3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
14 #include <linux/capability.h>
15 #include <linux/errno.h>
16 #include <linux/types.h>
17 #include <linux/kernel.h>
18 #include <linux/interrupt.h>
19 #include <linux/socket.h>
20 #include <linux/sockios.h>
21 #include <linux/in6.h>
22 #include <linux/ipv6.h>
23 #include <linux/route.h>
26 #include <net/ndisc.h>
27 #include <net/addrconf.h>
28 #include <net/transp_v6.h>
29 #include <net/ip6_route.h>
30 #include <net/tcp_states.h>
32 #include <linux/errqueue.h>
33 #include <asm/uaccess.h>
35 int ip6_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
37 struct sockaddr_in6 *usin = (struct sockaddr_in6 *) uaddr;
38 struct inet_sock *inet = inet_sk(sk);
39 struct ipv6_pinfo *np = inet6_sk(sk);
40 struct in6_addr *daddr, *final_p = NULL, final;
41 struct dst_entry *dst;
43 struct ip6_flowlabel *flowlabel = NULL;
47 if (usin->sin6_family == AF_INET) {
48 if (__ipv6_only_sock(sk))
50 err = ip4_datagram_connect(sk, uaddr, addr_len);
54 if (addr_len < SIN6_LEN_RFC2133)
57 if (usin->sin6_family != AF_INET6)
60 memset(&fl, 0, sizeof(fl));
62 fl.fl6_flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
63 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
64 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
65 if (flowlabel == NULL)
67 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
71 addr_type = ipv6_addr_type(&usin->sin6_addr);
73 if (addr_type == IPV6_ADDR_ANY) {
77 usin->sin6_addr.s6_addr[15] = 0x01;
80 daddr = &usin->sin6_addr;
82 if (addr_type == IPV6_ADDR_MAPPED) {
83 struct sockaddr_in sin;
85 if (__ipv6_only_sock(sk)) {
89 sin.sin_family = AF_INET;
90 sin.sin_addr.s_addr = daddr->s6_addr32[3];
91 sin.sin_port = usin->sin6_port;
93 err = ip4_datagram_connect(sk,
94 (struct sockaddr*) &sin,
101 ipv6_addr_set(&np->daddr, 0, 0, htonl(0x0000ffff), inet->daddr);
103 if (ipv6_addr_any(&np->saddr)) {
104 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000ffff),
108 if (ipv6_addr_any(&np->rcv_saddr)) {
109 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000ffff),
115 if (addr_type&IPV6_ADDR_LINKLOCAL) {
116 if (addr_len >= sizeof(struct sockaddr_in6) &&
117 usin->sin6_scope_id) {
118 if (sk->sk_bound_dev_if &&
119 sk->sk_bound_dev_if != usin->sin6_scope_id) {
123 sk->sk_bound_dev_if = usin->sin6_scope_id;
126 if (!sk->sk_bound_dev_if && (addr_type & IPV6_ADDR_MULTICAST))
127 sk->sk_bound_dev_if = np->mcast_oif;
129 /* Connect to link-local address requires an interface */
130 if (!sk->sk_bound_dev_if) {
136 ipv6_addr_copy(&np->daddr, daddr);
137 np->flow_label = fl.fl6_flowlabel;
139 inet->dport = usin->sin6_port;
142 * Check for a route to destination an obtain the
143 * destination cache for it.
146 fl.proto = sk->sk_protocol;
147 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
148 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
149 fl.oif = sk->sk_bound_dev_if;
150 fl.fl_ip_dport = inet->dport;
151 fl.fl_ip_sport = inet->sport;
153 if (!fl.oif && (addr_type&IPV6_ADDR_MULTICAST))
154 fl.oif = np->mcast_oif;
156 security_sk_classify_flow(sk, &fl);
159 if (flowlabel->opt && flowlabel->opt->srcrt) {
160 struct rt0_hdr *rt0 = (struct rt0_hdr *) flowlabel->opt->srcrt;
161 ipv6_addr_copy(&final, &fl.fl6_dst);
162 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
165 } else if (np->opt && np->opt->srcrt) {
166 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
167 ipv6_addr_copy(&final, &fl.fl6_dst);
168 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
172 err = ip6_dst_lookup(sk, &dst, &fl);
176 ipv6_addr_copy(&fl.fl6_dst, final_p);
178 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
181 err = ip6_dst_blackhole(sk, &dst, &fl);
186 /* source address lookup done in ip6_dst_lookup */
188 if (ipv6_addr_any(&np->saddr))
189 ipv6_addr_copy(&np->saddr, &fl.fl6_src);
191 if (ipv6_addr_any(&np->rcv_saddr)) {
192 ipv6_addr_copy(&np->rcv_saddr, &fl.fl6_src);
193 inet->rcv_saddr = LOOPBACK4_IPV6;
196 ip6_dst_store(sk, dst,
197 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
199 #ifdef CONFIG_IPV6_SUBTREES
200 ipv6_addr_equal(&fl.fl6_src, &np->saddr) ?
205 sk->sk_state = TCP_ESTABLISHED;
207 fl6_sock_release(flowlabel);
211 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
212 __be16 port, u32 info, u8 *payload)
214 struct ipv6_pinfo *np = inet6_sk(sk);
215 struct icmp6hdr *icmph = icmp6_hdr(skb);
216 struct sock_exterr_skb *serr;
221 skb = skb_clone(skb, GFP_ATOMIC);
225 serr = SKB_EXT_ERR(skb);
226 serr->ee.ee_errno = err;
227 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
228 serr->ee.ee_type = icmph->icmp6_type;
229 serr->ee.ee_code = icmph->icmp6_code;
231 serr->ee.ee_info = info;
232 serr->ee.ee_data = 0;
233 serr->addr_offset = (u8 *)&(((struct ipv6hdr *)(icmph + 1))->daddr) -
234 skb_network_header(skb);
237 __skb_pull(skb, payload - skb->data);
238 skb_reset_transport_header(skb);
240 if (sock_queue_err_skb(sk, skb))
244 void ipv6_local_error(struct sock *sk, int err, struct flowi *fl, u32 info)
246 struct ipv6_pinfo *np = inet6_sk(sk);
247 struct sock_exterr_skb *serr;
254 skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
258 skb_put(skb, sizeof(struct ipv6hdr));
259 skb_reset_network_header(skb);
261 ipv6_addr_copy(&iph->daddr, &fl->fl6_dst);
263 serr = SKB_EXT_ERR(skb);
264 serr->ee.ee_errno = err;
265 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
266 serr->ee.ee_type = 0;
267 serr->ee.ee_code = 0;
269 serr->ee.ee_info = info;
270 serr->ee.ee_data = 0;
271 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
272 serr->port = fl->fl_ip_dport;
274 __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
275 skb_reset_transport_header(skb);
277 if (sock_queue_err_skb(sk, skb))
282 * Handle MSG_ERRQUEUE
284 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len)
286 struct ipv6_pinfo *np = inet6_sk(sk);
287 struct sock_exterr_skb *serr;
288 struct sk_buff *skb, *skb2;
289 struct sockaddr_in6 *sin;
291 struct sock_extended_err ee;
292 struct sockaddr_in6 offender;
298 skb = skb_dequeue(&sk->sk_error_queue);
304 msg->msg_flags |= MSG_TRUNC;
307 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
311 sock_recv_timestamp(msg, sk, skb);
313 serr = SKB_EXT_ERR(skb);
315 sin = (struct sockaddr_in6 *)msg->msg_name;
317 const unsigned char *nh = skb_network_header(skb);
318 sin->sin6_family = AF_INET6;
319 sin->sin6_flowinfo = 0;
320 sin->sin6_port = serr->port;
321 sin->sin6_scope_id = 0;
322 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6) {
323 ipv6_addr_copy(&sin->sin6_addr,
324 (struct in6_addr *)(nh + serr->addr_offset));
327 (*(__be32 *)(nh + serr->addr_offset - 24) &
329 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
330 sin->sin6_scope_id = IP6CB(skb)->iif;
332 ipv6_addr_set(&sin->sin6_addr, 0, 0,
334 *(__be32 *)(nh + serr->addr_offset));
338 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
339 sin = &errhdr.offender;
340 sin->sin6_family = AF_UNSPEC;
341 if (serr->ee.ee_origin != SO_EE_ORIGIN_LOCAL) {
342 sin->sin6_family = AF_INET6;
343 sin->sin6_flowinfo = 0;
344 sin->sin6_scope_id = 0;
345 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6) {
346 ipv6_addr_copy(&sin->sin6_addr, &ipv6_hdr(skb)->saddr);
348 datagram_recv_ctl(sk, msg, skb);
349 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
350 sin->sin6_scope_id = IP6CB(skb)->iif;
352 struct inet_sock *inet = inet_sk(sk);
354 ipv6_addr_set(&sin->sin6_addr, 0, 0,
355 htonl(0xffff), ip_hdr(skb)->saddr);
356 if (inet->cmsg_flags)
357 ip_cmsg_recv(msg, skb);
361 put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
363 /* Now we could try to dump offended packet options */
365 msg->msg_flags |= MSG_ERRQUEUE;
368 /* Reset and regenerate socket error */
369 spin_lock_bh(&sk->sk_error_queue.lock);
371 if ((skb2 = skb_peek(&sk->sk_error_queue)) != NULL) {
372 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
373 spin_unlock_bh(&sk->sk_error_queue.lock);
374 sk->sk_error_report(sk);
376 spin_unlock_bh(&sk->sk_error_queue.lock);
387 int datagram_recv_ctl(struct sock *sk, struct msghdr *msg, struct sk_buff *skb)
389 struct ipv6_pinfo *np = inet6_sk(sk);
390 struct inet6_skb_parm *opt = IP6CB(skb);
391 unsigned char *nh = skb_network_header(skb);
393 if (np->rxopt.bits.rxinfo) {
394 struct in6_pktinfo src_info;
396 src_info.ipi6_ifindex = opt->iif;
397 ipv6_addr_copy(&src_info.ipi6_addr, &ipv6_hdr(skb)->daddr);
398 put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
401 if (np->rxopt.bits.rxhlim) {
402 int hlim = ipv6_hdr(skb)->hop_limit;
403 put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
406 if (np->rxopt.bits.rxtclass) {
407 int tclass = (ntohl(*(__be32 *)ipv6_hdr(skb)) >> 20) & 0xff;
408 put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
411 if (np->rxopt.bits.rxflow && (*(__be32 *)nh & IPV6_FLOWINFO_MASK)) {
412 __be32 flowinfo = *(__be32 *)nh & IPV6_FLOWINFO_MASK;
413 put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
416 /* HbH is allowed only once */
417 if (np->rxopt.bits.hopopts && opt->hop) {
418 u8 *ptr = nh + opt->hop;
419 put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
423 (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
425 * Silly enough, but we need to reparse in order to
426 * report extension headers (except for HbH)
429 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
430 * (and WILL NOT be) defined because
431 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
433 unsigned int off = sizeof(struct ipv6hdr);
434 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
436 while (off <= opt->lastopt) {
441 case IPPROTO_DSTOPTS:
443 len = (ptr[1] + 1) << 3;
444 if (np->rxopt.bits.dstopts)
445 put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
447 case IPPROTO_ROUTING:
449 len = (ptr[1] + 1) << 3;
450 if (np->rxopt.bits.srcrt)
451 put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
455 len = (ptr[1] + 2) << 2;
459 len = (ptr[1] + 1) << 3;
467 /* socket options in old style */
468 if (np->rxopt.bits.rxoinfo) {
469 struct in6_pktinfo src_info;
471 src_info.ipi6_ifindex = opt->iif;
472 ipv6_addr_copy(&src_info.ipi6_addr, &ipv6_hdr(skb)->daddr);
473 put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
475 if (np->rxopt.bits.rxohlim) {
476 int hlim = ipv6_hdr(skb)->hop_limit;
477 put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
479 if (np->rxopt.bits.ohopopts && opt->hop) {
480 u8 *ptr = nh + opt->hop;
481 put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
483 if (np->rxopt.bits.odstopts && opt->dst0) {
484 u8 *ptr = nh + opt->dst0;
485 put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
487 if (np->rxopt.bits.osrcrt && opt->srcrt) {
488 struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(nh + opt->srcrt);
489 put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
491 if (np->rxopt.bits.odstopts && opt->dst1) {
492 u8 *ptr = nh + opt->dst1;
493 put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
498 int datagram_send_ctl(struct net *net,
499 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)) {
513 if (!CMSG_OK(msg, cmsg)) {
518 if (cmsg->cmsg_level != SOL_IPV6)
521 switch (cmsg->cmsg_type) {
523 case IPV6_2292PKTINFO:
525 struct net_device *dev = NULL;
527 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
532 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
534 if (src_info->ipi6_ifindex) {
535 if (fl->oif && src_info->ipi6_ifindex != fl->oif)
537 fl->oif = src_info->ipi6_ifindex;
540 addr_type = __ipv6_addr_type(&src_info->ipi6_addr);
543 dev = dev_get_by_index(net, fl->oif);
546 } else if (addr_type & IPV6_ADDR_LINKLOCAL)
549 if (addr_type != IPV6_ADDR_ANY) {
550 int strict = __ipv6_addr_src_scope(addr_type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
551 if (!ipv6_chk_addr(net, &src_info->ipi6_addr,
552 strict ? dev : NULL, 0))
555 ipv6_addr_copy(&fl->fl6_src, &src_info->ipi6_addr);
568 if (cmsg->cmsg_len < CMSG_LEN(4)) {
573 if (fl->fl6_flowlabel&IPV6_FLOWINFO_MASK) {
574 if ((fl->fl6_flowlabel^*(__be32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
579 fl->fl6_flowlabel = IPV6_FLOWINFO_MASK & *(__be32 *)CMSG_DATA(cmsg);
582 case IPV6_2292HOPOPTS:
584 if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
589 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
590 len = ((hdr->hdrlen + 1) << 3);
591 if (cmsg->cmsg_len < CMSG_LEN(len)) {
595 if (!capable(CAP_NET_RAW)) {
599 opt->opt_nflen += len;
603 case IPV6_2292DSTOPTS:
604 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
609 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
610 len = ((hdr->hdrlen + 1) << 3);
611 if (cmsg->cmsg_len < CMSG_LEN(len)) {
615 if (!capable(CAP_NET_RAW)) {
623 opt->opt_flen += len;
628 case IPV6_RTHDRDSTOPTS:
629 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
634 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
635 len = ((hdr->hdrlen + 1) << 3);
636 if (cmsg->cmsg_len < CMSG_LEN(len)) {
640 if (!capable(CAP_NET_RAW)) {
644 if (cmsg->cmsg_type == IPV6_DSTOPTS) {
645 opt->opt_flen += len;
648 opt->opt_nflen += len;
655 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
660 rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
662 switch (rthdr->type) {
663 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
664 case IPV6_SRCRT_TYPE_2:
665 if (rthdr->hdrlen != 2 ||
666 rthdr->segments_left != 1) {
677 len = ((rthdr->hdrlen + 1) << 3);
679 if (cmsg->cmsg_len < CMSG_LEN(len)) {
684 /* segments left must also match */
685 if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
690 opt->opt_nflen += len;
693 if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
694 int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
696 opt->opt_nflen += dsthdrlen;
697 opt->dst0opt = opt->dst1opt;
699 opt->opt_flen -= dsthdrlen;
704 case IPV6_2292HOPLIMIT:
706 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
711 *hlimit = *(int *)CMSG_DATA(cmsg);
712 if (*hlimit < -1 || *hlimit > 0xff) {
724 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
728 tc = *(int *)CMSG_DATA(cmsg);
729 if (tc < -1 || tc > 0xff)
738 LIMIT_NETDEBUG(KERN_DEBUG "invalid cmsg type: %d\n",