]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * TCP over IPv6 | |
3 | * Linux INET6 implementation | |
4 | * | |
5 | * Authors: | |
6 | * Pedro Roque <roque@di.fc.ul.pt> | |
7 | * | |
8 | * Based on: | |
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 | ||
26 | #include <linux/bottom_half.h> | |
27 | #include <linux/module.h> | |
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> | |
41 | #include <linux/slab.h> | |
42 | #include <linux/uaccess.h> | |
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> | |
49 | #include <net/inet6_hashtables.h> | |
50 | #include <net/inet6_connection_sock.h> | |
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> | |
59 | #include <net/snmp.h> | |
60 | #include <net/dsfield.h> | |
61 | #include <net/timewait_sock.h> | |
62 | #include <net/inet_common.h> | |
63 | #include <net/secure_seq.h> | |
64 | #include <net/tcp_memcontrol.h> | |
65 | #include <net/busy_poll.h> | |
66 | ||
67 | #include <linux/proc_fs.h> | |
68 | #include <linux/seq_file.h> | |
69 | ||
70 | #include <linux/crypto.h> | |
71 | #include <linux/scatterlist.h> | |
72 | ||
73 | static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb); | |
74 | static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, | |
75 | struct request_sock *req); | |
76 | ||
77 | static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb); | |
78 | ||
79 | static const struct inet_connection_sock_af_ops ipv6_mapped; | |
80 | static const struct inet_connection_sock_af_ops ipv6_specific; | |
81 | #ifdef CONFIG_TCP_MD5SIG | |
82 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific; | |
83 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific; | |
84 | #else | |
85 | static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk, | |
86 | const struct in6_addr *addr) | |
87 | { | |
88 | return NULL; | |
89 | } | |
90 | #endif | |
91 | ||
92 | static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) | |
93 | { | |
94 | struct dst_entry *dst = skb_dst(skb); | |
95 | ||
96 | if (dst) { | |
97 | const struct rt6_info *rt = (const struct rt6_info *)dst; | |
98 | ||
99 | dst_hold(dst); | |
100 | sk->sk_rx_dst = dst; | |
101 | inet_sk(sk)->rx_dst_ifindex = skb->skb_iif; | |
102 | if (rt->rt6i_node) | |
103 | inet6_sk(sk)->rx_dst_cookie = rt->rt6i_node->fn_sernum; | |
104 | } | |
105 | } | |
106 | ||
107 | static void tcp_v6_hash(struct sock *sk) | |
108 | { | |
109 | if (sk->sk_state != TCP_CLOSE) { | |
110 | if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) { | |
111 | tcp_prot.hash(sk); | |
112 | return; | |
113 | } | |
114 | local_bh_disable(); | |
115 | __inet6_hash(sk, NULL); | |
116 | local_bh_enable(); | |
117 | } | |
118 | } | |
119 | ||
120 | static __u32 tcp_v6_init_sequence(const struct sk_buff *skb) | |
121 | { | |
122 | return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32, | |
123 | ipv6_hdr(skb)->saddr.s6_addr32, | |
124 | tcp_hdr(skb)->dest, | |
125 | tcp_hdr(skb)->source); | |
126 | } | |
127 | ||
128 | static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, | |
129 | int addr_len) | |
130 | { | |
131 | struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr; | |
132 | struct inet_sock *inet = inet_sk(sk); | |
133 | struct inet_connection_sock *icsk = inet_csk(sk); | |
134 | struct ipv6_pinfo *np = inet6_sk(sk); | |
135 | struct tcp_sock *tp = tcp_sk(sk); | |
136 | struct in6_addr *saddr = NULL, *final_p, final; | |
137 | struct rt6_info *rt; | |
138 | struct flowi6 fl6; | |
139 | struct dst_entry *dst; | |
140 | int addr_type; | |
141 | int err; | |
142 | ||
143 | if (addr_len < SIN6_LEN_RFC2133) | |
144 | return -EINVAL; | |
145 | ||
146 | if (usin->sin6_family != AF_INET6) | |
147 | return -EAFNOSUPPORT; | |
148 | ||
149 | memset(&fl6, 0, sizeof(fl6)); | |
150 | ||
151 | if (np->sndflow) { | |
152 | fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK; | |
153 | IP6_ECN_flow_init(fl6.flowlabel); | |
154 | if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) { | |
155 | struct ip6_flowlabel *flowlabel; | |
156 | flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); | |
157 | if (flowlabel == NULL) | |
158 | return -EINVAL; | |
159 | fl6_sock_release(flowlabel); | |
160 | } | |
161 | } | |
162 | ||
163 | /* | |
164 | * connect() to INADDR_ANY means loopback (BSD'ism). | |
165 | */ | |
166 | ||
167 | if (ipv6_addr_any(&usin->sin6_addr)) | |
168 | usin->sin6_addr.s6_addr[15] = 0x1; | |
169 | ||
170 | addr_type = ipv6_addr_type(&usin->sin6_addr); | |
171 | ||
172 | if (addr_type & IPV6_ADDR_MULTICAST) | |
173 | return -ENETUNREACH; | |
174 | ||
175 | if (addr_type&IPV6_ADDR_LINKLOCAL) { | |
176 | if (addr_len >= sizeof(struct sockaddr_in6) && | |
177 | usin->sin6_scope_id) { | |
178 | /* If interface is set while binding, indices | |
179 | * must coincide. | |
180 | */ | |
181 | if (sk->sk_bound_dev_if && | |
182 | sk->sk_bound_dev_if != usin->sin6_scope_id) | |
183 | return -EINVAL; | |
184 | ||
185 | sk->sk_bound_dev_if = usin->sin6_scope_id; | |
186 | } | |
187 | ||
188 | /* Connect to link-local address requires an interface */ | |
189 | if (!sk->sk_bound_dev_if) | |
190 | return -EINVAL; | |
191 | } | |
192 | ||
193 | if (tp->rx_opt.ts_recent_stamp && | |
194 | !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) { | |
195 | tp->rx_opt.ts_recent = 0; | |
196 | tp->rx_opt.ts_recent_stamp = 0; | |
197 | tp->write_seq = 0; | |
198 | } | |
199 | ||
200 | sk->sk_v6_daddr = usin->sin6_addr; | |
201 | np->flow_label = fl6.flowlabel; | |
202 | ||
203 | /* | |
204 | * TCP over IPv4 | |
205 | */ | |
206 | ||
207 | if (addr_type == IPV6_ADDR_MAPPED) { | |
208 | u32 exthdrlen = icsk->icsk_ext_hdr_len; | |
209 | struct sockaddr_in sin; | |
210 | ||
211 | SOCK_DEBUG(sk, "connect: ipv4 mapped\n"); | |
212 | ||
213 | if (__ipv6_only_sock(sk)) | |
214 | return -ENETUNREACH; | |
215 | ||
216 | sin.sin_family = AF_INET; | |
217 | sin.sin_port = usin->sin6_port; | |
218 | sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3]; | |
219 | ||
220 | icsk->icsk_af_ops = &ipv6_mapped; | |
221 | sk->sk_backlog_rcv = tcp_v4_do_rcv; | |
222 | #ifdef CONFIG_TCP_MD5SIG | |
223 | tp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
224 | #endif | |
225 | ||
226 | err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin)); | |
227 | ||
228 | if (err) { | |
229 | icsk->icsk_ext_hdr_len = exthdrlen; | |
230 | icsk->icsk_af_ops = &ipv6_specific; | |
231 | sk->sk_backlog_rcv = tcp_v6_do_rcv; | |
232 | #ifdef CONFIG_TCP_MD5SIG | |
233 | tp->af_specific = &tcp_sock_ipv6_specific; | |
234 | #endif | |
235 | goto failure; | |
236 | } else { | |
237 | ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr); | |
238 | ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, | |
239 | &sk->sk_v6_rcv_saddr); | |
240 | } | |
241 | ||
242 | return err; | |
243 | } | |
244 | ||
245 | if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) | |
246 | saddr = &sk->sk_v6_rcv_saddr; | |
247 | ||
248 | fl6.flowi6_proto = IPPROTO_TCP; | |
249 | fl6.daddr = sk->sk_v6_daddr; | |
250 | fl6.saddr = saddr ? *saddr : np->saddr; | |
251 | fl6.flowi6_oif = sk->sk_bound_dev_if; | |
252 | fl6.flowi6_mark = sk->sk_mark; | |
253 | fl6.fl6_dport = usin->sin6_port; | |
254 | fl6.fl6_sport = inet->inet_sport; | |
255 | ||
256 | final_p = fl6_update_dst(&fl6, np->opt, &final); | |
257 | ||
258 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); | |
259 | ||
260 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p); | |
261 | if (IS_ERR(dst)) { | |
262 | err = PTR_ERR(dst); | |
263 | goto failure; | |
264 | } | |
265 | ||
266 | if (saddr == NULL) { | |
267 | saddr = &fl6.saddr; | |
268 | sk->sk_v6_rcv_saddr = *saddr; | |
269 | } | |
270 | ||
271 | /* set the source address */ | |
272 | np->saddr = *saddr; | |
273 | inet->inet_rcv_saddr = LOOPBACK4_IPV6; | |
274 | ||
275 | sk->sk_gso_type = SKB_GSO_TCPV6; | |
276 | __ip6_dst_store(sk, dst, NULL, NULL); | |
277 | ||
278 | rt = (struct rt6_info *) dst; | |
279 | if (tcp_death_row.sysctl_tw_recycle && | |
280 | !tp->rx_opt.ts_recent_stamp && | |
281 | ipv6_addr_equal(&rt->rt6i_dst.addr, &sk->sk_v6_daddr)) | |
282 | tcp_fetch_timewait_stamp(sk, dst); | |
283 | ||
284 | icsk->icsk_ext_hdr_len = 0; | |
285 | if (np->opt) | |
286 | icsk->icsk_ext_hdr_len = (np->opt->opt_flen + | |
287 | np->opt->opt_nflen); | |
288 | ||
289 | tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); | |
290 | ||
291 | inet->inet_dport = usin->sin6_port; | |
292 | ||
293 | tcp_set_state(sk, TCP_SYN_SENT); | |
294 | err = inet6_hash_connect(&tcp_death_row, sk); | |
295 | if (err) | |
296 | goto late_failure; | |
297 | ||
298 | ip6_set_txhash(sk); | |
299 | ||
300 | if (!tp->write_seq && likely(!tp->repair)) | |
301 | tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32, | |
302 | sk->sk_v6_daddr.s6_addr32, | |
303 | inet->inet_sport, | |
304 | inet->inet_dport); | |
305 | ||
306 | err = tcp_connect(sk); | |
307 | if (err) | |
308 | goto late_failure; | |
309 | ||
310 | return 0; | |
311 | ||
312 | late_failure: | |
313 | tcp_set_state(sk, TCP_CLOSE); | |
314 | __sk_dst_reset(sk); | |
315 | failure: | |
316 | inet->inet_dport = 0; | |
317 | sk->sk_route_caps = 0; | |
318 | return err; | |
319 | } | |
320 | ||
321 | static void tcp_v6_mtu_reduced(struct sock *sk) | |
322 | { | |
323 | struct dst_entry *dst; | |
324 | ||
325 | if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) | |
326 | return; | |
327 | ||
328 | dst = inet6_csk_update_pmtu(sk, tcp_sk(sk)->mtu_info); | |
329 | if (!dst) | |
330 | return; | |
331 | ||
332 | if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) { | |
333 | tcp_sync_mss(sk, dst_mtu(dst)); | |
334 | tcp_simple_retransmit(sk); | |
335 | } | |
336 | } | |
337 | ||
338 | static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, | |
339 | u8 type, u8 code, int offset, __be32 info) | |
340 | { | |
341 | const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; | |
342 | const struct tcphdr *th = (struct tcphdr *)(skb->data+offset); | |
343 | struct ipv6_pinfo *np; | |
344 | struct sock *sk; | |
345 | int err; | |
346 | struct tcp_sock *tp; | |
347 | struct request_sock *fastopen; | |
348 | __u32 seq, snd_una; | |
349 | struct net *net = dev_net(skb->dev); | |
350 | ||
351 | sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr, | |
352 | th->dest, &hdr->saddr, th->source, skb->dev->ifindex); | |
353 | ||
354 | if (sk == NULL) { | |
355 | ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev), | |
356 | ICMP6_MIB_INERRORS); | |
357 | return; | |
358 | } | |
359 | ||
360 | if (sk->sk_state == TCP_TIME_WAIT) { | |
361 | inet_twsk_put(inet_twsk(sk)); | |
362 | return; | |
363 | } | |
364 | ||
365 | bh_lock_sock(sk); | |
366 | if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG) | |
367 | NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS); | |
368 | ||
369 | if (sk->sk_state == TCP_CLOSE) | |
370 | goto out; | |
371 | ||
372 | if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) { | |
373 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
374 | goto out; | |
375 | } | |
376 | ||
377 | tp = tcp_sk(sk); | |
378 | seq = ntohl(th->seq); | |
379 | /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */ | |
380 | fastopen = tp->fastopen_rsk; | |
381 | snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una; | |
382 | if (sk->sk_state != TCP_LISTEN && | |
383 | !between(seq, snd_una, tp->snd_nxt)) { | |
384 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); | |
385 | goto out; | |
386 | } | |
387 | ||
388 | np = inet6_sk(sk); | |
389 | ||
390 | if (type == NDISC_REDIRECT) { | |
391 | struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie); | |
392 | ||
393 | if (dst) | |
394 | dst->ops->redirect(dst, sk, skb); | |
395 | goto out; | |
396 | } | |
397 | ||
398 | if (type == ICMPV6_PKT_TOOBIG) { | |
399 | /* We are not interested in TCP_LISTEN and open_requests | |
400 | * (SYN-ACKs send out by Linux are always <576bytes so | |
401 | * they should go through unfragmented). | |
402 | */ | |
403 | if (sk->sk_state == TCP_LISTEN) | |
404 | goto out; | |
405 | ||
406 | if (!ip6_sk_accept_pmtu(sk)) | |
407 | goto out; | |
408 | ||
409 | tp->mtu_info = ntohl(info); | |
410 | if (!sock_owned_by_user(sk)) | |
411 | tcp_v6_mtu_reduced(sk); | |
412 | else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED, | |
413 | &tp->tsq_flags)) | |
414 | sock_hold(sk); | |
415 | goto out; | |
416 | } | |
417 | ||
418 | icmpv6_err_convert(type, code, &err); | |
419 | ||
420 | /* Might be for an request_sock */ | |
421 | switch (sk->sk_state) { | |
422 | struct request_sock *req, **prev; | |
423 | case TCP_LISTEN: | |
424 | if (sock_owned_by_user(sk)) | |
425 | goto out; | |
426 | ||
427 | /* Note : We use inet6_iif() here, not tcp_v6_iif() */ | |
428 | req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr, | |
429 | &hdr->saddr, inet6_iif(skb)); | |
430 | if (!req) | |
431 | goto out; | |
432 | ||
433 | /* ICMPs are not backlogged, hence we cannot get | |
434 | * an established socket here. | |
435 | */ | |
436 | WARN_ON(req->sk != NULL); | |
437 | ||
438 | if (seq != tcp_rsk(req)->snt_isn) { | |
439 | NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); | |
440 | goto out; | |
441 | } | |
442 | ||
443 | inet_csk_reqsk_queue_drop(sk, req, prev); | |
444 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); | |
445 | goto out; | |
446 | ||
447 | case TCP_SYN_SENT: | |
448 | case TCP_SYN_RECV: | |
449 | /* Only in fast or simultaneous open. If a fast open socket is | |
450 | * is already accepted it is treated as a connected one below. | |
451 | */ | |
452 | if (fastopen && fastopen->sk == NULL) | |
453 | break; | |
454 | ||
455 | if (!sock_owned_by_user(sk)) { | |
456 | sk->sk_err = err; | |
457 | sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */ | |
458 | ||
459 | tcp_done(sk); | |
460 | } else | |
461 | sk->sk_err_soft = err; | |
462 | goto out; | |
463 | } | |
464 | ||
465 | if (!sock_owned_by_user(sk) && np->recverr) { | |
466 | sk->sk_err = err; | |
467 | sk->sk_error_report(sk); | |
468 | } else | |
469 | sk->sk_err_soft = err; | |
470 | ||
471 | out: | |
472 | bh_unlock_sock(sk); | |
473 | sock_put(sk); | |
474 | } | |
475 | ||
476 | ||
477 | static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst, | |
478 | struct flowi *fl, | |
479 | struct request_sock *req, | |
480 | u16 queue_mapping, | |
481 | struct tcp_fastopen_cookie *foc) | |
482 | { | |
483 | struct inet_request_sock *ireq = inet_rsk(req); | |
484 | struct ipv6_pinfo *np = inet6_sk(sk); | |
485 | struct flowi6 *fl6 = &fl->u.ip6; | |
486 | struct sk_buff *skb; | |
487 | int err = -ENOMEM; | |
488 | ||
489 | /* First, grab a route. */ | |
490 | if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL) | |
491 | goto done; | |
492 | ||
493 | skb = tcp_make_synack(sk, dst, req, foc); | |
494 | ||
495 | if (skb) { | |
496 | __tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr, | |
497 | &ireq->ir_v6_rmt_addr); | |
498 | ||
499 | fl6->daddr = ireq->ir_v6_rmt_addr; | |
500 | if (np->repflow && (ireq->pktopts != NULL)) | |
501 | fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts)); | |
502 | ||
503 | skb_set_queue_mapping(skb, queue_mapping); | |
504 | err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass); | |
505 | err = net_xmit_eval(err); | |
506 | } | |
507 | ||
508 | done: | |
509 | return err; | |
510 | } | |
511 | ||
512 | ||
513 | static void tcp_v6_reqsk_destructor(struct request_sock *req) | |
514 | { | |
515 | kfree_skb(inet_rsk(req)->pktopts); | |
516 | } | |
517 | ||
518 | #ifdef CONFIG_TCP_MD5SIG | |
519 | static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk, | |
520 | const struct in6_addr *addr) | |
521 | { | |
522 | return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6); | |
523 | } | |
524 | ||
525 | static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk, | |
526 | struct sock *addr_sk) | |
527 | { | |
528 | return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr); | |
529 | } | |
530 | ||
531 | static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk, | |
532 | struct request_sock *req) | |
533 | { | |
534 | return tcp_v6_md5_do_lookup(sk, &inet_rsk(req)->ir_v6_rmt_addr); | |
535 | } | |
536 | ||
537 | static int tcp_v6_parse_md5_keys(struct sock *sk, char __user *optval, | |
538 | int optlen) | |
539 | { | |
540 | struct tcp_md5sig cmd; | |
541 | struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr; | |
542 | ||
543 | if (optlen < sizeof(cmd)) | |
544 | return -EINVAL; | |
545 | ||
546 | if (copy_from_user(&cmd, optval, sizeof(cmd))) | |
547 | return -EFAULT; | |
548 | ||
549 | if (sin6->sin6_family != AF_INET6) | |
550 | return -EINVAL; | |
551 | ||
552 | if (!cmd.tcpm_keylen) { | |
553 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) | |
554 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3], | |
555 | AF_INET); | |
556 | return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr, | |
557 | AF_INET6); | |
558 | } | |
559 | ||
560 | if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN) | |
561 | return -EINVAL; | |
562 | ||
563 | if (ipv6_addr_v4mapped(&sin6->sin6_addr)) | |
564 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3], | |
565 | AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL); | |
566 | ||
567 | return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr, | |
568 | AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL); | |
569 | } | |
570 | ||
571 | static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp, | |
572 | const struct in6_addr *daddr, | |
573 | const struct in6_addr *saddr, int nbytes) | |
574 | { | |
575 | struct tcp6_pseudohdr *bp; | |
576 | struct scatterlist sg; | |
577 | ||
578 | bp = &hp->md5_blk.ip6; | |
579 | /* 1. TCP pseudo-header (RFC2460) */ | |
580 | bp->saddr = *saddr; | |
581 | bp->daddr = *daddr; | |
582 | bp->protocol = cpu_to_be32(IPPROTO_TCP); | |
583 | bp->len = cpu_to_be32(nbytes); | |
584 | ||
585 | sg_init_one(&sg, bp, sizeof(*bp)); | |
586 | return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp)); | |
587 | } | |
588 | ||
589 | static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key, | |
590 | const struct in6_addr *daddr, struct in6_addr *saddr, | |
591 | const struct tcphdr *th) | |
592 | { | |
593 | struct tcp_md5sig_pool *hp; | |
594 | struct hash_desc *desc; | |
595 | ||
596 | hp = tcp_get_md5sig_pool(); | |
597 | if (!hp) | |
598 | goto clear_hash_noput; | |
599 | desc = &hp->md5_desc; | |
600 | ||
601 | if (crypto_hash_init(desc)) | |
602 | goto clear_hash; | |
603 | if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2)) | |
604 | goto clear_hash; | |
605 | if (tcp_md5_hash_header(hp, th)) | |
606 | goto clear_hash; | |
607 | if (tcp_md5_hash_key(hp, key)) | |
608 | goto clear_hash; | |
609 | if (crypto_hash_final(desc, md5_hash)) | |
610 | goto clear_hash; | |
611 | ||
612 | tcp_put_md5sig_pool(); | |
613 | return 0; | |
614 | ||
615 | clear_hash: | |
616 | tcp_put_md5sig_pool(); | |
617 | clear_hash_noput: | |
618 | memset(md5_hash, 0, 16); | |
619 | return 1; | |
620 | } | |
621 | ||
622 | static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key, | |
623 | const struct sock *sk, | |
624 | const struct request_sock *req, | |
625 | const struct sk_buff *skb) | |
626 | { | |
627 | const struct in6_addr *saddr, *daddr; | |
628 | struct tcp_md5sig_pool *hp; | |
629 | struct hash_desc *desc; | |
630 | const struct tcphdr *th = tcp_hdr(skb); | |
631 | ||
632 | if (sk) { | |
633 | saddr = &inet6_sk(sk)->saddr; | |
634 | daddr = &sk->sk_v6_daddr; | |
635 | } else if (req) { | |
636 | saddr = &inet_rsk(req)->ir_v6_loc_addr; | |
637 | daddr = &inet_rsk(req)->ir_v6_rmt_addr; | |
638 | } else { | |
639 | const struct ipv6hdr *ip6h = ipv6_hdr(skb); | |
640 | saddr = &ip6h->saddr; | |
641 | daddr = &ip6h->daddr; | |
642 | } | |
643 | ||
644 | hp = tcp_get_md5sig_pool(); | |
645 | if (!hp) | |
646 | goto clear_hash_noput; | |
647 | desc = &hp->md5_desc; | |
648 | ||
649 | if (crypto_hash_init(desc)) | |
650 | goto clear_hash; | |
651 | ||
652 | if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len)) | |
653 | goto clear_hash; | |
654 | if (tcp_md5_hash_header(hp, th)) | |
655 | goto clear_hash; | |
656 | if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2)) | |
657 | goto clear_hash; | |
658 | if (tcp_md5_hash_key(hp, key)) | |
659 | goto clear_hash; | |
660 | if (crypto_hash_final(desc, md5_hash)) | |
661 | goto clear_hash; | |
662 | ||
663 | tcp_put_md5sig_pool(); | |
664 | return 0; | |
665 | ||
666 | clear_hash: | |
667 | tcp_put_md5sig_pool(); | |
668 | clear_hash_noput: | |
669 | memset(md5_hash, 0, 16); | |
670 | return 1; | |
671 | } | |
672 | ||
673 | static int __tcp_v6_inbound_md5_hash(struct sock *sk, | |
674 | const struct sk_buff *skb) | |
675 | { | |
676 | const __u8 *hash_location = NULL; | |
677 | struct tcp_md5sig_key *hash_expected; | |
678 | const struct ipv6hdr *ip6h = ipv6_hdr(skb); | |
679 | const struct tcphdr *th = tcp_hdr(skb); | |
680 | int genhash; | |
681 | u8 newhash[16]; | |
682 | ||
683 | hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr); | |
684 | hash_location = tcp_parse_md5sig_option(th); | |
685 | ||
686 | /* We've parsed the options - do we have a hash? */ | |
687 | if (!hash_expected && !hash_location) | |
688 | return 0; | |
689 | ||
690 | if (hash_expected && !hash_location) { | |
691 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND); | |
692 | return 1; | |
693 | } | |
694 | ||
695 | if (!hash_expected && hash_location) { | |
696 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED); | |
697 | return 1; | |
698 | } | |
699 | ||
700 | /* check the signature */ | |
701 | genhash = tcp_v6_md5_hash_skb(newhash, | |
702 | hash_expected, | |
703 | NULL, NULL, skb); | |
704 | ||
705 | if (genhash || memcmp(hash_location, newhash, 16) != 0) { | |
706 | net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n", | |
707 | genhash ? "failed" : "mismatch", | |
708 | &ip6h->saddr, ntohs(th->source), | |
709 | &ip6h->daddr, ntohs(th->dest)); | |
710 | return 1; | |
711 | } | |
712 | return 0; | |
713 | } | |
714 | ||
715 | static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb) | |
716 | { | |
717 | int ret; | |
718 | ||
719 | rcu_read_lock(); | |
720 | ret = __tcp_v6_inbound_md5_hash(sk, skb); | |
721 | rcu_read_unlock(); | |
722 | ||
723 | return ret; | |
724 | } | |
725 | ||
726 | #endif | |
727 | ||
728 | static void tcp_v6_init_req(struct request_sock *req, struct sock *sk, | |
729 | struct sk_buff *skb) | |
730 | { | |
731 | struct inet_request_sock *ireq = inet_rsk(req); | |
732 | struct ipv6_pinfo *np = inet6_sk(sk); | |
733 | ||
734 | ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr; | |
735 | ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr; | |
736 | ||
737 | ireq->ir_iif = sk->sk_bound_dev_if; | |
738 | ||
739 | /* So that link locals have meaning */ | |
740 | if (!sk->sk_bound_dev_if && | |
741 | ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL) | |
742 | ireq->ir_iif = tcp_v6_iif(skb); | |
743 | ||
744 | if (!TCP_SKB_CB(skb)->tcp_tw_isn && | |
745 | (ipv6_opt_accepted(sk, skb, &TCP_SKB_CB(skb)->header.h6) || | |
746 | np->rxopt.bits.rxinfo || | |
747 | np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim || | |
748 | np->rxopt.bits.rxohlim || np->repflow)) { | |
749 | atomic_inc(&skb->users); | |
750 | ireq->pktopts = skb; | |
751 | } | |
752 | } | |
753 | ||
754 | static struct dst_entry *tcp_v6_route_req(struct sock *sk, struct flowi *fl, | |
755 | const struct request_sock *req, | |
756 | bool *strict) | |
757 | { | |
758 | if (strict) | |
759 | *strict = true; | |
760 | return inet6_csk_route_req(sk, &fl->u.ip6, req); | |
761 | } | |
762 | ||
763 | struct request_sock_ops tcp6_request_sock_ops __read_mostly = { | |
764 | .family = AF_INET6, | |
765 | .obj_size = sizeof(struct tcp6_request_sock), | |
766 | .rtx_syn_ack = tcp_rtx_synack, | |
767 | .send_ack = tcp_v6_reqsk_send_ack, | |
768 | .destructor = tcp_v6_reqsk_destructor, | |
769 | .send_reset = tcp_v6_send_reset, | |
770 | .syn_ack_timeout = tcp_syn_ack_timeout, | |
771 | }; | |
772 | ||
773 | static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = { | |
774 | .mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - | |
775 | sizeof(struct ipv6hdr), | |
776 | #ifdef CONFIG_TCP_MD5SIG | |
777 | .md5_lookup = tcp_v6_reqsk_md5_lookup, | |
778 | .calc_md5_hash = tcp_v6_md5_hash_skb, | |
779 | #endif | |
780 | .init_req = tcp_v6_init_req, | |
781 | #ifdef CONFIG_SYN_COOKIES | |
782 | .cookie_init_seq = cookie_v6_init_sequence, | |
783 | #endif | |
784 | .route_req = tcp_v6_route_req, | |
785 | .init_seq = tcp_v6_init_sequence, | |
786 | .send_synack = tcp_v6_send_synack, | |
787 | .queue_hash_add = inet6_csk_reqsk_queue_hash_add, | |
788 | }; | |
789 | ||
790 | static void tcp_v6_send_response(struct sock *sk, struct sk_buff *skb, u32 seq, | |
791 | u32 ack, u32 win, u32 tsval, u32 tsecr, | |
792 | int oif, struct tcp_md5sig_key *key, int rst, | |
793 | u8 tclass, u32 label) | |
794 | { | |
795 | const struct tcphdr *th = tcp_hdr(skb); | |
796 | struct tcphdr *t1; | |
797 | struct sk_buff *buff; | |
798 | struct flowi6 fl6; | |
799 | struct net *net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev); | |
800 | struct sock *ctl_sk = net->ipv6.tcp_sk; | |
801 | unsigned int tot_len = sizeof(struct tcphdr); | |
802 | struct dst_entry *dst; | |
803 | __be32 *topt; | |
804 | ||
805 | if (tsecr) | |
806 | tot_len += TCPOLEN_TSTAMP_ALIGNED; | |
807 | #ifdef CONFIG_TCP_MD5SIG | |
808 | if (key) | |
809 | tot_len += TCPOLEN_MD5SIG_ALIGNED; | |
810 | #endif | |
811 | ||
812 | buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len, | |
813 | GFP_ATOMIC); | |
814 | if (buff == NULL) | |
815 | return; | |
816 | ||
817 | skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len); | |
818 | ||
819 | t1 = (struct tcphdr *) skb_push(buff, tot_len); | |
820 | skb_reset_transport_header(buff); | |
821 | ||
822 | /* Swap the send and the receive. */ | |
823 | memset(t1, 0, sizeof(*t1)); | |
824 | t1->dest = th->source; | |
825 | t1->source = th->dest; | |
826 | t1->doff = tot_len / 4; | |
827 | t1->seq = htonl(seq); | |
828 | t1->ack_seq = htonl(ack); | |
829 | t1->ack = !rst || !th->ack; | |
830 | t1->rst = rst; | |
831 | t1->window = htons(win); | |
832 | ||
833 | topt = (__be32 *)(t1 + 1); | |
834 | ||
835 | if (tsecr) { | |
836 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | | |
837 | (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP); | |
838 | *topt++ = htonl(tsval); | |
839 | *topt++ = htonl(tsecr); | |
840 | } | |
841 | ||
842 | #ifdef CONFIG_TCP_MD5SIG | |
843 | if (key) { | |
844 | *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | | |
845 | (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG); | |
846 | tcp_v6_md5_hash_hdr((__u8 *)topt, key, | |
847 | &ipv6_hdr(skb)->saddr, | |
848 | &ipv6_hdr(skb)->daddr, t1); | |
849 | } | |
850 | #endif | |
851 | ||
852 | memset(&fl6, 0, sizeof(fl6)); | |
853 | fl6.daddr = ipv6_hdr(skb)->saddr; | |
854 | fl6.saddr = ipv6_hdr(skb)->daddr; | |
855 | fl6.flowlabel = label; | |
856 | ||
857 | buff->ip_summed = CHECKSUM_PARTIAL; | |
858 | buff->csum = 0; | |
859 | ||
860 | __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr); | |
861 | ||
862 | fl6.flowi6_proto = IPPROTO_TCP; | |
863 | if (rt6_need_strict(&fl6.daddr) && !oif) | |
864 | fl6.flowi6_oif = tcp_v6_iif(skb); | |
865 | else | |
866 | fl6.flowi6_oif = oif; | |
867 | fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark); | |
868 | fl6.fl6_dport = t1->dest; | |
869 | fl6.fl6_sport = t1->source; | |
870 | security_skb_classify_flow(skb, flowi6_to_flowi(&fl6)); | |
871 | ||
872 | /* Pass a socket to ip6_dst_lookup either it is for RST | |
873 | * Underlying function will use this to retrieve the network | |
874 | * namespace | |
875 | */ | |
876 | dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL); | |
877 | if (!IS_ERR(dst)) { | |
878 | skb_dst_set(buff, dst); | |
879 | ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass); | |
880 | TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS); | |
881 | if (rst) | |
882 | TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS); | |
883 | return; | |
884 | } | |
885 | ||
886 | kfree_skb(buff); | |
887 | } | |
888 | ||
889 | static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb) | |
890 | { | |
891 | const struct tcphdr *th = tcp_hdr(skb); | |
892 | u32 seq = 0, ack_seq = 0; | |
893 | struct tcp_md5sig_key *key = NULL; | |
894 | #ifdef CONFIG_TCP_MD5SIG | |
895 | const __u8 *hash_location = NULL; | |
896 | struct ipv6hdr *ipv6h = ipv6_hdr(skb); | |
897 | unsigned char newhash[16]; | |
898 | int genhash; | |
899 | struct sock *sk1 = NULL; | |
900 | #endif | |
901 | int oif; | |
902 | ||
903 | if (th->rst) | |
904 | return; | |
905 | ||
906 | /* If sk not NULL, it means we did a successful lookup and incoming | |
907 | * route had to be correct. prequeue might have dropped our dst. | |
908 | */ | |
909 | if (!sk && !ipv6_unicast_destination(skb)) | |
910 | return; | |
911 | ||
912 | #ifdef CONFIG_TCP_MD5SIG | |
913 | hash_location = tcp_parse_md5sig_option(th); | |
914 | if (!sk && hash_location) { | |
915 | /* | |
916 | * active side is lost. Try to find listening socket through | |
917 | * source port, and then find md5 key through listening socket. | |
918 | * we are not loose security here: | |
919 | * Incoming packet is checked with md5 hash with finding key, | |
920 | * no RST generated if md5 hash doesn't match. | |
921 | */ | |
922 | sk1 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev), | |
923 | &tcp_hashinfo, &ipv6h->saddr, | |
924 | th->source, &ipv6h->daddr, | |
925 | ntohs(th->source), tcp_v6_iif(skb)); | |
926 | if (!sk1) | |
927 | return; | |
928 | ||
929 | rcu_read_lock(); | |
930 | key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr); | |
931 | if (!key) | |
932 | goto release_sk1; | |
933 | ||
934 | genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, NULL, skb); | |
935 | if (genhash || memcmp(hash_location, newhash, 16) != 0) | |
936 | goto release_sk1; | |
937 | } else { | |
938 | key = sk ? tcp_v6_md5_do_lookup(sk, &ipv6h->saddr) : NULL; | |
939 | } | |
940 | #endif | |
941 | ||
942 | if (th->ack) | |
943 | seq = ntohl(th->ack_seq); | |
944 | else | |
945 | ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len - | |
946 | (th->doff << 2); | |
947 | ||
948 | oif = sk ? sk->sk_bound_dev_if : 0; | |
949 | tcp_v6_send_response(sk, skb, seq, ack_seq, 0, 0, 0, oif, key, 1, 0, 0); | |
950 | ||
951 | #ifdef CONFIG_TCP_MD5SIG | |
952 | release_sk1: | |
953 | if (sk1) { | |
954 | rcu_read_unlock(); | |
955 | sock_put(sk1); | |
956 | } | |
957 | #endif | |
958 | } | |
959 | ||
960 | static void tcp_v6_send_ack(struct sock *sk, struct sk_buff *skb, u32 seq, | |
961 | u32 ack, u32 win, u32 tsval, u32 tsecr, int oif, | |
962 | struct tcp_md5sig_key *key, u8 tclass, | |
963 | u32 label) | |
964 | { | |
965 | tcp_v6_send_response(sk, skb, seq, ack, win, tsval, tsecr, oif, key, 0, | |
966 | tclass, label); | |
967 | } | |
968 | ||
969 | static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb) | |
970 | { | |
971 | struct inet_timewait_sock *tw = inet_twsk(sk); | |
972 | struct tcp_timewait_sock *tcptw = tcp_twsk(sk); | |
973 | ||
974 | tcp_v6_send_ack(sk, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt, | |
975 | tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale, | |
976 | tcp_time_stamp + tcptw->tw_ts_offset, | |
977 | tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw), | |
978 | tw->tw_tclass, (tw->tw_flowlabel << 12)); | |
979 | ||
980 | inet_twsk_put(tw); | |
981 | } | |
982 | ||
983 | static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb, | |
984 | struct request_sock *req) | |
985 | { | |
986 | /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV | |
987 | * sk->sk_state == TCP_SYN_RECV -> for Fast Open. | |
988 | */ | |
989 | tcp_v6_send_ack(sk, skb, (sk->sk_state == TCP_LISTEN) ? | |
990 | tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt, | |
991 | tcp_rsk(req)->rcv_nxt, req->rcv_wnd, | |
992 | tcp_time_stamp, req->ts_recent, sk->sk_bound_dev_if, | |
993 | tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr), | |
994 | 0, 0); | |
995 | } | |
996 | ||
997 | ||
998 | static struct sock *tcp_v6_hnd_req(struct sock *sk, struct sk_buff *skb) | |
999 | { | |
1000 | struct request_sock *req, **prev; | |
1001 | const struct tcphdr *th = tcp_hdr(skb); | |
1002 | struct sock *nsk; | |
1003 | ||
1004 | /* Find possible connection requests. */ | |
1005 | req = inet6_csk_search_req(sk, &prev, th->source, | |
1006 | &ipv6_hdr(skb)->saddr, | |
1007 | &ipv6_hdr(skb)->daddr, tcp_v6_iif(skb)); | |
1008 | if (req) | |
1009 | return tcp_check_req(sk, skb, req, prev, false); | |
1010 | ||
1011 | nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo, | |
1012 | &ipv6_hdr(skb)->saddr, th->source, | |
1013 | &ipv6_hdr(skb)->daddr, ntohs(th->dest), | |
1014 | tcp_v6_iif(skb)); | |
1015 | ||
1016 | if (nsk) { | |
1017 | if (nsk->sk_state != TCP_TIME_WAIT) { | |
1018 | bh_lock_sock(nsk); | |
1019 | return nsk; | |
1020 | } | |
1021 | inet_twsk_put(inet_twsk(nsk)); | |
1022 | return NULL; | |
1023 | } | |
1024 | ||
1025 | #ifdef CONFIG_SYN_COOKIES | |
1026 | if (!th->syn) | |
1027 | sk = cookie_v6_check(sk, skb); | |
1028 | #endif | |
1029 | return sk; | |
1030 | } | |
1031 | ||
1032 | static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb) | |
1033 | { | |
1034 | if (skb->protocol == htons(ETH_P_IP)) | |
1035 | return tcp_v4_conn_request(sk, skb); | |
1036 | ||
1037 | if (!ipv6_unicast_destination(skb)) | |
1038 | goto drop; | |
1039 | ||
1040 | return tcp_conn_request(&tcp6_request_sock_ops, | |
1041 | &tcp_request_sock_ipv6_ops, sk, skb); | |
1042 | ||
1043 | drop: | |
1044 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); | |
1045 | return 0; /* don't send reset */ | |
1046 | } | |
1047 | ||
1048 | static struct sock *tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb, | |
1049 | struct request_sock *req, | |
1050 | struct dst_entry *dst) | |
1051 | { | |
1052 | struct inet_request_sock *ireq; | |
1053 | struct ipv6_pinfo *newnp, *np = inet6_sk(sk); | |
1054 | struct tcp6_sock *newtcp6sk; | |
1055 | struct inet_sock *newinet; | |
1056 | struct tcp_sock *newtp; | |
1057 | struct sock *newsk; | |
1058 | #ifdef CONFIG_TCP_MD5SIG | |
1059 | struct tcp_md5sig_key *key; | |
1060 | #endif | |
1061 | struct flowi6 fl6; | |
1062 | ||
1063 | if (skb->protocol == htons(ETH_P_IP)) { | |
1064 | /* | |
1065 | * v6 mapped | |
1066 | */ | |
1067 | ||
1068 | newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst); | |
1069 | ||
1070 | if (newsk == NULL) | |
1071 | return NULL; | |
1072 | ||
1073 | newtcp6sk = (struct tcp6_sock *)newsk; | |
1074 | inet_sk(newsk)->pinet6 = &newtcp6sk->inet6; | |
1075 | ||
1076 | newinet = inet_sk(newsk); | |
1077 | newnp = inet6_sk(newsk); | |
1078 | newtp = tcp_sk(newsk); | |
1079 | ||
1080 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1081 | ||
1082 | ipv6_addr_set_v4mapped(newinet->inet_daddr, &newsk->sk_v6_daddr); | |
1083 | ||
1084 | ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr); | |
1085 | ||
1086 | newsk->sk_v6_rcv_saddr = newnp->saddr; | |
1087 | ||
1088 | inet_csk(newsk)->icsk_af_ops = &ipv6_mapped; | |
1089 | newsk->sk_backlog_rcv = tcp_v4_do_rcv; | |
1090 | #ifdef CONFIG_TCP_MD5SIG | |
1091 | newtp->af_specific = &tcp_sock_ipv6_mapped_specific; | |
1092 | #endif | |
1093 | ||
1094 | newnp->ipv6_ac_list = NULL; | |
1095 | newnp->ipv6_fl_list = NULL; | |
1096 | newnp->pktoptions = NULL; | |
1097 | newnp->opt = NULL; | |
1098 | newnp->mcast_oif = tcp_v6_iif(skb); | |
1099 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; | |
1100 | newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb)); | |
1101 | if (np->repflow) | |
1102 | newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb)); | |
1103 | ||
1104 | /* | |
1105 | * No need to charge this sock to the relevant IPv6 refcnt debug socks count | |
1106 | * here, tcp_create_openreq_child now does this for us, see the comment in | |
1107 | * that function for the gory details. -acme | |
1108 | */ | |
1109 | ||
1110 | /* It is tricky place. Until this moment IPv4 tcp | |
1111 | worked with IPv6 icsk.icsk_af_ops. | |
1112 | Sync it now. | |
1113 | */ | |
1114 | tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie); | |
1115 | ||
1116 | return newsk; | |
1117 | } | |
1118 | ||
1119 | ireq = inet_rsk(req); | |
1120 | ||
1121 | if (sk_acceptq_is_full(sk)) | |
1122 | goto out_overflow; | |
1123 | ||
1124 | if (!dst) { | |
1125 | dst = inet6_csk_route_req(sk, &fl6, req); | |
1126 | if (!dst) | |
1127 | goto out; | |
1128 | } | |
1129 | ||
1130 | newsk = tcp_create_openreq_child(sk, req, skb); | |
1131 | if (newsk == NULL) | |
1132 | goto out_nonewsk; | |
1133 | ||
1134 | /* | |
1135 | * No need to charge this sock to the relevant IPv6 refcnt debug socks | |
1136 | * count here, tcp_create_openreq_child now does this for us, see the | |
1137 | * comment in that function for the gory details. -acme | |
1138 | */ | |
1139 | ||
1140 | newsk->sk_gso_type = SKB_GSO_TCPV6; | |
1141 | __ip6_dst_store(newsk, dst, NULL, NULL); | |
1142 | inet6_sk_rx_dst_set(newsk, skb); | |
1143 | ||
1144 | newtcp6sk = (struct tcp6_sock *)newsk; | |
1145 | inet_sk(newsk)->pinet6 = &newtcp6sk->inet6; | |
1146 | ||
1147 | newtp = tcp_sk(newsk); | |
1148 | newinet = inet_sk(newsk); | |
1149 | newnp = inet6_sk(newsk); | |
1150 | ||
1151 | memcpy(newnp, np, sizeof(struct ipv6_pinfo)); | |
1152 | ||
1153 | newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr; | |
1154 | newnp->saddr = ireq->ir_v6_loc_addr; | |
1155 | newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr; | |
1156 | newsk->sk_bound_dev_if = ireq->ir_iif; | |
1157 | ||
1158 | ip6_set_txhash(newsk); | |
1159 | ||
1160 | /* Now IPv6 options... | |
1161 | ||
1162 | First: no IPv4 options. | |
1163 | */ | |
1164 | newinet->inet_opt = NULL; | |
1165 | newnp->ipv6_ac_list = NULL; | |
1166 | newnp->ipv6_fl_list = NULL; | |
1167 | ||
1168 | /* Clone RX bits */ | |
1169 | newnp->rxopt.all = np->rxopt.all; | |
1170 | ||
1171 | /* Clone pktoptions received with SYN */ | |
1172 | newnp->pktoptions = NULL; | |
1173 | if (ireq->pktopts != NULL) { | |
1174 | newnp->pktoptions = skb_clone(ireq->pktopts, | |
1175 | sk_gfp_atomic(sk, GFP_ATOMIC)); | |
1176 | consume_skb(ireq->pktopts); | |
1177 | ireq->pktopts = NULL; | |
1178 | if (newnp->pktoptions) | |
1179 | skb_set_owner_r(newnp->pktoptions, newsk); | |
1180 | } | |
1181 | newnp->opt = NULL; | |
1182 | newnp->mcast_oif = tcp_v6_iif(skb); | |
1183 | newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; | |
1184 | newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb)); | |
1185 | if (np->repflow) | |
1186 | newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb)); | |
1187 | ||
1188 | /* Clone native IPv6 options from listening socket (if any) | |
1189 | ||
1190 | Yes, keeping reference count would be much more clever, | |
1191 | but we make one more one thing there: reattach optmem | |
1192 | to newsk. | |
1193 | */ | |
1194 | if (np->opt) | |
1195 | newnp->opt = ipv6_dup_options(newsk, np->opt); | |
1196 | ||
1197 | inet_csk(newsk)->icsk_ext_hdr_len = 0; | |
1198 | if (newnp->opt) | |
1199 | inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + | |
1200 | newnp->opt->opt_flen); | |
1201 | ||
1202 | tcp_sync_mss(newsk, dst_mtu(dst)); | |
1203 | newtp->advmss = dst_metric_advmss(dst); | |
1204 | if (tcp_sk(sk)->rx_opt.user_mss && | |
1205 | tcp_sk(sk)->rx_opt.user_mss < newtp->advmss) | |
1206 | newtp->advmss = tcp_sk(sk)->rx_opt.user_mss; | |
1207 | ||
1208 | tcp_initialize_rcv_mss(newsk); | |
1209 | ||
1210 | newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6; | |
1211 | newinet->inet_rcv_saddr = LOOPBACK4_IPV6; | |
1212 | ||
1213 | #ifdef CONFIG_TCP_MD5SIG | |
1214 | /* Copy over the MD5 key from the original socket */ | |
1215 | key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr); | |
1216 | if (key != NULL) { | |
1217 | /* We're using one, so create a matching key | |
1218 | * on the newsk structure. If we fail to get | |
1219 | * memory, then we end up not copying the key | |
1220 | * across. Shucks. | |
1221 | */ | |
1222 | tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr, | |
1223 | AF_INET6, key->key, key->keylen, | |
1224 | sk_gfp_atomic(sk, GFP_ATOMIC)); | |
1225 | } | |
1226 | #endif | |
1227 | ||
1228 | if (__inet_inherit_port(sk, newsk) < 0) { | |
1229 | inet_csk_prepare_forced_close(newsk); | |
1230 | tcp_done(newsk); | |
1231 | goto out; | |
1232 | } | |
1233 | __inet6_hash(newsk, NULL); | |
1234 | ||
1235 | return newsk; | |
1236 | ||
1237 | out_overflow: | |
1238 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); | |
1239 | out_nonewsk: | |
1240 | dst_release(dst); | |
1241 | out: | |
1242 | NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); | |
1243 | return NULL; | |
1244 | } | |
1245 | ||
1246 | /* The socket must have it's spinlock held when we get | |
1247 | * here. | |
1248 | * | |
1249 | * We have a potential double-lock case here, so even when | |
1250 | * doing backlog processing we use the BH locking scheme. | |
1251 | * This is because we cannot sleep with the original spinlock | |
1252 | * held. | |
1253 | */ | |
1254 | static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) | |
1255 | { | |
1256 | struct ipv6_pinfo *np = inet6_sk(sk); | |
1257 | struct tcp_sock *tp; | |
1258 | struct sk_buff *opt_skb = NULL; | |
1259 | ||
1260 | /* Imagine: socket is IPv6. IPv4 packet arrives, | |
1261 | goes to IPv4 receive handler and backlogged. | |
1262 | From backlog it always goes here. Kerboom... | |
1263 | Fortunately, tcp_rcv_established and rcv_established | |
1264 | handle them correctly, but it is not case with | |
1265 | tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK | |
1266 | */ | |
1267 | ||
1268 | if (skb->protocol == htons(ETH_P_IP)) | |
1269 | return tcp_v4_do_rcv(sk, skb); | |
1270 | ||
1271 | if (sk_filter(sk, skb)) | |
1272 | goto discard; | |
1273 | ||
1274 | /* | |
1275 | * socket locking is here for SMP purposes as backlog rcv | |
1276 | * is currently called with bh processing disabled. | |
1277 | */ | |
1278 | ||
1279 | /* Do Stevens' IPV6_PKTOPTIONS. | |
1280 | ||
1281 | Yes, guys, it is the only place in our code, where we | |
1282 | may make it not affecting IPv4. | |
1283 | The rest of code is protocol independent, | |
1284 | and I do not like idea to uglify IPv4. | |
1285 | ||
1286 | Actually, all the idea behind IPV6_PKTOPTIONS | |
1287 | looks not very well thought. For now we latch | |
1288 | options, received in the last packet, enqueued | |
1289 | by tcp. Feel free to propose better solution. | |
1290 | --ANK (980728) | |
1291 | */ | |
1292 | if (np->rxopt.all) | |
1293 | opt_skb = skb_clone(skb, sk_gfp_atomic(sk, GFP_ATOMIC)); | |
1294 | ||
1295 | if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */ | |
1296 | struct dst_entry *dst = sk->sk_rx_dst; | |
1297 | ||
1298 | sock_rps_save_rxhash(sk, skb); | |
1299 | sk_mark_napi_id(sk, skb); | |
1300 | if (dst) { | |
1301 | if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif || | |
1302 | dst->ops->check(dst, np->rx_dst_cookie) == NULL) { | |
1303 | dst_release(dst); | |
1304 | sk->sk_rx_dst = NULL; | |
1305 | } | |
1306 | } | |
1307 | ||
1308 | tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len); | |
1309 | if (opt_skb) | |
1310 | goto ipv6_pktoptions; | |
1311 | return 0; | |
1312 | } | |
1313 | ||
1314 | if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb)) | |
1315 | goto csum_err; | |
1316 | ||
1317 | if (sk->sk_state == TCP_LISTEN) { | |
1318 | struct sock *nsk = tcp_v6_hnd_req(sk, skb); | |
1319 | if (!nsk) | |
1320 | goto discard; | |
1321 | ||
1322 | /* | |
1323 | * Queue it on the new socket if the new socket is active, | |
1324 | * otherwise we just shortcircuit this and continue with | |
1325 | * the new socket.. | |
1326 | */ | |
1327 | if (nsk != sk) { | |
1328 | sock_rps_save_rxhash(nsk, skb); | |
1329 | sk_mark_napi_id(sk, skb); | |
1330 | if (tcp_child_process(sk, nsk, skb)) | |
1331 | goto reset; | |
1332 | if (opt_skb) | |
1333 | __kfree_skb(opt_skb); | |
1334 | return 0; | |
1335 | } | |
1336 | } else | |
1337 | sock_rps_save_rxhash(sk, skb); | |
1338 | ||
1339 | if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len)) | |
1340 | goto reset; | |
1341 | if (opt_skb) | |
1342 | goto ipv6_pktoptions; | |
1343 | return 0; | |
1344 | ||
1345 | reset: | |
1346 | tcp_v6_send_reset(sk, skb); | |
1347 | discard: | |
1348 | if (opt_skb) | |
1349 | __kfree_skb(opt_skb); | |
1350 | kfree_skb(skb); | |
1351 | return 0; | |
1352 | csum_err: | |
1353 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS); | |
1354 | TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS); | |
1355 | goto discard; | |
1356 | ||
1357 | ||
1358 | ipv6_pktoptions: | |
1359 | /* Do you ask, what is it? | |
1360 | ||
1361 | 1. skb was enqueued by tcp. | |
1362 | 2. skb is added to tail of read queue, rather than out of order. | |
1363 | 3. socket is not in passive state. | |
1364 | 4. Finally, it really contains options, which user wants to receive. | |
1365 | */ | |
1366 | tp = tcp_sk(sk); | |
1367 | if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt && | |
1368 | !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) { | |
1369 | if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo) | |
1370 | np->mcast_oif = tcp_v6_iif(opt_skb); | |
1371 | if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) | |
1372 | np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit; | |
1373 | if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass) | |
1374 | np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb)); | |
1375 | if (np->repflow) | |
1376 | np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb)); | |
1377 | if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->header.h6)) { | |
1378 | skb_set_owner_r(opt_skb, sk); | |
1379 | opt_skb = xchg(&np->pktoptions, opt_skb); | |
1380 | } else { | |
1381 | __kfree_skb(opt_skb); | |
1382 | opt_skb = xchg(&np->pktoptions, NULL); | |
1383 | } | |
1384 | } | |
1385 | ||
1386 | kfree_skb(opt_skb); | |
1387 | return 0; | |
1388 | } | |
1389 | ||
1390 | static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr, | |
1391 | const struct tcphdr *th) | |
1392 | { | |
1393 | /* This is tricky: we move IP6CB at its correct location into | |
1394 | * TCP_SKB_CB(). It must be done after xfrm6_policy_check(), because | |
1395 | * _decode_session6() uses IP6CB(). | |
1396 | * barrier() makes sure compiler won't play aliasing games. | |
1397 | */ | |
1398 | memmove(&TCP_SKB_CB(skb)->header.h6, IP6CB(skb), | |
1399 | sizeof(struct inet6_skb_parm)); | |
1400 | barrier(); | |
1401 | ||
1402 | TCP_SKB_CB(skb)->seq = ntohl(th->seq); | |
1403 | TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin + | |
1404 | skb->len - th->doff*4); | |
1405 | TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq); | |
1406 | TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th); | |
1407 | TCP_SKB_CB(skb)->tcp_tw_isn = 0; | |
1408 | TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr); | |
1409 | TCP_SKB_CB(skb)->sacked = 0; | |
1410 | } | |
1411 | ||
1412 | static int tcp_v6_rcv(struct sk_buff *skb) | |
1413 | { | |
1414 | const struct tcphdr *th; | |
1415 | const struct ipv6hdr *hdr; | |
1416 | struct sock *sk; | |
1417 | int ret; | |
1418 | struct net *net = dev_net(skb->dev); | |
1419 | ||
1420 | if (skb->pkt_type != PACKET_HOST) | |
1421 | goto discard_it; | |
1422 | ||
1423 | /* | |
1424 | * Count it even if it's bad. | |
1425 | */ | |
1426 | TCP_INC_STATS_BH(net, TCP_MIB_INSEGS); | |
1427 | ||
1428 | if (!pskb_may_pull(skb, sizeof(struct tcphdr))) | |
1429 | goto discard_it; | |
1430 | ||
1431 | th = tcp_hdr(skb); | |
1432 | ||
1433 | if (th->doff < sizeof(struct tcphdr)/4) | |
1434 | goto bad_packet; | |
1435 | if (!pskb_may_pull(skb, th->doff*4)) | |
1436 | goto discard_it; | |
1437 | ||
1438 | if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo)) | |
1439 | goto csum_error; | |
1440 | ||
1441 | th = tcp_hdr(skb); | |
1442 | hdr = ipv6_hdr(skb); | |
1443 | ||
1444 | sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest, | |
1445 | inet6_iif(skb)); | |
1446 | if (!sk) | |
1447 | goto no_tcp_socket; | |
1448 | ||
1449 | process: | |
1450 | if (sk->sk_state == TCP_TIME_WAIT) | |
1451 | goto do_time_wait; | |
1452 | ||
1453 | if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) { | |
1454 | NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP); | |
1455 | goto discard_and_relse; | |
1456 | } | |
1457 | ||
1458 | if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) | |
1459 | goto discard_and_relse; | |
1460 | ||
1461 | tcp_v6_fill_cb(skb, hdr, th); | |
1462 | ||
1463 | #ifdef CONFIG_TCP_MD5SIG | |
1464 | if (tcp_v6_inbound_md5_hash(sk, skb)) | |
1465 | goto discard_and_relse; | |
1466 | #endif | |
1467 | ||
1468 | if (sk_filter(sk, skb)) | |
1469 | goto discard_and_relse; | |
1470 | ||
1471 | sk_incoming_cpu_update(sk); | |
1472 | skb->dev = NULL; | |
1473 | ||
1474 | bh_lock_sock_nested(sk); | |
1475 | ret = 0; | |
1476 | if (!sock_owned_by_user(sk)) { | |
1477 | if (!tcp_prequeue(sk, skb)) | |
1478 | ret = tcp_v6_do_rcv(sk, skb); | |
1479 | } else if (unlikely(sk_add_backlog(sk, skb, | |
1480 | sk->sk_rcvbuf + sk->sk_sndbuf))) { | |
1481 | bh_unlock_sock(sk); | |
1482 | NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP); | |
1483 | goto discard_and_relse; | |
1484 | } | |
1485 | bh_unlock_sock(sk); | |
1486 | ||
1487 | sock_put(sk); | |
1488 | return ret ? -1 : 0; | |
1489 | ||
1490 | no_tcp_socket: | |
1491 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) | |
1492 | goto discard_it; | |
1493 | ||
1494 | tcp_v6_fill_cb(skb, hdr, th); | |
1495 | ||
1496 | if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) { | |
1497 | csum_error: | |
1498 | TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS); | |
1499 | bad_packet: | |
1500 | TCP_INC_STATS_BH(net, TCP_MIB_INERRS); | |
1501 | } else { | |
1502 | tcp_v6_send_reset(NULL, skb); | |
1503 | } | |
1504 | ||
1505 | discard_it: | |
1506 | kfree_skb(skb); | |
1507 | return 0; | |
1508 | ||
1509 | discard_and_relse: | |
1510 | sock_put(sk); | |
1511 | goto discard_it; | |
1512 | ||
1513 | do_time_wait: | |
1514 | if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { | |
1515 | inet_twsk_put(inet_twsk(sk)); | |
1516 | goto discard_it; | |
1517 | } | |
1518 | ||
1519 | tcp_v6_fill_cb(skb, hdr, th); | |
1520 | ||
1521 | if (skb->len < (th->doff<<2)) { | |
1522 | inet_twsk_put(inet_twsk(sk)); | |
1523 | goto bad_packet; | |
1524 | } | |
1525 | if (tcp_checksum_complete(skb)) { | |
1526 | inet_twsk_put(inet_twsk(sk)); | |
1527 | goto csum_error; | |
1528 | } | |
1529 | ||
1530 | switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) { | |
1531 | case TCP_TW_SYN: | |
1532 | { | |
1533 | struct sock *sk2; | |
1534 | ||
1535 | sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo, | |
1536 | &ipv6_hdr(skb)->saddr, th->source, | |
1537 | &ipv6_hdr(skb)->daddr, | |
1538 | ntohs(th->dest), tcp_v6_iif(skb)); | |
1539 | if (sk2 != NULL) { | |
1540 | struct inet_timewait_sock *tw = inet_twsk(sk); | |
1541 | inet_twsk_deschedule(tw, &tcp_death_row); | |
1542 | inet_twsk_put(tw); | |
1543 | sk = sk2; | |
1544 | goto process; | |
1545 | } | |
1546 | /* Fall through to ACK */ | |
1547 | } | |
1548 | case TCP_TW_ACK: | |
1549 | tcp_v6_timewait_ack(sk, skb); | |
1550 | break; | |
1551 | case TCP_TW_RST: | |
1552 | goto no_tcp_socket; | |
1553 | case TCP_TW_SUCCESS: | |
1554 | ; | |
1555 | } | |
1556 | goto discard_it; | |
1557 | } | |
1558 | ||
1559 | static void tcp_v6_early_demux(struct sk_buff *skb) | |
1560 | { | |
1561 | const struct ipv6hdr *hdr; | |
1562 | const struct tcphdr *th; | |
1563 | struct sock *sk; | |
1564 | ||
1565 | if (skb->pkt_type != PACKET_HOST) | |
1566 | return; | |
1567 | ||
1568 | if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr))) | |
1569 | return; | |
1570 | ||
1571 | hdr = ipv6_hdr(skb); | |
1572 | th = tcp_hdr(skb); | |
1573 | ||
1574 | if (th->doff < sizeof(struct tcphdr) / 4) | |
1575 | return; | |
1576 | ||
1577 | /* Note : We use inet6_iif() here, not tcp_v6_iif() */ | |
1578 | sk = __inet6_lookup_established(dev_net(skb->dev), &tcp_hashinfo, | |
1579 | &hdr->saddr, th->source, | |
1580 | &hdr->daddr, ntohs(th->dest), | |
1581 | inet6_iif(skb)); | |
1582 | if (sk) { | |
1583 | skb->sk = sk; | |
1584 | skb->destructor = sock_edemux; | |
1585 | if (sk->sk_state != TCP_TIME_WAIT) { | |
1586 | struct dst_entry *dst = sk->sk_rx_dst; | |
1587 | ||
1588 | if (dst) | |
1589 | dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie); | |
1590 | if (dst && | |
1591 | inet_sk(sk)->rx_dst_ifindex == skb->skb_iif) | |
1592 | skb_dst_set_noref(skb, dst); | |
1593 | } | |
1594 | } | |
1595 | } | |
1596 | ||
1597 | static struct timewait_sock_ops tcp6_timewait_sock_ops = { | |
1598 | .twsk_obj_size = sizeof(struct tcp6_timewait_sock), | |
1599 | .twsk_unique = tcp_twsk_unique, | |
1600 | .twsk_destructor = tcp_twsk_destructor, | |
1601 | }; | |
1602 | ||
1603 | static const struct inet_connection_sock_af_ops ipv6_specific = { | |
1604 | .queue_xmit = inet6_csk_xmit, | |
1605 | .send_check = tcp_v6_send_check, | |
1606 | .rebuild_header = inet6_sk_rebuild_header, | |
1607 | .sk_rx_dst_set = inet6_sk_rx_dst_set, | |
1608 | .conn_request = tcp_v6_conn_request, | |
1609 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
1610 | .net_header_len = sizeof(struct ipv6hdr), | |
1611 | .net_frag_header_len = sizeof(struct frag_hdr), | |
1612 | .setsockopt = ipv6_setsockopt, | |
1613 | .getsockopt = ipv6_getsockopt, | |
1614 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1615 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
1616 | .bind_conflict = inet6_csk_bind_conflict, | |
1617 | #ifdef CONFIG_COMPAT | |
1618 | .compat_setsockopt = compat_ipv6_setsockopt, | |
1619 | .compat_getsockopt = compat_ipv6_getsockopt, | |
1620 | #endif | |
1621 | .mtu_reduced = tcp_v6_mtu_reduced, | |
1622 | }; | |
1623 | ||
1624 | #ifdef CONFIG_TCP_MD5SIG | |
1625 | static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = { | |
1626 | .md5_lookup = tcp_v6_md5_lookup, | |
1627 | .calc_md5_hash = tcp_v6_md5_hash_skb, | |
1628 | .md5_parse = tcp_v6_parse_md5_keys, | |
1629 | }; | |
1630 | #endif | |
1631 | ||
1632 | /* | |
1633 | * TCP over IPv4 via INET6 API | |
1634 | */ | |
1635 | static const struct inet_connection_sock_af_ops ipv6_mapped = { | |
1636 | .queue_xmit = ip_queue_xmit, | |
1637 | .send_check = tcp_v4_send_check, | |
1638 | .rebuild_header = inet_sk_rebuild_header, | |
1639 | .sk_rx_dst_set = inet_sk_rx_dst_set, | |
1640 | .conn_request = tcp_v6_conn_request, | |
1641 | .syn_recv_sock = tcp_v6_syn_recv_sock, | |
1642 | .net_header_len = sizeof(struct iphdr), | |
1643 | .setsockopt = ipv6_setsockopt, | |
1644 | .getsockopt = ipv6_getsockopt, | |
1645 | .addr2sockaddr = inet6_csk_addr2sockaddr, | |
1646 | .sockaddr_len = sizeof(struct sockaddr_in6), | |
1647 | .bind_conflict = inet6_csk_bind_conflict, | |
1648 | #ifdef CONFIG_COMPAT | |
1649 | .compat_setsockopt = compat_ipv6_setsockopt, | |
1650 | .compat_getsockopt = compat_ipv6_getsockopt, | |
1651 | #endif | |
1652 | .mtu_reduced = tcp_v4_mtu_reduced, | |
1653 | }; | |
1654 | ||
1655 | #ifdef CONFIG_TCP_MD5SIG | |
1656 | static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = { | |
1657 | .md5_lookup = tcp_v4_md5_lookup, | |
1658 | .calc_md5_hash = tcp_v4_md5_hash_skb, | |
1659 | .md5_parse = tcp_v6_parse_md5_keys, | |
1660 | }; | |
1661 | #endif | |
1662 | ||
1663 | /* NOTE: A lot of things set to zero explicitly by call to | |
1664 | * sk_alloc() so need not be done here. | |
1665 | */ | |
1666 | static int tcp_v6_init_sock(struct sock *sk) | |
1667 | { | |
1668 | struct inet_connection_sock *icsk = inet_csk(sk); | |
1669 | ||
1670 | tcp_init_sock(sk); | |
1671 | ||
1672 | icsk->icsk_af_ops = &ipv6_specific; | |
1673 | ||
1674 | #ifdef CONFIG_TCP_MD5SIG | |
1675 | tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific; | |
1676 | #endif | |
1677 | ||
1678 | return 0; | |
1679 | } | |
1680 | ||
1681 | static void tcp_v6_destroy_sock(struct sock *sk) | |
1682 | { | |
1683 | tcp_v4_destroy_sock(sk); | |
1684 | inet6_destroy_sock(sk); | |
1685 | } | |
1686 | ||
1687 | #ifdef CONFIG_PROC_FS | |
1688 | /* Proc filesystem TCPv6 sock list dumping. */ | |
1689 | static void get_openreq6(struct seq_file *seq, | |
1690 | const struct sock *sk, struct request_sock *req, int i, kuid_t uid) | |
1691 | { | |
1692 | int ttd = req->expires - jiffies; | |
1693 | const struct in6_addr *src = &inet_rsk(req)->ir_v6_loc_addr; | |
1694 | const struct in6_addr *dest = &inet_rsk(req)->ir_v6_rmt_addr; | |
1695 | ||
1696 | if (ttd < 0) | |
1697 | ttd = 0; | |
1698 | ||
1699 | seq_printf(seq, | |
1700 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
1701 | "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n", | |
1702 | i, | |
1703 | src->s6_addr32[0], src->s6_addr32[1], | |
1704 | src->s6_addr32[2], src->s6_addr32[3], | |
1705 | inet_rsk(req)->ir_num, | |
1706 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1707 | dest->s6_addr32[2], dest->s6_addr32[3], | |
1708 | ntohs(inet_rsk(req)->ir_rmt_port), | |
1709 | TCP_SYN_RECV, | |
1710 | 0, 0, /* could print option size, but that is af dependent. */ | |
1711 | 1, /* timers active (only the expire timer) */ | |
1712 | jiffies_to_clock_t(ttd), | |
1713 | req->num_timeout, | |
1714 | from_kuid_munged(seq_user_ns(seq), uid), | |
1715 | 0, /* non standard timer */ | |
1716 | 0, /* open_requests have no inode */ | |
1717 | 0, req); | |
1718 | } | |
1719 | ||
1720 | static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i) | |
1721 | { | |
1722 | const struct in6_addr *dest, *src; | |
1723 | __u16 destp, srcp; | |
1724 | int timer_active; | |
1725 | unsigned long timer_expires; | |
1726 | const struct inet_sock *inet = inet_sk(sp); | |
1727 | const struct tcp_sock *tp = tcp_sk(sp); | |
1728 | const struct inet_connection_sock *icsk = inet_csk(sp); | |
1729 | struct fastopen_queue *fastopenq = icsk->icsk_accept_queue.fastopenq; | |
1730 | ||
1731 | dest = &sp->sk_v6_daddr; | |
1732 | src = &sp->sk_v6_rcv_saddr; | |
1733 | destp = ntohs(inet->inet_dport); | |
1734 | srcp = ntohs(inet->inet_sport); | |
1735 | ||
1736 | if (icsk->icsk_pending == ICSK_TIME_RETRANS) { | |
1737 | timer_active = 1; | |
1738 | timer_expires = icsk->icsk_timeout; | |
1739 | } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) { | |
1740 | timer_active = 4; | |
1741 | timer_expires = icsk->icsk_timeout; | |
1742 | } else if (timer_pending(&sp->sk_timer)) { | |
1743 | timer_active = 2; | |
1744 | timer_expires = sp->sk_timer.expires; | |
1745 | } else { | |
1746 | timer_active = 0; | |
1747 | timer_expires = jiffies; | |
1748 | } | |
1749 | ||
1750 | seq_printf(seq, | |
1751 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
1752 | "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n", | |
1753 | i, | |
1754 | src->s6_addr32[0], src->s6_addr32[1], | |
1755 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
1756 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1757 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
1758 | sp->sk_state, | |
1759 | tp->write_seq-tp->snd_una, | |
1760 | (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq), | |
1761 | timer_active, | |
1762 | jiffies_delta_to_clock_t(timer_expires - jiffies), | |
1763 | icsk->icsk_retransmits, | |
1764 | from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)), | |
1765 | icsk->icsk_probes_out, | |
1766 | sock_i_ino(sp), | |
1767 | atomic_read(&sp->sk_refcnt), sp, | |
1768 | jiffies_to_clock_t(icsk->icsk_rto), | |
1769 | jiffies_to_clock_t(icsk->icsk_ack.ato), | |
1770 | (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong, | |
1771 | tp->snd_cwnd, | |
1772 | sp->sk_state == TCP_LISTEN ? | |
1773 | (fastopenq ? fastopenq->max_qlen : 0) : | |
1774 | (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh) | |
1775 | ); | |
1776 | } | |
1777 | ||
1778 | static void get_timewait6_sock(struct seq_file *seq, | |
1779 | struct inet_timewait_sock *tw, int i) | |
1780 | { | |
1781 | const struct in6_addr *dest, *src; | |
1782 | __u16 destp, srcp; | |
1783 | s32 delta = tw->tw_ttd - inet_tw_time_stamp(); | |
1784 | ||
1785 | dest = &tw->tw_v6_daddr; | |
1786 | src = &tw->tw_v6_rcv_saddr; | |
1787 | destp = ntohs(tw->tw_dport); | |
1788 | srcp = ntohs(tw->tw_sport); | |
1789 | ||
1790 | seq_printf(seq, | |
1791 | "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X " | |
1792 | "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n", | |
1793 | i, | |
1794 | src->s6_addr32[0], src->s6_addr32[1], | |
1795 | src->s6_addr32[2], src->s6_addr32[3], srcp, | |
1796 | dest->s6_addr32[0], dest->s6_addr32[1], | |
1797 | dest->s6_addr32[2], dest->s6_addr32[3], destp, | |
1798 | tw->tw_substate, 0, 0, | |
1799 | 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0, | |
1800 | atomic_read(&tw->tw_refcnt), tw); | |
1801 | } | |
1802 | ||
1803 | static int tcp6_seq_show(struct seq_file *seq, void *v) | |
1804 | { | |
1805 | struct tcp_iter_state *st; | |
1806 | struct sock *sk = v; | |
1807 | ||
1808 | if (v == SEQ_START_TOKEN) { | |
1809 | seq_puts(seq, | |
1810 | " sl " | |
1811 | "local_address " | |
1812 | "remote_address " | |
1813 | "st tx_queue rx_queue tr tm->when retrnsmt" | |
1814 | " uid timeout inode\n"); | |
1815 | goto out; | |
1816 | } | |
1817 | st = seq->private; | |
1818 | ||
1819 | switch (st->state) { | |
1820 | case TCP_SEQ_STATE_LISTENING: | |
1821 | case TCP_SEQ_STATE_ESTABLISHED: | |
1822 | if (sk->sk_state == TCP_TIME_WAIT) | |
1823 | get_timewait6_sock(seq, v, st->num); | |
1824 | else | |
1825 | get_tcp6_sock(seq, v, st->num); | |
1826 | break; | |
1827 | case TCP_SEQ_STATE_OPENREQ: | |
1828 | get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid); | |
1829 | break; | |
1830 | } | |
1831 | out: | |
1832 | return 0; | |
1833 | } | |
1834 | ||
1835 | static const struct file_operations tcp6_afinfo_seq_fops = { | |
1836 | .owner = THIS_MODULE, | |
1837 | .open = tcp_seq_open, | |
1838 | .read = seq_read, | |
1839 | .llseek = seq_lseek, | |
1840 | .release = seq_release_net | |
1841 | }; | |
1842 | ||
1843 | static struct tcp_seq_afinfo tcp6_seq_afinfo = { | |
1844 | .name = "tcp6", | |
1845 | .family = AF_INET6, | |
1846 | .seq_fops = &tcp6_afinfo_seq_fops, | |
1847 | .seq_ops = { | |
1848 | .show = tcp6_seq_show, | |
1849 | }, | |
1850 | }; | |
1851 | ||
1852 | int __net_init tcp6_proc_init(struct net *net) | |
1853 | { | |
1854 | return tcp_proc_register(net, &tcp6_seq_afinfo); | |
1855 | } | |
1856 | ||
1857 | void tcp6_proc_exit(struct net *net) | |
1858 | { | |
1859 | tcp_proc_unregister(net, &tcp6_seq_afinfo); | |
1860 | } | |
1861 | #endif | |
1862 | ||
1863 | static void tcp_v6_clear_sk(struct sock *sk, int size) | |
1864 | { | |
1865 | struct inet_sock *inet = inet_sk(sk); | |
1866 | ||
1867 | /* we do not want to clear pinet6 field, because of RCU lookups */ | |
1868 | sk_prot_clear_nulls(sk, offsetof(struct inet_sock, pinet6)); | |
1869 | ||
1870 | size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6); | |
1871 | memset(&inet->pinet6 + 1, 0, size); | |
1872 | } | |
1873 | ||
1874 | struct proto tcpv6_prot = { | |
1875 | .name = "TCPv6", | |
1876 | .owner = THIS_MODULE, | |
1877 | .close = tcp_close, | |
1878 | .connect = tcp_v6_connect, | |
1879 | .disconnect = tcp_disconnect, | |
1880 | .accept = inet_csk_accept, | |
1881 | .ioctl = tcp_ioctl, | |
1882 | .init = tcp_v6_init_sock, | |
1883 | .destroy = tcp_v6_destroy_sock, | |
1884 | .shutdown = tcp_shutdown, | |
1885 | .setsockopt = tcp_setsockopt, | |
1886 | .getsockopt = tcp_getsockopt, | |
1887 | .recvmsg = tcp_recvmsg, | |
1888 | .sendmsg = tcp_sendmsg, | |
1889 | .sendpage = tcp_sendpage, | |
1890 | .backlog_rcv = tcp_v6_do_rcv, | |
1891 | .release_cb = tcp_release_cb, | |
1892 | .hash = tcp_v6_hash, | |
1893 | .unhash = inet_unhash, | |
1894 | .get_port = inet_csk_get_port, | |
1895 | .enter_memory_pressure = tcp_enter_memory_pressure, | |
1896 | .stream_memory_free = tcp_stream_memory_free, | |
1897 | .sockets_allocated = &tcp_sockets_allocated, | |
1898 | .memory_allocated = &tcp_memory_allocated, | |
1899 | .memory_pressure = &tcp_memory_pressure, | |
1900 | .orphan_count = &tcp_orphan_count, | |
1901 | .sysctl_mem = sysctl_tcp_mem, | |
1902 | .sysctl_wmem = sysctl_tcp_wmem, | |
1903 | .sysctl_rmem = sysctl_tcp_rmem, | |
1904 | .max_header = MAX_TCP_HEADER, | |
1905 | .obj_size = sizeof(struct tcp6_sock), | |
1906 | .slab_flags = SLAB_DESTROY_BY_RCU, | |
1907 | .twsk_prot = &tcp6_timewait_sock_ops, | |
1908 | .rsk_prot = &tcp6_request_sock_ops, | |
1909 | .h.hashinfo = &tcp_hashinfo, | |
1910 | .no_autobind = true, | |
1911 | #ifdef CONFIG_COMPAT | |
1912 | .compat_setsockopt = compat_tcp_setsockopt, | |
1913 | .compat_getsockopt = compat_tcp_getsockopt, | |
1914 | #endif | |
1915 | #ifdef CONFIG_MEMCG_KMEM | |
1916 | .proto_cgroup = tcp_proto_cgroup, | |
1917 | #endif | |
1918 | .clear_sk = tcp_v6_clear_sk, | |
1919 | }; | |
1920 | ||
1921 | static const struct inet6_protocol tcpv6_protocol = { | |
1922 | .early_demux = tcp_v6_early_demux, | |
1923 | .handler = tcp_v6_rcv, | |
1924 | .err_handler = tcp_v6_err, | |
1925 | .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL, | |
1926 | }; | |
1927 | ||
1928 | static struct inet_protosw tcpv6_protosw = { | |
1929 | .type = SOCK_STREAM, | |
1930 | .protocol = IPPROTO_TCP, | |
1931 | .prot = &tcpv6_prot, | |
1932 | .ops = &inet6_stream_ops, | |
1933 | .flags = INET_PROTOSW_PERMANENT | | |
1934 | INET_PROTOSW_ICSK, | |
1935 | }; | |
1936 | ||
1937 | static int __net_init tcpv6_net_init(struct net *net) | |
1938 | { | |
1939 | return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6, | |
1940 | SOCK_RAW, IPPROTO_TCP, net); | |
1941 | } | |
1942 | ||
1943 | static void __net_exit tcpv6_net_exit(struct net *net) | |
1944 | { | |
1945 | inet_ctl_sock_destroy(net->ipv6.tcp_sk); | |
1946 | } | |
1947 | ||
1948 | static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list) | |
1949 | { | |
1950 | inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6); | |
1951 | } | |
1952 | ||
1953 | static struct pernet_operations tcpv6_net_ops = { | |
1954 | .init = tcpv6_net_init, | |
1955 | .exit = tcpv6_net_exit, | |
1956 | .exit_batch = tcpv6_net_exit_batch, | |
1957 | }; | |
1958 | ||
1959 | int __init tcpv6_init(void) | |
1960 | { | |
1961 | int ret; | |
1962 | ||
1963 | ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
1964 | if (ret) | |
1965 | goto out; | |
1966 | ||
1967 | /* register inet6 protocol */ | |
1968 | ret = inet6_register_protosw(&tcpv6_protosw); | |
1969 | if (ret) | |
1970 | goto out_tcpv6_protocol; | |
1971 | ||
1972 | ret = register_pernet_subsys(&tcpv6_net_ops); | |
1973 | if (ret) | |
1974 | goto out_tcpv6_protosw; | |
1975 | out: | |
1976 | return ret; | |
1977 | ||
1978 | out_tcpv6_protosw: | |
1979 | inet6_unregister_protosw(&tcpv6_protosw); | |
1980 | out_tcpv6_protocol: | |
1981 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
1982 | goto out; | |
1983 | } | |
1984 | ||
1985 | void tcpv6_exit(void) | |
1986 | { | |
1987 | unregister_pernet_subsys(&tcpv6_net_ops); | |
1988 | inet6_unregister_protosw(&tcpv6_protosw); | |
1989 | inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP); | |
1990 | } |