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 __u32
tcp_v6_init_sequence(const struct sk_buff
*skb
)
109 return secure_tcpv6_sequence_number(ipv6_hdr(skb
)->daddr
.s6_addr32
,
110 ipv6_hdr(skb
)->saddr
.s6_addr32
,
112 tcp_hdr(skb
)->source
);
115 static int tcp_v6_connect(struct sock
*sk
, struct sockaddr
*uaddr
,
118 struct sockaddr_in6
*usin
= (struct sockaddr_in6
*) uaddr
;
119 struct inet_sock
*inet
= inet_sk(sk
);
120 struct inet_connection_sock
*icsk
= inet_csk(sk
);
121 struct ipv6_pinfo
*np
= inet6_sk(sk
);
122 struct tcp_sock
*tp
= tcp_sk(sk
);
123 struct in6_addr
*saddr
= NULL
, *final_p
, final
;
126 struct dst_entry
*dst
;
130 if (addr_len
< SIN6_LEN_RFC2133
)
133 if (usin
->sin6_family
!= AF_INET6
)
134 return -EAFNOSUPPORT
;
136 memset(&fl6
, 0, sizeof(fl6
));
139 fl6
.flowlabel
= usin
->sin6_flowinfo
&IPV6_FLOWINFO_MASK
;
140 IP6_ECN_flow_init(fl6
.flowlabel
);
141 if (fl6
.flowlabel
&IPV6_FLOWLABEL_MASK
) {
142 struct ip6_flowlabel
*flowlabel
;
143 flowlabel
= fl6_sock_lookup(sk
, fl6
.flowlabel
);
144 if (flowlabel
== NULL
)
146 fl6_sock_release(flowlabel
);
151 * connect() to INADDR_ANY means loopback (BSD'ism).
154 if (ipv6_addr_any(&usin
->sin6_addr
))
155 usin
->sin6_addr
.s6_addr
[15] = 0x1;
157 addr_type
= ipv6_addr_type(&usin
->sin6_addr
);
159 if (addr_type
& IPV6_ADDR_MULTICAST
)
162 if (addr_type
&IPV6_ADDR_LINKLOCAL
) {
163 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
164 usin
->sin6_scope_id
) {
165 /* If interface is set while binding, indices
168 if (sk
->sk_bound_dev_if
&&
169 sk
->sk_bound_dev_if
!= usin
->sin6_scope_id
)
172 sk
->sk_bound_dev_if
= usin
->sin6_scope_id
;
175 /* Connect to link-local address requires an interface */
176 if (!sk
->sk_bound_dev_if
)
180 if (tp
->rx_opt
.ts_recent_stamp
&&
181 !ipv6_addr_equal(&sk
->sk_v6_daddr
, &usin
->sin6_addr
)) {
182 tp
->rx_opt
.ts_recent
= 0;
183 tp
->rx_opt
.ts_recent_stamp
= 0;
187 sk
->sk_v6_daddr
= usin
->sin6_addr
;
188 np
->flow_label
= fl6
.flowlabel
;
194 if (addr_type
== IPV6_ADDR_MAPPED
) {
195 u32 exthdrlen
= icsk
->icsk_ext_hdr_len
;
196 struct sockaddr_in sin
;
198 SOCK_DEBUG(sk
, "connect: ipv4 mapped\n");
200 if (__ipv6_only_sock(sk
))
203 sin
.sin_family
= AF_INET
;
204 sin
.sin_port
= usin
->sin6_port
;
205 sin
.sin_addr
.s_addr
= usin
->sin6_addr
.s6_addr32
[3];
207 icsk
->icsk_af_ops
= &ipv6_mapped
;
208 sk
->sk_backlog_rcv
= tcp_v4_do_rcv
;
209 #ifdef CONFIG_TCP_MD5SIG
210 tp
->af_specific
= &tcp_sock_ipv6_mapped_specific
;
213 err
= tcp_v4_connect(sk
, (struct sockaddr
*)&sin
, sizeof(sin
));
216 icsk
->icsk_ext_hdr_len
= exthdrlen
;
217 icsk
->icsk_af_ops
= &ipv6_specific
;
218 sk
->sk_backlog_rcv
= tcp_v6_do_rcv
;
219 #ifdef CONFIG_TCP_MD5SIG
220 tp
->af_specific
= &tcp_sock_ipv6_specific
;
224 np
->saddr
= sk
->sk_v6_rcv_saddr
;
229 if (!ipv6_addr_any(&sk
->sk_v6_rcv_saddr
))
230 saddr
= &sk
->sk_v6_rcv_saddr
;
232 fl6
.flowi6_proto
= IPPROTO_TCP
;
233 fl6
.daddr
= sk
->sk_v6_daddr
;
234 fl6
.saddr
= saddr
? *saddr
: np
->saddr
;
235 fl6
.flowi6_oif
= sk
->sk_bound_dev_if
;
236 fl6
.flowi6_mark
= sk
->sk_mark
;
237 fl6
.fl6_dport
= usin
->sin6_port
;
238 fl6
.fl6_sport
= inet
->inet_sport
;
240 final_p
= fl6_update_dst(&fl6
, np
->opt
, &final
);
242 security_sk_classify_flow(sk
, flowi6_to_flowi(&fl6
));
244 dst
= ip6_dst_lookup_flow(sk
, &fl6
, final_p
);
252 sk
->sk_v6_rcv_saddr
= *saddr
;
255 /* set the source address */
257 inet
->inet_rcv_saddr
= LOOPBACK4_IPV6
;
259 sk
->sk_gso_type
= SKB_GSO_TCPV6
;
260 __ip6_dst_store(sk
, dst
, NULL
, NULL
);
262 rt
= (struct rt6_info
*) dst
;
263 if (tcp_death_row
.sysctl_tw_recycle
&&
264 !tp
->rx_opt
.ts_recent_stamp
&&
265 ipv6_addr_equal(&rt
->rt6i_dst
.addr
, &sk
->sk_v6_daddr
))
266 tcp_fetch_timewait_stamp(sk
, dst
);
268 icsk
->icsk_ext_hdr_len
= 0;
270 icsk
->icsk_ext_hdr_len
= (np
->opt
->opt_flen
+
273 tp
->rx_opt
.mss_clamp
= IPV6_MIN_MTU
- sizeof(struct tcphdr
) - sizeof(struct ipv6hdr
);
275 inet
->inet_dport
= usin
->sin6_port
;
277 tcp_set_state(sk
, TCP_SYN_SENT
);
278 err
= inet6_hash_connect(&tcp_death_row
, sk
);
284 if (!tp
->write_seq
&& likely(!tp
->repair
))
285 tp
->write_seq
= secure_tcpv6_sequence_number(np
->saddr
.s6_addr32
,
286 sk
->sk_v6_daddr
.s6_addr32
,
290 err
= tcp_connect(sk
);
297 tcp_set_state(sk
, TCP_CLOSE
);
300 inet
->inet_dport
= 0;
301 sk
->sk_route_caps
= 0;
305 static void tcp_v6_mtu_reduced(struct sock
*sk
)
307 struct dst_entry
*dst
;
309 if ((1 << sk
->sk_state
) & (TCPF_LISTEN
| TCPF_CLOSE
))
312 dst
= inet6_csk_update_pmtu(sk
, tcp_sk(sk
)->mtu_info
);
316 if (inet_csk(sk
)->icsk_pmtu_cookie
> dst_mtu(dst
)) {
317 tcp_sync_mss(sk
, dst_mtu(dst
));
318 tcp_simple_retransmit(sk
);
322 static void tcp_v6_err(struct sk_buff
*skb
, struct inet6_skb_parm
*opt
,
323 u8 type
, u8 code
, int offset
, __be32 info
)
325 const struct ipv6hdr
*hdr
= (const struct ipv6hdr
*)skb
->data
;
326 const struct tcphdr
*th
= (struct tcphdr
*)(skb
->data
+offset
);
327 struct net
*net
= dev_net(skb
->dev
);
328 struct request_sock
*fastopen
;
329 struct ipv6_pinfo
*np
;
335 sk
= __inet6_lookup_established(net
, &tcp_hashinfo
,
336 &hdr
->daddr
, th
->dest
,
337 &hdr
->saddr
, ntohs(th
->source
),
341 ICMP6_INC_STATS_BH(net
, __in6_dev_get(skb
->dev
),
346 if (sk
->sk_state
== TCP_TIME_WAIT
) {
347 inet_twsk_put(inet_twsk(sk
));
350 seq
= ntohl(th
->seq
);
351 if (sk
->sk_state
== TCP_NEW_SYN_RECV
)
352 return tcp_req_err(sk
, seq
);
355 if (sock_owned_by_user(sk
) && type
!= ICMPV6_PKT_TOOBIG
)
356 NET_INC_STATS_BH(net
, LINUX_MIB_LOCKDROPPEDICMPS
);
358 if (sk
->sk_state
== TCP_CLOSE
)
361 if (ipv6_hdr(skb
)->hop_limit
< inet6_sk(sk
)->min_hopcount
) {
362 NET_INC_STATS_BH(net
, LINUX_MIB_TCPMINTTLDROP
);
367 /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
368 fastopen
= tp
->fastopen_rsk
;
369 snd_una
= fastopen
? tcp_rsk(fastopen
)->snt_isn
: tp
->snd_una
;
370 if (sk
->sk_state
!= TCP_LISTEN
&&
371 !between(seq
, snd_una
, tp
->snd_nxt
)) {
372 NET_INC_STATS_BH(net
, LINUX_MIB_OUTOFWINDOWICMPS
);
378 if (type
== NDISC_REDIRECT
) {
379 struct dst_entry
*dst
= __sk_dst_check(sk
, np
->dst_cookie
);
382 dst
->ops
->redirect(dst
, sk
, skb
);
386 if (type
== ICMPV6_PKT_TOOBIG
) {
387 /* We are not interested in TCP_LISTEN and open_requests
388 * (SYN-ACKs send out by Linux are always <576bytes so
389 * they should go through unfragmented).
391 if (sk
->sk_state
== TCP_LISTEN
)
394 if (!ip6_sk_accept_pmtu(sk
))
397 tp
->mtu_info
= ntohl(info
);
398 if (!sock_owned_by_user(sk
))
399 tcp_v6_mtu_reduced(sk
);
400 else if (!test_and_set_bit(TCP_MTU_REDUCED_DEFERRED
,
406 icmpv6_err_convert(type
, code
, &err
);
408 /* Might be for an request_sock */
409 switch (sk
->sk_state
) {
412 /* Only in fast or simultaneous open. If a fast open socket is
413 * is already accepted it is treated as a connected one below.
415 if (fastopen
&& fastopen
->sk
== NULL
)
418 if (!sock_owned_by_user(sk
)) {
420 sk
->sk_error_report(sk
); /* Wake people up to see the error (see connect in sock.c) */
424 sk
->sk_err_soft
= err
;
428 if (!sock_owned_by_user(sk
) && np
->recverr
) {
430 sk
->sk_error_report(sk
);
432 sk
->sk_err_soft
= err
;
440 static int tcp_v6_send_synack(struct sock
*sk
, struct dst_entry
*dst
,
442 struct request_sock
*req
,
444 struct tcp_fastopen_cookie
*foc
)
446 struct inet_request_sock
*ireq
= inet_rsk(req
);
447 struct ipv6_pinfo
*np
= inet6_sk(sk
);
448 struct flowi6
*fl6
= &fl
->u
.ip6
;
452 /* First, grab a route. */
453 if (!dst
&& (dst
= inet6_csk_route_req(sk
, fl6
, req
)) == NULL
)
456 skb
= tcp_make_synack(sk
, dst
, req
, foc
);
459 __tcp_v6_send_check(skb
, &ireq
->ir_v6_loc_addr
,
460 &ireq
->ir_v6_rmt_addr
);
462 fl6
->daddr
= ireq
->ir_v6_rmt_addr
;
463 if (np
->repflow
&& (ireq
->pktopts
!= NULL
))
464 fl6
->flowlabel
= ip6_flowlabel(ipv6_hdr(ireq
->pktopts
));
466 skb_set_queue_mapping(skb
, queue_mapping
);
467 err
= ip6_xmit(sk
, skb
, fl6
, np
->opt
, np
->tclass
);
468 err
= net_xmit_eval(err
);
476 static void tcp_v6_reqsk_destructor(struct request_sock
*req
)
478 kfree_skb(inet_rsk(req
)->pktopts
);
481 #ifdef CONFIG_TCP_MD5SIG
482 static struct tcp_md5sig_key
*tcp_v6_md5_do_lookup(struct sock
*sk
,
483 const struct in6_addr
*addr
)
485 return tcp_md5_do_lookup(sk
, (union tcp_md5_addr
*)addr
, AF_INET6
);
488 static struct tcp_md5sig_key
*tcp_v6_md5_lookup(struct sock
*sk
,
489 struct sock
*addr_sk
)
491 return tcp_v6_md5_do_lookup(sk
, &addr_sk
->sk_v6_daddr
);
494 static struct tcp_md5sig_key
*tcp_v6_reqsk_md5_lookup(struct sock
*sk
,
495 struct request_sock
*req
)
497 return tcp_v6_md5_do_lookup(sk
, &inet_rsk(req
)->ir_v6_rmt_addr
);
500 static int tcp_v6_parse_md5_keys(struct sock
*sk
, char __user
*optval
,
503 struct tcp_md5sig cmd
;
504 struct sockaddr_in6
*sin6
= (struct sockaddr_in6
*)&cmd
.tcpm_addr
;
506 if (optlen
< sizeof(cmd
))
509 if (copy_from_user(&cmd
, optval
, sizeof(cmd
)))
512 if (sin6
->sin6_family
!= AF_INET6
)
515 if (!cmd
.tcpm_keylen
) {
516 if (ipv6_addr_v4mapped(&sin6
->sin6_addr
))
517 return tcp_md5_do_del(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
.s6_addr32
[3],
519 return tcp_md5_do_del(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
,
523 if (cmd
.tcpm_keylen
> TCP_MD5SIG_MAXKEYLEN
)
526 if (ipv6_addr_v4mapped(&sin6
->sin6_addr
))
527 return tcp_md5_do_add(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
.s6_addr32
[3],
528 AF_INET
, cmd
.tcpm_key
, cmd
.tcpm_keylen
, GFP_KERNEL
);
530 return tcp_md5_do_add(sk
, (union tcp_md5_addr
*)&sin6
->sin6_addr
,
531 AF_INET6
, cmd
.tcpm_key
, cmd
.tcpm_keylen
, GFP_KERNEL
);
534 static int tcp_v6_md5_hash_pseudoheader(struct tcp_md5sig_pool
*hp
,
535 const struct in6_addr
*daddr
,
536 const struct in6_addr
*saddr
, int nbytes
)
538 struct tcp6_pseudohdr
*bp
;
539 struct scatterlist sg
;
541 bp
= &hp
->md5_blk
.ip6
;
542 /* 1. TCP pseudo-header (RFC2460) */
545 bp
->protocol
= cpu_to_be32(IPPROTO_TCP
);
546 bp
->len
= cpu_to_be32(nbytes
);
548 sg_init_one(&sg
, bp
, sizeof(*bp
));
549 return crypto_hash_update(&hp
->md5_desc
, &sg
, sizeof(*bp
));
552 static int tcp_v6_md5_hash_hdr(char *md5_hash
, struct tcp_md5sig_key
*key
,
553 const struct in6_addr
*daddr
, struct in6_addr
*saddr
,
554 const struct tcphdr
*th
)
556 struct tcp_md5sig_pool
*hp
;
557 struct hash_desc
*desc
;
559 hp
= tcp_get_md5sig_pool();
561 goto clear_hash_noput
;
562 desc
= &hp
->md5_desc
;
564 if (crypto_hash_init(desc
))
566 if (tcp_v6_md5_hash_pseudoheader(hp
, daddr
, saddr
, th
->doff
<< 2))
568 if (tcp_md5_hash_header(hp
, th
))
570 if (tcp_md5_hash_key(hp
, key
))
572 if (crypto_hash_final(desc
, md5_hash
))
575 tcp_put_md5sig_pool();
579 tcp_put_md5sig_pool();
581 memset(md5_hash
, 0, 16);
585 static int tcp_v6_md5_hash_skb(char *md5_hash
, struct tcp_md5sig_key
*key
,
586 const struct sock
*sk
,
587 const struct request_sock
*req
,
588 const struct sk_buff
*skb
)
590 const struct in6_addr
*saddr
, *daddr
;
591 struct tcp_md5sig_pool
*hp
;
592 struct hash_desc
*desc
;
593 const struct tcphdr
*th
= tcp_hdr(skb
);
596 saddr
= &inet6_sk(sk
)->saddr
;
597 daddr
= &sk
->sk_v6_daddr
;
599 saddr
= &inet_rsk(req
)->ir_v6_loc_addr
;
600 daddr
= &inet_rsk(req
)->ir_v6_rmt_addr
;
602 const struct ipv6hdr
*ip6h
= ipv6_hdr(skb
);
603 saddr
= &ip6h
->saddr
;
604 daddr
= &ip6h
->daddr
;
607 hp
= tcp_get_md5sig_pool();
609 goto clear_hash_noput
;
610 desc
= &hp
->md5_desc
;
612 if (crypto_hash_init(desc
))
615 if (tcp_v6_md5_hash_pseudoheader(hp
, daddr
, saddr
, skb
->len
))
617 if (tcp_md5_hash_header(hp
, th
))
619 if (tcp_md5_hash_skb_data(hp
, skb
, th
->doff
<< 2))
621 if (tcp_md5_hash_key(hp
, key
))
623 if (crypto_hash_final(desc
, md5_hash
))
626 tcp_put_md5sig_pool();
630 tcp_put_md5sig_pool();
632 memset(md5_hash
, 0, 16);
636 static int __tcp_v6_inbound_md5_hash(struct sock
*sk
,
637 const struct sk_buff
*skb
)
639 const __u8
*hash_location
= NULL
;
640 struct tcp_md5sig_key
*hash_expected
;
641 const struct ipv6hdr
*ip6h
= ipv6_hdr(skb
);
642 const struct tcphdr
*th
= tcp_hdr(skb
);
646 hash_expected
= tcp_v6_md5_do_lookup(sk
, &ip6h
->saddr
);
647 hash_location
= tcp_parse_md5sig_option(th
);
649 /* We've parsed the options - do we have a hash? */
650 if (!hash_expected
&& !hash_location
)
653 if (hash_expected
&& !hash_location
) {
654 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_TCPMD5NOTFOUND
);
658 if (!hash_expected
&& hash_location
) {
659 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_TCPMD5UNEXPECTED
);
663 /* check the signature */
664 genhash
= tcp_v6_md5_hash_skb(newhash
,
668 if (genhash
|| memcmp(hash_location
, newhash
, 16) != 0) {
669 net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u\n",
670 genhash
? "failed" : "mismatch",
671 &ip6h
->saddr
, ntohs(th
->source
),
672 &ip6h
->daddr
, ntohs(th
->dest
));
678 static int tcp_v6_inbound_md5_hash(struct sock
*sk
, const struct sk_buff
*skb
)
683 ret
= __tcp_v6_inbound_md5_hash(sk
, skb
);
691 static void tcp_v6_init_req(struct request_sock
*req
, struct sock
*sk
,
694 struct inet_request_sock
*ireq
= inet_rsk(req
);
695 struct ipv6_pinfo
*np
= inet6_sk(sk
);
697 ireq
->ir_v6_rmt_addr
= ipv6_hdr(skb
)->saddr
;
698 ireq
->ir_v6_loc_addr
= ipv6_hdr(skb
)->daddr
;
700 /* So that link locals have meaning */
701 if (!sk
->sk_bound_dev_if
&&
702 ipv6_addr_type(&ireq
->ir_v6_rmt_addr
) & IPV6_ADDR_LINKLOCAL
)
703 ireq
->ir_iif
= tcp_v6_iif(skb
);
705 if (!TCP_SKB_CB(skb
)->tcp_tw_isn
&&
706 (ipv6_opt_accepted(sk
, skb
, &TCP_SKB_CB(skb
)->header
.h6
) ||
707 np
->rxopt
.bits
.rxinfo
||
708 np
->rxopt
.bits
.rxoinfo
|| np
->rxopt
.bits
.rxhlim
||
709 np
->rxopt
.bits
.rxohlim
|| np
->repflow
)) {
710 atomic_inc(&skb
->users
);
713 ireq
->ireq_family
= AF_INET6
;
716 static struct dst_entry
*tcp_v6_route_req(struct sock
*sk
, struct flowi
*fl
,
717 const struct request_sock
*req
,
722 return inet6_csk_route_req(sk
, &fl
->u
.ip6
, req
);
725 struct request_sock_ops tcp6_request_sock_ops __read_mostly
= {
727 .obj_size
= sizeof(struct tcp6_request_sock
),
728 .rtx_syn_ack
= tcp_rtx_synack
,
729 .send_ack
= tcp_v6_reqsk_send_ack
,
730 .destructor
= tcp_v6_reqsk_destructor
,
731 .send_reset
= tcp_v6_send_reset
,
732 .syn_ack_timeout
= tcp_syn_ack_timeout
,
735 static const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops
= {
736 .mss_clamp
= IPV6_MIN_MTU
- sizeof(struct tcphdr
) -
737 sizeof(struct ipv6hdr
),
738 #ifdef CONFIG_TCP_MD5SIG
739 .md5_lookup
= tcp_v6_reqsk_md5_lookup
,
740 .calc_md5_hash
= tcp_v6_md5_hash_skb
,
742 .init_req
= tcp_v6_init_req
,
743 #ifdef CONFIG_SYN_COOKIES
744 .cookie_init_seq
= cookie_v6_init_sequence
,
746 .route_req
= tcp_v6_route_req
,
747 .init_seq
= tcp_v6_init_sequence
,
748 .send_synack
= tcp_v6_send_synack
,
749 .queue_hash_add
= inet6_csk_reqsk_queue_hash_add
,
752 static void tcp_v6_send_response(struct sock
*sk
, struct sk_buff
*skb
, u32 seq
,
753 u32 ack
, u32 win
, u32 tsval
, u32 tsecr
,
754 int oif
, struct tcp_md5sig_key
*key
, int rst
,
755 u8 tclass
, u32 label
)
757 const struct tcphdr
*th
= tcp_hdr(skb
);
759 struct sk_buff
*buff
;
761 struct net
*net
= sk
? sock_net(sk
) : dev_net(skb_dst(skb
)->dev
);
762 struct sock
*ctl_sk
= net
->ipv6
.tcp_sk
;
763 unsigned int tot_len
= sizeof(struct tcphdr
);
764 struct dst_entry
*dst
;
768 tot_len
+= TCPOLEN_TSTAMP_ALIGNED
;
769 #ifdef CONFIG_TCP_MD5SIG
771 tot_len
+= TCPOLEN_MD5SIG_ALIGNED
;
774 buff
= alloc_skb(MAX_HEADER
+ sizeof(struct ipv6hdr
) + tot_len
,
779 skb_reserve(buff
, MAX_HEADER
+ sizeof(struct ipv6hdr
) + tot_len
);
781 t1
= (struct tcphdr
*) skb_push(buff
, tot_len
);
782 skb_reset_transport_header(buff
);
784 /* Swap the send and the receive. */
785 memset(t1
, 0, sizeof(*t1
));
786 t1
->dest
= th
->source
;
787 t1
->source
= th
->dest
;
788 t1
->doff
= tot_len
/ 4;
789 t1
->seq
= htonl(seq
);
790 t1
->ack_seq
= htonl(ack
);
791 t1
->ack
= !rst
|| !th
->ack
;
793 t1
->window
= htons(win
);
795 topt
= (__be32
*)(t1
+ 1);
798 *topt
++ = htonl((TCPOPT_NOP
<< 24) | (TCPOPT_NOP
<< 16) |
799 (TCPOPT_TIMESTAMP
<< 8) | TCPOLEN_TIMESTAMP
);
800 *topt
++ = htonl(tsval
);
801 *topt
++ = htonl(tsecr
);
804 #ifdef CONFIG_TCP_MD5SIG
806 *topt
++ = htonl((TCPOPT_NOP
<< 24) | (TCPOPT_NOP
<< 16) |
807 (TCPOPT_MD5SIG
<< 8) | TCPOLEN_MD5SIG
);
808 tcp_v6_md5_hash_hdr((__u8
*)topt
, key
,
809 &ipv6_hdr(skb
)->saddr
,
810 &ipv6_hdr(skb
)->daddr
, t1
);
814 memset(&fl6
, 0, sizeof(fl6
));
815 fl6
.daddr
= ipv6_hdr(skb
)->saddr
;
816 fl6
.saddr
= ipv6_hdr(skb
)->daddr
;
817 fl6
.flowlabel
= label
;
819 buff
->ip_summed
= CHECKSUM_PARTIAL
;
822 __tcp_v6_send_check(buff
, &fl6
.saddr
, &fl6
.daddr
);
824 fl6
.flowi6_proto
= IPPROTO_TCP
;
825 if (rt6_need_strict(&fl6
.daddr
) && !oif
)
826 fl6
.flowi6_oif
= tcp_v6_iif(skb
);
828 fl6
.flowi6_oif
= oif
;
829 fl6
.flowi6_mark
= IP6_REPLY_MARK(net
, skb
->mark
);
830 fl6
.fl6_dport
= t1
->dest
;
831 fl6
.fl6_sport
= t1
->source
;
832 security_skb_classify_flow(skb
, flowi6_to_flowi(&fl6
));
834 /* Pass a socket to ip6_dst_lookup either it is for RST
835 * Underlying function will use this to retrieve the network
838 dst
= ip6_dst_lookup_flow(ctl_sk
, &fl6
, NULL
);
840 skb_dst_set(buff
, dst
);
841 ip6_xmit(ctl_sk
, buff
, &fl6
, NULL
, tclass
);
842 TCP_INC_STATS_BH(net
, TCP_MIB_OUTSEGS
);
844 TCP_INC_STATS_BH(net
, TCP_MIB_OUTRSTS
);
851 static void tcp_v6_send_reset(struct sock
*sk
, struct sk_buff
*skb
)
853 const struct tcphdr
*th
= tcp_hdr(skb
);
854 u32 seq
= 0, ack_seq
= 0;
855 struct tcp_md5sig_key
*key
= NULL
;
856 #ifdef CONFIG_TCP_MD5SIG
857 const __u8
*hash_location
= NULL
;
858 struct ipv6hdr
*ipv6h
= ipv6_hdr(skb
);
859 unsigned char newhash
[16];
861 struct sock
*sk1
= NULL
;
868 /* If sk not NULL, it means we did a successful lookup and incoming
869 * route had to be correct. prequeue might have dropped our dst.
871 if (!sk
&& !ipv6_unicast_destination(skb
))
874 #ifdef CONFIG_TCP_MD5SIG
875 hash_location
= tcp_parse_md5sig_option(th
);
876 if (!sk
&& hash_location
) {
878 * active side is lost. Try to find listening socket through
879 * source port, and then find md5 key through listening socket.
880 * we are not loose security here:
881 * Incoming packet is checked with md5 hash with finding key,
882 * no RST generated if md5 hash doesn't match.
884 sk1
= inet6_lookup_listener(dev_net(skb_dst(skb
)->dev
),
885 &tcp_hashinfo
, &ipv6h
->saddr
,
886 th
->source
, &ipv6h
->daddr
,
887 ntohs(th
->source
), tcp_v6_iif(skb
));
892 key
= tcp_v6_md5_do_lookup(sk1
, &ipv6h
->saddr
);
896 genhash
= tcp_v6_md5_hash_skb(newhash
, key
, NULL
, NULL
, skb
);
897 if (genhash
|| memcmp(hash_location
, newhash
, 16) != 0)
900 key
= sk
? tcp_v6_md5_do_lookup(sk
, &ipv6h
->saddr
) : NULL
;
905 seq
= ntohl(th
->ack_seq
);
907 ack_seq
= ntohl(th
->seq
) + th
->syn
+ th
->fin
+ skb
->len
-
910 oif
= sk
? sk
->sk_bound_dev_if
: 0;
911 tcp_v6_send_response(sk
, skb
, seq
, ack_seq
, 0, 0, 0, oif
, key
, 1, 0, 0);
913 #ifdef CONFIG_TCP_MD5SIG
922 static void tcp_v6_send_ack(struct sock
*sk
, struct sk_buff
*skb
, u32 seq
,
923 u32 ack
, u32 win
, u32 tsval
, u32 tsecr
, int oif
,
924 struct tcp_md5sig_key
*key
, u8 tclass
,
927 tcp_v6_send_response(sk
, skb
, seq
, ack
, win
, tsval
, tsecr
, oif
, key
, 0,
931 static void tcp_v6_timewait_ack(struct sock
*sk
, struct sk_buff
*skb
)
933 struct inet_timewait_sock
*tw
= inet_twsk(sk
);
934 struct tcp_timewait_sock
*tcptw
= tcp_twsk(sk
);
936 tcp_v6_send_ack(sk
, skb
, tcptw
->tw_snd_nxt
, tcptw
->tw_rcv_nxt
,
937 tcptw
->tw_rcv_wnd
>> tw
->tw_rcv_wscale
,
938 tcp_time_stamp
+ tcptw
->tw_ts_offset
,
939 tcptw
->tw_ts_recent
, tw
->tw_bound_dev_if
, tcp_twsk_md5_key(tcptw
),
940 tw
->tw_tclass
, (tw
->tw_flowlabel
<< 12));
945 static void tcp_v6_reqsk_send_ack(struct sock
*sk
, struct sk_buff
*skb
,
946 struct request_sock
*req
)
948 /* sk->sk_state == TCP_LISTEN -> for regular TCP_SYN_RECV
949 * sk->sk_state == TCP_SYN_RECV -> for Fast Open.
951 tcp_v6_send_ack(sk
, skb
, (sk
->sk_state
== TCP_LISTEN
) ?
952 tcp_rsk(req
)->snt_isn
+ 1 : tcp_sk(sk
)->snd_nxt
,
953 tcp_rsk(req
)->rcv_nxt
, req
->rcv_wnd
,
954 tcp_time_stamp
, req
->ts_recent
, sk
->sk_bound_dev_if
,
955 tcp_v6_md5_do_lookup(sk
, &ipv6_hdr(skb
)->daddr
),
960 static struct sock
*tcp_v6_hnd_req(struct sock
*sk
, struct sk_buff
*skb
)
962 const struct tcphdr
*th
= tcp_hdr(skb
);
963 struct request_sock
*req
;
966 /* Find possible connection requests. */
967 req
= inet6_csk_search_req(sk
, th
->source
,
968 &ipv6_hdr(skb
)->saddr
,
969 &ipv6_hdr(skb
)->daddr
, tcp_v6_iif(skb
));
971 nsk
= tcp_check_req(sk
, skb
, req
, false);
975 nsk
= __inet6_lookup_established(sock_net(sk
), &tcp_hashinfo
,
976 &ipv6_hdr(skb
)->saddr
, th
->source
,
977 &ipv6_hdr(skb
)->daddr
, ntohs(th
->dest
),
981 if (nsk
->sk_state
!= TCP_TIME_WAIT
) {
985 inet_twsk_put(inet_twsk(nsk
));
989 #ifdef CONFIG_SYN_COOKIES
991 sk
= cookie_v6_check(sk
, skb
);
996 static int tcp_v6_conn_request(struct sock
*sk
, struct sk_buff
*skb
)
998 if (skb
->protocol
== htons(ETH_P_IP
))
999 return tcp_v4_conn_request(sk
, skb
);
1001 if (!ipv6_unicast_destination(skb
))
1004 return tcp_conn_request(&tcp6_request_sock_ops
,
1005 &tcp_request_sock_ipv6_ops
, sk
, skb
);
1008 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
1009 return 0; /* don't send reset */
1012 static struct sock
*tcp_v6_syn_recv_sock(struct sock
*sk
, struct sk_buff
*skb
,
1013 struct request_sock
*req
,
1014 struct dst_entry
*dst
)
1016 struct inet_request_sock
*ireq
;
1017 struct ipv6_pinfo
*newnp
, *np
= inet6_sk(sk
);
1018 struct tcp6_sock
*newtcp6sk
;
1019 struct inet_sock
*newinet
;
1020 struct tcp_sock
*newtp
;
1022 #ifdef CONFIG_TCP_MD5SIG
1023 struct tcp_md5sig_key
*key
;
1027 if (skb
->protocol
== htons(ETH_P_IP
)) {
1032 newsk
= tcp_v4_syn_recv_sock(sk
, skb
, req
, dst
);
1037 newtcp6sk
= (struct tcp6_sock
*)newsk
;
1038 inet_sk(newsk
)->pinet6
= &newtcp6sk
->inet6
;
1040 newinet
= inet_sk(newsk
);
1041 newnp
= inet6_sk(newsk
);
1042 newtp
= tcp_sk(newsk
);
1044 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
1046 newnp
->saddr
= newsk
->sk_v6_rcv_saddr
;
1048 inet_csk(newsk
)->icsk_af_ops
= &ipv6_mapped
;
1049 newsk
->sk_backlog_rcv
= tcp_v4_do_rcv
;
1050 #ifdef CONFIG_TCP_MD5SIG
1051 newtp
->af_specific
= &tcp_sock_ipv6_mapped_specific
;
1054 newnp
->ipv6_ac_list
= NULL
;
1055 newnp
->ipv6_fl_list
= NULL
;
1056 newnp
->pktoptions
= NULL
;
1058 newnp
->mcast_oif
= tcp_v6_iif(skb
);
1059 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
1060 newnp
->rcv_flowinfo
= ip6_flowinfo(ipv6_hdr(skb
));
1062 newnp
->flow_label
= ip6_flowlabel(ipv6_hdr(skb
));
1065 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
1066 * here, tcp_create_openreq_child now does this for us, see the comment in
1067 * that function for the gory details. -acme
1070 /* It is tricky place. Until this moment IPv4 tcp
1071 worked with IPv6 icsk.icsk_af_ops.
1074 tcp_sync_mss(newsk
, inet_csk(newsk
)->icsk_pmtu_cookie
);
1079 ireq
= inet_rsk(req
);
1081 if (sk_acceptq_is_full(sk
))
1085 dst
= inet6_csk_route_req(sk
, &fl6
, req
);
1090 newsk
= tcp_create_openreq_child(sk
, req
, skb
);
1095 * No need to charge this sock to the relevant IPv6 refcnt debug socks
1096 * count here, tcp_create_openreq_child now does this for us, see the
1097 * comment in that function for the gory details. -acme
1100 newsk
->sk_gso_type
= SKB_GSO_TCPV6
;
1101 __ip6_dst_store(newsk
, dst
, NULL
, NULL
);
1102 inet6_sk_rx_dst_set(newsk
, skb
);
1104 newtcp6sk
= (struct tcp6_sock
*)newsk
;
1105 inet_sk(newsk
)->pinet6
= &newtcp6sk
->inet6
;
1107 newtp
= tcp_sk(newsk
);
1108 newinet
= inet_sk(newsk
);
1109 newnp
= inet6_sk(newsk
);
1111 memcpy(newnp
, np
, sizeof(struct ipv6_pinfo
));
1113 newsk
->sk_v6_daddr
= ireq
->ir_v6_rmt_addr
;
1114 newnp
->saddr
= ireq
->ir_v6_loc_addr
;
1115 newsk
->sk_v6_rcv_saddr
= ireq
->ir_v6_loc_addr
;
1116 newsk
->sk_bound_dev_if
= ireq
->ir_iif
;
1118 ip6_set_txhash(newsk
);
1120 /* Now IPv6 options...
1122 First: no IPv4 options.
1124 newinet
->inet_opt
= NULL
;
1125 newnp
->ipv6_ac_list
= NULL
;
1126 newnp
->ipv6_fl_list
= NULL
;
1129 newnp
->rxopt
.all
= np
->rxopt
.all
;
1131 /* Clone pktoptions received with SYN */
1132 newnp
->pktoptions
= NULL
;
1133 if (ireq
->pktopts
!= NULL
) {
1134 newnp
->pktoptions
= skb_clone(ireq
->pktopts
,
1135 sk_gfp_atomic(sk
, GFP_ATOMIC
));
1136 consume_skb(ireq
->pktopts
);
1137 ireq
->pktopts
= NULL
;
1138 if (newnp
->pktoptions
)
1139 skb_set_owner_r(newnp
->pktoptions
, newsk
);
1142 newnp
->mcast_oif
= tcp_v6_iif(skb
);
1143 newnp
->mcast_hops
= ipv6_hdr(skb
)->hop_limit
;
1144 newnp
->rcv_flowinfo
= ip6_flowinfo(ipv6_hdr(skb
));
1146 newnp
->flow_label
= ip6_flowlabel(ipv6_hdr(skb
));
1148 /* Clone native IPv6 options from listening socket (if any)
1150 Yes, keeping reference count would be much more clever,
1151 but we make one more one thing there: reattach optmem
1155 newnp
->opt
= ipv6_dup_options(newsk
, np
->opt
);
1157 inet_csk(newsk
)->icsk_ext_hdr_len
= 0;
1159 inet_csk(newsk
)->icsk_ext_hdr_len
= (newnp
->opt
->opt_nflen
+
1160 newnp
->opt
->opt_flen
);
1162 tcp_ca_openreq_child(newsk
, dst
);
1164 tcp_sync_mss(newsk
, dst_mtu(dst
));
1165 newtp
->advmss
= dst_metric_advmss(dst
);
1166 if (tcp_sk(sk
)->rx_opt
.user_mss
&&
1167 tcp_sk(sk
)->rx_opt
.user_mss
< newtp
->advmss
)
1168 newtp
->advmss
= tcp_sk(sk
)->rx_opt
.user_mss
;
1170 tcp_initialize_rcv_mss(newsk
);
1172 newinet
->inet_daddr
= newinet
->inet_saddr
= LOOPBACK4_IPV6
;
1173 newinet
->inet_rcv_saddr
= LOOPBACK4_IPV6
;
1175 #ifdef CONFIG_TCP_MD5SIG
1176 /* Copy over the MD5 key from the original socket */
1177 key
= tcp_v6_md5_do_lookup(sk
, &newsk
->sk_v6_daddr
);
1179 /* We're using one, so create a matching key
1180 * on the newsk structure. If we fail to get
1181 * memory, then we end up not copying the key
1184 tcp_md5_do_add(newsk
, (union tcp_md5_addr
*)&newsk
->sk_v6_daddr
,
1185 AF_INET6
, key
->key
, key
->keylen
,
1186 sk_gfp_atomic(sk
, GFP_ATOMIC
));
1190 if (__inet_inherit_port(sk
, newsk
) < 0) {
1191 inet_csk_prepare_forced_close(newsk
);
1195 __inet_hash(newsk
, NULL
);
1200 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENOVERFLOWS
);
1204 NET_INC_STATS_BH(sock_net(sk
), LINUX_MIB_LISTENDROPS
);
1208 /* The socket must have it's spinlock held when we get
1211 * We have a potential double-lock case here, so even when
1212 * doing backlog processing we use the BH locking scheme.
1213 * This is because we cannot sleep with the original spinlock
1216 static int tcp_v6_do_rcv(struct sock
*sk
, struct sk_buff
*skb
)
1218 struct ipv6_pinfo
*np
= inet6_sk(sk
);
1219 struct tcp_sock
*tp
;
1220 struct sk_buff
*opt_skb
= NULL
;
1222 /* Imagine: socket is IPv6. IPv4 packet arrives,
1223 goes to IPv4 receive handler and backlogged.
1224 From backlog it always goes here. Kerboom...
1225 Fortunately, tcp_rcv_established and rcv_established
1226 handle them correctly, but it is not case with
1227 tcp_v6_hnd_req and tcp_v6_send_reset(). --ANK
1230 if (skb
->protocol
== htons(ETH_P_IP
))
1231 return tcp_v4_do_rcv(sk
, skb
);
1233 if (sk_filter(sk
, skb
))
1237 * socket locking is here for SMP purposes as backlog rcv
1238 * is currently called with bh processing disabled.
1241 /* Do Stevens' IPV6_PKTOPTIONS.
1243 Yes, guys, it is the only place in our code, where we
1244 may make it not affecting IPv4.
1245 The rest of code is protocol independent,
1246 and I do not like idea to uglify IPv4.
1248 Actually, all the idea behind IPV6_PKTOPTIONS
1249 looks not very well thought. For now we latch
1250 options, received in the last packet, enqueued
1251 by tcp. Feel free to propose better solution.
1255 opt_skb
= skb_clone(skb
, sk_gfp_atomic(sk
, GFP_ATOMIC
));
1257 if (sk
->sk_state
== TCP_ESTABLISHED
) { /* Fast path */
1258 struct dst_entry
*dst
= sk
->sk_rx_dst
;
1260 sock_rps_save_rxhash(sk
, skb
);
1261 sk_mark_napi_id(sk
, skb
);
1263 if (inet_sk(sk
)->rx_dst_ifindex
!= skb
->skb_iif
||
1264 dst
->ops
->check(dst
, np
->rx_dst_cookie
) == NULL
) {
1266 sk
->sk_rx_dst
= NULL
;
1270 tcp_rcv_established(sk
, skb
, tcp_hdr(skb
), skb
->len
);
1272 goto ipv6_pktoptions
;
1276 if (skb
->len
< tcp_hdrlen(skb
) || tcp_checksum_complete(skb
))
1279 if (sk
->sk_state
== TCP_LISTEN
) {
1280 struct sock
*nsk
= tcp_v6_hnd_req(sk
, skb
);
1285 * Queue it on the new socket if the new socket is active,
1286 * otherwise we just shortcircuit this and continue with
1290 sock_rps_save_rxhash(nsk
, skb
);
1291 sk_mark_napi_id(sk
, skb
);
1292 if (tcp_child_process(sk
, nsk
, skb
))
1295 __kfree_skb(opt_skb
);
1299 sock_rps_save_rxhash(sk
, skb
);
1301 if (tcp_rcv_state_process(sk
, skb
, tcp_hdr(skb
), skb
->len
))
1304 goto ipv6_pktoptions
;
1308 tcp_v6_send_reset(sk
, skb
);
1311 __kfree_skb(opt_skb
);
1315 TCP_INC_STATS_BH(sock_net(sk
), TCP_MIB_CSUMERRORS
);
1316 TCP_INC_STATS_BH(sock_net(sk
), TCP_MIB_INERRS
);
1321 /* Do you ask, what is it?
1323 1. skb was enqueued by tcp.
1324 2. skb is added to tail of read queue, rather than out of order.
1325 3. socket is not in passive state.
1326 4. Finally, it really contains options, which user wants to receive.
1329 if (TCP_SKB_CB(opt_skb
)->end_seq
== tp
->rcv_nxt
&&
1330 !((1 << sk
->sk_state
) & (TCPF_CLOSE
| TCPF_LISTEN
))) {
1331 if (np
->rxopt
.bits
.rxinfo
|| np
->rxopt
.bits
.rxoinfo
)
1332 np
->mcast_oif
= tcp_v6_iif(opt_skb
);
1333 if (np
->rxopt
.bits
.rxhlim
|| np
->rxopt
.bits
.rxohlim
)
1334 np
->mcast_hops
= ipv6_hdr(opt_skb
)->hop_limit
;
1335 if (np
->rxopt
.bits
.rxflow
|| np
->rxopt
.bits
.rxtclass
)
1336 np
->rcv_flowinfo
= ip6_flowinfo(ipv6_hdr(opt_skb
));
1338 np
->flow_label
= ip6_flowlabel(ipv6_hdr(opt_skb
));
1339 if (ipv6_opt_accepted(sk
, opt_skb
, &TCP_SKB_CB(opt_skb
)->header
.h6
)) {
1340 skb_set_owner_r(opt_skb
, sk
);
1341 opt_skb
= xchg(&np
->pktoptions
, opt_skb
);
1343 __kfree_skb(opt_skb
);
1344 opt_skb
= xchg(&np
->pktoptions
, NULL
);
1352 static void tcp_v6_fill_cb(struct sk_buff
*skb
, const struct ipv6hdr
*hdr
,
1353 const struct tcphdr
*th
)
1355 /* This is tricky: we move IP6CB at its correct location into
1356 * TCP_SKB_CB(). It must be done after xfrm6_policy_check(), because
1357 * _decode_session6() uses IP6CB().
1358 * barrier() makes sure compiler won't play aliasing games.
1360 memmove(&TCP_SKB_CB(skb
)->header
.h6
, IP6CB(skb
),
1361 sizeof(struct inet6_skb_parm
));
1364 TCP_SKB_CB(skb
)->seq
= ntohl(th
->seq
);
1365 TCP_SKB_CB(skb
)->end_seq
= (TCP_SKB_CB(skb
)->seq
+ th
->syn
+ th
->fin
+
1366 skb
->len
- th
->doff
*4);
1367 TCP_SKB_CB(skb
)->ack_seq
= ntohl(th
->ack_seq
);
1368 TCP_SKB_CB(skb
)->tcp_flags
= tcp_flag_byte(th
);
1369 TCP_SKB_CB(skb
)->tcp_tw_isn
= 0;
1370 TCP_SKB_CB(skb
)->ip_dsfield
= ipv6_get_dsfield(hdr
);
1371 TCP_SKB_CB(skb
)->sacked
= 0;
1374 static int tcp_v6_rcv(struct sk_buff
*skb
)
1376 const struct tcphdr
*th
;
1377 const struct ipv6hdr
*hdr
;
1380 struct net
*net
= dev_net(skb
->dev
);
1382 if (skb
->pkt_type
!= PACKET_HOST
)
1386 * Count it even if it's bad.
1388 TCP_INC_STATS_BH(net
, TCP_MIB_INSEGS
);
1390 if (!pskb_may_pull(skb
, sizeof(struct tcphdr
)))
1395 if (th
->doff
< sizeof(struct tcphdr
)/4)
1397 if (!pskb_may_pull(skb
, th
->doff
*4))
1400 if (skb_checksum_init(skb
, IPPROTO_TCP
, ip6_compute_pseudo
))
1404 hdr
= ipv6_hdr(skb
);
1406 sk
= __inet6_lookup_skb(&tcp_hashinfo
, skb
, th
->source
, th
->dest
,
1412 if (sk
->sk_state
== TCP_TIME_WAIT
)
1415 if (hdr
->hop_limit
< inet6_sk(sk
)->min_hopcount
) {
1416 NET_INC_STATS_BH(net
, LINUX_MIB_TCPMINTTLDROP
);
1417 goto discard_and_relse
;
1420 if (!xfrm6_policy_check(sk
, XFRM_POLICY_IN
, skb
))
1421 goto discard_and_relse
;
1423 tcp_v6_fill_cb(skb
, hdr
, th
);
1425 #ifdef CONFIG_TCP_MD5SIG
1426 if (tcp_v6_inbound_md5_hash(sk
, skb
))
1427 goto discard_and_relse
;
1430 if (sk_filter(sk
, skb
))
1431 goto discard_and_relse
;
1433 sk_incoming_cpu_update(sk
);
1436 bh_lock_sock_nested(sk
);
1438 if (!sock_owned_by_user(sk
)) {
1439 if (!tcp_prequeue(sk
, skb
))
1440 ret
= tcp_v6_do_rcv(sk
, skb
);
1441 } else if (unlikely(sk_add_backlog(sk
, skb
,
1442 sk
->sk_rcvbuf
+ sk
->sk_sndbuf
))) {
1444 NET_INC_STATS_BH(net
, LINUX_MIB_TCPBACKLOGDROP
);
1445 goto discard_and_relse
;
1450 return ret
? -1 : 0;
1453 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
))
1456 tcp_v6_fill_cb(skb
, hdr
, th
);
1458 if (skb
->len
< (th
->doff
<<2) || tcp_checksum_complete(skb
)) {
1460 TCP_INC_STATS_BH(net
, TCP_MIB_CSUMERRORS
);
1462 TCP_INC_STATS_BH(net
, TCP_MIB_INERRS
);
1464 tcp_v6_send_reset(NULL
, skb
);
1476 if (!xfrm6_policy_check(NULL
, XFRM_POLICY_IN
, skb
)) {
1477 inet_twsk_put(inet_twsk(sk
));
1481 tcp_v6_fill_cb(skb
, hdr
, th
);
1483 if (skb
->len
< (th
->doff
<<2)) {
1484 inet_twsk_put(inet_twsk(sk
));
1487 if (tcp_checksum_complete(skb
)) {
1488 inet_twsk_put(inet_twsk(sk
));
1492 switch (tcp_timewait_state_process(inet_twsk(sk
), skb
, th
)) {
1497 sk2
= inet6_lookup_listener(dev_net(skb
->dev
), &tcp_hashinfo
,
1498 &ipv6_hdr(skb
)->saddr
, th
->source
,
1499 &ipv6_hdr(skb
)->daddr
,
1500 ntohs(th
->dest
), tcp_v6_iif(skb
));
1502 struct inet_timewait_sock
*tw
= inet_twsk(sk
);
1503 inet_twsk_deschedule(tw
, &tcp_death_row
);
1508 /* Fall through to ACK */
1511 tcp_v6_timewait_ack(sk
, skb
);
1515 case TCP_TW_SUCCESS
:
1521 static void tcp_v6_early_demux(struct sk_buff
*skb
)
1523 const struct ipv6hdr
*hdr
;
1524 const struct tcphdr
*th
;
1527 if (skb
->pkt_type
!= PACKET_HOST
)
1530 if (!pskb_may_pull(skb
, skb_transport_offset(skb
) + sizeof(struct tcphdr
)))
1533 hdr
= ipv6_hdr(skb
);
1536 if (th
->doff
< sizeof(struct tcphdr
) / 4)
1539 /* Note : We use inet6_iif() here, not tcp_v6_iif() */
1540 sk
= __inet6_lookup_established(dev_net(skb
->dev
), &tcp_hashinfo
,
1541 &hdr
->saddr
, th
->source
,
1542 &hdr
->daddr
, ntohs(th
->dest
),
1546 skb
->destructor
= sock_edemux
;
1547 if (sk_fullsock(sk
)) {
1548 struct dst_entry
*dst
= sk
->sk_rx_dst
;
1551 dst
= dst_check(dst
, inet6_sk(sk
)->rx_dst_cookie
);
1553 inet_sk(sk
)->rx_dst_ifindex
== skb
->skb_iif
)
1554 skb_dst_set_noref(skb
, dst
);
1559 static struct timewait_sock_ops tcp6_timewait_sock_ops
= {
1560 .twsk_obj_size
= sizeof(struct tcp6_timewait_sock
),
1561 .twsk_unique
= tcp_twsk_unique
,
1562 .twsk_destructor
= tcp_twsk_destructor
,
1565 static const struct inet_connection_sock_af_ops ipv6_specific
= {
1566 .queue_xmit
= inet6_csk_xmit
,
1567 .send_check
= tcp_v6_send_check
,
1568 .rebuild_header
= inet6_sk_rebuild_header
,
1569 .sk_rx_dst_set
= inet6_sk_rx_dst_set
,
1570 .conn_request
= tcp_v6_conn_request
,
1571 .syn_recv_sock
= tcp_v6_syn_recv_sock
,
1572 .net_header_len
= sizeof(struct ipv6hdr
),
1573 .net_frag_header_len
= sizeof(struct frag_hdr
),
1574 .setsockopt
= ipv6_setsockopt
,
1575 .getsockopt
= ipv6_getsockopt
,
1576 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1577 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1578 .bind_conflict
= inet6_csk_bind_conflict
,
1579 #ifdef CONFIG_COMPAT
1580 .compat_setsockopt
= compat_ipv6_setsockopt
,
1581 .compat_getsockopt
= compat_ipv6_getsockopt
,
1583 .mtu_reduced
= tcp_v6_mtu_reduced
,
1586 #ifdef CONFIG_TCP_MD5SIG
1587 static const struct tcp_sock_af_ops tcp_sock_ipv6_specific
= {
1588 .md5_lookup
= tcp_v6_md5_lookup
,
1589 .calc_md5_hash
= tcp_v6_md5_hash_skb
,
1590 .md5_parse
= tcp_v6_parse_md5_keys
,
1595 * TCP over IPv4 via INET6 API
1597 static const struct inet_connection_sock_af_ops ipv6_mapped
= {
1598 .queue_xmit
= ip_queue_xmit
,
1599 .send_check
= tcp_v4_send_check
,
1600 .rebuild_header
= inet_sk_rebuild_header
,
1601 .sk_rx_dst_set
= inet_sk_rx_dst_set
,
1602 .conn_request
= tcp_v6_conn_request
,
1603 .syn_recv_sock
= tcp_v6_syn_recv_sock
,
1604 .net_header_len
= sizeof(struct iphdr
),
1605 .setsockopt
= ipv6_setsockopt
,
1606 .getsockopt
= ipv6_getsockopt
,
1607 .addr2sockaddr
= inet6_csk_addr2sockaddr
,
1608 .sockaddr_len
= sizeof(struct sockaddr_in6
),
1609 .bind_conflict
= inet6_csk_bind_conflict
,
1610 #ifdef CONFIG_COMPAT
1611 .compat_setsockopt
= compat_ipv6_setsockopt
,
1612 .compat_getsockopt
= compat_ipv6_getsockopt
,
1614 .mtu_reduced
= tcp_v4_mtu_reduced
,
1617 #ifdef CONFIG_TCP_MD5SIG
1618 static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific
= {
1619 .md5_lookup
= tcp_v4_md5_lookup
,
1620 .calc_md5_hash
= tcp_v4_md5_hash_skb
,
1621 .md5_parse
= tcp_v6_parse_md5_keys
,
1625 /* NOTE: A lot of things set to zero explicitly by call to
1626 * sk_alloc() so need not be done here.
1628 static int tcp_v6_init_sock(struct sock
*sk
)
1630 struct inet_connection_sock
*icsk
= inet_csk(sk
);
1634 icsk
->icsk_af_ops
= &ipv6_specific
;
1636 #ifdef CONFIG_TCP_MD5SIG
1637 tcp_sk(sk
)->af_specific
= &tcp_sock_ipv6_specific
;
1643 static void tcp_v6_destroy_sock(struct sock
*sk
)
1645 tcp_v4_destroy_sock(sk
);
1646 inet6_destroy_sock(sk
);
1649 #ifdef CONFIG_PROC_FS
1650 /* Proc filesystem TCPv6 sock list dumping. */
1651 static void get_openreq6(struct seq_file
*seq
,
1652 struct request_sock
*req
, int i
, kuid_t uid
)
1654 long ttd
= req
->rsk_timer
.expires
- jiffies
;
1655 const struct in6_addr
*src
= &inet_rsk(req
)->ir_v6_loc_addr
;
1656 const struct in6_addr
*dest
= &inet_rsk(req
)->ir_v6_rmt_addr
;
1662 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1663 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %d %d %pK\n",
1665 src
->s6_addr32
[0], src
->s6_addr32
[1],
1666 src
->s6_addr32
[2], src
->s6_addr32
[3],
1667 inet_rsk(req
)->ir_num
,
1668 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1669 dest
->s6_addr32
[2], dest
->s6_addr32
[3],
1670 ntohs(inet_rsk(req
)->ir_rmt_port
),
1672 0, 0, /* could print option size, but that is af dependent. */
1673 1, /* timers active (only the expire timer) */
1674 jiffies_to_clock_t(ttd
),
1676 from_kuid_munged(seq_user_ns(seq
), uid
),
1677 0, /* non standard timer */
1678 0, /* open_requests have no inode */
1682 static void get_tcp6_sock(struct seq_file
*seq
, struct sock
*sp
, int i
)
1684 const struct in6_addr
*dest
, *src
;
1687 unsigned long timer_expires
;
1688 const struct inet_sock
*inet
= inet_sk(sp
);
1689 const struct tcp_sock
*tp
= tcp_sk(sp
);
1690 const struct inet_connection_sock
*icsk
= inet_csk(sp
);
1691 struct fastopen_queue
*fastopenq
= icsk
->icsk_accept_queue
.fastopenq
;
1693 dest
= &sp
->sk_v6_daddr
;
1694 src
= &sp
->sk_v6_rcv_saddr
;
1695 destp
= ntohs(inet
->inet_dport
);
1696 srcp
= ntohs(inet
->inet_sport
);
1698 if (icsk
->icsk_pending
== ICSK_TIME_RETRANS
) {
1700 timer_expires
= icsk
->icsk_timeout
;
1701 } else if (icsk
->icsk_pending
== ICSK_TIME_PROBE0
) {
1703 timer_expires
= icsk
->icsk_timeout
;
1704 } else if (timer_pending(&sp
->sk_timer
)) {
1706 timer_expires
= sp
->sk_timer
.expires
;
1709 timer_expires
= jiffies
;
1713 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1714 "%02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %lu %lu %u %u %d\n",
1716 src
->s6_addr32
[0], src
->s6_addr32
[1],
1717 src
->s6_addr32
[2], src
->s6_addr32
[3], srcp
,
1718 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1719 dest
->s6_addr32
[2], dest
->s6_addr32
[3], destp
,
1721 tp
->write_seq
-tp
->snd_una
,
1722 (sp
->sk_state
== TCP_LISTEN
) ? sp
->sk_ack_backlog
: (tp
->rcv_nxt
- tp
->copied_seq
),
1724 jiffies_delta_to_clock_t(timer_expires
- jiffies
),
1725 icsk
->icsk_retransmits
,
1726 from_kuid_munged(seq_user_ns(seq
), sock_i_uid(sp
)),
1727 icsk
->icsk_probes_out
,
1729 atomic_read(&sp
->sk_refcnt
), sp
,
1730 jiffies_to_clock_t(icsk
->icsk_rto
),
1731 jiffies_to_clock_t(icsk
->icsk_ack
.ato
),
1732 (icsk
->icsk_ack
.quick
<< 1) | icsk
->icsk_ack
.pingpong
,
1734 sp
->sk_state
== TCP_LISTEN
?
1735 (fastopenq
? fastopenq
->max_qlen
: 0) :
1736 (tcp_in_initial_slowstart(tp
) ? -1 : tp
->snd_ssthresh
)
1740 static void get_timewait6_sock(struct seq_file
*seq
,
1741 struct inet_timewait_sock
*tw
, int i
)
1743 const struct in6_addr
*dest
, *src
;
1745 s32 delta
= tw
->tw_ttd
- inet_tw_time_stamp();
1747 dest
= &tw
->tw_v6_daddr
;
1748 src
= &tw
->tw_v6_rcv_saddr
;
1749 destp
= ntohs(tw
->tw_dport
);
1750 srcp
= ntohs(tw
->tw_sport
);
1753 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1754 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %d %d %pK\n",
1756 src
->s6_addr32
[0], src
->s6_addr32
[1],
1757 src
->s6_addr32
[2], src
->s6_addr32
[3], srcp
,
1758 dest
->s6_addr32
[0], dest
->s6_addr32
[1],
1759 dest
->s6_addr32
[2], dest
->s6_addr32
[3], destp
,
1760 tw
->tw_substate
, 0, 0,
1761 3, jiffies_delta_to_clock_t(delta
), 0, 0, 0, 0,
1762 atomic_read(&tw
->tw_refcnt
), tw
);
1765 static int tcp6_seq_show(struct seq_file
*seq
, void *v
)
1767 struct tcp_iter_state
*st
;
1768 struct sock
*sk
= v
;
1770 if (v
== SEQ_START_TOKEN
) {
1775 "st tx_queue rx_queue tr tm->when retrnsmt"
1776 " uid timeout inode\n");
1781 switch (st
->state
) {
1782 case TCP_SEQ_STATE_LISTENING
:
1783 case TCP_SEQ_STATE_ESTABLISHED
:
1784 if (sk
->sk_state
== TCP_TIME_WAIT
)
1785 get_timewait6_sock(seq
, v
, st
->num
);
1787 get_tcp6_sock(seq
, v
, st
->num
);
1789 case TCP_SEQ_STATE_OPENREQ
:
1790 get_openreq6(seq
, v
, st
->num
, st
->uid
);
1797 static const struct file_operations tcp6_afinfo_seq_fops
= {
1798 .owner
= THIS_MODULE
,
1799 .open
= tcp_seq_open
,
1801 .llseek
= seq_lseek
,
1802 .release
= seq_release_net
1805 static struct tcp_seq_afinfo tcp6_seq_afinfo
= {
1808 .seq_fops
= &tcp6_afinfo_seq_fops
,
1810 .show
= tcp6_seq_show
,
1814 int __net_init
tcp6_proc_init(struct net
*net
)
1816 return tcp_proc_register(net
, &tcp6_seq_afinfo
);
1819 void tcp6_proc_exit(struct net
*net
)
1821 tcp_proc_unregister(net
, &tcp6_seq_afinfo
);
1825 static void tcp_v6_clear_sk(struct sock
*sk
, int size
)
1827 struct inet_sock
*inet
= inet_sk(sk
);
1829 /* we do not want to clear pinet6 field, because of RCU lookups */
1830 sk_prot_clear_nulls(sk
, offsetof(struct inet_sock
, pinet6
));
1832 size
-= offsetof(struct inet_sock
, pinet6
) + sizeof(inet
->pinet6
);
1833 memset(&inet
->pinet6
+ 1, 0, size
);
1836 struct proto tcpv6_prot
= {
1838 .owner
= THIS_MODULE
,
1840 .connect
= tcp_v6_connect
,
1841 .disconnect
= tcp_disconnect
,
1842 .accept
= inet_csk_accept
,
1844 .init
= tcp_v6_init_sock
,
1845 .destroy
= tcp_v6_destroy_sock
,
1846 .shutdown
= tcp_shutdown
,
1847 .setsockopt
= tcp_setsockopt
,
1848 .getsockopt
= tcp_getsockopt
,
1849 .recvmsg
= tcp_recvmsg
,
1850 .sendmsg
= tcp_sendmsg
,
1851 .sendpage
= tcp_sendpage
,
1852 .backlog_rcv
= tcp_v6_do_rcv
,
1853 .release_cb
= tcp_release_cb
,
1855 .unhash
= inet_unhash
,
1856 .get_port
= inet_csk_get_port
,
1857 .enter_memory_pressure
= tcp_enter_memory_pressure
,
1858 .stream_memory_free
= tcp_stream_memory_free
,
1859 .sockets_allocated
= &tcp_sockets_allocated
,
1860 .memory_allocated
= &tcp_memory_allocated
,
1861 .memory_pressure
= &tcp_memory_pressure
,
1862 .orphan_count
= &tcp_orphan_count
,
1863 .sysctl_mem
= sysctl_tcp_mem
,
1864 .sysctl_wmem
= sysctl_tcp_wmem
,
1865 .sysctl_rmem
= sysctl_tcp_rmem
,
1866 .max_header
= MAX_TCP_HEADER
,
1867 .obj_size
= sizeof(struct tcp6_sock
),
1868 .slab_flags
= SLAB_DESTROY_BY_RCU
,
1869 .twsk_prot
= &tcp6_timewait_sock_ops
,
1870 .rsk_prot
= &tcp6_request_sock_ops
,
1871 .h
.hashinfo
= &tcp_hashinfo
,
1872 .no_autobind
= true,
1873 #ifdef CONFIG_COMPAT
1874 .compat_setsockopt
= compat_tcp_setsockopt
,
1875 .compat_getsockopt
= compat_tcp_getsockopt
,
1877 #ifdef CONFIG_MEMCG_KMEM
1878 .proto_cgroup
= tcp_proto_cgroup
,
1880 .clear_sk
= tcp_v6_clear_sk
,
1883 static const struct inet6_protocol tcpv6_protocol
= {
1884 .early_demux
= tcp_v6_early_demux
,
1885 .handler
= tcp_v6_rcv
,
1886 .err_handler
= tcp_v6_err
,
1887 .flags
= INET6_PROTO_NOPOLICY
|INET6_PROTO_FINAL
,
1890 static struct inet_protosw tcpv6_protosw
= {
1891 .type
= SOCK_STREAM
,
1892 .protocol
= IPPROTO_TCP
,
1893 .prot
= &tcpv6_prot
,
1894 .ops
= &inet6_stream_ops
,
1895 .flags
= INET_PROTOSW_PERMANENT
|
1899 static int __net_init
tcpv6_net_init(struct net
*net
)
1901 return inet_ctl_sock_create(&net
->ipv6
.tcp_sk
, PF_INET6
,
1902 SOCK_RAW
, IPPROTO_TCP
, net
);
1905 static void __net_exit
tcpv6_net_exit(struct net
*net
)
1907 inet_ctl_sock_destroy(net
->ipv6
.tcp_sk
);
1910 static void __net_exit
tcpv6_net_exit_batch(struct list_head
*net_exit_list
)
1912 inet_twsk_purge(&tcp_hashinfo
, &tcp_death_row
, AF_INET6
);
1915 static struct pernet_operations tcpv6_net_ops
= {
1916 .init
= tcpv6_net_init
,
1917 .exit
= tcpv6_net_exit
,
1918 .exit_batch
= tcpv6_net_exit_batch
,
1921 int __init
tcpv6_init(void)
1925 ret
= inet6_add_protocol(&tcpv6_protocol
, IPPROTO_TCP
);
1929 /* register inet6 protocol */
1930 ret
= inet6_register_protosw(&tcpv6_protosw
);
1932 goto out_tcpv6_protocol
;
1934 ret
= register_pernet_subsys(&tcpv6_net_ops
);
1936 goto out_tcpv6_protosw
;
1941 inet6_unregister_protosw(&tcpv6_protosw
);
1943 inet6_del_protocol(&tcpv6_protocol
, IPPROTO_TCP
);
1947 void tcpv6_exit(void)
1949 unregister_pernet_subsys(&tcpv6_net_ops
);
1950 inet6_unregister_protosw(&tcpv6_protosw
);
1951 inet6_del_protocol(&tcpv6_protocol
, IPPROTO_TCP
);