3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
10 * linux/net/ipv4/tcp_input.c
11 * linux/net/ipv4/tcp_output.c
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>
35 #include <linux/in6.h>
36 #include <linux/netdevice.h>
37 #include <linux/init.h>
38 #include <linux/jhash.h>
39 #include <linux/ipsec.h>
40 #include <linux/times.h>
41 #include <linux/slab.h>
42 #include <linux/uaccess.h>
43 #include <linux/ipv6.h>
44 #include <linux/icmpv6.h>
45 #include <linux/random.h>
48 #include <net/ndisc.h>
49 #include <net/inet6_hashtables.h>
50 #include <net/inet6_connection_sock.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>
60 #include <net/dsfield.h>
61 #include <net/timewait_sock.h>
62 #include <net/inet_common.h>
63 #include <net/secure_seq.h>
64 #include <net/tcp_memcontrol.h>
65 #include <net/busy_poll.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
;
85 static struct tcp_md5sig_key
*tcp_v6_md5_do_lookup(struct sock
*sk
,
86 const struct in6_addr
*addr
)
92 static void inet6_sk_rx_dst_set(struct sock
*sk
, const struct sk_buff
*skb
)
94 struct dst_entry
*dst
= skb_dst(skb
);
97 const struct rt6_info
*rt
= (const struct rt6_info
*)dst
;
101 inet_sk(sk
)->rx_dst_ifindex
= skb
->skb_iif
;
103 inet6_sk(sk
)->rx_dst_cookie
= rt
->rt6i_node
->fn_sernum
;
107 static void tcp_v6_hash(struct sock
*sk
)
109 if (sk
->sk_state
!= TCP_CLOSE
) {
110 if (inet_csk(sk
)->icsk_af_ops
== &ipv6_mapped
) {
115 __inet6_hash(sk
, NULL
);
120 static __u32
tcp_v6_init_sequence(const struct sk_buff
*skb
)
122 return secure_tcpv6_sequence_number(ipv6_hdr(skb
)->daddr
.s6_addr32
,
123 ipv6_hdr(skb
)->saddr
.s6_addr32
,
125 tcp_hdr(skb
)->source
);
128 static int tcp_v6_connect(struct sock
*sk
, struct sockaddr
*uaddr
,
131 struct sockaddr_in6
*usin
= (struct sockaddr_in6
*) uaddr
;
132 struct inet_sock
*inet
= inet_sk(sk
);
133 struct inet_connection_sock
*icsk
= inet_csk(sk
);
134 struct ipv6_pinfo
*np
= inet6_sk(sk
);
135 struct tcp_sock
*tp
= tcp_sk(sk
);
136 struct in6_addr
*saddr
= NULL
, *final_p
, final
;
139 struct dst_entry
*dst
;
143 if (addr_len
< SIN6_LEN_RFC2133
)
146 if (usin
->sin6_family
!= AF_INET6
)
147 return -EAFNOSUPPORT
;
149 memset(&fl6
, 0, sizeof(fl6
));
152 fl6
.flowlabel
= usin
->sin6_flowinfo
&IPV6_FLOWINFO_MASK
;
153 IP6_ECN_flow_init(fl6
.flowlabel
);
154 if (fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) {
155 struct ip6_flowlabel
*flowlabel
;
156 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
157 if (flowlabel
== NULL
)
159 fl6_sock_release(flowlabel
);
164 * connect() to INADDR_ANY means loopback (BSD'ism).
167 if (ipv6_addr_any(&usin
->sin6_addr
))
168 usin
->sin6_addr
.s6_addr
[15] = 0x1;
170 addr_type
= ipv6_addr_type(&usin
->sin6_addr
);
172 if (addr_type
& IPV6_ADDR_MULTICAST
)
175 if (addr_type
&IPV6_ADDR_LINKLOCAL
) {
176 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
177 usin
->sin6_scope_id
) {
178 /* If interface is set while binding, indices
181 if (sk
->sk_bound_dev_if
&&
182 sk
->sk_bound_dev_if
!= usin
->sin6_scope_id
)
185 sk
->sk_bound_dev_if
= usin
->sin6_scope_id
;
188 /* Connect to link-local address requires an interface */
189 if (!sk
->sk_bound_dev_if
)
193 if (tp
->rx_opt
.ts_recent_stamp
&&
194 !ipv6_addr_equal(&sk
->sk_v6_daddr
, &usin
->sin6_addr
)) {
195 tp
->rx_opt
.ts_recent
= 0;
196 tp
->rx_opt
.ts_recent_stamp
= 0;
200 sk
->sk_v6_daddr
= usin
->sin6_addr
;
201 np
->flow_label
= fl6
.flowlabel
;
209 if (addr_type
== IPV6_ADDR_MAPPED
) {
210 u32 exthdrlen
= icsk
->icsk_ext_hdr_len
;
211 struct sockaddr_in sin
;
213 SOCK_DEBUG(sk
, "connect: ipv4 mapped\n");
215 if (__ipv6_only_sock(sk
))
218 sin
.sin_family
= AF_INET
;
219 sin
.sin_port
= usin
->sin6_port
;
220 sin
.sin_addr
.s_addr
= usin
->sin6_addr
.s6_addr32
[3];
222 icsk
->icsk_af_ops
= &ipv6_mapped
;
223 sk
->sk_backlog_rcv
= tcp_v4_do_rcv
;
224 #ifdef CONFIG_TCP_MD5SIG
225 tp
->af_specific
= &tcp_sock_ipv6_mapped_specific
;
228 err
= tcp_v4_connect(sk
, (struct sockaddr
*)&sin
, sizeof(sin
));
231 icsk
->icsk_ext_hdr_len
= exthdrlen
;
232 icsk
->icsk_af_ops
= &ipv6_specific
;
233 sk
->sk_backlog_rcv
= tcp_v6_do_rcv
;
234 #ifdef CONFIG_TCP_MD5SIG
235 tp
->af_specific
= &tcp_sock_ipv6_specific
;
239 ipv6_addr_set_v4mapped(inet
->inet_saddr
, &np
->saddr
);
240 ipv6_addr_set_v4mapped(inet
->inet_rcv_saddr
,
241 &sk
->sk_v6_rcv_saddr
);
247 if (!ipv6_addr_any(&sk
->sk_v6_rcv_saddr
))
248 saddr
= &sk
->sk_v6_rcv_saddr
;
250 fl6
.flowi6_proto
= IPPROTO_TCP
;
251 fl6
.daddr
= sk
->sk_v6_daddr
;
252 fl6
.saddr
= saddr
? *saddr
: np
->saddr
;
253 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
254 fl6
.flowi6_mark
= sk
->sk_mark
;
255 fl6
.fl6_dport
= usin
->sin6_port
;
256 fl6
.fl6_sport
= inet
->inet_sport
;
258 final_p
= fl6_update_dst(&fl6
, np
->opt
, &final
);
260 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
262 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
);
270 sk
->sk_v6_rcv_saddr
= *saddr
;
273 /* set the source address */
275 inet
->inet_rcv_saddr
= LOOPBACK4_IPV6
;
277 sk
->sk_gso_type
= SKB_GSO_TCPV6
;
278 __ip6_dst_store(sk
, dst
, NULL
, NULL
);
280 rt
= (struct rt6_info
*) dst
;
281 if (tcp_death_row
.sysctl_tw_recycle
&&
282 !tp
->rx_opt
.ts_recent_stamp
&&
283 ipv6_addr_equal(&rt
->rt6i_dst
.addr
, &sk
->sk_v6_daddr
))
284 tcp_fetch_timewait_stamp(sk
, dst
);
286 icsk
->icsk_ext_hdr_len
= 0;
288 icsk
->icsk_ext_hdr_len
= (np
->opt
->opt_flen
+
291 tp
->rx_opt
.mss_clamp
= IPV6_MIN_MTU
- sizeof(struct tcphdr
) - sizeof(struct ipv6hdr
);
293 inet
->inet_dport
= usin
->sin6_port
;
295 tcp_set_state(sk
, TCP_SYN_SENT
);
296 err
= inet6_hash_connect(&tcp_death_row
, sk
);
300 if (!tp
->write_seq
&& likely(!tp
->repair
))
301 tp
->write_seq
= secure_tcpv6_sequence_number(np
->saddr
.s6_addr32
,
302 sk
->sk_v6_daddr
.s6_addr32
,
306 err
= tcp_connect(sk
);
313 tcp_set_state(sk
, TCP_CLOSE
);
316 inet
->inet_dport
= 0;
317 sk
->sk_route_caps
= 0;
321 static void tcp_v6_mtu_reduced(struct sock
*sk
)
323 struct dst_entry
*dst
;
325 if ((1 << sk
->sk_state
) & (TCPF_LISTEN
| TCPF_CLOSE
))
328 dst
= inet6_csk_update_pmtu(sk
, tcp_sk(sk
)->mtu_info
);
332 if (inet_csk(sk
)->icsk_pmtu_cookie
> dst_mtu(dst
)) {
333 tcp_sync_mss(sk
, dst_mtu(dst
));
334 tcp_simple_retransmit(sk
);
338 static void tcp_v6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
339 u8 type
, u8 code
, int offset
, __be32 info
)
341 const struct ipv6hdr
*hdr
= (const struct ipv6hdr
*)skb
->data
;
342 const struct tcphdr
*th
= (struct tcphdr
*)(skb
->data
+offset
);
343 struct ipv6_pinfo
*np
;
347 struct request_sock
*fastopen
;
349 struct net
*net
= dev_net(skb
->dev
);
351 sk
= inet6_lookup(net
, &tcp_hashinfo
, &hdr
->daddr
,
352 th
->dest
, &hdr
->saddr
, th
->source
, skb
->dev
->ifindex
);
355 ICMP6_INC_STATS_BH(net
, __in6_dev_get(skb
->dev
),
360 if (sk
->sk_state
== TCP_TIME_WAIT
) {
361 inet_twsk_put(inet_twsk(sk
));
366 if (sock_owned_by_user(sk
) && type
!= ICMPV6_PKT_TOOBIG
)
367 NET_INC_STATS_BH(net
, LINUX_MIB_LOCKDROPPEDICMPS
);
369 if (sk
->sk_state
== TCP_CLOSE
)
372 if (ipv6_hdr(skb
)->hop_limit
< inet6_sk(sk
)->min_hopcount
) {
373 NET_INC_STATS_BH(net
, LINUX_MIB_TCPMINTTLDROP
);
378 seq
= ntohl(th
->seq
);
379 /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
380 fastopen
= tp
->fastopen_rsk
;
381 snd_una
= fastopen
? tcp_rsk(fastopen
)->snt_isn
: tp
->snd_una
;
382 if (sk
->sk_state
!= TCP_LISTEN
&&
383 !between(seq
, snd_una
, tp
->snd_nxt
)) {
384 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
390 if (type
== NDISC_REDIRECT
) {
391 struct dst_entry
*dst
= __sk_dst_check(sk
, np
->dst_cookie
);
394 dst
->ops
->redirect(dst
, sk
, skb
);
398 if (type
== ICMPV6_PKT_TOOBIG
) {
399 /* We are not interested in TCP_LISTEN and open_requests
400 * (SYN-ACKs send out by Linux are always <576bytes so
401 * they should go through unfragmented).
403 if (sk
->sk_state
== TCP_LISTEN
)
406 if (!ip6_sk_accept_pmtu(sk
))
409 tp
->mtu_info
= ntohl(info
);
410 if (!sock_owned_by_user(sk
))
411 tcp_v6_mtu_reduced(sk
);
412 else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED
,
418 icmpv6_err_convert(type
, code
, &err
);
420 /* Might be for an request_sock */
421 switch (sk
->sk_state
) {
422 struct request_sock
*req
, **prev
;
424 if (sock_owned_by_user(sk
))
427 req
= inet6_csk_search_req(sk
, &prev
, th
->dest
, &hdr
->daddr
,
428 &hdr
->saddr
, inet6_iif(skb
));
432 /* ICMPs are not backlogged, hence we cannot get
433 * an established socket here.
435 WARN_ON(req
->sk
!= NULL
);
437 if (seq
!= tcp_rsk(req
)->snt_isn
) {
438 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
442 inet_csk_reqsk_queue_drop(sk
, req
, prev
);
443 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
448 /* Only in fast or simultaneous open. If a fast open socket is
449 * is already accepted it is treated as a connected one below.
451 if (fastopen
&& fastopen
->sk
== NULL
)
454 if (!sock_owned_by_user(sk
)) {
456 sk
->sk_error_report(sk
); /* Wake people up to see the error (see connect in sock.c) */
460 sk
->sk_err_soft
= err
;
464 if (!sock_owned_by_user(sk
) && np
->recverr
) {
466 sk
->sk_error_report(sk
);
468 sk
->sk_err_soft
= err
;
476 static int tcp_v6_send_synack(struct sock
*sk
, struct dst_entry
*dst
,
478 struct request_sock
*req
,
480 struct tcp_fastopen_cookie
*foc
)
482 struct inet_request_sock
*ireq
= inet_rsk(req
);
483 struct ipv6_pinfo
*np
= inet6_sk(sk
);
484 struct flowi6
*fl6
= &fl
->u
.ip6
;
488 /* First, grab a route. */
489 if (!dst
&& (dst
= inet6_csk_route_req(sk
, fl6
, req
)) == NULL
)
492 skb
= tcp_make_synack(sk
, dst
, req
, foc
);
495 __tcp_v6_send_check(skb
, &ireq
->ir_v6_loc_addr
,
496 &ireq
->ir_v6_rmt_addr
);
498 fl6
->daddr
= ireq
->ir_v6_rmt_addr
;
499 if (np
->repflow
&& (ireq
->pktopts
!= NULL
))
500 fl6
->flowlabel
= ip6_flowlabel(ipv6_hdr(ireq
->pktopts
));
502 skb_set_queue_mapping(skb
, queue_mapping
);
503 err
= ip6_xmit(sk
, skb
, fl6
, np
->opt
, np
->tclass
);
504 err
= net_xmit_eval(err
);
512 static void tcp_v6_reqsk_destructor(struct request_sock
*req
)
514 kfree_skb(inet_rsk(req
)->pktopts
);
517 #ifdef CONFIG_TCP_MD5SIG
518 static struct tcp_md5sig_key
*tcp_v6_md5_do_lookup(struct sock
*sk
,
519 const struct in6_addr
*addr
)
521 return tcp_md5_do_lookup(sk
, (union tcp_md5_addr
*)addr
, AF_INET6
);
524 static struct tcp_md5sig_key
*tcp_v6_md5_lookup(struct sock
*sk
,
525 struct sock
*addr_sk
)
527 return tcp_v6_md5_do_lookup(sk
, &addr_sk
->sk_v6_daddr
);
530 static struct tcp_md5sig_key
*tcp_v6_reqsk_md5_lookup(struct sock
*sk
,
531 struct request_sock
*req
)
533 return tcp_v6_md5_do_lookup(sk
, &inet_rsk(req
)->ir_v6_rmt_addr
);
536 static int tcp_v6_parse_md5_keys(struct sock
*sk
, char __user
*optval
,
539 struct tcp_md5sig cmd
;
540 struct sockaddr_in6
*sin6
= (struct sockaddr_in6
*)&cmd
.tcpm_addr
;
542 if (optlen
< sizeof(cmd
))
545 if (copy_from_user(&cmd
, optval
, sizeof(cmd
)))
548 if (sin6
->sin6_family
!= AF_INET6
)
551 if (!cmd
.tcpm_keylen
) {
552 if (ipv6_addr_v4mapped(&sin6
->sin6_addr
))
553 return tcp_md5_do_del(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
.s6_addr32
[3],
555 return tcp_md5_do_del(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
,
559 if (cmd
.tcpm_keylen
> TCP_MD5SIG_MAXKEYLEN
)
562 if (ipv6_addr_v4mapped(&sin6
->sin6_addr
))
563 return tcp_md5_do_add(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
.s6_addr32
[3],
564 AF_INET
, cmd
.tcpm_key
, cmd
.tcpm_keylen
, GFP_KERNEL
);
566 return tcp_md5_do_add(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
,
567 AF_INET6
, cmd
.tcpm_key
, cmd
.tcpm_keylen
, GFP_KERNEL
);
570 static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool
*hp
,
571 const struct in6_addr
*daddr
,
572 const struct in6_addr
*saddr
, int nbytes
)
574 struct tcp6_pseudohdr
*bp
;
575 struct scatterlist sg
;
577 bp
= &hp
->md5_blk
.ip6
;
578 /* 1. TCP pseudo-header (RFC2460) */
581 bp
->protocol
= cpu_to_be32(IPPROTO_TCP
);
582 bp
->len
= cpu_to_be32(nbytes
);
584 sg_init_one(&sg
, bp
, sizeof(*bp
));
585 return crypto_hash_update(&hp
->md5_desc
, &sg
, sizeof(*bp
));
588 static int tcp_v6_md5_hash_hdr(char *md5_hash
, struct tcp_md5sig_key
*key
,
589 const struct in6_addr
*daddr
, struct in6_addr
*saddr
,
590 const struct tcphdr
*th
)
592 struct tcp_md5sig_pool
*hp
;
593 struct hash_desc
*desc
;
595 hp
= tcp_get_md5sig_pool();
597 goto clear_hash_noput
;
598 desc
= &hp
->md5_desc
;
600 if (crypto_hash_init(desc
))
602 if (tcp_v6_md5_hash_pseudoheader(hp
, daddr
, saddr
, th
->doff
<< 2))
604 if (tcp_md5_hash_header(hp
, th
))
606 if (tcp_md5_hash_key(hp
, key
))
608 if (crypto_hash_final(desc
, md5_hash
))
611 tcp_put_md5sig_pool();
615 tcp_put_md5sig_pool();
617 memset(md5_hash
, 0, 16);
621 static int tcp_v6_md5_hash_skb(char *md5_hash
, struct tcp_md5sig_key
*key
,
622 const struct sock
*sk
,
623 const struct request_sock
*req
,
624 const struct sk_buff
*skb
)
626 const struct in6_addr
*saddr
, *daddr
;
627 struct tcp_md5sig_pool
*hp
;
628 struct hash_desc
*desc
;
629 const struct tcphdr
*th
= tcp_hdr(skb
);
632 saddr
= &inet6_sk(sk
)->saddr
;
633 daddr
= &sk
->sk_v6_daddr
;
635 saddr
= &inet_rsk(req
)->ir_v6_loc_addr
;
636 daddr
= &inet_rsk(req
)->ir_v6_rmt_addr
;
638 const struct ipv6hdr
*ip6h
= ipv6_hdr(skb
);
639 saddr
= &ip6h
->saddr
;
640 daddr
= &ip6h
->daddr
;
643 hp
= tcp_get_md5sig_pool();
645 goto clear_hash_noput
;
646 desc
= &hp
->md5_desc
;
648 if (crypto_hash_init(desc
))
651 if (tcp_v6_md5_hash_pseudoheader(hp
, daddr
, saddr
, skb
->len
))
653 if (tcp_md5_hash_header(hp
, th
))
655 if (tcp_md5_hash_skb_data(hp
, skb
, th
->doff
<< 2))
657 if (tcp_md5_hash_key(hp
, key
))
659 if (crypto_hash_final(desc
, md5_hash
))
662 tcp_put_md5sig_pool();
666 tcp_put_md5sig_pool();
668 memset(md5_hash
, 0, 16);
672 static int __tcp_v6_inbound_md5_hash(struct sock
*sk
,
673 const struct sk_buff
*skb
)
675 const __u8
*hash_location
= NULL
;
676 struct tcp_md5sig_key
*hash_expected
;
677 const struct ipv6hdr
*ip6h
= ipv6_hdr(skb
);
678 const struct tcphdr
*th
= tcp_hdr(skb
);
682 hash_expected
= tcp_v6_md5_do_lookup(sk
, &ip6h
->saddr
);
683 hash_location
= tcp_parse_md5sig_option(th
);
685 /* We've parsed the options - do we have a hash? */
686 if (!hash_expected
&& !hash_location
)
689 if (hash_expected
&& !hash_location
) {
690 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_TCPMD5NOTFOUND
);
694 if (!hash_expected
&& hash_location
) {
695 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_TCPMD5UNEXPECTED
);
699 /* check the signature */
700 genhash
= tcp_v6_md5_hash_skb(newhash
,
704 if (genhash
|| memcmp(hash_location
, newhash
, 16) != 0) {
705 net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n",
706 genhash
? "failed" : "mismatch",
707 &ip6h
->saddr
, ntohs(th
->source
),
708 &ip6h
->daddr
, ntohs(th
->dest
));
714 static int tcp_v6_inbound_md5_hash(struct sock
*sk
, const struct sk_buff
*skb
)
719 ret
= __tcp_v6_inbound_md5_hash(sk
, skb
);
727 static void tcp_v6_init_req(struct request_sock
*req
, struct sock
*sk
,
730 struct inet_request_sock
*ireq
= inet_rsk(req
);
731 struct ipv6_pinfo
*np
= inet6_sk(sk
);
733 ireq
->ir_v6_rmt_addr
= ipv6_hdr(skb
)->saddr
;
734 ireq
->ir_v6_loc_addr
= ipv6_hdr(skb
)->daddr
;
736 ireq
->ir_iif
= sk
->sk_bound_dev_if
;
738 /* So that link locals have meaning */
739 if (!sk
->sk_bound_dev_if
&&
740 ipv6_addr_type(&ireq
->ir_v6_rmt_addr
) & IPV6_ADDR_LINKLOCAL
)
741 ireq
->ir_iif
= inet6_iif(skb
);
743 if (!TCP_SKB_CB(skb
)->tcp_tw_isn
&&
744 (ipv6_opt_accepted(sk
, skb
, &TCP_SKB_CB(skb
)->header
.h6
) ||
745 np
->rxopt
.bits
.rxinfo
||
746 np
->rxopt
.bits
.rxoinfo
|| np
->rxopt
.bits
.rxhlim
||
747 np
->rxopt
.bits
.rxohlim
|| np
->repflow
)) {
748 atomic_inc(&skb
->users
);
753 static struct dst_entry
*tcp_v6_route_req(struct sock
*sk
, struct flowi
*fl
,
754 const struct request_sock
*req
,
759 return inet6_csk_route_req(sk
, &fl
->u
.ip6
, req
);
762 struct request_sock_ops tcp6_request_sock_ops __read_mostly
= {
764 .obj_size
= sizeof(struct tcp6_request_sock
),
765 .rtx_syn_ack
= tcp_rtx_synack
,
766 .send_ack
= tcp_v6_reqsk_send_ack
,
767 .destructor
= tcp_v6_reqsk_destructor
,
768 .send_reset
= tcp_v6_send_reset
,
769 .syn_ack_timeout
= tcp_syn_ack_timeout
,
772 static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops
= {
773 .mss_clamp
= IPV6_MIN_MTU
- sizeof(struct tcphdr
) -
774 sizeof(struct ipv6hdr
),
775 #ifdef CONFIG_TCP_MD5SIG
776 .md5_lookup
= tcp_v6_reqsk_md5_lookup
,
777 .calc_md5_hash
= tcp_v6_md5_hash_skb
,
779 .init_req
= tcp_v6_init_req
,
780 #ifdef CONFIG_SYN_COOKIES
781 .cookie_init_seq
= cookie_v6_init_sequence
,
783 .route_req
= tcp_v6_route_req
,
784 .init_seq
= tcp_v6_init_sequence
,
785 .send_synack
= tcp_v6_send_synack
,
786 .queue_hash_add
= inet6_csk_reqsk_queue_hash_add
,
789 static void tcp_v6_send_response(struct sk_buff
*skb
, u32 seq
, u32 ack
, u32 win
,
790 u32 tsval
, u32 tsecr
, int oif
,
791 struct tcp_md5sig_key
*key
, int rst
, u8 tclass
,
794 const struct tcphdr
*th
= tcp_hdr(skb
);
796 struct sk_buff
*buff
;
798 struct net
*net
= dev_net(skb_dst(skb
)->dev
);
799 struct sock
*ctl_sk
= net
->ipv6
.tcp_sk
;
800 unsigned int tot_len
= sizeof(struct tcphdr
);
801 struct dst_entry
*dst
;
805 tot_len
+= TCPOLEN_TSTAMP_ALIGNED
;
806 #ifdef CONFIG_TCP_MD5SIG
808 tot_len
+= TCPOLEN_MD5SIG_ALIGNED
;
811 buff
= alloc_skb(MAX_HEADER
+ sizeof(struct ipv6hdr
) + tot_len
,
816 skb_reserve(buff
, MAX_HEADER
+ sizeof(struct ipv6hdr
) + tot_len
);
818 t1
= (struct tcphdr
*) skb_push(buff
, tot_len
);
819 skb_reset_transport_header(buff
);
821 /* Swap the send and the receive. */
822 memset(t1
, 0, sizeof(*t1
));
823 t1
->dest
= th
->source
;
824 t1
->source
= th
->dest
;
825 t1
->doff
= tot_len
/ 4;
826 t1
->seq
= htonl(seq
);
827 t1
->ack_seq
= htonl(ack
);
828 t1
->ack
= !rst
|| !th
->ack
;
830 t1
->window
= htons(win
);
832 topt
= (__be32
*)(t1
+ 1);
835 *topt
++ = htonl((TCPOPT_NOP
<< 24) | (TCPOPT_NOP
<< 16) |
836 (TCPOPT_TIMESTAMP
<< 8) | TCPOLEN_TIMESTAMP
);
837 *topt
++ = htonl(tsval
);
838 *topt
++ = htonl(tsecr
);
841 #ifdef CONFIG_TCP_MD5SIG
843 *topt
++ = htonl((TCPOPT_NOP
<< 24) | (TCPOPT_NOP
<< 16) |
844 (TCPOPT_MD5SIG
<< 8) | TCPOLEN_MD5SIG
);
845 tcp_v6_md5_hash_hdr((__u8
*)topt
, key
,
846 &ipv6_hdr(skb
)->saddr
,
847 &ipv6_hdr(skb
)->daddr
, t1
);
851 memset(&fl6
, 0, sizeof(fl6
));
852 fl6
.daddr
= ipv6_hdr(skb
)->saddr
;
853 fl6
.saddr
= ipv6_hdr(skb
)->daddr
;
854 fl6
.flowlabel
= label
;
856 buff
->ip_summed
= CHECKSUM_PARTIAL
;
859 __tcp_v6_send_check(buff
, &fl6
.saddr
, &fl6
.daddr
);
861 fl6
.flowi6_proto
= IPPROTO_TCP
;
862 if (rt6_need_strict(&fl6
.daddr
) && !oif
)
863 fl6
.flowi6_oif
= inet6_iif(skb
);
865 fl6
.flowi6_oif
= oif
;
866 fl6
.flowi6_mark
= IP6_REPLY_MARK(net
, skb
->mark
);
867 fl6
.fl6_dport
= t1
->dest
;
868 fl6
.fl6_sport
= t1
->source
;
869 security_skb_classify_flow(skb
, flowi6_to_flowi(&fl6
));
871 /* Pass a socket to ip6_dst_lookup either it is for RST
872 * Underlying function will use this to retrieve the network
875 dst
= ip6_dst_lookup_flow(ctl_sk
, &fl6
, NULL
);
877 skb_dst_set(buff
, dst
);
878 ip6_xmit(ctl_sk
, buff
, &fl6
, NULL
, tclass
);
879 TCP_INC_STATS_BH(net
, TCP_MIB_OUTSEGS
);
881 TCP_INC_STATS_BH(net
, TCP_MIB_OUTRSTS
);
888 static void tcp_v6_send_reset(struct sock
*sk
, struct sk_buff
*skb
)
890 const struct tcphdr
*th
= tcp_hdr(skb
);
891 u32 seq
= 0, ack_seq
= 0;
892 struct tcp_md5sig_key
*key
= NULL
;
893 #ifdef CONFIG_TCP_MD5SIG
894 const __u8
*hash_location
= NULL
;
895 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
896 unsigned char newhash
[16];
898 struct sock
*sk1
= NULL
;
905 if (!ipv6_unicast_destination(skb
))
908 #ifdef CONFIG_TCP_MD5SIG
909 hash_location
= tcp_parse_md5sig_option(th
);
910 if (!sk
&& hash_location
) {
912 * active side is lost. Try to find listening socket through
913 * source port, and then find md5 key through listening socket.
914 * we are not loose security here:
915 * Incoming packet is checked with md5 hash with finding key,
916 * no RST generated if md5 hash doesn't match.
918 sk1
= inet6_lookup_listener(dev_net(skb_dst(skb
)->dev
),
919 &tcp_hashinfo
, &ipv6h
->saddr
,
920 th
->source
, &ipv6h
->daddr
,
921 ntohs(th
->source
), inet6_iif(skb
));
926 key
= tcp_v6_md5_do_lookup(sk1
, &ipv6h
->saddr
);
930 genhash
= tcp_v6_md5_hash_skb(newhash
, key
, NULL
, NULL
, skb
);
931 if (genhash
|| memcmp(hash_location
, newhash
, 16) != 0)
934 key
= sk
? tcp_v6_md5_do_lookup(sk
, &ipv6h
->saddr
) : NULL
;
939 seq
= ntohl(th
->ack_seq
);
941 ack_seq
= ntohl(th
->seq
) + th
->syn
+ th
->fin
+ skb
->len
-
944 oif
= sk
? sk
->sk_bound_dev_if
: 0;
945 tcp_v6_send_response(skb
, seq
, ack_seq
, 0, 0, 0, oif
, key
, 1, 0, 0);
947 #ifdef CONFIG_TCP_MD5SIG
956 static void tcp_v6_send_ack(struct sk_buff
*skb
, u32 seq
, u32 ack
,
957 u32 win
, u32 tsval
, u32 tsecr
, int oif
,
958 struct tcp_md5sig_key
*key
, u8 tclass
,
961 tcp_v6_send_response(skb
, seq
, ack
, win
, tsval
, tsecr
, oif
, key
, 0, tclass
,
965 static void tcp_v6_timewait_ack(struct sock
*sk
, struct sk_buff
*skb
)
967 struct inet_timewait_sock
*tw
= inet_twsk(sk
);
968 struct tcp_timewait_sock
*tcptw
= tcp_twsk(sk
);
970 tcp_v6_send_ack(skb
, tcptw
->tw_snd_nxt
, tcptw
->tw_rcv_nxt
,
971 tcptw
->tw_rcv_wnd
>> tw
->tw_rcv_wscale
,
972 tcp_time_stamp
+ tcptw
->tw_ts_offset
,
973 tcptw
->tw_ts_recent
, tw
->tw_bound_dev_if
, tcp_twsk_md5_key(tcptw
),
974 tw
->tw_tclass
, (tw
->tw_flowlabel
<< 12));
979 static void tcp_v6_reqsk_send_ack(struct sock
*sk
, struct sk_buff
*skb
,
980 struct request_sock
*req
)
982 /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
983 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
985 tcp_v6_send_ack(skb
, (sk
->sk_state
== TCP_LISTEN
) ?
986 tcp_rsk(req
)->snt_isn
+ 1 : tcp_sk(sk
)->snd_nxt
,
987 tcp_rsk(req
)->rcv_nxt
,
988 req
->rcv_wnd
, tcp_time_stamp
, req
->ts_recent
, sk
->sk_bound_dev_if
,
989 tcp_v6_md5_do_lookup(sk
, &ipv6_hdr(skb
)->daddr
),
994 static struct sock
*tcp_v6_hnd_req(struct sock
*sk
, struct sk_buff
*skb
)
996 struct request_sock
*req
, **prev
;
997 const struct tcphdr
*th
= tcp_hdr(skb
);
1000 /* Find possible connection requests. */
1001 req
= inet6_csk_search_req(sk
, &prev
, th
->source
,
1002 &ipv6_hdr(skb
)->saddr
,
1003 &ipv6_hdr(skb
)->daddr
, inet6_iif(skb
));
1005 return tcp_check_req(sk
, skb
, req
, prev
, false);
1007 nsk
= __inet6_lookup_established(sock_net(sk
), &tcp_hashinfo
,
1008 &ipv6_hdr(skb
)->saddr
, th
->source
,
1009 &ipv6_hdr(skb
)->daddr
, ntohs(th
->dest
), inet6_iif(skb
));
1012 if (nsk
->sk_state
!= TCP_TIME_WAIT
) {
1016 inet_twsk_put(inet_twsk(nsk
));
1020 #ifdef CONFIG_SYN_COOKIES
1022 sk
= cookie_v6_check(sk
, skb
);
1027 static int tcp_v6_conn_request(struct sock
*sk
, struct sk_buff
*skb
)
1029 if (skb
->protocol
== htons(ETH_P_IP
))
1030 return tcp_v4_conn_request(sk
, skb
);
1032 if (!ipv6_unicast_destination(skb
))
1035 return tcp_conn_request(&tcp6_request_sock_ops
,
1036 &tcp_request_sock_ipv6_ops
, sk
, skb
);
1039 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
1040 return 0; /* don't send reset */
1043 static struct sock
*tcp_v6_syn_recv_sock(struct sock
*sk
, struct sk_buff
*skb
,
1044 struct request_sock
*req
,
1045 struct dst_entry
*dst
)
1047 struct inet_request_sock
*ireq
;
1048 struct ipv6_pinfo
*newnp
, *np
= inet6_sk(sk
);
1049 struct tcp6_sock
*newtcp6sk
;
1050 struct inet_sock
*newinet
;
1051 struct tcp_sock
*newtp
;
1053 #ifdef CONFIG_TCP_MD5SIG
1054 struct tcp_md5sig_key
*key
;
1058 if (skb
->protocol
== htons(ETH_P_IP
)) {
1063 newsk
= tcp_v4_syn_recv_sock(sk
, skb
, req
, dst
);
1068 newtcp6sk
= (struct tcp6_sock
*)newsk
;
1069 inet_sk(newsk
)->pinet6
= &newtcp6sk
->inet6
;
1071 newinet
= inet_sk(newsk
);
1072 newnp
= inet6_sk(newsk
);
1073 newtp
= tcp_sk(newsk
);
1075 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
1077 ipv6_addr_set_v4mapped(newinet
->inet_daddr
, &newsk
->sk_v6_daddr
);
1079 ipv6_addr_set_v4mapped(newinet
->inet_saddr
, &newnp
->saddr
);
1081 newsk
->sk_v6_rcv_saddr
= newnp
->saddr
;
1083 inet_csk(newsk
)->icsk_af_ops
= &ipv6_mapped
;
1084 newsk
->sk_backlog_rcv
= tcp_v4_do_rcv
;
1085 #ifdef CONFIG_TCP_MD5SIG
1086 newtp
->af_specific
= &tcp_sock_ipv6_mapped_specific
;
1089 newnp
->ipv6_ac_list
= NULL
;
1090 newnp
->ipv6_fl_list
= NULL
;
1091 newnp
->pktoptions
= NULL
;
1093 newnp
->mcast_oif
= inet6_iif(skb
);
1094 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
1095 newnp
->rcv_flowinfo
= ip6_flowinfo(ipv6_hdr(skb
));
1097 newnp
->flow_label
= ip6_flowlabel(ipv6_hdr(skb
));
1100 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1101 * here, tcp_create_openreq_child now does this for us, see the comment in
1102 * that function for the gory details. -acme
1105 /* It is tricky place. Until this moment IPv4 tcp
1106 worked with IPv6 icsk.icsk_af_ops.
1109 tcp_sync_mss(newsk
, inet_csk(newsk
)->icsk_pmtu_cookie
);
1114 ireq
= inet_rsk(req
);
1116 if (sk_acceptq_is_full(sk
))
1120 dst
= inet6_csk_route_req(sk
, &fl6
, req
);
1125 newsk
= tcp_create_openreq_child(sk
, req
, skb
);
1130 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1131 * count here, tcp_create_openreq_child now does this for us, see the
1132 * comment in that function for the gory details. -acme
1135 newsk
->sk_gso_type
= SKB_GSO_TCPV6
;
1136 __ip6_dst_store(newsk
, dst
, NULL
, NULL
);
1137 inet6_sk_rx_dst_set(newsk
, skb
);
1139 newtcp6sk
= (struct tcp6_sock
*)newsk
;
1140 inet_sk(newsk
)->pinet6
= &newtcp6sk
->inet6
;
1142 newtp
= tcp_sk(newsk
);
1143 newinet
= inet_sk(newsk
);
1144 newnp
= inet6_sk(newsk
);
1146 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
1148 newsk
->sk_v6_daddr
= ireq
->ir_v6_rmt_addr
;
1149 newnp
->saddr
= ireq
->ir_v6_loc_addr
;
1150 newsk
->sk_v6_rcv_saddr
= ireq
->ir_v6_loc_addr
;
1151 newsk
->sk_bound_dev_if
= ireq
->ir_iif
;
1153 ip6_set_txhash(newsk
);
1155 /* Now IPv6 options...
1157 First: no IPv4 options.
1159 newinet
->inet_opt
= NULL
;
1160 newnp
->ipv6_ac_list
= NULL
;
1161 newnp
->ipv6_fl_list
= NULL
;
1164 newnp
->rxopt
.all
= np
->rxopt
.all
;
1166 /* Clone pktoptions received with SYN */
1167 newnp
->pktoptions
= NULL
;
1168 if (ireq
->pktopts
!= NULL
) {
1169 newnp
->pktoptions
= skb_clone(ireq
->pktopts
,
1170 sk_gfp_atomic(sk
, GFP_ATOMIC
));
1171 consume_skb(ireq
->pktopts
);
1172 ireq
->pktopts
= NULL
;
1173 if (newnp
->pktoptions
)
1174 skb_set_owner_r(newnp
->pktoptions
, newsk
);
1177 newnp
->mcast_oif
= inet6_iif(skb
);
1178 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
1179 newnp
->rcv_flowinfo
= ip6_flowinfo(ipv6_hdr(skb
));
1181 newnp
->flow_label
= ip6_flowlabel(ipv6_hdr(skb
));
1183 /* Clone native IPv6 options from listening socket (if any)
1185 Yes, keeping reference count would be much more clever,
1186 but we make one more one thing there: reattach optmem
1190 newnp
->opt
= ipv6_dup_options(newsk
, np
->opt
);
1192 inet_csk(newsk
)->icsk_ext_hdr_len
= 0;
1194 inet_csk(newsk
)->icsk_ext_hdr_len
= (newnp
->opt
->opt_nflen
+
1195 newnp
->opt
->opt_flen
);
1197 tcp_sync_mss(newsk
, dst_mtu(dst
));
1198 newtp
->advmss
= dst_metric_advmss(dst
);
1199 if (tcp_sk(sk
)->rx_opt
.user_mss
&&
1200 tcp_sk(sk
)->rx_opt
.user_mss
< newtp
->advmss
)
1201 newtp
->advmss
= tcp_sk(sk
)->rx_opt
.user_mss
;
1203 tcp_initialize_rcv_mss(newsk
);
1205 newinet
->inet_daddr
= newinet
->inet_saddr
= LOOPBACK4_IPV6
;
1206 newinet
->inet_rcv_saddr
= LOOPBACK4_IPV6
;
1208 #ifdef CONFIG_TCP_MD5SIG
1209 /* Copy over the MD5 key from the original socket */
1210 key
= tcp_v6_md5_do_lookup(sk
, &newsk
->sk_v6_daddr
);
1212 /* We're using one, so create a matching key
1213 * on the newsk structure. If we fail to get
1214 * memory, then we end up not copying the key
1217 tcp_md5_do_add(newsk
, (union tcp_md5_addr
*)&newsk
->sk_v6_daddr
,
1218 AF_INET6
, key
->key
, key
->keylen
,
1219 sk_gfp_atomic(sk
, GFP_ATOMIC
));
1223 if (__inet_inherit_port(sk
, newsk
) < 0) {
1224 inet_csk_prepare_forced_close(newsk
);
1228 __inet6_hash(newsk
, NULL
);
1233 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENOVERFLOWS
);
1237 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
1241 /* The socket must have it's spinlock held when we get
1244 * We have a potential double-lock case here, so even when
1245 * doing backlog processing we use the BH locking scheme.
1246 * This is because we cannot sleep with the original spinlock
1249 static int tcp_v6_do_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1251 struct ipv6_pinfo
*np
= inet6_sk(sk
);
1252 struct tcp_sock
*tp
;
1253 struct sk_buff
*opt_skb
= NULL
;
1255 /* Imagine: socket is IPv6. IPv4 packet arrives,
1256 goes to IPv4 receive handler and backlogged.
1257 From backlog it always goes here. Kerboom...
1258 Fortunately, tcp_rcv_established and rcv_established
1259 handle them correctly, but it is not case with
1260 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1263 if (skb
->protocol
== htons(ETH_P_IP
))
1264 return tcp_v4_do_rcv(sk
, skb
);
1266 if (sk_filter(sk
, skb
))
1270 * socket locking is here for SMP purposes as backlog rcv
1271 * is currently called with bh processing disabled.
1274 /* Do Stevens' IPV6_PKTOPTIONS.
1276 Yes, guys, it is the only place in our code, where we
1277 may make it not affecting IPv4.
1278 The rest of code is protocol independent,
1279 and I do not like idea to uglify IPv4.
1281 Actually, all the idea behind IPV6_PKTOPTIONS
1282 looks not very well thought. For now we latch
1283 options, received in the last packet, enqueued
1284 by tcp. Feel free to propose better solution.
1288 opt_skb
= skb_clone(skb
, sk_gfp_atomic(sk
, GFP_ATOMIC
));
1290 if (sk
->sk_state
== TCP_ESTABLISHED
) { /* Fast path */
1291 struct dst_entry
*dst
= sk
->sk_rx_dst
;
1293 sock_rps_save_rxhash(sk
, skb
);
1295 if (inet_sk(sk
)->rx_dst_ifindex
!= skb
->skb_iif
||
1296 dst
->ops
->check(dst
, np
->rx_dst_cookie
) == NULL
) {
1298 sk
->sk_rx_dst
= NULL
;
1302 tcp_rcv_established(sk
, skb
, tcp_hdr(skb
), skb
->len
);
1304 goto ipv6_pktoptions
;
1308 if (skb
->len
< tcp_hdrlen(skb
) || tcp_checksum_complete(skb
))
1311 if (sk
->sk_state
== TCP_LISTEN
) {
1312 struct sock
*nsk
= tcp_v6_hnd_req(sk
, skb
);
1317 * Queue it on the new socket if the new socket is active,
1318 * otherwise we just shortcircuit this and continue with
1322 sock_rps_save_rxhash(nsk
, skb
);
1323 if (tcp_child_process(sk
, nsk
, skb
))
1326 __kfree_skb(opt_skb
);
1330 sock_rps_save_rxhash(sk
, skb
);
1332 if (tcp_rcv_state_process(sk
, skb
, tcp_hdr(skb
), skb
->len
))
1335 goto ipv6_pktoptions
;
1339 tcp_v6_send_reset(sk
, skb
);
1342 __kfree_skb(opt_skb
);
1346 TCP_INC_STATS_BH(sock_net(sk
), TCP_MIB_CSUMERRORS
);
1347 TCP_INC_STATS_BH(sock_net(sk
), TCP_MIB_INERRS
);
1352 /* Do you ask, what is it?
1354 1. skb was enqueued by tcp.
1355 2. skb is added to tail of read queue, rather than out of order.
1356 3. socket is not in passive state.
1357 4. Finally, it really contains options, which user wants to receive.
1360 if (TCP_SKB_CB(opt_skb
)->end_seq
== tp
->rcv_nxt
&&
1361 !((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_LISTEN
))) {
1362 if (np
->rxopt
.bits
.rxinfo
|| np
->rxopt
.bits
.rxoinfo
)
1363 np
->mcast_oif
= inet6_iif(opt_skb
);
1364 if (np
->rxopt
.bits
.rxhlim
|| np
->rxopt
.bits
.rxohlim
)
1365 np
->mcast_hops
= ipv6_hdr(opt_skb
)->hop_limit
;
1366 if (np
->rxopt
.bits
.rxflow
|| np
->rxopt
.bits
.rxtclass
)
1367 np
->rcv_flowinfo
= ip6_flowinfo(ipv6_hdr(opt_skb
));
1369 np
->flow_label
= ip6_flowlabel(ipv6_hdr(opt_skb
));
1370 if (ipv6_opt_accepted(sk
, opt_skb
, &TCP_SKB_CB(opt_skb
)->header
.h6
)) {
1371 skb_set_owner_r(opt_skb
, sk
);
1372 opt_skb
= xchg(&np
->pktoptions
, opt_skb
);
1374 __kfree_skb(opt_skb
);
1375 opt_skb
= xchg(&np
->pktoptions
, NULL
);
1383 static int tcp_v6_rcv(struct sk_buff
*skb
)
1385 const struct tcphdr
*th
;
1386 const struct ipv6hdr
*hdr
;
1389 struct net
*net
= dev_net(skb
->dev
);
1391 if (skb
->pkt_type
!= PACKET_HOST
)
1395 * Count it even if it's bad.
1397 TCP_INC_STATS_BH(net
, TCP_MIB_INSEGS
);
1399 if (!pskb_may_pull(skb
, sizeof(struct tcphdr
)))
1404 if (th
->doff
< sizeof(struct tcphdr
)/4)
1406 if (!pskb_may_pull(skb
, th
->doff
*4))
1409 if (skb_checksum_init(skb
, IPPROTO_TCP
, ip6_compute_pseudo
))
1413 hdr
= ipv6_hdr(skb
);
1414 /* This is tricky : We move IPCB at its correct location into TCP_SKB_CB()
1415 * barrier() makes sure compiler wont play fool^Waliasing games.
1417 memmove(&TCP_SKB_CB(skb
)->header
.h6
, IP6CB(skb
),
1418 sizeof(struct inet6_skb_parm
));
1421 TCP_SKB_CB(skb
)->seq
= ntohl(th
->seq
);
1422 TCP_SKB_CB(skb
)->end_seq
= (TCP_SKB_CB(skb
)->seq
+ th
->syn
+ th
->fin
+
1423 skb
->len
- th
->doff
*4);
1424 TCP_SKB_CB(skb
)->ack_seq
= ntohl(th
->ack_seq
);
1425 TCP_SKB_CB(skb
)->tcp_flags
= tcp_flag_byte(th
);
1426 TCP_SKB_CB(skb
)->tcp_tw_isn
= 0;
1427 TCP_SKB_CB(skb
)->ip_dsfield
= ipv6_get_dsfield(hdr
);
1428 TCP_SKB_CB(skb
)->sacked
= 0;
1430 sk
= __inet6_lookup_skb(&tcp_hashinfo
, skb
, th
->source
, th
->dest
);
1435 if (sk
->sk_state
== TCP_TIME_WAIT
)
1438 if (hdr
->hop_limit
< inet6_sk(sk
)->min_hopcount
) {
1439 NET_INC_STATS_BH(net
, LINUX_MIB_TCPMINTTLDROP
);
1440 goto discard_and_relse
;
1443 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
))
1444 goto discard_and_relse
;
1446 #ifdef CONFIG_TCP_MD5SIG
1447 if (tcp_v6_inbound_md5_hash(sk
, skb
))
1448 goto discard_and_relse
;
1451 if (sk_filter(sk
, skb
))
1452 goto discard_and_relse
;
1454 sk_mark_napi_id(sk
, skb
);
1457 bh_lock_sock_nested(sk
);
1459 if (!sock_owned_by_user(sk
)) {
1460 if (!tcp_prequeue(sk
, skb
))
1461 ret
= tcp_v6_do_rcv(sk
, skb
);
1462 } else if (unlikely(sk_add_backlog(sk
, skb
,
1463 sk
->sk_rcvbuf
+ sk
->sk_sndbuf
))) {
1465 NET_INC_STATS_BH(net
, LINUX_MIB_TCPBACKLOGDROP
);
1466 goto discard_and_relse
;
1471 return ret
? -1 : 0;
1474 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
1477 if (skb
->len
< (th
->doff
<<2) || tcp_checksum_complete(skb
)) {
1479 TCP_INC_STATS_BH(net
, TCP_MIB_CSUMERRORS
);
1481 TCP_INC_STATS_BH(net
, TCP_MIB_INERRS
);
1483 tcp_v6_send_reset(NULL
, skb
);
1495 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
)) {
1496 inet_twsk_put(inet_twsk(sk
));
1500 if (skb
->len
< (th
->doff
<<2)) {
1501 inet_twsk_put(inet_twsk(sk
));
1504 if (tcp_checksum_complete(skb
)) {
1505 inet_twsk_put(inet_twsk(sk
));
1509 switch (tcp_timewait_state_process(inet_twsk(sk
), skb
, th
)) {
1514 sk2
= inet6_lookup_listener(dev_net(skb
->dev
), &tcp_hashinfo
,
1515 &ipv6_hdr(skb
)->saddr
, th
->source
,
1516 &ipv6_hdr(skb
)->daddr
,
1517 ntohs(th
->dest
), inet6_iif(skb
));
1519 struct inet_timewait_sock
*tw
= inet_twsk(sk
);
1520 inet_twsk_deschedule(tw
, &tcp_death_row
);
1525 /* Fall through to ACK */
1528 tcp_v6_timewait_ack(sk
, skb
);
1532 case TCP_TW_SUCCESS
:
1538 static void tcp_v6_early_demux(struct sk_buff
*skb
)
1540 const struct ipv6hdr
*hdr
;
1541 const struct tcphdr
*th
;
1544 if (skb
->pkt_type
!= PACKET_HOST
)
1547 if (!pskb_may_pull(skb
, skb_transport_offset(skb
) + sizeof(struct tcphdr
)))
1550 hdr
= ipv6_hdr(skb
);
1553 if (th
->doff
< sizeof(struct tcphdr
) / 4)
1556 sk
= __inet6_lookup_established(dev_net(skb
->dev
), &tcp_hashinfo
,
1557 &hdr
->saddr
, th
->source
,
1558 &hdr
->daddr
, ntohs(th
->dest
),
1562 skb
->destructor
= sock_edemux
;
1563 if (sk
->sk_state
!= TCP_TIME_WAIT
) {
1564 struct dst_entry
*dst
= sk
->sk_rx_dst
;
1567 dst
= dst_check(dst
, inet6_sk(sk
)->rx_dst_cookie
);
1569 inet_sk(sk
)->rx_dst_ifindex
== skb
->skb_iif
)
1570 skb_dst_set_noref(skb
, dst
);
1575 static struct timewait_sock_ops tcp6_timewait_sock_ops
= {
1576 .twsk_obj_size
= sizeof(struct tcp6_timewait_sock
),
1577 .twsk_unique
= tcp_twsk_unique
,
1578 .twsk_destructor
= tcp_twsk_destructor
,
1581 static const struct inet_connection_sock_af_ops ipv6_specific
= {
1582 .queue_xmit
= inet6_csk_xmit
,
1583 .send_check
= tcp_v6_send_check
,
1584 .rebuild_header
= inet6_sk_rebuild_header
,
1585 .sk_rx_dst_set
= inet6_sk_rx_dst_set
,
1586 .conn_request
= tcp_v6_conn_request
,
1587 .syn_recv_sock
= tcp_v6_syn_recv_sock
,
1588 .net_header_len
= sizeof(struct ipv6hdr
),
1589 .net_frag_header_len
= sizeof(struct frag_hdr
),
1590 .setsockopt
= ipv6_setsockopt
,
1591 .getsockopt
= ipv6_getsockopt
,
1592 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1593 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1594 .bind_conflict
= inet6_csk_bind_conflict
,
1595 #ifdef CONFIG_COMPAT
1596 .compat_setsockopt
= compat_ipv6_setsockopt
,
1597 .compat_getsockopt
= compat_ipv6_getsockopt
,
1599 .mtu_reduced
= tcp_v6_mtu_reduced
,
1602 #ifdef CONFIG_TCP_MD5SIG
1603 static const struct tcp_sock_af_ops tcp_sock_ipv6_specific
= {
1604 .md5_lookup
= tcp_v6_md5_lookup
,
1605 .calc_md5_hash
= tcp_v6_md5_hash_skb
,
1606 .md5_parse
= tcp_v6_parse_md5_keys
,
1611 * TCP over IPv4 via INET6 API
1613 static const struct inet_connection_sock_af_ops ipv6_mapped
= {
1614 .queue_xmit
= ip_queue_xmit
,
1615 .send_check
= tcp_v4_send_check
,
1616 .rebuild_header
= inet_sk_rebuild_header
,
1617 .sk_rx_dst_set
= inet_sk_rx_dst_set
,
1618 .conn_request
= tcp_v6_conn_request
,
1619 .syn_recv_sock
= tcp_v6_syn_recv_sock
,
1620 .net_header_len
= sizeof(struct iphdr
),
1621 .setsockopt
= ipv6_setsockopt
,
1622 .getsockopt
= ipv6_getsockopt
,
1623 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1624 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1625 .bind_conflict
= inet6_csk_bind_conflict
,
1626 #ifdef CONFIG_COMPAT
1627 .compat_setsockopt
= compat_ipv6_setsockopt
,
1628 .compat_getsockopt
= compat_ipv6_getsockopt
,
1630 .mtu_reduced
= tcp_v4_mtu_reduced
,
1633 #ifdef CONFIG_TCP_MD5SIG
1634 static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific
= {
1635 .md5_lookup
= tcp_v4_md5_lookup
,
1636 .calc_md5_hash
= tcp_v4_md5_hash_skb
,
1637 .md5_parse
= tcp_v6_parse_md5_keys
,
1641 /* NOTE: A lot of things set to zero explicitly by call to
1642 * sk_alloc() so need not be done here.
1644 static int tcp_v6_init_sock(struct sock
*sk
)
1646 struct inet_connection_sock
*icsk
= inet_csk(sk
);
1650 icsk
->icsk_af_ops
= &ipv6_specific
;
1652 #ifdef CONFIG_TCP_MD5SIG
1653 tcp_sk(sk
)->af_specific
= &tcp_sock_ipv6_specific
;
1659 static void tcp_v6_destroy_sock(struct sock
*sk
)
1661 tcp_v4_destroy_sock(sk
);
1662 inet6_destroy_sock(sk
);
1665 #ifdef CONFIG_PROC_FS
1666 /* Proc filesystem TCPv6 sock list dumping. */
1667 static void get_openreq6(struct seq_file
*seq
,
1668 const struct sock
*sk
, struct request_sock
*req
, int i
, kuid_t uid
)
1670 int ttd
= req
->expires
- jiffies
;
1671 const struct in6_addr
*src
= &inet_rsk(req
)->ir_v6_loc_addr
;
1672 const struct in6_addr
*dest
= &inet_rsk(req
)->ir_v6_rmt_addr
;
1678 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1679 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
1681 src
->s6_addr32
[0], src
->s6_addr32
[1],
1682 src
->s6_addr32
[2], src
->s6_addr32
[3],
1683 inet_rsk(req
)->ir_num
,
1684 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1685 dest
->s6_addr32
[2], dest
->s6_addr32
[3],
1686 ntohs(inet_rsk(req
)->ir_rmt_port
),
1688 0, 0, /* could print option size, but that is af dependent. */
1689 1, /* timers active (only the expire timer) */
1690 jiffies_to_clock_t(ttd
),
1692 from_kuid_munged(seq_user_ns(seq
), uid
),
1693 0, /* non standard timer */
1694 0, /* open_requests have no inode */
1698 static void get_tcp6_sock(struct seq_file
*seq
, struct sock
*sp
, int i
)
1700 const struct in6_addr
*dest
, *src
;
1703 unsigned long timer_expires
;
1704 const struct inet_sock
*inet
= inet_sk(sp
);
1705 const struct tcp_sock
*tp
= tcp_sk(sp
);
1706 const struct inet_connection_sock
*icsk
= inet_csk(sp
);
1707 struct fastopen_queue
*fastopenq
= icsk
->icsk_accept_queue
.fastopenq
;
1709 dest
= &sp
->sk_v6_daddr
;
1710 src
= &sp
->sk_v6_rcv_saddr
;
1711 destp
= ntohs(inet
->inet_dport
);
1712 srcp
= ntohs(inet
->inet_sport
);
1714 if (icsk
->icsk_pending
== ICSK_TIME_RETRANS
) {
1716 timer_expires
= icsk
->icsk_timeout
;
1717 } else if (icsk
->icsk_pending
== ICSK_TIME_PROBE0
) {
1719 timer_expires
= icsk
->icsk_timeout
;
1720 } else if (timer_pending(&sp
->sk_timer
)) {
1722 timer_expires
= sp
->sk_timer
.expires
;
1725 timer_expires
= jiffies
;
1729 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1730 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
1732 src
->s6_addr32
[0], src
->s6_addr32
[1],
1733 src
->s6_addr32
[2], src
->s6_addr32
[3], srcp
,
1734 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1735 dest
->s6_addr32
[2], dest
->s6_addr32
[3], destp
,
1737 tp
->write_seq
-tp
->snd_una
,
1738 (sp
->sk_state
== TCP_LISTEN
) ? sp
->sk_ack_backlog
: (tp
->rcv_nxt
- tp
->copied_seq
),
1740 jiffies_delta_to_clock_t(timer_expires
- jiffies
),
1741 icsk
->icsk_retransmits
,
1742 from_kuid_munged(seq_user_ns(seq
), sock_i_uid(sp
)),
1743 icsk
->icsk_probes_out
,
1745 atomic_read(&sp
->sk_refcnt
), sp
,
1746 jiffies_to_clock_t(icsk
->icsk_rto
),
1747 jiffies_to_clock_t(icsk
->icsk_ack
.ato
),
1748 (icsk
->icsk_ack
.quick
<< 1) | icsk
->icsk_ack
.pingpong
,
1750 sp
->sk_state
== TCP_LISTEN
?
1751 (fastopenq
? fastopenq
->max_qlen
: 0) :
1752 (tcp_in_initial_slowstart(tp
) ? -1 : tp
->snd_ssthresh
)
1756 static void get_timewait6_sock(struct seq_file
*seq
,
1757 struct inet_timewait_sock
*tw
, int i
)
1759 const struct in6_addr
*dest
, *src
;
1761 s32 delta
= tw
->tw_ttd
- inet_tw_time_stamp();
1763 dest
= &tw
->tw_v6_daddr
;
1764 src
= &tw
->tw_v6_rcv_saddr
;
1765 destp
= ntohs(tw
->tw_dport
);
1766 srcp
= ntohs(tw
->tw_sport
);
1769 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1770 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
1772 src
->s6_addr32
[0], src
->s6_addr32
[1],
1773 src
->s6_addr32
[2], src
->s6_addr32
[3], srcp
,
1774 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1775 dest
->s6_addr32
[2], dest
->s6_addr32
[3], destp
,
1776 tw
->tw_substate
, 0, 0,
1777 3, jiffies_delta_to_clock_t(delta
), 0, 0, 0, 0,
1778 atomic_read(&tw
->tw_refcnt
), tw
);
1781 static int tcp6_seq_show(struct seq_file
*seq
, void *v
)
1783 struct tcp_iter_state
*st
;
1784 struct sock
*sk
= v
;
1786 if (v
== SEQ_START_TOKEN
) {
1791 "st tx_queue rx_queue tr tm->when retrnsmt"
1792 " uid timeout inode\n");
1797 switch (st
->state
) {
1798 case TCP_SEQ_STATE_LISTENING
:
1799 case TCP_SEQ_STATE_ESTABLISHED
:
1800 if (sk
->sk_state
== TCP_TIME_WAIT
)
1801 get_timewait6_sock(seq
, v
, st
->num
);
1803 get_tcp6_sock(seq
, v
, st
->num
);
1805 case TCP_SEQ_STATE_OPENREQ
:
1806 get_openreq6(seq
, st
->syn_wait_sk
, v
, st
->num
, st
->uid
);
1813 static const struct file_operations tcp6_afinfo_seq_fops
= {
1814 .owner
= THIS_MODULE
,
1815 .open
= tcp_seq_open
,
1817 .llseek
= seq_lseek
,
1818 .release
= seq_release_net
1821 static struct tcp_seq_afinfo tcp6_seq_afinfo
= {
1824 .seq_fops
= &tcp6_afinfo_seq_fops
,
1826 .show
= tcp6_seq_show
,
1830 int __net_init
tcp6_proc_init(struct net
*net
)
1832 return tcp_proc_register(net
, &tcp6_seq_afinfo
);
1835 void tcp6_proc_exit(struct net
*net
)
1837 tcp_proc_unregister(net
, &tcp6_seq_afinfo
);
1841 static void tcp_v6_clear_sk(struct sock
*sk
, int size
)
1843 struct inet_sock
*inet
= inet_sk(sk
);
1845 /* we do not want to clear pinet6 field, because of RCU lookups */
1846 sk_prot_clear_nulls(sk
, offsetof(struct inet_sock
, pinet6
));
1848 size
-= offsetof(struct inet_sock
, pinet6
) + sizeof(inet
->pinet6
);
1849 memset(&inet
->pinet6
+ 1, 0, size
);
1852 struct proto tcpv6_prot
= {
1854 .owner
= THIS_MODULE
,
1856 .connect
= tcp_v6_connect
,
1857 .disconnect
= tcp_disconnect
,
1858 .accept
= inet_csk_accept
,
1860 .init
= tcp_v6_init_sock
,
1861 .destroy
= tcp_v6_destroy_sock
,
1862 .shutdown
= tcp_shutdown
,
1863 .setsockopt
= tcp_setsockopt
,
1864 .getsockopt
= tcp_getsockopt
,
1865 .recvmsg
= tcp_recvmsg
,
1866 .sendmsg
= tcp_sendmsg
,
1867 .sendpage
= tcp_sendpage
,
1868 .backlog_rcv
= tcp_v6_do_rcv
,
1869 .release_cb
= tcp_release_cb
,
1870 .hash
= tcp_v6_hash
,
1871 .unhash
= inet_unhash
,
1872 .get_port
= inet_csk_get_port
,
1873 .enter_memory_pressure
= tcp_enter_memory_pressure
,
1874 .stream_memory_free
= tcp_stream_memory_free
,
1875 .sockets_allocated
= &tcp_sockets_allocated
,
1876 .memory_allocated
= &tcp_memory_allocated
,
1877 .memory_pressure
= &tcp_memory_pressure
,
1878 .orphan_count
= &tcp_orphan_count
,
1879 .sysctl_mem
= sysctl_tcp_mem
,
1880 .sysctl_wmem
= sysctl_tcp_wmem
,
1881 .sysctl_rmem
= sysctl_tcp_rmem
,
1882 .max_header
= MAX_TCP_HEADER
,
1883 .obj_size
= sizeof(struct tcp6_sock
),
1884 .slab_flags
= SLAB_DESTROY_BY_RCU
,
1885 .twsk_prot
= &tcp6_timewait_sock_ops
,
1886 .rsk_prot
= &tcp6_request_sock_ops
,
1887 .h
.hashinfo
= &tcp_hashinfo
,
1888 .no_autobind
= true,
1889 #ifdef CONFIG_COMPAT
1890 .compat_setsockopt
= compat_tcp_setsockopt
,
1891 .compat_getsockopt
= compat_tcp_getsockopt
,
1893 #ifdef CONFIG_MEMCG_KMEM
1894 .proto_cgroup
= tcp_proto_cgroup
,
1896 .clear_sk
= tcp_v6_clear_sk
,
1899 static const struct inet6_protocol tcpv6_protocol
= {
1900 .early_demux
= tcp_v6_early_demux
,
1901 .handler
= tcp_v6_rcv
,
1902 .err_handler
= tcp_v6_err
,
1903 .flags
= INET6_PROTO_NOPOLICY
|INET6_PROTO_FINAL
,
1906 static struct inet_protosw tcpv6_protosw
= {
1907 .type
= SOCK_STREAM
,
1908 .protocol
= IPPROTO_TCP
,
1909 .prot
= &tcpv6_prot
,
1910 .ops
= &inet6_stream_ops
,
1911 .flags
= INET_PROTOSW_PERMANENT
|
1915 static int __net_init
tcpv6_net_init(struct net
*net
)
1917 return inet_ctl_sock_create(&net
->ipv6
.tcp_sk
, PF_INET6
,
1918 SOCK_RAW
, IPPROTO_TCP
, net
);
1921 static void __net_exit
tcpv6_net_exit(struct net
*net
)
1923 inet_ctl_sock_destroy(net
->ipv6
.tcp_sk
);
1926 static void __net_exit
tcpv6_net_exit_batch(struct list_head
*net_exit_list
)
1928 inet_twsk_purge(&tcp_hashinfo
, &tcp_death_row
, AF_INET6
);
1931 static struct pernet_operations tcpv6_net_ops
= {
1932 .init
= tcpv6_net_init
,
1933 .exit
= tcpv6_net_exit
,
1934 .exit_batch
= tcpv6_net_exit_batch
,
1937 int __init
tcpv6_init(void)
1941 ret
= inet6_add_protocol(&tcpv6_protocol
, IPPROTO_TCP
);
1945 /* register inet6 protocol */
1946 ret
= inet6_register_protosw(&tcpv6_protosw
);
1948 goto out_tcpv6_protocol
;
1950 ret
= register_pernet_subsys(&tcpv6_net_ops
);
1952 goto out_tcpv6_protosw
;
1957 inet6_unregister_protosw(&tcpv6_protosw
);
1959 inet6_del_protocol(&tcpv6_protocol
, IPPROTO_TCP
);
1963 void tcpv6_exit(void)
1965 unregister_pernet_subsys(&tcpv6_net_ops
);
1966 inet6_unregister_protosw(&tcpv6_protosw
);
1967 inet6_del_protocol(&tcpv6_protocol
, IPPROTO_TCP
);