Staging: Merge branch 'tidspbridge-for-2.6.39' of git://dev.omapzoom.org/pub/scm...
[zen-stable.git] / net / ipv6 / raw.c
blob86c39526ba5ecb581e197d245855c9edabb53e99
1 /*
2 * RAW sockets for IPv6
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Adapted from linux/net/ipv4/raw.c
10 * Fixes:
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
13 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
15 * This program is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU General Public License
17 * as published by the Free Software Foundation; either version
18 * 2 of the License, or (at your option) any later version.
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/slab.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/netfilter.h>
32 #include <linux/netfilter_ipv6.h>
33 #include <linux/skbuff.h>
34 #include <asm/uaccess.h>
35 #include <asm/ioctls.h>
37 #include <net/net_namespace.h>
38 #include <net/ip.h>
39 #include <net/sock.h>
40 #include <net/snmp.h>
42 #include <net/ipv6.h>
43 #include <net/ndisc.h>
44 #include <net/protocol.h>
45 #include <net/ip6_route.h>
46 #include <net/ip6_checksum.h>
47 #include <net/addrconf.h>
48 #include <net/transp_v6.h>
49 #include <net/udp.h>
50 #include <net/inet_common.h>
51 #include <net/tcp_states.h>
52 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
53 #include <net/mip6.h>
54 #endif
55 #include <linux/mroute6.h>
57 #include <net/raw.h>
58 #include <net/rawv6.h>
59 #include <net/xfrm.h>
61 #include <linux/proc_fs.h>
62 #include <linux/seq_file.h>
64 static struct raw_hashinfo raw_v6_hashinfo = {
65 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
68 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
69 unsigned short num, struct in6_addr *loc_addr,
70 struct in6_addr *rmt_addr, int dif)
72 struct hlist_node *node;
73 int is_multicast = ipv6_addr_is_multicast(loc_addr);
75 sk_for_each_from(sk, node)
76 if (inet_sk(sk)->inet_num == num) {
77 struct ipv6_pinfo *np = inet6_sk(sk);
79 if (!net_eq(sock_net(sk), net))
80 continue;
82 if (!ipv6_addr_any(&np->daddr) &&
83 !ipv6_addr_equal(&np->daddr, rmt_addr))
84 continue;
86 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
87 continue;
89 if (!ipv6_addr_any(&np->rcv_saddr)) {
90 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
91 goto found;
92 if (is_multicast &&
93 inet6_mc_check(sk, loc_addr, rmt_addr))
94 goto found;
95 continue;
97 goto found;
99 sk = NULL;
100 found:
101 return sk;
105 * 0 - deliver
106 * 1 - block
108 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
110 struct icmp6hdr *icmph;
111 struct raw6_sock *rp = raw6_sk(sk);
113 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
114 __u32 *data = &rp->filter.data[0];
115 int bit_nr;
117 icmph = (struct icmp6hdr *) skb->data;
118 bit_nr = icmph->icmp6_type;
120 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
122 return 0;
125 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
126 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
128 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
129 struct sk_buff *skb))
131 rcu_assign_pointer(mh_filter, filter);
132 return 0;
134 EXPORT_SYMBOL(rawv6_mh_filter_register);
136 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
137 struct sk_buff *skb))
139 rcu_assign_pointer(mh_filter, NULL);
140 synchronize_rcu();
141 return 0;
143 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
145 #endif
148 * demultiplex raw sockets.
149 * (should consider queueing the skb in the sock receive_queue
150 * without calling rawv6.c)
152 * Caller owns SKB so we must make clones.
154 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
156 struct in6_addr *saddr;
157 struct in6_addr *daddr;
158 struct sock *sk;
159 int delivered = 0;
160 __u8 hash;
161 struct net *net;
163 saddr = &ipv6_hdr(skb)->saddr;
164 daddr = saddr + 1;
166 hash = nexthdr & (MAX_INET_PROTOS - 1);
168 read_lock(&raw_v6_hashinfo.lock);
169 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
171 if (sk == NULL)
172 goto out;
174 net = dev_net(skb->dev);
175 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
177 while (sk) {
178 int filtered;
180 delivered = 1;
181 switch (nexthdr) {
182 case IPPROTO_ICMPV6:
183 filtered = icmpv6_filter(sk, skb);
184 break;
186 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
187 case IPPROTO_MH:
189 /* XXX: To validate MH only once for each packet,
190 * this is placed here. It should be after checking
191 * xfrm policy, however it doesn't. The checking xfrm
192 * policy is placed in rawv6_rcv() because it is
193 * required for each socket.
195 int (*filter)(struct sock *sock, struct sk_buff *skb);
197 filter = rcu_dereference(mh_filter);
198 filtered = filter ? filter(sk, skb) : 0;
199 break;
201 #endif
202 default:
203 filtered = 0;
204 break;
207 if (filtered < 0)
208 break;
209 if (filtered == 0) {
210 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
212 /* Not releasing hash table! */
213 if (clone) {
214 nf_reset(clone);
215 rawv6_rcv(sk, clone);
218 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
219 IP6CB(skb)->iif);
221 out:
222 read_unlock(&raw_v6_hashinfo.lock);
223 return delivered;
226 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
228 struct sock *raw_sk;
230 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
231 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
232 raw_sk = NULL;
234 return raw_sk != NULL;
237 /* This cleans up af_inet6 a bit. -DaveM */
238 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
240 struct inet_sock *inet = inet_sk(sk);
241 struct ipv6_pinfo *np = inet6_sk(sk);
242 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
243 __be32 v4addr = 0;
244 int addr_type;
245 int err;
247 if (addr_len < SIN6_LEN_RFC2133)
248 return -EINVAL;
249 addr_type = ipv6_addr_type(&addr->sin6_addr);
251 /* Raw sockets are IPv6 only */
252 if (addr_type == IPV6_ADDR_MAPPED)
253 return -EADDRNOTAVAIL;
255 lock_sock(sk);
257 err = -EINVAL;
258 if (sk->sk_state != TCP_CLOSE)
259 goto out;
261 rcu_read_lock();
262 /* Check if the address belongs to the host. */
263 if (addr_type != IPV6_ADDR_ANY) {
264 struct net_device *dev = NULL;
266 if (addr_type & IPV6_ADDR_LINKLOCAL) {
267 if (addr_len >= sizeof(struct sockaddr_in6) &&
268 addr->sin6_scope_id) {
269 /* Override any existing binding, if another
270 * one is supplied by user.
272 sk->sk_bound_dev_if = addr->sin6_scope_id;
275 /* Binding to link-local address requires an interface */
276 if (!sk->sk_bound_dev_if)
277 goto out_unlock;
279 err = -ENODEV;
280 dev = dev_get_by_index_rcu(sock_net(sk),
281 sk->sk_bound_dev_if);
282 if (!dev)
283 goto out_unlock;
286 /* ipv4 addr of the socket is invalid. Only the
287 * unspecified and mapped address have a v4 equivalent.
289 v4addr = LOOPBACK4_IPV6;
290 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
291 err = -EADDRNOTAVAIL;
292 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
293 dev, 0)) {
294 goto out_unlock;
299 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
300 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
301 if (!(addr_type & IPV6_ADDR_MULTICAST))
302 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
303 err = 0;
304 out_unlock:
305 rcu_read_unlock();
306 out:
307 release_sock(sk);
308 return err;
311 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
312 struct inet6_skb_parm *opt,
313 u8 type, u8 code, int offset, __be32 info)
315 struct inet_sock *inet = inet_sk(sk);
316 struct ipv6_pinfo *np = inet6_sk(sk);
317 int err;
318 int harderr;
320 /* Report error on raw socket, if:
321 1. User requested recverr.
322 2. Socket is connected (otherwise the error indication
323 is useless without recverr and error is hard.
325 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
326 return;
328 harderr = icmpv6_err_convert(type, code, &err);
329 if (type == ICMPV6_PKT_TOOBIG)
330 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
332 if (np->recverr) {
333 u8 *payload = skb->data;
334 if (!inet->hdrincl)
335 payload += offset;
336 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
339 if (np->recverr || harderr) {
340 sk->sk_err = err;
341 sk->sk_error_report(sk);
345 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
346 u8 type, u8 code, int inner_offset, __be32 info)
348 struct sock *sk;
349 int hash;
350 struct in6_addr *saddr, *daddr;
351 struct net *net;
353 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
355 read_lock(&raw_v6_hashinfo.lock);
356 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
357 if (sk != NULL) {
358 /* Note: ipv6_hdr(skb) != skb->data */
359 struct ipv6hdr *ip6h = (struct ipv6hdr *)skb->data;
360 saddr = &ip6h->saddr;
361 daddr = &ip6h->daddr;
362 net = dev_net(skb->dev);
364 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
365 IP6CB(skb)->iif))) {
366 rawv6_err(sk, skb, NULL, type, code,
367 inner_offset, info);
368 sk = sk_next(sk);
371 read_unlock(&raw_v6_hashinfo.lock);
374 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
376 if ((raw6_sk(sk)->checksum || rcu_dereference_raw(sk->sk_filter)) &&
377 skb_checksum_complete(skb)) {
378 atomic_inc(&sk->sk_drops);
379 kfree_skb(skb);
380 return NET_RX_DROP;
383 /* Charge it to the socket. */
384 if (ip_queue_rcv_skb(sk, skb) < 0) {
385 kfree_skb(skb);
386 return NET_RX_DROP;
389 return 0;
393 * This is next to useless...
394 * if we demultiplex in network layer we don't need the extra call
395 * just to queue the skb...
396 * maybe we could have the network decide upon a hint if it
397 * should call raw_rcv for demultiplexing
399 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
401 struct inet_sock *inet = inet_sk(sk);
402 struct raw6_sock *rp = raw6_sk(sk);
404 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
405 atomic_inc(&sk->sk_drops);
406 kfree_skb(skb);
407 return NET_RX_DROP;
410 if (!rp->checksum)
411 skb->ip_summed = CHECKSUM_UNNECESSARY;
413 if (skb->ip_summed == CHECKSUM_COMPLETE) {
414 skb_postpull_rcsum(skb, skb_network_header(skb),
415 skb_network_header_len(skb));
416 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
417 &ipv6_hdr(skb)->daddr,
418 skb->len, inet->inet_num, skb->csum))
419 skb->ip_summed = CHECKSUM_UNNECESSARY;
421 if (!skb_csum_unnecessary(skb))
422 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
423 &ipv6_hdr(skb)->daddr,
424 skb->len,
425 inet->inet_num, 0));
427 if (inet->hdrincl) {
428 if (skb_checksum_complete(skb)) {
429 atomic_inc(&sk->sk_drops);
430 kfree_skb(skb);
431 return NET_RX_DROP;
435 rawv6_rcv_skb(sk, skb);
436 return 0;
441 * This should be easy, if there is something there
442 * we return it, otherwise we block.
445 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
446 struct msghdr *msg, size_t len,
447 int noblock, int flags, int *addr_len)
449 struct ipv6_pinfo *np = inet6_sk(sk);
450 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
451 struct sk_buff *skb;
452 size_t copied;
453 int err;
455 if (flags & MSG_OOB)
456 return -EOPNOTSUPP;
458 if (addr_len)
459 *addr_len=sizeof(*sin6);
461 if (flags & MSG_ERRQUEUE)
462 return ipv6_recv_error(sk, msg, len);
464 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
465 return ipv6_recv_rxpmtu(sk, msg, len);
467 skb = skb_recv_datagram(sk, flags, noblock, &err);
468 if (!skb)
469 goto out;
471 copied = skb->len;
472 if (copied > len) {
473 copied = len;
474 msg->msg_flags |= MSG_TRUNC;
477 if (skb_csum_unnecessary(skb)) {
478 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
479 } else if (msg->msg_flags&MSG_TRUNC) {
480 if (__skb_checksum_complete(skb))
481 goto csum_copy_err;
482 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
483 } else {
484 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
485 if (err == -EINVAL)
486 goto csum_copy_err;
488 if (err)
489 goto out_free;
491 /* Copy the address. */
492 if (sin6) {
493 sin6->sin6_family = AF_INET6;
494 sin6->sin6_port = 0;
495 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
496 sin6->sin6_flowinfo = 0;
497 sin6->sin6_scope_id = 0;
498 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
499 sin6->sin6_scope_id = IP6CB(skb)->iif;
502 sock_recv_ts_and_drops(msg, sk, skb);
504 if (np->rxopt.all)
505 datagram_recv_ctl(sk, msg, skb);
507 err = copied;
508 if (flags & MSG_TRUNC)
509 err = skb->len;
511 out_free:
512 skb_free_datagram(sk, skb);
513 out:
514 return err;
516 csum_copy_err:
517 skb_kill_datagram(sk, skb, flags);
519 /* Error for blocking case is chosen to masquerade
520 as some normal condition.
522 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
523 goto out;
526 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
527 struct raw6_sock *rp)
529 struct sk_buff *skb;
530 int err = 0;
531 int offset;
532 int len;
533 int total_len;
534 __wsum tmp_csum;
535 __sum16 csum;
537 if (!rp->checksum)
538 goto send;
540 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
541 goto out;
543 offset = rp->offset;
544 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
545 skb->data);
546 if (offset >= total_len - 1) {
547 err = -EINVAL;
548 ip6_flush_pending_frames(sk);
549 goto out;
552 /* should be check HW csum miyazawa */
553 if (skb_queue_len(&sk->sk_write_queue) == 1) {
555 * Only one fragment on the socket.
557 tmp_csum = skb->csum;
558 } else {
559 struct sk_buff *csum_skb = NULL;
560 tmp_csum = 0;
562 skb_queue_walk(&sk->sk_write_queue, skb) {
563 tmp_csum = csum_add(tmp_csum, skb->csum);
565 if (csum_skb)
566 continue;
568 len = skb->len - skb_transport_offset(skb);
569 if (offset >= len) {
570 offset -= len;
571 continue;
574 csum_skb = skb;
577 skb = csum_skb;
580 offset += skb_transport_offset(skb);
581 if (skb_copy_bits(skb, offset, &csum, 2))
582 BUG();
584 /* in case cksum was not initialized */
585 if (unlikely(csum))
586 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
588 csum = csum_ipv6_magic(&fl->fl6_src,
589 &fl->fl6_dst,
590 total_len, fl->proto, tmp_csum);
592 if (csum == 0 && fl->proto == IPPROTO_UDP)
593 csum = CSUM_MANGLED_0;
595 if (skb_store_bits(skb, offset, &csum, 2))
596 BUG();
598 send:
599 err = ip6_push_pending_frames(sk);
600 out:
601 return err;
604 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
605 struct flowi *fl, struct dst_entry **dstp,
606 unsigned int flags)
608 struct ipv6_pinfo *np = inet6_sk(sk);
609 struct ipv6hdr *iph;
610 struct sk_buff *skb;
611 int err;
612 struct rt6_info *rt = (struct rt6_info *)*dstp;
614 if (length > rt->dst.dev->mtu) {
615 ipv6_local_error(sk, EMSGSIZE, fl, rt->dst.dev->mtu);
616 return -EMSGSIZE;
618 if (flags&MSG_PROBE)
619 goto out;
621 skb = sock_alloc_send_skb(sk,
622 length + LL_ALLOCATED_SPACE(rt->dst.dev) + 15,
623 flags & MSG_DONTWAIT, &err);
624 if (skb == NULL)
625 goto error;
626 skb_reserve(skb, LL_RESERVED_SPACE(rt->dst.dev));
628 skb->priority = sk->sk_priority;
629 skb->mark = sk->sk_mark;
630 skb_dst_set(skb, &rt->dst);
631 *dstp = NULL;
633 skb_put(skb, length);
634 skb_reset_network_header(skb);
635 iph = ipv6_hdr(skb);
637 skb->ip_summed = CHECKSUM_NONE;
639 skb->transport_header = skb->network_header;
640 err = memcpy_fromiovecend((void *)iph, from, 0, length);
641 if (err)
642 goto error_fault;
644 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
645 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
646 rt->dst.dev, dst_output);
647 if (err > 0)
648 err = net_xmit_errno(err);
649 if (err)
650 goto error;
651 out:
652 return 0;
654 error_fault:
655 err = -EFAULT;
656 kfree_skb(skb);
657 error:
658 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
659 if (err == -ENOBUFS && !np->recverr)
660 err = 0;
661 return err;
664 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
666 struct iovec *iov;
667 u8 __user *type = NULL;
668 u8 __user *code = NULL;
669 u8 len = 0;
670 int probed = 0;
671 int i;
673 if (!msg->msg_iov)
674 return 0;
676 for (i = 0; i < msg->msg_iovlen; i++) {
677 iov = &msg->msg_iov[i];
678 if (!iov)
679 continue;
681 switch (fl->proto) {
682 case IPPROTO_ICMPV6:
683 /* check if one-byte field is readable or not. */
684 if (iov->iov_base && iov->iov_len < 1)
685 break;
687 if (!type) {
688 type = iov->iov_base;
689 /* check if code field is readable or not. */
690 if (iov->iov_len > 1)
691 code = type + 1;
692 } else if (!code)
693 code = iov->iov_base;
695 if (type && code) {
696 if (get_user(fl->fl_icmp_type, type) ||
697 get_user(fl->fl_icmp_code, code))
698 return -EFAULT;
699 probed = 1;
701 break;
702 case IPPROTO_MH:
703 if (iov->iov_base && iov->iov_len < 1)
704 break;
705 /* check if type field is readable or not. */
706 if (iov->iov_len > 2 - len) {
707 u8 __user *p = iov->iov_base;
708 if (get_user(fl->fl_mh_type, &p[2 - len]))
709 return -EFAULT;
710 probed = 1;
711 } else
712 len += iov->iov_len;
714 break;
715 default:
716 probed = 1;
717 break;
719 if (probed)
720 break;
722 return 0;
725 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
726 struct msghdr *msg, size_t len)
728 struct ipv6_txoptions opt_space;
729 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
730 struct in6_addr *daddr, *final_p, final;
731 struct inet_sock *inet = inet_sk(sk);
732 struct ipv6_pinfo *np = inet6_sk(sk);
733 struct raw6_sock *rp = raw6_sk(sk);
734 struct ipv6_txoptions *opt = NULL;
735 struct ip6_flowlabel *flowlabel = NULL;
736 struct dst_entry *dst = NULL;
737 struct flowi fl;
738 int addr_len = msg->msg_namelen;
739 int hlimit = -1;
740 int tclass = -1;
741 int dontfrag = -1;
742 u16 proto;
743 int err;
745 /* Rough check on arithmetic overflow,
746 better check is made in ip6_append_data().
748 if (len > INT_MAX)
749 return -EMSGSIZE;
751 /* Mirror BSD error message compatibility */
752 if (msg->msg_flags & MSG_OOB)
753 return -EOPNOTSUPP;
756 * Get and verify the address.
758 memset(&fl, 0, sizeof(fl));
760 fl.mark = sk->sk_mark;
762 if (sin6) {
763 if (addr_len < SIN6_LEN_RFC2133)
764 return -EINVAL;
766 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
767 return -EAFNOSUPPORT;
769 /* port is the proto value [0..255] carried in nexthdr */
770 proto = ntohs(sin6->sin6_port);
772 if (!proto)
773 proto = inet->inet_num;
774 else if (proto != inet->inet_num)
775 return -EINVAL;
777 if (proto > 255)
778 return -EINVAL;
780 daddr = &sin6->sin6_addr;
781 if (np->sndflow) {
782 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
783 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
784 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
785 if (flowlabel == NULL)
786 return -EINVAL;
787 daddr = &flowlabel->dst;
792 * Otherwise it will be difficult to maintain
793 * sk->sk_dst_cache.
795 if (sk->sk_state == TCP_ESTABLISHED &&
796 ipv6_addr_equal(daddr, &np->daddr))
797 daddr = &np->daddr;
799 if (addr_len >= sizeof(struct sockaddr_in6) &&
800 sin6->sin6_scope_id &&
801 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
802 fl.oif = sin6->sin6_scope_id;
803 } else {
804 if (sk->sk_state != TCP_ESTABLISHED)
805 return -EDESTADDRREQ;
807 proto = inet->inet_num;
808 daddr = &np->daddr;
809 fl.fl6_flowlabel = np->flow_label;
812 if (fl.oif == 0)
813 fl.oif = sk->sk_bound_dev_if;
815 if (msg->msg_controllen) {
816 opt = &opt_space;
817 memset(opt, 0, sizeof(struct ipv6_txoptions));
818 opt->tot_len = sizeof(struct ipv6_txoptions);
820 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit,
821 &tclass, &dontfrag);
822 if (err < 0) {
823 fl6_sock_release(flowlabel);
824 return err;
826 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
827 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
828 if (flowlabel == NULL)
829 return -EINVAL;
831 if (!(opt->opt_nflen|opt->opt_flen))
832 opt = NULL;
834 if (opt == NULL)
835 opt = np->opt;
836 if (flowlabel)
837 opt = fl6_merge_options(&opt_space, flowlabel, opt);
838 opt = ipv6_fixup_options(&opt_space, opt);
840 fl.proto = proto;
841 err = rawv6_probe_proto_opt(&fl, msg);
842 if (err)
843 goto out;
845 if (!ipv6_addr_any(daddr))
846 ipv6_addr_copy(&fl.fl6_dst, daddr);
847 else
848 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
849 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
850 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
852 final_p = fl6_update_dst(&fl, opt, &final);
854 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
855 fl.oif = np->mcast_oif;
856 security_sk_classify_flow(sk, &fl);
858 err = ip6_dst_lookup(sk, &dst, &fl);
859 if (err)
860 goto out;
861 if (final_p)
862 ipv6_addr_copy(&fl.fl6_dst, final_p);
864 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
865 if (err < 0) {
866 if (err == -EREMOTE)
867 err = ip6_dst_blackhole(sk, &dst, &fl);
868 if (err < 0)
869 goto out;
872 if (hlimit < 0) {
873 if (ipv6_addr_is_multicast(&fl.fl6_dst))
874 hlimit = np->mcast_hops;
875 else
876 hlimit = np->hop_limit;
877 if (hlimit < 0)
878 hlimit = ip6_dst_hoplimit(dst);
881 if (tclass < 0)
882 tclass = np->tclass;
884 if (dontfrag < 0)
885 dontfrag = np->dontfrag;
887 if (msg->msg_flags&MSG_CONFIRM)
888 goto do_confirm;
890 back_from_confirm:
891 if (inet->hdrincl)
892 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, &dst, msg->msg_flags);
893 else {
894 lock_sock(sk);
895 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
896 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
897 msg->msg_flags, dontfrag);
899 if (err)
900 ip6_flush_pending_frames(sk);
901 else if (!(msg->msg_flags & MSG_MORE))
902 err = rawv6_push_pending_frames(sk, &fl, rp);
903 release_sock(sk);
905 done:
906 dst_release(dst);
907 out:
908 fl6_sock_release(flowlabel);
909 return err<0?err:len;
910 do_confirm:
911 dst_confirm(dst);
912 if (!(msg->msg_flags & MSG_PROBE) || len)
913 goto back_from_confirm;
914 err = 0;
915 goto done;
918 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
919 char __user *optval, int optlen)
921 switch (optname) {
922 case ICMPV6_FILTER:
923 if (optlen > sizeof(struct icmp6_filter))
924 optlen = sizeof(struct icmp6_filter);
925 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
926 return -EFAULT;
927 return 0;
928 default:
929 return -ENOPROTOOPT;
932 return 0;
935 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
936 char __user *optval, int __user *optlen)
938 int len;
940 switch (optname) {
941 case ICMPV6_FILTER:
942 if (get_user(len, optlen))
943 return -EFAULT;
944 if (len < 0)
945 return -EINVAL;
946 if (len > sizeof(struct icmp6_filter))
947 len = sizeof(struct icmp6_filter);
948 if (put_user(len, optlen))
949 return -EFAULT;
950 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
951 return -EFAULT;
952 return 0;
953 default:
954 return -ENOPROTOOPT;
957 return 0;
961 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
962 char __user *optval, unsigned int optlen)
964 struct raw6_sock *rp = raw6_sk(sk);
965 int val;
967 if (get_user(val, (int __user *)optval))
968 return -EFAULT;
970 switch (optname) {
971 case IPV6_CHECKSUM:
972 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
973 level == IPPROTO_IPV6) {
975 * RFC3542 tells that IPV6_CHECKSUM socket
976 * option in the IPPROTO_IPV6 level is not
977 * allowed on ICMPv6 sockets.
978 * If you want to set it, use IPPROTO_RAW
979 * level IPV6_CHECKSUM socket option
980 * (Linux extension).
982 return -EINVAL;
985 /* You may get strange result with a positive odd offset;
986 RFC2292bis agrees with me. */
987 if (val > 0 && (val&1))
988 return -EINVAL;
989 if (val < 0) {
990 rp->checksum = 0;
991 } else {
992 rp->checksum = 1;
993 rp->offset = val;
996 return 0;
997 break;
999 default:
1000 return -ENOPROTOOPT;
1004 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1005 char __user *optval, unsigned int optlen)
1007 switch(level) {
1008 case SOL_RAW:
1009 break;
1011 case SOL_ICMPV6:
1012 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1013 return -EOPNOTSUPP;
1014 return rawv6_seticmpfilter(sk, level, optname, optval,
1015 optlen);
1016 case SOL_IPV6:
1017 if (optname == IPV6_CHECKSUM)
1018 break;
1019 default:
1020 return ipv6_setsockopt(sk, level, optname, optval,
1021 optlen);
1024 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1027 #ifdef CONFIG_COMPAT
1028 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1029 char __user *optval, unsigned int optlen)
1031 switch (level) {
1032 case SOL_RAW:
1033 break;
1034 case SOL_ICMPV6:
1035 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1036 return -EOPNOTSUPP;
1037 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1038 case SOL_IPV6:
1039 if (optname == IPV6_CHECKSUM)
1040 break;
1041 default:
1042 return compat_ipv6_setsockopt(sk, level, optname,
1043 optval, optlen);
1045 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1047 #endif
1049 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1050 char __user *optval, int __user *optlen)
1052 struct raw6_sock *rp = raw6_sk(sk);
1053 int val, len;
1055 if (get_user(len,optlen))
1056 return -EFAULT;
1058 switch (optname) {
1059 case IPV6_CHECKSUM:
1061 * We allow getsockopt() for IPPROTO_IPV6-level
1062 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1063 * since RFC3542 is silent about it.
1065 if (rp->checksum == 0)
1066 val = -1;
1067 else
1068 val = rp->offset;
1069 break;
1071 default:
1072 return -ENOPROTOOPT;
1075 len = min_t(unsigned int, sizeof(int), len);
1077 if (put_user(len, optlen))
1078 return -EFAULT;
1079 if (copy_to_user(optval,&val,len))
1080 return -EFAULT;
1081 return 0;
1084 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1085 char __user *optval, int __user *optlen)
1087 switch(level) {
1088 case SOL_RAW:
1089 break;
1091 case SOL_ICMPV6:
1092 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1093 return -EOPNOTSUPP;
1094 return rawv6_geticmpfilter(sk, level, optname, optval,
1095 optlen);
1096 case SOL_IPV6:
1097 if (optname == IPV6_CHECKSUM)
1098 break;
1099 default:
1100 return ipv6_getsockopt(sk, level, optname, optval,
1101 optlen);
1104 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1107 #ifdef CONFIG_COMPAT
1108 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1109 char __user *optval, int __user *optlen)
1111 switch (level) {
1112 case SOL_RAW:
1113 break;
1114 case SOL_ICMPV6:
1115 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1116 return -EOPNOTSUPP;
1117 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1118 case SOL_IPV6:
1119 if (optname == IPV6_CHECKSUM)
1120 break;
1121 default:
1122 return compat_ipv6_getsockopt(sk, level, optname,
1123 optval, optlen);
1125 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1127 #endif
1129 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1131 switch(cmd) {
1132 case SIOCOUTQ:
1134 int amount = sk_wmem_alloc_get(sk);
1136 return put_user(amount, (int __user *)arg);
1138 case SIOCINQ:
1140 struct sk_buff *skb;
1141 int amount = 0;
1143 spin_lock_bh(&sk->sk_receive_queue.lock);
1144 skb = skb_peek(&sk->sk_receive_queue);
1145 if (skb != NULL)
1146 amount = skb->tail - skb->transport_header;
1147 spin_unlock_bh(&sk->sk_receive_queue.lock);
1148 return put_user(amount, (int __user *)arg);
1151 default:
1152 #ifdef CONFIG_IPV6_MROUTE
1153 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1154 #else
1155 return -ENOIOCTLCMD;
1156 #endif
1160 static void rawv6_close(struct sock *sk, long timeout)
1162 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1163 ip6_ra_control(sk, -1);
1164 ip6mr_sk_done(sk);
1165 sk_common_release(sk);
1168 static void raw6_destroy(struct sock *sk)
1170 lock_sock(sk);
1171 ip6_flush_pending_frames(sk);
1172 release_sock(sk);
1174 inet6_destroy_sock(sk);
1177 static int rawv6_init_sk(struct sock *sk)
1179 struct raw6_sock *rp = raw6_sk(sk);
1181 switch (inet_sk(sk)->inet_num) {
1182 case IPPROTO_ICMPV6:
1183 rp->checksum = 1;
1184 rp->offset = 2;
1185 break;
1186 case IPPROTO_MH:
1187 rp->checksum = 1;
1188 rp->offset = 4;
1189 break;
1190 default:
1191 break;
1193 return 0;
1196 struct proto rawv6_prot = {
1197 .name = "RAWv6",
1198 .owner = THIS_MODULE,
1199 .close = rawv6_close,
1200 .destroy = raw6_destroy,
1201 .connect = ip6_datagram_connect,
1202 .disconnect = udp_disconnect,
1203 .ioctl = rawv6_ioctl,
1204 .init = rawv6_init_sk,
1205 .setsockopt = rawv6_setsockopt,
1206 .getsockopt = rawv6_getsockopt,
1207 .sendmsg = rawv6_sendmsg,
1208 .recvmsg = rawv6_recvmsg,
1209 .bind = rawv6_bind,
1210 .backlog_rcv = rawv6_rcv_skb,
1211 .hash = raw_hash_sk,
1212 .unhash = raw_unhash_sk,
1213 .obj_size = sizeof(struct raw6_sock),
1214 .h.raw_hash = &raw_v6_hashinfo,
1215 #ifdef CONFIG_COMPAT
1216 .compat_setsockopt = compat_rawv6_setsockopt,
1217 .compat_getsockopt = compat_rawv6_getsockopt,
1218 #endif
1221 #ifdef CONFIG_PROC_FS
1222 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1224 struct ipv6_pinfo *np = inet6_sk(sp);
1225 struct in6_addr *dest, *src;
1226 __u16 destp, srcp;
1228 dest = &np->daddr;
1229 src = &np->rcv_saddr;
1230 destp = 0;
1231 srcp = inet_sk(sp)->inet_num;
1232 seq_printf(seq,
1233 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1234 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1236 src->s6_addr32[0], src->s6_addr32[1],
1237 src->s6_addr32[2], src->s6_addr32[3], srcp,
1238 dest->s6_addr32[0], dest->s6_addr32[1],
1239 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1240 sp->sk_state,
1241 sk_wmem_alloc_get(sp),
1242 sk_rmem_alloc_get(sp),
1243 0, 0L, 0,
1244 sock_i_uid(sp), 0,
1245 sock_i_ino(sp),
1246 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1249 static int raw6_seq_show(struct seq_file *seq, void *v)
1251 if (v == SEQ_START_TOKEN)
1252 seq_printf(seq,
1253 " sl "
1254 "local_address "
1255 "remote_address "
1256 "st tx_queue rx_queue tr tm->when retrnsmt"
1257 " uid timeout inode ref pointer drops\n");
1258 else
1259 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1260 return 0;
1263 static const struct seq_operations raw6_seq_ops = {
1264 .start = raw_seq_start,
1265 .next = raw_seq_next,
1266 .stop = raw_seq_stop,
1267 .show = raw6_seq_show,
1270 static int raw6_seq_open(struct inode *inode, struct file *file)
1272 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1275 static const struct file_operations raw6_seq_fops = {
1276 .owner = THIS_MODULE,
1277 .open = raw6_seq_open,
1278 .read = seq_read,
1279 .llseek = seq_lseek,
1280 .release = seq_release_net,
1283 static int __net_init raw6_init_net(struct net *net)
1285 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1286 return -ENOMEM;
1288 return 0;
1291 static void __net_exit raw6_exit_net(struct net *net)
1293 proc_net_remove(net, "raw6");
1296 static struct pernet_operations raw6_net_ops = {
1297 .init = raw6_init_net,
1298 .exit = raw6_exit_net,
1301 int __init raw6_proc_init(void)
1303 return register_pernet_subsys(&raw6_net_ops);
1306 void raw6_proc_exit(void)
1308 unregister_pernet_subsys(&raw6_net_ops);
1310 #endif /* CONFIG_PROC_FS */
1312 /* Same as inet6_dgram_ops, sans udp_poll. */
1313 static const struct proto_ops inet6_sockraw_ops = {
1314 .family = PF_INET6,
1315 .owner = THIS_MODULE,
1316 .release = inet6_release,
1317 .bind = inet6_bind,
1318 .connect = inet_dgram_connect, /* ok */
1319 .socketpair = sock_no_socketpair, /* a do nothing */
1320 .accept = sock_no_accept, /* a do nothing */
1321 .getname = inet6_getname,
1322 .poll = datagram_poll, /* ok */
1323 .ioctl = inet6_ioctl, /* must change */
1324 .listen = sock_no_listen, /* ok */
1325 .shutdown = inet_shutdown, /* ok */
1326 .setsockopt = sock_common_setsockopt, /* ok */
1327 .getsockopt = sock_common_getsockopt, /* ok */
1328 .sendmsg = inet_sendmsg, /* ok */
1329 .recvmsg = sock_common_recvmsg, /* ok */
1330 .mmap = sock_no_mmap,
1331 .sendpage = sock_no_sendpage,
1332 #ifdef CONFIG_COMPAT
1333 .compat_setsockopt = compat_sock_common_setsockopt,
1334 .compat_getsockopt = compat_sock_common_getsockopt,
1335 #endif
1338 static struct inet_protosw rawv6_protosw = {
1339 .type = SOCK_RAW,
1340 .protocol = IPPROTO_IP, /* wild card */
1341 .prot = &rawv6_prot,
1342 .ops = &inet6_sockraw_ops,
1343 .no_check = UDP_CSUM_DEFAULT,
1344 .flags = INET_PROTOSW_REUSE,
1347 int __init rawv6_init(void)
1349 int ret;
1351 ret = inet6_register_protosw(&rawv6_protosw);
1352 if (ret)
1353 goto out;
1354 out:
1355 return ret;
1358 void rawv6_exit(void)
1360 inet6_unregister_protosw(&rawv6_protosw);