init from v2.6.32.60
[mach-moxart.git] / net / ipv6 / tcp_ipv6.c
blobfaae6dfd6cf370c047dbea268aa550063f138d29
1 /*
2 * TCP over IPv6
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Based on:
9 * linux/net/ipv4/tcp.c
10 * linux/net/ipv4/tcp_input.c
11 * linux/net/ipv4/tcp_output.c
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.
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.
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>
42 #include <linux/ipv6.h>
43 #include <linux/icmpv6.h>
44 #include <linux/random.h>
46 #include <net/tcp.h>
47 #include <net/ndisc.h>
48 #include <net/inet6_hashtables.h>
49 #include <net/inet6_connection_sock.h>
50 #include <net/ipv6.h>
51 #include <net/transp_v6.h>
52 #include <net/addrconf.h>
53 #include <net/ip6_route.h>
54 #include <net/ip6_checksum.h>
55 #include <net/inet_ecn.h>
56 #include <net/protocol.h>
57 #include <net/xfrm.h>
58 #include <net/snmp.h>
59 #include <net/dsfield.h>
60 #include <net/timewait_sock.h>
61 #include <net/netdma.h>
62 #include <net/inet_common.h>
63 #include <net/secure_seq.h>
65 #include <asm/uaccess.h>
67 #include <linux/proc_fs.h>
68 #include <linux/seq_file.h>
70 #include <linux/crypto.h>
71 #include <linux/scatterlist.h>
73 static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb);
74 static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
75 struct request_sock *req);
77 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
79 static const struct inet_connection_sock_af_ops ipv6_mapped;
80 static const struct inet_connection_sock_af_ops ipv6_specific;
81 #ifdef CONFIG_TCP_MD5SIG
82 static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
83 static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
84 #else
85 static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
86 struct in6_addr *addr)
88 return NULL;
90 #endif
92 static void tcp_v6_hash(struct sock *sk)
94 if (sk->sk_state != TCP_CLOSE) {
95 if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
96 tcp_prot.hash(sk);
97 return;
99 local_bh_disable();
100 __inet6_hash(sk);
101 local_bh_enable();
105 static __inline__ __sum16 tcp_v6_check(int len,
106 struct in6_addr *saddr,
107 struct in6_addr *daddr,
108 __wsum base)
110 return csum_ipv6_magic(saddr, daddr, len, IPPROTO_TCP, base);
113 static __u32 tcp_v6_init_sequence(struct sk_buff *skb)
115 return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
116 ipv6_hdr(skb)->saddr.s6_addr32,
117 tcp_hdr(skb)->dest,
118 tcp_hdr(skb)->source);
121 static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
122 int addr_len)
124 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
125 struct inet_sock *inet = inet_sk(sk);
126 struct inet_connection_sock *icsk = inet_csk(sk);
127 struct ipv6_pinfo *np = inet6_sk(sk);
128 struct tcp_sock *tp = tcp_sk(sk);
129 struct in6_addr *saddr = NULL, *final_p = NULL, final;
130 struct flowi fl;
131 struct dst_entry *dst;
132 int addr_type;
133 int err;
135 if (addr_len < SIN6_LEN_RFC2133)
136 return -EINVAL;
138 if (usin->sin6_family != AF_INET6)
139 return(-EAFNOSUPPORT);
141 memset(&fl, 0, sizeof(fl));
143 if (np->sndflow) {
144 fl.fl6_flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
145 IP6_ECN_flow_init(fl.fl6_flowlabel);
146 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
147 struct ip6_flowlabel *flowlabel;
148 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
149 if (flowlabel == NULL)
150 return -EINVAL;
151 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
152 fl6_sock_release(flowlabel);
157 * connect() to INADDR_ANY means loopback (BSD'ism).
160 if(ipv6_addr_any(&usin->sin6_addr))
161 usin->sin6_addr.s6_addr[15] = 0x1;
163 addr_type = ipv6_addr_type(&usin->sin6_addr);
165 if(addr_type & IPV6_ADDR_MULTICAST)
166 return -ENETUNREACH;
168 if (addr_type&IPV6_ADDR_LINKLOCAL) {
169 if (addr_len >= sizeof(struct sockaddr_in6) &&
170 usin->sin6_scope_id) {
171 /* If interface is set while binding, indices
172 * must coincide.
174 if (sk->sk_bound_dev_if &&
175 sk->sk_bound_dev_if != usin->sin6_scope_id)
176 return -EINVAL;
178 sk->sk_bound_dev_if = usin->sin6_scope_id;
181 /* Connect to link-local address requires an interface */
182 if (!sk->sk_bound_dev_if)
183 return -EINVAL;
186 if (tp->rx_opt.ts_recent_stamp &&
187 !ipv6_addr_equal(&np->daddr, &usin->sin6_addr)) {
188 tp->rx_opt.ts_recent = 0;
189 tp->rx_opt.ts_recent_stamp = 0;
190 tp->write_seq = 0;
193 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
194 np->flow_label = fl.fl6_flowlabel;
197 * TCP over IPv4
200 if (addr_type == IPV6_ADDR_MAPPED) {
201 u32 exthdrlen = icsk->icsk_ext_hdr_len;
202 struct sockaddr_in sin;
204 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
206 if (__ipv6_only_sock(sk))
207 return -ENETUNREACH;
209 sin.sin_family = AF_INET;
210 sin.sin_port = usin->sin6_port;
211 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
213 icsk->icsk_af_ops = &ipv6_mapped;
214 sk->sk_backlog_rcv = tcp_v4_do_rcv;
215 #ifdef CONFIG_TCP_MD5SIG
216 tp->af_specific = &tcp_sock_ipv6_mapped_specific;
217 #endif
219 err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
221 if (err) {
222 icsk->icsk_ext_hdr_len = exthdrlen;
223 icsk->icsk_af_ops = &ipv6_specific;
224 sk->sk_backlog_rcv = tcp_v6_do_rcv;
225 #ifdef CONFIG_TCP_MD5SIG
226 tp->af_specific = &tcp_sock_ipv6_specific;
227 #endif
228 goto failure;
229 } else {
230 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
231 inet->saddr);
232 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
233 inet->rcv_saddr);
236 return err;
239 if (!ipv6_addr_any(&np->rcv_saddr))
240 saddr = &np->rcv_saddr;
242 fl.proto = IPPROTO_TCP;
243 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
244 ipv6_addr_copy(&fl.fl6_src,
245 (saddr ? saddr : &np->saddr));
246 fl.oif = sk->sk_bound_dev_if;
247 fl.fl_ip_dport = usin->sin6_port;
248 fl.fl_ip_sport = inet->sport;
250 if (np->opt && np->opt->srcrt) {
251 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
252 ipv6_addr_copy(&final, &fl.fl6_dst);
253 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
254 final_p = &final;
257 security_sk_classify_flow(sk, &fl);
259 err = ip6_dst_lookup(sk, &dst, &fl);
260 if (err)
261 goto failure;
262 if (final_p)
263 ipv6_addr_copy(&fl.fl6_dst, final_p);
265 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
266 if (err < 0) {
267 if (err == -EREMOTE)
268 err = ip6_dst_blackhole(sk, &dst, &fl);
269 if (err < 0)
270 goto failure;
273 if (saddr == NULL) {
274 saddr = &fl.fl6_src;
275 ipv6_addr_copy(&np->rcv_saddr, saddr);
278 /* set the source address */
279 ipv6_addr_copy(&np->saddr, saddr);
280 inet->rcv_saddr = LOOPBACK4_IPV6;
282 sk->sk_gso_type = SKB_GSO_TCPV6;
283 __ip6_dst_store(sk, dst, NULL, NULL);
285 icsk->icsk_ext_hdr_len = 0;
286 if (np->opt)
287 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
288 np->opt->opt_nflen);
290 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
292 inet->dport = usin->sin6_port;
294 tcp_set_state(sk, TCP_SYN_SENT);
295 err = inet6_hash_connect(&tcp_death_row, sk);
296 if (err)
297 goto late_failure;
299 if (!tp->write_seq)
300 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
301 np->daddr.s6_addr32,
302 inet->sport,
303 inet->dport);
305 err = tcp_connect(sk);
306 if (err)
307 goto late_failure;
309 return 0;
311 late_failure:
312 tcp_set_state(sk, TCP_CLOSE);
313 __sk_dst_reset(sk);
314 failure:
315 inet->dport = 0;
316 sk->sk_route_caps = 0;
317 return err;
320 static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
321 u8 type, u8 code, int offset, __be32 info)
323 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
324 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
325 struct ipv6_pinfo *np;
326 struct sock *sk;
327 int err;
328 struct tcp_sock *tp;
329 __u32 seq;
330 struct net *net = dev_net(skb->dev);
332 sk = inet6_lookup(net, &tcp_hashinfo, &hdr->daddr,
333 th->dest, &hdr->saddr, th->source, skb->dev->ifindex);
335 if (sk == NULL) {
336 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
337 ICMP6_MIB_INERRORS);
338 return;
341 if (sk->sk_state == TCP_TIME_WAIT) {
342 inet_twsk_put(inet_twsk(sk));
343 return;
346 bh_lock_sock(sk);
347 if (sock_owned_by_user(sk))
348 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
350 if (sk->sk_state == TCP_CLOSE)
351 goto out;
353 tp = tcp_sk(sk);
354 seq = ntohl(th->seq);
355 if (sk->sk_state != TCP_LISTEN &&
356 !between(seq, tp->snd_una, tp->snd_nxt)) {
357 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
358 goto out;
361 np = inet6_sk(sk);
363 if (type == ICMPV6_PKT_TOOBIG) {
364 struct dst_entry *dst = NULL;
366 if (sock_owned_by_user(sk))
367 goto out;
368 if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
369 goto out;
371 /* icmp should have updated the destination cache entry */
372 dst = __sk_dst_check(sk, np->dst_cookie);
374 if (dst == NULL) {
375 struct inet_sock *inet = inet_sk(sk);
376 struct flowi fl;
378 /* BUGGG_FUTURE: Again, it is not clear how
379 to handle rthdr case. Ignore this complexity
380 for now.
382 memset(&fl, 0, sizeof(fl));
383 fl.proto = IPPROTO_TCP;
384 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
385 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
386 fl.oif = sk->sk_bound_dev_if;
387 fl.fl_ip_dport = inet->dport;
388 fl.fl_ip_sport = inet->sport;
389 security_skb_classify_flow(skb, &fl);
391 if ((err = ip6_dst_lookup(sk, &dst, &fl))) {
392 sk->sk_err_soft = -err;
393 goto out;
396 if ((err = xfrm_lookup(net, &dst, &fl, sk, 0)) < 0) {
397 sk->sk_err_soft = -err;
398 goto out;
401 } else
402 dst_hold(dst);
404 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) {
405 tcp_sync_mss(sk, dst_mtu(dst));
406 tcp_simple_retransmit(sk);
407 } /* else let the usual retransmit timer handle it */
408 dst_release(dst);
409 goto out;
412 icmpv6_err_convert(type, code, &err);
414 /* Might be for an request_sock */
415 switch (sk->sk_state) {
416 struct request_sock *req, **prev;
417 case TCP_LISTEN:
418 if (sock_owned_by_user(sk))
419 goto out;
421 req = inet6_csk_search_req(sk, &prev, th->dest, &hdr->daddr,
422 &hdr->saddr, inet6_iif(skb));
423 if (!req)
424 goto out;
426 /* ICMPs are not backlogged, hence we cannot get
427 * an established socket here.
429 WARN_ON(req->sk != NULL);
431 if (seq != tcp_rsk(req)->snt_isn) {
432 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
433 goto out;
436 inet_csk_reqsk_queue_drop(sk, req, prev);
437 goto out;
439 case TCP_SYN_SENT:
440 case TCP_SYN_RECV: /* Cannot happen.
441 It can, it SYNs are crossed. --ANK */
442 if (!sock_owned_by_user(sk)) {
443 sk->sk_err = err;
444 sk->sk_error_report(sk); /* Wake people up to see the error (see connect in sock.c) */
446 tcp_done(sk);
447 } else
448 sk->sk_err_soft = err;
449 goto out;
452 if (!sock_owned_by_user(sk) && np->recverr) {
453 sk->sk_err = err;
454 sk->sk_error_report(sk);
455 } else
456 sk->sk_err_soft = err;
458 out:
459 bh_unlock_sock(sk);
460 sock_put(sk);
464 static int tcp_v6_send_synack(struct sock *sk, struct request_sock *req)
466 struct inet6_request_sock *treq = inet6_rsk(req);
467 struct ipv6_pinfo *np = inet6_sk(sk);
468 struct sk_buff * skb;
469 struct ipv6_txoptions *opt = NULL;
470 struct in6_addr * final_p = NULL, final;
471 struct flowi fl;
472 struct dst_entry *dst;
473 int err = -1;
475 memset(&fl, 0, sizeof(fl));
476 fl.proto = IPPROTO_TCP;
477 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
478 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
479 fl.fl6_flowlabel = 0;
480 fl.oif = treq->iif;
481 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
482 fl.fl_ip_sport = inet_rsk(req)->loc_port;
483 security_req_classify_flow(req, &fl);
485 opt = np->opt;
486 if (opt && opt->srcrt) {
487 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
488 ipv6_addr_copy(&final, &fl.fl6_dst);
489 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
490 final_p = &final;
493 err = ip6_dst_lookup(sk, &dst, &fl);
494 if (err)
495 goto done;
496 if (final_p)
497 ipv6_addr_copy(&fl.fl6_dst, final_p);
498 if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0)
499 goto done;
501 skb = tcp_make_synack(sk, dst, req);
502 if (skb) {
503 struct tcphdr *th = tcp_hdr(skb);
505 th->check = tcp_v6_check(skb->len,
506 &treq->loc_addr, &treq->rmt_addr,
507 csum_partial(th, skb->len, skb->csum));
509 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
510 err = ip6_xmit(sk, skb, &fl, opt, 0);
511 err = net_xmit_eval(err);
514 done:
515 if (opt && opt != np->opt)
516 sock_kfree_s(sk, opt, opt->tot_len);
517 dst_release(dst);
518 return err;
521 static inline void syn_flood_warning(struct sk_buff *skb)
523 #ifdef CONFIG_SYN_COOKIES
524 if (sysctl_tcp_syncookies)
525 printk(KERN_INFO
526 "TCPv6: Possible SYN flooding on port %d. "
527 "Sending cookies.\n", ntohs(tcp_hdr(skb)->dest));
528 else
529 #endif
530 printk(KERN_INFO
531 "TCPv6: Possible SYN flooding on port %d. "
532 "Dropping request.\n", ntohs(tcp_hdr(skb)->dest));
535 static void tcp_v6_reqsk_destructor(struct request_sock *req)
537 kfree_skb(inet6_rsk(req)->pktopts);
540 #ifdef CONFIG_TCP_MD5SIG
541 static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(struct sock *sk,
542 struct in6_addr *addr)
544 struct tcp_sock *tp = tcp_sk(sk);
545 int i;
547 BUG_ON(tp == NULL);
549 if (!tp->md5sig_info || !tp->md5sig_info->entries6)
550 return NULL;
552 for (i = 0; i < tp->md5sig_info->entries6; i++) {
553 if (ipv6_addr_equal(&tp->md5sig_info->keys6[i].addr, addr))
554 return &tp->md5sig_info->keys6[i].base;
556 return NULL;
559 static struct tcp_md5sig_key *tcp_v6_md5_lookup(struct sock *sk,
560 struct sock *addr_sk)
562 return tcp_v6_md5_do_lookup(sk, &inet6_sk(addr_sk)->daddr);
565 static struct tcp_md5sig_key *tcp_v6_reqsk_md5_lookup(struct sock *sk,
566 struct request_sock *req)
568 return tcp_v6_md5_do_lookup(sk, &inet6_rsk(req)->rmt_addr);
571 static int tcp_v6_md5_do_add(struct sock *sk, struct in6_addr *peer,
572 char *newkey, u8 newkeylen)
574 /* Add key to the list */
575 struct tcp_md5sig_key *key;
576 struct tcp_sock *tp = tcp_sk(sk);
577 struct tcp6_md5sig_key *keys;
579 key = tcp_v6_md5_do_lookup(sk, peer);
580 if (key) {
581 /* modify existing entry - just update that one */
582 kfree(key->key);
583 key->key = newkey;
584 key->keylen = newkeylen;
585 } else {
586 /* reallocate new list if current one is full. */
587 if (!tp->md5sig_info) {
588 tp->md5sig_info = kzalloc(sizeof(*tp->md5sig_info), GFP_ATOMIC);
589 if (!tp->md5sig_info) {
590 kfree(newkey);
591 return -ENOMEM;
593 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
595 if (tcp_alloc_md5sig_pool(sk) == NULL) {
596 kfree(newkey);
597 return -ENOMEM;
599 if (tp->md5sig_info->alloced6 == tp->md5sig_info->entries6) {
600 keys = kmalloc((sizeof (tp->md5sig_info->keys6[0]) *
601 (tp->md5sig_info->entries6 + 1)), GFP_ATOMIC);
603 if (!keys) {
604 tcp_free_md5sig_pool();
605 kfree(newkey);
606 return -ENOMEM;
609 if (tp->md5sig_info->entries6)
610 memmove(keys, tp->md5sig_info->keys6,
611 (sizeof (tp->md5sig_info->keys6[0]) *
612 tp->md5sig_info->entries6));
614 kfree(tp->md5sig_info->keys6);
615 tp->md5sig_info->keys6 = keys;
616 tp->md5sig_info->alloced6++;
619 ipv6_addr_copy(&tp->md5sig_info->keys6[tp->md5sig_info->entries6].addr,
620 peer);
621 tp->md5sig_info->keys6[tp->md5sig_info->entries6].base.key = newkey;
622 tp->md5sig_info->keys6[tp->md5sig_info->entries6].base.keylen = newkeylen;
624 tp->md5sig_info->entries6++;
626 return 0;
629 static int tcp_v6_md5_add_func(struct sock *sk, struct sock *addr_sk,
630 u8 *newkey, __u8 newkeylen)
632 return tcp_v6_md5_do_add(sk, &inet6_sk(addr_sk)->daddr,
633 newkey, newkeylen);
636 static int tcp_v6_md5_do_del(struct sock *sk, struct in6_addr *peer)
638 struct tcp_sock *tp = tcp_sk(sk);
639 int i;
641 for (i = 0; i < tp->md5sig_info->entries6; i++) {
642 if (ipv6_addr_equal(&tp->md5sig_info->keys6[i].addr, peer)) {
643 /* Free the key */
644 kfree(tp->md5sig_info->keys6[i].base.key);
645 tp->md5sig_info->entries6--;
647 if (tp->md5sig_info->entries6 == 0) {
648 kfree(tp->md5sig_info->keys6);
649 tp->md5sig_info->keys6 = NULL;
650 tp->md5sig_info->alloced6 = 0;
651 } else {
652 /* shrink the database */
653 if (tp->md5sig_info->entries6 != i)
654 memmove(&tp->md5sig_info->keys6[i],
655 &tp->md5sig_info->keys6[i+1],
656 (tp->md5sig_info->entries6 - i)
657 * sizeof (tp->md5sig_info->keys6[0]));
659 tcp_free_md5sig_pool();
660 return 0;
663 return -ENOENT;
666 static void tcp_v6_clear_md5_list (struct sock *sk)
668 struct tcp_sock *tp = tcp_sk(sk);
669 int i;
671 if (tp->md5sig_info->entries6) {
672 for (i = 0; i < tp->md5sig_info->entries6; i++)
673 kfree(tp->md5sig_info->keys6[i].base.key);
674 tp->md5sig_info->entries6 = 0;
675 tcp_free_md5sig_pool();
678 kfree(tp->md5sig_info->keys6);
679 tp->md5sig_info->keys6 = NULL;
680 tp->md5sig_info->alloced6 = 0;
682 if (tp->md5sig_info->entries4) {
683 for (i = 0; i < tp->md5sig_info->entries4; i++)
684 kfree(tp->md5sig_info->keys4[i].base.key);
685 tp->md5sig_info->entries4 = 0;
686 tcp_free_md5sig_pool();
689 kfree(tp->md5sig_info->keys4);
690 tp->md5sig_info->keys4 = NULL;
691 tp->md5sig_info->alloced4 = 0;
694 static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval,
695 int optlen)
697 struct tcp_md5sig cmd;
698 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
699 u8 *newkey;
701 if (optlen < sizeof(cmd))
702 return -EINVAL;
704 if (copy_from_user(&cmd, optval, sizeof(cmd)))
705 return -EFAULT;
707 if (sin6->sin6_family != AF_INET6)
708 return -EINVAL;
710 if (!cmd.tcpm_keylen) {
711 if (!tcp_sk(sk)->md5sig_info)
712 return -ENOENT;
713 if (ipv6_addr_v4mapped(&sin6->sin6_addr))
714 return tcp_v4_md5_do_del(sk, sin6->sin6_addr.s6_addr32[3]);
715 return tcp_v6_md5_do_del(sk, &sin6->sin6_addr);
718 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
719 return -EINVAL;
721 if (!tcp_sk(sk)->md5sig_info) {
722 struct tcp_sock *tp = tcp_sk(sk);
723 struct tcp_md5sig_info *p;
725 p = kzalloc(sizeof(struct tcp_md5sig_info), GFP_KERNEL);
726 if (!p)
727 return -ENOMEM;
729 tp->md5sig_info = p;
730 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
733 newkey = kmemdup(cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
734 if (!newkey)
735 return -ENOMEM;
736 if (ipv6_addr_v4mapped(&sin6->sin6_addr)) {
737 return tcp_v4_md5_do_add(sk, sin6->sin6_addr.s6_addr32[3],
738 newkey, cmd.tcpm_keylen);
740 return tcp_v6_md5_do_add(sk, &sin6->sin6_addr, newkey, cmd.tcpm_keylen);
743 static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool *hp,
744 struct in6_addr *daddr,
745 struct in6_addr *saddr, int nbytes)
747 struct tcp6_pseudohdr *bp;
748 struct scatterlist sg;
750 bp = &hp->md5_blk.ip6;
751 /* 1. TCP pseudo-header (RFC2460) */
752 ipv6_addr_copy(&bp->saddr, saddr);
753 ipv6_addr_copy(&bp->daddr, daddr);
754 bp->protocol = cpu_to_be32(IPPROTO_TCP);
755 bp->len = cpu_to_be32(nbytes);
757 sg_init_one(&sg, bp, sizeof(*bp));
758 return crypto_hash_update(&hp->md5_desc, &sg, sizeof(*bp));
761 static int tcp_v6_md5_hash_hdr(char *md5_hash, struct tcp_md5sig_key *key,
762 struct in6_addr *daddr, struct in6_addr *saddr,
763 struct tcphdr *th)
765 struct tcp_md5sig_pool *hp;
766 struct hash_desc *desc;
768 hp = tcp_get_md5sig_pool();
769 if (!hp)
770 goto clear_hash_noput;
771 desc = &hp->md5_desc;
773 if (crypto_hash_init(desc))
774 goto clear_hash;
775 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, th->doff << 2))
776 goto clear_hash;
777 if (tcp_md5_hash_header(hp, th))
778 goto clear_hash;
779 if (tcp_md5_hash_key(hp, key))
780 goto clear_hash;
781 if (crypto_hash_final(desc, md5_hash))
782 goto clear_hash;
784 tcp_put_md5sig_pool();
785 return 0;
787 clear_hash:
788 tcp_put_md5sig_pool();
789 clear_hash_noput:
790 memset(md5_hash, 0, 16);
791 return 1;
794 static int tcp_v6_md5_hash_skb(char *md5_hash, struct tcp_md5sig_key *key,
795 struct sock *sk, struct request_sock *req,
796 struct sk_buff *skb)
798 struct in6_addr *saddr, *daddr;
799 struct tcp_md5sig_pool *hp;
800 struct hash_desc *desc;
801 struct tcphdr *th = tcp_hdr(skb);
803 if (sk) {
804 saddr = &inet6_sk(sk)->saddr;
805 daddr = &inet6_sk(sk)->daddr;
806 } else if (req) {
807 saddr = &inet6_rsk(req)->loc_addr;
808 daddr = &inet6_rsk(req)->rmt_addr;
809 } else {
810 struct ipv6hdr *ip6h = ipv6_hdr(skb);
811 saddr = &ip6h->saddr;
812 daddr = &ip6h->daddr;
815 hp = tcp_get_md5sig_pool();
816 if (!hp)
817 goto clear_hash_noput;
818 desc = &hp->md5_desc;
820 if (crypto_hash_init(desc))
821 goto clear_hash;
823 if (tcp_v6_md5_hash_pseudoheader(hp, daddr, saddr, skb->len))
824 goto clear_hash;
825 if (tcp_md5_hash_header(hp, th))
826 goto clear_hash;
827 if (tcp_md5_hash_skb_data(hp, skb, th->doff << 2))
828 goto clear_hash;
829 if (tcp_md5_hash_key(hp, key))
830 goto clear_hash;
831 if (crypto_hash_final(desc, md5_hash))
832 goto clear_hash;
834 tcp_put_md5sig_pool();
835 return 0;
837 clear_hash:
838 tcp_put_md5sig_pool();
839 clear_hash_noput:
840 memset(md5_hash, 0, 16);
841 return 1;
844 static int tcp_v6_inbound_md5_hash (struct sock *sk, struct sk_buff *skb)
846 __u8 *hash_location = NULL;
847 struct tcp_md5sig_key *hash_expected;
848 struct ipv6hdr *ip6h = ipv6_hdr(skb);
849 struct tcphdr *th = tcp_hdr(skb);
850 int genhash;
851 u8 newhash[16];
853 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
854 hash_location = tcp_parse_md5sig_option(th);
856 /* We've parsed the options - do we have a hash? */
857 if (!hash_expected && !hash_location)
858 return 0;
860 if (hash_expected && !hash_location) {
861 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
862 return 1;
865 if (!hash_expected && hash_location) {
866 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
867 return 1;
870 /* check the signature */
871 genhash = tcp_v6_md5_hash_skb(newhash,
872 hash_expected,
873 NULL, NULL, skb);
875 if (genhash || memcmp(hash_location, newhash, 16) != 0) {
876 if (net_ratelimit()) {
877 printk(KERN_INFO "MD5 Hash %s for (%pI6, %u)->(%pI6, %u)\n",
878 genhash ? "failed" : "mismatch",
879 &ip6h->saddr, ntohs(th->source),
880 &ip6h->daddr, ntohs(th->dest));
882 return 1;
884 return 0;
886 #endif
888 struct request_sock_ops tcp6_request_sock_ops __read_mostly = {
889 .family = AF_INET6,
890 .obj_size = sizeof(struct tcp6_request_sock),
891 .rtx_syn_ack = tcp_v6_send_synack,
892 .send_ack = tcp_v6_reqsk_send_ack,
893 .destructor = tcp_v6_reqsk_destructor,
894 .send_reset = tcp_v6_send_reset
897 #ifdef CONFIG_TCP_MD5SIG
898 static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
899 .md5_lookup = tcp_v6_reqsk_md5_lookup,
900 .calc_md5_hash = tcp_v6_md5_hash_skb,
902 #endif
904 static struct timewait_sock_ops tcp6_timewait_sock_ops = {
905 .twsk_obj_size = sizeof(struct tcp6_timewait_sock),
906 .twsk_unique = tcp_twsk_unique,
907 .twsk_destructor= tcp_twsk_destructor,
910 static void tcp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
912 struct ipv6_pinfo *np = inet6_sk(sk);
913 struct tcphdr *th = tcp_hdr(skb);
915 if (skb->ip_summed == CHECKSUM_PARTIAL) {
916 th->check = ~csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP, 0);
917 skb->csum_start = skb_transport_header(skb) - skb->head;
918 skb->csum_offset = offsetof(struct tcphdr, check);
919 } else {
920 th->check = csum_ipv6_magic(&np->saddr, &np->daddr, len, IPPROTO_TCP,
921 csum_partial(th, th->doff<<2,
922 skb->csum));
926 static int tcp_v6_gso_send_check(struct sk_buff *skb)
928 struct ipv6hdr *ipv6h;
929 struct tcphdr *th;
931 if (!pskb_may_pull(skb, sizeof(*th)))
932 return -EINVAL;
934 ipv6h = ipv6_hdr(skb);
935 th = tcp_hdr(skb);
937 th->check = 0;
938 th->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
939 IPPROTO_TCP, 0);
940 skb->csum_start = skb_transport_header(skb) - skb->head;
941 skb->csum_offset = offsetof(struct tcphdr, check);
942 skb->ip_summed = CHECKSUM_PARTIAL;
943 return 0;
946 static struct sk_buff **tcp6_gro_receive(struct sk_buff **head,
947 struct sk_buff *skb)
949 struct ipv6hdr *iph = skb_gro_network_header(skb);
951 switch (skb->ip_summed) {
952 case CHECKSUM_COMPLETE:
953 if (!tcp_v6_check(skb_gro_len(skb), &iph->saddr, &iph->daddr,
954 skb->csum)) {
955 skb->ip_summed = CHECKSUM_UNNECESSARY;
956 break;
959 /* fall through */
960 case CHECKSUM_NONE:
961 NAPI_GRO_CB(skb)->flush = 1;
962 return NULL;
965 return tcp_gro_receive(head, skb);
968 static int tcp6_gro_complete(struct sk_buff *skb)
970 struct ipv6hdr *iph = ipv6_hdr(skb);
971 struct tcphdr *th = tcp_hdr(skb);
973 th->check = ~tcp_v6_check(skb->len - skb_transport_offset(skb),
974 &iph->saddr, &iph->daddr, 0);
975 skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
977 return tcp_gro_complete(skb);
980 static void tcp_v6_send_response(struct sk_buff *skb, u32 seq, u32 ack, u32 win,
981 u32 ts, struct tcp_md5sig_key *key, int rst)
983 struct tcphdr *th = tcp_hdr(skb), *t1;
984 struct sk_buff *buff;
985 struct flowi fl;
986 struct net *net = dev_net(skb_dst(skb)->dev);
987 struct sock *ctl_sk = net->ipv6.tcp_sk;
988 unsigned int tot_len = sizeof(struct tcphdr);
989 struct dst_entry *dst;
990 __be32 *topt;
992 if (ts)
993 tot_len += TCPOLEN_TSTAMP_ALIGNED;
994 #ifdef CONFIG_TCP_MD5SIG
995 if (key)
996 tot_len += TCPOLEN_MD5SIG_ALIGNED;
997 #endif
999 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1000 GFP_ATOMIC);
1001 if (buff == NULL)
1002 return;
1004 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1006 t1 = (struct tcphdr *) skb_push(buff, tot_len);
1007 skb_reset_transport_header(skb);
1009 /* Swap the send and the receive. */
1010 memset(t1, 0, sizeof(*t1));
1011 t1->dest = th->source;
1012 t1->source = th->dest;
1013 t1->doff = tot_len / 4;
1014 t1->seq = htonl(seq);
1015 t1->ack_seq = htonl(ack);
1016 t1->ack = !rst || !th->ack;
1017 t1->rst = rst;
1018 t1->window = htons(win);
1020 topt = (__be32 *)(t1 + 1);
1022 if (ts) {
1023 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1024 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
1025 *topt++ = htonl(tcp_time_stamp);
1026 *topt++ = htonl(ts);
1029 #ifdef CONFIG_TCP_MD5SIG
1030 if (key) {
1031 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1032 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
1033 tcp_v6_md5_hash_hdr((__u8 *)topt, key,
1034 &ipv6_hdr(skb)->saddr,
1035 &ipv6_hdr(skb)->daddr, t1);
1037 #endif
1039 buff->csum = csum_partial(t1, tot_len, 0);
1041 memset(&fl, 0, sizeof(fl));
1042 ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
1043 ipv6_addr_copy(&fl.fl6_src, &ipv6_hdr(skb)->daddr);
1045 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1046 tot_len, IPPROTO_TCP,
1047 buff->csum);
1049 fl.proto = IPPROTO_TCP;
1050 fl.oif = inet6_iif(skb);
1051 fl.fl_ip_dport = t1->dest;
1052 fl.fl_ip_sport = t1->source;
1053 security_skb_classify_flow(skb, &fl);
1055 /* Pass a socket to ip6_dst_lookup either it is for RST
1056 * Underlying function will use this to retrieve the network
1057 * namespace
1059 if (!ip6_dst_lookup(ctl_sk, &dst, &fl)) {
1060 if (xfrm_lookup(net, &dst, &fl, NULL, 0) >= 0) {
1061 skb_dst_set(buff, dst);
1062 ip6_xmit(ctl_sk, buff, &fl, NULL, 0);
1063 TCP_INC_STATS_BH(net, TCP_MIB_OUTSEGS);
1064 if (rst)
1065 TCP_INC_STATS_BH(net, TCP_MIB_OUTRSTS);
1066 return;
1070 kfree_skb(buff);
1073 static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
1075 struct tcphdr *th = tcp_hdr(skb);
1076 u32 seq = 0, ack_seq = 0;
1077 struct tcp_md5sig_key *key = NULL;
1079 if (th->rst)
1080 return;
1082 if (!ipv6_unicast_destination(skb))
1083 return;
1085 #ifdef CONFIG_TCP_MD5SIG
1086 if (sk)
1087 key = tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr);
1088 #endif
1090 if (th->ack)
1091 seq = ntohl(th->ack_seq);
1092 else
1093 ack_seq = ntohl(th->seq) + th->syn + th->fin + skb->len -
1094 (th->doff << 2);
1096 tcp_v6_send_response(skb, seq, ack_seq, 0, 0, key, 1);
1099 static void tcp_v6_send_ack(struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts,
1100 struct tcp_md5sig_key *key)
1102 tcp_v6_send_response(skb, seq, ack, win, ts, key, 0);
1105 static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
1107 struct inet_timewait_sock *tw = inet_twsk(sk);
1108 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
1110 tcp_v6_send_ack(skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
1111 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
1112 tcptw->tw_ts_recent, tcp_twsk_md5_key(tcptw));
1114 inet_twsk_put(tw);
1117 static void tcp_v6_reqsk_send_ack(struct sock *sk, struct sk_buff *skb,
1118 struct request_sock *req)
1120 tcp_v6_send_ack(skb, tcp_rsk(req)->snt_isn + 1, tcp_rsk(req)->rcv_isn + 1, req->rcv_wnd, req->ts_recent,
1121 tcp_v6_md5_do_lookup(sk, &ipv6_hdr(skb)->daddr));
1125 static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
1127 struct request_sock *req, **prev;
1128 const struct tcphdr *th = tcp_hdr(skb);
1129 struct sock *nsk;
1131 /* Find possible connection requests. */
1132 req = inet6_csk_search_req(sk, &prev, th->source,
1133 &ipv6_hdr(skb)->saddr,
1134 &ipv6_hdr(skb)->daddr, inet6_iif(skb));
1135 if (req)
1136 return tcp_check_req(sk, skb, req, prev);
1138 nsk = __inet6_lookup_established(sock_net(sk), &tcp_hashinfo,
1139 &ipv6_hdr(skb)->saddr, th->source,
1140 &ipv6_hdr(skb)->daddr, ntohs(th->dest), inet6_iif(skb));
1142 if (nsk) {
1143 if (nsk->sk_state != TCP_TIME_WAIT) {
1144 bh_lock_sock(nsk);
1145 return nsk;
1147 inet_twsk_put(inet_twsk(nsk));
1148 return NULL;
1151 #ifdef CONFIG_SYN_COOKIES
1152 if (!th->rst && !th->syn && th->ack)
1153 sk = cookie_v6_check(sk, skb);
1154 #endif
1155 return sk;
1158 /* FIXME: this is substantially similar to the ipv4 code.
1159 * Can some kind of merge be done? -- erics
1161 static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
1163 struct inet6_request_sock *treq;
1164 struct ipv6_pinfo *np = inet6_sk(sk);
1165 struct tcp_options_received tmp_opt;
1166 struct tcp_sock *tp = tcp_sk(sk);
1167 struct request_sock *req = NULL;
1168 __u32 isn = TCP_SKB_CB(skb)->when;
1169 #ifdef CONFIG_SYN_COOKIES
1170 int want_cookie = 0;
1171 #else
1172 #define want_cookie 0
1173 #endif
1175 if (skb->protocol == htons(ETH_P_IP))
1176 return tcp_v4_conn_request(sk, skb);
1178 if (!ipv6_unicast_destination(skb))
1179 goto drop;
1181 if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
1182 if (net_ratelimit())
1183 syn_flood_warning(skb);
1184 #ifdef CONFIG_SYN_COOKIES
1185 if (sysctl_tcp_syncookies)
1186 want_cookie = 1;
1187 else
1188 #endif
1189 goto drop;
1192 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
1193 goto drop;
1195 req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
1196 if (req == NULL)
1197 goto drop;
1199 #ifdef CONFIG_TCP_MD5SIG
1200 tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
1201 #endif
1203 tcp_clear_options(&tmp_opt);
1204 tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
1205 tmp_opt.user_mss = tp->rx_opt.user_mss;
1207 tcp_parse_options(skb, &tmp_opt, 0);
1209 if (want_cookie && !tmp_opt.saw_tstamp)
1210 tcp_clear_options(&tmp_opt);
1212 tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
1213 tcp_openreq_init(req, &tmp_opt, skb);
1215 treq = inet6_rsk(req);
1216 ipv6_addr_copy(&treq->rmt_addr, &ipv6_hdr(skb)->saddr);
1217 ipv6_addr_copy(&treq->loc_addr, &ipv6_hdr(skb)->daddr);
1218 if (!want_cookie)
1219 TCP_ECN_create_request(req, tcp_hdr(skb));
1221 if (want_cookie) {
1222 isn = cookie_v6_init_sequence(sk, skb, &req->mss);
1223 req->cookie_ts = tmp_opt.tstamp_ok;
1224 } else if (!isn) {
1225 if (ipv6_opt_accepted(sk, skb) ||
1226 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
1227 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
1228 atomic_inc(&skb->users);
1229 treq->pktopts = skb;
1231 treq->iif = sk->sk_bound_dev_if;
1233 /* So that link locals have meaning */
1234 if (!sk->sk_bound_dev_if &&
1235 ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL)
1236 treq->iif = inet6_iif(skb);
1238 isn = tcp_v6_init_sequence(skb);
1241 tcp_rsk(req)->snt_isn = isn;
1243 security_inet_conn_request(sk, skb, req);
1245 if (tcp_v6_send_synack(sk, req))
1246 goto drop;
1248 if (!want_cookie) {
1249 inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
1250 return 0;
1253 drop:
1254 if (req)
1255 reqsk_free(req);
1257 return 0; /* don't send reset */
1260 static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
1261 struct request_sock *req,
1262 struct dst_entry *dst)
1264 struct inet6_request_sock *treq;
1265 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
1266 struct tcp6_sock *newtcp6sk;
1267 struct inet_sock *newinet;
1268 struct tcp_sock *newtp;
1269 struct sock *newsk;
1270 struct ipv6_txoptions *opt;
1271 #ifdef CONFIG_TCP_MD5SIG
1272 struct tcp_md5sig_key *key;
1273 #endif
1275 if (skb->protocol == htons(ETH_P_IP)) {
1277 * v6 mapped
1280 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1282 if (newsk == NULL)
1283 return NULL;
1285 newtcp6sk = (struct tcp6_sock *)newsk;
1286 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1288 newinet = inet_sk(newsk);
1289 newnp = inet6_sk(newsk);
1290 newtp = tcp_sk(newsk);
1292 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1294 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
1295 newinet->daddr);
1297 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
1298 newinet->saddr);
1300 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
1302 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1303 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1304 #ifdef CONFIG_TCP_MD5SIG
1305 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1306 #endif
1308 newnp->pktoptions = NULL;
1309 newnp->opt = NULL;
1310 newnp->mcast_oif = inet6_iif(skb);
1311 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1314 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1315 * here, tcp_create_openreq_child now does this for us, see the comment in
1316 * that function for the gory details. -acme
1319 /* It is tricky place. Until this moment IPv4 tcp
1320 worked with IPv6 icsk.icsk_af_ops.
1321 Sync it now.
1323 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1325 return newsk;
1328 treq = inet6_rsk(req);
1329 opt = np->opt;
1331 if (sk_acceptq_is_full(sk))
1332 goto out_overflow;
1334 if (dst == NULL) {
1335 struct in6_addr *final_p = NULL, final;
1336 struct flowi fl;
1338 memset(&fl, 0, sizeof(fl));
1339 fl.proto = IPPROTO_TCP;
1340 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
1341 if (opt && opt->srcrt) {
1342 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
1343 ipv6_addr_copy(&final, &fl.fl6_dst);
1344 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
1345 final_p = &final;
1347 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
1348 fl.oif = sk->sk_bound_dev_if;
1349 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
1350 fl.fl_ip_sport = inet_rsk(req)->loc_port;
1351 security_req_classify_flow(req, &fl);
1353 if (ip6_dst_lookup(sk, &dst, &fl))
1354 goto out;
1356 if (final_p)
1357 ipv6_addr_copy(&fl.fl6_dst, final_p);
1359 if ((xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0)
1360 goto out;
1363 newsk = tcp_create_openreq_child(sk, req, skb);
1364 if (newsk == NULL)
1365 goto out;
1368 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1369 * count here, tcp_create_openreq_child now does this for us, see the
1370 * comment in that function for the gory details. -acme
1373 newsk->sk_gso_type = SKB_GSO_TCPV6;
1374 __ip6_dst_store(newsk, dst, NULL, NULL);
1376 newtcp6sk = (struct tcp6_sock *)newsk;
1377 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1379 newtp = tcp_sk(newsk);
1380 newinet = inet_sk(newsk);
1381 newnp = inet6_sk(newsk);
1383 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1385 ipv6_addr_copy(&newnp->daddr, &treq->rmt_addr);
1386 ipv6_addr_copy(&newnp->saddr, &treq->loc_addr);
1387 ipv6_addr_copy(&newnp->rcv_saddr, &treq->loc_addr);
1388 newsk->sk_bound_dev_if = treq->iif;
1390 /* Now IPv6 options...
1392 First: no IPv4 options.
1394 newinet->opt = NULL;
1395 newnp->ipv6_fl_list = NULL;
1397 /* Clone RX bits */
1398 newnp->rxopt.all = np->rxopt.all;
1400 /* Clone pktoptions received with SYN */
1401 newnp->pktoptions = NULL;
1402 if (treq->pktopts != NULL) {
1403 newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC);
1404 kfree_skb(treq->pktopts);
1405 treq->pktopts = NULL;
1406 if (newnp->pktoptions)
1407 skb_set_owner_r(newnp->pktoptions, newsk);
1409 newnp->opt = NULL;
1410 newnp->mcast_oif = inet6_iif(skb);
1411 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1413 /* Clone native IPv6 options from listening socket (if any)
1415 Yes, keeping reference count would be much more clever,
1416 but we make one more one thing there: reattach optmem
1417 to newsk.
1419 if (opt) {
1420 newnp->opt = ipv6_dup_options(newsk, opt);
1421 if (opt != np->opt)
1422 sock_kfree_s(sk, opt, opt->tot_len);
1425 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1426 if (newnp->opt)
1427 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1428 newnp->opt->opt_flen);
1430 tcp_mtup_init(newsk);
1431 tcp_sync_mss(newsk, dst_mtu(dst));
1432 newtp->advmss = dst_metric(dst, RTAX_ADVMSS);
1433 tcp_initialize_rcv_mss(newsk);
1435 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
1437 #ifdef CONFIG_TCP_MD5SIG
1438 /* Copy over the MD5 key from the original socket */
1439 if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) {
1440 /* We're using one, so create a matching key
1441 * on the newsk structure. If we fail to get
1442 * memory, then we end up not copying the key
1443 * across. Shucks.
1445 char *newkey = kmemdup(key->key, key->keylen, GFP_ATOMIC);
1446 if (newkey != NULL)
1447 tcp_v6_md5_do_add(newsk, &newnp->daddr,
1448 newkey, key->keylen);
1450 #endif
1452 __inet6_hash(newsk);
1453 __inet_inherit_port(sk, newsk);
1455 return newsk;
1457 out_overflow:
1458 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
1459 out:
1460 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
1461 if (opt && opt != np->opt)
1462 sock_kfree_s(sk, opt, opt->tot_len);
1463 dst_release(dst);
1464 return NULL;
1467 static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
1469 if (skb->ip_summed == CHECKSUM_COMPLETE) {
1470 if (!tcp_v6_check(skb->len, &ipv6_hdr(skb)->saddr,
1471 &ipv6_hdr(skb)->daddr, skb->csum)) {
1472 skb->ip_summed = CHECKSUM_UNNECESSARY;
1473 return 0;
1477 skb->csum = ~csum_unfold(tcp_v6_check(skb->len,
1478 &ipv6_hdr(skb)->saddr,
1479 &ipv6_hdr(skb)->daddr, 0));
1481 if (skb->len <= 76) {
1482 return __skb_checksum_complete(skb);
1484 return 0;
1487 /* The socket must have it's spinlock held when we get
1488 * here.
1490 * We have a potential double-lock case here, so even when
1491 * doing backlog processing we use the BH locking scheme.
1492 * This is because we cannot sleep with the original spinlock
1493 * held.
1495 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1497 struct ipv6_pinfo *np = inet6_sk(sk);
1498 struct tcp_sock *tp;
1499 struct sk_buff *opt_skb = NULL;
1501 /* Imagine: socket is IPv6. IPv4 packet arrives,
1502 goes to IPv4 receive handler and backlogged.
1503 From backlog it always goes here. Kerboom...
1504 Fortunately, tcp_rcv_established and rcv_established
1505 handle them correctly, but it is not case with
1506 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1509 if (skb->protocol == htons(ETH_P_IP))
1510 return tcp_v4_do_rcv(sk, skb);
1512 #ifdef CONFIG_TCP_MD5SIG
1513 if (tcp_v6_inbound_md5_hash (sk, skb))
1514 goto discard;
1515 #endif
1517 if (sk_filter(sk, skb))
1518 goto discard;
1521 * socket locking is here for SMP purposes as backlog rcv
1522 * is currently called with bh processing disabled.
1525 /* Do Stevens' IPV6_PKTOPTIONS.
1527 Yes, guys, it is the only place in our code, where we
1528 may make it not affecting IPv4.
1529 The rest of code is protocol independent,
1530 and I do not like idea to uglify IPv4.
1532 Actually, all the idea behind IPV6_PKTOPTIONS
1533 looks not very well thought. For now we latch
1534 options, received in the last packet, enqueued
1535 by tcp. Feel free to propose better solution.
1536 --ANK (980728)
1538 if (np->rxopt.all)
1539 opt_skb = skb_clone(skb, GFP_ATOMIC);
1541 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1542 TCP_CHECK_TIMER(sk);
1543 if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len))
1544 goto reset;
1545 TCP_CHECK_TIMER(sk);
1546 if (opt_skb)
1547 goto ipv6_pktoptions;
1548 return 0;
1551 if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
1552 goto csum_err;
1554 if (sk->sk_state == TCP_LISTEN) {
1555 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1556 if (!nsk)
1557 goto discard;
1560 * Queue it on the new socket if the new socket is active,
1561 * otherwise we just shortcircuit this and continue with
1562 * the new socket..
1564 if(nsk != sk) {
1565 if (tcp_child_process(sk, nsk, skb))
1566 goto reset;
1567 if (opt_skb)
1568 __kfree_skb(opt_skb);
1569 return 0;
1573 TCP_CHECK_TIMER(sk);
1574 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
1575 goto reset;
1576 TCP_CHECK_TIMER(sk);
1577 if (opt_skb)
1578 goto ipv6_pktoptions;
1579 return 0;
1581 reset:
1582 tcp_v6_send_reset(sk, skb);
1583 discard:
1584 if (opt_skb)
1585 __kfree_skb(opt_skb);
1586 kfree_skb(skb);
1587 return 0;
1588 csum_err:
1589 TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
1590 goto discard;
1593 ipv6_pktoptions:
1594 /* Do you ask, what is it?
1596 1. skb was enqueued by tcp.
1597 2. skb is added to tail of read queue, rather than out of order.
1598 3. socket is not in passive state.
1599 4. Finally, it really contains options, which user wants to receive.
1601 tp = tcp_sk(sk);
1602 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1603 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
1604 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
1605 np->mcast_oif = inet6_iif(opt_skb);
1606 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1607 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1608 if (ipv6_opt_accepted(sk, opt_skb)) {
1609 skb_set_owner_r(opt_skb, sk);
1610 opt_skb = xchg(&np->pktoptions, opt_skb);
1611 } else {
1612 __kfree_skb(opt_skb);
1613 opt_skb = xchg(&np->pktoptions, NULL);
1617 kfree_skb(opt_skb);
1618 return 0;
1621 static int tcp_v6_rcv(struct sk_buff *skb)
1623 struct tcphdr *th;
1624 struct sock *sk;
1625 int ret;
1626 struct net *net = dev_net(skb->dev);
1628 if (skb->pkt_type != PACKET_HOST)
1629 goto discard_it;
1632 * Count it even if it's bad.
1634 TCP_INC_STATS_BH(net, TCP_MIB_INSEGS);
1636 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1637 goto discard_it;
1639 th = tcp_hdr(skb);
1641 if (th->doff < sizeof(struct tcphdr)/4)
1642 goto bad_packet;
1643 if (!pskb_may_pull(skb, th->doff*4))
1644 goto discard_it;
1646 if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
1647 goto bad_packet;
1649 th = tcp_hdr(skb);
1650 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1651 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1652 skb->len - th->doff*4);
1653 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1654 TCP_SKB_CB(skb)->when = 0;
1655 TCP_SKB_CB(skb)->flags = ipv6_get_dsfield(ipv6_hdr(skb));
1656 TCP_SKB_CB(skb)->sacked = 0;
1658 sk = __inet6_lookup_skb(&tcp_hashinfo, skb, th->source, th->dest);
1659 if (!sk)
1660 goto no_tcp_socket;
1662 process:
1663 if (sk->sk_state == TCP_TIME_WAIT)
1664 goto do_time_wait;
1666 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1667 goto discard_and_relse;
1669 if (sk_filter(sk, skb))
1670 goto discard_and_relse;
1672 skb->dev = NULL;
1674 bh_lock_sock_nested(sk);
1675 ret = 0;
1676 if (!sock_owned_by_user(sk)) {
1677 #ifdef CONFIG_NET_DMA
1678 struct tcp_sock *tp = tcp_sk(sk);
1679 if (!tp->ucopy.dma_chan && tp->ucopy.pinned_list)
1680 tp->ucopy.dma_chan = dma_find_channel(DMA_MEMCPY);
1681 if (tp->ucopy.dma_chan)
1682 ret = tcp_v6_do_rcv(sk, skb);
1683 else
1684 #endif
1686 if (!tcp_prequeue(sk, skb))
1687 ret = tcp_v6_do_rcv(sk, skb);
1689 } else
1690 sk_add_backlog(sk, skb);
1691 bh_unlock_sock(sk);
1693 sock_put(sk);
1694 return ret ? -1 : 0;
1696 no_tcp_socket:
1697 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1698 goto discard_it;
1700 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1701 bad_packet:
1702 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1703 } else {
1704 tcp_v6_send_reset(NULL, skb);
1707 discard_it:
1710 * Discard frame
1713 kfree_skb(skb);
1714 return 0;
1716 discard_and_relse:
1717 sock_put(sk);
1718 goto discard_it;
1720 do_time_wait:
1721 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1722 inet_twsk_put(inet_twsk(sk));
1723 goto discard_it;
1726 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1727 TCP_INC_STATS_BH(net, TCP_MIB_INERRS);
1728 inet_twsk_put(inet_twsk(sk));
1729 goto discard_it;
1732 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1733 case TCP_TW_SYN:
1735 struct sock *sk2;
1737 sk2 = inet6_lookup_listener(dev_net(skb->dev), &tcp_hashinfo,
1738 &ipv6_hdr(skb)->daddr,
1739 ntohs(th->dest), inet6_iif(skb));
1740 if (sk2 != NULL) {
1741 struct inet_timewait_sock *tw = inet_twsk(sk);
1742 inet_twsk_deschedule(tw, &tcp_death_row);
1743 inet_twsk_put(tw);
1744 sk = sk2;
1745 goto process;
1747 /* Fall through to ACK */
1749 case TCP_TW_ACK:
1750 tcp_v6_timewait_ack(sk, skb);
1751 break;
1752 case TCP_TW_RST:
1753 goto no_tcp_socket;
1754 case TCP_TW_SUCCESS:;
1756 goto discard_it;
1759 static int tcp_v6_remember_stamp(struct sock *sk)
1761 /* Alas, not yet... */
1762 return 0;
1765 static const struct inet_connection_sock_af_ops ipv6_specific = {
1766 .queue_xmit = inet6_csk_xmit,
1767 .send_check = tcp_v6_send_check,
1768 .rebuild_header = inet6_sk_rebuild_header,
1769 .conn_request = tcp_v6_conn_request,
1770 .syn_recv_sock = tcp_v6_syn_recv_sock,
1771 .remember_stamp = tcp_v6_remember_stamp,
1772 .net_header_len = sizeof(struct ipv6hdr),
1773 .setsockopt = ipv6_setsockopt,
1774 .getsockopt = ipv6_getsockopt,
1775 .addr2sockaddr = inet6_csk_addr2sockaddr,
1776 .sockaddr_len = sizeof(struct sockaddr_in6),
1777 .bind_conflict = inet6_csk_bind_conflict,
1778 #ifdef CONFIG_COMPAT
1779 .compat_setsockopt = compat_ipv6_setsockopt,
1780 .compat_getsockopt = compat_ipv6_getsockopt,
1781 #endif
1784 #ifdef CONFIG_TCP_MD5SIG
1785 static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1786 .md5_lookup = tcp_v6_md5_lookup,
1787 .calc_md5_hash = tcp_v6_md5_hash_skb,
1788 .md5_add = tcp_v6_md5_add_func,
1789 .md5_parse = tcp_v6_parse_md5_keys,
1791 #endif
1794 * TCP over IPv4 via INET6 API
1797 static const struct inet_connection_sock_af_ops ipv6_mapped = {
1798 .queue_xmit = ip_queue_xmit,
1799 .send_check = tcp_v4_send_check,
1800 .rebuild_header = inet_sk_rebuild_header,
1801 .conn_request = tcp_v6_conn_request,
1802 .syn_recv_sock = tcp_v6_syn_recv_sock,
1803 .remember_stamp = tcp_v4_remember_stamp,
1804 .net_header_len = sizeof(struct iphdr),
1805 .setsockopt = ipv6_setsockopt,
1806 .getsockopt = ipv6_getsockopt,
1807 .addr2sockaddr = inet6_csk_addr2sockaddr,
1808 .sockaddr_len = sizeof(struct sockaddr_in6),
1809 .bind_conflict = inet6_csk_bind_conflict,
1810 #ifdef CONFIG_COMPAT
1811 .compat_setsockopt = compat_ipv6_setsockopt,
1812 .compat_getsockopt = compat_ipv6_getsockopt,
1813 #endif
1816 #ifdef CONFIG_TCP_MD5SIG
1817 static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1818 .md5_lookup = tcp_v4_md5_lookup,
1819 .calc_md5_hash = tcp_v4_md5_hash_skb,
1820 .md5_add = tcp_v6_md5_add_func,
1821 .md5_parse = tcp_v6_parse_md5_keys,
1823 #endif
1825 /* NOTE: A lot of things set to zero explicitly by call to
1826 * sk_alloc() so need not be done here.
1828 static int tcp_v6_init_sock(struct sock *sk)
1830 struct inet_connection_sock *icsk = inet_csk(sk);
1831 struct tcp_sock *tp = tcp_sk(sk);
1833 skb_queue_head_init(&tp->out_of_order_queue);
1834 tcp_init_xmit_timers(sk);
1835 tcp_prequeue_init(tp);
1837 icsk->icsk_rto = TCP_TIMEOUT_INIT;
1838 tp->mdev = TCP_TIMEOUT_INIT;
1840 /* So many TCP implementations out there (incorrectly) count the
1841 * initial SYN frame in their delayed-ACK and congestion control
1842 * algorithms that we must have the following bandaid to talk
1843 * efficiently to them. -DaveM
1845 tp->snd_cwnd = 2;
1847 /* See draft-stevens-tcpca-spec-01 for discussion of the
1848 * initialization of these values.
1850 tp->snd_ssthresh = TCP_INFINITE_SSTHRESH;
1851 tp->snd_cwnd_clamp = ~0;
1852 tp->mss_cache = 536;
1854 tp->reordering = sysctl_tcp_reordering;
1856 sk->sk_state = TCP_CLOSE;
1858 icsk->icsk_af_ops = &ipv6_specific;
1859 icsk->icsk_ca_ops = &tcp_init_congestion_ops;
1860 icsk->icsk_sync_mss = tcp_sync_mss;
1861 sk->sk_write_space = sk_stream_write_space;
1862 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
1864 #ifdef CONFIG_TCP_MD5SIG
1865 tp->af_specific = &tcp_sock_ipv6_specific;
1866 #endif
1868 sk->sk_sndbuf = sysctl_tcp_wmem[1];
1869 sk->sk_rcvbuf = sysctl_tcp_rmem[1];
1871 local_bh_disable();
1872 percpu_counter_inc(&tcp_sockets_allocated);
1873 local_bh_enable();
1875 return 0;
1878 static void tcp_v6_destroy_sock(struct sock *sk)
1880 #ifdef CONFIG_TCP_MD5SIG
1881 /* Clean up the MD5 key list */
1882 if (tcp_sk(sk)->md5sig_info)
1883 tcp_v6_clear_md5_list(sk);
1884 #endif
1885 tcp_v4_destroy_sock(sk);
1886 inet6_destroy_sock(sk);
1889 #ifdef CONFIG_PROC_FS
1890 /* Proc filesystem TCPv6 sock list dumping. */
1891 static void get_openreq6(struct seq_file *seq,
1892 struct sock *sk, struct request_sock *req, int i, int uid)
1894 int ttd = req->expires - jiffies;
1895 struct in6_addr *src = &inet6_rsk(req)->loc_addr;
1896 struct in6_addr *dest = &inet6_rsk(req)->rmt_addr;
1898 if (ttd < 0)
1899 ttd = 0;
1901 seq_printf(seq,
1902 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1903 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
1905 src->s6_addr32[0], src->s6_addr32[1],
1906 src->s6_addr32[2], src->s6_addr32[3],
1907 ntohs(inet_rsk(req)->loc_port),
1908 dest->s6_addr32[0], dest->s6_addr32[1],
1909 dest->s6_addr32[2], dest->s6_addr32[3],
1910 ntohs(inet_rsk(req)->rmt_port),
1911 TCP_SYN_RECV,
1912 0,0, /* could print option size, but that is af dependent. */
1913 1, /* timers active (only the expire timer) */
1914 jiffies_to_clock_t(ttd),
1915 req->retrans,
1916 uid,
1917 0, /* non standard timer */
1918 0, /* open_requests have no inode */
1919 0, req);
1922 static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1924 struct in6_addr *dest, *src;
1925 __u16 destp, srcp;
1926 int timer_active;
1927 unsigned long timer_expires;
1928 struct inet_sock *inet = inet_sk(sp);
1929 struct tcp_sock *tp = tcp_sk(sp);
1930 const struct inet_connection_sock *icsk = inet_csk(sp);
1931 struct ipv6_pinfo *np = inet6_sk(sp);
1933 dest = &np->daddr;
1934 src = &np->rcv_saddr;
1935 destp = ntohs(inet->dport);
1936 srcp = ntohs(inet->sport);
1938 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
1939 timer_active = 1;
1940 timer_expires = icsk->icsk_timeout;
1941 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
1942 timer_active = 4;
1943 timer_expires = icsk->icsk_timeout;
1944 } else if (timer_pending(&sp->sk_timer)) {
1945 timer_active = 2;
1946 timer_expires = sp->sk_timer.expires;
1947 } else {
1948 timer_active = 0;
1949 timer_expires = jiffies;
1952 seq_printf(seq,
1953 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1954 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %lu %lu %u %u %d\n",
1956 src->s6_addr32[0], src->s6_addr32[1],
1957 src->s6_addr32[2], src->s6_addr32[3], srcp,
1958 dest->s6_addr32[0], dest->s6_addr32[1],
1959 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1960 sp->sk_state,
1961 tp->write_seq-tp->snd_una,
1962 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
1963 timer_active,
1964 jiffies_to_clock_t(timer_expires - jiffies),
1965 icsk->icsk_retransmits,
1966 sock_i_uid(sp),
1967 icsk->icsk_probes_out,
1968 sock_i_ino(sp),
1969 atomic_read(&sp->sk_refcnt), sp,
1970 jiffies_to_clock_t(icsk->icsk_rto),
1971 jiffies_to_clock_t(icsk->icsk_ack.ato),
1972 (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
1973 tp->snd_cwnd,
1974 tcp_in_initial_slowstart(tp) ? -1 : tp->snd_ssthresh
1978 static void get_timewait6_sock(struct seq_file *seq,
1979 struct inet_timewait_sock *tw, int i)
1981 struct in6_addr *dest, *src;
1982 __u16 destp, srcp;
1983 struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
1984 int ttd = tw->tw_ttd - jiffies;
1986 if (ttd < 0)
1987 ttd = 0;
1989 dest = &tw6->tw_v6_daddr;
1990 src = &tw6->tw_v6_rcv_saddr;
1991 destp = ntohs(tw->tw_dport);
1992 srcp = ntohs(tw->tw_sport);
1994 seq_printf(seq,
1995 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1996 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
1998 src->s6_addr32[0], src->s6_addr32[1],
1999 src->s6_addr32[2], src->s6_addr32[3], srcp,
2000 dest->s6_addr32[0], dest->s6_addr32[1],
2001 dest->s6_addr32[2], dest->s6_addr32[3], destp,
2002 tw->tw_substate, 0, 0,
2003 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
2004 atomic_read(&tw->tw_refcnt), tw);
2007 static int tcp6_seq_show(struct seq_file *seq, void *v)
2009 struct tcp_iter_state *st;
2011 if (v == SEQ_START_TOKEN) {
2012 seq_puts(seq,
2013 " sl "
2014 "local_address "
2015 "remote_address "
2016 "st tx_queue rx_queue tr tm->when retrnsmt"
2017 " uid timeout inode\n");
2018 goto out;
2020 st = seq->private;
2022 switch (st->state) {
2023 case TCP_SEQ_STATE_LISTENING:
2024 case TCP_SEQ_STATE_ESTABLISHED:
2025 get_tcp6_sock(seq, v, st->num);
2026 break;
2027 case TCP_SEQ_STATE_OPENREQ:
2028 get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
2029 break;
2030 case TCP_SEQ_STATE_TIME_WAIT:
2031 get_timewait6_sock(seq, v, st->num);
2032 break;
2034 out:
2035 return 0;
2038 static struct tcp_seq_afinfo tcp6_seq_afinfo = {
2039 .name = "tcp6",
2040 .family = AF_INET6,
2041 .seq_fops = {
2042 .owner = THIS_MODULE,
2044 .seq_ops = {
2045 .show = tcp6_seq_show,
2049 int tcp6_proc_init(struct net *net)
2051 return tcp_proc_register(net, &tcp6_seq_afinfo);
2054 void tcp6_proc_exit(struct net *net)
2056 tcp_proc_unregister(net, &tcp6_seq_afinfo);
2058 #endif
2060 struct proto tcpv6_prot = {
2061 .name = "TCPv6",
2062 .owner = THIS_MODULE,
2063 .close = tcp_close,
2064 .connect = tcp_v6_connect,
2065 .disconnect = tcp_disconnect,
2066 .accept = inet_csk_accept,
2067 .ioctl = tcp_ioctl,
2068 .init = tcp_v6_init_sock,
2069 .destroy = tcp_v6_destroy_sock,
2070 .shutdown = tcp_shutdown,
2071 .setsockopt = tcp_setsockopt,
2072 .getsockopt = tcp_getsockopt,
2073 .recvmsg = tcp_recvmsg,
2074 .backlog_rcv = tcp_v6_do_rcv,
2075 .hash = tcp_v6_hash,
2076 .unhash = inet_unhash,
2077 .get_port = inet_csk_get_port,
2078 .enter_memory_pressure = tcp_enter_memory_pressure,
2079 .sockets_allocated = &tcp_sockets_allocated,
2080 .memory_allocated = &tcp_memory_allocated,
2081 .memory_pressure = &tcp_memory_pressure,
2082 .orphan_count = &tcp_orphan_count,
2083 .sysctl_mem = sysctl_tcp_mem,
2084 .sysctl_wmem = sysctl_tcp_wmem,
2085 .sysctl_rmem = sysctl_tcp_rmem,
2086 .max_header = MAX_TCP_HEADER,
2087 .obj_size = sizeof(struct tcp6_sock),
2088 .slab_flags = SLAB_DESTROY_BY_RCU,
2089 .twsk_prot = &tcp6_timewait_sock_ops,
2090 .rsk_prot = &tcp6_request_sock_ops,
2091 .h.hashinfo = &tcp_hashinfo,
2092 #ifdef CONFIG_COMPAT
2093 .compat_setsockopt = compat_tcp_setsockopt,
2094 .compat_getsockopt = compat_tcp_getsockopt,
2095 #endif
2098 static const struct inet6_protocol tcpv6_protocol = {
2099 .handler = tcp_v6_rcv,
2100 .err_handler = tcp_v6_err,
2101 .gso_send_check = tcp_v6_gso_send_check,
2102 .gso_segment = tcp_tso_segment,
2103 .gro_receive = tcp6_gro_receive,
2104 .gro_complete = tcp6_gro_complete,
2105 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
2108 static struct inet_protosw tcpv6_protosw = {
2109 .type = SOCK_STREAM,
2110 .protocol = IPPROTO_TCP,
2111 .prot = &tcpv6_prot,
2112 .ops = &inet6_stream_ops,
2113 .capability = -1,
2114 .no_check = 0,
2115 .flags = INET_PROTOSW_PERMANENT |
2116 INET_PROTOSW_ICSK,
2119 static int tcpv6_net_init(struct net *net)
2121 return inet_ctl_sock_create(&net->ipv6.tcp_sk, PF_INET6,
2122 SOCK_RAW, IPPROTO_TCP, net);
2125 static void tcpv6_net_exit(struct net *net)
2127 inet_ctl_sock_destroy(net->ipv6.tcp_sk);
2128 inet_twsk_purge(net, &tcp_hashinfo, &tcp_death_row, AF_INET6);
2131 static struct pernet_operations tcpv6_net_ops = {
2132 .init = tcpv6_net_init,
2133 .exit = tcpv6_net_exit,
2136 int __init tcpv6_init(void)
2138 int ret;
2140 ret = inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP);
2141 if (ret)
2142 goto out;
2144 /* register inet6 protocol */
2145 ret = inet6_register_protosw(&tcpv6_protosw);
2146 if (ret)
2147 goto out_tcpv6_protocol;
2149 ret = register_pernet_subsys(&tcpv6_net_ops);
2150 if (ret)
2151 goto out_tcpv6_protosw;
2152 out:
2153 return ret;
2155 out_tcpv6_protocol:
2156 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);
2157 out_tcpv6_protosw:
2158 inet6_unregister_protosw(&tcpv6_protosw);
2159 goto out;
2162 void tcpv6_exit(void)
2164 unregister_pernet_subsys(&tcpv6_net_ops);
2165 inet6_unregister_protosw(&tcpv6_protosw);
2166 inet6_del_protocol(&tcpv6_protocol, IPPROTO_TCP);