]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - net/ipv6/tcp_ipv6.c
net: replace remaining __FUNCTION__ occurrences
[mirror_ubuntu-artful-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 * $Id: tcp_ipv6.c,v 1.144 2002/02/01 22:01:04 davem Exp $
9 *
10 * Based on:
11 * linux/net/ipv4/tcp.c
12 * linux/net/ipv4/tcp_input.c
13 * linux/net/ipv4/tcp_output.c
14 *
15 * Fixes:
16 * Hideaki YOSHIFUJI : sin6_scope_id support
17 * YOSHIFUJI Hideaki @USAGI and: Support IPV6_V6ONLY socket option, which
18 * Alexey Kuznetsov allow both IPv4 and IPv6 sockets to bind
19 * a single port at the same time.
20 * YOSHIFUJI Hideaki @USAGI: convert /proc/net/tcp6 to seq_file.
21 *
22 * This program is free software; you can redistribute it and/or
23 * modify it under the terms of the GNU General Public License
24 * as published by the Free Software Foundation; either version
25 * 2 of the License, or (at your option) any later version.
26 */
27
28 #include <linux/module.h>
29 #include <linux/errno.h>
30 #include <linux/types.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/net.h>
34 #include <linux/jiffies.h>
35 #include <linux/in.h>
36 #include <linux/in6.h>
37 #include <linux/netdevice.h>
38 #include <linux/init.h>
39 #include <linux/jhash.h>
40 #include <linux/ipsec.h>
41 #include <linux/times.h>
42
43 #include <linux/ipv6.h>
44 #include <linux/icmpv6.h>
45 #include <linux/random.h>
46
47 #include <net/tcp.h>
48 #include <net/ndisc.h>
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
64 #include <asm/uaccess.h>
65
66 #include <linux/proc_fs.h>
67 #include <linux/seq_file.h>
68
69 #include <linux/crypto.h>
70 #include <linux/scatterlist.h>
71
72 /* Socket used for sending RSTs and ACKs */
73 static struct socket *tcp6_socket;
74
75 static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb);
76 static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req);
77 static void tcp_v6_send_check(struct sock *sk, int len,
78 struct sk_buff *skb);
79
80 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
81
82 static struct inet_connection_sock_af_ops ipv6_mapped;
83 static struct inet_connection_sock_af_ops ipv6_specific;
84 #ifdef CONFIG_TCP_MD5SIG
85 static struct tcp_sock_af_ops tcp_sock_ipv6_specific;
86 static struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
87 #endif
88
89 static void tcp_v6_hash(struct sock *sk)
90 {
91 if (sk->sk_state != TCP_CLOSE) {
92 if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
93 tcp_prot.hash(sk);
94 return;
95 }
96 local_bh_disable();
97 __inet6_hash(sk);
98 local_bh_enable();
99 }
100 }
101
102 static __inline__ __sum16 tcp_v6_check(struct tcphdr *th, int len,
103 struct in6_addr *saddr,
104 struct in6_addr *daddr,
105 __wsum base)
106 {
107 return csum_ipv6_magic(saddr, daddr, len, IPPROTO_TCP, base);
108 }
109
110 static __u32 tcp_v6_init_sequence(struct sk_buff *skb)
111 {
112 return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
113 ipv6_hdr(skb)->saddr.s6_addr32,
114 tcp_hdr(skb)->dest,
115 tcp_hdr(skb)->source);
116 }
117
118 static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
119 int addr_len)
120 {
121 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
122 struct inet_sock *inet = inet_sk(sk);
123 struct inet_connection_sock *icsk = inet_csk(sk);
124 struct ipv6_pinfo *np = inet6_sk(sk);
125 struct tcp_sock *tp = tcp_sk(sk);
126 struct in6_addr *saddr = NULL, *final_p = NULL, final;
127 struct flowi fl;
128 struct dst_entry *dst;
129 int addr_type;
130 int err;
131
132 if (addr_len < SIN6_LEN_RFC2133)
133 return -EINVAL;
134
135 if (usin->sin6_family != AF_INET6)
136 return(-EAFNOSUPPORT);
137
138 memset(&fl, 0, sizeof(fl));
139
140 if (np->sndflow) {
141 fl.fl6_flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
142 IP6_ECN_flow_init(fl.fl6_flowlabel);
143 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
144 struct ip6_flowlabel *flowlabel;
145 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
146 if (flowlabel == NULL)
147 return -EINVAL;
148 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
149 fl6_sock_release(flowlabel);
150 }
151 }
152
153 /*
154 * connect() to INADDR_ANY means loopback (BSD'ism).
155 */
156
157 if(ipv6_addr_any(&usin->sin6_addr))
158 usin->sin6_addr.s6_addr[15] = 0x1;
159
160 addr_type = ipv6_addr_type(&usin->sin6_addr);
161
162 if(addr_type & IPV6_ADDR_MULTICAST)
163 return -ENETUNREACH;
164
165 if (addr_type&IPV6_ADDR_LINKLOCAL) {
166 if (addr_len >= sizeof(struct sockaddr_in6) &&
167 usin->sin6_scope_id) {
168 /* If interface is set while binding, indices
169 * must coincide.
170 */
171 if (sk->sk_bound_dev_if &&
172 sk->sk_bound_dev_if != usin->sin6_scope_id)
173 return -EINVAL;
174
175 sk->sk_bound_dev_if = usin->sin6_scope_id;
176 }
177
178 /* Connect to link-local address requires an interface */
179 if (!sk->sk_bound_dev_if)
180 return -EINVAL;
181 }
182
183 if (tp->rx_opt.ts_recent_stamp &&
184 !ipv6_addr_equal(&np->daddr, &usin->sin6_addr)) {
185 tp->rx_opt.ts_recent = 0;
186 tp->rx_opt.ts_recent_stamp = 0;
187 tp->write_seq = 0;
188 }
189
190 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
191 np->flow_label = fl.fl6_flowlabel;
192
193 /*
194 * TCP over IPv4
195 */
196
197 if (addr_type == IPV6_ADDR_MAPPED) {
198 u32 exthdrlen = icsk->icsk_ext_hdr_len;
199 struct sockaddr_in sin;
200
201 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
202
203 if (__ipv6_only_sock(sk))
204 return -ENETUNREACH;
205
206 sin.sin_family = AF_INET;
207 sin.sin_port = usin->sin6_port;
208 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
209
210 icsk->icsk_af_ops = &ipv6_mapped;
211 sk->sk_backlog_rcv = tcp_v4_do_rcv;
212 #ifdef CONFIG_TCP_MD5SIG
213 tp->af_specific = &tcp_sock_ipv6_mapped_specific;
214 #endif
215
216 err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
217
218 if (err) {
219 icsk->icsk_ext_hdr_len = exthdrlen;
220 icsk->icsk_af_ops = &ipv6_specific;
221 sk->sk_backlog_rcv = tcp_v6_do_rcv;
222 #ifdef CONFIG_TCP_MD5SIG
223 tp->af_specific = &tcp_sock_ipv6_specific;
224 #endif
225 goto failure;
226 } else {
227 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
228 inet->saddr);
229 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
230 inet->rcv_saddr);
231 }
232
233 return err;
234 }
235
236 if (!ipv6_addr_any(&np->rcv_saddr))
237 saddr = &np->rcv_saddr;
238
239 fl.proto = IPPROTO_TCP;
240 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
241 ipv6_addr_copy(&fl.fl6_src,
242 (saddr ? saddr : &np->saddr));
243 fl.oif = sk->sk_bound_dev_if;
244 fl.fl_ip_dport = usin->sin6_port;
245 fl.fl_ip_sport = inet->sport;
246
247 if (np->opt && np->opt->srcrt) {
248 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
249 ipv6_addr_copy(&final, &fl.fl6_dst);
250 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
251 final_p = &final;
252 }
253
254 security_sk_classify_flow(sk, &fl);
255
256 err = ip6_dst_lookup(sk, &dst, &fl);
257 if (err)
258 goto failure;
259 if (final_p)
260 ipv6_addr_copy(&fl.fl6_dst, final_p);
261
262 if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 0) {
263 if (err == -EREMOTE)
264 err = ip6_dst_blackhole(sk, &dst, &fl);
265 if (err < 0)
266 goto failure;
267 }
268
269 if (saddr == NULL) {
270 saddr = &fl.fl6_src;
271 ipv6_addr_copy(&np->rcv_saddr, saddr);
272 }
273
274 /* set the source address */
275 ipv6_addr_copy(&np->saddr, saddr);
276 inet->rcv_saddr = LOOPBACK4_IPV6;
277
278 sk->sk_gso_type = SKB_GSO_TCPV6;
279 __ip6_dst_store(sk, dst, NULL, NULL);
280
281 icsk->icsk_ext_hdr_len = 0;
282 if (np->opt)
283 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
284 np->opt->opt_nflen);
285
286 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
287
288 inet->dport = usin->sin6_port;
289
290 tcp_set_state(sk, TCP_SYN_SENT);
291 err = inet6_hash_connect(&tcp_death_row, sk);
292 if (err)
293 goto late_failure;
294
295 if (!tp->write_seq)
296 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
297 np->daddr.s6_addr32,
298 inet->sport,
299 inet->dport);
300
301 err = tcp_connect(sk);
302 if (err)
303 goto late_failure;
304
305 return 0;
306
307 late_failure:
308 tcp_set_state(sk, TCP_CLOSE);
309 __sk_dst_reset(sk);
310 failure:
311 inet->dport = 0;
312 sk->sk_route_caps = 0;
313 return err;
314 }
315
316 static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
317 int type, int code, int offset, __be32 info)
318 {
319 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
320 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
321 struct ipv6_pinfo *np;
322 struct sock *sk;
323 int err;
324 struct tcp_sock *tp;
325 __u32 seq;
326
327 sk = inet6_lookup(skb->dev->nd_net, &tcp_hashinfo, &hdr->daddr,
328 th->dest, &hdr->saddr, th->source, skb->dev->ifindex);
329
330 if (sk == NULL) {
331 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS);
332 return;
333 }
334
335 if (sk->sk_state == TCP_TIME_WAIT) {
336 inet_twsk_put(inet_twsk(sk));
337 return;
338 }
339
340 bh_lock_sock(sk);
341 if (sock_owned_by_user(sk))
342 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS);
343
344 if (sk->sk_state == TCP_CLOSE)
345 goto out;
346
347 tp = tcp_sk(sk);
348 seq = ntohl(th->seq);
349 if (sk->sk_state != TCP_LISTEN &&
350 !between(seq, tp->snd_una, tp->snd_nxt)) {
351 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
352 goto out;
353 }
354
355 np = inet6_sk(sk);
356
357 if (type == ICMPV6_PKT_TOOBIG) {
358 struct dst_entry *dst = NULL;
359
360 if (sock_owned_by_user(sk))
361 goto out;
362 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
363 goto out;
364
365 /* icmp should have updated the destination cache entry */
366 dst = __sk_dst_check(sk, np->dst_cookie);
367
368 if (dst == NULL) {
369 struct inet_sock *inet = inet_sk(sk);
370 struct flowi fl;
371
372 /* BUGGG_FUTURE: Again, it is not clear how
373 to handle rthdr case. Ignore this complexity
374 for now.
375 */
376 memset(&fl, 0, sizeof(fl));
377 fl.proto = IPPROTO_TCP;
378 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
379 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
380 fl.oif = sk->sk_bound_dev_if;
381 fl.fl_ip_dport = inet->dport;
382 fl.fl_ip_sport = inet->sport;
383 security_skb_classify_flow(skb, &fl);
384
385 if ((err = ip6_dst_lookup(sk, &dst, &fl))) {
386 sk->sk_err_soft = -err;
387 goto out;
388 }
389
390 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
391 sk->sk_err_soft = -err;
392 goto out;
393 }
394
395 } else
396 dst_hold(dst);
397
398 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
399 tcp_sync_mss(sk, dst_mtu(dst));
400 tcp_simple_retransmit(sk);
401 } /* else let the usual retransmit timer handle it */
402 dst_release(dst);
403 goto out;
404 }
405
406 icmpv6_err_convert(type, code, &err);
407
408 /* Might be for an request_sock */
409 switch (sk->sk_state) {
410 struct request_sock *req, **prev;
411 case TCP_LISTEN:
412 if (sock_owned_by_user(sk))
413 goto out;
414
415 req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
416 &hdr->saddr, inet6_iif(skb));
417 if (!req)
418 goto out;
419
420 /* ICMPs are not backlogged, hence we cannot get
421 * an established socket here.
422 */
423 BUG_TRAP(req->sk == NULL);
424
425 if (seq != tcp_rsk(req)->snt_isn) {
426 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS);
427 goto out;
428 }
429
430 inet_csk_reqsk_queue_drop(sk, req, prev);
431 goto out;
432
433 case TCP_SYN_SENT:
434 case TCP_SYN_RECV: /* Cannot happen.
435 It can, it SYNs are crossed. --ANK */
436 if (!sock_owned_by_user(sk)) {
437 sk->sk_err = err;
438 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
439
440 tcp_done(sk);
441 } else
442 sk->sk_err_soft = err;
443 goto out;
444 }
445
446 if (!sock_owned_by_user(sk) && np->recverr) {
447 sk->sk_err = err;
448 sk->sk_error_report(sk);
449 } else
450 sk->sk_err_soft = err;
451
452 out:
453 bh_unlock_sock(sk);
454 sock_put(sk);
455 }
456
457
458 static int tcp_v6_send_synack(struct sock *sk, struct request_sock *req)
459 {
460 struct inet6_request_sock *treq = inet6_rsk(req);
461 struct ipv6_pinfo *np = inet6_sk(sk);
462 struct sk_buff * skb;
463 struct ipv6_txoptions *opt = NULL;
464 struct in6_addr * final_p = NULL, final;
465 struct flowi fl;
466 struct dst_entry *dst;
467 int err = -1;
468
469 memset(&fl, 0, sizeof(fl));
470 fl.proto = IPPROTO_TCP;
471 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
472 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
473 fl.fl6_flowlabel = 0;
474 fl.oif = treq->iif;
475 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
476 fl.fl_ip_sport = inet_sk(sk)->sport;
477 security_req_classify_flow(req, &fl);
478
479 opt = np->opt;
480 if (opt && opt->srcrt) {
481 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
482 ipv6_addr_copy(&final, &fl.fl6_dst);
483 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
484 final_p = &final;
485 }
486
487 err = ip6_dst_lookup(sk, &dst, &fl);
488 if (err)
489 goto done;
490 if (final_p)
491 ipv6_addr_copy(&fl.fl6_dst, final_p);
492 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
493 goto done;
494
495 skb = tcp_make_synack(sk, dst, req);
496 if (skb) {
497 struct tcphdr *th = tcp_hdr(skb);
498
499 th->check = tcp_v6_check(th, skb->len,
500 &treq->loc_addr, &treq->rmt_addr,
501 csum_partial((char *)th, skb->len, skb->csum));
502
503 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
504 err = ip6_xmit(sk, skb, &fl, opt, 0);
505 err = net_xmit_eval(err);
506 }
507
508 done:
509 if (opt && opt != np->opt)
510 sock_kfree_s(sk, opt, opt->tot_len);
511 dst_release(dst);
512 return err;
513 }
514
515 static inline void syn_flood_warning(struct sk_buff *skb)
516 {
517 #ifdef CONFIG_SYN_COOKIES
518 if (sysctl_tcp_syncookies)
519 printk(KERN_INFO
520 "TCPv6: Possible SYN flooding on port %d. "
521 "Sending cookies.\n", ntohs(tcp_hdr(skb)->dest));
522 else
523 #endif
524 printk(KERN_INFO
525 "TCPv6: Possible SYN flooding on port %d. "
526 "Dropping request.\n", ntohs(tcp_hdr(skb)->dest));
527 }
528
529 static void tcp_v6_reqsk_destructor(struct request_sock *req)
530 {
531 if (inet6_rsk(req)->pktopts)
532 kfree_skb(inet6_rsk(req)->pktopts);
533 }
534
535 #ifdef CONFIG_TCP_MD5SIG
536 static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
537 struct in6_addr *addr)
538 {
539 struct tcp_sock *tp = tcp_sk(sk);
540 int i;
541
542 BUG_ON(tp == NULL);
543
544 if (!tp->md5sig_info || !tp->md5sig_info->entries6)
545 return NULL;
546
547 for (i = 0; i < tp->md5sig_info->entries6; i++) {
548 if (ipv6_addr_cmp(&tp->md5sig_info->keys6[i].addr, addr) == 0)
549 return &tp->md5sig_info->keys6[i].base;
550 }
551 return NULL;
552 }
553
554 static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
555 struct sock *addr_sk)
556 {
557 return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr);
558 }
559
560 static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
561 struct request_sock *req)
562 {
563 return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr);
564 }
565
566 static int tcp_v6_md5_do_add(struct sock *sk, struct in6_addr *peer,
567 char *newkey, u8 newkeylen)
568 {
569 /* Add key to the list */
570 struct tcp_md5sig_key *key;
571 struct tcp_sock *tp = tcp_sk(sk);
572 struct tcp6_md5sig_key *keys;
573
574 key = tcp_v6_md5_do_lookup(sk, peer);
575 if (key) {
576 /* modify existing entry - just update that one */
577 kfree(key->key);
578 key->key = newkey;
579 key->keylen = newkeylen;
580 } else {
581 /* reallocate new list if current one is full. */
582 if (!tp->md5sig_info) {
583 tp->md5sig_info = kzalloc(sizeof(*tp->md5sig_info), GFP_ATOMIC);
584 if (!tp->md5sig_info) {
585 kfree(newkey);
586 return -ENOMEM;
587 }
588 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
589 }
590 if (tcp_alloc_md5sig_pool() == NULL) {
591 kfree(newkey);
592 return -ENOMEM;
593 }
594 if (tp->md5sig_info->alloced6 == tp->md5sig_info->entries6) {
595 keys = kmalloc((sizeof (tp->md5sig_info->keys6[0]) *
596 (tp->md5sig_info->entries6 + 1)), GFP_ATOMIC);
597
598 if (!keys) {
599 tcp_free_md5sig_pool();
600 kfree(newkey);
601 return -ENOMEM;
602 }
603
604 if (tp->md5sig_info->entries6)
605 memmove(keys, tp->md5sig_info->keys6,
606 (sizeof (tp->md5sig_info->keys6[0]) *
607 tp->md5sig_info->entries6));
608
609 kfree(tp->md5sig_info->keys6);
610 tp->md5sig_info->keys6 = keys;
611 tp->md5sig_info->alloced6++;
612 }
613
614 ipv6_addr_copy(&tp->md5sig_info->keys6[tp->md5sig_info->entries6].addr,
615 peer);
616 tp->md5sig_info->keys6[tp->md5sig_info->entries6].base.key = newkey;
617 tp->md5sig_info->keys6[tp->md5sig_info->entries6].base.keylen = newkeylen;
618
619 tp->md5sig_info->entries6++;
620 }
621 return 0;
622 }
623
624 static int tcp_v6_md5_add_func(struct sock *sk, struct sock *addr_sk,
625 u8 *newkey, __u8 newkeylen)
626 {
627 return tcp_v6_md5_do_add(sk, &inet6_sk(addr_sk)->daddr,
628 newkey, newkeylen);
629 }
630
631 static int tcp_v6_md5_do_del(struct sock *sk, struct in6_addr *peer)
632 {
633 struct tcp_sock *tp = tcp_sk(sk);
634 int i;
635
636 for (i = 0; i < tp->md5sig_info->entries6; i++) {
637 if (ipv6_addr_cmp(&tp->md5sig_info->keys6[i].addr, peer) == 0) {
638 /* Free the key */
639 kfree(tp->md5sig_info->keys6[i].base.key);
640 tp->md5sig_info->entries6--;
641
642 if (tp->md5sig_info->entries6 == 0) {
643 kfree(tp->md5sig_info->keys6);
644 tp->md5sig_info->keys6 = NULL;
645 tp->md5sig_info->alloced6 = 0;
646 } else {
647 /* shrink the database */
648 if (tp->md5sig_info->entries6 != i)
649 memmove(&tp->md5sig_info->keys6[i],
650 &tp->md5sig_info->keys6[i+1],
651 (tp->md5sig_info->entries6 - i)
652 * sizeof (tp->md5sig_info->keys6[0]));
653 }
654 tcp_free_md5sig_pool();
655 return 0;
656 }
657 }
658 return -ENOENT;
659 }
660
661 static void tcp_v6_clear_md5_list (struct sock *sk)
662 {
663 struct tcp_sock *tp = tcp_sk(sk);
664 int i;
665
666 if (tp->md5sig_info->entries6) {
667 for (i = 0; i < tp->md5sig_info->entries6; i++)
668 kfree(tp->md5sig_info->keys6[i].base.key);
669 tp->md5sig_info->entries6 = 0;
670 tcp_free_md5sig_pool();
671 }
672
673 kfree(tp->md5sig_info->keys6);
674 tp->md5sig_info->keys6 = NULL;
675 tp->md5sig_info->alloced6 = 0;
676
677 if (tp->md5sig_info->entries4) {
678 for (i = 0; i < tp->md5sig_info->entries4; i++)
679 kfree(tp->md5sig_info->keys4[i].base.key);
680 tp->md5sig_info->entries4 = 0;
681 tcp_free_md5sig_pool();
682 }
683
684 kfree(tp->md5sig_info->keys4);
685 tp->md5sig_info->keys4 = NULL;
686 tp->md5sig_info->alloced4 = 0;
687 }
688
689 static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval,
690 int optlen)
691 {
692 struct tcp_md5sig cmd;
693 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
694 u8 *newkey;
695
696 if (optlen < sizeof(cmd))
697 return -EINVAL;
698
699 if (copy_from_user(&cmd, optval, sizeof(cmd)))
700 return -EFAULT;
701
702 if (sin6->sin6_family != AF_INET6)
703 return -EINVAL;
704
705 if (!cmd.tcpm_keylen) {
706 if (!tcp_sk(sk)->md5sig_info)
707 return -ENOENT;
708 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
709 return tcp_v4_md5_do_del(sk, sin6->sin6_addr.s6_addr32[3]);
710 return tcp_v6_md5_do_del(sk, &sin6->sin6_addr);
711 }
712
713 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
714 return -EINVAL;
715
716 if (!tcp_sk(sk)->md5sig_info) {
717 struct tcp_sock *tp = tcp_sk(sk);
718 struct tcp_md5sig_info *p;
719
720 p = kzalloc(sizeof(struct tcp_md5sig_info), GFP_KERNEL);
721 if (!p)
722 return -ENOMEM;
723
724 tp->md5sig_info = p;
725 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
726 }
727
728 newkey = kmemdup(cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
729 if (!newkey)
730 return -ENOMEM;
731 if (ipv6_addr_v4mapped(&sin6->sin6_addr)) {
732 return tcp_v4_md5_do_add(sk, sin6->sin6_addr.s6_addr32[3],
733 newkey, cmd.tcpm_keylen);
734 }
735 return tcp_v6_md5_do_add(sk, &sin6->sin6_addr, newkey, cmd.tcpm_keylen);
736 }
737
738 static int tcp_v6_do_calc_md5_hash(char *md5_hash, struct tcp_md5sig_key *key,
739 struct in6_addr *saddr,
740 struct in6_addr *daddr,
741 struct tcphdr *th, int protocol,
742 unsigned int tcplen)
743 {
744 struct scatterlist sg[4];
745 __u16 data_len;
746 int block = 0;
747 __sum16 cksum;
748 struct tcp_md5sig_pool *hp;
749 struct tcp6_pseudohdr *bp;
750 struct hash_desc *desc;
751 int err;
752 unsigned int nbytes = 0;
753
754 hp = tcp_get_md5sig_pool();
755 if (!hp) {
756 printk(KERN_WARNING "%s(): hash pool not found...\n", __func__);
757 goto clear_hash_noput;
758 }
759 bp = &hp->md5_blk.ip6;
760 desc = &hp->md5_desc;
761
762 /* 1. TCP pseudo-header (RFC2460) */
763 ipv6_addr_copy(&bp->saddr, saddr);
764 ipv6_addr_copy(&bp->daddr, daddr);
765 bp->len = htonl(tcplen);
766 bp->protocol = htonl(protocol);
767
768 sg_init_table(sg, 4);
769
770 sg_set_buf(&sg[block++], bp, sizeof(*bp));
771 nbytes += sizeof(*bp);
772
773 /* 2. TCP header, excluding options */
774 cksum = th->check;
775 th->check = 0;
776 sg_set_buf(&sg[block++], th, sizeof(*th));
777 nbytes += sizeof(*th);
778
779 /* 3. TCP segment data (if any) */
780 data_len = tcplen - (th->doff << 2);
781 if (data_len > 0) {
782 u8 *data = (u8 *)th + (th->doff << 2);
783 sg_set_buf(&sg[block++], data, data_len);
784 nbytes += data_len;
785 }
786
787 /* 4. shared key */
788 sg_set_buf(&sg[block++], key->key, key->keylen);
789 nbytes += key->keylen;
790
791 sg_mark_end(&sg[block - 1]);
792
793 /* Now store the hash into the packet */
794 err = crypto_hash_init(desc);
795 if (err) {
796 printk(KERN_WARNING "%s(): hash_init failed\n", __func__);
797 goto clear_hash;
798 }
799 err = crypto_hash_update(desc, sg, nbytes);
800 if (err) {
801 printk(KERN_WARNING "%s(): hash_update failed\n", __func__);
802 goto clear_hash;
803 }
804 err = crypto_hash_final(desc, md5_hash);
805 if (err) {
806 printk(KERN_WARNING "%s(): hash_final failed\n", __func__);
807 goto clear_hash;
808 }
809
810 /* Reset header, and free up the crypto */
811 tcp_put_md5sig_pool();
812 th->check = cksum;
813 out:
814 return 0;
815 clear_hash:
816 tcp_put_md5sig_pool();
817 clear_hash_noput:
818 memset(md5_hash, 0, 16);
819 goto out;
820 }
821
822 static int tcp_v6_calc_md5_hash(char *md5_hash, struct tcp_md5sig_key *key,
823 struct sock *sk,
824 struct dst_entry *dst,
825 struct request_sock *req,
826 struct tcphdr *th, int protocol,
827 unsigned int tcplen)
828 {
829 struct in6_addr *saddr, *daddr;
830
831 if (sk) {
832 saddr = &inet6_sk(sk)->saddr;
833 daddr = &inet6_sk(sk)->daddr;
834 } else {
835 saddr = &inet6_rsk(req)->loc_addr;
836 daddr = &inet6_rsk(req)->rmt_addr;
837 }
838 return tcp_v6_do_calc_md5_hash(md5_hash, key,
839 saddr, daddr,
840 th, protocol, tcplen);
841 }
842
843 static int tcp_v6_inbound_md5_hash (struct sock *sk, struct sk_buff *skb)
844 {
845 __u8 *hash_location = NULL;
846 struct tcp_md5sig_key *hash_expected;
847 struct ipv6hdr *ip6h = ipv6_hdr(skb);
848 struct tcphdr *th = tcp_hdr(skb);
849 int length = (th->doff << 2) - sizeof (*th);
850 int genhash;
851 u8 *ptr;
852 u8 newhash[16];
853
854 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
855
856 /* If the TCP option is too short, we can short cut */
857 if (length < TCPOLEN_MD5SIG)
858 return hash_expected ? 1 : 0;
859
860 /* parse options */
861 ptr = (u8*)(th + 1);
862 while (length > 0) {
863 int opcode = *ptr++;
864 int opsize;
865
866 switch(opcode) {
867 case TCPOPT_EOL:
868 goto done_opts;
869 case TCPOPT_NOP:
870 length--;
871 continue;
872 default:
873 opsize = *ptr++;
874 if (opsize < 2 || opsize > length)
875 goto done_opts;
876 if (opcode == TCPOPT_MD5SIG) {
877 hash_location = ptr;
878 goto done_opts;
879 }
880 }
881 ptr += opsize - 2;
882 length -= opsize;
883 }
884
885 done_opts:
886 /* do we have a hash as expected? */
887 if (!hash_expected) {
888 if (!hash_location)
889 return 0;
890 if (net_ratelimit()) {
891 printk(KERN_INFO "MD5 Hash NOT expected but found "
892 "(" NIP6_FMT ", %u)->"
893 "(" NIP6_FMT ", %u)\n",
894 NIP6(ip6h->saddr), ntohs(th->source),
895 NIP6(ip6h->daddr), ntohs(th->dest));
896 }
897 return 1;
898 }
899
900 if (!hash_location) {
901 if (net_ratelimit()) {
902 printk(KERN_INFO "MD5 Hash expected but NOT found "
903 "(" NIP6_FMT ", %u)->"
904 "(" NIP6_FMT ", %u)\n",
905 NIP6(ip6h->saddr), ntohs(th->source),
906 NIP6(ip6h->daddr), ntohs(th->dest));
907 }
908 return 1;
909 }
910
911 /* check the signature */
912 genhash = tcp_v6_do_calc_md5_hash(newhash,
913 hash_expected,
914 &ip6h->saddr, &ip6h->daddr,
915 th, sk->sk_protocol,
916 skb->len);
917 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
918 if (net_ratelimit()) {
919 printk(KERN_INFO "MD5 Hash %s for "
920 "(" NIP6_FMT ", %u)->"
921 "(" NIP6_FMT ", %u)\n",
922 genhash ? "failed" : "mismatch",
923 NIP6(ip6h->saddr), ntohs(th->source),
924 NIP6(ip6h->daddr), ntohs(th->dest));
925 }
926 return 1;
927 }
928 return 0;
929 }
930 #endif
931
932 struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
933 .family = AF_INET6,
934 .obj_size = sizeof(struct tcp6_request_sock),
935 .rtx_syn_ack = tcp_v6_send_synack,
936 .send_ack = tcp_v6_reqsk_send_ack,
937 .destructor = tcp_v6_reqsk_destructor,
938 .send_reset = tcp_v6_send_reset
939 };
940
941 #ifdef CONFIG_TCP_MD5SIG
942 static struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
943 .md5_lookup = tcp_v6_reqsk_md5_lookup,
944 };
945 #endif
946
947 static struct timewait_sock_ops tcp6_timewait_sock_ops = {
948 .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
949 .twsk_unique = tcp_twsk_unique,
950 .twsk_destructor= tcp_twsk_destructor,
951 };
952
953 static void tcp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
954 {
955 struct ipv6_pinfo *np = inet6_sk(sk);
956 struct tcphdr *th = tcp_hdr(skb);
957
958 if (skb->ip_summed == CHECKSUM_PARTIAL) {
959 th->check = ~csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP, 0);
960 skb->csum_start = skb_transport_header(skb) - skb->head;
961 skb->csum_offset = offsetof(struct tcphdr, check);
962 } else {
963 th->check = csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP,
964 csum_partial((char *)th, th->doff<<2,
965 skb->csum));
966 }
967 }
968
969 static int tcp_v6_gso_send_check(struct sk_buff *skb)
970 {
971 struct ipv6hdr *ipv6h;
972 struct tcphdr *th;
973
974 if (!pskb_may_pull(skb, sizeof(*th)))
975 return -EINVAL;
976
977 ipv6h = ipv6_hdr(skb);
978 th = tcp_hdr(skb);
979
980 th->check = 0;
981 th->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
982 IPPROTO_TCP, 0);
983 skb->csum_start = skb_transport_header(skb) - skb->head;
984 skb->csum_offset = offsetof(struct tcphdr, check);
985 skb->ip_summed = CHECKSUM_PARTIAL;
986 return 0;
987 }
988
989 static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
990 {
991 struct tcphdr *th = tcp_hdr(skb), *t1;
992 struct sk_buff *buff;
993 struct flowi fl;
994 unsigned int tot_len = sizeof(*th);
995 #ifdef CONFIG_TCP_MD5SIG
996 struct tcp_md5sig_key *key;
997 #endif
998
999 if (th->rst)
1000 return;
1001
1002 if (!ipv6_unicast_destination(skb))
1003 return;
1004
1005 #ifdef CONFIG_TCP_MD5SIG
1006 if (sk)
1007 key = tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr);
1008 else
1009 key = NULL;
1010
1011 if (key)
1012 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1013 #endif
1014
1015 /*
1016 * We need to grab some memory, and put together an RST,
1017 * and then put it into the queue to be sent.
1018 */
1019
1020 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1021 GFP_ATOMIC);
1022 if (buff == NULL)
1023 return;
1024
1025 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1026
1027 t1 = (struct tcphdr *) skb_push(buff, tot_len);
1028
1029 /* Swap the send and the receive. */
1030 memset(t1, 0, sizeof(*t1));
1031 t1->dest = th->source;
1032 t1->source = th->dest;
1033 t1->doff = tot_len / 4;
1034 t1->rst = 1;
1035
1036 if(th->ack) {
1037 t1->seq = th->ack_seq;
1038 } else {
1039 t1->ack = 1;
1040 t1->ack_seq = htonl(ntohl(th->seq) + th->syn + th->fin
1041 + skb->len - (th->doff<<2));
1042 }
1043
1044 #ifdef CONFIG_TCP_MD5SIG
1045 if (key) {
1046 __be32 *opt = (__be32*)(t1 + 1);
1047 opt[0] = htonl((TCPOPT_NOP << 24) |
1048 (TCPOPT_NOP << 16) |
1049 (TCPOPT_MD5SIG << 8) |
1050 TCPOLEN_MD5SIG);
1051 tcp_v6_do_calc_md5_hash((__u8 *)&opt[1], key,
1052 &ipv6_hdr(skb)->daddr,
1053 &ipv6_hdr(skb)->saddr,
1054 t1, IPPROTO_TCP, tot_len);
1055 }
1056 #endif
1057
1058 buff->csum = csum_partial((char *)t1, sizeof(*t1), 0);
1059
1060 memset(&fl, 0, sizeof(fl));
1061 ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
1062 ipv6_addr_copy(&fl.fl6_src, &ipv6_hdr(skb)->daddr);
1063
1064 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1065 sizeof(*t1), IPPROTO_TCP,
1066 buff->csum);
1067
1068 fl.proto = IPPROTO_TCP;
1069 fl.oif = inet6_iif(skb);
1070 fl.fl_ip_dport = t1->dest;
1071 fl.fl_ip_sport = t1->source;
1072 security_skb_classify_flow(skb, &fl);
1073
1074 /* Pass a socket to ip6_dst_lookup either it is for RST
1075 * Underlying function will use this to retrieve the network
1076 * namespace
1077 */
1078 if (!ip6_dst_lookup(tcp6_socket->sk, &buff->dst, &fl)) {
1079
1080 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
1081 ip6_xmit(tcp6_socket->sk, buff, &fl, NULL, 0);
1082 TCP_INC_STATS_BH(TCP_MIB_OUTSEGS);
1083 TCP_INC_STATS_BH(TCP_MIB_OUTRSTS);
1084 return;
1085 }
1086 }
1087
1088 kfree_skb(buff);
1089 }
1090
1091 static void tcp_v6_send_ack(struct tcp_timewait_sock *tw,
1092 struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts)
1093 {
1094 struct tcphdr *th = tcp_hdr(skb), *t1;
1095 struct sk_buff *buff;
1096 struct flowi fl;
1097 unsigned int tot_len = sizeof(struct tcphdr);
1098 __be32 *topt;
1099 #ifdef CONFIG_TCP_MD5SIG
1100 struct tcp_md5sig_key *key;
1101 struct tcp_md5sig_key tw_key;
1102 #endif
1103
1104 #ifdef CONFIG_TCP_MD5SIG
1105 if (!tw && skb->sk) {
1106 key = tcp_v6_md5_do_lookup(skb->sk, &ipv6_hdr(skb)->daddr);
1107 } else if (tw && tw->tw_md5_keylen) {
1108 tw_key.key = tw->tw_md5_key;
1109 tw_key.keylen = tw->tw_md5_keylen;
1110 key = &tw_key;
1111 } else {
1112 key = NULL;
1113 }
1114 #endif
1115
1116 if (ts)
1117 tot_len += TCPOLEN_TSTAMP_ALIGNED;
1118 #ifdef CONFIG_TCP_MD5SIG
1119 if (key)
1120 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1121 #endif
1122
1123 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1124 GFP_ATOMIC);
1125 if (buff == NULL)
1126 return;
1127
1128 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1129
1130 t1 = (struct tcphdr *) skb_push(buff,tot_len);
1131
1132 /* Swap the send and the receive. */
1133 memset(t1, 0, sizeof(*t1));
1134 t1->dest = th->source;
1135 t1->source = th->dest;
1136 t1->doff = tot_len/4;
1137 t1->seq = htonl(seq);
1138 t1->ack_seq = htonl(ack);
1139 t1->ack = 1;
1140 t1->window = htons(win);
1141
1142 topt = (__be32 *)(t1 + 1);
1143
1144 if (ts) {
1145 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1146 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
1147 *topt++ = htonl(tcp_time_stamp);
1148 *topt = htonl(ts);
1149 }
1150
1151 #ifdef CONFIG_TCP_MD5SIG
1152 if (key) {
1153 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1154 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
1155 tcp_v6_do_calc_md5_hash((__u8 *)topt, key,
1156 &ipv6_hdr(skb)->daddr,
1157 &ipv6_hdr(skb)->saddr,
1158 t1, IPPROTO_TCP, tot_len);
1159 }
1160 #endif
1161
1162 buff->csum = csum_partial((char *)t1, tot_len, 0);
1163
1164 memset(&fl, 0, sizeof(fl));
1165 ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
1166 ipv6_addr_copy(&fl.fl6_src, &ipv6_hdr(skb)->daddr);
1167
1168 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1169 tot_len, IPPROTO_TCP,
1170 buff->csum);
1171
1172 fl.proto = IPPROTO_TCP;
1173 fl.oif = inet6_iif(skb);
1174 fl.fl_ip_dport = t1->dest;
1175 fl.fl_ip_sport = t1->source;
1176 security_skb_classify_flow(skb, &fl);
1177
1178 if (!ip6_dst_lookup(tcp6_socket->sk, &buff->dst, &fl)) {
1179 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
1180 ip6_xmit(tcp6_socket->sk, buff, &fl, NULL, 0);
1181 TCP_INC_STATS_BH(TCP_MIB_OUTSEGS);
1182 return;
1183 }
1184 }
1185
1186 kfree_skb(buff);
1187 }
1188
1189 static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
1190 {
1191 struct inet_timewait_sock *tw = inet_twsk(sk);
1192 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
1193
1194 tcp_v6_send_ack(tcptw, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
1195 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
1196 tcptw->tw_ts_recent);
1197
1198 inet_twsk_put(tw);
1199 }
1200
1201 static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req)
1202 {
1203 tcp_v6_send_ack(NULL, skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, req->ts_recent);
1204 }
1205
1206
1207 static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
1208 {
1209 struct request_sock *req, **prev;
1210 const struct tcphdr *th = tcp_hdr(skb);
1211 struct sock *nsk;
1212
1213 /* Find possible connection requests. */
1214 req = inet6_csk_search_req(sk, &prev, th->source,
1215 &ipv6_hdr(skb)->saddr,
1216 &ipv6_hdr(skb)->daddr, inet6_iif(skb));
1217 if (req)
1218 return tcp_check_req(sk, skb, req, prev);
1219
1220 nsk = __inet6_lookup_established(sk->sk_net, &tcp_hashinfo,
1221 &ipv6_hdr(skb)->saddr, th->source,
1222 &ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb));
1223
1224 if (nsk) {
1225 if (nsk->sk_state != TCP_TIME_WAIT) {
1226 bh_lock_sock(nsk);
1227 return nsk;
1228 }
1229 inet_twsk_put(inet_twsk(nsk));
1230 return NULL;
1231 }
1232
1233 #ifdef CONFIG_SYN_COOKIES
1234 if (!th->rst && !th->syn && th->ack)
1235 sk = cookie_v6_check(sk, skb);
1236 #endif
1237 return sk;
1238 }
1239
1240 /* FIXME: this is substantially similar to the ipv4 code.
1241 * Can some kind of merge be done? -- erics
1242 */
1243 static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
1244 {
1245 struct inet6_request_sock *treq;
1246 struct ipv6_pinfo *np = inet6_sk(sk);
1247 struct tcp_options_received tmp_opt;
1248 struct tcp_sock *tp = tcp_sk(sk);
1249 struct request_sock *req = NULL;
1250 __u32 isn = TCP_SKB_CB(skb)->when;
1251 #ifdef CONFIG_SYN_COOKIES
1252 int want_cookie = 0;
1253 #else
1254 #define want_cookie 0
1255 #endif
1256
1257 if (skb->protocol == htons(ETH_P_IP))
1258 return tcp_v4_conn_request(sk, skb);
1259
1260 if (!ipv6_unicast_destination(skb))
1261 goto drop;
1262
1263 if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
1264 if (net_ratelimit())
1265 syn_flood_warning(skb);
1266 #ifdef CONFIG_SYN_COOKIES
1267 if (sysctl_tcp_syncookies)
1268 want_cookie = 1;
1269 else
1270 #endif
1271 goto drop;
1272 }
1273
1274 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
1275 goto drop;
1276
1277 req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
1278 if (req == NULL)
1279 goto drop;
1280
1281 #ifdef CONFIG_TCP_MD5SIG
1282 tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
1283 #endif
1284
1285 tcp_clear_options(&tmp_opt);
1286 tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
1287 tmp_opt.user_mss = tp->rx_opt.user_mss;
1288
1289 tcp_parse_options(skb, &tmp_opt, 0);
1290
1291 if (want_cookie) {
1292 tcp_clear_options(&tmp_opt);
1293 tmp_opt.saw_tstamp = 0;
1294 }
1295
1296 tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
1297 tcp_openreq_init(req, &tmp_opt, skb);
1298
1299 treq = inet6_rsk(req);
1300 ipv6_addr_copy(&treq->rmt_addr, &ipv6_hdr(skb)->saddr);
1301 ipv6_addr_copy(&treq->loc_addr, &ipv6_hdr(skb)->daddr);
1302 treq->pktopts = NULL;
1303 if (!want_cookie)
1304 TCP_ECN_create_request(req, tcp_hdr(skb));
1305
1306 if (want_cookie) {
1307 isn = cookie_v6_init_sequence(sk, skb, &req->mss);
1308 } else if (!isn) {
1309 if (ipv6_opt_accepted(sk, skb) ||
1310 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
1311 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
1312 atomic_inc(&skb->users);
1313 treq->pktopts = skb;
1314 }
1315 treq->iif = sk->sk_bound_dev_if;
1316
1317 /* So that link locals have meaning */
1318 if (!sk->sk_bound_dev_if &&
1319 ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL)
1320 treq->iif = inet6_iif(skb);
1321
1322 isn = tcp_v6_init_sequence(skb);
1323 }
1324
1325 tcp_rsk(req)->snt_isn = isn;
1326
1327 security_inet_conn_request(sk, skb, req);
1328
1329 if (tcp_v6_send_synack(sk, req))
1330 goto drop;
1331
1332 if (!want_cookie) {
1333 inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
1334 return 0;
1335 }
1336
1337 drop:
1338 if (req)
1339 reqsk_free(req);
1340
1341 return 0; /* don't send reset */
1342 }
1343
1344 static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
1345 struct request_sock *req,
1346 struct dst_entry *dst)
1347 {
1348 struct inet6_request_sock *treq = inet6_rsk(req);
1349 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
1350 struct tcp6_sock *newtcp6sk;
1351 struct inet_sock *newinet;
1352 struct tcp_sock *newtp;
1353 struct sock *newsk;
1354 struct ipv6_txoptions *opt;
1355 #ifdef CONFIG_TCP_MD5SIG
1356 struct tcp_md5sig_key *key;
1357 #endif
1358
1359 if (skb->protocol == htons(ETH_P_IP)) {
1360 /*
1361 * v6 mapped
1362 */
1363
1364 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1365
1366 if (newsk == NULL)
1367 return NULL;
1368
1369 newtcp6sk = (struct tcp6_sock *)newsk;
1370 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1371
1372 newinet = inet_sk(newsk);
1373 newnp = inet6_sk(newsk);
1374 newtp = tcp_sk(newsk);
1375
1376 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1377
1378 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
1379 newinet->daddr);
1380
1381 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
1382 newinet->saddr);
1383
1384 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
1385
1386 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1387 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1388 #ifdef CONFIG_TCP_MD5SIG
1389 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1390 #endif
1391
1392 newnp->pktoptions = NULL;
1393 newnp->opt = NULL;
1394 newnp->mcast_oif = inet6_iif(skb);
1395 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1396
1397 /*
1398 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1399 * here, tcp_create_openreq_child now does this for us, see the comment in
1400 * that function for the gory details. -acme
1401 */
1402
1403 /* It is tricky place. Until this moment IPv4 tcp
1404 worked with IPv6 icsk.icsk_af_ops.
1405 Sync it now.
1406 */
1407 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1408
1409 return newsk;
1410 }
1411
1412 opt = np->opt;
1413
1414 if (sk_acceptq_is_full(sk))
1415 goto out_overflow;
1416
1417 if (dst == NULL) {
1418 struct in6_addr *final_p = NULL, final;
1419 struct flowi fl;
1420
1421 memset(&fl, 0, sizeof(fl));
1422 fl.proto = IPPROTO_TCP;
1423 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
1424 if (opt && opt->srcrt) {
1425 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
1426 ipv6_addr_copy(&final, &fl.fl6_dst);
1427 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
1428 final_p = &final;
1429 }
1430 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
1431 fl.oif = sk->sk_bound_dev_if;
1432 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
1433 fl.fl_ip_sport = inet_sk(sk)->sport;
1434 security_req_classify_flow(req, &fl);
1435
1436 if (ip6_dst_lookup(sk, &dst, &fl))
1437 goto out;
1438
1439 if (final_p)
1440 ipv6_addr_copy(&fl.fl6_dst, final_p);
1441
1442 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
1443 goto out;
1444 }
1445
1446 newsk = tcp_create_openreq_child(sk, req, skb);
1447 if (newsk == NULL)
1448 goto out;
1449
1450 /*
1451 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1452 * count here, tcp_create_openreq_child now does this for us, see the
1453 * comment in that function for the gory details. -acme
1454 */
1455
1456 newsk->sk_gso_type = SKB_GSO_TCPV6;
1457 __ip6_dst_store(newsk, dst, NULL, NULL);
1458
1459 newtcp6sk = (struct tcp6_sock *)newsk;
1460 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1461
1462 newtp = tcp_sk(newsk);
1463 newinet = inet_sk(newsk);
1464 newnp = inet6_sk(newsk);
1465
1466 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1467
1468 ipv6_addr_copy(&newnp->daddr, &treq->rmt_addr);
1469 ipv6_addr_copy(&newnp->saddr, &treq->loc_addr);
1470 ipv6_addr_copy(&newnp->rcv_saddr, &treq->loc_addr);
1471 newsk->sk_bound_dev_if = treq->iif;
1472
1473 /* Now IPv6 options...
1474
1475 First: no IPv4 options.
1476 */
1477 newinet->opt = NULL;
1478 newnp->ipv6_fl_list = NULL;
1479
1480 /* Clone RX bits */
1481 newnp->rxopt.all = np->rxopt.all;
1482
1483 /* Clone pktoptions received with SYN */
1484 newnp->pktoptions = NULL;
1485 if (treq->pktopts != NULL) {
1486 newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC);
1487 kfree_skb(treq->pktopts);
1488 treq->pktopts = NULL;
1489 if (newnp->pktoptions)
1490 skb_set_owner_r(newnp->pktoptions, newsk);
1491 }
1492 newnp->opt = NULL;
1493 newnp->mcast_oif = inet6_iif(skb);
1494 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1495
1496 /* Clone native IPv6 options from listening socket (if any)
1497
1498 Yes, keeping reference count would be much more clever,
1499 but we make one more one thing there: reattach optmem
1500 to newsk.
1501 */
1502 if (opt) {
1503 newnp->opt = ipv6_dup_options(newsk, opt);
1504 if (opt != np->opt)
1505 sock_kfree_s(sk, opt, opt->tot_len);
1506 }
1507
1508 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1509 if (newnp->opt)
1510 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1511 newnp->opt->opt_flen);
1512
1513 tcp_mtup_init(newsk);
1514 tcp_sync_mss(newsk, dst_mtu(dst));
1515 newtp->advmss = dst_metric(dst, RTAX_ADVMSS);
1516 tcp_initialize_rcv_mss(newsk);
1517
1518 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
1519
1520 #ifdef CONFIG_TCP_MD5SIG
1521 /* Copy over the MD5 key from the original socket */
1522 if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) {
1523 /* We're using one, so create a matching key
1524 * on the newsk structure. If we fail to get
1525 * memory, then we end up not copying the key
1526 * across. Shucks.
1527 */
1528 char *newkey = kmemdup(key->key, key->keylen, GFP_ATOMIC);
1529 if (newkey != NULL)
1530 tcp_v6_md5_do_add(newsk, &inet6_sk(sk)->daddr,
1531 newkey, key->keylen);
1532 }
1533 #endif
1534
1535 __inet6_hash(newsk);
1536 inet_inherit_port(sk, newsk);
1537
1538 return newsk;
1539
1540 out_overflow:
1541 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
1542 out:
1543 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
1544 if (opt && opt != np->opt)
1545 sock_kfree_s(sk, opt, opt->tot_len);
1546 dst_release(dst);
1547 return NULL;
1548 }
1549
1550 static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
1551 {
1552 if (skb->ip_summed == CHECKSUM_COMPLETE) {
1553 if (!tcp_v6_check(tcp_hdr(skb), skb->len, &ipv6_hdr(skb)->saddr,
1554 &ipv6_hdr(skb)->daddr, skb->csum)) {
1555 skb->ip_summed = CHECKSUM_UNNECESSARY;
1556 return 0;
1557 }
1558 }
1559
1560 skb->csum = ~csum_unfold(tcp_v6_check(tcp_hdr(skb), skb->len,
1561 &ipv6_hdr(skb)->saddr,
1562 &ipv6_hdr(skb)->daddr, 0));
1563
1564 if (skb->len <= 76) {
1565 return __skb_checksum_complete(skb);
1566 }
1567 return 0;
1568 }
1569
1570 /* The socket must have it's spinlock held when we get
1571 * here.
1572 *
1573 * We have a potential double-lock case here, so even when
1574 * doing backlog processing we use the BH locking scheme.
1575 * This is because we cannot sleep with the original spinlock
1576 * held.
1577 */
1578 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1579 {
1580 struct ipv6_pinfo *np = inet6_sk(sk);
1581 struct tcp_sock *tp;
1582 struct sk_buff *opt_skb = NULL;
1583
1584 /* Imagine: socket is IPv6. IPv4 packet arrives,
1585 goes to IPv4 receive handler and backlogged.
1586 From backlog it always goes here. Kerboom...
1587 Fortunately, tcp_rcv_established and rcv_established
1588 handle them correctly, but it is not case with
1589 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1590 */
1591
1592 if (skb->protocol == htons(ETH_P_IP))
1593 return tcp_v4_do_rcv(sk, skb);
1594
1595 #ifdef CONFIG_TCP_MD5SIG
1596 if (tcp_v6_inbound_md5_hash (sk, skb))
1597 goto discard;
1598 #endif
1599
1600 if (sk_filter(sk, skb))
1601 goto discard;
1602
1603 /*
1604 * socket locking is here for SMP purposes as backlog rcv
1605 * is currently called with bh processing disabled.
1606 */
1607
1608 /* Do Stevens' IPV6_PKTOPTIONS.
1609
1610 Yes, guys, it is the only place in our code, where we
1611 may make it not affecting IPv4.
1612 The rest of code is protocol independent,
1613 and I do not like idea to uglify IPv4.
1614
1615 Actually, all the idea behind IPV6_PKTOPTIONS
1616 looks not very well thought. For now we latch
1617 options, received in the last packet, enqueued
1618 by tcp. Feel free to propose better solution.
1619 --ANK (980728)
1620 */
1621 if (np->rxopt.all)
1622 opt_skb = skb_clone(skb, GFP_ATOMIC);
1623
1624 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1625 TCP_CHECK_TIMER(sk);
1626 if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len))
1627 goto reset;
1628 TCP_CHECK_TIMER(sk);
1629 if (opt_skb)
1630 goto ipv6_pktoptions;
1631 return 0;
1632 }
1633
1634 if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
1635 goto csum_err;
1636
1637 if (sk->sk_state == TCP_LISTEN) {
1638 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1639 if (!nsk)
1640 goto discard;
1641
1642 /*
1643 * Queue it on the new socket if the new socket is active,
1644 * otherwise we just shortcircuit this and continue with
1645 * the new socket..
1646 */
1647 if(nsk != sk) {
1648 if (tcp_child_process(sk, nsk, skb))
1649 goto reset;
1650 if (opt_skb)
1651 __kfree_skb(opt_skb);
1652 return 0;
1653 }
1654 }
1655
1656 TCP_CHECK_TIMER(sk);
1657 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
1658 goto reset;
1659 TCP_CHECK_TIMER(sk);
1660 if (opt_skb)
1661 goto ipv6_pktoptions;
1662 return 0;
1663
1664 reset:
1665 tcp_v6_send_reset(sk, skb);
1666 discard:
1667 if (opt_skb)
1668 __kfree_skb(opt_skb);
1669 kfree_skb(skb);
1670 return 0;
1671 csum_err:
1672 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1673 goto discard;
1674
1675
1676 ipv6_pktoptions:
1677 /* Do you ask, what is it?
1678
1679 1. skb was enqueued by tcp.
1680 2. skb is added to tail of read queue, rather than out of order.
1681 3. socket is not in passive state.
1682 4. Finally, it really contains options, which user wants to receive.
1683 */
1684 tp = tcp_sk(sk);
1685 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1686 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
1687 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
1688 np->mcast_oif = inet6_iif(opt_skb);
1689 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1690 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1691 if (ipv6_opt_accepted(sk, opt_skb)) {
1692 skb_set_owner_r(opt_skb, sk);
1693 opt_skb = xchg(&np->pktoptions, opt_skb);
1694 } else {
1695 __kfree_skb(opt_skb);
1696 opt_skb = xchg(&np->pktoptions, NULL);
1697 }
1698 }
1699
1700 if (opt_skb)
1701 kfree_skb(opt_skb);
1702 return 0;
1703 }
1704
1705 static int tcp_v6_rcv(struct sk_buff *skb)
1706 {
1707 struct tcphdr *th;
1708 struct sock *sk;
1709 int ret;
1710
1711 if (skb->pkt_type != PACKET_HOST)
1712 goto discard_it;
1713
1714 /*
1715 * Count it even if it's bad.
1716 */
1717 TCP_INC_STATS_BH(TCP_MIB_INSEGS);
1718
1719 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1720 goto discard_it;
1721
1722 th = tcp_hdr(skb);
1723
1724 if (th->doff < sizeof(struct tcphdr)/4)
1725 goto bad_packet;
1726 if (!pskb_may_pull(skb, th->doff*4))
1727 goto discard_it;
1728
1729 if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
1730 goto bad_packet;
1731
1732 th = tcp_hdr(skb);
1733 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1734 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1735 skb->len - th->doff*4);
1736 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1737 TCP_SKB_CB(skb)->when = 0;
1738 TCP_SKB_CB(skb)->flags = ipv6_get_dsfield(ipv6_hdr(skb));
1739 TCP_SKB_CB(skb)->sacked = 0;
1740
1741 sk = __inet6_lookup(skb->dev->nd_net, &tcp_hashinfo,
1742 &ipv6_hdr(skb)->saddr, th->source,
1743 &ipv6_hdr(skb)->daddr, ntohs(th->dest),
1744 inet6_iif(skb));
1745
1746 if (!sk)
1747 goto no_tcp_socket;
1748
1749 process:
1750 if (sk->sk_state == TCP_TIME_WAIT)
1751 goto do_time_wait;
1752
1753 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1754 goto discard_and_relse;
1755
1756 if (sk_filter(sk, skb))
1757 goto discard_and_relse;
1758
1759 skb->dev = NULL;
1760
1761 bh_lock_sock_nested(sk);
1762 ret = 0;
1763 if (!sock_owned_by_user(sk)) {
1764 #ifdef CONFIG_NET_DMA
1765 struct tcp_sock *tp = tcp_sk(sk);
1766 if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
1767 tp->ucopy.dma_chan = get_softnet_dma();
1768 if (tp->ucopy.dma_chan)
1769 ret = tcp_v6_do_rcv(sk, skb);
1770 else
1771 #endif
1772 {
1773 if (!tcp_prequeue(sk, skb))
1774 ret = tcp_v6_do_rcv(sk, skb);
1775 }
1776 } else
1777 sk_add_backlog(sk, skb);
1778 bh_unlock_sock(sk);
1779
1780 sock_put(sk);
1781 return ret ? -1 : 0;
1782
1783 no_tcp_socket:
1784 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1785 goto discard_it;
1786
1787 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1788 bad_packet:
1789 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1790 } else {
1791 tcp_v6_send_reset(NULL, skb);
1792 }
1793
1794 discard_it:
1795
1796 /*
1797 * Discard frame
1798 */
1799
1800 kfree_skb(skb);
1801 return 0;
1802
1803 discard_and_relse:
1804 sock_put(sk);
1805 goto discard_it;
1806
1807 do_time_wait:
1808 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1809 inet_twsk_put(inet_twsk(sk));
1810 goto discard_it;
1811 }
1812
1813 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1814 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1815 inet_twsk_put(inet_twsk(sk));
1816 goto discard_it;
1817 }
1818
1819 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1820 case TCP_TW_SYN:
1821 {
1822 struct sock *sk2;
1823
1824 sk2 = inet6_lookup_listener(skb->dev->nd_net, &tcp_hashinfo,
1825 &ipv6_hdr(skb)->daddr,
1826 ntohs(th->dest), inet6_iif(skb));
1827 if (sk2 != NULL) {
1828 struct inet_timewait_sock *tw = inet_twsk(sk);
1829 inet_twsk_deschedule(tw, &tcp_death_row);
1830 inet_twsk_put(tw);
1831 sk = sk2;
1832 goto process;
1833 }
1834 /* Fall through to ACK */
1835 }
1836 case TCP_TW_ACK:
1837 tcp_v6_timewait_ack(sk, skb);
1838 break;
1839 case TCP_TW_RST:
1840 goto no_tcp_socket;
1841 case TCP_TW_SUCCESS:;
1842 }
1843 goto discard_it;
1844 }
1845
1846 static int tcp_v6_remember_stamp(struct sock *sk)
1847 {
1848 /* Alas, not yet... */
1849 return 0;
1850 }
1851
1852 static struct inet_connection_sock_af_ops ipv6_specific = {
1853 .queue_xmit = inet6_csk_xmit,
1854 .send_check = tcp_v6_send_check,
1855 .rebuild_header = inet6_sk_rebuild_header,
1856 .conn_request = tcp_v6_conn_request,
1857 .syn_recv_sock = tcp_v6_syn_recv_sock,
1858 .remember_stamp = tcp_v6_remember_stamp,
1859 .net_header_len = sizeof(struct ipv6hdr),
1860 .setsockopt = ipv6_setsockopt,
1861 .getsockopt = ipv6_getsockopt,
1862 .addr2sockaddr = inet6_csk_addr2sockaddr,
1863 .sockaddr_len = sizeof(struct sockaddr_in6),
1864 .bind_conflict = inet6_csk_bind_conflict,
1865 #ifdef CONFIG_COMPAT
1866 .compat_setsockopt = compat_ipv6_setsockopt,
1867 .compat_getsockopt = compat_ipv6_getsockopt,
1868 #endif
1869 };
1870
1871 #ifdef CONFIG_TCP_MD5SIG
1872 static struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1873 .md5_lookup = tcp_v6_md5_lookup,
1874 .calc_md5_hash = tcp_v6_calc_md5_hash,
1875 .md5_add = tcp_v6_md5_add_func,
1876 .md5_parse = tcp_v6_parse_md5_keys,
1877 };
1878 #endif
1879
1880 /*
1881 * TCP over IPv4 via INET6 API
1882 */
1883
1884 static struct inet_connection_sock_af_ops ipv6_mapped = {
1885 .queue_xmit = ip_queue_xmit,
1886 .send_check = tcp_v4_send_check,
1887 .rebuild_header = inet_sk_rebuild_header,
1888 .conn_request = tcp_v6_conn_request,
1889 .syn_recv_sock = tcp_v6_syn_recv_sock,
1890 .remember_stamp = tcp_v4_remember_stamp,
1891 .net_header_len = sizeof(struct iphdr),
1892 .setsockopt = ipv6_setsockopt,
1893 .getsockopt = ipv6_getsockopt,
1894 .addr2sockaddr = inet6_csk_addr2sockaddr,
1895 .sockaddr_len = sizeof(struct sockaddr_in6),
1896 .bind_conflict = inet6_csk_bind_conflict,
1897 #ifdef CONFIG_COMPAT
1898 .compat_setsockopt = compat_ipv6_setsockopt,
1899 .compat_getsockopt = compat_ipv6_getsockopt,
1900 #endif
1901 };
1902
1903 #ifdef CONFIG_TCP_MD5SIG
1904 static struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1905 .md5_lookup = tcp_v4_md5_lookup,
1906 .calc_md5_hash = tcp_v4_calc_md5_hash,
1907 .md5_add = tcp_v6_md5_add_func,
1908 .md5_parse = tcp_v6_parse_md5_keys,
1909 };
1910 #endif
1911
1912 /* NOTE: A lot of things set to zero explicitly by call to
1913 * sk_alloc() so need not be done here.
1914 */
1915 static int tcp_v6_init_sock(struct sock *sk)
1916 {
1917 struct inet_connection_sock *icsk = inet_csk(sk);
1918 struct tcp_sock *tp = tcp_sk(sk);
1919
1920 skb_queue_head_init(&tp->out_of_order_queue);
1921 tcp_init_xmit_timers(sk);
1922 tcp_prequeue_init(tp);
1923
1924 icsk->icsk_rto = TCP_TIMEOUT_INIT;
1925 tp->mdev = TCP_TIMEOUT_INIT;
1926
1927 /* So many TCP implementations out there (incorrectly) count the
1928 * initial SYN frame in their delayed-ACK and congestion control
1929 * algorithms that we must have the following bandaid to talk
1930 * efficiently to them. -DaveM
1931 */
1932 tp->snd_cwnd = 2;
1933
1934 /* See draft-stevens-tcpca-spec-01 for discussion of the
1935 * initialization of these values.
1936 */
1937 tp->snd_ssthresh = 0x7fffffff;
1938 tp->snd_cwnd_clamp = ~0;
1939 tp->mss_cache = 536;
1940
1941 tp->reordering = sysctl_tcp_reordering;
1942
1943 sk->sk_state = TCP_CLOSE;
1944
1945 icsk->icsk_af_ops = &ipv6_specific;
1946 icsk->icsk_ca_ops = &tcp_init_congestion_ops;
1947 icsk->icsk_sync_mss = tcp_sync_mss;
1948 sk->sk_write_space = sk_stream_write_space;
1949 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
1950
1951 #ifdef CONFIG_TCP_MD5SIG
1952 tp->af_specific = &tcp_sock_ipv6_specific;
1953 #endif
1954
1955 sk->sk_sndbuf = sysctl_tcp_wmem[1];
1956 sk->sk_rcvbuf = sysctl_tcp_rmem[1];
1957
1958 atomic_inc(&tcp_sockets_allocated);
1959
1960 return 0;
1961 }
1962
1963 static int tcp_v6_destroy_sock(struct sock *sk)
1964 {
1965 #ifdef CONFIG_TCP_MD5SIG
1966 /* Clean up the MD5 key list */
1967 if (tcp_sk(sk)->md5sig_info)
1968 tcp_v6_clear_md5_list(sk);
1969 #endif
1970 tcp_v4_destroy_sock(sk);
1971 return inet6_destroy_sock(sk);
1972 }
1973
1974 #ifdef CONFIG_PROC_FS
1975 /* Proc filesystem TCPv6 sock list dumping. */
1976 static void get_openreq6(struct seq_file *seq,
1977 struct sock *sk, struct request_sock *req, int i, int uid)
1978 {
1979 int ttd = req->expires - jiffies;
1980 struct in6_addr *src = &inet6_rsk(req)->loc_addr;
1981 struct in6_addr *dest = &inet6_rsk(req)->rmt_addr;
1982
1983 if (ttd < 0)
1984 ttd = 0;
1985
1986 seq_printf(seq,
1987 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1988 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
1989 i,
1990 src->s6_addr32[0], src->s6_addr32[1],
1991 src->s6_addr32[2], src->s6_addr32[3],
1992 ntohs(inet_sk(sk)->sport),
1993 dest->s6_addr32[0], dest->s6_addr32[1],
1994 dest->s6_addr32[2], dest->s6_addr32[3],
1995 ntohs(inet_rsk(req)->rmt_port),
1996 TCP_SYN_RECV,
1997 0,0, /* could print option size, but that is af dependent. */
1998 1, /* timers active (only the expire timer) */
1999 jiffies_to_clock_t(ttd),
2000 req->retrans,
2001 uid,
2002 0, /* non standard timer */
2003 0, /* open_requests have no inode */
2004 0, req);
2005 }
2006
2007 static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
2008 {
2009 struct in6_addr *dest, *src;
2010 __u16 destp, srcp;
2011 int timer_active;
2012 unsigned long timer_expires;
2013 struct inet_sock *inet = inet_sk(sp);
2014 struct tcp_sock *tp = tcp_sk(sp);
2015 const struct inet_connection_sock *icsk = inet_csk(sp);
2016 struct ipv6_pinfo *np = inet6_sk(sp);
2017
2018 dest = &np->daddr;
2019 src = &np->rcv_saddr;
2020 destp = ntohs(inet->dport);
2021 srcp = ntohs(inet->sport);
2022
2023 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
2024 timer_active = 1;
2025 timer_expires = icsk->icsk_timeout;
2026 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
2027 timer_active = 4;
2028 timer_expires = icsk->icsk_timeout;
2029 } else if (timer_pending(&sp->sk_timer)) {
2030 timer_active = 2;
2031 timer_expires = sp->sk_timer.expires;
2032 } else {
2033 timer_active = 0;
2034 timer_expires = jiffies;
2035 }
2036
2037 seq_printf(seq,
2038 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2039 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %u %u %u %u %d\n",
2040 i,
2041 src->s6_addr32[0], src->s6_addr32[1],
2042 src->s6_addr32[2], src->s6_addr32[3], srcp,
2043 dest->s6_addr32[0], dest->s6_addr32[1],
2044 dest->s6_addr32[2], dest->s6_addr32[3], destp,
2045 sp->sk_state,
2046 tp->write_seq-tp->snd_una,
2047 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
2048 timer_active,
2049 jiffies_to_clock_t(timer_expires - jiffies),
2050 icsk->icsk_retransmits,
2051 sock_i_uid(sp),
2052 icsk->icsk_probes_out,
2053 sock_i_ino(sp),
2054 atomic_read(&sp->sk_refcnt), sp,
2055 icsk->icsk_rto,
2056 icsk->icsk_ack.ato,
2057 (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
2058 tp->snd_cwnd, tp->snd_ssthresh>=0xFFFF?-1:tp->snd_ssthresh
2059 );
2060 }
2061
2062 static void get_timewait6_sock(struct seq_file *seq,
2063 struct inet_timewait_sock *tw, int i)
2064 {
2065 struct in6_addr *dest, *src;
2066 __u16 destp, srcp;
2067 struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
2068 int ttd = tw->tw_ttd - jiffies;
2069
2070 if (ttd < 0)
2071 ttd = 0;
2072
2073 dest = &tw6->tw_v6_daddr;
2074 src = &tw6->tw_v6_rcv_saddr;
2075 destp = ntohs(tw->tw_dport);
2076 srcp = ntohs(tw->tw_sport);
2077
2078 seq_printf(seq,
2079 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2080 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
2081 i,
2082 src->s6_addr32[0], src->s6_addr32[1],
2083 src->s6_addr32[2], src->s6_addr32[3], srcp,
2084 dest->s6_addr32[0], dest->s6_addr32[1],
2085 dest->s6_addr32[2], dest->s6_addr32[3], destp,
2086 tw->tw_substate, 0, 0,
2087 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
2088 atomic_read(&tw->tw_refcnt), tw);
2089 }
2090
2091 static int tcp6_seq_show(struct seq_file *seq, void *v)
2092 {
2093 struct tcp_iter_state *st;
2094
2095 if (v == SEQ_START_TOKEN) {
2096 seq_puts(seq,
2097 " sl "
2098 "local_address "
2099 "remote_address "
2100 "st tx_queue rx_queue tr tm->when retrnsmt"
2101 " uid timeout inode\n");
2102 goto out;
2103 }
2104 st = seq->private;
2105
2106 switch (st->state) {
2107 case TCP_SEQ_STATE_LISTENING:
2108 case TCP_SEQ_STATE_ESTABLISHED:
2109 get_tcp6_sock(seq, v, st->num);
2110 break;
2111 case TCP_SEQ_STATE_OPENREQ:
2112 get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
2113 break;
2114 case TCP_SEQ_STATE_TIME_WAIT:
2115 get_timewait6_sock(seq, v, st->num);
2116 break;
2117 }
2118 out:
2119 return 0;
2120 }
2121
2122 static struct file_operations tcp6_seq_fops;
2123 static struct tcp_seq_afinfo tcp6_seq_afinfo = {
2124 .owner = THIS_MODULE,
2125 .name = "tcp6",
2126 .family = AF_INET6,
2127 .seq_show = tcp6_seq_show,
2128 .seq_fops = &tcp6_seq_fops,
2129 };
2130
2131 int __init tcp6_proc_init(void)
2132 {
2133 return tcp_proc_register(&tcp6_seq_afinfo);
2134 }
2135
2136 void tcp6_proc_exit(void)
2137 {
2138 tcp_proc_unregister(&tcp6_seq_afinfo);
2139 }
2140 #endif
2141
2142 DEFINE_PROTO_INUSE(tcpv6)
2143
2144 struct proto tcpv6_prot = {
2145 .name = "TCPv6",
2146 .owner = THIS_MODULE,
2147 .close = tcp_close,
2148 .connect = tcp_v6_connect,
2149 .disconnect = tcp_disconnect,
2150 .accept = inet_csk_accept,
2151 .ioctl = tcp_ioctl,
2152 .init = tcp_v6_init_sock,
2153 .destroy = tcp_v6_destroy_sock,
2154 .shutdown = tcp_shutdown,
2155 .setsockopt = tcp_setsockopt,
2156 .getsockopt = tcp_getsockopt,
2157 .recvmsg = tcp_recvmsg,
2158 .backlog_rcv = tcp_v6_do_rcv,
2159 .hash = tcp_v6_hash,
2160 .unhash = inet_unhash,
2161 .get_port = inet_csk_get_port,
2162 .enter_memory_pressure = tcp_enter_memory_pressure,
2163 .sockets_allocated = &tcp_sockets_allocated,
2164 .memory_allocated = &tcp_memory_allocated,
2165 .memory_pressure = &tcp_memory_pressure,
2166 .orphan_count = &tcp_orphan_count,
2167 .sysctl_mem = sysctl_tcp_mem,
2168 .sysctl_wmem = sysctl_tcp_wmem,
2169 .sysctl_rmem = sysctl_tcp_rmem,
2170 .max_header = MAX_TCP_HEADER,
2171 .obj_size = sizeof(struct tcp6_sock),
2172 .twsk_prot = &tcp6_timewait_sock_ops,
2173 .rsk_prot = &tcp6_request_sock_ops,
2174 .hashinfo = &tcp_hashinfo,
2175 #ifdef CONFIG_COMPAT
2176 .compat_setsockopt = compat_tcp_setsockopt,
2177 .compat_getsockopt = compat_tcp_getsockopt,
2178 #endif
2179 REF_PROTO_INUSE(tcpv6)
2180 };
2181
2182 static struct inet6_protocol tcpv6_protocol = {
2183 .handler = tcp_v6_rcv,
2184 .err_handler = tcp_v6_err,
2185 .gso_send_check = tcp_v6_gso_send_check,
2186 .gso_segment = tcp_tso_segment,
2187 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
2188 };
2189
2190 static struct inet_protosw tcpv6_protosw = {
2191 .type = SOCK_STREAM,
2192 .protocol = IPPROTO_TCP,
2193 .prot = &tcpv6_prot,
2194 .ops = &inet6_stream_ops,
2195 .capability = -1,
2196 .no_check = 0,
2197 .flags = INET_PROTOSW_PERMANENT |
2198 INET_PROTOSW_ICSK,
2199 };
2200
2201 int __init tcpv6_init(void)
2202 {
2203 int ret;
2204
2205 ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
2206 if (ret)
2207 goto out;
2208
2209 /* register inet6 protocol */
2210 ret = inet6_register_protosw(&tcpv6_protosw);
2211 if (ret)
2212 goto out_tcpv6_protocol;
2213
2214 ret = inet_csk_ctl_sock_create(&tcp6_socket, PF_INET6,
2215 SOCK_RAW, IPPROTO_TCP);
2216 if (ret)
2217 goto out_tcpv6_protosw;
2218 out:
2219 return ret;
2220
2221 out_tcpv6_protocol:
2222 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2223 out_tcpv6_protosw:
2224 inet6_unregister_protosw(&tcpv6_protosw);
2225 goto out;
2226 }
2227
2228 void tcpv6_exit(void)
2229 {
2230 sock_release(tcp6_socket);
2231 inet6_unregister_protosw(&tcpv6_protosw);
2232 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2233 }