3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * $Id: datagram.c,v 1.24 2002/02/01 22:01:04 davem Exp $
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 #include <linux/capability.h>
17 #include <linux/errno.h>
18 #include <linux/types.h>
19 #include <linux/kernel.h>
20 #include <linux/interrupt.h>
21 #include <linux/socket.h>
22 #include <linux/sockios.h>
23 #include <linux/in6.h>
24 #include <linux/ipv6.h>
25 #include <linux/route.h>
28 #include <net/ndisc.h>
29 #include <net/addrconf.h>
30 #include <net/transp_v6.h>
31 #include <net/ip6_route.h>
32 #include <net/tcp_states.h>
34 #include <linux/errqueue.h>
35 #include <asm/uaccess.h>
37 int ip6_datagram_connect(struct sock
*sk
, struct sockaddr
*uaddr
, int addr_len
)
39 struct sockaddr_in6
*usin
= (struct sockaddr_in6
*) uaddr
;
40 struct inet_sock
*inet
= inet_sk(sk
);
41 struct ipv6_pinfo
*np
= inet6_sk(sk
);
42 struct in6_addr
*daddr
, *final_p
= NULL
, final
;
43 struct dst_entry
*dst
;
45 struct ip6_flowlabel
*flowlabel
= NULL
;
49 if (usin
->sin6_family
== AF_INET
) {
50 if (__ipv6_only_sock(sk
))
52 err
= ip4_datagram_connect(sk
, uaddr
, addr_len
);
56 if (addr_len
< SIN6_LEN_RFC2133
)
59 if (usin
->sin6_family
!= AF_INET6
)
62 memset(&fl
, 0, sizeof(fl
));
64 fl
.fl6_flowlabel
= usin
->sin6_flowinfo
&IPV6_FLOWINFO_MASK
;
65 if (fl
.fl6_flowlabel
&IPV6_FLOWLABEL_MASK
) {
66 flowlabel
= fl6_sock_lookup(sk
, fl
.fl6_flowlabel
);
67 if (flowlabel
== NULL
)
69 ipv6_addr_copy(&usin
->sin6_addr
, &flowlabel
->dst
);
73 addr_type
= ipv6_addr_type(&usin
->sin6_addr
);
75 if (addr_type
== IPV6_ADDR_ANY
) {
79 usin
->sin6_addr
.s6_addr
[15] = 0x01;
82 daddr
= &usin
->sin6_addr
;
84 if (addr_type
== IPV6_ADDR_MAPPED
) {
85 struct sockaddr_in sin
;
87 if (__ipv6_only_sock(sk
)) {
91 sin
.sin_family
= AF_INET
;
92 sin
.sin_addr
.s_addr
= daddr
->s6_addr32
[3];
93 sin
.sin_port
= usin
->sin6_port
;
95 err
= ip4_datagram_connect(sk
,
96 (struct sockaddr
*) &sin
,
103 ipv6_addr_set(&np
->daddr
, 0, 0, htonl(0x0000ffff), inet
->daddr
);
105 if (ipv6_addr_any(&np
->saddr
)) {
106 ipv6_addr_set(&np
->saddr
, 0, 0, htonl(0x0000ffff),
110 if (ipv6_addr_any(&np
->rcv_saddr
)) {
111 ipv6_addr_set(&np
->rcv_saddr
, 0, 0, htonl(0x0000ffff),
117 if (addr_type
&IPV6_ADDR_LINKLOCAL
) {
118 if (addr_len
>= sizeof(struct sockaddr_in6
) &&
119 usin
->sin6_scope_id
) {
120 if (sk
->sk_bound_dev_if
&&
121 sk
->sk_bound_dev_if
!= usin
->sin6_scope_id
) {
125 sk
->sk_bound_dev_if
= usin
->sin6_scope_id
;
128 if (!sk
->sk_bound_dev_if
&& (addr_type
& IPV6_ADDR_MULTICAST
))
129 sk
->sk_bound_dev_if
= np
->mcast_oif
;
131 /* Connect to link-local address requires an interface */
132 if (!sk
->sk_bound_dev_if
) {
138 ipv6_addr_copy(&np
->daddr
, daddr
);
139 np
->flow_label
= fl
.fl6_flowlabel
;
141 inet
->dport
= usin
->sin6_port
;
144 * Check for a route to destination an obtain the
145 * destination cache for it.
148 fl
.proto
= sk
->sk_protocol
;
149 ipv6_addr_copy(&fl
.fl6_dst
, &np
->daddr
);
150 ipv6_addr_copy(&fl
.fl6_src
, &np
->saddr
);
151 fl
.oif
= sk
->sk_bound_dev_if
;
152 fl
.fl_ip_dport
= inet
->dport
;
153 fl
.fl_ip_sport
= inet
->sport
;
155 if (!fl
.oif
&& (addr_type
&IPV6_ADDR_MULTICAST
))
156 fl
.oif
= np
->mcast_oif
;
158 security_sk_classify_flow(sk
, &fl
);
161 if (flowlabel
->opt
&& flowlabel
->opt
->srcrt
) {
162 struct rt0_hdr
*rt0
= (struct rt0_hdr
*) flowlabel
->opt
->srcrt
;
163 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
164 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
167 } else if (np
->opt
&& np
->opt
->srcrt
) {
168 struct rt0_hdr
*rt0
= (struct rt0_hdr
*)np
->opt
->srcrt
;
169 ipv6_addr_copy(&final
, &fl
.fl6_dst
);
170 ipv6_addr_copy(&fl
.fl6_dst
, rt0
->addr
);
174 err
= ip6_dst_lookup(sk
, &dst
, &fl
);
178 ipv6_addr_copy(&fl
.fl6_dst
, final_p
);
180 if ((err
= __xfrm_lookup(&dst
, &fl
, sk
, 1)) < 0) {
182 err
= ip6_dst_blackhole(sk
, &dst
, &fl
);
187 /* source address lookup done in ip6_dst_lookup */
189 if (ipv6_addr_any(&np
->saddr
))
190 ipv6_addr_copy(&np
->saddr
, &fl
.fl6_src
);
192 if (ipv6_addr_any(&np
->rcv_saddr
)) {
193 ipv6_addr_copy(&np
->rcv_saddr
, &fl
.fl6_src
);
194 inet
->rcv_saddr
= LOOPBACK4_IPV6
;
197 ip6_dst_store(sk
, dst
,
198 ipv6_addr_equal(&fl
.fl6_dst
, &np
->daddr
) ?
200 #ifdef CONFIG_IPV6_SUBTREES
201 ipv6_addr_equal(&fl
.fl6_src
, &np
->saddr
) ?
206 sk
->sk_state
= TCP_ESTABLISHED
;
208 fl6_sock_release(flowlabel
);
212 void ipv6_icmp_error(struct sock
*sk
, struct sk_buff
*skb
, int err
,
213 __be16 port
, u32 info
, u8
*payload
)
215 struct ipv6_pinfo
*np
= inet6_sk(sk
);
216 struct icmp6hdr
*icmph
= icmp6_hdr(skb
);
217 struct sock_exterr_skb
*serr
;
222 skb
= skb_clone(skb
, GFP_ATOMIC
);
226 serr
= SKB_EXT_ERR(skb
);
227 serr
->ee
.ee_errno
= err
;
228 serr
->ee
.ee_origin
= SO_EE_ORIGIN_ICMP6
;
229 serr
->ee
.ee_type
= icmph
->icmp6_type
;
230 serr
->ee
.ee_code
= icmph
->icmp6_code
;
232 serr
->ee
.ee_info
= info
;
233 serr
->ee
.ee_data
= 0;
234 serr
->addr_offset
= (u8
*)&(((struct ipv6hdr
*)(icmph
+ 1))->daddr
) -
235 skb_network_header(skb
);
238 __skb_pull(skb
, payload
- skb
->data
);
239 skb_reset_transport_header(skb
);
241 if (sock_queue_err_skb(sk
, skb
))
245 void ipv6_local_error(struct sock
*sk
, int err
, struct flowi
*fl
, u32 info
)
247 struct ipv6_pinfo
*np
= inet6_sk(sk
);
248 struct sock_exterr_skb
*serr
;
255 skb
= alloc_skb(sizeof(struct ipv6hdr
), GFP_ATOMIC
);
259 skb_put(skb
, sizeof(struct ipv6hdr
));
260 skb_reset_network_header(skb
);
262 ipv6_addr_copy(&iph
->daddr
, &fl
->fl6_dst
);
264 serr
= SKB_EXT_ERR(skb
);
265 serr
->ee
.ee_errno
= err
;
266 serr
->ee
.ee_origin
= SO_EE_ORIGIN_LOCAL
;
267 serr
->ee
.ee_type
= 0;
268 serr
->ee
.ee_code
= 0;
270 serr
->ee
.ee_info
= info
;
271 serr
->ee
.ee_data
= 0;
272 serr
->addr_offset
= (u8
*)&iph
->daddr
- skb_network_header(skb
);
273 serr
->port
= fl
->fl_ip_dport
;
275 __skb_pull(skb
, skb_tail_pointer(skb
) - skb
->data
);
276 skb_reset_transport_header(skb
);
278 if (sock_queue_err_skb(sk
, skb
))
283 * Handle MSG_ERRQUEUE
285 int ipv6_recv_error(struct sock
*sk
, struct msghdr
*msg
, int len
)
287 struct ipv6_pinfo
*np
= inet6_sk(sk
);
288 struct sock_exterr_skb
*serr
;
289 struct sk_buff
*skb
, *skb2
;
290 struct sockaddr_in6
*sin
;
292 struct sock_extended_err ee
;
293 struct sockaddr_in6 offender
;
299 skb
= skb_dequeue(&sk
->sk_error_queue
);
305 msg
->msg_flags
|= MSG_TRUNC
;
308 err
= skb_copy_datagram_iovec(skb
, 0, msg
->msg_iov
, copied
);
312 sock_recv_timestamp(msg
, sk
, skb
);
314 serr
= SKB_EXT_ERR(skb
);
316 sin
= (struct sockaddr_in6
*)msg
->msg_name
;
318 const unsigned char *nh
= skb_network_header(skb
);
319 sin
->sin6_family
= AF_INET6
;
320 sin
->sin6_flowinfo
= 0;
321 sin
->sin6_port
= serr
->port
;
322 sin
->sin6_scope_id
= 0;
323 if (serr
->ee
.ee_origin
== SO_EE_ORIGIN_ICMP6
) {
324 ipv6_addr_copy(&sin
->sin6_addr
,
325 (struct in6_addr
*)(nh
+ serr
->addr_offset
));
328 (*(__be32
*)(nh
+ serr
->addr_offset
- 24) &
330 if (ipv6_addr_type(&sin
->sin6_addr
) & IPV6_ADDR_LINKLOCAL
)
331 sin
->sin6_scope_id
= IP6CB(skb
)->iif
;
333 ipv6_addr_set(&sin
->sin6_addr
, 0, 0,
335 *(__be32
*)(nh
+ serr
->addr_offset
));
339 memcpy(&errhdr
.ee
, &serr
->ee
, sizeof(struct sock_extended_err
));
340 sin
= &errhdr
.offender
;
341 sin
->sin6_family
= AF_UNSPEC
;
342 if (serr
->ee
.ee_origin
!= SO_EE_ORIGIN_LOCAL
) {
343 sin
->sin6_family
= AF_INET6
;
344 sin
->sin6_flowinfo
= 0;
345 sin
->sin6_scope_id
= 0;
346 if (serr
->ee
.ee_origin
== SO_EE_ORIGIN_ICMP6
) {
347 ipv6_addr_copy(&sin
->sin6_addr
, &ipv6_hdr(skb
)->saddr
);
349 datagram_recv_ctl(sk
, msg
, skb
);
350 if (ipv6_addr_type(&sin
->sin6_addr
) & IPV6_ADDR_LINKLOCAL
)
351 sin
->sin6_scope_id
= IP6CB(skb
)->iif
;
353 struct inet_sock
*inet
= inet_sk(sk
);
355 ipv6_addr_set(&sin
->sin6_addr
, 0, 0,
356 htonl(0xffff), ip_hdr(skb
)->saddr
);
357 if (inet
->cmsg_flags
)
358 ip_cmsg_recv(msg
, skb
);
362 put_cmsg(msg
, SOL_IPV6
, IPV6_RECVERR
, sizeof(errhdr
), &errhdr
);
364 /* Now we could try to dump offended packet options */
366 msg
->msg_flags
|= MSG_ERRQUEUE
;
369 /* Reset and regenerate socket error */
370 spin_lock_bh(&sk
->sk_error_queue
.lock
);
372 if ((skb2
= skb_peek(&sk
->sk_error_queue
)) != NULL
) {
373 sk
->sk_err
= SKB_EXT_ERR(skb2
)->ee
.ee_errno
;
374 spin_unlock_bh(&sk
->sk_error_queue
.lock
);
375 sk
->sk_error_report(sk
);
377 spin_unlock_bh(&sk
->sk_error_queue
.lock
);
388 int datagram_recv_ctl(struct sock
*sk
, struct msghdr
*msg
, struct sk_buff
*skb
)
390 struct ipv6_pinfo
*np
= inet6_sk(sk
);
391 struct inet6_skb_parm
*opt
= IP6CB(skb
);
392 unsigned char *nh
= skb_network_header(skb
);
394 if (np
->rxopt
.bits
.rxinfo
) {
395 struct in6_pktinfo src_info
;
397 src_info
.ipi6_ifindex
= opt
->iif
;
398 ipv6_addr_copy(&src_info
.ipi6_addr
, &ipv6_hdr(skb
)->daddr
);
399 put_cmsg(msg
, SOL_IPV6
, IPV6_PKTINFO
, sizeof(src_info
), &src_info
);
402 if (np
->rxopt
.bits
.rxhlim
) {
403 int hlim
= ipv6_hdr(skb
)->hop_limit
;
404 put_cmsg(msg
, SOL_IPV6
, IPV6_HOPLIMIT
, sizeof(hlim
), &hlim
);
407 if (np
->rxopt
.bits
.rxtclass
) {
408 int tclass
= (ntohl(*(__be32
*)ipv6_hdr(skb
)) >> 20) & 0xff;
409 put_cmsg(msg
, SOL_IPV6
, IPV6_TCLASS
, sizeof(tclass
), &tclass
);
412 if (np
->rxopt
.bits
.rxflow
&& (*(__be32
*)nh
& IPV6_FLOWINFO_MASK
)) {
413 __be32 flowinfo
= *(__be32
*)nh
& IPV6_FLOWINFO_MASK
;
414 put_cmsg(msg
, SOL_IPV6
, IPV6_FLOWINFO
, sizeof(flowinfo
), &flowinfo
);
417 /* HbH is allowed only once */
418 if (np
->rxopt
.bits
.hopopts
&& opt
->hop
) {
419 u8
*ptr
= nh
+ opt
->hop
;
420 put_cmsg(msg
, SOL_IPV6
, IPV6_HOPOPTS
, (ptr
[1]+1)<<3, ptr
);
424 (np
->rxopt
.bits
.dstopts
|| np
->rxopt
.bits
.srcrt
)) {
426 * Silly enough, but we need to reparse in order to
427 * report extension headers (except for HbH)
430 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
431 * (and WILL NOT be) defined because
432 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
434 unsigned int off
= sizeof(struct ipv6hdr
);
435 u8 nexthdr
= ipv6_hdr(skb
)->nexthdr
;
437 while (off
<= opt
->lastopt
) {
442 case IPPROTO_DSTOPTS
:
444 len
= (ptr
[1] + 1) << 3;
445 if (np
->rxopt
.bits
.dstopts
)
446 put_cmsg(msg
, SOL_IPV6
, IPV6_DSTOPTS
, len
, ptr
);
448 case IPPROTO_ROUTING
:
450 len
= (ptr
[1] + 1) << 3;
451 if (np
->rxopt
.bits
.srcrt
)
452 put_cmsg(msg
, SOL_IPV6
, IPV6_RTHDR
, len
, ptr
);
456 len
= (ptr
[1] + 2) << 2;
460 len
= (ptr
[1] + 1) << 3;
468 /* socket options in old style */
469 if (np
->rxopt
.bits
.rxoinfo
) {
470 struct in6_pktinfo src_info
;
472 src_info
.ipi6_ifindex
= opt
->iif
;
473 ipv6_addr_copy(&src_info
.ipi6_addr
, &ipv6_hdr(skb
)->daddr
);
474 put_cmsg(msg
, SOL_IPV6
, IPV6_2292PKTINFO
, sizeof(src_info
), &src_info
);
476 if (np
->rxopt
.bits
.rxohlim
) {
477 int hlim
= ipv6_hdr(skb
)->hop_limit
;
478 put_cmsg(msg
, SOL_IPV6
, IPV6_2292HOPLIMIT
, sizeof(hlim
), &hlim
);
480 if (np
->rxopt
.bits
.ohopopts
&& opt
->hop
) {
481 u8
*ptr
= nh
+ opt
->hop
;
482 put_cmsg(msg
, SOL_IPV6
, IPV6_2292HOPOPTS
, (ptr
[1]+1)<<3, ptr
);
484 if (np
->rxopt
.bits
.odstopts
&& opt
->dst0
) {
485 u8
*ptr
= nh
+ opt
->dst0
;
486 put_cmsg(msg
, SOL_IPV6
, IPV6_2292DSTOPTS
, (ptr
[1]+1)<<3, ptr
);
488 if (np
->rxopt
.bits
.osrcrt
&& opt
->srcrt
) {
489 struct ipv6_rt_hdr
*rthdr
= (struct ipv6_rt_hdr
*)(nh
+ opt
->srcrt
);
490 put_cmsg(msg
, SOL_IPV6
, IPV6_2292RTHDR
, (rthdr
->hdrlen
+1) << 3, rthdr
);
492 if (np
->rxopt
.bits
.odstopts
&& opt
->dst1
) {
493 u8
*ptr
= nh
+ opt
->dst1
;
494 put_cmsg(msg
, SOL_IPV6
, IPV6_2292DSTOPTS
, (ptr
[1]+1)<<3, ptr
);
499 int datagram_send_ctl(struct msghdr
*msg
, struct flowi
*fl
,
500 struct ipv6_txoptions
*opt
,
501 int *hlimit
, int *tclass
)
503 struct in6_pktinfo
*src_info
;
504 struct cmsghdr
*cmsg
;
505 struct ipv6_rt_hdr
*rthdr
;
506 struct ipv6_opt_hdr
*hdr
;
510 for (cmsg
= CMSG_FIRSTHDR(msg
); cmsg
; cmsg
= CMSG_NXTHDR(msg
, cmsg
)) {
512 struct net_device
*dev
= NULL
;
514 if (!CMSG_OK(msg
, cmsg
)) {
519 if (cmsg
->cmsg_level
!= SOL_IPV6
)
522 switch (cmsg
->cmsg_type
) {
524 case IPV6_2292PKTINFO
:
525 if (cmsg
->cmsg_len
< CMSG_LEN(sizeof(struct in6_pktinfo
))) {
530 src_info
= (struct in6_pktinfo
*)CMSG_DATA(cmsg
);
532 if (src_info
->ipi6_ifindex
) {
533 if (fl
->oif
&& src_info
->ipi6_ifindex
!= fl
->oif
)
535 fl
->oif
= src_info
->ipi6_ifindex
;
538 addr_type
= ipv6_addr_type(&src_info
->ipi6_addr
);
540 if (addr_type
== IPV6_ADDR_ANY
)
543 if (addr_type
& IPV6_ADDR_LINKLOCAL
) {
544 if (!src_info
->ipi6_ifindex
)
547 dev
= dev_get_by_index(&init_net
, src_info
->ipi6_ifindex
);
552 if (!ipv6_chk_addr(&src_info
->ipi6_addr
, dev
, 0)) {
561 ipv6_addr_copy(&fl
->fl6_src
, &src_info
->ipi6_addr
);
565 if (cmsg
->cmsg_len
< CMSG_LEN(4)) {
570 if (fl
->fl6_flowlabel
&IPV6_FLOWINFO_MASK
) {
571 if ((fl
->fl6_flowlabel
^*(__be32
*)CMSG_DATA(cmsg
))&~IPV6_FLOWINFO_MASK
) {
576 fl
->fl6_flowlabel
= IPV6_FLOWINFO_MASK
& *(__be32
*)CMSG_DATA(cmsg
);
579 case IPV6_2292HOPOPTS
:
581 if (opt
->hopopt
|| cmsg
->cmsg_len
< CMSG_LEN(sizeof(struct ipv6_opt_hdr
))) {
586 hdr
= (struct ipv6_opt_hdr
*)CMSG_DATA(cmsg
);
587 len
= ((hdr
->hdrlen
+ 1) << 3);
588 if (cmsg
->cmsg_len
< CMSG_LEN(len
)) {
592 if (!capable(CAP_NET_RAW
)) {
596 opt
->opt_nflen
+= len
;
600 case IPV6_2292DSTOPTS
:
601 if (cmsg
->cmsg_len
< CMSG_LEN(sizeof(struct ipv6_opt_hdr
))) {
606 hdr
= (struct ipv6_opt_hdr
*)CMSG_DATA(cmsg
);
607 len
= ((hdr
->hdrlen
+ 1) << 3);
608 if (cmsg
->cmsg_len
< CMSG_LEN(len
)) {
612 if (!capable(CAP_NET_RAW
)) {
620 opt
->opt_flen
+= len
;
625 case IPV6_RTHDRDSTOPTS
:
626 if (cmsg
->cmsg_len
< CMSG_LEN(sizeof(struct ipv6_opt_hdr
))) {
631 hdr
= (struct ipv6_opt_hdr
*)CMSG_DATA(cmsg
);
632 len
= ((hdr
->hdrlen
+ 1) << 3);
633 if (cmsg
->cmsg_len
< CMSG_LEN(len
)) {
637 if (!capable(CAP_NET_RAW
)) {
641 if (cmsg
->cmsg_type
== IPV6_DSTOPTS
) {
642 opt
->opt_flen
+= len
;
645 opt
->opt_nflen
+= len
;
652 if (cmsg
->cmsg_len
< CMSG_LEN(sizeof(struct ipv6_rt_hdr
))) {
657 rthdr
= (struct ipv6_rt_hdr
*)CMSG_DATA(cmsg
);
659 switch (rthdr
->type
) {
660 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
661 case IPV6_SRCRT_TYPE_2
:
669 len
= ((rthdr
->hdrlen
+ 1) << 3);
671 if (cmsg
->cmsg_len
< CMSG_LEN(len
)) {
676 /* segments left must also match */
677 if ((rthdr
->hdrlen
>> 1) != rthdr
->segments_left
) {
682 opt
->opt_nflen
+= len
;
685 if (cmsg
->cmsg_type
== IPV6_2292RTHDR
&& opt
->dst1opt
) {
686 int dsthdrlen
= ((opt
->dst1opt
->hdrlen
+1)<<3);
688 opt
->opt_nflen
+= dsthdrlen
;
689 opt
->dst0opt
= opt
->dst1opt
;
691 opt
->opt_flen
-= dsthdrlen
;
696 case IPV6_2292HOPLIMIT
:
698 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(int))) {
703 *hlimit
= *(int *)CMSG_DATA(cmsg
);
711 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(int))) {
715 tc
= *(int *)CMSG_DATA(cmsg
);
716 if (tc
< -1 || tc
> 0xff)
725 LIMIT_NETDEBUG(KERN_DEBUG
"invalid cmsg type: %d\n",