Merge tag 'dm-3.17-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device...
[deliverable/linux.git] / net / ipv6 / tcp_ipv6.c
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/netdma.h>
63 #include <net/inet_common.h>
64 #include <net/secure_seq.h>
65 #include <net/tcp_memcontrol.h>
66 #include <net/busy_poll.h>
67
68 #include <linux/proc_fs.h>
69 #include <linux/seq_file.h>
70
71 #include <linux/crypto.h>
72 #include <linux/scatterlist.h>
73
74 static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb);
75 static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
76 struct request_sock *req);
77
78 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
79
80 static const struct inet_connection_sock_af_ops ipv6_mapped;
81 static const struct inet_connection_sock_af_ops ipv6_specific;
82 #ifdef CONFIG_TCP_MD5SIG
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;
85 #else
86 static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
87 const struct in6_addr *addr)
88 {
89 return NULL;
90 }
91 #endif
92
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
105 static void tcp_v6_hash(struct sock *sk)
106 {
107 if (sk->sk_state != TCP_CLOSE) {
108 if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
109 tcp_prot.hash(sk);
110 return;
111 }
112 local_bh_disable();
113 __inet6_hash(sk, NULL);
114 local_bh_enable();
115 }
116 }
117
118 static __u32 tcp_v6_init_sequence(const struct sk_buff *skb)
119 {
120 return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
121 ipv6_hdr(skb)->saddr.s6_addr32,
122 tcp_hdr(skb)->dest,
123 tcp_hdr(skb)->source);
124 }
125
126 static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
127 int addr_len)
128 {
129 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
130 struct inet_sock *inet = inet_sk(sk);
131 struct inet_connection_sock *icsk = inet_csk(sk);
132 struct ipv6_pinfo *np = inet6_sk(sk);
133 struct tcp_sock *tp = tcp_sk(sk);
134 struct in6_addr *saddr = NULL, *final_p, final;
135 struct rt6_info *rt;
136 struct flowi6 fl6;
137 struct dst_entry *dst;
138 int addr_type;
139 int err;
140
141 if (addr_len < SIN6_LEN_RFC2133)
142 return -EINVAL;
143
144 if (usin->sin6_family != AF_INET6)
145 return -EAFNOSUPPORT;
146
147 memset(&fl6, 0, sizeof(fl6));
148
149 if (np->sndflow) {
150 fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
151 IP6_ECN_flow_init(fl6.flowlabel);
152 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
153 struct ip6_flowlabel *flowlabel;
154 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
155 if (flowlabel == NULL)
156 return -EINVAL;
157 fl6_sock_release(flowlabel);
158 }
159 }
160
161 /*
162 * connect() to INADDR_ANY means loopback (BSD'ism).
163 */
164
165 if (ipv6_addr_any(&usin->sin6_addr))
166 usin->sin6_addr.s6_addr[15] = 0x1;
167
168 addr_type = ipv6_addr_type(&usin->sin6_addr);
169
170 if (addr_type & IPV6_ADDR_MULTICAST)
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 &&
192 !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) {
193 tp->rx_opt.ts_recent = 0;
194 tp->rx_opt.ts_recent_stamp = 0;
195 tp->write_seq = 0;
196 }
197
198 sk->sk_v6_daddr = usin->sin6_addr;
199 np->flow_label = fl6.flowlabel;
200
201 ip6_set_txhash(sk);
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 if (!tp->write_seq && likely(!tp->repair))
299 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
300 sk->sk_v6_daddr.s6_addr32,
301 inet->inet_sport,
302 inet->inet_dport);
303
304 err = tcp_connect(sk);
305 if (err)
306 goto late_failure;
307
308 return 0;
309
310 late_failure:
311 tcp_set_state(sk, TCP_CLOSE);
312 __sk_dst_reset(sk);
313 failure:
314 inet->inet_dport = 0;
315 sk->sk_route_caps = 0;
316 return err;
317 }
318
319 static void tcp_v6_mtu_reduced(struct sock *sk)
320 {
321 struct dst_entry *dst;
322
323 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
324 return;
325
326 dst = inet6_csk_update_pmtu(sk, tcp_sk(sk)->mtu_info);
327 if (!dst)
328 return;
329
330 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
331 tcp_sync_mss(sk, dst_mtu(dst));
332 tcp_simple_retransmit(sk);
333 }
334 }
335
336 static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
337 u8 type, u8 code, int offset, __be32 info)
338 {
339 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
340 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
341 struct ipv6_pinfo *np;
342 struct sock *sk;
343 int err;
344 struct tcp_sock *tp;
345 struct request_sock *fastopen;
346 __u32 seq, snd_una;
347 struct net *net = dev_net(skb->dev);
348
349 sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr,
350 th->dest, &hdr->saddr, th->source, skb->dev->ifindex);
351
352 if (sk == NULL) {
353 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
354 ICMP6_MIB_INERRORS);
355 return;
356 }
357
358 if (sk->sk_state == TCP_TIME_WAIT) {
359 inet_twsk_put(inet_twsk(sk));
360 return;
361 }
362
363 bh_lock_sock(sk);
364 if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG)
365 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
366
367 if (sk->sk_state == TCP_CLOSE)
368 goto out;
369
370 if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) {
371 NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
372 goto out;
373 }
374
375 tp = tcp_sk(sk);
376 seq = ntohl(th->seq);
377 /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
378 fastopen = tp->fastopen_rsk;
379 snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
380 if (sk->sk_state != TCP_LISTEN &&
381 !between(seq, snd_una, tp->snd_nxt)) {
382 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
383 goto out;
384 }
385
386 np = inet6_sk(sk);
387
388 if (type == NDISC_REDIRECT) {
389 struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
390
391 if (dst)
392 dst->ops->redirect(dst, sk, skb);
393 goto out;
394 }
395
396 if (type == ICMPV6_PKT_TOOBIG) {
397 /* We are not interested in TCP_LISTEN and open_requests
398 * (SYN-ACKs send out by Linux are always <576bytes so
399 * they should go through unfragmented).
400 */
401 if (sk->sk_state == TCP_LISTEN)
402 goto out;
403
404 if (!ip6_sk_accept_pmtu(sk))
405 goto out;
406
407 tp->mtu_info = ntohl(info);
408 if (!sock_owned_by_user(sk))
409 tcp_v6_mtu_reduced(sk);
410 else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED,
411 &tp->tsq_flags))
412 sock_hold(sk);
413 goto out;
414 }
415
416 icmpv6_err_convert(type, code, &err);
417
418 /* Might be for an request_sock */
419 switch (sk->sk_state) {
420 struct request_sock *req, **prev;
421 case TCP_LISTEN:
422 if (sock_owned_by_user(sk))
423 goto out;
424
425 req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
426 &hdr->saddr, inet6_iif(skb));
427 if (!req)
428 goto out;
429
430 /* ICMPs are not backlogged, hence we cannot get
431 * an established socket here.
432 */
433 WARN_ON(req->sk != NULL);
434
435 if (seq != tcp_rsk(req)->snt_isn) {
436 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
437 goto out;
438 }
439
440 inet_csk_reqsk_queue_drop(sk, req, prev);
441 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
442 goto out;
443
444 case TCP_SYN_SENT:
445 case TCP_SYN_RECV:
446 /* Only in fast or simultaneous open. If a fast open socket is
447 * is already accepted it is treated as a connected one below.
448 */
449 if (fastopen && fastopen->sk == NULL)
450 break;
451
452 if (!sock_owned_by_user(sk)) {
453 sk->sk_err = err;
454 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
455
456 tcp_done(sk);
457 } else
458 sk->sk_err_soft = err;
459 goto out;
460 }
461
462 if (!sock_owned_by_user(sk) && np->recverr) {
463 sk->sk_err = err;
464 sk->sk_error_report(sk);
465 } else
466 sk->sk_err_soft = err;
467
468 out:
469 bh_unlock_sock(sk);
470 sock_put(sk);
471 }
472
473
474 static int tcp_v6_send_synack(struct sock *sk, struct dst_entry *dst,
475 struct flowi *fl,
476 struct request_sock *req,
477 u16 queue_mapping,
478 struct tcp_fastopen_cookie *foc)
479 {
480 struct inet_request_sock *ireq = inet_rsk(req);
481 struct ipv6_pinfo *np = inet6_sk(sk);
482 struct flowi6 *fl6 = &fl->u.ip6;
483 struct sk_buff *skb;
484 int err = -ENOMEM;
485
486 /* First, grab a route. */
487 if (!dst && (dst = inet6_csk_route_req(sk, fl6, req)) == NULL)
488 goto done;
489
490 skb = tcp_make_synack(sk, dst, req, foc);
491
492 if (skb) {
493 __tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr,
494 &ireq->ir_v6_rmt_addr);
495
496 fl6->daddr = ireq->ir_v6_rmt_addr;
497 if (np->repflow && (ireq->pktopts != NULL))
498 fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts));
499
500 skb_set_queue_mapping(skb, queue_mapping);
501 err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass);
502 err = net_xmit_eval(err);
503 }
504
505 done:
506 return err;
507 }
508
509
510 static void tcp_v6_reqsk_destructor(struct request_sock *req)
511 {
512 kfree_skb(inet_rsk(req)->pktopts);
513 }
514
515 #ifdef CONFIG_TCP_MD5SIG
516 static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
517 const struct in6_addr *addr)
518 {
519 return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
520 }
521
522 static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
523 struct sock *addr_sk)
524 {
525 return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr);
526 }
527
528 static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
529 struct request_sock *req)
530 {
531 return tcp_v6_md5_do_lookup(sk, &inet_rsk(req)->ir_v6_rmt_addr);
532 }
533
534 static int tcp_v6_parse_md5_keys(struct sock *sk, char __user *optval,
535 int optlen)
536 {
537 struct tcp_md5sig cmd;
538 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
539
540 if (optlen < sizeof(cmd))
541 return -EINVAL;
542
543 if (copy_from_user(&cmd, optval, sizeof(cmd)))
544 return -EFAULT;
545
546 if (sin6->sin6_family != AF_INET6)
547 return -EINVAL;
548
549 if (!cmd.tcpm_keylen) {
550 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
551 return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
552 AF_INET);
553 return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
554 AF_INET6);
555 }
556
557 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
558 return -EINVAL;
559
560 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
561 return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
562 AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
563
564 return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
565 AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
566 }
567
568 static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
569 const struct in6_addr *daddr,
570 const struct in6_addr *saddr, int nbytes)
571 {
572 struct tcp6_pseudohdr *bp;
573 struct scatterlist sg;
574
575 bp = &hp->md5_blk.ip6;
576 /* 1. TCP pseudo-header (RFC2460) */
577 bp->saddr = *saddr;
578 bp->daddr = *daddr;
579 bp->protocol = cpu_to_be32(IPPROTO_TCP);
580 bp->len = cpu_to_be32(nbytes);
581
582 sg_init_one(&sg, bp, sizeof(*bp));
583 return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
584 }
585
586 static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
587 const struct in6_addr *daddr, struct in6_addr *saddr,
588 const struct tcphdr *th)
589 {
590 struct tcp_md5sig_pool *hp;
591 struct hash_desc *desc;
592
593 hp = tcp_get_md5sig_pool();
594 if (!hp)
595 goto clear_hash_noput;
596 desc = &hp->md5_desc;
597
598 if (crypto_hash_init(desc))
599 goto clear_hash;
600 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2))
601 goto clear_hash;
602 if (tcp_md5_hash_header(hp, th))
603 goto clear_hash;
604 if (tcp_md5_hash_key(hp, key))
605 goto clear_hash;
606 if (crypto_hash_final(desc, md5_hash))
607 goto clear_hash;
608
609 tcp_put_md5sig_pool();
610 return 0;
611
612 clear_hash:
613 tcp_put_md5sig_pool();
614 clear_hash_noput:
615 memset(md5_hash, 0, 16);
616 return 1;
617 }
618
619 static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
620 const struct sock *sk,
621 const struct request_sock *req,
622 const struct sk_buff *skb)
623 {
624 const struct in6_addr *saddr, *daddr;
625 struct tcp_md5sig_pool *hp;
626 struct hash_desc *desc;
627 const struct tcphdr *th = tcp_hdr(skb);
628
629 if (sk) {
630 saddr = &inet6_sk(sk)->saddr;
631 daddr = &sk->sk_v6_daddr;
632 } else if (req) {
633 saddr = &inet_rsk(req)->ir_v6_loc_addr;
634 daddr = &inet_rsk(req)->ir_v6_rmt_addr;
635 } else {
636 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
637 saddr = &ip6h->saddr;
638 daddr = &ip6h->daddr;
639 }
640
641 hp = tcp_get_md5sig_pool();
642 if (!hp)
643 goto clear_hash_noput;
644 desc = &hp->md5_desc;
645
646 if (crypto_hash_init(desc))
647 goto clear_hash;
648
649 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len))
650 goto clear_hash;
651 if (tcp_md5_hash_header(hp, th))
652 goto clear_hash;
653 if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
654 goto clear_hash;
655 if (tcp_md5_hash_key(hp, key))
656 goto clear_hash;
657 if (crypto_hash_final(desc, md5_hash))
658 goto clear_hash;
659
660 tcp_put_md5sig_pool();
661 return 0;
662
663 clear_hash:
664 tcp_put_md5sig_pool();
665 clear_hash_noput:
666 memset(md5_hash, 0, 16);
667 return 1;
668 }
669
670 static int __tcp_v6_inbound_md5_hash(struct sock *sk,
671 const struct sk_buff *skb)
672 {
673 const __u8 *hash_location = NULL;
674 struct tcp_md5sig_key *hash_expected;
675 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
676 const struct tcphdr *th = tcp_hdr(skb);
677 int genhash;
678 u8 newhash[16];
679
680 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
681 hash_location = tcp_parse_md5sig_option(th);
682
683 /* We've parsed the options - do we have a hash? */
684 if (!hash_expected && !hash_location)
685 return 0;
686
687 if (hash_expected && !hash_location) {
688 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
689 return 1;
690 }
691
692 if (!hash_expected && hash_location) {
693 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
694 return 1;
695 }
696
697 /* check the signature */
698 genhash = tcp_v6_md5_hash_skb(newhash,
699 hash_expected,
700 NULL, NULL, skb);
701
702 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
703 net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n",
704 genhash ? "failed" : "mismatch",
705 &ip6h->saddr, ntohs(th->source),
706 &ip6h->daddr, ntohs(th->dest));
707 return 1;
708 }
709 return 0;
710 }
711
712 static int tcp_v6_inbound_md5_hash(struct sock *sk, const struct sk_buff *skb)
713 {
714 int ret;
715
716 rcu_read_lock();
717 ret = __tcp_v6_inbound_md5_hash(sk, skb);
718 rcu_read_unlock();
719
720 return ret;
721 }
722
723 #endif
724
725 static void tcp_v6_init_req(struct request_sock *req, struct sock *sk,
726 struct sk_buff *skb)
727 {
728 struct inet_request_sock *ireq = inet_rsk(req);
729 struct ipv6_pinfo *np = inet6_sk(sk);
730
731 ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
732 ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
733
734 ireq->ir_iif = sk->sk_bound_dev_if;
735
736 /* So that link locals have meaning */
737 if (!sk->sk_bound_dev_if &&
738 ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
739 ireq->ir_iif = inet6_iif(skb);
740
741 if (!TCP_SKB_CB(skb)->when &&
742 (ipv6_opt_accepted(sk, skb) || np->rxopt.bits.rxinfo ||
743 np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim ||
744 np->rxopt.bits.rxohlim || np->repflow)) {
745 atomic_inc(&skb->users);
746 ireq->pktopts = skb;
747 }
748 }
749
750 static struct dst_entry *tcp_v6_route_req(struct sock *sk, struct flowi *fl,
751 const struct request_sock *req,
752 bool *strict)
753 {
754 if (strict)
755 *strict = true;
756 return inet6_csk_route_req(sk, &fl->u.ip6, req);
757 }
758
759 struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
760 .family = AF_INET6,
761 .obj_size = sizeof(struct tcp6_request_sock),
762 .rtx_syn_ack = tcp_rtx_synack,
763 .send_ack = tcp_v6_reqsk_send_ack,
764 .destructor = tcp_v6_reqsk_destructor,
765 .send_reset = tcp_v6_send_reset,
766 .syn_ack_timeout = tcp_syn_ack_timeout,
767 };
768
769 static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
770 .mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) -
771 sizeof(struct ipv6hdr),
772 #ifdef CONFIG_TCP_MD5SIG
773 .md5_lookup = tcp_v6_reqsk_md5_lookup,
774 .calc_md5_hash = tcp_v6_md5_hash_skb,
775 #endif
776 .init_req = tcp_v6_init_req,
777 #ifdef CONFIG_SYN_COOKIES
778 .cookie_init_seq = cookie_v6_init_sequence,
779 #endif
780 .route_req = tcp_v6_route_req,
781 .init_seq = tcp_v6_init_sequence,
782 .send_synack = tcp_v6_send_synack,
783 .queue_hash_add = inet6_csk_reqsk_queue_hash_add,
784 };
785
786 static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win,
787 u32 tsval, u32 tsecr, int oif,
788 struct tcp_md5sig_key *key, int rst, u8 tclass,
789 u32 label)
790 {
791 const struct tcphdr *th = tcp_hdr(skb);
792 struct tcphdr *t1;
793 struct sk_buff *buff;
794 struct flowi6 fl6;
795 struct net *net = dev_net(skb_dst(skb)->dev);
796 struct sock *ctl_sk = net->ipv6.tcp_sk;
797 unsigned int tot_len = sizeof(struct tcphdr);
798 struct dst_entry *dst;
799 __be32 *topt;
800
801 if (tsecr)
802 tot_len += TCPOLEN_TSTAMP_ALIGNED;
803 #ifdef CONFIG_TCP_MD5SIG
804 if (key)
805 tot_len += TCPOLEN_MD5SIG_ALIGNED;
806 #endif
807
808 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
809 GFP_ATOMIC);
810 if (buff == NULL)
811 return;
812
813 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
814
815 t1 = (struct tcphdr *) skb_push(buff, tot_len);
816 skb_reset_transport_header(buff);
817
818 /* Swap the send and the receive. */
819 memset(t1, 0, sizeof(*t1));
820 t1->dest = th->source;
821 t1->source = th->dest;
822 t1->doff = tot_len / 4;
823 t1->seq = htonl(seq);
824 t1->ack_seq = htonl(ack);
825 t1->ack = !rst || !th->ack;
826 t1->rst = rst;
827 t1->window = htons(win);
828
829 topt = (__be32 *)(t1 + 1);
830
831 if (tsecr) {
832 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
833 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
834 *topt++ = htonl(tsval);
835 *topt++ = htonl(tsecr);
836 }
837
838 #ifdef CONFIG_TCP_MD5SIG
839 if (key) {
840 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
841 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
842 tcp_v6_md5_hash_hdr((__u8 *)topt, key,
843 &ipv6_hdr(skb)->saddr,
844 &ipv6_hdr(skb)->daddr, t1);
845 }
846 #endif
847
848 memset(&fl6, 0, sizeof(fl6));
849 fl6.daddr = ipv6_hdr(skb)->saddr;
850 fl6.saddr = ipv6_hdr(skb)->daddr;
851 fl6.flowlabel = label;
852
853 buff->ip_summed = CHECKSUM_PARTIAL;
854 buff->csum = 0;
855
856 __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
857
858 fl6.flowi6_proto = IPPROTO_TCP;
859 if (rt6_need_strict(&fl6.daddr) && !oif)
860 fl6.flowi6_oif = inet6_iif(skb);
861 else
862 fl6.flowi6_oif = oif;
863 fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark);
864 fl6.fl6_dport = t1->dest;
865 fl6.fl6_sport = t1->source;
866 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
867
868 /* Pass a socket to ip6_dst_lookup either it is for RST
869 * Underlying function will use this to retrieve the network
870 * namespace
871 */
872 dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
873 if (!IS_ERR(dst)) {
874 skb_dst_set(buff, dst);
875 ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass);
876 TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
877 if (rst)
878 TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS);
879 return;
880 }
881
882 kfree_skb(buff);
883 }
884
885 static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
886 {
887 const struct tcphdr *th = tcp_hdr(skb);
888 u32 seq = 0, ack_seq = 0;
889 struct tcp_md5sig_key *key = NULL;
890 #ifdef CONFIG_TCP_MD5SIG
891 const __u8 *hash_location = NULL;
892 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
893 unsigned char newhash[16];
894 int genhash;
895 struct sock *sk1 = NULL;
896 #endif
897 int oif;
898
899 if (th->rst)
900 return;
901
902 if (!ipv6_unicast_destination(skb))
903 return;
904
905 #ifdef CONFIG_TCP_MD5SIG
906 hash_location = tcp_parse_md5sig_option(th);
907 if (!sk && hash_location) {
908 /*
909 * active side is lost. Try to find listening socket through
910 * source port, and then find md5 key through listening socket.
911 * we are not loose security here:
912 * Incoming packet is checked with md5 hash with finding key,
913 * no RST generated if md5 hash doesn't match.
914 */
915 sk1 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev),
916 &tcp_hashinfo, &ipv6h->saddr,
917 th->source, &ipv6h->daddr,
918 ntohs(th->source), inet6_iif(skb));
919 if (!sk1)
920 return;
921
922 rcu_read_lock();
923 key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr);
924 if (!key)
925 goto release_sk1;
926
927 genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, NULL, skb);
928 if (genhash || memcmp(hash_location, newhash, 16) != 0)
929 goto release_sk1;
930 } else {
931 key = sk ? tcp_v6_md5_do_lookup(sk, &ipv6h->saddr) : NULL;
932 }
933 #endif
934
935 if (th->ack)
936 seq = ntohl(th->ack_seq);
937 else
938 ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len -
939 (th->doff << 2);
940
941 oif = sk ? sk->sk_bound_dev_if : 0;
942 tcp_v6_send_response(skb, seq, ack_seq, 0, 0, 0, oif, key, 1, 0, 0);
943
944 #ifdef CONFIG_TCP_MD5SIG
945 release_sk1:
946 if (sk1) {
947 rcu_read_unlock();
948 sock_put(sk1);
949 }
950 #endif
951 }
952
953 static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack,
954 u32 win, u32 tsval, u32 tsecr, int oif,
955 struct tcp_md5sig_key *key, u8 tclass,
956 u32 label)
957 {
958 tcp_v6_send_response(skb, seq, ack, win, tsval, tsecr, oif, key, 0, tclass,
959 label);
960 }
961
962 static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
963 {
964 struct inet_timewait_sock *tw = inet_twsk(sk);
965 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
966
967 tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
968 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
969 tcp_time_stamp + tcptw->tw_ts_offset,
970 tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw),
971 tw->tw_tclass, (tw->tw_flowlabel << 12));
972
973 inet_twsk_put(tw);
974 }
975
976 static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
977 struct request_sock *req)
978 {
979 /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
980 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
981 */
982 tcp_v6_send_ack(skb, (sk->sk_state == TCP_LISTEN) ?
983 tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt,
984 tcp_rsk(req)->rcv_nxt,
985 req->rcv_wnd, tcp_time_stamp, req->ts_recent, sk->sk_bound_dev_if,
986 tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr),
987 0, 0);
988 }
989
990
991 static struct sock *tcp_v6_hnd_req(struct sock *sk, struct sk_buff *skb)
992 {
993 struct request_sock *req, **prev;
994 const struct tcphdr *th = tcp_hdr(skb);
995 struct sock *nsk;
996
997 /* Find possible connection requests. */
998 req = inet6_csk_search_req(sk, &prev, th->source,
999 &ipv6_hdr(skb)->saddr,
1000 &ipv6_hdr(skb)->daddr, inet6_iif(skb));
1001 if (req)
1002 return tcp_check_req(sk, skb, req, prev, false);
1003
1004 nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
1005 &ipv6_hdr(skb)->saddr, th->source,
1006 &ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb));
1007
1008 if (nsk) {
1009 if (nsk->sk_state != TCP_TIME_WAIT) {
1010 bh_lock_sock(nsk);
1011 return nsk;
1012 }
1013 inet_twsk_put(inet_twsk(nsk));
1014 return NULL;
1015 }
1016
1017 #ifdef CONFIG_SYN_COOKIES
1018 if (!th->syn)
1019 sk = cookie_v6_check(sk, skb);
1020 #endif
1021 return sk;
1022 }
1023
1024 static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
1025 {
1026 if (skb->protocol == htons(ETH_P_IP))
1027 return tcp_v4_conn_request(sk, skb);
1028
1029 if (!ipv6_unicast_destination(skb))
1030 goto drop;
1031
1032 return tcp_conn_request(&tcp6_request_sock_ops,
1033 &tcp_request_sock_ipv6_ops, sk, skb);
1034
1035 drop:
1036 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
1037 return 0; /* don't send reset */
1038 }
1039
1040 static struct sock *tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
1041 struct request_sock *req,
1042 struct dst_entry *dst)
1043 {
1044 struct inet_request_sock *ireq;
1045 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
1046 struct tcp6_sock *newtcp6sk;
1047 struct inet_sock *newinet;
1048 struct tcp_sock *newtp;
1049 struct sock *newsk;
1050 #ifdef CONFIG_TCP_MD5SIG
1051 struct tcp_md5sig_key *key;
1052 #endif
1053 struct flowi6 fl6;
1054
1055 if (skb->protocol == htons(ETH_P_IP)) {
1056 /*
1057 * v6 mapped
1058 */
1059
1060 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1061
1062 if (newsk == NULL)
1063 return NULL;
1064
1065 newtcp6sk = (struct tcp6_sock *)newsk;
1066 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1067
1068 newinet = inet_sk(newsk);
1069 newnp = inet6_sk(newsk);
1070 newtp = tcp_sk(newsk);
1071
1072 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1073
1074 ipv6_addr_set_v4mapped(newinet->inet_daddr, &newsk->sk_v6_daddr);
1075
1076 ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr);
1077
1078 newsk->sk_v6_rcv_saddr = newnp->saddr;
1079
1080 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1081 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1082 #ifdef CONFIG_TCP_MD5SIG
1083 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1084 #endif
1085
1086 newnp->ipv6_ac_list = NULL;
1087 newnp->ipv6_fl_list = NULL;
1088 newnp->pktoptions = NULL;
1089 newnp->opt = NULL;
1090 newnp->mcast_oif = inet6_iif(skb);
1091 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1092 newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1093 if (np->repflow)
1094 newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
1095
1096 /*
1097 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1098 * here, tcp_create_openreq_child now does this for us, see the comment in
1099 * that function for the gory details. -acme
1100 */
1101
1102 /* It is tricky place. Until this moment IPv4 tcp
1103 worked with IPv6 icsk.icsk_af_ops.
1104 Sync it now.
1105 */
1106 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1107
1108 return newsk;
1109 }
1110
1111 ireq = inet_rsk(req);
1112
1113 if (sk_acceptq_is_full(sk))
1114 goto out_overflow;
1115
1116 if (!dst) {
1117 dst = inet6_csk_route_req(sk, &fl6, req);
1118 if (!dst)
1119 goto out;
1120 }
1121
1122 newsk = tcp_create_openreq_child(sk, req, skb);
1123 if (newsk == NULL)
1124 goto out_nonewsk;
1125
1126 /*
1127 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1128 * count here, tcp_create_openreq_child now does this for us, see the
1129 * comment in that function for the gory details. -acme
1130 */
1131
1132 newsk->sk_gso_type = SKB_GSO_TCPV6;
1133 __ip6_dst_store(newsk, dst, NULL, NULL);
1134 inet6_sk_rx_dst_set(newsk, skb);
1135
1136 newtcp6sk = (struct tcp6_sock *)newsk;
1137 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1138
1139 newtp = tcp_sk(newsk);
1140 newinet = inet_sk(newsk);
1141 newnp = inet6_sk(newsk);
1142
1143 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1144
1145 newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr;
1146 newnp->saddr = ireq->ir_v6_loc_addr;
1147 newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr;
1148 newsk->sk_bound_dev_if = ireq->ir_iif;
1149
1150 ip6_set_txhash(newsk);
1151
1152 /* Now IPv6 options...
1153
1154 First: no IPv4 options.
1155 */
1156 newinet->inet_opt = NULL;
1157 newnp->ipv6_ac_list = NULL;
1158 newnp->ipv6_fl_list = NULL;
1159
1160 /* Clone RX bits */
1161 newnp->rxopt.all = np->rxopt.all;
1162
1163 /* Clone pktoptions received with SYN */
1164 newnp->pktoptions = NULL;
1165 if (ireq->pktopts != NULL) {
1166 newnp->pktoptions = skb_clone(ireq->pktopts,
1167 sk_gfp_atomic(sk, GFP_ATOMIC));
1168 consume_skb(ireq->pktopts);
1169 ireq->pktopts = NULL;
1170 if (newnp->pktoptions)
1171 skb_set_owner_r(newnp->pktoptions, newsk);
1172 }
1173 newnp->opt = NULL;
1174 newnp->mcast_oif = inet6_iif(skb);
1175 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1176 newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1177 if (np->repflow)
1178 newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
1179
1180 /* Clone native IPv6 options from listening socket (if any)
1181
1182 Yes, keeping reference count would be much more clever,
1183 but we make one more one thing there: reattach optmem
1184 to newsk.
1185 */
1186 if (np->opt)
1187 newnp->opt = ipv6_dup_options(newsk, np->opt);
1188
1189 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1190 if (newnp->opt)
1191 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1192 newnp->opt->opt_flen);
1193
1194 tcp_sync_mss(newsk, dst_mtu(dst));
1195 newtp->advmss = dst_metric_advmss(dst);
1196 if (tcp_sk(sk)->rx_opt.user_mss &&
1197 tcp_sk(sk)->rx_opt.user_mss < newtp->advmss)
1198 newtp->advmss = tcp_sk(sk)->rx_opt.user_mss;
1199
1200 tcp_initialize_rcv_mss(newsk);
1201
1202 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
1203 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
1204
1205 #ifdef CONFIG_TCP_MD5SIG
1206 /* Copy over the MD5 key from the original socket */
1207 key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr);
1208 if (key != NULL) {
1209 /* We're using one, so create a matching key
1210 * on the newsk structure. If we fail to get
1211 * memory, then we end up not copying the key
1212 * across. Shucks.
1213 */
1214 tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
1215 AF_INET6, key->key, key->keylen,
1216 sk_gfp_atomic(sk, GFP_ATOMIC));
1217 }
1218 #endif
1219
1220 if (__inet_inherit_port(sk, newsk) < 0) {
1221 inet_csk_prepare_forced_close(newsk);
1222 tcp_done(newsk);
1223 goto out;
1224 }
1225 __inet6_hash(newsk, NULL);
1226
1227 return newsk;
1228
1229 out_overflow:
1230 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1231 out_nonewsk:
1232 dst_release(dst);
1233 out:
1234 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
1235 return NULL;
1236 }
1237
1238 /* The socket must have it's spinlock held when we get
1239 * here.
1240 *
1241 * We have a potential double-lock case here, so even when
1242 * doing backlog processing we use the BH locking scheme.
1243 * This is because we cannot sleep with the original spinlock
1244 * held.
1245 */
1246 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1247 {
1248 struct ipv6_pinfo *np = inet6_sk(sk);
1249 struct tcp_sock *tp;
1250 struct sk_buff *opt_skb = NULL;
1251
1252 /* Imagine: socket is IPv6. IPv4 packet arrives,
1253 goes to IPv4 receive handler and backlogged.
1254 From backlog it always goes here. Kerboom...
1255 Fortunately, tcp_rcv_established and rcv_established
1256 handle them correctly, but it is not case with
1257 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1258 */
1259
1260 if (skb->protocol == htons(ETH_P_IP))
1261 return tcp_v4_do_rcv(sk, skb);
1262
1263 if (sk_filter(sk, skb))
1264 goto discard;
1265
1266 /*
1267 * socket locking is here for SMP purposes as backlog rcv
1268 * is currently called with bh processing disabled.
1269 */
1270
1271 /* Do Stevens' IPV6_PKTOPTIONS.
1272
1273 Yes, guys, it is the only place in our code, where we
1274 may make it not affecting IPv4.
1275 The rest of code is protocol independent,
1276 and I do not like idea to uglify IPv4.
1277
1278 Actually, all the idea behind IPV6_PKTOPTIONS
1279 looks not very well thought. For now we latch
1280 options, received in the last packet, enqueued
1281 by tcp. Feel free to propose better solution.
1282 --ANK (980728)
1283 */
1284 if (np->rxopt.all)
1285 opt_skb = skb_clone(skb, sk_gfp_atomic(sk, GFP_ATOMIC));
1286
1287 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1288 struct dst_entry *dst = sk->sk_rx_dst;
1289
1290 sock_rps_save_rxhash(sk, skb);
1291 if (dst) {
1292 if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif ||
1293 dst->ops->check(dst, np->rx_dst_cookie) == NULL) {
1294 dst_release(dst);
1295 sk->sk_rx_dst = NULL;
1296 }
1297 }
1298
1299 tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
1300 if (opt_skb)
1301 goto ipv6_pktoptions;
1302 return 0;
1303 }
1304
1305 if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
1306 goto csum_err;
1307
1308 if (sk->sk_state == TCP_LISTEN) {
1309 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1310 if (!nsk)
1311 goto discard;
1312
1313 /*
1314 * Queue it on the new socket if the new socket is active,
1315 * otherwise we just shortcircuit this and continue with
1316 * the new socket..
1317 */
1318 if (nsk != sk) {
1319 sock_rps_save_rxhash(nsk, skb);
1320 if (tcp_child_process(sk, nsk, skb))
1321 goto reset;
1322 if (opt_skb)
1323 __kfree_skb(opt_skb);
1324 return 0;
1325 }
1326 } else
1327 sock_rps_save_rxhash(sk, skb);
1328
1329 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
1330 goto reset;
1331 if (opt_skb)
1332 goto ipv6_pktoptions;
1333 return 0;
1334
1335 reset:
1336 tcp_v6_send_reset(sk, skb);
1337 discard:
1338 if (opt_skb)
1339 __kfree_skb(opt_skb);
1340 kfree_skb(skb);
1341 return 0;
1342 csum_err:
1343 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_CSUMERRORS);
1344 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
1345 goto discard;
1346
1347
1348 ipv6_pktoptions:
1349 /* Do you ask, what is it?
1350
1351 1. skb was enqueued by tcp.
1352 2. skb is added to tail of read queue, rather than out of order.
1353 3. socket is not in passive state.
1354 4. Finally, it really contains options, which user wants to receive.
1355 */
1356 tp = tcp_sk(sk);
1357 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1358 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
1359 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
1360 np->mcast_oif = inet6_iif(opt_skb);
1361 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1362 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1363 if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
1364 np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
1365 if (np->repflow)
1366 np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
1367 if (ipv6_opt_accepted(sk, opt_skb)) {
1368 skb_set_owner_r(opt_skb, sk);
1369 opt_skb = xchg(&np->pktoptions, opt_skb);
1370 } else {
1371 __kfree_skb(opt_skb);
1372 opt_skb = xchg(&np->pktoptions, NULL);
1373 }
1374 }
1375
1376 kfree_skb(opt_skb);
1377 return 0;
1378 }
1379
1380 static int tcp_v6_rcv(struct sk_buff *skb)
1381 {
1382 const struct tcphdr *th;
1383 const struct ipv6hdr *hdr;
1384 struct sock *sk;
1385 int ret;
1386 struct net *net = dev_net(skb->dev);
1387
1388 if (skb->pkt_type != PACKET_HOST)
1389 goto discard_it;
1390
1391 /*
1392 * Count it even if it's bad.
1393 */
1394 TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
1395
1396 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1397 goto discard_it;
1398
1399 th = tcp_hdr(skb);
1400
1401 if (th->doff < sizeof(struct tcphdr)/4)
1402 goto bad_packet;
1403 if (!pskb_may_pull(skb, th->doff*4))
1404 goto discard_it;
1405
1406 if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo))
1407 goto csum_error;
1408
1409 th = tcp_hdr(skb);
1410 hdr = ipv6_hdr(skb);
1411 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1412 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1413 skb->len - th->doff*4);
1414 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1415 TCP_SKB_CB(skb)->when = 0;
1416 TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
1417 TCP_SKB_CB(skb)->sacked = 0;
1418
1419 sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest);
1420 if (!sk)
1421 goto no_tcp_socket;
1422
1423 process:
1424 if (sk->sk_state == TCP_TIME_WAIT)
1425 goto do_time_wait;
1426
1427 if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) {
1428 NET_INC_STATS_BH(net, LINUX_MIB_TCPMINTTLDROP);
1429 goto discard_and_relse;
1430 }
1431
1432 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1433 goto discard_and_relse;
1434
1435 #ifdef CONFIG_TCP_MD5SIG
1436 if (tcp_v6_inbound_md5_hash(sk, skb))
1437 goto discard_and_relse;
1438 #endif
1439
1440 if (sk_filter(sk, skb))
1441 goto discard_and_relse;
1442
1443 sk_mark_napi_id(sk, skb);
1444 skb->dev = NULL;
1445
1446 bh_lock_sock_nested(sk);
1447 ret = 0;
1448 if (!sock_owned_by_user(sk)) {
1449 #ifdef CONFIG_NET_DMA
1450 struct tcp_sock *tp = tcp_sk(sk);
1451 if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
1452 tp->ucopy.dma_chan = net_dma_find_channel();
1453 if (tp->ucopy.dma_chan)
1454 ret = tcp_v6_do_rcv(sk, skb);
1455 else
1456 #endif
1457 {
1458 if (!tcp_prequeue(sk, skb))
1459 ret = tcp_v6_do_rcv(sk, skb);
1460 }
1461 } else if (unlikely(sk_add_backlog(sk, skb,
1462 sk->sk_rcvbuf + sk->sk_sndbuf))) {
1463 bh_unlock_sock(sk);
1464 NET_INC_STATS_BH(net, LINUX_MIB_TCPBACKLOGDROP);
1465 goto discard_and_relse;
1466 }
1467 bh_unlock_sock(sk);
1468
1469 sock_put(sk);
1470 return ret ? -1 : 0;
1471
1472 no_tcp_socket:
1473 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1474 goto discard_it;
1475
1476 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1477 csum_error:
1478 TCP_INC_STATS_BH(net, TCP_MIB_CSUMERRORS);
1479 bad_packet:
1480 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1481 } else {
1482 tcp_v6_send_reset(NULL, skb);
1483 }
1484
1485 discard_it:
1486 kfree_skb(skb);
1487 return 0;
1488
1489 discard_and_relse:
1490 sock_put(sk);
1491 goto discard_it;
1492
1493 do_time_wait:
1494 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1495 inet_twsk_put(inet_twsk(sk));
1496 goto discard_it;
1497 }
1498
1499 if (skb->len < (th->doff<<2)) {
1500 inet_twsk_put(inet_twsk(sk));
1501 goto bad_packet;
1502 }
1503 if (tcp_checksum_complete(skb)) {
1504 inet_twsk_put(inet_twsk(sk));
1505 goto csum_error;
1506 }
1507
1508 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1509 case TCP_TW_SYN:
1510 {
1511 struct sock *sk2;
1512
1513 sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1514 &ipv6_hdr(skb)->saddr, th->source,
1515 &ipv6_hdr(skb)->daddr,
1516 ntohs(th->dest), inet6_iif(skb));
1517 if (sk2 != NULL) {
1518 struct inet_timewait_sock *tw = inet_twsk(sk);
1519 inet_twsk_deschedule(tw, &tcp_death_row);
1520 inet_twsk_put(tw);
1521 sk = sk2;
1522 goto process;
1523 }
1524 /* Fall through to ACK */
1525 }
1526 case TCP_TW_ACK:
1527 tcp_v6_timewait_ack(sk, skb);
1528 break;
1529 case TCP_TW_RST:
1530 goto no_tcp_socket;
1531 case TCP_TW_SUCCESS:
1532 ;
1533 }
1534 goto discard_it;
1535 }
1536
1537 static void tcp_v6_early_demux(struct sk_buff *skb)
1538 {
1539 const struct ipv6hdr *hdr;
1540 const struct tcphdr *th;
1541 struct sock *sk;
1542
1543 if (skb->pkt_type != PACKET_HOST)
1544 return;
1545
1546 if (!pskb_may_pull(skb, skb_transport_offset(skb) + sizeof(struct tcphdr)))
1547 return;
1548
1549 hdr = ipv6_hdr(skb);
1550 th = tcp_hdr(skb);
1551
1552 if (th->doff < sizeof(struct tcphdr) / 4)
1553 return;
1554
1555 sk = __inet6_lookup_established(dev_net(skb->dev), &tcp_hashinfo,
1556 &hdr->saddr, th->source,
1557 &hdr->daddr, ntohs(th->dest),
1558 inet6_iif(skb));
1559 if (sk) {
1560 skb->sk = sk;
1561 skb->destructor = sock_edemux;
1562 if (sk->sk_state != TCP_TIME_WAIT) {
1563 struct dst_entry *dst = sk->sk_rx_dst;
1564
1565 if (dst)
1566 dst = dst_check(dst, inet6_sk(sk)->rx_dst_cookie);
1567 if (dst &&
1568 inet_sk(sk)->rx_dst_ifindex == skb->skb_iif)
1569 skb_dst_set_noref(skb, dst);
1570 }
1571 }
1572 }
1573
1574 static struct timewait_sock_ops tcp6_timewait_sock_ops = {
1575 .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
1576 .twsk_unique = tcp_twsk_unique,
1577 .twsk_destructor = tcp_twsk_destructor,
1578 };
1579
1580 static const struct inet_connection_sock_af_ops ipv6_specific = {
1581 .queue_xmit = inet6_csk_xmit,
1582 .send_check = tcp_v6_send_check,
1583 .rebuild_header = inet6_sk_rebuild_header,
1584 .sk_rx_dst_set = inet6_sk_rx_dst_set,
1585 .conn_request = tcp_v6_conn_request,
1586 .syn_recv_sock = tcp_v6_syn_recv_sock,
1587 .net_header_len = sizeof(struct ipv6hdr),
1588 .net_frag_header_len = sizeof(struct frag_hdr),
1589 .setsockopt = ipv6_setsockopt,
1590 .getsockopt = ipv6_getsockopt,
1591 .addr2sockaddr = inet6_csk_addr2sockaddr,
1592 .sockaddr_len = sizeof(struct sockaddr_in6),
1593 .bind_conflict = inet6_csk_bind_conflict,
1594 #ifdef CONFIG_COMPAT
1595 .compat_setsockopt = compat_ipv6_setsockopt,
1596 .compat_getsockopt = compat_ipv6_getsockopt,
1597 #endif
1598 };
1599
1600 #ifdef CONFIG_TCP_MD5SIG
1601 static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1602 .md5_lookup = tcp_v6_md5_lookup,
1603 .calc_md5_hash = tcp_v6_md5_hash_skb,
1604 .md5_parse = tcp_v6_parse_md5_keys,
1605 };
1606 #endif
1607
1608 /*
1609 * TCP over IPv4 via INET6 API
1610 */
1611 static const struct inet_connection_sock_af_ops ipv6_mapped = {
1612 .queue_xmit = ip_queue_xmit,
1613 .send_check = tcp_v4_send_check,
1614 .rebuild_header = inet_sk_rebuild_header,
1615 .sk_rx_dst_set = inet_sk_rx_dst_set,
1616 .conn_request = tcp_v6_conn_request,
1617 .syn_recv_sock = tcp_v6_syn_recv_sock,
1618 .net_header_len = sizeof(struct iphdr),
1619 .setsockopt = ipv6_setsockopt,
1620 .getsockopt = ipv6_getsockopt,
1621 .addr2sockaddr = inet6_csk_addr2sockaddr,
1622 .sockaddr_len = sizeof(struct sockaddr_in6),
1623 .bind_conflict = inet6_csk_bind_conflict,
1624 #ifdef CONFIG_COMPAT
1625 .compat_setsockopt = compat_ipv6_setsockopt,
1626 .compat_getsockopt = compat_ipv6_getsockopt,
1627 #endif
1628 };
1629
1630 #ifdef CONFIG_TCP_MD5SIG
1631 static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1632 .md5_lookup = tcp_v4_md5_lookup,
1633 .calc_md5_hash = tcp_v4_md5_hash_skb,
1634 .md5_parse = tcp_v6_parse_md5_keys,
1635 };
1636 #endif
1637
1638 /* NOTE: A lot of things set to zero explicitly by call to
1639 * sk_alloc() so need not be done here.
1640 */
1641 static int tcp_v6_init_sock(struct sock *sk)
1642 {
1643 struct inet_connection_sock *icsk = inet_csk(sk);
1644
1645 tcp_init_sock(sk);
1646
1647 icsk->icsk_af_ops = &ipv6_specific;
1648
1649 #ifdef CONFIG_TCP_MD5SIG
1650 tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
1651 #endif
1652
1653 return 0;
1654 }
1655
1656 static void tcp_v6_destroy_sock(struct sock *sk)
1657 {
1658 tcp_v4_destroy_sock(sk);
1659 inet6_destroy_sock(sk);
1660 }
1661
1662 #ifdef CONFIG_PROC_FS
1663 /* Proc filesystem TCPv6 sock list dumping. */
1664 static void get_openreq6(struct seq_file *seq,
1665 const struct sock *sk, struct request_sock *req, int i, kuid_t uid)
1666 {
1667 int ttd = req->expires - jiffies;
1668 const struct in6_addr *src = &inet_rsk(req)->ir_v6_loc_addr;
1669 const struct in6_addr *dest = &inet_rsk(req)->ir_v6_rmt_addr;
1670
1671 if (ttd < 0)
1672 ttd = 0;
1673
1674 seq_printf(seq,
1675 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1676 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
1677 i,
1678 src->s6_addr32[0], src->s6_addr32[1],
1679 src->s6_addr32[2], src->s6_addr32[3],
1680 inet_rsk(req)->ir_num,
1681 dest->s6_addr32[0], dest->s6_addr32[1],
1682 dest->s6_addr32[2], dest->s6_addr32[3],
1683 ntohs(inet_rsk(req)->ir_rmt_port),
1684 TCP_SYN_RECV,
1685 0, 0, /* could print option size, but that is af dependent. */
1686 1, /* timers active (only the expire timer) */
1687 jiffies_to_clock_t(ttd),
1688 req->num_timeout,
1689 from_kuid_munged(seq_user_ns(seq), uid),
1690 0, /* non standard timer */
1691 0, /* open_requests have no inode */
1692 0, req);
1693 }
1694
1695 static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1696 {
1697 const struct in6_addr *dest, *src;
1698 __u16 destp, srcp;
1699 int timer_active;
1700 unsigned long timer_expires;
1701 const struct inet_sock *inet = inet_sk(sp);
1702 const struct tcp_sock *tp = tcp_sk(sp);
1703 const struct inet_connection_sock *icsk = inet_csk(sp);
1704 struct fastopen_queue *fastopenq = icsk->icsk_accept_queue.fastopenq;
1705
1706 dest = &sp->sk_v6_daddr;
1707 src = &sp->sk_v6_rcv_saddr;
1708 destp = ntohs(inet->inet_dport);
1709 srcp = ntohs(inet->inet_sport);
1710
1711 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
1712 timer_active = 1;
1713 timer_expires = icsk->icsk_timeout;
1714 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
1715 timer_active = 4;
1716 timer_expires = icsk->icsk_timeout;
1717 } else if (timer_pending(&sp->sk_timer)) {
1718 timer_active = 2;
1719 timer_expires = sp->sk_timer.expires;
1720 } else {
1721 timer_active = 0;
1722 timer_expires = jiffies;
1723 }
1724
1725 seq_printf(seq,
1726 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1727 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
1728 i,
1729 src->s6_addr32[0], src->s6_addr32[1],
1730 src->s6_addr32[2], src->s6_addr32[3], srcp,
1731 dest->s6_addr32[0], dest->s6_addr32[1],
1732 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1733 sp->sk_state,
1734 tp->write_seq-tp->snd_una,
1735 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
1736 timer_active,
1737 jiffies_delta_to_clock_t(timer_expires - jiffies),
1738 icsk->icsk_retransmits,
1739 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1740 icsk->icsk_probes_out,
1741 sock_i_ino(sp),
1742 atomic_read(&sp->sk_refcnt), sp,
1743 jiffies_to_clock_t(icsk->icsk_rto),
1744 jiffies_to_clock_t(icsk->icsk_ack.ato),
1745 (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
1746 tp->snd_cwnd,
1747 sp->sk_state == TCP_LISTEN ?
1748 (fastopenq ? fastopenq->max_qlen : 0) :
1749 (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)
1750 );
1751 }
1752
1753 static void get_timewait6_sock(struct seq_file *seq,
1754 struct inet_timewait_sock *tw, int i)
1755 {
1756 const struct in6_addr *dest, *src;
1757 __u16 destp, srcp;
1758 s32 delta = tw->tw_ttd - inet_tw_time_stamp();
1759
1760 dest = &tw->tw_v6_daddr;
1761 src = &tw->tw_v6_rcv_saddr;
1762 destp = ntohs(tw->tw_dport);
1763 srcp = ntohs(tw->tw_sport);
1764
1765 seq_printf(seq,
1766 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1767 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
1768 i,
1769 src->s6_addr32[0], src->s6_addr32[1],
1770 src->s6_addr32[2], src->s6_addr32[3], srcp,
1771 dest->s6_addr32[0], dest->s6_addr32[1],
1772 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1773 tw->tw_substate, 0, 0,
1774 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
1775 atomic_read(&tw->tw_refcnt), tw);
1776 }
1777
1778 static int tcp6_seq_show(struct seq_file *seq, void *v)
1779 {
1780 struct tcp_iter_state *st;
1781 struct sock *sk = v;
1782
1783 if (v == SEQ_START_TOKEN) {
1784 seq_puts(seq,
1785 " sl "
1786 "local_address "
1787 "remote_address "
1788 "st tx_queue rx_queue tr tm->when retrnsmt"
1789 " uid timeout inode\n");
1790 goto out;
1791 }
1792 st = seq->private;
1793
1794 switch (st->state) {
1795 case TCP_SEQ_STATE_LISTENING:
1796 case TCP_SEQ_STATE_ESTABLISHED:
1797 if (sk->sk_state == TCP_TIME_WAIT)
1798 get_timewait6_sock(seq, v, st->num);
1799 else
1800 get_tcp6_sock(seq, v, st->num);
1801 break;
1802 case TCP_SEQ_STATE_OPENREQ:
1803 get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
1804 break;
1805 }
1806 out:
1807 return 0;
1808 }
1809
1810 static const struct file_operations tcp6_afinfo_seq_fops = {
1811 .owner = THIS_MODULE,
1812 .open = tcp_seq_open,
1813 .read = seq_read,
1814 .llseek = seq_lseek,
1815 .release = seq_release_net
1816 };
1817
1818 static struct tcp_seq_afinfo tcp6_seq_afinfo = {
1819 .name = "tcp6",
1820 .family = AF_INET6,
1821 .seq_fops = &tcp6_afinfo_seq_fops,
1822 .seq_ops = {
1823 .show = tcp6_seq_show,
1824 },
1825 };
1826
1827 int __net_init tcp6_proc_init(struct net *net)
1828 {
1829 return tcp_proc_register(net, &tcp6_seq_afinfo);
1830 }
1831
1832 void tcp6_proc_exit(struct net *net)
1833 {
1834 tcp_proc_unregister(net, &tcp6_seq_afinfo);
1835 }
1836 #endif
1837
1838 static void tcp_v6_clear_sk(struct sock *sk, int size)
1839 {
1840 struct inet_sock *inet = inet_sk(sk);
1841
1842 /* we do not want to clear pinet6 field, because of RCU lookups */
1843 sk_prot_clear_nulls(sk, offsetof(struct inet_sock, pinet6));
1844
1845 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1846 memset(&inet->pinet6 + 1, 0, size);
1847 }
1848
1849 struct proto tcpv6_prot = {
1850 .name = "TCPv6",
1851 .owner = THIS_MODULE,
1852 .close = tcp_close,
1853 .connect = tcp_v6_connect,
1854 .disconnect = tcp_disconnect,
1855 .accept = inet_csk_accept,
1856 .ioctl = tcp_ioctl,
1857 .init = tcp_v6_init_sock,
1858 .destroy = tcp_v6_destroy_sock,
1859 .shutdown = tcp_shutdown,
1860 .setsockopt = tcp_setsockopt,
1861 .getsockopt = tcp_getsockopt,
1862 .recvmsg = tcp_recvmsg,
1863 .sendmsg = tcp_sendmsg,
1864 .sendpage = tcp_sendpage,
1865 .backlog_rcv = tcp_v6_do_rcv,
1866 .release_cb = tcp_release_cb,
1867 .mtu_reduced = tcp_v6_mtu_reduced,
1868 .hash = tcp_v6_hash,
1869 .unhash = inet_unhash,
1870 .get_port = inet_csk_get_port,
1871 .enter_memory_pressure = tcp_enter_memory_pressure,
1872 .stream_memory_free = tcp_stream_memory_free,
1873 .sockets_allocated = &tcp_sockets_allocated,
1874 .memory_allocated = &tcp_memory_allocated,
1875 .memory_pressure = &tcp_memory_pressure,
1876 .orphan_count = &tcp_orphan_count,
1877 .sysctl_mem = sysctl_tcp_mem,
1878 .sysctl_wmem = sysctl_tcp_wmem,
1879 .sysctl_rmem = sysctl_tcp_rmem,
1880 .max_header = MAX_TCP_HEADER,
1881 .obj_size = sizeof(struct tcp6_sock),
1882 .slab_flags = SLAB_DESTROY_BY_RCU,
1883 .twsk_prot = &tcp6_timewait_sock_ops,
1884 .rsk_prot = &tcp6_request_sock_ops,
1885 .h.hashinfo = &tcp_hashinfo,
1886 .no_autobind = true,
1887 #ifdef CONFIG_COMPAT
1888 .compat_setsockopt = compat_tcp_setsockopt,
1889 .compat_getsockopt = compat_tcp_getsockopt,
1890 #endif
1891 #ifdef CONFIG_MEMCG_KMEM
1892 .proto_cgroup = tcp_proto_cgroup,
1893 #endif
1894 .clear_sk = tcp_v6_clear_sk,
1895 };
1896
1897 static const struct inet6_protocol tcpv6_protocol = {
1898 .early_demux = tcp_v6_early_demux,
1899 .handler = tcp_v6_rcv,
1900 .err_handler = tcp_v6_err,
1901 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1902 };
1903
1904 static struct inet_protosw tcpv6_protosw = {
1905 .type = SOCK_STREAM,
1906 .protocol = IPPROTO_TCP,
1907 .prot = &tcpv6_prot,
1908 .ops = &inet6_stream_ops,
1909 .flags = INET_PROTOSW_PERMANENT |
1910 INET_PROTOSW_ICSK,
1911 };
1912
1913 static int __net_init tcpv6_net_init(struct net *net)
1914 {
1915 return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
1916 SOCK_RAW, IPPROTO_TCP, net);
1917 }
1918
1919 static void __net_exit tcpv6_net_exit(struct net *net)
1920 {
1921 inet_ctl_sock_destroy(net->ipv6.tcp_sk);
1922 }
1923
1924 static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
1925 {
1926 inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6);
1927 }
1928
1929 static struct pernet_operations tcpv6_net_ops = {
1930 .init = tcpv6_net_init,
1931 .exit = tcpv6_net_exit,
1932 .exit_batch = tcpv6_net_exit_batch,
1933 };
1934
1935 int __init tcpv6_init(void)
1936 {
1937 int ret;
1938
1939 ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
1940 if (ret)
1941 goto out;
1942
1943 /* register inet6 protocol */
1944 ret = inet6_register_protosw(&tcpv6_protosw);
1945 if (ret)
1946 goto out_tcpv6_protocol;
1947
1948 ret = register_pernet_subsys(&tcpv6_net_ops);
1949 if (ret)
1950 goto out_tcpv6_protosw;
1951 out:
1952 return ret;
1953
1954 out_tcpv6_protosw:
1955 inet6_unregister_protosw(&tcpv6_protosw);
1956 out_tcpv6_protocol:
1957 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
1958 goto out;
1959 }
1960
1961 void tcpv6_exit(void)
1962 {
1963 unregister_pernet_subsys(&tcpv6_net_ops);
1964 inet6_unregister_protosw(&tcpv6_protosw);
1965 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
1966 }
This page took 0.109121 seconds and 6 git commands to generate.