]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - net/ipv6/tcp_ipv6.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'for-linus-4.8' of git://git.kernel.org/pub/scm/linux/kernel/git/mason...
[mirror_ubuntu-bionic-kernel.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/inet_common.h>
63 #include <net/secure_seq.h>
64 #include <net/busy_poll.h>
65
66 #include <linux/proc_fs.h>
67 #include <linux/seq_file.h>
68
69 #include <crypto/hash.h>
70 #include <linux/scatterlist.h>
71
72 static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb);
73 static void tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
74 struct request_sock *req);
75
76 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
77
78 static const struct inet_connection_sock_af_ops ipv6_mapped;
79 static const struct inet_connection_sock_af_ops ipv6_specific;
80 #ifdef CONFIG_TCP_MD5SIG
81 static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
82 static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
83 #else
84 static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk,
85 const struct in6_addr *addr)
86 {
87 return NULL;
88 }
89 #endif
90
91 static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb)
92 {
93 struct dst_entry *dst = skb_dst(skb);
94
95 if (dst && dst_hold_safe(dst)) {
96 const struct rt6_info *rt = (const struct rt6_info *)dst;
97
98 sk->sk_rx_dst = dst;
99 inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
100 inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt);
101 }
102 }
103
104 static __u32 tcp_v6_init_sequence(const struct sk_buff *skb)
105 {
106 return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
107 ipv6_hdr(skb)->saddr.s6_addr32,
108 tcp_hdr(skb)->dest,
109 tcp_hdr(skb)->source);
110 }
111
112 static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
113 int addr_len)
114 {
115 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
116 struct inet_sock *inet = inet_sk(sk);
117 struct inet_connection_sock *icsk = inet_csk(sk);
118 struct ipv6_pinfo *np = inet6_sk(sk);
119 struct tcp_sock *tp = tcp_sk(sk);
120 struct in6_addr *saddr = NULL, *final_p, final;
121 struct ipv6_txoptions *opt;
122 struct flowi6 fl6;
123 struct dst_entry *dst;
124 int addr_type;
125 int err;
126
127 if (addr_len < SIN6_LEN_RFC2133)
128 return -EINVAL;
129
130 if (usin->sin6_family != AF_INET6)
131 return -EAFNOSUPPORT;
132
133 memset(&fl6, 0, sizeof(fl6));
134
135 if (np->sndflow) {
136 fl6.flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
137 IP6_ECN_flow_init(fl6.flowlabel);
138 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
139 struct ip6_flowlabel *flowlabel;
140 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
141 if (!flowlabel)
142 return -EINVAL;
143 fl6_sock_release(flowlabel);
144 }
145 }
146
147 /*
148 * connect() to INADDR_ANY means loopback (BSD'ism).
149 */
150
151 if (ipv6_addr_any(&usin->sin6_addr))
152 usin->sin6_addr.s6_addr[15] = 0x1;
153
154 addr_type = ipv6_addr_type(&usin->sin6_addr);
155
156 if (addr_type & IPV6_ADDR_MULTICAST)
157 return -ENETUNREACH;
158
159 if (addr_type&IPV6_ADDR_LINKLOCAL) {
160 if (addr_len >= sizeof(struct sockaddr_in6) &&
161 usin->sin6_scope_id) {
162 /* If interface is set while binding, indices
163 * must coincide.
164 */
165 if (sk->sk_bound_dev_if &&
166 sk->sk_bound_dev_if != usin->sin6_scope_id)
167 return -EINVAL;
168
169 sk->sk_bound_dev_if = usin->sin6_scope_id;
170 }
171
172 /* Connect to link-local address requires an interface */
173 if (!sk->sk_bound_dev_if)
174 return -EINVAL;
175 }
176
177 if (tp->rx_opt.ts_recent_stamp &&
178 !ipv6_addr_equal(&sk->sk_v6_daddr, &usin->sin6_addr)) {
179 tp->rx_opt.ts_recent = 0;
180 tp->rx_opt.ts_recent_stamp = 0;
181 tp->write_seq = 0;
182 }
183
184 sk->sk_v6_daddr = usin->sin6_addr;
185 np->flow_label = fl6.flowlabel;
186
187 /*
188 * TCP over IPv4
189 */
190
191 if (addr_type == IPV6_ADDR_MAPPED) {
192 u32 exthdrlen = icsk->icsk_ext_hdr_len;
193 struct sockaddr_in sin;
194
195 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
196
197 if (__ipv6_only_sock(sk))
198 return -ENETUNREACH;
199
200 sin.sin_family = AF_INET;
201 sin.sin_port = usin->sin6_port;
202 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
203
204 icsk->icsk_af_ops = &ipv6_mapped;
205 sk->sk_backlog_rcv = tcp_v4_do_rcv;
206 #ifdef CONFIG_TCP_MD5SIG
207 tp->af_specific = &tcp_sock_ipv6_mapped_specific;
208 #endif
209
210 err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
211
212 if (err) {
213 icsk->icsk_ext_hdr_len = exthdrlen;
214 icsk->icsk_af_ops = &ipv6_specific;
215 sk->sk_backlog_rcv = tcp_v6_do_rcv;
216 #ifdef CONFIG_TCP_MD5SIG
217 tp->af_specific = &tcp_sock_ipv6_specific;
218 #endif
219 goto failure;
220 }
221 np->saddr = sk->sk_v6_rcv_saddr;
222
223 return err;
224 }
225
226 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
227 saddr = &sk->sk_v6_rcv_saddr;
228
229 fl6.flowi6_proto = IPPROTO_TCP;
230 fl6.daddr = sk->sk_v6_daddr;
231 fl6.saddr = saddr ? *saddr : np->saddr;
232 fl6.flowi6_oif = sk->sk_bound_dev_if;
233 fl6.flowi6_mark = sk->sk_mark;
234 fl6.fl6_dport = usin->sin6_port;
235 fl6.fl6_sport = inet->inet_sport;
236
237 opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
238 final_p = fl6_update_dst(&fl6, opt, &final);
239
240 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
241
242 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
243 if (IS_ERR(dst)) {
244 err = PTR_ERR(dst);
245 goto failure;
246 }
247
248 if (!saddr) {
249 saddr = &fl6.saddr;
250 sk->sk_v6_rcv_saddr = *saddr;
251 }
252
253 /* set the source address */
254 np->saddr = *saddr;
255 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
256
257 sk->sk_gso_type = SKB_GSO_TCPV6;
258 ip6_dst_store(sk, dst, NULL, NULL);
259
260 if (tcp_death_row.sysctl_tw_recycle &&
261 !tp->rx_opt.ts_recent_stamp &&
262 ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr))
263 tcp_fetch_timewait_stamp(sk, dst);
264
265 icsk->icsk_ext_hdr_len = 0;
266 if (opt)
267 icsk->icsk_ext_hdr_len = opt->opt_flen +
268 opt->opt_nflen;
269
270 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
271
272 inet->inet_dport = usin->sin6_port;
273
274 tcp_set_state(sk, TCP_SYN_SENT);
275 err = inet6_hash_connect(&tcp_death_row, sk);
276 if (err)
277 goto late_failure;
278
279 sk_set_txhash(sk);
280
281 if (!tp->write_seq && likely(!tp->repair))
282 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
283 sk->sk_v6_daddr.s6_addr32,
284 inet->inet_sport,
285 inet->inet_dport);
286
287 err = tcp_connect(sk);
288 if (err)
289 goto late_failure;
290
291 return 0;
292
293 late_failure:
294 tcp_set_state(sk, TCP_CLOSE);
295 __sk_dst_reset(sk);
296 failure:
297 inet->inet_dport = 0;
298 sk->sk_route_caps = 0;
299 return err;
300 }
301
302 static void tcp_v6_mtu_reduced(struct sock *sk)
303 {
304 struct dst_entry *dst;
305
306 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
307 return;
308
309 dst = inet6_csk_update_pmtu(sk, tcp_sk(sk)->mtu_info);
310 if (!dst)
311 return;
312
313 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
314 tcp_sync_mss(sk, dst_mtu(dst));
315 tcp_simple_retransmit(sk);
316 }
317 }
318
319 static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
320 u8 type, u8 code, int offset, __be32 info)
321 {
322 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
323 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
324 struct net *net = dev_net(skb->dev);
325 struct request_sock *fastopen;
326 struct ipv6_pinfo *np;
327 struct tcp_sock *tp;
328 __u32 seq, snd_una;
329 struct sock *sk;
330 bool fatal;
331 int err;
332
333 sk = __inet6_lookup_established(net, &tcp_hashinfo,
334 &hdr->daddr, th->dest,
335 &hdr->saddr, ntohs(th->source),
336 skb->dev->ifindex);
337
338 if (!sk) {
339 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
340 ICMP6_MIB_INERRORS);
341 return;
342 }
343
344 if (sk->sk_state == TCP_TIME_WAIT) {
345 inet_twsk_put(inet_twsk(sk));
346 return;
347 }
348 seq = ntohl(th->seq);
349 fatal = icmpv6_err_convert(type, code, &err);
350 if (sk->sk_state == TCP_NEW_SYN_RECV)
351 return tcp_req_err(sk, seq, fatal);
352
353 bh_lock_sock(sk);
354 if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG)
355 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
356
357 if (sk->sk_state == TCP_CLOSE)
358 goto out;
359
360 if (ipv6_hdr(skb)->hop_limit < inet6_sk(sk)->min_hopcount) {
361 __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
362 goto out;
363 }
364
365 tp = tcp_sk(sk);
366 /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
367 fastopen = tp->fastopen_rsk;
368 snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
369 if (sk->sk_state != TCP_LISTEN &&
370 !between(seq, snd_una, tp->snd_nxt)) {
371 __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
372 goto out;
373 }
374
375 np = inet6_sk(sk);
376
377 if (type == NDISC_REDIRECT) {
378 struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
379
380 if (dst)
381 dst->ops->redirect(dst, sk, skb);
382 goto out;
383 }
384
385 if (type == ICMPV6_PKT_TOOBIG) {
386 /* We are not interested in TCP_LISTEN and open_requests
387 * (SYN-ACKs send out by Linux are always <576bytes so
388 * they should go through unfragmented).
389 */
390 if (sk->sk_state == TCP_LISTEN)
391 goto out;
392
393 if (!ip6_sk_accept_pmtu(sk))
394 goto out;
395
396 tp->mtu_info = ntohl(info);
397 if (!sock_owned_by_user(sk))
398 tcp_v6_mtu_reduced(sk);
399 else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED,
400 &tp->tsq_flags))
401 sock_hold(sk);
402 goto out;
403 }
404
405
406 /* Might be for an request_sock */
407 switch (sk->sk_state) {
408 case TCP_SYN_SENT:
409 case TCP_SYN_RECV:
410 /* Only in fast or simultaneous open. If a fast open socket is
411 * is already accepted it is treated as a connected one below.
412 */
413 if (fastopen && !fastopen->sk)
414 break;
415
416 if (!sock_owned_by_user(sk)) {
417 sk->sk_err = err;
418 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
419
420 tcp_done(sk);
421 } else
422 sk->sk_err_soft = err;
423 goto out;
424 }
425
426 if (!sock_owned_by_user(sk) && np->recverr) {
427 sk->sk_err = err;
428 sk->sk_error_report(sk);
429 } else
430 sk->sk_err_soft = err;
431
432 out:
433 bh_unlock_sock(sk);
434 sock_put(sk);
435 }
436
437
438 static int tcp_v6_send_synack(const struct sock *sk, struct dst_entry *dst,
439 struct flowi *fl,
440 struct request_sock *req,
441 struct tcp_fastopen_cookie *foc,
442 enum tcp_synack_type synack_type)
443 {
444 struct inet_request_sock *ireq = inet_rsk(req);
445 struct ipv6_pinfo *np = inet6_sk(sk);
446 struct ipv6_txoptions *opt;
447 struct flowi6 *fl6 = &fl->u.ip6;
448 struct sk_buff *skb;
449 int err = -ENOMEM;
450
451 /* First, grab a route. */
452 if (!dst && (dst = inet6_csk_route_req(sk, fl6, req,
453 IPPROTO_TCP)) == NULL)
454 goto done;
455
456 skb = tcp_make_synack(sk, dst, req, foc, synack_type);
457
458 if (skb) {
459 __tcp_v6_send_check(skb, &ireq->ir_v6_loc_addr,
460 &ireq->ir_v6_rmt_addr);
461
462 fl6->daddr = ireq->ir_v6_rmt_addr;
463 if (np->repflow && ireq->pktopts)
464 fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts));
465
466 rcu_read_lock();
467 opt = ireq->ipv6_opt;
468 if (!opt)
469 opt = rcu_dereference(np->opt);
470 err = ip6_xmit(sk, skb, fl6, opt, np->tclass);
471 rcu_read_unlock();
472 err = net_xmit_eval(err);
473 }
474
475 done:
476 return err;
477 }
478
479
480 static void tcp_v6_reqsk_destructor(struct request_sock *req)
481 {
482 kfree(inet_rsk(req)->ipv6_opt);
483 kfree_skb(inet_rsk(req)->pktopts);
484 }
485
486 #ifdef CONFIG_TCP_MD5SIG
487 static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk,
488 const struct in6_addr *addr)
489 {
490 return tcp_md5_do_lookup(sk, (union tcp_md5_addr *)addr, AF_INET6);
491 }
492
493 static struct tcp_md5sig_key *tcp_v6_md5_lookup(const struct sock *sk,
494 const struct sock *addr_sk)
495 {
496 return tcp_v6_md5_do_lookup(sk, &addr_sk->sk_v6_daddr);
497 }
498
499 static int tcp_v6_parse_md5_keys(struct sock *sk, char __user *optval,
500 int optlen)
501 {
502 struct tcp_md5sig cmd;
503 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
504
505 if (optlen < sizeof(cmd))
506 return -EINVAL;
507
508 if (copy_from_user(&cmd, optval, sizeof(cmd)))
509 return -EFAULT;
510
511 if (sin6->sin6_family != AF_INET6)
512 return -EINVAL;
513
514 if (!cmd.tcpm_keylen) {
515 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
516 return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
517 AF_INET);
518 return tcp_md5_do_del(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
519 AF_INET6);
520 }
521
522 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
523 return -EINVAL;
524
525 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
526 return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
527 AF_INET, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
528
529 return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
530 AF_INET6, cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
531 }
532
533 static int tcp_v6_md5_hash_headers(struct tcp_md5sig_pool *hp,
534 const struct in6_addr *daddr,
535 const struct in6_addr *saddr,
536 const struct tcphdr *th, int nbytes)
537 {
538 struct tcp6_pseudohdr *bp;
539 struct scatterlist sg;
540 struct tcphdr *_th;
541
542 bp = hp->scratch;
543 /* 1. TCP pseudo-header (RFC2460) */
544 bp->saddr = *saddr;
545 bp->daddr = *daddr;
546 bp->protocol = cpu_to_be32(IPPROTO_TCP);
547 bp->len = cpu_to_be32(nbytes);
548
549 _th = (struct tcphdr *)(bp + 1);
550 memcpy(_th, th, sizeof(*th));
551 _th->check = 0;
552
553 sg_init_one(&sg, bp, sizeof(*bp) + sizeof(*th));
554 ahash_request_set_crypt(hp->md5_req, &sg, NULL,
555 sizeof(*bp) + sizeof(*th));
556 return crypto_ahash_update(hp->md5_req);
557 }
558
559 static int tcp_v6_md5_hash_hdr(char *md5_hash, const struct tcp_md5sig_key *key,
560 const struct in6_addr *daddr, struct in6_addr *saddr,
561 const struct tcphdr *th)
562 {
563 struct tcp_md5sig_pool *hp;
564 struct ahash_request *req;
565
566 hp = tcp_get_md5sig_pool();
567 if (!hp)
568 goto clear_hash_noput;
569 req = hp->md5_req;
570
571 if (crypto_ahash_init(req))
572 goto clear_hash;
573 if (tcp_v6_md5_hash_headers(hp, daddr, saddr, th, th->doff << 2))
574 goto clear_hash;
575 if (tcp_md5_hash_key(hp, key))
576 goto clear_hash;
577 ahash_request_set_crypt(req, NULL, md5_hash, 0);
578 if (crypto_ahash_final(req))
579 goto clear_hash;
580
581 tcp_put_md5sig_pool();
582 return 0;
583
584 clear_hash:
585 tcp_put_md5sig_pool();
586 clear_hash_noput:
587 memset(md5_hash, 0, 16);
588 return 1;
589 }
590
591 static int tcp_v6_md5_hash_skb(char *md5_hash,
592 const struct tcp_md5sig_key *key,
593 const struct sock *sk,
594 const struct sk_buff *skb)
595 {
596 const struct in6_addr *saddr, *daddr;
597 struct tcp_md5sig_pool *hp;
598 struct ahash_request *req;
599 const struct tcphdr *th = tcp_hdr(skb);
600
601 if (sk) { /* valid for establish/request sockets */
602 saddr = &sk->sk_v6_rcv_saddr;
603 daddr = &sk->sk_v6_daddr;
604 } else {
605 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
606 saddr = &ip6h->saddr;
607 daddr = &ip6h->daddr;
608 }
609
610 hp = tcp_get_md5sig_pool();
611 if (!hp)
612 goto clear_hash_noput;
613 req = hp->md5_req;
614
615 if (crypto_ahash_init(req))
616 goto clear_hash;
617
618 if (tcp_v6_md5_hash_headers(hp, daddr, saddr, th, skb->len))
619 goto clear_hash;
620 if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
621 goto clear_hash;
622 if (tcp_md5_hash_key(hp, key))
623 goto clear_hash;
624 ahash_request_set_crypt(req, NULL, md5_hash, 0);
625 if (crypto_ahash_final(req))
626 goto clear_hash;
627
628 tcp_put_md5sig_pool();
629 return 0;
630
631 clear_hash:
632 tcp_put_md5sig_pool();
633 clear_hash_noput:
634 memset(md5_hash, 0, 16);
635 return 1;
636 }
637
638 #endif
639
640 static bool tcp_v6_inbound_md5_hash(const struct sock *sk,
641 const struct sk_buff *skb)
642 {
643 #ifdef CONFIG_TCP_MD5SIG
644 const __u8 *hash_location = NULL;
645 struct tcp_md5sig_key *hash_expected;
646 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
647 const struct tcphdr *th = tcp_hdr(skb);
648 int genhash;
649 u8 newhash[16];
650
651 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
652 hash_location = tcp_parse_md5sig_option(th);
653
654 /* We've parsed the options - do we have a hash? */
655 if (!hash_expected && !hash_location)
656 return false;
657
658 if (hash_expected && !hash_location) {
659 NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
660 return true;
661 }
662
663 if (!hash_expected && hash_location) {
664 NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
665 return true;
666 }
667
668 /* check the signature */
669 genhash = tcp_v6_md5_hash_skb(newhash,
670 hash_expected,
671 NULL, skb);
672
673 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
674 net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n",
675 genhash ? "failed" : "mismatch",
676 &ip6h->saddr, ntohs(th->source),
677 &ip6h->daddr, ntohs(th->dest));
678 return true;
679 }
680 #endif
681 return false;
682 }
683
684 static void tcp_v6_init_req(struct request_sock *req,
685 const struct sock *sk_listener,
686 struct sk_buff *skb)
687 {
688 struct inet_request_sock *ireq = inet_rsk(req);
689 const struct ipv6_pinfo *np = inet6_sk(sk_listener);
690
691 ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
692 ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
693
694 /* So that link locals have meaning */
695 if (!sk_listener->sk_bound_dev_if &&
696 ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
697 ireq->ir_iif = tcp_v6_iif(skb);
698
699 if (!TCP_SKB_CB(skb)->tcp_tw_isn &&
700 (ipv6_opt_accepted(sk_listener, skb, &TCP_SKB_CB(skb)->header.h6) ||
701 np->rxopt.bits.rxinfo ||
702 np->rxopt.bits.rxoinfo || np->rxopt.bits.rxhlim ||
703 np->rxopt.bits.rxohlim || np->repflow)) {
704 atomic_inc(&skb->users);
705 ireq->pktopts = skb;
706 }
707 }
708
709 static struct dst_entry *tcp_v6_route_req(const struct sock *sk,
710 struct flowi *fl,
711 const struct request_sock *req,
712 bool *strict)
713 {
714 if (strict)
715 *strict = true;
716 return inet6_csk_route_req(sk, &fl->u.ip6, req, IPPROTO_TCP);
717 }
718
719 struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
720 .family = AF_INET6,
721 .obj_size = sizeof(struct tcp6_request_sock),
722 .rtx_syn_ack = tcp_rtx_synack,
723 .send_ack = tcp_v6_reqsk_send_ack,
724 .destructor = tcp_v6_reqsk_destructor,
725 .send_reset = tcp_v6_send_reset,
726 .syn_ack_timeout = tcp_syn_ack_timeout,
727 };
728
729 static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
730 .mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) -
731 sizeof(struct ipv6hdr),
732 #ifdef CONFIG_TCP_MD5SIG
733 .req_md5_lookup = tcp_v6_md5_lookup,
734 .calc_md5_hash = tcp_v6_md5_hash_skb,
735 #endif
736 .init_req = tcp_v6_init_req,
737 #ifdef CONFIG_SYN_COOKIES
738 .cookie_init_seq = cookie_v6_init_sequence,
739 #endif
740 .route_req = tcp_v6_route_req,
741 .init_seq = tcp_v6_init_sequence,
742 .send_synack = tcp_v6_send_synack,
743 };
744
745 static void tcp_v6_send_response(const struct sock *sk, struct sk_buff *skb, u32 seq,
746 u32 ack, u32 win, u32 tsval, u32 tsecr,
747 int oif, struct tcp_md5sig_key *key, int rst,
748 u8 tclass, __be32 label)
749 {
750 const struct tcphdr *th = tcp_hdr(skb);
751 struct tcphdr *t1;
752 struct sk_buff *buff;
753 struct flowi6 fl6;
754 struct net *net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
755 struct sock *ctl_sk = net->ipv6.tcp_sk;
756 unsigned int tot_len = sizeof(struct tcphdr);
757 struct dst_entry *dst;
758 __be32 *topt;
759
760 if (tsecr)
761 tot_len += TCPOLEN_TSTAMP_ALIGNED;
762 #ifdef CONFIG_TCP_MD5SIG
763 if (key)
764 tot_len += TCPOLEN_MD5SIG_ALIGNED;
765 #endif
766
767 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
768 GFP_ATOMIC);
769 if (!buff)
770 return;
771
772 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
773
774 t1 = (struct tcphdr *) skb_push(buff, tot_len);
775 skb_reset_transport_header(buff);
776
777 /* Swap the send and the receive. */
778 memset(t1, 0, sizeof(*t1));
779 t1->dest = th->source;
780 t1->source = th->dest;
781 t1->doff = tot_len / 4;
782 t1->seq = htonl(seq);
783 t1->ack_seq = htonl(ack);
784 t1->ack = !rst || !th->ack;
785 t1->rst = rst;
786 t1->window = htons(win);
787
788 topt = (__be32 *)(t1 + 1);
789
790 if (tsecr) {
791 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
792 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
793 *topt++ = htonl(tsval);
794 *topt++ = htonl(tsecr);
795 }
796
797 #ifdef CONFIG_TCP_MD5SIG
798 if (key) {
799 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
800 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
801 tcp_v6_md5_hash_hdr((__u8 *)topt, key,
802 &ipv6_hdr(skb)->saddr,
803 &ipv6_hdr(skb)->daddr, t1);
804 }
805 #endif
806
807 memset(&fl6, 0, sizeof(fl6));
808 fl6.daddr = ipv6_hdr(skb)->saddr;
809 fl6.saddr = ipv6_hdr(skb)->daddr;
810 fl6.flowlabel = label;
811
812 buff->ip_summed = CHECKSUM_PARTIAL;
813 buff->csum = 0;
814
815 __tcp_v6_send_check(buff, &fl6.saddr, &fl6.daddr);
816
817 fl6.flowi6_proto = IPPROTO_TCP;
818 if (rt6_need_strict(&fl6.daddr) && !oif)
819 fl6.flowi6_oif = tcp_v6_iif(skb);
820 else {
821 if (!oif && netif_index_is_l3_master(net, skb->skb_iif))
822 oif = skb->skb_iif;
823
824 fl6.flowi6_oif = oif;
825 }
826
827 fl6.flowi6_mark = IP6_REPLY_MARK(net, skb->mark);
828 fl6.fl6_dport = t1->dest;
829 fl6.fl6_sport = t1->source;
830 security_skb_classify_flow(skb, flowi6_to_flowi(&fl6));
831
832 /* Pass a socket to ip6_dst_lookup either it is for RST
833 * Underlying function will use this to retrieve the network
834 * namespace
835 */
836 dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
837 if (!IS_ERR(dst)) {
838 skb_dst_set(buff, dst);
839 ip6_xmit(ctl_sk, buff, &fl6, NULL, tclass);
840 TCP_INC_STATS(net, TCP_MIB_OUTSEGS);
841 if (rst)
842 TCP_INC_STATS(net, TCP_MIB_OUTRSTS);
843 return;
844 }
845
846 kfree_skb(buff);
847 }
848
849 static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb)
850 {
851 const struct tcphdr *th = tcp_hdr(skb);
852 u32 seq = 0, ack_seq = 0;
853 struct tcp_md5sig_key *key = NULL;
854 #ifdef CONFIG_TCP_MD5SIG
855 const __u8 *hash_location = NULL;
856 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
857 unsigned char newhash[16];
858 int genhash;
859 struct sock *sk1 = NULL;
860 #endif
861 int oif;
862
863 if (th->rst)
864 return;
865
866 /* If sk not NULL, it means we did a successful lookup and incoming
867 * route had to be correct. prequeue might have dropped our dst.
868 */
869 if (!sk && !ipv6_unicast_destination(skb))
870 return;
871
872 #ifdef CONFIG_TCP_MD5SIG
873 rcu_read_lock();
874 hash_location = tcp_parse_md5sig_option(th);
875 if (sk && sk_fullsock(sk)) {
876 key = tcp_v6_md5_do_lookup(sk, &ipv6h->saddr);
877 } else if (hash_location) {
878 /*
879 * active side is lost. Try to find listening socket through
880 * source port, and then find md5 key through listening socket.
881 * we are not loose security here:
882 * Incoming packet is checked with md5 hash with finding key,
883 * no RST generated if md5 hash doesn't match.
884 */
885 sk1 = inet6_lookup_listener(dev_net(skb_dst(skb)->dev),
886 &tcp_hashinfo, NULL, 0,
887 &ipv6h->saddr,
888 th->source, &ipv6h->daddr,
889 ntohs(th->source), tcp_v6_iif(skb));
890 if (!sk1)
891 goto out;
892
893 key = tcp_v6_md5_do_lookup(sk1, &ipv6h->saddr);
894 if (!key)
895 goto out;
896
897 genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, skb);
898 if (genhash || memcmp(hash_location, newhash, 16) != 0)
899 goto out;
900 }
901 #endif
902
903 if (th->ack)
904 seq = ntohl(th->ack_seq);
905 else
906 ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len -
907 (th->doff << 2);
908
909 oif = sk ? sk->sk_bound_dev_if : 0;
910 tcp_v6_send_response(sk, skb, seq, ack_seq, 0, 0, 0, oif, key, 1, 0, 0);
911
912 #ifdef CONFIG_TCP_MD5SIG
913 out:
914 rcu_read_unlock();
915 #endif
916 }
917
918 static void tcp_v6_send_ack(const struct sock *sk, struct sk_buff *skb, u32 seq,
919 u32 ack, u32 win, u32 tsval, u32 tsecr, int oif,
920 struct tcp_md5sig_key *key, u8 tclass,
921 __be32 label)
922 {
923 tcp_v6_send_response(sk, skb, seq, ack, win, tsval, tsecr, oif, key, 0,
924 tclass, label);
925 }
926
927 static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
928 {
929 struct inet_timewait_sock *tw = inet_twsk(sk);
930 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
931
932 tcp_v6_send_ack(sk, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
933 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
934 tcp_time_stamp + tcptw->tw_ts_offset,
935 tcptw->tw_ts_recent, tw->tw_bound_dev_if, tcp_twsk_md5_key(tcptw),
936 tw->tw_tclass, cpu_to_be32(tw->tw_flowlabel));
937
938 inet_twsk_put(tw);
939 }
940
941 static void tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
942 struct request_sock *req)
943 {
944 /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
945 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
946 */
947 /* RFC 7323 2.3
948 * The window field (SEG.WND) of every outgoing segment, with the
949 * exception of <SYN> segments, MUST be right-shifted by
950 * Rcv.Wind.Shift bits:
951 */
952 tcp_v6_send_ack(sk, skb, (sk->sk_state == TCP_LISTEN) ?
953 tcp_rsk(req)->snt_isn + 1 : tcp_sk(sk)->snd_nxt,
954 tcp_rsk(req)->rcv_nxt,
955 req->rsk_rcv_wnd >> inet_rsk(req)->rcv_wscale,
956 tcp_time_stamp, req->ts_recent, sk->sk_bound_dev_if,
957 tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr),
958 0, 0);
959 }
960
961
962 static struct sock *tcp_v6_cookie_check(struct sock *sk, struct sk_buff *skb)
963 {
964 #ifdef CONFIG_SYN_COOKIES
965 const struct tcphdr *th = tcp_hdr(skb);
966
967 if (!th->syn)
968 sk = cookie_v6_check(sk, skb);
969 #endif
970 return sk;
971 }
972
973 static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
974 {
975 if (skb->protocol == htons(ETH_P_IP))
976 return tcp_v4_conn_request(sk, skb);
977
978 if (!ipv6_unicast_destination(skb))
979 goto drop;
980
981 return tcp_conn_request(&tcp6_request_sock_ops,
982 &tcp_request_sock_ipv6_ops, sk, skb);
983
984 drop:
985 tcp_listendrop(sk);
986 return 0; /* don't send reset */
987 }
988
989 static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
990 struct request_sock *req,
991 struct dst_entry *dst,
992 struct request_sock *req_unhash,
993 bool *own_req)
994 {
995 struct inet_request_sock *ireq;
996 struct ipv6_pinfo *newnp;
997 const struct ipv6_pinfo *np = inet6_sk(sk);
998 struct ipv6_txoptions *opt;
999 struct tcp6_sock *newtcp6sk;
1000 struct inet_sock *newinet;
1001 struct tcp_sock *newtp;
1002 struct sock *newsk;
1003 #ifdef CONFIG_TCP_MD5SIG
1004 struct tcp_md5sig_key *key;
1005 #endif
1006 struct flowi6 fl6;
1007
1008 if (skb->protocol == htons(ETH_P_IP)) {
1009 /*
1010 * v6 mapped
1011 */
1012
1013 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst,
1014 req_unhash, own_req);
1015
1016 if (!newsk)
1017 return NULL;
1018
1019 newtcp6sk = (struct tcp6_sock *)newsk;
1020 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1021
1022 newinet = inet_sk(newsk);
1023 newnp = inet6_sk(newsk);
1024 newtp = tcp_sk(newsk);
1025
1026 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1027
1028 newnp->saddr = newsk->sk_v6_rcv_saddr;
1029
1030 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1031 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1032 #ifdef CONFIG_TCP_MD5SIG
1033 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1034 #endif
1035
1036 newnp->ipv6_ac_list = NULL;
1037 newnp->ipv6_fl_list = NULL;
1038 newnp->pktoptions = NULL;
1039 newnp->opt = NULL;
1040 newnp->mcast_oif = tcp_v6_iif(skb);
1041 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1042 newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1043 if (np->repflow)
1044 newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
1045
1046 /*
1047 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1048 * here, tcp_create_openreq_child now does this for us, see the comment in
1049 * that function for the gory details. -acme
1050 */
1051
1052 /* It is tricky place. Until this moment IPv4 tcp
1053 worked with IPv6 icsk.icsk_af_ops.
1054 Sync it now.
1055 */
1056 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1057
1058 return newsk;
1059 }
1060
1061 ireq = inet_rsk(req);
1062
1063 if (sk_acceptq_is_full(sk))
1064 goto out_overflow;
1065
1066 if (!dst) {
1067 dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_TCP);
1068 if (!dst)
1069 goto out;
1070 }
1071
1072 newsk = tcp_create_openreq_child(sk, req, skb);
1073 if (!newsk)
1074 goto out_nonewsk;
1075
1076 /*
1077 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1078 * count here, tcp_create_openreq_child now does this for us, see the
1079 * comment in that function for the gory details. -acme
1080 */
1081
1082 newsk->sk_gso_type = SKB_GSO_TCPV6;
1083 ip6_dst_store(newsk, dst, NULL, NULL);
1084 inet6_sk_rx_dst_set(newsk, skb);
1085
1086 newtcp6sk = (struct tcp6_sock *)newsk;
1087 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1088
1089 newtp = tcp_sk(newsk);
1090 newinet = inet_sk(newsk);
1091 newnp = inet6_sk(newsk);
1092
1093 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1094
1095 newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr;
1096 newnp->saddr = ireq->ir_v6_loc_addr;
1097 newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr;
1098 newsk->sk_bound_dev_if = ireq->ir_iif;
1099
1100 /* Now IPv6 options...
1101
1102 First: no IPv4 options.
1103 */
1104 newinet->inet_opt = NULL;
1105 newnp->ipv6_ac_list = NULL;
1106 newnp->ipv6_fl_list = NULL;
1107
1108 /* Clone RX bits */
1109 newnp->rxopt.all = np->rxopt.all;
1110
1111 newnp->pktoptions = NULL;
1112 newnp->opt = NULL;
1113 newnp->mcast_oif = tcp_v6_iif(skb);
1114 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1115 newnp->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(skb));
1116 if (np->repflow)
1117 newnp->flow_label = ip6_flowlabel(ipv6_hdr(skb));
1118
1119 /* Clone native IPv6 options from listening socket (if any)
1120
1121 Yes, keeping reference count would be much more clever,
1122 but we make one more one thing there: reattach optmem
1123 to newsk.
1124 */
1125 opt = ireq->ipv6_opt;
1126 if (!opt)
1127 opt = rcu_dereference(np->opt);
1128 if (opt) {
1129 opt = ipv6_dup_options(newsk, opt);
1130 RCU_INIT_POINTER(newnp->opt, opt);
1131 }
1132 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1133 if (opt)
1134 inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
1135 opt->opt_flen;
1136
1137 tcp_ca_openreq_child(newsk, dst);
1138
1139 tcp_sync_mss(newsk, dst_mtu(dst));
1140 newtp->advmss = dst_metric_advmss(dst);
1141 if (tcp_sk(sk)->rx_opt.user_mss &&
1142 tcp_sk(sk)->rx_opt.user_mss < newtp->advmss)
1143 newtp->advmss = tcp_sk(sk)->rx_opt.user_mss;
1144
1145 tcp_initialize_rcv_mss(newsk);
1146
1147 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
1148 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
1149
1150 #ifdef CONFIG_TCP_MD5SIG
1151 /* Copy over the MD5 key from the original socket */
1152 key = tcp_v6_md5_do_lookup(sk, &newsk->sk_v6_daddr);
1153 if (key) {
1154 /* We're using one, so create a matching key
1155 * on the newsk structure. If we fail to get
1156 * memory, then we end up not copying the key
1157 * across. Shucks.
1158 */
1159 tcp_md5_do_add(newsk, (union tcp_md5_addr *)&newsk->sk_v6_daddr,
1160 AF_INET6, key->key, key->keylen,
1161 sk_gfp_mask(sk, GFP_ATOMIC));
1162 }
1163 #endif
1164
1165 if (__inet_inherit_port(sk, newsk) < 0) {
1166 inet_csk_prepare_forced_close(newsk);
1167 tcp_done(newsk);
1168 goto out;
1169 }
1170 *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
1171 if (*own_req) {
1172 tcp_move_syn(newtp, req);
1173
1174 /* Clone pktoptions received with SYN, if we own the req */
1175 if (ireq->pktopts) {
1176 newnp->pktoptions = skb_clone(ireq->pktopts,
1177 sk_gfp_mask(sk, GFP_ATOMIC));
1178 consume_skb(ireq->pktopts);
1179 ireq->pktopts = NULL;
1180 if (newnp->pktoptions)
1181 skb_set_owner_r(newnp->pktoptions, newsk);
1182 }
1183 }
1184
1185 return newsk;
1186
1187 out_overflow:
1188 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1189 out_nonewsk:
1190 dst_release(dst);
1191 out:
1192 tcp_listendrop(sk);
1193 return NULL;
1194 }
1195
1196 /* The socket must have it's spinlock held when we get
1197 * here, unless it is a TCP_LISTEN socket.
1198 *
1199 * We have a potential double-lock case here, so even when
1200 * doing backlog processing we use the BH locking scheme.
1201 * This is because we cannot sleep with the original spinlock
1202 * held.
1203 */
1204 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1205 {
1206 struct ipv6_pinfo *np = inet6_sk(sk);
1207 struct tcp_sock *tp;
1208 struct sk_buff *opt_skb = NULL;
1209
1210 /* Imagine: socket is IPv6. IPv4 packet arrives,
1211 goes to IPv4 receive handler and backlogged.
1212 From backlog it always goes here. Kerboom...
1213 Fortunately, tcp_rcv_established and rcv_established
1214 handle them correctly, but it is not case with
1215 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1216 */
1217
1218 if (skb->protocol == htons(ETH_P_IP))
1219 return tcp_v4_do_rcv(sk, skb);
1220
1221 if (sk_filter(sk, skb))
1222 goto discard;
1223
1224 /*
1225 * socket locking is here for SMP purposes as backlog rcv
1226 * is currently called with bh processing disabled.
1227 */
1228
1229 /* Do Stevens' IPV6_PKTOPTIONS.
1230
1231 Yes, guys, it is the only place in our code, where we
1232 may make it not affecting IPv4.
1233 The rest of code is protocol independent,
1234 and I do not like idea to uglify IPv4.
1235
1236 Actually, all the idea behind IPV6_PKTOPTIONS
1237 looks not very well thought. For now we latch
1238 options, received in the last packet, enqueued
1239 by tcp. Feel free to propose better solution.
1240 --ANK (980728)
1241 */
1242 if (np->rxopt.all)
1243 opt_skb = skb_clone(skb, sk_gfp_mask(sk, GFP_ATOMIC));
1244
1245 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1246 struct dst_entry *dst = sk->sk_rx_dst;
1247
1248 sock_rps_save_rxhash(sk, skb);
1249 sk_mark_napi_id(sk, skb);
1250 if (dst) {
1251 if (inet_sk(sk)->rx_dst_ifindex != skb->skb_iif ||
1252 dst->ops->check(dst, np->rx_dst_cookie) == NULL) {
1253 dst_release(dst);
1254 sk->sk_rx_dst = NULL;
1255 }
1256 }
1257
1258 tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len);
1259 if (opt_skb)
1260 goto ipv6_pktoptions;
1261 return 0;
1262 }
1263
1264 if (tcp_checksum_complete(skb))
1265 goto csum_err;
1266
1267 if (sk->sk_state == TCP_LISTEN) {
1268 struct sock *nsk = tcp_v6_cookie_check(sk, skb);
1269
1270 if (!nsk)
1271 goto discard;
1272
1273 if (nsk != sk) {
1274 sock_rps_save_rxhash(nsk, skb);
1275 sk_mark_napi_id(nsk, skb);
1276 if (tcp_child_process(sk, nsk, skb))
1277 goto reset;
1278 if (opt_skb)
1279 __kfree_skb(opt_skb);
1280 return 0;
1281 }
1282 } else
1283 sock_rps_save_rxhash(sk, skb);
1284
1285 if (tcp_rcv_state_process(sk, skb))
1286 goto reset;
1287 if (opt_skb)
1288 goto ipv6_pktoptions;
1289 return 0;
1290
1291 reset:
1292 tcp_v6_send_reset(sk, skb);
1293 discard:
1294 if (opt_skb)
1295 __kfree_skb(opt_skb);
1296 kfree_skb(skb);
1297 return 0;
1298 csum_err:
1299 TCP_INC_STATS(sock_net(sk), TCP_MIB_CSUMERRORS);
1300 TCP_INC_STATS(sock_net(sk), TCP_MIB_INERRS);
1301 goto discard;
1302
1303
1304 ipv6_pktoptions:
1305 /* Do you ask, what is it?
1306
1307 1. skb was enqueued by tcp.
1308 2. skb is added to tail of read queue, rather than out of order.
1309 3. socket is not in passive state.
1310 4. Finally, it really contains options, which user wants to receive.
1311 */
1312 tp = tcp_sk(sk);
1313 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1314 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
1315 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
1316 np->mcast_oif = tcp_v6_iif(opt_skb);
1317 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1318 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1319 if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
1320 np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
1321 if (np->repflow)
1322 np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
1323 if (ipv6_opt_accepted(sk, opt_skb, &TCP_SKB_CB(opt_skb)->header.h6)) {
1324 skb_set_owner_r(opt_skb, sk);
1325 opt_skb = xchg(&np->pktoptions, opt_skb);
1326 } else {
1327 __kfree_skb(opt_skb);
1328 opt_skb = xchg(&np->pktoptions, NULL);
1329 }
1330 }
1331
1332 kfree_skb(opt_skb);
1333 return 0;
1334 }
1335
1336 static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
1337 const struct tcphdr *th)
1338 {
1339 /* This is tricky: we move IP6CB at its correct location into
1340 * TCP_SKB_CB(). It must be done after xfrm6_policy_check(), because
1341 * _decode_session6() uses IP6CB().
1342 * barrier() makes sure compiler won't play aliasing games.
1343 */
1344 memmove(&TCP_SKB_CB(skb)->header.h6, IP6CB(skb),
1345 sizeof(struct inet6_skb_parm));
1346 barrier();
1347
1348 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1349 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1350 skb->len - th->doff*4);
1351 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1352 TCP_SKB_CB(skb)->tcp_flags = tcp_flag_byte(th);
1353 TCP_SKB_CB(skb)->tcp_tw_isn = 0;
1354 TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
1355 TCP_SKB_CB(skb)->sacked = 0;
1356 }
1357
1358 static void tcp_v6_restore_cb(struct sk_buff *skb)
1359 {
1360 /* We need to move header back to the beginning if xfrm6_policy_check()
1361 * and tcp_v6_fill_cb() are going to be called again.
1362 */
1363 memmove(IP6CB(skb), &TCP_SKB_CB(skb)->header.h6,
1364 sizeof(struct inet6_skb_parm));
1365 }
1366
1367 static int tcp_v6_rcv(struct sk_buff *skb)
1368 {
1369 const struct tcphdr *th;
1370 const struct ipv6hdr *hdr;
1371 bool refcounted;
1372 struct sock *sk;
1373 int ret;
1374 struct net *net = dev_net(skb->dev);
1375
1376 if (skb->pkt_type != PACKET_HOST)
1377 goto discard_it;
1378
1379 /*
1380 * Count it even if it's bad.
1381 */
1382 __TCP_INC_STATS(net, TCP_MIB_INSEGS);
1383
1384 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1385 goto discard_it;
1386
1387 th = (const struct tcphdr *)skb->data;
1388
1389 if (unlikely(th->doff < sizeof(struct tcphdr)/4))
1390 goto bad_packet;
1391 if (!pskb_may_pull(skb, th->doff*4))
1392 goto discard_it;
1393
1394 if (skb_checksum_init(skb, IPPROTO_TCP, ip6_compute_pseudo))
1395 goto csum_error;
1396
1397 th = (const struct tcphdr *)skb->data;
1398 hdr = ipv6_hdr(skb);
1399
1400 lookup:
1401 sk = __inet6_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th),
1402 th->source, th->dest, inet6_iif(skb),
1403 &refcounted);
1404 if (!sk)
1405 goto no_tcp_socket;
1406
1407 process:
1408 if (sk->sk_state == TCP_TIME_WAIT)
1409 goto do_time_wait;
1410
1411 if (sk->sk_state == TCP_NEW_SYN_RECV) {
1412 struct request_sock *req = inet_reqsk(sk);
1413 struct sock *nsk;
1414
1415 sk = req->rsk_listener;
1416 tcp_v6_fill_cb(skb, hdr, th);
1417 if (tcp_v6_inbound_md5_hash(sk, skb)) {
1418 reqsk_put(req);
1419 goto discard_it;
1420 }
1421 if (unlikely(sk->sk_state != TCP_LISTEN)) {
1422 inet_csk_reqsk_queue_drop_and_put(sk, req);
1423 goto lookup;
1424 }
1425 sock_hold(sk);
1426 refcounted = true;
1427 nsk = tcp_check_req(sk, skb, req, false);
1428 if (!nsk) {
1429 reqsk_put(req);
1430 goto discard_and_relse;
1431 }
1432 if (nsk == sk) {
1433 reqsk_put(req);
1434 tcp_v6_restore_cb(skb);
1435 } else if (tcp_child_process(sk, nsk, skb)) {
1436 tcp_v6_send_reset(nsk, skb);
1437 goto discard_and_relse;
1438 } else {
1439 sock_put(sk);
1440 return 0;
1441 }
1442 }
1443 if (hdr->hop_limit < inet6_sk(sk)->min_hopcount) {
1444 __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
1445 goto discard_and_relse;
1446 }
1447
1448 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1449 goto discard_and_relse;
1450
1451 tcp_v6_fill_cb(skb, hdr, th);
1452
1453 if (tcp_v6_inbound_md5_hash(sk, skb))
1454 goto discard_and_relse;
1455
1456 if (sk_filter(sk, skb))
1457 goto discard_and_relse;
1458
1459 skb->dev = NULL;
1460
1461 if (sk->sk_state == TCP_LISTEN) {
1462 ret = tcp_v6_do_rcv(sk, skb);
1463 goto put_and_return;
1464 }
1465
1466 sk_incoming_cpu_update(sk);
1467
1468 bh_lock_sock_nested(sk);
1469 tcp_segs_in(tcp_sk(sk), skb);
1470 ret = 0;
1471 if (!sock_owned_by_user(sk)) {
1472 if (!tcp_prequeue(sk, skb))
1473 ret = tcp_v6_do_rcv(sk, skb);
1474 } else if (unlikely(sk_add_backlog(sk, skb,
1475 sk->sk_rcvbuf + sk->sk_sndbuf))) {
1476 bh_unlock_sock(sk);
1477 __NET_INC_STATS(net, LINUX_MIB_TCPBACKLOGDROP);
1478 goto discard_and_relse;
1479 }
1480 bh_unlock_sock(sk);
1481
1482 put_and_return:
1483 if (refcounted)
1484 sock_put(sk);
1485 return ret ? -1 : 0;
1486
1487 no_tcp_socket:
1488 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1489 goto discard_it;
1490
1491 tcp_v6_fill_cb(skb, hdr, th);
1492
1493 if (tcp_checksum_complete(skb)) {
1494 csum_error:
1495 __TCP_INC_STATS(net, TCP_MIB_CSUMERRORS);
1496 bad_packet:
1497 __TCP_INC_STATS(net, TCP_MIB_INERRS);
1498 } else {
1499 tcp_v6_send_reset(NULL, skb);
1500 }
1501
1502 discard_it:
1503 kfree_skb(skb);
1504 return 0;
1505
1506 discard_and_relse:
1507 sk_drops_add(sk, skb);
1508 if (refcounted)
1509 sock_put(sk);
1510 goto discard_it;
1511
1512 do_time_wait:
1513 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1514 inet_twsk_put(inet_twsk(sk));
1515 goto discard_it;
1516 }
1517
1518 tcp_v6_fill_cb(skb, hdr, th);
1519
1520 if (tcp_checksum_complete(skb)) {
1521 inet_twsk_put(inet_twsk(sk));
1522 goto csum_error;
1523 }
1524
1525 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1526 case TCP_TW_SYN:
1527 {
1528 struct sock *sk2;
1529
1530 sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1531 skb, __tcp_hdrlen(th),
1532 &ipv6_hdr(skb)->saddr, th->source,
1533 &ipv6_hdr(skb)->daddr,
1534 ntohs(th->dest), tcp_v6_iif(skb));
1535 if (sk2) {
1536 struct inet_timewait_sock *tw = inet_twsk(sk);
1537 inet_twsk_deschedule_put(tw);
1538 sk = sk2;
1539 tcp_v6_restore_cb(skb);
1540 refcounted = false;
1541 goto process;
1542 }
1543 /* Fall through to ACK */
1544 }
1545 case TCP_TW_ACK:
1546 tcp_v6_timewait_ack(sk, skb);
1547 break;
1548 case TCP_TW_RST:
1549 tcp_v6_restore_cb(skb);
1550 tcp_v6_send_reset(sk, skb);
1551 inet_twsk_deschedule_put(inet_twsk(sk));
1552 goto discard_it;
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_fullsock(sk)) {
1586 struct dst_entry *dst = READ_ONCE(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 request_sock *req, int i)
1691 {
1692 long ttd = req->rsk_timer.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),
1715 sock_i_uid(req->rsk_listener)),
1716 0, /* non standard timer */
1717 0, /* open_requests have no inode */
1718 0, req);
1719 }
1720
1721 static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1722 {
1723 const struct in6_addr *dest, *src;
1724 __u16 destp, srcp;
1725 int timer_active;
1726 unsigned long timer_expires;
1727 const struct inet_sock *inet = inet_sk(sp);
1728 const struct tcp_sock *tp = tcp_sk(sp);
1729 const struct inet_connection_sock *icsk = inet_csk(sp);
1730 const struct fastopen_queue *fastopenq = &icsk->icsk_accept_queue.fastopenq;
1731 int rx_queue;
1732 int state;
1733
1734 dest = &sp->sk_v6_daddr;
1735 src = &sp->sk_v6_rcv_saddr;
1736 destp = ntohs(inet->inet_dport);
1737 srcp = ntohs(inet->inet_sport);
1738
1739 if (icsk->icsk_pending == ICSK_TIME_RETRANS ||
1740 icsk->icsk_pending == ICSK_TIME_EARLY_RETRANS ||
1741 icsk->icsk_pending == ICSK_TIME_LOSS_PROBE) {
1742 timer_active = 1;
1743 timer_expires = icsk->icsk_timeout;
1744 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
1745 timer_active = 4;
1746 timer_expires = icsk->icsk_timeout;
1747 } else if (timer_pending(&sp->sk_timer)) {
1748 timer_active = 2;
1749 timer_expires = sp->sk_timer.expires;
1750 } else {
1751 timer_active = 0;
1752 timer_expires = jiffies;
1753 }
1754
1755 state = sk_state_load(sp);
1756 if (state == TCP_LISTEN)
1757 rx_queue = sp->sk_ack_backlog;
1758 else
1759 /* Because we don't lock the socket,
1760 * we might find a transient negative value.
1761 */
1762 rx_queue = max_t(int, tp->rcv_nxt - tp->copied_seq, 0);
1763
1764 seq_printf(seq,
1765 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1766 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
1767 i,
1768 src->s6_addr32[0], src->s6_addr32[1],
1769 src->s6_addr32[2], src->s6_addr32[3], srcp,
1770 dest->s6_addr32[0], dest->s6_addr32[1],
1771 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1772 state,
1773 tp->write_seq - tp->snd_una,
1774 rx_queue,
1775 timer_active,
1776 jiffies_delta_to_clock_t(timer_expires - jiffies),
1777 icsk->icsk_retransmits,
1778 from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1779 icsk->icsk_probes_out,
1780 sock_i_ino(sp),
1781 atomic_read(&sp->sk_refcnt), sp,
1782 jiffies_to_clock_t(icsk->icsk_rto),
1783 jiffies_to_clock_t(icsk->icsk_ack.ato),
1784 (icsk->icsk_ack.quick << 1) | icsk->icsk_ack.pingpong,
1785 tp->snd_cwnd,
1786 state == TCP_LISTEN ?
1787 fastopenq->max_qlen :
1788 (tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh)
1789 );
1790 }
1791
1792 static void get_timewait6_sock(struct seq_file *seq,
1793 struct inet_timewait_sock *tw, int i)
1794 {
1795 long delta = tw->tw_timer.expires - jiffies;
1796 const struct in6_addr *dest, *src;
1797 __u16 destp, srcp;
1798
1799 dest = &tw->tw_v6_daddr;
1800 src = &tw->tw_v6_rcv_saddr;
1801 destp = ntohs(tw->tw_dport);
1802 srcp = ntohs(tw->tw_sport);
1803
1804 seq_printf(seq,
1805 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1806 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
1807 i,
1808 src->s6_addr32[0], src->s6_addr32[1],
1809 src->s6_addr32[2], src->s6_addr32[3], srcp,
1810 dest->s6_addr32[0], dest->s6_addr32[1],
1811 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1812 tw->tw_substate, 0, 0,
1813 3, jiffies_delta_to_clock_t(delta), 0, 0, 0, 0,
1814 atomic_read(&tw->tw_refcnt), tw);
1815 }
1816
1817 static int tcp6_seq_show(struct seq_file *seq, void *v)
1818 {
1819 struct tcp_iter_state *st;
1820 struct sock *sk = v;
1821
1822 if (v == SEQ_START_TOKEN) {
1823 seq_puts(seq,
1824 " sl "
1825 "local_address "
1826 "remote_address "
1827 "st tx_queue rx_queue tr tm->when retrnsmt"
1828 " uid timeout inode\n");
1829 goto out;
1830 }
1831 st = seq->private;
1832
1833 if (sk->sk_state == TCP_TIME_WAIT)
1834 get_timewait6_sock(seq, v, st->num);
1835 else if (sk->sk_state == TCP_NEW_SYN_RECV)
1836 get_openreq6(seq, v, st->num);
1837 else
1838 get_tcp6_sock(seq, v, st->num);
1839 out:
1840 return 0;
1841 }
1842
1843 static const struct file_operations tcp6_afinfo_seq_fops = {
1844 .owner = THIS_MODULE,
1845 .open = tcp_seq_open,
1846 .read = seq_read,
1847 .llseek = seq_lseek,
1848 .release = seq_release_net
1849 };
1850
1851 static struct tcp_seq_afinfo tcp6_seq_afinfo = {
1852 .name = "tcp6",
1853 .family = AF_INET6,
1854 .seq_fops = &tcp6_afinfo_seq_fops,
1855 .seq_ops = {
1856 .show = tcp6_seq_show,
1857 },
1858 };
1859
1860 int __net_init tcp6_proc_init(struct net *net)
1861 {
1862 return tcp_proc_register(net, &tcp6_seq_afinfo);
1863 }
1864
1865 void tcp6_proc_exit(struct net *net)
1866 {
1867 tcp_proc_unregister(net, &tcp6_seq_afinfo);
1868 }
1869 #endif
1870
1871 static void tcp_v6_clear_sk(struct sock *sk, int size)
1872 {
1873 struct inet_sock *inet = inet_sk(sk);
1874
1875 /* we do not want to clear pinet6 field, because of RCU lookups */
1876 sk_prot_clear_nulls(sk, offsetof(struct inet_sock, pinet6));
1877
1878 size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1879 memset(&inet->pinet6 + 1, 0, size);
1880 }
1881
1882 struct proto tcpv6_prot = {
1883 .name = "TCPv6",
1884 .owner = THIS_MODULE,
1885 .close = tcp_close,
1886 .connect = tcp_v6_connect,
1887 .disconnect = tcp_disconnect,
1888 .accept = inet_csk_accept,
1889 .ioctl = tcp_ioctl,
1890 .init = tcp_v6_init_sock,
1891 .destroy = tcp_v6_destroy_sock,
1892 .shutdown = tcp_shutdown,
1893 .setsockopt = tcp_setsockopt,
1894 .getsockopt = tcp_getsockopt,
1895 .recvmsg = tcp_recvmsg,
1896 .sendmsg = tcp_sendmsg,
1897 .sendpage = tcp_sendpage,
1898 .backlog_rcv = tcp_v6_do_rcv,
1899 .release_cb = tcp_release_cb,
1900 .hash = inet6_hash,
1901 .unhash = inet_unhash,
1902 .get_port = inet_csk_get_port,
1903 .enter_memory_pressure = tcp_enter_memory_pressure,
1904 .stream_memory_free = tcp_stream_memory_free,
1905 .sockets_allocated = &tcp_sockets_allocated,
1906 .memory_allocated = &tcp_memory_allocated,
1907 .memory_pressure = &tcp_memory_pressure,
1908 .orphan_count = &tcp_orphan_count,
1909 .sysctl_mem = sysctl_tcp_mem,
1910 .sysctl_wmem = sysctl_tcp_wmem,
1911 .sysctl_rmem = sysctl_tcp_rmem,
1912 .max_header = MAX_TCP_HEADER,
1913 .obj_size = sizeof(struct tcp6_sock),
1914 .slab_flags = SLAB_DESTROY_BY_RCU,
1915 .twsk_prot = &tcp6_timewait_sock_ops,
1916 .rsk_prot = &tcp6_request_sock_ops,
1917 .h.hashinfo = &tcp_hashinfo,
1918 .no_autobind = true,
1919 #ifdef CONFIG_COMPAT
1920 .compat_setsockopt = compat_tcp_setsockopt,
1921 .compat_getsockopt = compat_tcp_getsockopt,
1922 #endif
1923 .clear_sk = tcp_v6_clear_sk,
1924 .diag_destroy = tcp_abort,
1925 };
1926
1927 static const struct inet6_protocol tcpv6_protocol = {
1928 .early_demux = tcp_v6_early_demux,
1929 .handler = tcp_v6_rcv,
1930 .err_handler = tcp_v6_err,
1931 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1932 };
1933
1934 static struct inet_protosw tcpv6_protosw = {
1935 .type = SOCK_STREAM,
1936 .protocol = IPPROTO_TCP,
1937 .prot = &tcpv6_prot,
1938 .ops = &inet6_stream_ops,
1939 .flags = INET_PROTOSW_PERMANENT |
1940 INET_PROTOSW_ICSK,
1941 };
1942
1943 static int __net_init tcpv6_net_init(struct net *net)
1944 {
1945 return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
1946 SOCK_RAW, IPPROTO_TCP, net);
1947 }
1948
1949 static void __net_exit tcpv6_net_exit(struct net *net)
1950 {
1951 inet_ctl_sock_destroy(net->ipv6.tcp_sk);
1952 }
1953
1954 static void __net_exit tcpv6_net_exit_batch(struct list_head *net_exit_list)
1955 {
1956 inet_twsk_purge(&tcp_hashinfo, &tcp_death_row, AF_INET6);
1957 }
1958
1959 static struct pernet_operations tcpv6_net_ops = {
1960 .init = tcpv6_net_init,
1961 .exit = tcpv6_net_exit,
1962 .exit_batch = tcpv6_net_exit_batch,
1963 };
1964
1965 int __init tcpv6_init(void)
1966 {
1967 int ret;
1968
1969 ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
1970 if (ret)
1971 goto out;
1972
1973 /* register inet6 protocol */
1974 ret = inet6_register_protosw(&tcpv6_protosw);
1975 if (ret)
1976 goto out_tcpv6_protocol;
1977
1978 ret = register_pernet_subsys(&tcpv6_net_ops);
1979 if (ret)
1980 goto out_tcpv6_protosw;
1981 out:
1982 return ret;
1983
1984 out_tcpv6_protosw:
1985 inet6_unregister_protosw(&tcpv6_protosw);
1986 out_tcpv6_protocol:
1987 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
1988 goto out;
1989 }
1990
1991 void tcpv6_exit(void)
1992 {
1993 unregister_pernet_subsys(&tcpv6_net_ops);
1994 inet6_unregister_protosw(&tcpv6_protosw);
1995 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
1996 }
ubuntu-bionic-kernel mirror