ACPI: thinkpad-acpi: make EC-based thermal readings non-experimental
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / tcp_ipv6.c
blob9b81cbce547a73e3f9b65718a13a40fff57a2fd7
1 /*
2 * TCP over IPv6
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * $Id: tcp_ipv6.c,v 1.144 2002/02/01 22:01:04 davem Exp $
10 * Based on:
11 * linux/net/ipv4/tcp.c
12 * linux/net/ipv4/tcp_input.c
13 * linux/net/ipv4/tcp_output.c
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.
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.
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>
43 #include <linux/ipv6.h>
44 #include <linux/icmpv6.h>
45 #include <linux/random.h>
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/addrconf.h>
60 #include <net/snmp.h>
61 #include <net/dsfield.h>
62 #include <net/timewait_sock.h>
64 #include <asm/uaccess.h>
66 #include <linux/proc_fs.h>
67 #include <linux/seq_file.h>
69 #include <linux/crypto.h>
70 #include <linux/scatterlist.h>
72 /* Socket used for sending RSTs and ACKs */
73 static struct socket *tcp6_socket;
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);
80 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
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
89 static int tcp_v6_get_port(struct sock *sk, unsigned short snum)
91 return inet_csk_get_port(&tcp_hashinfo, sk, snum,
92 inet6_csk_bind_conflict);
95 static void tcp_v6_hash(struct sock *sk)
97 if (sk->sk_state != TCP_CLOSE) {
98 if (inet_csk(sk)->icsk_af_ops == &ipv6_mapped) {
99 tcp_prot.hash(sk);
100 return;
102 local_bh_disable();
103 __inet6_hash(&tcp_hashinfo, sk);
104 local_bh_enable();
108 static __inline__ __sum16 tcp_v6_check(struct tcphdr *th, int len,
109 struct in6_addr *saddr,
110 struct in6_addr *daddr,
111 __wsum base)
113 return csum_ipv6_magic(saddr, daddr, len, IPPROTO_TCP, base);
116 static __u32 tcp_v6_init_sequence(struct sk_buff *skb)
118 return secure_tcpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
119 ipv6_hdr(skb)->saddr.s6_addr32,
120 tcp_hdr(skb)->dest,
121 tcp_hdr(skb)->source);
124 static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
125 int addr_len)
127 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
128 struct inet_sock *inet = inet_sk(sk);
129 struct inet_connection_sock *icsk = inet_csk(sk);
130 struct ipv6_pinfo *np = inet6_sk(sk);
131 struct tcp_sock *tp = tcp_sk(sk);
132 struct in6_addr *saddr = NULL, *final_p = NULL, final;
133 struct flowi fl;
134 struct dst_entry *dst;
135 int addr_type;
136 int err;
138 if (addr_len < SIN6_LEN_RFC2133)
139 return -EINVAL;
141 if (usin->sin6_family != AF_INET6)
142 return(-EAFNOSUPPORT);
144 memset(&fl, 0, sizeof(fl));
146 if (np->sndflow) {
147 fl.fl6_flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
148 IP6_ECN_flow_init(fl.fl6_flowlabel);
149 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
150 struct ip6_flowlabel *flowlabel;
151 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
152 if (flowlabel == NULL)
153 return -EINVAL;
154 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
155 fl6_sock_release(flowlabel);
160 * connect() to INADDR_ANY means loopback (BSD'ism).
163 if(ipv6_addr_any(&usin->sin6_addr))
164 usin->sin6_addr.s6_addr[15] = 0x1;
166 addr_type = ipv6_addr_type(&usin->sin6_addr);
168 if(addr_type & IPV6_ADDR_MULTICAST)
169 return -ENETUNREACH;
171 if (addr_type&IPV6_ADDR_LINKLOCAL) {
172 if (addr_len >= sizeof(struct sockaddr_in6) &&
173 usin->sin6_scope_id) {
174 /* If interface is set while binding, indices
175 * must coincide.
177 if (sk->sk_bound_dev_if &&
178 sk->sk_bound_dev_if != usin->sin6_scope_id)
179 return -EINVAL;
181 sk->sk_bound_dev_if = usin->sin6_scope_id;
184 /* Connect to link-local address requires an interface */
185 if (!sk->sk_bound_dev_if)
186 return -EINVAL;
189 if (tp->rx_opt.ts_recent_stamp &&
190 !ipv6_addr_equal(&np->daddr, &usin->sin6_addr)) {
191 tp->rx_opt.ts_recent = 0;
192 tp->rx_opt.ts_recent_stamp = 0;
193 tp->write_seq = 0;
196 ipv6_addr_copy(&np->daddr, &usin->sin6_addr);
197 np->flow_label = fl.fl6_flowlabel;
200 * TCP over IPv4
203 if (addr_type == IPV6_ADDR_MAPPED) {
204 u32 exthdrlen = icsk->icsk_ext_hdr_len;
205 struct sockaddr_in sin;
207 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
209 if (__ipv6_only_sock(sk))
210 return -ENETUNREACH;
212 sin.sin_family = AF_INET;
213 sin.sin_port = usin->sin6_port;
214 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
216 icsk->icsk_af_ops = &ipv6_mapped;
217 sk->sk_backlog_rcv = tcp_v4_do_rcv;
218 #ifdef CONFIG_TCP_MD5SIG
219 tp->af_specific = &tcp_sock_ipv6_mapped_specific;
220 #endif
222 err = tcp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
224 if (err) {
225 icsk->icsk_ext_hdr_len = exthdrlen;
226 icsk->icsk_af_ops = &ipv6_specific;
227 sk->sk_backlog_rcv = tcp_v6_do_rcv;
228 #ifdef CONFIG_TCP_MD5SIG
229 tp->af_specific = &tcp_sock_ipv6_specific;
230 #endif
231 goto failure;
232 } else {
233 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF),
234 inet->saddr);
235 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF),
236 inet->rcv_saddr);
239 return err;
242 if (!ipv6_addr_any(&np->rcv_saddr))
243 saddr = &np->rcv_saddr;
245 fl.proto = IPPROTO_TCP;
246 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
247 ipv6_addr_copy(&fl.fl6_src,
248 (saddr ? saddr : &np->saddr));
249 fl.oif = sk->sk_bound_dev_if;
250 fl.fl_ip_dport = usin->sin6_port;
251 fl.fl_ip_sport = inet->sport;
253 if (np->opt && np->opt->srcrt) {
254 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
255 ipv6_addr_copy(&final, &fl.fl6_dst);
256 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
257 final_p = &final;
260 security_sk_classify_flow(sk, &fl);
262 err = ip6_dst_lookup(sk, &dst, &fl);
263 if (err)
264 goto failure;
265 if (final_p)
266 ipv6_addr_copy(&fl.fl6_dst, final_p);
268 if ((err = __xfrm_lookup(&dst, &fl, sk, 1)) < 0) {
269 if (err == -EREMOTE)
270 err = ip6_dst_blackhole(sk, &dst, &fl);
271 if (err < 0)
272 goto failure;
275 if (saddr == NULL) {
276 saddr = &fl.fl6_src;
277 ipv6_addr_copy(&np->rcv_saddr, saddr);
280 /* set the source address */
281 ipv6_addr_copy(&np->saddr, saddr);
282 inet->rcv_saddr = LOOPBACK4_IPV6;
284 sk->sk_gso_type = SKB_GSO_TCPV6;
285 __ip6_dst_store(sk, dst, NULL, NULL);
287 icsk->icsk_ext_hdr_len = 0;
288 if (np->opt)
289 icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
290 np->opt->opt_nflen);
292 tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
294 inet->dport = usin->sin6_port;
296 tcp_set_state(sk, TCP_SYN_SENT);
297 err = inet6_hash_connect(&tcp_death_row, sk);
298 if (err)
299 goto late_failure;
301 if (!tp->write_seq)
302 tp->write_seq = secure_tcpv6_sequence_number(np->saddr.s6_addr32,
303 np->daddr.s6_addr32,
304 inet->sport,
305 inet->dport);
307 err = tcp_connect(sk);
308 if (err)
309 goto late_failure;
311 return 0;
313 late_failure:
314 tcp_set_state(sk, TCP_CLOSE);
315 __sk_dst_reset(sk);
316 failure:
317 inet->dport = 0;
318 sk->sk_route_caps = 0;
319 return err;
322 static void tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
323 int type, int code, int offset, __be32 info)
325 struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
326 const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
327 struct ipv6_pinfo *np;
328 struct sock *sk;
329 int err;
330 struct tcp_sock *tp;
331 __u32 seq;
333 sk = inet6_lookup(&tcp_hashinfo, &hdr->daddr, th->dest, &hdr->saddr,
334 th->source, skb->dev->ifindex);
336 if (sk == NULL) {
337 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), 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(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(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(&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 BUG_TRAP(req->sk == NULL);
431 if (seq != tcp_rsk(req)->snt_isn) {
432 NET_INC_STATS_BH(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,
465 struct dst_entry *dst)
467 struct inet6_request_sock *treq = inet6_rsk(req);
468 struct ipv6_pinfo *np = inet6_sk(sk);
469 struct sk_buff * skb;
470 struct ipv6_txoptions *opt = NULL;
471 struct in6_addr * final_p = NULL, final;
472 struct flowi fl;
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_sk(sk)->sport;
483 security_req_classify_flow(req, &fl);
485 if (dst == NULL) {
486 opt = np->opt;
487 if (opt == NULL &&
488 np->rxopt.bits.osrcrt == 2 &&
489 treq->pktopts) {
490 struct sk_buff *pktopts = treq->pktopts;
491 struct inet6_skb_parm *rxopt = IP6CB(pktopts);
492 if (rxopt->srcrt)
493 opt = ipv6_invert_rthdr(sk,
494 (struct ipv6_rt_hdr *)(skb_network_header(pktopts) +
495 rxopt->srcrt));
498 if (opt && opt->srcrt) {
499 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
500 ipv6_addr_copy(&final, &fl.fl6_dst);
501 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
502 final_p = &final;
505 err = ip6_dst_lookup(sk, &dst, &fl);
506 if (err)
507 goto done;
508 if (final_p)
509 ipv6_addr_copy(&fl.fl6_dst, final_p);
510 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
511 goto done;
514 skb = tcp_make_synack(sk, dst, req);
515 if (skb) {
516 struct tcphdr *th = tcp_hdr(skb);
518 th->check = tcp_v6_check(th, skb->len,
519 &treq->loc_addr, &treq->rmt_addr,
520 csum_partial((char *)th, skb->len, skb->csum));
522 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
523 err = ip6_xmit(sk, skb, &fl, opt, 0);
524 err = net_xmit_eval(err);
527 done:
528 if (opt && opt != np->opt)
529 sock_kfree_s(sk, opt, opt->tot_len);
530 dst_release(dst);
531 return err;
534 static void tcp_v6_reqsk_destructor(struct request_sock *req)
536 if (inet6_rsk(req)->pktopts)
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_cmp(&tp->md5sig_info->keys6[i].addr, addr) == 0)
554 return (struct tcp_md5sig_key *)&tp->md5sig_info->keys6[i];
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 tcp6_md5sig_key *key;
576 struct tcp_sock *tp = tcp_sk(sk);
577 struct tcp6_md5sig_key *keys;
579 key = (struct tcp6_md5sig_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 tcp_alloc_md5sig_pool();
596 if (tp->md5sig_info->alloced6 == tp->md5sig_info->entries6) {
597 keys = kmalloc((sizeof (tp->md5sig_info->keys6[0]) *
598 (tp->md5sig_info->entries6 + 1)), GFP_ATOMIC);
600 if (!keys) {
601 tcp_free_md5sig_pool();
602 kfree(newkey);
603 return -ENOMEM;
606 if (tp->md5sig_info->entries6)
607 memmove(keys, tp->md5sig_info->keys6,
608 (sizeof (tp->md5sig_info->keys6[0]) *
609 tp->md5sig_info->entries6));
611 kfree(tp->md5sig_info->keys6);
612 tp->md5sig_info->keys6 = keys;
613 tp->md5sig_info->alloced6++;
616 ipv6_addr_copy(&tp->md5sig_info->keys6[tp->md5sig_info->entries6].addr,
617 peer);
618 tp->md5sig_info->keys6[tp->md5sig_info->entries6].key = newkey;
619 tp->md5sig_info->keys6[tp->md5sig_info->entries6].keylen = newkeylen;
621 tp->md5sig_info->entries6++;
623 return 0;
626 static int tcp_v6_md5_add_func(struct sock *sk, struct sock *addr_sk,
627 u8 *newkey, __u8 newkeylen)
629 return tcp_v6_md5_do_add(sk, &inet6_sk(addr_sk)->daddr,
630 newkey, newkeylen);
633 static int tcp_v6_md5_do_del(struct sock *sk, struct in6_addr *peer)
635 struct tcp_sock *tp = tcp_sk(sk);
636 int i;
638 for (i = 0; i < tp->md5sig_info->entries6; i++) {
639 if (ipv6_addr_cmp(&tp->md5sig_info->keys6[i].addr, peer) == 0) {
640 /* Free the key */
641 kfree(tp->md5sig_info->keys6[i].key);
642 tp->md5sig_info->entries6--;
644 if (tp->md5sig_info->entries6 == 0) {
645 kfree(tp->md5sig_info->keys6);
646 tp->md5sig_info->keys6 = NULL;
647 tp->md5sig_info->alloced6 = 0;
649 tcp_free_md5sig_pool();
651 return 0;
652 } else {
653 /* shrink the database */
654 if (tp->md5sig_info->entries6 != i)
655 memmove(&tp->md5sig_info->keys6[i],
656 &tp->md5sig_info->keys6[i+1],
657 (tp->md5sig_info->entries6 - i)
658 * sizeof (tp->md5sig_info->keys6[0]));
662 return -ENOENT;
665 static void tcp_v6_clear_md5_list (struct sock *sk)
667 struct tcp_sock *tp = tcp_sk(sk);
668 int i;
670 if (tp->md5sig_info->entries6) {
671 for (i = 0; i < tp->md5sig_info->entries6; i++)
672 kfree(tp->md5sig_info->keys6[i].key);
673 tp->md5sig_info->entries6 = 0;
674 tcp_free_md5sig_pool();
677 kfree(tp->md5sig_info->keys6);
678 tp->md5sig_info->keys6 = NULL;
679 tp->md5sig_info->alloced6 = 0;
681 if (tp->md5sig_info->entries4) {
682 for (i = 0; i < tp->md5sig_info->entries4; i++)
683 kfree(tp->md5sig_info->keys4[i].key);
684 tp->md5sig_info->entries4 = 0;
685 tcp_free_md5sig_pool();
688 kfree(tp->md5sig_info->keys4);
689 tp->md5sig_info->keys4 = NULL;
690 tp->md5sig_info->alloced4 = 0;
693 static int tcp_v6_parse_md5_keys (struct sock *sk, char __user *optval,
694 int optlen)
696 struct tcp_md5sig cmd;
697 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
698 u8 *newkey;
700 if (optlen < sizeof(cmd))
701 return -EINVAL;
703 if (copy_from_user(&cmd, optval, sizeof(cmd)))
704 return -EFAULT;
706 if (sin6->sin6_family != AF_INET6)
707 return -EINVAL;
709 if (!cmd.tcpm_keylen) {
710 if (!tcp_sk(sk)->md5sig_info)
711 return -ENOENT;
712 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_MAPPED)
713 return tcp_v4_md5_do_del(sk, sin6->sin6_addr.s6_addr32[3]);
714 return tcp_v6_md5_do_del(sk, &sin6->sin6_addr);
717 if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
718 return -EINVAL;
720 if (!tcp_sk(sk)->md5sig_info) {
721 struct tcp_sock *tp = tcp_sk(sk);
722 struct tcp_md5sig_info *p;
724 p = kzalloc(sizeof(struct tcp_md5sig_info), GFP_KERNEL);
725 if (!p)
726 return -ENOMEM;
728 tp->md5sig_info = p;
729 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
732 newkey = kmemdup(cmd.tcpm_key, cmd.tcpm_keylen, GFP_KERNEL);
733 if (!newkey)
734 return -ENOMEM;
735 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_MAPPED) {
736 return tcp_v4_md5_do_add(sk, sin6->sin6_addr.s6_addr32[3],
737 newkey, cmd.tcpm_keylen);
739 return tcp_v6_md5_do_add(sk, &sin6->sin6_addr, newkey, cmd.tcpm_keylen);
742 static int tcp_v6_do_calc_md5_hash(char *md5_hash, struct tcp_md5sig_key *key,
743 struct in6_addr *saddr,
744 struct in6_addr *daddr,
745 struct tcphdr *th, int protocol,
746 int tcplen)
748 struct scatterlist sg[4];
749 __u16 data_len;
750 int block = 0;
751 __sum16 cksum;
752 struct tcp_md5sig_pool *hp;
753 struct tcp6_pseudohdr *bp;
754 struct hash_desc *desc;
755 int err;
756 unsigned int nbytes = 0;
758 hp = tcp_get_md5sig_pool();
759 if (!hp) {
760 printk(KERN_WARNING "%s(): hash pool not found...\n", __FUNCTION__);
761 goto clear_hash_noput;
763 bp = &hp->md5_blk.ip6;
764 desc = &hp->md5_desc;
766 /* 1. TCP pseudo-header (RFC2460) */
767 ipv6_addr_copy(&bp->saddr, saddr);
768 ipv6_addr_copy(&bp->daddr, daddr);
769 bp->len = htonl(tcplen);
770 bp->protocol = htonl(protocol);
772 sg_set_buf(&sg[block++], bp, sizeof(*bp));
773 nbytes += sizeof(*bp);
775 /* 2. TCP header, excluding options */
776 cksum = th->check;
777 th->check = 0;
778 sg_set_buf(&sg[block++], th, sizeof(*th));
779 nbytes += sizeof(*th);
781 /* 3. TCP segment data (if any) */
782 data_len = tcplen - (th->doff << 2);
783 if (data_len > 0) {
784 u8 *data = (u8 *)th + (th->doff << 2);
785 sg_set_buf(&sg[block++], data, data_len);
786 nbytes += data_len;
789 /* 4. shared key */
790 sg_set_buf(&sg[block++], key->key, key->keylen);
791 nbytes += key->keylen;
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", __FUNCTION__);
797 goto clear_hash;
799 err = crypto_hash_update(desc, sg, nbytes);
800 if (err) {
801 printk(KERN_WARNING "%s(): hash_update failed\n", __FUNCTION__);
802 goto clear_hash;
804 err = crypto_hash_final(desc, md5_hash);
805 if (err) {
806 printk(KERN_WARNING "%s(): hash_final failed\n", __FUNCTION__);
807 goto clear_hash;
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;
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 int tcplen)
829 struct in6_addr *saddr, *daddr;
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;
838 return tcp_v6_do_calc_md5_hash(md5_hash, key,
839 saddr, daddr,
840 th, protocol, tcplen);
843 static int tcp_v6_inbound_md5_hash (struct sock *sk, struct sk_buff *skb)
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];
854 hash_expected = tcp_v6_md5_do_lookup(sk, &ip6h->saddr);
856 /* If the TCP option is too short, we can short cut */
857 if (length < TCPOLEN_MD5SIG)
858 return hash_expected ? 1 : 0;
860 /* parse options */
861 ptr = (u8*)(th + 1);
862 while (length > 0) {
863 int opcode = *ptr++;
864 int opsize;
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;
881 ptr += opsize - 2;
882 length -= opsize;
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));
897 return 1;
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));
908 return 1;
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));
926 return 1;
928 return 0;
930 #endif
932 static 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
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,
945 #endif
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,
953 static void tcp_v6_send_check(struct sock *sk, int len, struct sk_buff *skb)
955 struct ipv6_pinfo *np = inet6_sk(sk);
956 struct tcphdr *th = tcp_hdr(skb);
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));
969 static int tcp_v6_gso_send_check(struct sk_buff *skb)
971 struct ipv6hdr *ipv6h;
972 struct tcphdr *th;
974 if (!pskb_may_pull(skb, sizeof(*th)))
975 return -EINVAL;
977 ipv6h = ipv6_hdr(skb);
978 th = tcp_hdr(skb);
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;
989 static void tcp_v6_send_reset(struct sock *sk, struct sk_buff *skb)
991 struct tcphdr *th = tcp_hdr(skb), *t1;
992 struct sk_buff *buff;
993 struct flowi fl;
994 int tot_len = sizeof(*th);
995 #ifdef CONFIG_TCP_MD5SIG
996 struct tcp_md5sig_key *key;
997 #endif
999 if (th->rst)
1000 return;
1002 if (!ipv6_unicast_destination(skb))
1003 return;
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;
1011 if (key)
1012 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1013 #endif
1016 * We need to grab some memory, and put together an RST,
1017 * and then put it into the queue to be sent.
1020 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1021 GFP_ATOMIC);
1022 if (buff == NULL)
1023 return;
1025 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1027 t1 = (struct tcphdr *) skb_push(buff, tot_len);
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;
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));
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);
1056 #endif
1058 buff->csum = csum_partial((char *)t1, sizeof(*t1), 0);
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);
1064 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1065 sizeof(*t1), IPPROTO_TCP,
1066 buff->csum);
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);
1074 /* sk = NULL, but it is safe for now. RST socket required. */
1075 if (!ip6_dst_lookup(NULL, &buff->dst, &fl)) {
1077 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
1078 ip6_xmit(tcp6_socket->sk, buff, &fl, NULL, 0);
1079 TCP_INC_STATS_BH(TCP_MIB_OUTSEGS);
1080 TCP_INC_STATS_BH(TCP_MIB_OUTRSTS);
1081 return;
1085 kfree_skb(buff);
1088 static void tcp_v6_send_ack(struct tcp_timewait_sock *tw,
1089 struct sk_buff *skb, u32 seq, u32 ack, u32 win, u32 ts)
1091 struct tcphdr *th = tcp_hdr(skb), *t1;
1092 struct sk_buff *buff;
1093 struct flowi fl;
1094 int tot_len = sizeof(struct tcphdr);
1095 __be32 *topt;
1096 #ifdef CONFIG_TCP_MD5SIG
1097 struct tcp_md5sig_key *key;
1098 struct tcp_md5sig_key tw_key;
1099 #endif
1101 #ifdef CONFIG_TCP_MD5SIG
1102 if (!tw && skb->sk) {
1103 key = tcp_v6_md5_do_lookup(skb->sk, &ipv6_hdr(skb)->daddr);
1104 } else if (tw && tw->tw_md5_keylen) {
1105 tw_key.key = tw->tw_md5_key;
1106 tw_key.keylen = tw->tw_md5_keylen;
1107 key = &tw_key;
1108 } else {
1109 key = NULL;
1111 #endif
1113 if (ts)
1114 tot_len += TCPOLEN_TSTAMP_ALIGNED;
1115 #ifdef CONFIG_TCP_MD5SIG
1116 if (key)
1117 tot_len += TCPOLEN_MD5SIG_ALIGNED;
1118 #endif
1120 buff = alloc_skb(MAX_HEADER + sizeof(struct ipv6hdr) + tot_len,
1121 GFP_ATOMIC);
1122 if (buff == NULL)
1123 return;
1125 skb_reserve(buff, MAX_HEADER + sizeof(struct ipv6hdr) + tot_len);
1127 t1 = (struct tcphdr *) skb_push(buff,tot_len);
1129 /* Swap the send and the receive. */
1130 memset(t1, 0, sizeof(*t1));
1131 t1->dest = th->source;
1132 t1->source = th->dest;
1133 t1->doff = tot_len/4;
1134 t1->seq = htonl(seq);
1135 t1->ack_seq = htonl(ack);
1136 t1->ack = 1;
1137 t1->window = htons(win);
1139 topt = (__be32 *)(t1 + 1);
1141 if (ts) {
1142 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1143 (TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP);
1144 *topt++ = htonl(tcp_time_stamp);
1145 *topt = htonl(ts);
1148 #ifdef CONFIG_TCP_MD5SIG
1149 if (key) {
1150 *topt++ = htonl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) |
1151 (TCPOPT_MD5SIG << 8) | TCPOLEN_MD5SIG);
1152 tcp_v6_do_calc_md5_hash((__u8 *)topt, key,
1153 &ipv6_hdr(skb)->daddr,
1154 &ipv6_hdr(skb)->saddr,
1155 t1, IPPROTO_TCP, tot_len);
1157 #endif
1159 buff->csum = csum_partial((char *)t1, tot_len, 0);
1161 memset(&fl, 0, sizeof(fl));
1162 ipv6_addr_copy(&fl.fl6_dst, &ipv6_hdr(skb)->saddr);
1163 ipv6_addr_copy(&fl.fl6_src, &ipv6_hdr(skb)->daddr);
1165 t1->check = csum_ipv6_magic(&fl.fl6_src, &fl.fl6_dst,
1166 tot_len, IPPROTO_TCP,
1167 buff->csum);
1169 fl.proto = IPPROTO_TCP;
1170 fl.oif = inet6_iif(skb);
1171 fl.fl_ip_dport = t1->dest;
1172 fl.fl_ip_sport = t1->source;
1173 security_skb_classify_flow(skb, &fl);
1175 if (!ip6_dst_lookup(NULL, &buff->dst, &fl)) {
1176 if (xfrm_lookup(&buff->dst, &fl, NULL, 0) >= 0) {
1177 ip6_xmit(tcp6_socket->sk, buff, &fl, NULL, 0);
1178 TCP_INC_STATS_BH(TCP_MIB_OUTSEGS);
1179 return;
1183 kfree_skb(buff);
1186 static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
1188 struct inet_timewait_sock *tw = inet_twsk(sk);
1189 struct tcp_timewait_sock *tcptw = tcp_twsk(sk);
1191 tcp_v6_send_ack(tcptw, skb, tcptw->tw_snd_nxt, tcptw->tw_rcv_nxt,
1192 tcptw->tw_rcv_wnd >> tw->tw_rcv_wscale,
1193 tcptw->tw_ts_recent);
1195 inet_twsk_put(tw);
1198 static void tcp_v6_reqsk_send_ack(struct sk_buff *skb, struct request_sock *req)
1200 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 static struct sock *tcp_v6_hnd_req(struct sock *sk,struct sk_buff *skb)
1206 struct request_sock *req, **prev;
1207 const struct tcphdr *th = tcp_hdr(skb);
1208 struct sock *nsk;
1210 /* Find possible connection requests. */
1211 req = inet6_csk_search_req(sk, &prev, th->source,
1212 &ipv6_hdr(skb)->saddr,
1213 &ipv6_hdr(skb)->daddr, inet6_iif(skb));
1214 if (req)
1215 return tcp_check_req(sk, skb, req, prev);
1217 nsk = __inet6_lookup_established(&tcp_hashinfo, &ipv6_hdr(skb)->saddr,
1218 th->source, &ipv6_hdr(skb)->daddr,
1219 ntohs(th->dest), inet6_iif(skb));
1221 if (nsk) {
1222 if (nsk->sk_state != TCP_TIME_WAIT) {
1223 bh_lock_sock(nsk);
1224 return nsk;
1226 inet_twsk_put(inet_twsk(nsk));
1227 return NULL;
1230 #if 0 /*def CONFIG_SYN_COOKIES*/
1231 if (!th->rst && !th->syn && th->ack)
1232 sk = cookie_v6_check(sk, skb, &(IPCB(skb)->opt));
1233 #endif
1234 return sk;
1237 /* FIXME: this is substantially similar to the ipv4 code.
1238 * Can some kind of merge be done? -- erics
1240 static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
1242 struct inet6_request_sock *treq;
1243 struct ipv6_pinfo *np = inet6_sk(sk);
1244 struct tcp_options_received tmp_opt;
1245 struct tcp_sock *tp = tcp_sk(sk);
1246 struct request_sock *req = NULL;
1247 __u32 isn = TCP_SKB_CB(skb)->when;
1249 if (skb->protocol == htons(ETH_P_IP))
1250 return tcp_v4_conn_request(sk, skb);
1252 if (!ipv6_unicast_destination(skb))
1253 goto drop;
1256 * There are no SYN attacks on IPv6, yet...
1258 if (inet_csk_reqsk_queue_is_full(sk) && !isn) {
1259 if (net_ratelimit())
1260 printk(KERN_INFO "TCPv6: dropping request, synflood is possible\n");
1261 goto drop;
1264 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
1265 goto drop;
1267 req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
1268 if (req == NULL)
1269 goto drop;
1271 #ifdef CONFIG_TCP_MD5SIG
1272 tcp_rsk(req)->af_specific = &tcp_request_sock_ipv6_ops;
1273 #endif
1275 tcp_clear_options(&tmp_opt);
1276 tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
1277 tmp_opt.user_mss = tp->rx_opt.user_mss;
1279 tcp_parse_options(skb, &tmp_opt, 0);
1281 tmp_opt.tstamp_ok = tmp_opt.saw_tstamp;
1282 tcp_openreq_init(req, &tmp_opt, skb);
1284 treq = inet6_rsk(req);
1285 ipv6_addr_copy(&treq->rmt_addr, &ipv6_hdr(skb)->saddr);
1286 ipv6_addr_copy(&treq->loc_addr, &ipv6_hdr(skb)->daddr);
1287 TCP_ECN_create_request(req, tcp_hdr(skb));
1288 treq->pktopts = NULL;
1289 if (ipv6_opt_accepted(sk, skb) ||
1290 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
1291 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
1292 atomic_inc(&skb->users);
1293 treq->pktopts = skb;
1295 treq->iif = sk->sk_bound_dev_if;
1297 /* So that link locals have meaning */
1298 if (!sk->sk_bound_dev_if &&
1299 ipv6_addr_type(&treq->rmt_addr) & IPV6_ADDR_LINKLOCAL)
1300 treq->iif = inet6_iif(skb);
1302 if (isn == 0)
1303 isn = tcp_v6_init_sequence(skb);
1305 tcp_rsk(req)->snt_isn = isn;
1307 security_inet_conn_request(sk, skb, req);
1309 if (tcp_v6_send_synack(sk, req, NULL))
1310 goto drop;
1312 inet6_csk_reqsk_queue_hash_add(sk, req, TCP_TIMEOUT_INIT);
1313 return 0;
1315 drop:
1316 if (req)
1317 reqsk_free(req);
1319 return 0; /* don't send reset */
1322 static struct sock * tcp_v6_syn_recv_sock(struct sock *sk, struct sk_buff *skb,
1323 struct request_sock *req,
1324 struct dst_entry *dst)
1326 struct inet6_request_sock *treq = inet6_rsk(req);
1327 struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
1328 struct tcp6_sock *newtcp6sk;
1329 struct inet_sock *newinet;
1330 struct tcp_sock *newtp;
1331 struct sock *newsk;
1332 struct ipv6_txoptions *opt;
1333 #ifdef CONFIG_TCP_MD5SIG
1334 struct tcp_md5sig_key *key;
1335 #endif
1337 if (skb->protocol == htons(ETH_P_IP)) {
1339 * v6 mapped
1342 newsk = tcp_v4_syn_recv_sock(sk, skb, req, dst);
1344 if (newsk == NULL)
1345 return NULL;
1347 newtcp6sk = (struct tcp6_sock *)newsk;
1348 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1350 newinet = inet_sk(newsk);
1351 newnp = inet6_sk(newsk);
1352 newtp = tcp_sk(newsk);
1354 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1356 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF),
1357 newinet->daddr);
1359 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF),
1360 newinet->saddr);
1362 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr);
1364 inet_csk(newsk)->icsk_af_ops = &ipv6_mapped;
1365 newsk->sk_backlog_rcv = tcp_v4_do_rcv;
1366 #ifdef CONFIG_TCP_MD5SIG
1367 newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
1368 #endif
1370 newnp->pktoptions = NULL;
1371 newnp->opt = NULL;
1372 newnp->mcast_oif = inet6_iif(skb);
1373 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1376 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1377 * here, tcp_create_openreq_child now does this for us, see the comment in
1378 * that function for the gory details. -acme
1381 /* It is tricky place. Until this moment IPv4 tcp
1382 worked with IPv6 icsk.icsk_af_ops.
1383 Sync it now.
1385 tcp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
1387 return newsk;
1390 opt = np->opt;
1392 if (sk_acceptq_is_full(sk))
1393 goto out_overflow;
1395 if (np->rxopt.bits.osrcrt == 2 &&
1396 opt == NULL && treq->pktopts) {
1397 struct inet6_skb_parm *rxopt = IP6CB(treq->pktopts);
1398 if (rxopt->srcrt)
1399 opt = ipv6_invert_rthdr(sk,
1400 (struct ipv6_rt_hdr *)(skb_network_header(treq->pktopts) +
1401 rxopt->srcrt));
1404 if (dst == NULL) {
1405 struct in6_addr *final_p = NULL, final;
1406 struct flowi fl;
1408 memset(&fl, 0, sizeof(fl));
1409 fl.proto = IPPROTO_TCP;
1410 ipv6_addr_copy(&fl.fl6_dst, &treq->rmt_addr);
1411 if (opt && opt->srcrt) {
1412 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
1413 ipv6_addr_copy(&final, &fl.fl6_dst);
1414 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
1415 final_p = &final;
1417 ipv6_addr_copy(&fl.fl6_src, &treq->loc_addr);
1418 fl.oif = sk->sk_bound_dev_if;
1419 fl.fl_ip_dport = inet_rsk(req)->rmt_port;
1420 fl.fl_ip_sport = inet_sk(sk)->sport;
1421 security_req_classify_flow(req, &fl);
1423 if (ip6_dst_lookup(sk, &dst, &fl))
1424 goto out;
1426 if (final_p)
1427 ipv6_addr_copy(&fl.fl6_dst, final_p);
1429 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0)
1430 goto out;
1433 newsk = tcp_create_openreq_child(sk, req, skb);
1434 if (newsk == NULL)
1435 goto out;
1438 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1439 * count here, tcp_create_openreq_child now does this for us, see the
1440 * comment in that function for the gory details. -acme
1443 newsk->sk_gso_type = SKB_GSO_TCPV6;
1444 __ip6_dst_store(newsk, dst, NULL, NULL);
1446 newtcp6sk = (struct tcp6_sock *)newsk;
1447 inet_sk(newsk)->pinet6 = &newtcp6sk->inet6;
1449 newtp = tcp_sk(newsk);
1450 newinet = inet_sk(newsk);
1451 newnp = inet6_sk(newsk);
1453 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
1455 ipv6_addr_copy(&newnp->daddr, &treq->rmt_addr);
1456 ipv6_addr_copy(&newnp->saddr, &treq->loc_addr);
1457 ipv6_addr_copy(&newnp->rcv_saddr, &treq->loc_addr);
1458 newsk->sk_bound_dev_if = treq->iif;
1460 /* Now IPv6 options...
1462 First: no IPv4 options.
1464 newinet->opt = NULL;
1465 newnp->ipv6_fl_list = NULL;
1467 /* Clone RX bits */
1468 newnp->rxopt.all = np->rxopt.all;
1470 /* Clone pktoptions received with SYN */
1471 newnp->pktoptions = NULL;
1472 if (treq->pktopts != NULL) {
1473 newnp->pktoptions = skb_clone(treq->pktopts, GFP_ATOMIC);
1474 kfree_skb(treq->pktopts);
1475 treq->pktopts = NULL;
1476 if (newnp->pktoptions)
1477 skb_set_owner_r(newnp->pktoptions, newsk);
1479 newnp->opt = NULL;
1480 newnp->mcast_oif = inet6_iif(skb);
1481 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
1483 /* Clone native IPv6 options from listening socket (if any)
1485 Yes, keeping reference count would be much more clever,
1486 but we make one more one thing there: reattach optmem
1487 to newsk.
1489 if (opt) {
1490 newnp->opt = ipv6_dup_options(newsk, opt);
1491 if (opt != np->opt)
1492 sock_kfree_s(sk, opt, opt->tot_len);
1495 inet_csk(newsk)->icsk_ext_hdr_len = 0;
1496 if (newnp->opt)
1497 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
1498 newnp->opt->opt_flen);
1500 tcp_mtup_init(newsk);
1501 tcp_sync_mss(newsk, dst_mtu(dst));
1502 newtp->advmss = dst_metric(dst, RTAX_ADVMSS);
1503 tcp_initialize_rcv_mss(newsk);
1505 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6;
1507 #ifdef CONFIG_TCP_MD5SIG
1508 /* Copy over the MD5 key from the original socket */
1509 if ((key = tcp_v6_md5_do_lookup(sk, &newnp->daddr)) != NULL) {
1510 /* We're using one, so create a matching key
1511 * on the newsk structure. If we fail to get
1512 * memory, then we end up not copying the key
1513 * across. Shucks.
1515 char *newkey = kmemdup(key->key, key->keylen, GFP_ATOMIC);
1516 if (newkey != NULL)
1517 tcp_v6_md5_do_add(newsk, &inet6_sk(sk)->daddr,
1518 newkey, key->keylen);
1520 #endif
1522 __inet6_hash(&tcp_hashinfo, newsk);
1523 inet_inherit_port(&tcp_hashinfo, sk, newsk);
1525 return newsk;
1527 out_overflow:
1528 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS);
1529 out:
1530 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS);
1531 if (opt && opt != np->opt)
1532 sock_kfree_s(sk, opt, opt->tot_len);
1533 dst_release(dst);
1534 return NULL;
1537 static __sum16 tcp_v6_checksum_init(struct sk_buff *skb)
1539 if (skb->ip_summed == CHECKSUM_COMPLETE) {
1540 if (!tcp_v6_check(tcp_hdr(skb), skb->len, &ipv6_hdr(skb)->saddr,
1541 &ipv6_hdr(skb)->daddr, skb->csum)) {
1542 skb->ip_summed = CHECKSUM_UNNECESSARY;
1543 return 0;
1547 skb->csum = ~csum_unfold(tcp_v6_check(tcp_hdr(skb), skb->len,
1548 &ipv6_hdr(skb)->saddr,
1549 &ipv6_hdr(skb)->daddr, 0));
1551 if (skb->len <= 76) {
1552 return __skb_checksum_complete(skb);
1554 return 0;
1557 /* The socket must have it's spinlock held when we get
1558 * here.
1560 * We have a potential double-lock case here, so even when
1561 * doing backlog processing we use the BH locking scheme.
1562 * This is because we cannot sleep with the original spinlock
1563 * held.
1565 static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
1567 struct ipv6_pinfo *np = inet6_sk(sk);
1568 struct tcp_sock *tp;
1569 struct sk_buff *opt_skb = NULL;
1571 /* Imagine: socket is IPv6. IPv4 packet arrives,
1572 goes to IPv4 receive handler and backlogged.
1573 From backlog it always goes here. Kerboom...
1574 Fortunately, tcp_rcv_established and rcv_established
1575 handle them correctly, but it is not case with
1576 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1579 if (skb->protocol == htons(ETH_P_IP))
1580 return tcp_v4_do_rcv(sk, skb);
1582 #ifdef CONFIG_TCP_MD5SIG
1583 if (tcp_v6_inbound_md5_hash (sk, skb))
1584 goto discard;
1585 #endif
1587 if (sk_filter(sk, skb))
1588 goto discard;
1591 * socket locking is here for SMP purposes as backlog rcv
1592 * is currently called with bh processing disabled.
1595 /* Do Stevens' IPV6_PKTOPTIONS.
1597 Yes, guys, it is the only place in our code, where we
1598 may make it not affecting IPv4.
1599 The rest of code is protocol independent,
1600 and I do not like idea to uglify IPv4.
1602 Actually, all the idea behind IPV6_PKTOPTIONS
1603 looks not very well thought. For now we latch
1604 options, received in the last packet, enqueued
1605 by tcp. Feel free to propose better solution.
1606 --ANK (980728)
1608 if (np->rxopt.all)
1609 opt_skb = skb_clone(skb, GFP_ATOMIC);
1611 if (sk->sk_state == TCP_ESTABLISHED) { /* Fast path */
1612 TCP_CHECK_TIMER(sk);
1613 if (tcp_rcv_established(sk, skb, tcp_hdr(skb), skb->len))
1614 goto reset;
1615 TCP_CHECK_TIMER(sk);
1616 if (opt_skb)
1617 goto ipv6_pktoptions;
1618 return 0;
1621 if (skb->len < tcp_hdrlen(skb) || tcp_checksum_complete(skb))
1622 goto csum_err;
1624 if (sk->sk_state == TCP_LISTEN) {
1625 struct sock *nsk = tcp_v6_hnd_req(sk, skb);
1626 if (!nsk)
1627 goto discard;
1630 * Queue it on the new socket if the new socket is active,
1631 * otherwise we just shortcircuit this and continue with
1632 * the new socket..
1634 if(nsk != sk) {
1635 if (tcp_child_process(sk, nsk, skb))
1636 goto reset;
1637 if (opt_skb)
1638 __kfree_skb(opt_skb);
1639 return 0;
1643 TCP_CHECK_TIMER(sk);
1644 if (tcp_rcv_state_process(sk, skb, tcp_hdr(skb), skb->len))
1645 goto reset;
1646 TCP_CHECK_TIMER(sk);
1647 if (opt_skb)
1648 goto ipv6_pktoptions;
1649 return 0;
1651 reset:
1652 tcp_v6_send_reset(sk, skb);
1653 discard:
1654 if (opt_skb)
1655 __kfree_skb(opt_skb);
1656 kfree_skb(skb);
1657 return 0;
1658 csum_err:
1659 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1660 goto discard;
1663 ipv6_pktoptions:
1664 /* Do you ask, what is it?
1666 1. skb was enqueued by tcp.
1667 2. skb is added to tail of read queue, rather than out of order.
1668 3. socket is not in passive state.
1669 4. Finally, it really contains options, which user wants to receive.
1671 tp = tcp_sk(sk);
1672 if (TCP_SKB_CB(opt_skb)->end_seq == tp->rcv_nxt &&
1673 !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
1674 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
1675 np->mcast_oif = inet6_iif(opt_skb);
1676 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
1677 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
1678 if (ipv6_opt_accepted(sk, opt_skb)) {
1679 skb_set_owner_r(opt_skb, sk);
1680 opt_skb = xchg(&np->pktoptions, opt_skb);
1681 } else {
1682 __kfree_skb(opt_skb);
1683 opt_skb = xchg(&np->pktoptions, NULL);
1687 if (opt_skb)
1688 kfree_skb(opt_skb);
1689 return 0;
1692 static int tcp_v6_rcv(struct sk_buff **pskb)
1694 struct sk_buff *skb = *pskb;
1695 struct tcphdr *th;
1696 struct sock *sk;
1697 int ret;
1699 if (skb->pkt_type != PACKET_HOST)
1700 goto discard_it;
1703 * Count it even if it's bad.
1705 TCP_INC_STATS_BH(TCP_MIB_INSEGS);
1707 if (!pskb_may_pull(skb, sizeof(struct tcphdr)))
1708 goto discard_it;
1710 th = tcp_hdr(skb);
1712 if (th->doff < sizeof(struct tcphdr)/4)
1713 goto bad_packet;
1714 if (!pskb_may_pull(skb, th->doff*4))
1715 goto discard_it;
1717 if (!skb_csum_unnecessary(skb) && tcp_v6_checksum_init(skb))
1718 goto bad_packet;
1720 th = tcp_hdr(skb);
1721 TCP_SKB_CB(skb)->seq = ntohl(th->seq);
1722 TCP_SKB_CB(skb)->end_seq = (TCP_SKB_CB(skb)->seq + th->syn + th->fin +
1723 skb->len - th->doff*4);
1724 TCP_SKB_CB(skb)->ack_seq = ntohl(th->ack_seq);
1725 TCP_SKB_CB(skb)->when = 0;
1726 TCP_SKB_CB(skb)->flags = ipv6_get_dsfield(ipv6_hdr(skb));
1727 TCP_SKB_CB(skb)->sacked = 0;
1729 sk = __inet6_lookup(&tcp_hashinfo, &ipv6_hdr(skb)->saddr, th->source,
1730 &ipv6_hdr(skb)->daddr, ntohs(th->dest),
1731 inet6_iif(skb));
1733 if (!sk)
1734 goto no_tcp_socket;
1736 process:
1737 if (sk->sk_state == TCP_TIME_WAIT)
1738 goto do_time_wait;
1740 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
1741 goto discard_and_relse;
1743 if (sk_filter(sk, skb))
1744 goto discard_and_relse;
1746 skb->dev = NULL;
1748 bh_lock_sock_nested(sk);
1749 ret = 0;
1750 if (!sock_owned_by_user(sk)) {
1751 #ifdef CONFIG_NET_DMA
1752 struct tcp_sock *tp = tcp_sk(sk);
1753 if (tp->ucopy.dma_chan)
1754 ret = tcp_v6_do_rcv(sk, skb);
1755 else
1756 #endif
1758 if (!tcp_prequeue(sk, skb))
1759 ret = tcp_v6_do_rcv(sk, skb);
1761 } else
1762 sk_add_backlog(sk, skb);
1763 bh_unlock_sock(sk);
1765 sock_put(sk);
1766 return ret ? -1 : 0;
1768 no_tcp_socket:
1769 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
1770 goto discard_it;
1772 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1773 bad_packet:
1774 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1775 } else {
1776 tcp_v6_send_reset(NULL, skb);
1779 discard_it:
1782 * Discard frame
1785 kfree_skb(skb);
1786 return 0;
1788 discard_and_relse:
1789 sock_put(sk);
1790 goto discard_it;
1792 do_time_wait:
1793 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1794 inet_twsk_put(inet_twsk(sk));
1795 goto discard_it;
1798 if (skb->len < (th->doff<<2) || tcp_checksum_complete(skb)) {
1799 TCP_INC_STATS_BH(TCP_MIB_INERRS);
1800 inet_twsk_put(inet_twsk(sk));
1801 goto discard_it;
1804 switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
1805 case TCP_TW_SYN:
1807 struct sock *sk2;
1809 sk2 = inet6_lookup_listener(&tcp_hashinfo,
1810 &ipv6_hdr(skb)->daddr,
1811 ntohs(th->dest), inet6_iif(skb));
1812 if (sk2 != NULL) {
1813 struct inet_timewait_sock *tw = inet_twsk(sk);
1814 inet_twsk_deschedule(tw, &tcp_death_row);
1815 inet_twsk_put(tw);
1816 sk = sk2;
1817 goto process;
1819 /* Fall through to ACK */
1821 case TCP_TW_ACK:
1822 tcp_v6_timewait_ack(sk, skb);
1823 break;
1824 case TCP_TW_RST:
1825 goto no_tcp_socket;
1826 case TCP_TW_SUCCESS:;
1828 goto discard_it;
1831 static int tcp_v6_remember_stamp(struct sock *sk)
1833 /* Alas, not yet... */
1834 return 0;
1837 static struct inet_connection_sock_af_ops ipv6_specific = {
1838 .queue_xmit = inet6_csk_xmit,
1839 .send_check = tcp_v6_send_check,
1840 .rebuild_header = inet6_sk_rebuild_header,
1841 .conn_request = tcp_v6_conn_request,
1842 .syn_recv_sock = tcp_v6_syn_recv_sock,
1843 .remember_stamp = tcp_v6_remember_stamp,
1844 .net_header_len = sizeof(struct ipv6hdr),
1845 .setsockopt = ipv6_setsockopt,
1846 .getsockopt = ipv6_getsockopt,
1847 .addr2sockaddr = inet6_csk_addr2sockaddr,
1848 .sockaddr_len = sizeof(struct sockaddr_in6),
1849 #ifdef CONFIG_COMPAT
1850 .compat_setsockopt = compat_ipv6_setsockopt,
1851 .compat_getsockopt = compat_ipv6_getsockopt,
1852 #endif
1855 #ifdef CONFIG_TCP_MD5SIG
1856 static struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
1857 .md5_lookup = tcp_v6_md5_lookup,
1858 .calc_md5_hash = tcp_v6_calc_md5_hash,
1859 .md5_add = tcp_v6_md5_add_func,
1860 .md5_parse = tcp_v6_parse_md5_keys,
1862 #endif
1865 * TCP over IPv4 via INET6 API
1868 static struct inet_connection_sock_af_ops ipv6_mapped = {
1869 .queue_xmit = ip_queue_xmit,
1870 .send_check = tcp_v4_send_check,
1871 .rebuild_header = inet_sk_rebuild_header,
1872 .conn_request = tcp_v6_conn_request,
1873 .syn_recv_sock = tcp_v6_syn_recv_sock,
1874 .remember_stamp = tcp_v4_remember_stamp,
1875 .net_header_len = sizeof(struct iphdr),
1876 .setsockopt = ipv6_setsockopt,
1877 .getsockopt = ipv6_getsockopt,
1878 .addr2sockaddr = inet6_csk_addr2sockaddr,
1879 .sockaddr_len = sizeof(struct sockaddr_in6),
1880 #ifdef CONFIG_COMPAT
1881 .compat_setsockopt = compat_ipv6_setsockopt,
1882 .compat_getsockopt = compat_ipv6_getsockopt,
1883 #endif
1886 #ifdef CONFIG_TCP_MD5SIG
1887 static struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
1888 .md5_lookup = tcp_v4_md5_lookup,
1889 .calc_md5_hash = tcp_v4_calc_md5_hash,
1890 .md5_add = tcp_v6_md5_add_func,
1891 .md5_parse = tcp_v6_parse_md5_keys,
1893 #endif
1895 /* NOTE: A lot of things set to zero explicitly by call to
1896 * sk_alloc() so need not be done here.
1898 static int tcp_v6_init_sock(struct sock *sk)
1900 struct inet_connection_sock *icsk = inet_csk(sk);
1901 struct tcp_sock *tp = tcp_sk(sk);
1903 skb_queue_head_init(&tp->out_of_order_queue);
1904 tcp_init_xmit_timers(sk);
1905 tcp_prequeue_init(tp);
1907 icsk->icsk_rto = TCP_TIMEOUT_INIT;
1908 tp->mdev = TCP_TIMEOUT_INIT;
1910 /* So many TCP implementations out there (incorrectly) count the
1911 * initial SYN frame in their delayed-ACK and congestion control
1912 * algorithms that we must have the following bandaid to talk
1913 * efficiently to them. -DaveM
1915 tp->snd_cwnd = 2;
1917 /* See draft-stevens-tcpca-spec-01 for discussion of the
1918 * initialization of these values.
1920 tp->snd_ssthresh = 0x7fffffff;
1921 tp->snd_cwnd_clamp = ~0;
1922 tp->mss_cache = 536;
1924 tp->reordering = sysctl_tcp_reordering;
1926 sk->sk_state = TCP_CLOSE;
1928 icsk->icsk_af_ops = &ipv6_specific;
1929 icsk->icsk_ca_ops = &tcp_init_congestion_ops;
1930 icsk->icsk_sync_mss = tcp_sync_mss;
1931 sk->sk_write_space = sk_stream_write_space;
1932 sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
1934 #ifdef CONFIG_TCP_MD5SIG
1935 tp->af_specific = &tcp_sock_ipv6_specific;
1936 #endif
1938 sk->sk_sndbuf = sysctl_tcp_wmem[1];
1939 sk->sk_rcvbuf = sysctl_tcp_rmem[1];
1941 atomic_inc(&tcp_sockets_allocated);
1943 return 0;
1946 static int tcp_v6_destroy_sock(struct sock *sk)
1948 #ifdef CONFIG_TCP_MD5SIG
1949 /* Clean up the MD5 key list */
1950 if (tcp_sk(sk)->md5sig_info)
1951 tcp_v6_clear_md5_list(sk);
1952 #endif
1953 tcp_v4_destroy_sock(sk);
1954 return inet6_destroy_sock(sk);
1957 #ifdef CONFIG_PROC_FS
1958 /* Proc filesystem TCPv6 sock list dumping. */
1959 static void get_openreq6(struct seq_file *seq,
1960 struct sock *sk, struct request_sock *req, int i, int uid)
1962 int ttd = req->expires - jiffies;
1963 struct in6_addr *src = &inet6_rsk(req)->loc_addr;
1964 struct in6_addr *dest = &inet6_rsk(req)->rmt_addr;
1966 if (ttd < 0)
1967 ttd = 0;
1969 seq_printf(seq,
1970 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1971 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
1973 src->s6_addr32[0], src->s6_addr32[1],
1974 src->s6_addr32[2], src->s6_addr32[3],
1975 ntohs(inet_sk(sk)->sport),
1976 dest->s6_addr32[0], dest->s6_addr32[1],
1977 dest->s6_addr32[2], dest->s6_addr32[3],
1978 ntohs(inet_rsk(req)->rmt_port),
1979 TCP_SYN_RECV,
1980 0,0, /* could print option size, but that is af dependent. */
1981 1, /* timers active (only the expire timer) */
1982 jiffies_to_clock_t(ttd),
1983 req->retrans,
1984 uid,
1985 0, /* non standard timer */
1986 0, /* open_requests have no inode */
1987 0, req);
1990 static void get_tcp6_sock(struct seq_file *seq, struct sock *sp, int i)
1992 struct in6_addr *dest, *src;
1993 __u16 destp, srcp;
1994 int timer_active;
1995 unsigned long timer_expires;
1996 struct inet_sock *inet = inet_sk(sp);
1997 struct tcp_sock *tp = tcp_sk(sp);
1998 const struct inet_connection_sock *icsk = inet_csk(sp);
1999 struct ipv6_pinfo *np = inet6_sk(sp);
2001 dest = &np->daddr;
2002 src = &np->rcv_saddr;
2003 destp = ntohs(inet->dport);
2004 srcp = ntohs(inet->sport);
2006 if (icsk->icsk_pending == ICSK_TIME_RETRANS) {
2007 timer_active = 1;
2008 timer_expires = icsk->icsk_timeout;
2009 } else if (icsk->icsk_pending == ICSK_TIME_PROBE0) {
2010 timer_active = 4;
2011 timer_expires = icsk->icsk_timeout;
2012 } else if (timer_pending(&sp->sk_timer)) {
2013 timer_active = 2;
2014 timer_expires = sp->sk_timer.expires;
2015 } else {
2016 timer_active = 0;
2017 timer_expires = jiffies;
2020 seq_printf(seq,
2021 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2022 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %u %u %u %u %d\n",
2024 src->s6_addr32[0], src->s6_addr32[1],
2025 src->s6_addr32[2], src->s6_addr32[3], srcp,
2026 dest->s6_addr32[0], dest->s6_addr32[1],
2027 dest->s6_addr32[2], dest->s6_addr32[3], destp,
2028 sp->sk_state,
2029 tp->write_seq-tp->snd_una,
2030 (sp->sk_state == TCP_LISTEN) ? sp->sk_ack_backlog : (tp->rcv_nxt - tp->copied_seq),
2031 timer_active,
2032 jiffies_to_clock_t(timer_expires - jiffies),
2033 icsk->icsk_retransmits,
2034 sock_i_uid(sp),
2035 icsk->icsk_probes_out,
2036 sock_i_ino(sp),
2037 atomic_read(&sp->sk_refcnt), sp,
2038 icsk->icsk_rto,
2039 icsk->icsk_ack.ato,
2040 (icsk->icsk_ack.quick << 1 ) | icsk->icsk_ack.pingpong,
2041 tp->snd_cwnd, tp->snd_ssthresh>=0xFFFF?-1:tp->snd_ssthresh
2045 static void get_timewait6_sock(struct seq_file *seq,
2046 struct inet_timewait_sock *tw, int i)
2048 struct in6_addr *dest, *src;
2049 __u16 destp, srcp;
2050 struct inet6_timewait_sock *tw6 = inet6_twsk((struct sock *)tw);
2051 int ttd = tw->tw_ttd - jiffies;
2053 if (ttd < 0)
2054 ttd = 0;
2056 dest = &tw6->tw_v6_daddr;
2057 src = &tw6->tw_v6_rcv_saddr;
2058 destp = ntohs(tw->tw_dport);
2059 srcp = ntohs(tw->tw_sport);
2061 seq_printf(seq,
2062 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
2063 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %p\n",
2065 src->s6_addr32[0], src->s6_addr32[1],
2066 src->s6_addr32[2], src->s6_addr32[3], srcp,
2067 dest->s6_addr32[0], dest->s6_addr32[1],
2068 dest->s6_addr32[2], dest->s6_addr32[3], destp,
2069 tw->tw_substate, 0, 0,
2070 3, jiffies_to_clock_t(ttd), 0, 0, 0, 0,
2071 atomic_read(&tw->tw_refcnt), tw);
2074 static int tcp6_seq_show(struct seq_file *seq, void *v)
2076 struct tcp_iter_state *st;
2078 if (v == SEQ_START_TOKEN) {
2079 seq_puts(seq,
2080 " sl "
2081 "local_address "
2082 "remote_address "
2083 "st tx_queue rx_queue tr tm->when retrnsmt"
2084 " uid timeout inode\n");
2085 goto out;
2087 st = seq->private;
2089 switch (st->state) {
2090 case TCP_SEQ_STATE_LISTENING:
2091 case TCP_SEQ_STATE_ESTABLISHED:
2092 get_tcp6_sock(seq, v, st->num);
2093 break;
2094 case TCP_SEQ_STATE_OPENREQ:
2095 get_openreq6(seq, st->syn_wait_sk, v, st->num, st->uid);
2096 break;
2097 case TCP_SEQ_STATE_TIME_WAIT:
2098 get_timewait6_sock(seq, v, st->num);
2099 break;
2101 out:
2102 return 0;
2105 static struct file_operations tcp6_seq_fops;
2106 static struct tcp_seq_afinfo tcp6_seq_afinfo = {
2107 .owner = THIS_MODULE,
2108 .name = "tcp6",
2109 .family = AF_INET6,
2110 .seq_show = tcp6_seq_show,
2111 .seq_fops = &tcp6_seq_fops,
2114 int __init tcp6_proc_init(void)
2116 return tcp_proc_register(&tcp6_seq_afinfo);
2119 void tcp6_proc_exit(void)
2121 tcp_proc_unregister(&tcp6_seq_afinfo);
2123 #endif
2125 struct proto tcpv6_prot = {
2126 .name = "TCPv6",
2127 .owner = THIS_MODULE,
2128 .close = tcp_close,
2129 .connect = tcp_v6_connect,
2130 .disconnect = tcp_disconnect,
2131 .accept = inet_csk_accept,
2132 .ioctl = tcp_ioctl,
2133 .init = tcp_v6_init_sock,
2134 .destroy = tcp_v6_destroy_sock,
2135 .shutdown = tcp_shutdown,
2136 .setsockopt = tcp_setsockopt,
2137 .getsockopt = tcp_getsockopt,
2138 .recvmsg = tcp_recvmsg,
2139 .backlog_rcv = tcp_v6_do_rcv,
2140 .hash = tcp_v6_hash,
2141 .unhash = tcp_unhash,
2142 .get_port = tcp_v6_get_port,
2143 .enter_memory_pressure = tcp_enter_memory_pressure,
2144 .sockets_allocated = &tcp_sockets_allocated,
2145 .memory_allocated = &tcp_memory_allocated,
2146 .memory_pressure = &tcp_memory_pressure,
2147 .orphan_count = &tcp_orphan_count,
2148 .sysctl_mem = sysctl_tcp_mem,
2149 .sysctl_wmem = sysctl_tcp_wmem,
2150 .sysctl_rmem = sysctl_tcp_rmem,
2151 .max_header = MAX_TCP_HEADER,
2152 .obj_size = sizeof(struct tcp6_sock),
2153 .twsk_prot = &tcp6_timewait_sock_ops,
2154 .rsk_prot = &tcp6_request_sock_ops,
2155 #ifdef CONFIG_COMPAT
2156 .compat_setsockopt = compat_tcp_setsockopt,
2157 .compat_getsockopt = compat_tcp_getsockopt,
2158 #endif
2161 static struct inet6_protocol tcpv6_protocol = {
2162 .handler = tcp_v6_rcv,
2163 .err_handler = tcp_v6_err,
2164 .gso_send_check = tcp_v6_gso_send_check,
2165 .gso_segment = tcp_tso_segment,
2166 .flags = INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
2169 static struct inet_protosw tcpv6_protosw = {
2170 .type = SOCK_STREAM,
2171 .protocol = IPPROTO_TCP,
2172 .prot = &tcpv6_prot,
2173 .ops = &inet6_stream_ops,
2174 .capability = -1,
2175 .no_check = 0,
2176 .flags = INET_PROTOSW_PERMANENT |
2177 INET_PROTOSW_ICSK,
2180 void __init tcpv6_init(void)
2182 /* register inet6 protocol */
2183 if (inet6_add_protocol(&tcpv6_protocol, IPPROTO_TCP) < 0)
2184 printk(KERN_ERR "tcpv6_init: Could not register protocol\n");
2185 inet6_register_protosw(&tcpv6_protosw);
2187 if (inet_csk_ctl_sock_create(&tcp6_socket, PF_INET6, SOCK_RAW,
2188 IPPROTO_TCP) < 0)
2189 panic("Failed to create the TCPv6 control socket.\n");