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/errno.h>
17 #include <linux/types.h>
18 #include <linux/kernel.h>
19 #include <linux/sched.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;
126 if (!sk->sk_bound_dev_if &&
127 (addr_type & IPV6_ADDR_MULTICAST))
128 fl.oif = 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;
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 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
181 /* source address lookup done in ip6_dst_lookup */
183 if (ipv6_addr_any(&np->saddr))
184 ipv6_addr_copy(&np->saddr, &fl.fl6_src);
186 if (ipv6_addr_any(&np->rcv_saddr)) {
187 ipv6_addr_copy(&np->rcv_saddr, &fl.fl6_src);
188 inet->rcv_saddr = LOOPBACK4_IPV6;
191 ip6_dst_store(sk, dst,
192 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
195 sk->sk_state = TCP_ESTABLISHED;
197 fl6_sock_release(flowlabel);
201 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
202 u16 port, u32 info, u8 *payload)
204 struct ipv6_pinfo *np = inet6_sk(sk);
205 struct icmp6hdr *icmph = (struct icmp6hdr *)skb->h.raw;
206 struct sock_exterr_skb *serr;
211 skb = skb_clone(skb, GFP_ATOMIC);
215 serr = SKB_EXT_ERR(skb);
216 serr->ee.ee_errno = err;
217 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP6;
218 serr->ee.ee_type = icmph->icmp6_type;
219 serr->ee.ee_code = icmph->icmp6_code;
221 serr->ee.ee_info = info;
222 serr->ee.ee_data = 0;
223 serr->addr_offset = (u8*)&(((struct ipv6hdr*)(icmph+1))->daddr) - skb->nh.raw;
226 skb->h.raw = payload;
227 __skb_pull(skb, payload - skb->data);
229 if (sock_queue_err_skb(sk, skb))
233 void ipv6_local_error(struct sock *sk, int err, struct flowi *fl, u32 info)
235 struct ipv6_pinfo *np = inet6_sk(sk);
236 struct sock_exterr_skb *serr;
243 skb = alloc_skb(sizeof(struct ipv6hdr), GFP_ATOMIC);
247 iph = (struct ipv6hdr*)skb_put(skb, sizeof(struct ipv6hdr));
249 ipv6_addr_copy(&iph->daddr, &fl->fl6_dst);
251 serr = SKB_EXT_ERR(skb);
252 serr->ee.ee_errno = err;
253 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
254 serr->ee.ee_type = 0;
255 serr->ee.ee_code = 0;
257 serr->ee.ee_info = info;
258 serr->ee.ee_data = 0;
259 serr->addr_offset = (u8*)&iph->daddr - skb->nh.raw;
260 serr->port = fl->fl_ip_dport;
262 skb->h.raw = skb->tail;
263 __skb_pull(skb, skb->tail - skb->data);
265 if (sock_queue_err_skb(sk, skb))
270 * Handle MSG_ERRQUEUE
272 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len)
274 struct ipv6_pinfo *np = inet6_sk(sk);
275 struct sock_exterr_skb *serr;
276 struct sk_buff *skb, *skb2;
277 struct sockaddr_in6 *sin;
279 struct sock_extended_err ee;
280 struct sockaddr_in6 offender;
286 skb = skb_dequeue(&sk->sk_error_queue);
292 msg->msg_flags |= MSG_TRUNC;
295 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
299 sock_recv_timestamp(msg, sk, skb);
301 serr = SKB_EXT_ERR(skb);
303 sin = (struct sockaddr_in6 *)msg->msg_name;
305 sin->sin6_family = AF_INET6;
306 sin->sin6_flowinfo = 0;
307 sin->sin6_port = serr->port;
308 sin->sin6_scope_id = 0;
309 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6) {
310 ipv6_addr_copy(&sin->sin6_addr,
311 (struct in6_addr *)(skb->nh.raw + serr->addr_offset));
313 sin->sin6_flowinfo = *(u32*)(skb->nh.raw + serr->addr_offset - 24) & IPV6_FLOWINFO_MASK;
314 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
315 sin->sin6_scope_id = IP6CB(skb)->iif;
317 ipv6_addr_set(&sin->sin6_addr, 0, 0,
319 *(u32*)(skb->nh.raw + serr->addr_offset));
323 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
324 sin = &errhdr.offender;
325 sin->sin6_family = AF_UNSPEC;
326 if (serr->ee.ee_origin != SO_EE_ORIGIN_LOCAL) {
327 sin->sin6_family = AF_INET6;
328 sin->sin6_flowinfo = 0;
329 sin->sin6_scope_id = 0;
330 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP6) {
331 ipv6_addr_copy(&sin->sin6_addr, &skb->nh.ipv6h->saddr);
333 datagram_recv_ctl(sk, msg, skb);
334 if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
335 sin->sin6_scope_id = IP6CB(skb)->iif;
337 struct inet_sock *inet = inet_sk(sk);
339 ipv6_addr_set(&sin->sin6_addr, 0, 0,
342 if (inet->cmsg_flags)
343 ip_cmsg_recv(msg, skb);
347 put_cmsg(msg, SOL_IPV6, IPV6_RECVERR, sizeof(errhdr), &errhdr);
349 /* Now we could try to dump offended packet options */
351 msg->msg_flags |= MSG_ERRQUEUE;
354 /* Reset and regenerate socket error */
355 spin_lock_bh(&sk->sk_error_queue.lock);
357 if ((skb2 = skb_peek(&sk->sk_error_queue)) != NULL) {
358 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
359 spin_unlock_bh(&sk->sk_error_queue.lock);
360 sk->sk_error_report(sk);
362 spin_unlock_bh(&sk->sk_error_queue.lock);
373 int datagram_recv_ctl(struct sock *sk, struct msghdr *msg, struct sk_buff *skb)
375 struct ipv6_pinfo *np = inet6_sk(sk);
376 struct inet6_skb_parm *opt = IP6CB(skb);
378 if (np->rxopt.bits.rxinfo) {
379 struct in6_pktinfo src_info;
381 src_info.ipi6_ifindex = opt->iif;
382 ipv6_addr_copy(&src_info.ipi6_addr, &skb->nh.ipv6h->daddr);
383 put_cmsg(msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
386 if (np->rxopt.bits.rxhlim) {
387 int hlim = skb->nh.ipv6h->hop_limit;
388 put_cmsg(msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
391 if (np->rxopt.bits.rxtclass) {
392 int tclass = (ntohl(*(u32 *)skb->nh.ipv6h) >> 20) & 0xff;
393 put_cmsg(msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
396 if (np->rxopt.bits.rxflow && (*(u32*)skb->nh.raw & IPV6_FLOWINFO_MASK)) {
397 u32 flowinfo = *(u32*)skb->nh.raw & IPV6_FLOWINFO_MASK;
398 put_cmsg(msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
401 /* HbH is allowed only once */
402 if (np->rxopt.bits.hopopts && opt->hop) {
403 u8 *ptr = skb->nh.raw + opt->hop;
404 put_cmsg(msg, SOL_IPV6, IPV6_HOPOPTS, (ptr[1]+1)<<3, ptr);
408 (np->rxopt.bits.dstopts || np->rxopt.bits.srcrt)) {
410 * Silly enough, but we need to reparse in order to
411 * report extension headers (except for HbH)
414 * Also note that IPV6_RECVRTHDRDSTOPTS is NOT
415 * (and WILL NOT be) defined because
416 * IPV6_RECVDSTOPTS is more generic. --yoshfuji
418 unsigned int off = sizeof(struct ipv6hdr);
419 u8 nexthdr = skb->nh.ipv6h->nexthdr;
421 while (off <= opt->lastopt) {
423 u8 *ptr = skb->nh.raw + off;
426 case IPPROTO_DSTOPTS:
428 len = (ptr[1] + 1) << 3;
429 if (np->rxopt.bits.dstopts)
430 put_cmsg(msg, SOL_IPV6, IPV6_DSTOPTS, len, ptr);
432 case IPPROTO_ROUTING:
434 len = (ptr[1] + 1) << 3;
435 if (np->rxopt.bits.srcrt)
436 put_cmsg(msg, SOL_IPV6, IPV6_RTHDR, len, ptr);
440 len = (ptr[1] + 2) << 2;
444 len = (ptr[1] + 1) << 3;
452 /* socket options in old style */
453 if (np->rxopt.bits.rxoinfo) {
454 struct in6_pktinfo src_info;
456 src_info.ipi6_ifindex = opt->iif;
457 ipv6_addr_copy(&src_info.ipi6_addr, &skb->nh.ipv6h->daddr);
458 put_cmsg(msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
460 if (np->rxopt.bits.rxohlim) {
461 int hlim = skb->nh.ipv6h->hop_limit;
462 put_cmsg(msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
464 if (np->rxopt.bits.ohopopts && opt->hop) {
465 u8 *ptr = skb->nh.raw + opt->hop;
466 put_cmsg(msg, SOL_IPV6, IPV6_2292HOPOPTS, (ptr[1]+1)<<3, ptr);
468 if (np->rxopt.bits.odstopts && opt->dst0) {
469 u8 *ptr = skb->nh.raw + opt->dst0;
470 put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
472 if (np->rxopt.bits.osrcrt && opt->srcrt) {
473 struct ipv6_rt_hdr *rthdr = (struct ipv6_rt_hdr *)(skb->nh.raw + opt->srcrt);
474 put_cmsg(msg, SOL_IPV6, IPV6_2292RTHDR, (rthdr->hdrlen+1) << 3, rthdr);
476 if (np->rxopt.bits.odstopts && opt->dst1) {
477 u8 *ptr = skb->nh.raw + opt->dst1;
478 put_cmsg(msg, SOL_IPV6, IPV6_2292DSTOPTS, (ptr[1]+1)<<3, ptr);
483 int datagram_send_ctl(struct msghdr *msg, struct flowi *fl,
484 struct ipv6_txoptions *opt,
485 int *hlimit, int *tclass)
487 struct in6_pktinfo *src_info;
488 struct cmsghdr *cmsg;
489 struct ipv6_rt_hdr *rthdr;
490 struct ipv6_opt_hdr *hdr;
494 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
496 struct net_device *dev = NULL;
498 if (!CMSG_OK(msg, cmsg)) {
503 if (cmsg->cmsg_level != SOL_IPV6)
506 switch (cmsg->cmsg_type) {
508 case IPV6_2292PKTINFO:
509 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct in6_pktinfo))) {
514 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
516 if (src_info->ipi6_ifindex) {
517 if (fl->oif && src_info->ipi6_ifindex != fl->oif)
519 fl->oif = src_info->ipi6_ifindex;
522 addr_type = ipv6_addr_type(&src_info->ipi6_addr);
524 if (addr_type == IPV6_ADDR_ANY)
527 if (addr_type & IPV6_ADDR_LINKLOCAL) {
528 if (!src_info->ipi6_ifindex)
531 dev = dev_get_by_index(src_info->ipi6_ifindex);
536 if (!ipv6_chk_addr(&src_info->ipi6_addr, dev, 0)) {
545 ipv6_addr_copy(&fl->fl6_src, &src_info->ipi6_addr);
549 if (cmsg->cmsg_len < CMSG_LEN(4)) {
554 if (fl->fl6_flowlabel&IPV6_FLOWINFO_MASK) {
555 if ((fl->fl6_flowlabel^*(u32 *)CMSG_DATA(cmsg))&~IPV6_FLOWINFO_MASK) {
560 fl->fl6_flowlabel = IPV6_FLOWINFO_MASK & *(u32 *)CMSG_DATA(cmsg);
563 case IPV6_2292HOPOPTS:
565 if (opt->hopopt || cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
570 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
571 len = ((hdr->hdrlen + 1) << 3);
572 if (cmsg->cmsg_len < CMSG_LEN(len)) {
576 if (!capable(CAP_NET_RAW)) {
580 opt->opt_nflen += len;
584 case IPV6_2292DSTOPTS:
585 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
590 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
591 len = ((hdr->hdrlen + 1) << 3);
592 if (cmsg->cmsg_len < CMSG_LEN(len)) {
596 if (!capable(CAP_NET_RAW)) {
604 opt->opt_flen += len;
609 case IPV6_RTHDRDSTOPTS:
610 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_opt_hdr))) {
615 hdr = (struct ipv6_opt_hdr *)CMSG_DATA(cmsg);
616 len = ((hdr->hdrlen + 1) << 3);
617 if (cmsg->cmsg_len < CMSG_LEN(len)) {
621 if (!capable(CAP_NET_RAW)) {
625 if (cmsg->cmsg_type == IPV6_DSTOPTS) {
626 opt->opt_flen += len;
629 opt->opt_nflen += len;
636 if (cmsg->cmsg_len < CMSG_LEN(sizeof(struct ipv6_rt_hdr))) {
641 rthdr = (struct ipv6_rt_hdr *)CMSG_DATA(cmsg);
651 len = ((rthdr->hdrlen + 1) << 3);
653 if (cmsg->cmsg_len < CMSG_LEN(len)) {
658 /* segments left must also match */
659 if ((rthdr->hdrlen >> 1) != rthdr->segments_left) {
664 opt->opt_nflen += len;
667 if (cmsg->cmsg_type == IPV6_2292RTHDR && opt->dst1opt) {
668 int dsthdrlen = ((opt->dst1opt->hdrlen+1)<<3);
670 opt->opt_nflen += dsthdrlen;
671 opt->dst0opt = opt->dst1opt;
673 opt->opt_flen -= dsthdrlen;
678 case IPV6_2292HOPLIMIT:
680 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
685 *hlimit = *(int *)CMSG_DATA(cmsg);
693 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) {
697 tc = *(int *)CMSG_DATA(cmsg);
698 if (tc < 0 || tc > 0xff)
707 LIMIT_NETDEBUG(KERN_DEBUG "invalid cmsg type: %d\n",