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
;
102 inet6_sk(sk
)->rx_dst_cookie
= rt6_get_cookie(rt
);
106 static __u32
tcp_v6_init_sequence(const struct sk_buff
*skb
)
108 return secure_tcpv6_sequence_number(ipv6_hdr(skb
)->daddr
.s6_addr32
,
109 ipv6_hdr(skb
)->saddr
.s6_addr32
,
111 tcp_hdr(skb
)->source
);
114 static int tcp_v6_connect(struct sock
*sk
, struct sockaddr
*uaddr
,
117 struct sockaddr_in6
*usin
= (struct sockaddr_in6
*) uaddr
;
118 struct inet_sock
*inet
= inet_sk(sk
);
119 struct inet_connection_sock
*icsk
= inet_csk(sk
);
120 struct ipv6_pinfo
*np
= inet6_sk(sk
);
121 struct tcp_sock
*tp
= tcp_sk(sk
);
122 struct in6_addr
*saddr
= NULL
, *final_p
, final
;
125 struct dst_entry
*dst
;
129 if (addr_len
< SIN6_LEN_RFC2133
)
132 if (usin
->sin6_family
!= AF_INET6
)
133 return -EAFNOSUPPORT
;
135 memset(&fl6
, 0, sizeof(fl6
));
138 fl6
.flowlabel
= usin
->sin6_flowinfo
&IPV6_FLOWINFO_MASK
;
139 IP6_ECN_flow_init(fl6
.flowlabel
);
140 if (fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) {
141 struct ip6_flowlabel
*flowlabel
;
142 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
145 fl6_sock_release(flowlabel
);
150 * connect() to INADDR_ANY means loopback (BSD'ism).
153 if (ipv6_addr_any(&usin
->sin6_addr
))
154 usin
->sin6_addr
.s6_addr
[15] = 0x1;
156 addr_type
= ipv6_addr_type(&usin
->sin6_addr
);
158 if (addr_type
& IPV6_ADDR_MULTICAST
)
161 if (addr_type
&IPV6_ADDR_LINKLOCAL
) {
162 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
163 usin
->sin6_scope_id
) {
164 /* If interface is set while binding, indices
167 if (sk
->sk_bound_dev_if
&&
168 sk
->sk_bound_dev_if
!= usin
->sin6_scope_id
)
171 sk
->sk_bound_dev_if
= usin
->sin6_scope_id
;
174 /* Connect to link-local address requires an interface */
175 if (!sk
->sk_bound_dev_if
)
179 if (tp
->rx_opt
.ts_recent_stamp
&&
180 !ipv6_addr_equal(&sk
->sk_v6_daddr
, &usin
->sin6_addr
)) {
181 tp
->rx_opt
.ts_recent
= 0;
182 tp
->rx_opt
.ts_recent_stamp
= 0;
186 sk
->sk_v6_daddr
= usin
->sin6_addr
;
187 np
->flow_label
= fl6
.flowlabel
;
193 if (addr_type
== IPV6_ADDR_MAPPED
) {
194 u32 exthdrlen
= icsk
->icsk_ext_hdr_len
;
195 struct sockaddr_in sin
;
197 SOCK_DEBUG(sk
, "connect: ipv4 mapped\n");
199 if (__ipv6_only_sock(sk
))
202 sin
.sin_family
= AF_INET
;
203 sin
.sin_port
= usin
->sin6_port
;
204 sin
.sin_addr
.s_addr
= usin
->sin6_addr
.s6_addr32
[3];
206 icsk
->icsk_af_ops
= &ipv6_mapped
;
207 sk
->sk_backlog_rcv
= tcp_v4_do_rcv
;
208 #ifdef CONFIG_TCP_MD5SIG
209 tp
->af_specific
= &tcp_sock_ipv6_mapped_specific
;
212 err
= tcp_v4_connect(sk
, (struct sockaddr
*)&sin
, sizeof(sin
));
215 icsk
->icsk_ext_hdr_len
= exthdrlen
;
216 icsk
->icsk_af_ops
= &ipv6_specific
;
217 sk
->sk_backlog_rcv
= tcp_v6_do_rcv
;
218 #ifdef CONFIG_TCP_MD5SIG
219 tp
->af_specific
= &tcp_sock_ipv6_specific
;
223 np
->saddr
= sk
->sk_v6_rcv_saddr
;
228 if (!ipv6_addr_any(&sk
->sk_v6_rcv_saddr
))
229 saddr
= &sk
->sk_v6_rcv_saddr
;
231 fl6
.flowi6_proto
= IPPROTO_TCP
;
232 fl6
.daddr
= sk
->sk_v6_daddr
;
233 fl6
.saddr
= saddr
? *saddr
: np
->saddr
;
234 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
235 fl6
.flowi6_mark
= sk
->sk_mark
;
236 fl6
.fl6_dport
= usin
->sin6_port
;
237 fl6
.fl6_sport
= inet
->inet_sport
;
239 final_p
= fl6_update_dst(&fl6
, np
->opt
, &final
);
241 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
243 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
);
251 sk
->sk_v6_rcv_saddr
= *saddr
;
254 /* set the source address */
256 inet
->inet_rcv_saddr
= LOOPBACK4_IPV6
;
258 sk
->sk_gso_type
= SKB_GSO_TCPV6
;
259 __ip6_dst_store(sk
, dst
, NULL
, NULL
);
261 rt
= (struct rt6_info
*) dst
;
262 if (tcp_death_row
.sysctl_tw_recycle
&&
263 !tp
->rx_opt
.ts_recent_stamp
&&
264 ipv6_addr_equal(&fl6
.daddr
, &sk
->sk_v6_daddr
))
265 tcp_fetch_timewait_stamp(sk
, dst
);
267 icsk
->icsk_ext_hdr_len
= 0;
269 icsk
->icsk_ext_hdr_len
= (np
->opt
->opt_flen
+
272 tp
->rx_opt
.mss_clamp
= IPV6_MIN_MTU
- sizeof(struct tcphdr
) - sizeof(struct ipv6hdr
);
274 inet
->inet_dport
= usin
->sin6_port
;
276 tcp_set_state(sk
, TCP_SYN_SENT
);
277 err
= inet6_hash_connect(&tcp_death_row
, sk
);
283 if (!tp
->write_seq
&& likely(!tp
->repair
))
284 tp
->write_seq
= secure_tcpv6_sequence_number(np
->saddr
.s6_addr32
,
285 sk
->sk_v6_daddr
.s6_addr32
,
289 err
= tcp_connect(sk
);
296 tcp_set_state(sk
, TCP_CLOSE
);
299 inet
->inet_dport
= 0;
300 sk
->sk_route_caps
= 0;
304 static void tcp_v6_mtu_reduced(struct sock
*sk
)
306 struct dst_entry
*dst
;
308 if ((1 << sk
->sk_state
) & (TCPF_LISTEN
| TCPF_CLOSE
))
311 dst
= inet6_csk_update_pmtu(sk
, tcp_sk(sk
)->mtu_info
);
315 if (inet_csk(sk
)->icsk_pmtu_cookie
> dst_mtu(dst
)) {
316 tcp_sync_mss(sk
, dst_mtu(dst
));
317 tcp_simple_retransmit(sk
);
321 static void tcp_v6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
322 u8 type
, u8 code
, int offset
, __be32 info
)
324 const struct ipv6hdr
*hdr
= (const struct ipv6hdr
*)skb
->data
;
325 const struct tcphdr
*th
= (struct tcphdr
*)(skb
->data
+offset
);
326 struct net
*net
= dev_net(skb
->dev
);
327 struct request_sock
*fastopen
;
328 struct ipv6_pinfo
*np
;
334 sk
= __inet6_lookup_established(net
, &tcp_hashinfo
,
335 &hdr
->daddr
, th
->dest
,
336 &hdr
->saddr
, ntohs(th
->source
),
340 ICMP6_INC_STATS_BH(net
, __in6_dev_get(skb
->dev
),
345 if (sk
->sk_state
== TCP_TIME_WAIT
) {
346 inet_twsk_put(inet_twsk(sk
));
349 seq
= ntohl(th
->seq
);
350 if (sk
->sk_state
== TCP_NEW_SYN_RECV
)
351 return tcp_req_err(sk
, seq
);
354 if (sock_owned_by_user(sk
) && type
!= ICMPV6_PKT_TOOBIG
)
355 NET_INC_STATS_BH(net
, LINUX_MIB_LOCKDROPPEDICMPS
);
357 if (sk
->sk_state
== TCP_CLOSE
)
360 if (ipv6_hdr(skb
)->hop_limit
< inet6_sk(sk
)->min_hopcount
) {
361 NET_INC_STATS_BH(net
, LINUX_MIB_TCPMINTTLDROP
);
366 /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
367 fastopen
= tp
->fastopen_rsk
;
368 snd_una
= fastopen
? tcp_rsk(fastopen
)->snt_isn
: tp
->snd_una
;
369 if (sk
->sk_state
!= TCP_LISTEN
&&
370 !between(seq
, snd_una
, tp
->snd_nxt
)) {
371 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
377 if (type
== NDISC_REDIRECT
) {
378 struct dst_entry
*dst
= __sk_dst_check(sk
, np
->dst_cookie
);
381 dst
->ops
->redirect(dst
, sk
, skb
);
385 if (type
== ICMPV6_PKT_TOOBIG
) {
386 /* We are not interested in TCP_LISTEN and open_requests
387 * (SYN-ACKs send out by Linux are always <576bytes so
388 * they should go through unfragmented).
390 if (sk
->sk_state
== TCP_LISTEN
)
393 if (!ip6_sk_accept_pmtu(sk
))
396 tp
->mtu_info
= ntohl(info
);
397 if (!sock_owned_by_user(sk
))
398 tcp_v6_mtu_reduced(sk
);
399 else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED
,
405 icmpv6_err_convert(type
, code
, &err
);
407 /* Might be for an request_sock */
408 switch (sk
->sk_state
) {
411 /* Only in fast or simultaneous open. If a fast open socket is
412 * is already accepted it is treated as a connected one below.
414 if (fastopen
&& !fastopen
->sk
)
417 if (!sock_owned_by_user(sk
)) {
419 sk
->sk_error_report(sk
); /* Wake people up to see the error (see connect in sock.c) */
423 sk
->sk_err_soft
= err
;
427 if (!sock_owned_by_user(sk
) && np
->recverr
) {
429 sk
->sk_error_report(sk
);
431 sk
->sk_err_soft
= err
;
439 static int tcp_v6_send_synack(struct sock
*sk
, struct dst_entry
*dst
,
441 struct request_sock
*req
,
443 struct tcp_fastopen_cookie
*foc
)
445 struct inet_request_sock
*ireq
= inet_rsk(req
);
446 struct ipv6_pinfo
*np
= inet6_sk(sk
);
447 struct flowi6
*fl6
= &fl
->u
.ip6
;
451 /* First, grab a route. */
452 if (!dst
&& (dst
= inet6_csk_route_req(sk
, fl6
, req
)) == NULL
)
455 skb
= tcp_make_synack(sk
, dst
, req
, foc
);
458 __tcp_v6_send_check(skb
, &ireq
->ir_v6_loc_addr
,
459 &ireq
->ir_v6_rmt_addr
);
461 fl6
->daddr
= ireq
->ir_v6_rmt_addr
;
462 if (np
->repflow
&& ireq
->pktopts
)
463 fl6
->flowlabel
= ip6_flowlabel(ipv6_hdr(ireq
->pktopts
));
465 skb_set_queue_mapping(skb
, queue_mapping
);
466 err
= ip6_xmit(sk
, skb
, fl6
, np
->opt
, np
->tclass
);
467 err
= net_xmit_eval(err
);
475 static void tcp_v6_reqsk_destructor(struct request_sock
*req
)
477 kfree_skb(inet_rsk(req
)->pktopts
);
480 #ifdef CONFIG_TCP_MD5SIG
481 static struct tcp_md5sig_key
*tcp_v6_md5_do_lookup(struct sock
*sk
,
482 const struct in6_addr
*addr
)
484 return tcp_md5_do_lookup(sk
, (union tcp_md5_addr
*)addr
, AF_INET6
);
487 static struct tcp_md5sig_key
*tcp_v6_md5_lookup(struct sock
*sk
,
488 const struct sock
*addr_sk
)
490 return tcp_v6_md5_do_lookup(sk
, &addr_sk
->sk_v6_daddr
);
493 static int tcp_v6_parse_md5_keys(struct sock
*sk
, char __user
*optval
,
496 struct tcp_md5sig cmd
;
497 struct sockaddr_in6
*sin6
= (struct sockaddr_in6
*)&cmd
.tcpm_addr
;
499 if (optlen
< sizeof(cmd
))
502 if (copy_from_user(&cmd
, optval
, sizeof(cmd
)))
505 if (sin6
->sin6_family
!= AF_INET6
)
508 if (!cmd
.tcpm_keylen
) {
509 if (ipv6_addr_v4mapped(&sin6
->sin6_addr
))
510 return tcp_md5_do_del(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
.s6_addr32
[3],
512 return tcp_md5_do_del(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
,
516 if (cmd
.tcpm_keylen
> TCP_MD5SIG_MAXKEYLEN
)
519 if (ipv6_addr_v4mapped(&sin6
->sin6_addr
))
520 return tcp_md5_do_add(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
.s6_addr32
[3],
521 AF_INET
, cmd
.tcpm_key
, cmd
.tcpm_keylen
, GFP_KERNEL
);
523 return tcp_md5_do_add(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
,
524 AF_INET6
, cmd
.tcpm_key
, cmd
.tcpm_keylen
, GFP_KERNEL
);
527 static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool
*hp
,
528 const struct in6_addr
*daddr
,
529 const struct in6_addr
*saddr
, int nbytes
)
531 struct tcp6_pseudohdr
*bp
;
532 struct scatterlist sg
;
534 bp
= &hp
->md5_blk
.ip6
;
535 /* 1. TCP pseudo-header (RFC2460) */
538 bp
->protocol
= cpu_to_be32(IPPROTO_TCP
);
539 bp
->len
= cpu_to_be32(nbytes
);
541 sg_init_one(&sg
, bp
, sizeof(*bp
));
542 return crypto_hash_update(&hp
->md5_desc
, &sg
, sizeof(*bp
));
545 static int tcp_v6_md5_hash_hdr(char *md5_hash
, struct tcp_md5sig_key
*key
,
546 const struct in6_addr
*daddr
, struct in6_addr
*saddr
,
547 const struct tcphdr
*th
)
549 struct tcp_md5sig_pool
*hp
;
550 struct hash_desc
*desc
;
552 hp
= tcp_get_md5sig_pool();
554 goto clear_hash_noput
;
555 desc
= &hp
->md5_desc
;
557 if (crypto_hash_init(desc
))
559 if (tcp_v6_md5_hash_pseudoheader(hp
, daddr
, saddr
, th
->doff
<< 2))
561 if (tcp_md5_hash_header(hp
, th
))
563 if (tcp_md5_hash_key(hp
, key
))
565 if (crypto_hash_final(desc
, md5_hash
))
568 tcp_put_md5sig_pool();
572 tcp_put_md5sig_pool();
574 memset(md5_hash
, 0, 16);
578 static int tcp_v6_md5_hash_skb(char *md5_hash
,
579 const struct tcp_md5sig_key
*key
,
580 const struct sock
*sk
,
581 const struct sk_buff
*skb
)
583 const struct in6_addr
*saddr
, *daddr
;
584 struct tcp_md5sig_pool
*hp
;
585 struct hash_desc
*desc
;
586 const struct tcphdr
*th
= tcp_hdr(skb
);
588 if (sk
) { /* valid for establish/request sockets */
589 saddr
= &sk
->sk_v6_rcv_saddr
;
590 daddr
= &sk
->sk_v6_daddr
;
592 const struct ipv6hdr
*ip6h
= ipv6_hdr(skb
);
593 saddr
= &ip6h
->saddr
;
594 daddr
= &ip6h
->daddr
;
597 hp
= tcp_get_md5sig_pool();
599 goto clear_hash_noput
;
600 desc
= &hp
->md5_desc
;
602 if (crypto_hash_init(desc
))
605 if (tcp_v6_md5_hash_pseudoheader(hp
, daddr
, saddr
, skb
->len
))
607 if (tcp_md5_hash_header(hp
, th
))
609 if (tcp_md5_hash_skb_data(hp
, skb
, th
->doff
<< 2))
611 if (tcp_md5_hash_key(hp
, key
))
613 if (crypto_hash_final(desc
, md5_hash
))
616 tcp_put_md5sig_pool();
620 tcp_put_md5sig_pool();
622 memset(md5_hash
, 0, 16);
626 static bool tcp_v6_inbound_md5_hash(struct sock
*sk
, const struct sk_buff
*skb
)
628 const __u8
*hash_location
= NULL
;
629 struct tcp_md5sig_key
*hash_expected
;
630 const struct ipv6hdr
*ip6h
= ipv6_hdr(skb
);
631 const struct tcphdr
*th
= tcp_hdr(skb
);
635 hash_expected
= tcp_v6_md5_do_lookup(sk
, &ip6h
->saddr
);
636 hash_location
= tcp_parse_md5sig_option(th
);
638 /* We've parsed the options - do we have a hash? */
639 if (!hash_expected
&& !hash_location
)
642 if (hash_expected
&& !hash_location
) {
643 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_TCPMD5NOTFOUND
);
647 if (!hash_expected
&& hash_location
) {
648 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_TCPMD5UNEXPECTED
);
652 /* check the signature */
653 genhash
= tcp_v6_md5_hash_skb(newhash
,
657 if (genhash
|| memcmp(hash_location
, newhash
, 16) != 0) {
658 net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n",
659 genhash
? "failed" : "mismatch",
660 &ip6h
->saddr
, ntohs(th
->source
),
661 &ip6h
->daddr
, ntohs(th
->dest
));
668 static void tcp_v6_init_req(struct request_sock
*req
, struct sock
*sk
,
671 struct inet_request_sock
*ireq
= inet_rsk(req
);
672 struct ipv6_pinfo
*np
= inet6_sk(sk
);
674 ireq
->ir_v6_rmt_addr
= ipv6_hdr(skb
)->saddr
;
675 ireq
->ir_v6_loc_addr
= ipv6_hdr(skb
)->daddr
;
677 /* So that link locals have meaning */
678 if (!sk
->sk_bound_dev_if
&&
679 ipv6_addr_type(&ireq
->ir_v6_rmt_addr
) & IPV6_ADDR_LINKLOCAL
)
680 ireq
->ir_iif
= tcp_v6_iif(skb
);
682 if (!TCP_SKB_CB(skb
)->tcp_tw_isn
&&
683 (ipv6_opt_accepted(sk
, skb
, &TCP_SKB_CB(skb
)->header
.h6
) ||
684 np
->rxopt
.bits
.rxinfo
||
685 np
->rxopt
.bits
.rxoinfo
|| np
->rxopt
.bits
.rxhlim
||
686 np
->rxopt
.bits
.rxohlim
|| np
->repflow
)) {
687 atomic_inc(&skb
->users
);
692 static struct dst_entry
*tcp_v6_route_req(struct sock
*sk
, struct flowi
*fl
,
693 const struct request_sock
*req
,
698 return inet6_csk_route_req(sk
, &fl
->u
.ip6
, req
);
701 struct request_sock_ops tcp6_request_sock_ops __read_mostly
= {
703 .obj_size
= sizeof(struct tcp6_request_sock
),
704 .rtx_syn_ack
= tcp_rtx_synack
,
705 .send_ack
= tcp_v6_reqsk_send_ack
,
706 .destructor
= tcp_v6_reqsk_destructor
,
707 .send_reset
= tcp_v6_send_reset
,
708 .syn_ack_timeout
= tcp_syn_ack_timeout
,
711 static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops
= {
712 .mss_clamp
= IPV6_MIN_MTU
- sizeof(struct tcphdr
) -
713 sizeof(struct ipv6hdr
),
714 #ifdef CONFIG_TCP_MD5SIG
715 .req_md5_lookup
= tcp_v6_md5_lookup
,
716 .calc_md5_hash
= tcp_v6_md5_hash_skb
,
718 .init_req
= tcp_v6_init_req
,
719 #ifdef CONFIG_SYN_COOKIES
720 .cookie_init_seq
= cookie_v6_init_sequence
,
722 .route_req
= tcp_v6_route_req
,
723 .init_seq
= tcp_v6_init_sequence
,
724 .send_synack
= tcp_v6_send_synack
,
725 .queue_hash_add
= inet6_csk_reqsk_queue_hash_add
,
728 static void tcp_v6_send_response(struct sock
*sk
, struct sk_buff
*skb
, u32 seq
,
729 u32 ack
, u32 win
, u32 tsval
, u32 tsecr
,
730 int oif
, struct tcp_md5sig_key
*key
, int rst
,
731 u8 tclass
, u32 label
)
733 const struct tcphdr
*th
= tcp_hdr(skb
);
735 struct sk_buff
*buff
;
737 struct net
*net
= sk
? sock_net(sk
) : dev_net(skb_dst(skb
)->dev
);
738 struct sock
*ctl_sk
= net
->ipv6
.tcp_sk
;
739 unsigned int tot_len
= sizeof(struct tcphdr
);
740 struct dst_entry
*dst
;
744 tot_len
+= TCPOLEN_TSTAMP_ALIGNED
;
745 #ifdef CONFIG_TCP_MD5SIG
747 tot_len
+= TCPOLEN_MD5SIG_ALIGNED
;
750 buff
= alloc_skb(MAX_HEADER
+ sizeof(struct ipv6hdr
) + tot_len
,
755 skb_reserve(buff
, MAX_HEADER
+ sizeof(struct ipv6hdr
) + tot_len
);
757 t1
= (struct tcphdr
*) skb_push(buff
, tot_len
);
758 skb_reset_transport_header(buff
);
760 /* Swap the send and the receive. */
761 memset(t1
, 0, sizeof(*t1
));
762 t1
->dest
= th
->source
;
763 t1
->source
= th
->dest
;
764 t1
->doff
= tot_len
/ 4;
765 t1
->seq
= htonl(seq
);
766 t1
->ack_seq
= htonl(ack
);
767 t1
->ack
= !rst
|| !th
->ack
;
769 t1
->window
= htons(win
);
771 topt
= (__be32
*)(t1
+ 1);
774 *topt
++ = htonl((TCPOPT_NOP
<< 24) | (TCPOPT_NOP
<< 16) |
775 (TCPOPT_TIMESTAMP
<< 8) | TCPOLEN_TIMESTAMP
);
776 *topt
++ = htonl(tsval
);
777 *topt
++ = htonl(tsecr
);
780 #ifdef CONFIG_TCP_MD5SIG
782 *topt
++ = htonl((TCPOPT_NOP
<< 24) | (TCPOPT_NOP
<< 16) |
783 (TCPOPT_MD5SIG
<< 8) | TCPOLEN_MD5SIG
);
784 tcp_v6_md5_hash_hdr((__u8
*)topt
, key
,
785 &ipv6_hdr(skb
)->saddr
,
786 &ipv6_hdr(skb
)->daddr
, t1
);
790 memset(&fl6
, 0, sizeof(fl6
));
791 fl6
.daddr
= ipv6_hdr(skb
)->saddr
;
792 fl6
.saddr
= ipv6_hdr(skb
)->daddr
;
793 fl6
.flowlabel
= label
;
795 buff
->ip_summed
= CHECKSUM_PARTIAL
;
798 __tcp_v6_send_check(buff
, &fl6
.saddr
, &fl6
.daddr
);
800 fl6
.flowi6_proto
= IPPROTO_TCP
;
801 if (rt6_need_strict(&fl6
.daddr
) && !oif
)
802 fl6
.flowi6_oif
= tcp_v6_iif(skb
);
804 fl6
.flowi6_oif
= oif
;
805 fl6
.flowi6_mark
= IP6_REPLY_MARK(net
, skb
->mark
);
806 fl6
.fl6_dport
= t1
->dest
;
807 fl6
.fl6_sport
= t1
->source
;
808 security_skb_classify_flow(skb
, flowi6_to_flowi(&fl6
));
810 /* Pass a socket to ip6_dst_lookup either it is for RST
811 * Underlying function will use this to retrieve the network
814 dst
= ip6_dst_lookup_flow(ctl_sk
, &fl6
, NULL
);
816 skb_dst_set(buff
, dst
);
817 ip6_xmit(ctl_sk
, buff
, &fl6
, NULL
, tclass
);
818 TCP_INC_STATS_BH(net
, TCP_MIB_OUTSEGS
);
820 TCP_INC_STATS_BH(net
, TCP_MIB_OUTRSTS
);
827 static void tcp_v6_send_reset(struct sock
*sk
, struct sk_buff
*skb
)
829 const struct tcphdr
*th
= tcp_hdr(skb
);
830 u32 seq
= 0, ack_seq
= 0;
831 struct tcp_md5sig_key
*key
= NULL
;
832 #ifdef CONFIG_TCP_MD5SIG
833 const __u8
*hash_location
= NULL
;
834 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
835 unsigned char newhash
[16];
837 struct sock
*sk1
= NULL
;
844 /* If sk not NULL, it means we did a successful lookup and incoming
845 * route had to be correct. prequeue might have dropped our dst.
847 if (!sk
&& !ipv6_unicast_destination(skb
))
850 #ifdef CONFIG_TCP_MD5SIG
851 hash_location
= tcp_parse_md5sig_option(th
);
852 if (!sk
&& hash_location
) {
854 * active side is lost. Try to find listening socket through
855 * source port, and then find md5 key through listening socket.
856 * we are not loose security here:
857 * Incoming packet is checked with md5 hash with finding key,
858 * no RST generated if md5 hash doesn't match.
860 sk1
= inet6_lookup_listener(dev_net(skb_dst(skb
)->dev
),
861 &tcp_hashinfo
, &ipv6h
->saddr
,
862 th
->source
, &ipv6h
->daddr
,
863 ntohs(th
->source
), tcp_v6_iif(skb
));
868 key
= tcp_v6_md5_do_lookup(sk1
, &ipv6h
->saddr
);
872 genhash
= tcp_v6_md5_hash_skb(newhash
, key
, NULL
, skb
);
873 if (genhash
|| memcmp(hash_location
, newhash
, 16) != 0)
876 key
= sk
? tcp_v6_md5_do_lookup(sk
, &ipv6h
->saddr
) : NULL
;
881 seq
= ntohl(th
->ack_seq
);
883 ack_seq
= ntohl(th
->seq
) + th
->syn
+ th
->fin
+ skb
->len
-
886 oif
= sk
? sk
->sk_bound_dev_if
: 0;
887 tcp_v6_send_response(sk
, skb
, seq
, ack_seq
, 0, 0, 0, oif
, key
, 1, 0, 0);
889 #ifdef CONFIG_TCP_MD5SIG
898 static void tcp_v6_send_ack(struct sock
*sk
, struct sk_buff
*skb
, u32 seq
,
899 u32 ack
, u32 win
, u32 tsval
, u32 tsecr
, int oif
,
900 struct tcp_md5sig_key
*key
, u8 tclass
,
903 tcp_v6_send_response(sk
, skb
, seq
, ack
, win
, tsval
, tsecr
, oif
, key
, 0,
907 static void tcp_v6_timewait_ack(struct sock
*sk
, struct sk_buff
*skb
)
909 struct inet_timewait_sock
*tw
= inet_twsk(sk
);
910 struct tcp_timewait_sock
*tcptw
= tcp_twsk(sk
);
912 tcp_v6_send_ack(sk
, skb
, tcptw
->tw_snd_nxt
, tcptw
->tw_rcv_nxt
,
913 tcptw
->tw_rcv_wnd
>> tw
->tw_rcv_wscale
,
914 tcp_time_stamp
+ tcptw
->tw_ts_offset
,
915 tcptw
->tw_ts_recent
, tw
->tw_bound_dev_if
, tcp_twsk_md5_key(tcptw
),
916 tw
->tw_tclass
, cpu_to_be32(tw
->tw_flowlabel
));
921 static void tcp_v6_reqsk_send_ack(struct sock
*sk
, struct sk_buff
*skb
,
922 struct request_sock
*req
)
924 /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
925 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
927 tcp_v6_send_ack(sk
, skb
, (sk
->sk_state
== TCP_LISTEN
) ?
928 tcp_rsk(req
)->snt_isn
+ 1 : tcp_sk(sk
)->snd_nxt
,
929 tcp_rsk(req
)->rcv_nxt
, req
->rcv_wnd
,
930 tcp_time_stamp
, req
->ts_recent
, sk
->sk_bound_dev_if
,
931 tcp_v6_md5_do_lookup(sk
, &ipv6_hdr(skb
)->daddr
),
936 static struct sock
*tcp_v6_hnd_req(struct sock
*sk
, struct sk_buff
*skb
)
938 const struct tcphdr
*th
= tcp_hdr(skb
);
939 struct request_sock
*req
;
942 /* Find possible connection requests. */
943 req
= inet6_csk_search_req(sk
, th
->source
,
944 &ipv6_hdr(skb
)->saddr
,
945 &ipv6_hdr(skb
)->daddr
, tcp_v6_iif(skb
));
947 nsk
= tcp_check_req(sk
, skb
, req
, false);
952 nsk
= __inet6_lookup_established(sock_net(sk
), &tcp_hashinfo
,
953 &ipv6_hdr(skb
)->saddr
, th
->source
,
954 &ipv6_hdr(skb
)->daddr
, ntohs(th
->dest
),
958 if (nsk
->sk_state
!= TCP_TIME_WAIT
) {
962 inet_twsk_put(inet_twsk(nsk
));
966 #ifdef CONFIG_SYN_COOKIES
968 sk
= cookie_v6_check(sk
, skb
);
973 static int tcp_v6_conn_request(struct sock
*sk
, struct sk_buff
*skb
)
975 if (skb
->protocol
== htons(ETH_P_IP
))
976 return tcp_v4_conn_request(sk
, skb
);
978 if (!ipv6_unicast_destination(skb
))
981 return tcp_conn_request(&tcp6_request_sock_ops
,
982 &tcp_request_sock_ipv6_ops
, sk
, skb
);
985 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
986 return 0; /* don't send reset */
989 static struct sock
*tcp_v6_syn_recv_sock(struct sock
*sk
, struct sk_buff
*skb
,
990 struct request_sock
*req
,
991 struct dst_entry
*dst
)
993 struct inet_request_sock
*ireq
;
994 struct ipv6_pinfo
*newnp
, *np
= inet6_sk(sk
);
995 struct tcp6_sock
*newtcp6sk
;
996 struct inet_sock
*newinet
;
997 struct tcp_sock
*newtp
;
999 #ifdef CONFIG_TCP_MD5SIG
1000 struct tcp_md5sig_key
*key
;
1004 if (skb
->protocol
== htons(ETH_P_IP
)) {
1009 newsk
= tcp_v4_syn_recv_sock(sk
, skb
, req
, dst
);
1014 newtcp6sk
= (struct tcp6_sock
*)newsk
;
1015 inet_sk(newsk
)->pinet6
= &newtcp6sk
->inet6
;
1017 newinet
= inet_sk(newsk
);
1018 newnp
= inet6_sk(newsk
);
1019 newtp
= tcp_sk(newsk
);
1021 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
1023 newnp
->saddr
= newsk
->sk_v6_rcv_saddr
;
1025 inet_csk(newsk
)->icsk_af_ops
= &ipv6_mapped
;
1026 newsk
->sk_backlog_rcv
= tcp_v4_do_rcv
;
1027 #ifdef CONFIG_TCP_MD5SIG
1028 newtp
->af_specific
= &tcp_sock_ipv6_mapped_specific
;
1031 newnp
->ipv6_ac_list
= NULL
;
1032 newnp
->ipv6_fl_list
= NULL
;
1033 newnp
->pktoptions
= NULL
;
1035 newnp
->mcast_oif
= tcp_v6_iif(skb
);
1036 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
1037 newnp
->rcv_flowinfo
= ip6_flowinfo(ipv6_hdr(skb
));
1039 newnp
->flow_label
= ip6_flowlabel(ipv6_hdr(skb
));
1042 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1043 * here, tcp_create_openreq_child now does this for us, see the comment in
1044 * that function for the gory details. -acme
1047 /* It is tricky place. Until this moment IPv4 tcp
1048 worked with IPv6 icsk.icsk_af_ops.
1051 tcp_sync_mss(newsk
, inet_csk(newsk
)->icsk_pmtu_cookie
);
1056 ireq
= inet_rsk(req
);
1058 if (sk_acceptq_is_full(sk
))
1062 dst
= inet6_csk_route_req(sk
, &fl6
, req
);
1067 newsk
= tcp_create_openreq_child(sk
, req
, skb
);
1072 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1073 * count here, tcp_create_openreq_child now does this for us, see the
1074 * comment in that function for the gory details. -acme
1077 newsk
->sk_gso_type
= SKB_GSO_TCPV6
;
1078 __ip6_dst_store(newsk
, dst
, NULL
, NULL
);
1079 inet6_sk_rx_dst_set(newsk
, skb
);
1081 newtcp6sk
= (struct tcp6_sock
*)newsk
;
1082 inet_sk(newsk
)->pinet6
= &newtcp6sk
->inet6
;
1084 newtp
= tcp_sk(newsk
);
1085 newinet
= inet_sk(newsk
);
1086 newnp
= inet6_sk(newsk
);
1088 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
1090 newsk
->sk_v6_daddr
= ireq
->ir_v6_rmt_addr
;
1091 newnp
->saddr
= ireq
->ir_v6_loc_addr
;
1092 newsk
->sk_v6_rcv_saddr
= ireq
->ir_v6_loc_addr
;
1093 newsk
->sk_bound_dev_if
= ireq
->ir_iif
;
1095 ip6_set_txhash(newsk
);
1097 /* Now IPv6 options...
1099 First: no IPv4 options.
1101 newinet
->inet_opt
= NULL
;
1102 newnp
->ipv6_ac_list
= NULL
;
1103 newnp
->ipv6_fl_list
= NULL
;
1106 newnp
->rxopt
.all
= np
->rxopt
.all
;
1108 /* Clone pktoptions received with SYN */
1109 newnp
->pktoptions
= NULL
;
1110 if (ireq
->pktopts
) {
1111 newnp
->pktoptions
= skb_clone(ireq
->pktopts
,
1112 sk_gfp_atomic(sk
, GFP_ATOMIC
));
1113 consume_skb(ireq
->pktopts
);
1114 ireq
->pktopts
= NULL
;
1115 if (newnp
->pktoptions
)
1116 skb_set_owner_r(newnp
->pktoptions
, newsk
);
1119 newnp
->mcast_oif
= tcp_v6_iif(skb
);
1120 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
1121 newnp
->rcv_flowinfo
= ip6_flowinfo(ipv6_hdr(skb
));
1123 newnp
->flow_label
= ip6_flowlabel(ipv6_hdr(skb
));
1125 /* Clone native IPv6 options from listening socket (if any)
1127 Yes, keeping reference count would be much more clever,
1128 but we make one more one thing there: reattach optmem
1132 newnp
->opt
= ipv6_dup_options(newsk
, np
->opt
);
1134 inet_csk(newsk
)->icsk_ext_hdr_len
= 0;
1136 inet_csk(newsk
)->icsk_ext_hdr_len
= (newnp
->opt
->opt_nflen
+
1137 newnp
->opt
->opt_flen
);
1139 tcp_ca_openreq_child(newsk
, dst
);
1141 tcp_sync_mss(newsk
, dst_mtu(dst
));
1142 newtp
->advmss
= dst_metric_advmss(dst
);
1143 if (tcp_sk(sk
)->rx_opt
.user_mss
&&
1144 tcp_sk(sk
)->rx_opt
.user_mss
< newtp
->advmss
)
1145 newtp
->advmss
= tcp_sk(sk
)->rx_opt
.user_mss
;
1147 tcp_initialize_rcv_mss(newsk
);
1149 newinet
->inet_daddr
= newinet
->inet_saddr
= LOOPBACK4_IPV6
;
1150 newinet
->inet_rcv_saddr
= LOOPBACK4_IPV6
;
1152 #ifdef CONFIG_TCP_MD5SIG
1153 /* Copy over the MD5 key from the original socket */
1154 key
= tcp_v6_md5_do_lookup(sk
, &newsk
->sk_v6_daddr
);
1156 /* We're using one, so create a matching key
1157 * on the newsk structure. If we fail to get
1158 * memory, then we end up not copying the key
1161 tcp_md5_do_add(newsk
, (union tcp_md5_addr
*)&newsk
->sk_v6_daddr
,
1162 AF_INET6
, key
->key
, key
->keylen
,
1163 sk_gfp_atomic(sk
, GFP_ATOMIC
));
1167 if (__inet_inherit_port(sk
, newsk
) < 0) {
1168 inet_csk_prepare_forced_close(newsk
);
1172 __inet_hash(newsk
, NULL
);
1177 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENOVERFLOWS
);
1181 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
1185 /* The socket must have it's spinlock held when we get
1188 * We have a potential double-lock case here, so even when
1189 * doing backlog processing we use the BH locking scheme.
1190 * This is because we cannot sleep with the original spinlock
1193 static int tcp_v6_do_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1195 struct ipv6_pinfo
*np
= inet6_sk(sk
);
1196 struct tcp_sock
*tp
;
1197 struct sk_buff
*opt_skb
= NULL
;
1199 /* Imagine: socket is IPv6. IPv4 packet arrives,
1200 goes to IPv4 receive handler and backlogged.
1201 From backlog it always goes here. Kerboom...
1202 Fortunately, tcp_rcv_established and rcv_established
1203 handle them correctly, but it is not case with
1204 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1207 if (skb
->protocol
== htons(ETH_P_IP
))
1208 return tcp_v4_do_rcv(sk
, skb
);
1210 if (sk_filter(sk
, skb
))
1214 * socket locking is here for SMP purposes as backlog rcv
1215 * is currently called with bh processing disabled.
1218 /* Do Stevens' IPV6_PKTOPTIONS.
1220 Yes, guys, it is the only place in our code, where we
1221 may make it not affecting IPv4.
1222 The rest of code is protocol independent,
1223 and I do not like idea to uglify IPv4.
1225 Actually, all the idea behind IPV6_PKTOPTIONS
1226 looks not very well thought. For now we latch
1227 options, received in the last packet, enqueued
1228 by tcp. Feel free to propose better solution.
1232 opt_skb
= skb_clone(skb
, sk_gfp_atomic(sk
, GFP_ATOMIC
));
1234 if (sk
->sk_state
== TCP_ESTABLISHED
) { /* Fast path */
1235 struct dst_entry
*dst
= sk
->sk_rx_dst
;
1237 sock_rps_save_rxhash(sk
, skb
);
1238 sk_mark_napi_id(sk
, skb
);
1240 if (inet_sk(sk
)->rx_dst_ifindex
!= skb
->skb_iif
||
1241 dst
->ops
->check(dst
, np
->rx_dst_cookie
) == NULL
) {
1243 sk
->sk_rx_dst
= NULL
;
1247 tcp_rcv_established(sk
, skb
, tcp_hdr(skb
), skb
->len
);
1249 goto ipv6_pktoptions
;
1253 if (tcp_checksum_complete(skb
))
1256 if (sk
->sk_state
== TCP_LISTEN
) {
1257 struct sock
*nsk
= tcp_v6_hnd_req(sk
, skb
);
1262 * Queue it on the new socket if the new socket is active,
1263 * otherwise we just shortcircuit this and continue with
1267 sock_rps_save_rxhash(nsk
, skb
);
1268 sk_mark_napi_id(sk
, skb
);
1269 if (tcp_child_process(sk
, nsk
, skb
))
1272 __kfree_skb(opt_skb
);
1276 sock_rps_save_rxhash(sk
, skb
);
1278 if (tcp_rcv_state_process(sk
, skb
, tcp_hdr(skb
), skb
->len
))
1281 goto ipv6_pktoptions
;
1285 tcp_v6_send_reset(sk
, skb
);
1288 __kfree_skb(opt_skb
);
1292 TCP_INC_STATS_BH(sock_net(sk
), TCP_MIB_CSUMERRORS
);
1293 TCP_INC_STATS_BH(sock_net(sk
), TCP_MIB_INERRS
);
1298 /* Do you ask, what is it?
1300 1. skb was enqueued by tcp.
1301 2. skb is added to tail of read queue, rather than out of order.
1302 3. socket is not in passive state.
1303 4. Finally, it really contains options, which user wants to receive.
1306 if (TCP_SKB_CB(opt_skb
)->end_seq
== tp
->rcv_nxt
&&
1307 !((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_LISTEN
))) {
1308 if (np
->rxopt
.bits
.rxinfo
|| np
->rxopt
.bits
.rxoinfo
)
1309 np
->mcast_oif
= tcp_v6_iif(opt_skb
);
1310 if (np
->rxopt
.bits
.rxhlim
|| np
->rxopt
.bits
.rxohlim
)
1311 np
->mcast_hops
= ipv6_hdr(opt_skb
)->hop_limit
;
1312 if (np
->rxopt
.bits
.rxflow
|| np
->rxopt
.bits
.rxtclass
)
1313 np
->rcv_flowinfo
= ip6_flowinfo(ipv6_hdr(opt_skb
));
1315 np
->flow_label
= ip6_flowlabel(ipv6_hdr(opt_skb
));
1316 if (ipv6_opt_accepted(sk
, opt_skb
, &TCP_SKB_CB(opt_skb
)->header
.h6
)) {
1317 skb_set_owner_r(opt_skb
, sk
);
1318 opt_skb
= xchg(&np
->pktoptions
, opt_skb
);
1320 __kfree_skb(opt_skb
);
1321 opt_skb
= xchg(&np
->pktoptions
, NULL
);
1329 static void tcp_v6_fill_cb(struct sk_buff
*skb
, const struct ipv6hdr
*hdr
,
1330 const struct tcphdr
*th
)
1332 /* This is tricky: we move IP6CB at its correct location into
1333 * TCP_SKB_CB(). It must be done after xfrm6_policy_check(), because
1334 * _decode_session6() uses IP6CB().
1335 * barrier() makes sure compiler won't play aliasing games.
1337 memmove(&TCP_SKB_CB(skb
)->header
.h6
, IP6CB(skb
),
1338 sizeof(struct inet6_skb_parm
));
1341 TCP_SKB_CB(skb
)->seq
= ntohl(th
->seq
);
1342 TCP_SKB_CB(skb
)->end_seq
= (TCP_SKB_CB(skb
)->seq
+ th
->syn
+ th
->fin
+
1343 skb
->len
- th
->doff
*4);
1344 TCP_SKB_CB(skb
)->ack_seq
= ntohl(th
->ack_seq
);
1345 TCP_SKB_CB(skb
)->tcp_flags
= tcp_flag_byte(th
);
1346 TCP_SKB_CB(skb
)->tcp_tw_isn
= 0;
1347 TCP_SKB_CB(skb
)->ip_dsfield
= ipv6_get_dsfield(hdr
);
1348 TCP_SKB_CB(skb
)->sacked
= 0;
1351 static void tcp_v6_restore_cb(struct sk_buff
*skb
)
1353 /* We need to move header back to the beginning if xfrm6_policy_check()
1354 * and tcp_v6_fill_cb() are going to be called again.
1356 memmove(IP6CB(skb
), &TCP_SKB_CB(skb
)->header
.h6
,
1357 sizeof(struct inet6_skb_parm
));
1360 static int tcp_v6_rcv(struct sk_buff
*skb
)
1362 const struct tcphdr
*th
;
1363 const struct ipv6hdr
*hdr
;
1366 struct net
*net
= dev_net(skb
->dev
);
1368 if (skb
->pkt_type
!= PACKET_HOST
)
1372 * Count it even if it's bad.
1374 TCP_INC_STATS_BH(net
, TCP_MIB_INSEGS
);
1376 if (!pskb_may_pull(skb
, sizeof(struct tcphdr
)))
1381 if (th
->doff
< sizeof(struct tcphdr
)/4)
1383 if (!pskb_may_pull(skb
, th
->doff
*4))
1386 if (skb_checksum_init(skb
, IPPROTO_TCP
, ip6_compute_pseudo
))
1390 hdr
= ipv6_hdr(skb
);
1392 sk
= __inet6_lookup_skb(&tcp_hashinfo
, skb
, th
->source
, th
->dest
,
1398 if (sk
->sk_state
== TCP_TIME_WAIT
)
1401 if (hdr
->hop_limit
< inet6_sk(sk
)->min_hopcount
) {
1402 NET_INC_STATS_BH(net
, LINUX_MIB_TCPMINTTLDROP
);
1403 goto discard_and_relse
;
1406 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
))
1407 goto discard_and_relse
;
1409 tcp_v6_fill_cb(skb
, hdr
, th
);
1411 #ifdef CONFIG_TCP_MD5SIG
1412 if (tcp_v6_inbound_md5_hash(sk
, skb
))
1413 goto discard_and_relse
;
1416 if (sk_filter(sk
, skb
))
1417 goto discard_and_relse
;
1419 sk_incoming_cpu_update(sk
);
1422 bh_lock_sock_nested(sk
);
1423 tcp_sk(sk
)->segs_in
+= max_t(u16
, 1, skb_shinfo(skb
)->gso_segs
);
1425 if (!sock_owned_by_user(sk
)) {
1426 if (!tcp_prequeue(sk
, skb
))
1427 ret
= tcp_v6_do_rcv(sk
, skb
);
1428 } else if (unlikely(sk_add_backlog(sk
, skb
,
1429 sk
->sk_rcvbuf
+ sk
->sk_sndbuf
))) {
1431 NET_INC_STATS_BH(net
, LINUX_MIB_TCPBACKLOGDROP
);
1432 goto discard_and_relse
;
1437 return ret
? -1 : 0;
1440 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
1443 tcp_v6_fill_cb(skb
, hdr
, th
);
1445 if (tcp_checksum_complete(skb
)) {
1447 TCP_INC_STATS_BH(net
, TCP_MIB_CSUMERRORS
);
1449 TCP_INC_STATS_BH(net
, TCP_MIB_INERRS
);
1451 tcp_v6_send_reset(NULL
, skb
);
1463 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
)) {
1464 inet_twsk_put(inet_twsk(sk
));
1468 tcp_v6_fill_cb(skb
, hdr
, th
);
1470 if (skb
->len
< (th
->doff
<<2)) {
1471 inet_twsk_put(inet_twsk(sk
));
1474 if (tcp_checksum_complete(skb
)) {
1475 inet_twsk_put(inet_twsk(sk
));
1479 switch (tcp_timewait_state_process(inet_twsk(sk
), skb
, th
)) {
1484 sk2
= inet6_lookup_listener(dev_net(skb
->dev
), &tcp_hashinfo
,
1485 &ipv6_hdr(skb
)->saddr
, th
->source
,
1486 &ipv6_hdr(skb
)->daddr
,
1487 ntohs(th
->dest
), tcp_v6_iif(skb
));
1489 struct inet_timewait_sock
*tw
= inet_twsk(sk
);
1490 inet_twsk_deschedule(tw
);
1493 tcp_v6_restore_cb(skb
);
1496 /* Fall through to ACK */
1499 tcp_v6_timewait_ack(sk
, skb
);
1502 tcp_v6_restore_cb(skb
);
1504 case TCP_TW_SUCCESS
:
1510 static void tcp_v6_early_demux(struct sk_buff
*skb
)
1512 const struct ipv6hdr
*hdr
;
1513 const struct tcphdr
*th
;
1516 if (skb
->pkt_type
!= PACKET_HOST
)
1519 if (!pskb_may_pull(skb
, skb_transport_offset(skb
) + sizeof(struct tcphdr
)))
1522 hdr
= ipv6_hdr(skb
);
1525 if (th
->doff
< sizeof(struct tcphdr
) / 4)
1528 /* Note : We use inet6_iif() here, not tcp_v6_iif() */
1529 sk
= __inet6_lookup_established(dev_net(skb
->dev
), &tcp_hashinfo
,
1530 &hdr
->saddr
, th
->source
,
1531 &hdr
->daddr
, ntohs(th
->dest
),
1535 skb
->destructor
= sock_edemux
;
1536 if (sk_fullsock(sk
)) {
1537 struct dst_entry
*dst
= READ_ONCE(sk
->sk_rx_dst
);
1540 dst
= dst_check(dst
, inet6_sk(sk
)->rx_dst_cookie
);
1542 inet_sk(sk
)->rx_dst_ifindex
== skb
->skb_iif
)
1543 skb_dst_set_noref(skb
, dst
);
1548 static struct timewait_sock_ops tcp6_timewait_sock_ops
= {
1549 .twsk_obj_size
= sizeof(struct tcp6_timewait_sock
),
1550 .twsk_unique
= tcp_twsk_unique
,
1551 .twsk_destructor
= tcp_twsk_destructor
,
1554 static const struct inet_connection_sock_af_ops ipv6_specific
= {
1555 .queue_xmit
= inet6_csk_xmit
,
1556 .send_check
= tcp_v6_send_check
,
1557 .rebuild_header
= inet6_sk_rebuild_header
,
1558 .sk_rx_dst_set
= inet6_sk_rx_dst_set
,
1559 .conn_request
= tcp_v6_conn_request
,
1560 .syn_recv_sock
= tcp_v6_syn_recv_sock
,
1561 .net_header_len
= sizeof(struct ipv6hdr
),
1562 .net_frag_header_len
= sizeof(struct frag_hdr
),
1563 .setsockopt
= ipv6_setsockopt
,
1564 .getsockopt
= ipv6_getsockopt
,
1565 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1566 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1567 .bind_conflict
= inet6_csk_bind_conflict
,
1568 #ifdef CONFIG_COMPAT
1569 .compat_setsockopt
= compat_ipv6_setsockopt
,
1570 .compat_getsockopt
= compat_ipv6_getsockopt
,
1572 .mtu_reduced
= tcp_v6_mtu_reduced
,
1575 #ifdef CONFIG_TCP_MD5SIG
1576 static const struct tcp_sock_af_ops tcp_sock_ipv6_specific
= {
1577 .md5_lookup
= tcp_v6_md5_lookup
,
1578 .calc_md5_hash
= tcp_v6_md5_hash_skb
,
1579 .md5_parse
= tcp_v6_parse_md5_keys
,
1584 * TCP over IPv4 via INET6 API
1586 static const struct inet_connection_sock_af_ops ipv6_mapped
= {
1587 .queue_xmit
= ip_queue_xmit
,
1588 .send_check
= tcp_v4_send_check
,
1589 .rebuild_header
= inet_sk_rebuild_header
,
1590 .sk_rx_dst_set
= inet_sk_rx_dst_set
,
1591 .conn_request
= tcp_v6_conn_request
,
1592 .syn_recv_sock
= tcp_v6_syn_recv_sock
,
1593 .net_header_len
= sizeof(struct iphdr
),
1594 .setsockopt
= ipv6_setsockopt
,
1595 .getsockopt
= ipv6_getsockopt
,
1596 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1597 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1598 .bind_conflict
= inet6_csk_bind_conflict
,
1599 #ifdef CONFIG_COMPAT
1600 .compat_setsockopt
= compat_ipv6_setsockopt
,
1601 .compat_getsockopt
= compat_ipv6_getsockopt
,
1603 .mtu_reduced
= tcp_v4_mtu_reduced
,
1606 #ifdef CONFIG_TCP_MD5SIG
1607 static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific
= {
1608 .md5_lookup
= tcp_v4_md5_lookup
,
1609 .calc_md5_hash
= tcp_v4_md5_hash_skb
,
1610 .md5_parse
= tcp_v6_parse_md5_keys
,
1614 /* NOTE: A lot of things set to zero explicitly by call to
1615 * sk_alloc() so need not be done here.
1617 static int tcp_v6_init_sock(struct sock
*sk
)
1619 struct inet_connection_sock
*icsk
= inet_csk(sk
);
1623 icsk
->icsk_af_ops
= &ipv6_specific
;
1625 #ifdef CONFIG_TCP_MD5SIG
1626 tcp_sk(sk
)->af_specific
= &tcp_sock_ipv6_specific
;
1632 static void tcp_v6_destroy_sock(struct sock
*sk
)
1634 tcp_v4_destroy_sock(sk
);
1635 inet6_destroy_sock(sk
);
1638 #ifdef CONFIG_PROC_FS
1639 /* Proc filesystem TCPv6 sock list dumping. */
1640 static void get_openreq6(struct seq_file
*seq
,
1641 struct request_sock
*req
, int i
, kuid_t uid
)
1643 long ttd
= req
->rsk_timer
.expires
- jiffies
;
1644 const struct in6_addr
*src
= &inet_rsk(req
)->ir_v6_loc_addr
;
1645 const struct in6_addr
*dest
= &inet_rsk(req
)->ir_v6_rmt_addr
;
1651 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1652 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
1654 src
->s6_addr32
[0], src
->s6_addr32
[1],
1655 src
->s6_addr32
[2], src
->s6_addr32
[3],
1656 inet_rsk(req
)->ir_num
,
1657 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1658 dest
->s6_addr32
[2], dest
->s6_addr32
[3],
1659 ntohs(inet_rsk(req
)->ir_rmt_port
),
1661 0, 0, /* could print option size, but that is af dependent. */
1662 1, /* timers active (only the expire timer) */
1663 jiffies_to_clock_t(ttd
),
1665 from_kuid_munged(seq_user_ns(seq
), uid
),
1666 0, /* non standard timer */
1667 0, /* open_requests have no inode */
1671 static void get_tcp6_sock(struct seq_file
*seq
, struct sock
*sp
, int i
)
1673 const struct in6_addr
*dest
, *src
;
1676 unsigned long timer_expires
;
1677 const struct inet_sock
*inet
= inet_sk(sp
);
1678 const struct tcp_sock
*tp
= tcp_sk(sp
);
1679 const struct inet_connection_sock
*icsk
= inet_csk(sp
);
1680 struct fastopen_queue
*fastopenq
= icsk
->icsk_accept_queue
.fastopenq
;
1682 dest
= &sp
->sk_v6_daddr
;
1683 src
= &sp
->sk_v6_rcv_saddr
;
1684 destp
= ntohs(inet
->inet_dport
);
1685 srcp
= ntohs(inet
->inet_sport
);
1687 if (icsk
->icsk_pending
== ICSK_TIME_RETRANS
) {
1689 timer_expires
= icsk
->icsk_timeout
;
1690 } else if (icsk
->icsk_pending
== ICSK_TIME_PROBE0
) {
1692 timer_expires
= icsk
->icsk_timeout
;
1693 } else if (timer_pending(&sp
->sk_timer
)) {
1695 timer_expires
= sp
->sk_timer
.expires
;
1698 timer_expires
= jiffies
;
1702 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1703 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
1705 src
->s6_addr32
[0], src
->s6_addr32
[1],
1706 src
->s6_addr32
[2], src
->s6_addr32
[3], srcp
,
1707 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1708 dest
->s6_addr32
[2], dest
->s6_addr32
[3], destp
,
1710 tp
->write_seq
-tp
->snd_una
,
1711 (sp
->sk_state
== TCP_LISTEN
) ? sp
->sk_ack_backlog
: (tp
->rcv_nxt
- tp
->copied_seq
),
1713 jiffies_delta_to_clock_t(timer_expires
- jiffies
),
1714 icsk
->icsk_retransmits
,
1715 from_kuid_munged(seq_user_ns(seq
), sock_i_uid(sp
)),
1716 icsk
->icsk_probes_out
,
1718 atomic_read(&sp
->sk_refcnt
), sp
,
1719 jiffies_to_clock_t(icsk
->icsk_rto
),
1720 jiffies_to_clock_t(icsk
->icsk_ack
.ato
),
1721 (icsk
->icsk_ack
.quick
<< 1) | icsk
->icsk_ack
.pingpong
,
1723 sp
->sk_state
== TCP_LISTEN
?
1724 (fastopenq
? fastopenq
->max_qlen
: 0) :
1725 (tcp_in_initial_slowstart(tp
) ? -1 : tp
->snd_ssthresh
)
1729 static void get_timewait6_sock(struct seq_file
*seq
,
1730 struct inet_timewait_sock
*tw
, int i
)
1732 long delta
= tw
->tw_timer
.expires
- jiffies
;
1733 const struct in6_addr
*dest
, *src
;
1736 dest
= &tw
->tw_v6_daddr
;
1737 src
= &tw
->tw_v6_rcv_saddr
;
1738 destp
= ntohs(tw
->tw_dport
);
1739 srcp
= ntohs(tw
->tw_sport
);
1742 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1743 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
1745 src
->s6_addr32
[0], src
->s6_addr32
[1],
1746 src
->s6_addr32
[2], src
->s6_addr32
[3], srcp
,
1747 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1748 dest
->s6_addr32
[2], dest
->s6_addr32
[3], destp
,
1749 tw
->tw_substate
, 0, 0,
1750 3, jiffies_delta_to_clock_t(delta
), 0, 0, 0, 0,
1751 atomic_read(&tw
->tw_refcnt
), tw
);
1754 static int tcp6_seq_show(struct seq_file
*seq
, void *v
)
1756 struct tcp_iter_state
*st
;
1757 struct sock
*sk
= v
;
1759 if (v
== SEQ_START_TOKEN
) {
1764 "st tx_queue rx_queue tr tm->when retrnsmt"
1765 " uid timeout inode\n");
1770 switch (st
->state
) {
1771 case TCP_SEQ_STATE_LISTENING
:
1772 case TCP_SEQ_STATE_ESTABLISHED
:
1773 if (sk
->sk_state
== TCP_TIME_WAIT
)
1774 get_timewait6_sock(seq
, v
, st
->num
);
1776 get_tcp6_sock(seq
, v
, st
->num
);
1778 case TCP_SEQ_STATE_OPENREQ
:
1779 get_openreq6(seq
, v
, st
->num
, st
->uid
);
1786 static const struct file_operations tcp6_afinfo_seq_fops
= {
1787 .owner
= THIS_MODULE
,
1788 .open
= tcp_seq_open
,
1790 .llseek
= seq_lseek
,
1791 .release
= seq_release_net
1794 static struct tcp_seq_afinfo tcp6_seq_afinfo
= {
1797 .seq_fops
= &tcp6_afinfo_seq_fops
,
1799 .show
= tcp6_seq_show
,
1803 int __net_init
tcp6_proc_init(struct net
*net
)
1805 return tcp_proc_register(net
, &tcp6_seq_afinfo
);
1808 void tcp6_proc_exit(struct net
*net
)
1810 tcp_proc_unregister(net
, &tcp6_seq_afinfo
);
1814 static void tcp_v6_clear_sk(struct sock
*sk
, int size
)
1816 struct inet_sock
*inet
= inet_sk(sk
);
1818 /* we do not want to clear pinet6 field, because of RCU lookups */
1819 sk_prot_clear_nulls(sk
, offsetof(struct inet_sock
, pinet6
));
1821 size
-= offsetof(struct inet_sock
, pinet6
) + sizeof(inet
->pinet6
);
1822 memset(&inet
->pinet6
+ 1, 0, size
);
1825 struct proto tcpv6_prot
= {
1827 .owner
= THIS_MODULE
,
1829 .connect
= tcp_v6_connect
,
1830 .disconnect
= tcp_disconnect
,
1831 .accept
= inet_csk_accept
,
1833 .init
= tcp_v6_init_sock
,
1834 .destroy
= tcp_v6_destroy_sock
,
1835 .shutdown
= tcp_shutdown
,
1836 .setsockopt
= tcp_setsockopt
,
1837 .getsockopt
= tcp_getsockopt
,
1838 .recvmsg
= tcp_recvmsg
,
1839 .sendmsg
= tcp_sendmsg
,
1840 .sendpage
= tcp_sendpage
,
1841 .backlog_rcv
= tcp_v6_do_rcv
,
1842 .release_cb
= tcp_release_cb
,
1844 .unhash
= inet_unhash
,
1845 .get_port
= inet_csk_get_port
,
1846 .enter_memory_pressure
= tcp_enter_memory_pressure
,
1847 .stream_memory_free
= tcp_stream_memory_free
,
1848 .sockets_allocated
= &tcp_sockets_allocated
,
1849 .memory_allocated
= &tcp_memory_allocated
,
1850 .memory_pressure
= &tcp_memory_pressure
,
1851 .orphan_count
= &tcp_orphan_count
,
1852 .sysctl_mem
= sysctl_tcp_mem
,
1853 .sysctl_wmem
= sysctl_tcp_wmem
,
1854 .sysctl_rmem
= sysctl_tcp_rmem
,
1855 .max_header
= MAX_TCP_HEADER
,
1856 .obj_size
= sizeof(struct tcp6_sock
),
1857 .slab_flags
= SLAB_DESTROY_BY_RCU
,
1858 .twsk_prot
= &tcp6_timewait_sock_ops
,
1859 .rsk_prot
= &tcp6_request_sock_ops
,
1860 .h
.hashinfo
= &tcp_hashinfo
,
1861 .no_autobind
= true,
1862 #ifdef CONFIG_COMPAT
1863 .compat_setsockopt
= compat_tcp_setsockopt
,
1864 .compat_getsockopt
= compat_tcp_getsockopt
,
1866 #ifdef CONFIG_MEMCG_KMEM
1867 .proto_cgroup
= tcp_proto_cgroup
,
1869 .clear_sk
= tcp_v6_clear_sk
,
1872 static const struct inet6_protocol tcpv6_protocol
= {
1873 .early_demux
= tcp_v6_early_demux
,
1874 .handler
= tcp_v6_rcv
,
1875 .err_handler
= tcp_v6_err
,
1876 .flags
= INET6_PROTO_NOPOLICY
|INET6_PROTO_FINAL
,
1879 static struct inet_protosw tcpv6_protosw
= {
1880 .type
= SOCK_STREAM
,
1881 .protocol
= IPPROTO_TCP
,
1882 .prot
= &tcpv6_prot
,
1883 .ops
= &inet6_stream_ops
,
1884 .flags
= INET_PROTOSW_PERMANENT
|
1888 static int __net_init
tcpv6_net_init(struct net
*net
)
1890 return inet_ctl_sock_create(&net
->ipv6
.tcp_sk
, PF_INET6
,
1891 SOCK_RAW
, IPPROTO_TCP
, net
);
1894 static void __net_exit
tcpv6_net_exit(struct net
*net
)
1896 inet_ctl_sock_destroy(net
->ipv6
.tcp_sk
);
1899 static void __net_exit
tcpv6_net_exit_batch(struct list_head
*net_exit_list
)
1901 inet_twsk_purge(&tcp_hashinfo
, &tcp_death_row
, AF_INET6
);
1904 static struct pernet_operations tcpv6_net_ops
= {
1905 .init
= tcpv6_net_init
,
1906 .exit
= tcpv6_net_exit
,
1907 .exit_batch
= tcpv6_net_exit_batch
,
1910 int __init
tcpv6_init(void)
1914 ret
= inet6_add_protocol(&tcpv6_protocol
, IPPROTO_TCP
);
1918 /* register inet6 protocol */
1919 ret
= inet6_register_protosw(&tcpv6_protosw
);
1921 goto out_tcpv6_protocol
;
1923 ret
= register_pernet_subsys(&tcpv6_net_ops
);
1925 goto out_tcpv6_protosw
;
1930 inet6_unregister_protosw(&tcpv6_protosw
);
1932 inet6_del_protocol(&tcpv6_protocol
, IPPROTO_TCP
);
1936 void tcpv6_exit(void)
1938 unregister_pernet_subsys(&tcpv6_net_ops
);
1939 inet6_unregister_protosw(&tcpv6_protosw
);
1940 inet6_del_protocol(&tcpv6_protocol
, IPPROTO_TCP
);