcp_seq_afinfo tcp6_seq_afinfo = { |
1da177e4 LT |
1917 | .name = "tcp6", |
1918 | .family = AF_INET6, | |
73cb88ec | 1919 | .seq_fops = &tcp6_afinfo_seq_fops, |
9427c4b3 DL |
1920 | .seq_ops = { |
1921 | .show = tcp6_seq_show, | |
1922 | }, | |
1da177e4 LT |
1923 | }; |
1924 | ||
2c8c1e72 | 1925 | int __net_init tcp6_proc_init(struct net *net) |
1da177e4 | 1926 | { |
6f8b13bc | 1927 | return tcp_proc_register(net, &tcp6_seq_afinfo); |
1da177e4 LT |
1928 | } |
1929 | ||
6f8b13bc | 1930 | void tcp6_proc_exit(struct net *net) |
1da177e4 | 1931 | { |
6f8b13bc | 1932 | tcp_proc_unregister(net, &tcp6_seq_afinfo); |
1da177e4 LT |
1933 | } |
1934 | #endif | |
1935 | ||
f77d6021 ED |
1936 | static void tcp_v6_clear_sk(struct sock *sk, int size) |
1937 | { | |
1938 | struct inet_sock *inet = inet_sk(sk); | |
1939 | ||
1940 | /* we do not want to clear pinet6 field, because of RCU lookups */ | |
1941 | sk_prot_clear_nulls(sk, offsetof(struct inet_sock, pinet6)); | |
1942 | ||
1943 | size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6); | |
1944 | memset(&inet->pinet6 + 1, 0, size); | |
1945 | } | |
1946 | ||
1da177e4 LT |
1947 | struct proto tcpv6_prot = { |
1948 | .name = "TCPv6", | |
1949 | .owner = THIS_MODULE, | |
1950 | .close = tcp_close, | |
1951 | .connect = tcp_v6_connect, | |
1952 | .disconnect = tcp_disconnect, | |
463c84b9 | 1953 | .accept = inet_csk_accept, |
1da177e4 LT |
1954 | .ioctl = tcp_ioctl, |
1955 | .init = tcp_v6_init_sock, | |
1956 | .destroy = tcp_v6_destroy_sock, | |
1957 | .shutdown = tcp_shutdown, | |
1958 | .setsockopt = tcp_setsockopt, | |
1959 | .getsockopt = tcp_getsockopt, | |
1da177e4 | 1960 | .recvmsg = tcp_recvmsg, |
7ba42910 CG |
1961 | .sendmsg = tcp_sendmsg, |
1962 | .sendpage = tcp_sendpage, | |
1da177e4 | 1963 | .backlog_rcv = tcp_v6_do_rcv, |
46d3ceab | 1964 | .release_cb = tcp_release_cb, |
563d34d0 | 1965 | .mtu_reduced = tcp_v6_mtu_reduced, |
1da177e4 | 1966 | .hash = tcp_v6_hash, |
ab1e0a13 ACM |
1967 | .unhash = inet_unhash, |
1968 | .get_port = inet_csk_get_port, | |
1da177e4 | 1969 | .enter_memory_pressure = tcp_enter_memory_pressure, |
c9bee3b7 | 1970 | .stream_memory_free = tcp_stream_memory_free, |
1da177e4 LT |
1971 | .sockets_allocated = &tcp_sockets_allocated, |
1972 | .memory_allocated = &tcp_memory_allocated, | |
1973 | .memory_pressure = &tcp_memory_pressure, | |
0a5578cf | 1974 | .orphan_count = &tcp_orphan_count, |
a4fe34bf | 1975 | .sysctl_mem = sysctl_tcp_mem, |
1da177e4 LT |
1976 | .sysctl_wmem = sysctl_tcp_wmem, |
1977 | .sysctl_rmem = sysctl_tcp_rmem, | |
1978 | .max_header = MAX_TCP_HEADER, | |
1979 | .obj_size = sizeof(struct tcp6_sock), | |
3ab5aee7 | 1980 | .slab_flags = SLAB_DESTROY_BY_RCU, |
6d6ee43e | 1981 | .twsk_prot = &tcp6_timewait_sock_ops, |
60236fdd | 1982 | .rsk_prot = &tcp6_request_sock_ops, |
39d8cda7 | 1983 | .h.hashinfo = &tcp_hashinfo, |
7ba42910 | 1984 | .no_autobind = true, |
543d9cfe ACM |
1985 | #ifdef CONFIG_COMPAT |
1986 | .compat_setsockopt = compat_tcp_setsockopt, | |
1987 | .compat_getsockopt = compat_tcp_getsockopt, | |
1988 | #endif | |
c255a458 | 1989 | #ifdef CONFIG_MEMCG_KMEM |
d1a4c0b3 GC |
1990 | .proto_cgroup = tcp_proto_cgroup, |
1991 | #endif | |
f77d6021 | 1992 | .clear_sk = tcp_v6_clear_sk, |
1da177e4 LT |
1993 | }; |
1994 | ||
41135cc8 | 1995 | static const struct inet6_protocol tcpv6_protocol = { |
c7109986 | 1996 | .early_demux = tcp_v6_early_demux, |
1da177e4 LT |
1997 | .handler = tcp_v6_rcv, |
1998 | .err_handler = tcp_v6_err, | |
1999 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, | |
2000 | }; | |
2001 | ||
1da177e4 LT |
2002 | static struct inet_protosw tcpv6_protosw = { |
2003 | .type = SOCK_STREAM, | |
2004 | .protocol = IPPROTO_TCP, | |
2005 | .prot = &tcpv6_prot, | |
2006 | .ops = &inet6_stream_ops, | |
d83d8461 ACM |
2007 | .flags = INET_PROTOSW_PERMANENT | |
2008 | INET_PROTOSW_ICSK, | |
1da177e4 LT |
2009 | }; |
2010 | ||
2c8c1e72 | 2011 | static int __net_init tcpv6_net_init(struct net *net) |
93ec926b | 2012 | { |
5677242f DL |
2013 | return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6, |
2014 | SOCK_RAW, IPPROTO_TCP, net); | |
93ec926b DL |
2015 | } |
2016 | ||
2c8c1e72 | 2017 | static void __net_exit tcpv6_net_exit(struct net *net) |
93ec926b | 2018 | { |
5677242f | 2019 | inet_ctl_sock_destroy(net->ipv6.tcp_sk); |
b099ce26 EB |
2020 | } |
2021 | ||
2c8c1e72 | 2022 | static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list) |
b099ce26 EB |
2023 | { |
2024 | inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6); | |
93ec926b DL |
2025 | } |
2026 | ||
2027 | static struct pernet_operations tcpv6_net_ops = { | |
b099ce26 EB |
2028 | .init = tcpv6_net_init, |
2029 | .exit = tcpv6_net_exit, | |
2030 | .exit_batch = tcpv6_net_exit_batch, | |
93ec926b DL |
2031 | }; |
2032 | ||
7f4e4868 | 2033 | int __init tcpv6_init(void) |
1da177e4 | 2034 | { |
7f4e4868 DL |
2035 | int ret; |
2036 | ||
3336288a VY |
2037 | ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP); |
2038 | if (ret) | |
c6b641a4 | 2039 | goto out; |
3336288a | 2040 | |
1da177e4 | 2041 | /* register inet6 protocol */ |
7f4e4868 DL |
2042 | ret = inet6_register_protosw(&tcpv6_protosw); |
2043 | if (ret) | |
2044 | goto out_tcpv6_protocol; | |
2045 | ||
93ec926b | 2046 | ret = register_pernet_subsys(&tcpv6_net_ops); |
7f4e4868 DL |
2047 | if (ret) |
2048 | goto out_tcpv6_protosw; | |
2049 | out: | |
2050 | return ret; | |
ae0f7d5f | 2051 | |
7f4e4868 DL |
2052 | out_tcpv6_protosw: |
2053 | inet6_unregister_protosw(&tcpv6_protosw); | |
3336288a VY |
2054 | out_tcpv6_protocol: |
2055 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
7f4e4868 DL |
2056 | goto out; |
2057 | } | |
2058 | ||
09f7709f | 2059 | void tcpv6_exit(void) |
7f4e4868 | 2060 | { |
93ec926b | 2061 | unregister_pernet_subsys(&tcpv6_net_ops); |
7f4e4868 DL |
2062 | inet6_unregister_protosw(&tcpv6_protosw); |
2063 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
1da177e4 | 2064 | } |