net caif: Register properly as a pernet subsystem.
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / ipv6 / raw.c
blob331af3b882ac211ca65cb59042f14c381fe2d244
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 <linux/compat.h>
35 #include <asm/uaccess.h>
36 #include <asm/ioctls.h>
38 #include <net/net_namespace.h>
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
50 #include <net/udp.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
54 #include <net/mip6.h>
55 #endif
56 #include <linux/mroute6.h>
58 #include <net/raw.h>
59 #include <net/rawv6.h>
60 #include <net/xfrm.h>
62 #include <linux/proc_fs.h>
63 #include <linux/seq_file.h>
64 #include <linux/export.h>
66 static struct raw_hashinfo raw_v6_hashinfo = {
67 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
70 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
71 unsigned short num, const struct in6_addr *loc_addr,
72 const struct in6_addr *rmt_addr, int dif)
74 struct hlist_node *node;
75 int is_multicast = ipv6_addr_is_multicast(loc_addr);
77 sk_for_each_from(sk, node)
78 if (inet_sk(sk)->inet_num == num) {
79 struct ipv6_pinfo *np = inet6_sk(sk);
81 if (!net_eq(sock_net(sk), net))
82 continue;
84 if (!ipv6_addr_any(&np->daddr) &&
85 !ipv6_addr_equal(&np->daddr, rmt_addr))
86 continue;
88 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
89 continue;
91 if (!ipv6_addr_any(&np->rcv_saddr)) {
92 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
93 goto found;
94 if (is_multicast &&
95 inet6_mc_check(sk, loc_addr, rmt_addr))
96 goto found;
97 continue;
99 goto found;
101 sk = NULL;
102 found:
103 return sk;
107 * 0 - deliver
108 * 1 - block
110 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
112 struct icmp6hdr *icmph;
113 struct raw6_sock *rp = raw6_sk(sk);
115 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
116 __u32 *data = &rp->filter.data[0];
117 int bit_nr;
119 icmph = (struct icmp6hdr *) skb->data;
120 bit_nr = icmph->icmp6_type;
122 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
124 return 0;
127 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
128 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
130 static mh_filter_t __rcu *mh_filter __read_mostly;
132 int rawv6_mh_filter_register(mh_filter_t filter)
134 RCU_INIT_POINTER(mh_filter, filter);
135 return 0;
137 EXPORT_SYMBOL(rawv6_mh_filter_register);
139 int rawv6_mh_filter_unregister(mh_filter_t filter)
141 RCU_INIT_POINTER(mh_filter, NULL);
142 synchronize_rcu();
143 return 0;
145 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
147 #endif
150 * demultiplex raw sockets.
151 * (should consider queueing the skb in the sock receive_queue
152 * without calling rawv6.c)
154 * Caller owns SKB so we must make clones.
156 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
158 const struct in6_addr *saddr;
159 const struct in6_addr *daddr;
160 struct sock *sk;
161 int delivered = 0;
162 __u8 hash;
163 struct net *net;
165 saddr = &ipv6_hdr(skb)->saddr;
166 daddr = saddr + 1;
168 hash = nexthdr & (MAX_INET_PROTOS - 1);
170 read_lock(&raw_v6_hashinfo.lock);
171 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
173 if (sk == NULL)
174 goto out;
176 net = dev_net(skb->dev);
177 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
179 while (sk) {
180 int filtered;
182 delivered = 1;
183 switch (nexthdr) {
184 case IPPROTO_ICMPV6:
185 filtered = icmpv6_filter(sk, skb);
186 break;
188 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
189 case IPPROTO_MH:
191 /* XXX: To validate MH only once for each packet,
192 * this is placed here. It should be after checking
193 * xfrm policy, however it doesn't. The checking xfrm
194 * policy is placed in rawv6_rcv() because it is
195 * required for each socket.
197 mh_filter_t *filter;
199 filter = rcu_dereference(mh_filter);
200 filtered = filter ? (*filter)(sk, skb) : 0;
201 break;
203 #endif
204 default:
205 filtered = 0;
206 break;
209 if (filtered < 0)
210 break;
211 if (filtered == 0) {
212 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
214 /* Not releasing hash table! */
215 if (clone) {
216 nf_reset(clone);
217 rawv6_rcv(sk, clone);
220 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
221 IP6CB(skb)->iif);
223 out:
224 read_unlock(&raw_v6_hashinfo.lock);
225 return delivered;
228 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
230 struct sock *raw_sk;
232 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
233 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
234 raw_sk = NULL;
236 return raw_sk != NULL;
239 /* This cleans up af_inet6 a bit. -DaveM */
240 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
242 struct inet_sock *inet = inet_sk(sk);
243 struct ipv6_pinfo *np = inet6_sk(sk);
244 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
245 __be32 v4addr = 0;
246 int addr_type;
247 int err;
249 if (addr_len < SIN6_LEN_RFC2133)
250 return -EINVAL;
251 addr_type = ipv6_addr_type(&addr->sin6_addr);
253 /* Raw sockets are IPv6 only */
254 if (addr_type == IPV6_ADDR_MAPPED)
255 return -EADDRNOTAVAIL;
257 lock_sock(sk);
259 err = -EINVAL;
260 if (sk->sk_state != TCP_CLOSE)
261 goto out;
263 rcu_read_lock();
264 /* Check if the address belongs to the host. */
265 if (addr_type != IPV6_ADDR_ANY) {
266 struct net_device *dev = NULL;
268 if (addr_type & IPV6_ADDR_LINKLOCAL) {
269 if (addr_len >= sizeof(struct sockaddr_in6) &&
270 addr->sin6_scope_id) {
271 /* Override any existing binding, if another
272 * one is supplied by user.
274 sk->sk_bound_dev_if = addr->sin6_scope_id;
277 /* Binding to link-local address requires an interface */
278 if (!sk->sk_bound_dev_if)
279 goto out_unlock;
281 err = -ENODEV;
282 dev = dev_get_by_index_rcu(sock_net(sk),
283 sk->sk_bound_dev_if);
284 if (!dev)
285 goto out_unlock;
288 /* ipv4 addr of the socket is invalid. Only the
289 * unspecified and mapped address have a v4 equivalent.
291 v4addr = LOOPBACK4_IPV6;
292 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
293 err = -EADDRNOTAVAIL;
294 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
295 dev, 0)) {
296 goto out_unlock;
301 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
302 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
303 if (!(addr_type & IPV6_ADDR_MULTICAST))
304 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
305 err = 0;
306 out_unlock:
307 rcu_read_unlock();
308 out:
309 release_sock(sk);
310 return err;
313 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
314 struct inet6_skb_parm *opt,
315 u8 type, u8 code, int offset, __be32 info)
317 struct inet_sock *inet = inet_sk(sk);
318 struct ipv6_pinfo *np = inet6_sk(sk);
319 int err;
320 int harderr;
322 /* Report error on raw socket, if:
323 1. User requested recverr.
324 2. Socket is connected (otherwise the error indication
325 is useless without recverr and error is hard.
327 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
328 return;
330 harderr = icmpv6_err_convert(type, code, &err);
331 if (type == ICMPV6_PKT_TOOBIG)
332 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
334 if (np->recverr) {
335 u8 *payload = skb->data;
336 if (!inet->hdrincl)
337 payload += offset;
338 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
341 if (np->recverr || harderr) {
342 sk->sk_err = err;
343 sk->sk_error_report(sk);
347 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
348 u8 type, u8 code, int inner_offset, __be32 info)
350 struct sock *sk;
351 int hash;
352 const struct in6_addr *saddr, *daddr;
353 struct net *net;
355 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
357 read_lock(&raw_v6_hashinfo.lock);
358 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
359 if (sk != NULL) {
360 /* Note: ipv6_hdr(skb) != skb->data */
361 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
362 saddr = &ip6h->saddr;
363 daddr = &ip6h->daddr;
364 net = dev_net(skb->dev);
366 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
367 IP6CB(skb)->iif))) {
368 rawv6_err(sk, skb, NULL, type, code,
369 inner_offset, info);
370 sk = sk_next(sk);
373 read_unlock(&raw_v6_hashinfo.lock);
376 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
378 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
379 skb_checksum_complete(skb)) {
380 atomic_inc(&sk->sk_drops);
381 kfree_skb(skb);
382 return NET_RX_DROP;
385 /* Charge it to the socket. */
386 if (ip_queue_rcv_skb(sk, skb) < 0) {
387 kfree_skb(skb);
388 return NET_RX_DROP;
391 return 0;
395 * This is next to useless...
396 * if we demultiplex in network layer we don't need the extra call
397 * just to queue the skb...
398 * maybe we could have the network decide upon a hint if it
399 * should call raw_rcv for demultiplexing
401 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
403 struct inet_sock *inet = inet_sk(sk);
404 struct raw6_sock *rp = raw6_sk(sk);
406 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
407 atomic_inc(&sk->sk_drops);
408 kfree_skb(skb);
409 return NET_RX_DROP;
412 if (!rp->checksum)
413 skb->ip_summed = CHECKSUM_UNNECESSARY;
415 if (skb->ip_summed == CHECKSUM_COMPLETE) {
416 skb_postpull_rcsum(skb, skb_network_header(skb),
417 skb_network_header_len(skb));
418 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
419 &ipv6_hdr(skb)->daddr,
420 skb->len, inet->inet_num, skb->csum))
421 skb->ip_summed = CHECKSUM_UNNECESSARY;
423 if (!skb_csum_unnecessary(skb))
424 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
425 &ipv6_hdr(skb)->daddr,
426 skb->len,
427 inet->inet_num, 0));
429 if (inet->hdrincl) {
430 if (skb_checksum_complete(skb)) {
431 atomic_inc(&sk->sk_drops);
432 kfree_skb(skb);
433 return NET_RX_DROP;
437 rawv6_rcv_skb(sk, skb);
438 return 0;
443 * This should be easy, if there is something there
444 * we return it, otherwise we block.
447 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
448 struct msghdr *msg, size_t len,
449 int noblock, int flags, int *addr_len)
451 struct ipv6_pinfo *np = inet6_sk(sk);
452 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
453 struct sk_buff *skb;
454 size_t copied;
455 int err;
457 if (flags & MSG_OOB)
458 return -EOPNOTSUPP;
460 if (addr_len)
461 *addr_len=sizeof(*sin6);
463 if (flags & MSG_ERRQUEUE)
464 return ipv6_recv_error(sk, msg, len);
466 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
467 return ipv6_recv_rxpmtu(sk, msg, len);
469 skb = skb_recv_datagram(sk, flags, noblock, &err);
470 if (!skb)
471 goto out;
473 copied = skb->len;
474 if (copied > len) {
475 copied = len;
476 msg->msg_flags |= MSG_TRUNC;
479 if (skb_csum_unnecessary(skb)) {
480 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
481 } else if (msg->msg_flags&MSG_TRUNC) {
482 if (__skb_checksum_complete(skb))
483 goto csum_copy_err;
484 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
485 } else {
486 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
487 if (err == -EINVAL)
488 goto csum_copy_err;
490 if (err)
491 goto out_free;
493 /* Copy the address. */
494 if (sin6) {
495 sin6->sin6_family = AF_INET6;
496 sin6->sin6_port = 0;
497 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
498 sin6->sin6_flowinfo = 0;
499 sin6->sin6_scope_id = 0;
500 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
501 sin6->sin6_scope_id = IP6CB(skb)->iif;
504 sock_recv_ts_and_drops(msg, sk, skb);
506 if (np->rxopt.all)
507 datagram_recv_ctl(sk, msg, skb);
509 err = copied;
510 if (flags & MSG_TRUNC)
511 err = skb->len;
513 out_free:
514 skb_free_datagram(sk, skb);
515 out:
516 return err;
518 csum_copy_err:
519 skb_kill_datagram(sk, skb, flags);
521 /* Error for blocking case is chosen to masquerade
522 as some normal condition.
524 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
525 goto out;
528 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
529 struct raw6_sock *rp)
531 struct sk_buff *skb;
532 int err = 0;
533 int offset;
534 int len;
535 int total_len;
536 __wsum tmp_csum;
537 __sum16 csum;
539 if (!rp->checksum)
540 goto send;
542 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
543 goto out;
545 offset = rp->offset;
546 total_len = inet_sk(sk)->cork.base.length;
547 if (offset >= total_len - 1) {
548 err = -EINVAL;
549 ip6_flush_pending_frames(sk);
550 goto out;
553 /* should be check HW csum miyazawa */
554 if (skb_queue_len(&sk->sk_write_queue) == 1) {
556 * Only one fragment on the socket.
558 tmp_csum = skb->csum;
559 } else {
560 struct sk_buff *csum_skb = NULL;
561 tmp_csum = 0;
563 skb_queue_walk(&sk->sk_write_queue, skb) {
564 tmp_csum = csum_add(tmp_csum, skb->csum);
566 if (csum_skb)
567 continue;
569 len = skb->len - skb_transport_offset(skb);
570 if (offset >= len) {
571 offset -= len;
572 continue;
575 csum_skb = skb;
578 skb = csum_skb;
581 offset += skb_transport_offset(skb);
582 if (skb_copy_bits(skb, offset, &csum, 2))
583 BUG();
585 /* in case cksum was not initialized */
586 if (unlikely(csum))
587 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
589 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
590 total_len, fl6->flowi6_proto, tmp_csum);
592 if (csum == 0 && fl6->flowi6_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 flowi6 *fl6, 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, fl6, 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 flowi6 *fl6, 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 (fl6->flowi6_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(fl6->fl6_icmp_type, type) ||
697 get_user(fl6->fl6_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(fl6->fl6_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 flowi6 fl6;
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(&fl6, 0, sizeof(fl6));
760 fl6.flowi6_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 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
783 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
784 flowlabel = fl6_sock_lookup(sk, 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 fl6.flowi6_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 fl6.flowlabel = np->flow_label;
812 if (fl6.flowi6_oif == 0)
813 fl6.flowi6_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), sk, msg, &fl6, opt,
821 &hlimit, &tclass, &dontfrag);
822 if (err < 0) {
823 fl6_sock_release(flowlabel);
824 return err;
826 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
827 flowlabel = fl6_sock_lookup(sk, 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 fl6.flowi6_proto = proto;
841 err = rawv6_probe_proto_opt(&fl6, msg);
842 if (err)
843 goto out;
845 if (!ipv6_addr_any(daddr))
846 ipv6_addr_copy(&fl6.daddr, daddr);
847 else
848 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
849 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
850 ipv6_addr_copy(&fl6.saddr, &np->saddr);
852 final_p = fl6_update_dst(&fl6, opt, &final);
854 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
855 fl6.flowi6_oif = np->mcast_oif;
856 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
858 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
859 if (IS_ERR(dst)) {
860 err = PTR_ERR(dst);
861 goto out;
863 if (hlimit < 0) {
864 if (ipv6_addr_is_multicast(&fl6.daddr))
865 hlimit = np->mcast_hops;
866 else
867 hlimit = np->hop_limit;
868 if (hlimit < 0)
869 hlimit = ip6_dst_hoplimit(dst);
872 if (tclass < 0)
873 tclass = np->tclass;
875 if (dontfrag < 0)
876 dontfrag = np->dontfrag;
878 if (msg->msg_flags&MSG_CONFIRM)
879 goto do_confirm;
881 back_from_confirm:
882 if (inet->hdrincl)
883 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags);
884 else {
885 lock_sock(sk);
886 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
887 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
888 msg->msg_flags, dontfrag);
890 if (err)
891 ip6_flush_pending_frames(sk);
892 else if (!(msg->msg_flags & MSG_MORE))
893 err = rawv6_push_pending_frames(sk, &fl6, rp);
894 release_sock(sk);
896 done:
897 dst_release(dst);
898 out:
899 fl6_sock_release(flowlabel);
900 return err<0?err:len;
901 do_confirm:
902 dst_confirm(dst);
903 if (!(msg->msg_flags & MSG_PROBE) || len)
904 goto back_from_confirm;
905 err = 0;
906 goto done;
909 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
910 char __user *optval, int optlen)
912 switch (optname) {
913 case ICMPV6_FILTER:
914 if (optlen > sizeof(struct icmp6_filter))
915 optlen = sizeof(struct icmp6_filter);
916 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
917 return -EFAULT;
918 return 0;
919 default:
920 return -ENOPROTOOPT;
923 return 0;
926 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
927 char __user *optval, int __user *optlen)
929 int len;
931 switch (optname) {
932 case ICMPV6_FILTER:
933 if (get_user(len, optlen))
934 return -EFAULT;
935 if (len < 0)
936 return -EINVAL;
937 if (len > sizeof(struct icmp6_filter))
938 len = sizeof(struct icmp6_filter);
939 if (put_user(len, optlen))
940 return -EFAULT;
941 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
942 return -EFAULT;
943 return 0;
944 default:
945 return -ENOPROTOOPT;
948 return 0;
952 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
953 char __user *optval, unsigned int optlen)
955 struct raw6_sock *rp = raw6_sk(sk);
956 int val;
958 if (get_user(val, (int __user *)optval))
959 return -EFAULT;
961 switch (optname) {
962 case IPV6_CHECKSUM:
963 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
964 level == IPPROTO_IPV6) {
966 * RFC3542 tells that IPV6_CHECKSUM socket
967 * option in the IPPROTO_IPV6 level is not
968 * allowed on ICMPv6 sockets.
969 * If you want to set it, use IPPROTO_RAW
970 * level IPV6_CHECKSUM socket option
971 * (Linux extension).
973 return -EINVAL;
976 /* You may get strange result with a positive odd offset;
977 RFC2292bis agrees with me. */
978 if (val > 0 && (val&1))
979 return -EINVAL;
980 if (val < 0) {
981 rp->checksum = 0;
982 } else {
983 rp->checksum = 1;
984 rp->offset = val;
987 return 0;
989 default:
990 return -ENOPROTOOPT;
994 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
995 char __user *optval, unsigned int optlen)
997 switch (level) {
998 case SOL_RAW:
999 break;
1001 case SOL_ICMPV6:
1002 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1003 return -EOPNOTSUPP;
1004 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1005 case SOL_IPV6:
1006 if (optname == IPV6_CHECKSUM)
1007 break;
1008 default:
1009 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1012 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1015 #ifdef CONFIG_COMPAT
1016 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1017 char __user *optval, unsigned int optlen)
1019 switch (level) {
1020 case SOL_RAW:
1021 break;
1022 case SOL_ICMPV6:
1023 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1024 return -EOPNOTSUPP;
1025 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1026 case SOL_IPV6:
1027 if (optname == IPV6_CHECKSUM)
1028 break;
1029 default:
1030 return compat_ipv6_setsockopt(sk, level, optname,
1031 optval, optlen);
1033 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1035 #endif
1037 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1038 char __user *optval, int __user *optlen)
1040 struct raw6_sock *rp = raw6_sk(sk);
1041 int val, len;
1043 if (get_user(len,optlen))
1044 return -EFAULT;
1046 switch (optname) {
1047 case IPV6_CHECKSUM:
1049 * We allow getsockopt() for IPPROTO_IPV6-level
1050 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1051 * since RFC3542 is silent about it.
1053 if (rp->checksum == 0)
1054 val = -1;
1055 else
1056 val = rp->offset;
1057 break;
1059 default:
1060 return -ENOPROTOOPT;
1063 len = min_t(unsigned int, sizeof(int), len);
1065 if (put_user(len, optlen))
1066 return -EFAULT;
1067 if (copy_to_user(optval,&val,len))
1068 return -EFAULT;
1069 return 0;
1072 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1073 char __user *optval, int __user *optlen)
1075 switch (level) {
1076 case SOL_RAW:
1077 break;
1079 case SOL_ICMPV6:
1080 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1081 return -EOPNOTSUPP;
1082 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1083 case SOL_IPV6:
1084 if (optname == IPV6_CHECKSUM)
1085 break;
1086 default:
1087 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1090 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1093 #ifdef CONFIG_COMPAT
1094 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1095 char __user *optval, int __user *optlen)
1097 switch (level) {
1098 case SOL_RAW:
1099 break;
1100 case SOL_ICMPV6:
1101 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1102 return -EOPNOTSUPP;
1103 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1104 case SOL_IPV6:
1105 if (optname == IPV6_CHECKSUM)
1106 break;
1107 default:
1108 return compat_ipv6_getsockopt(sk, level, optname,
1109 optval, optlen);
1111 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1113 #endif
1115 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1117 switch (cmd) {
1118 case SIOCOUTQ: {
1119 int amount = sk_wmem_alloc_get(sk);
1121 return put_user(amount, (int __user *)arg);
1123 case SIOCINQ: {
1124 struct sk_buff *skb;
1125 int amount = 0;
1127 spin_lock_bh(&sk->sk_receive_queue.lock);
1128 skb = skb_peek(&sk->sk_receive_queue);
1129 if (skb != NULL)
1130 amount = skb->tail - skb->transport_header;
1131 spin_unlock_bh(&sk->sk_receive_queue.lock);
1132 return put_user(amount, (int __user *)arg);
1135 default:
1136 #ifdef CONFIG_IPV6_MROUTE
1137 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1138 #else
1139 return -ENOIOCTLCMD;
1140 #endif
1144 #ifdef CONFIG_COMPAT
1145 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1147 switch (cmd) {
1148 case SIOCOUTQ:
1149 case SIOCINQ:
1150 return -ENOIOCTLCMD;
1151 default:
1152 #ifdef CONFIG_IPV6_MROUTE
1153 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1154 #else
1155 return -ENOIOCTLCMD;
1156 #endif
1159 #endif
1161 static void rawv6_close(struct sock *sk, long timeout)
1163 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1164 ip6_ra_control(sk, -1);
1165 ip6mr_sk_done(sk);
1166 sk_common_release(sk);
1169 static void raw6_destroy(struct sock *sk)
1171 lock_sock(sk);
1172 ip6_flush_pending_frames(sk);
1173 release_sock(sk);
1175 inet6_destroy_sock(sk);
1178 static int rawv6_init_sk(struct sock *sk)
1180 struct raw6_sock *rp = raw6_sk(sk);
1182 switch (inet_sk(sk)->inet_num) {
1183 case IPPROTO_ICMPV6:
1184 rp->checksum = 1;
1185 rp->offset = 2;
1186 break;
1187 case IPPROTO_MH:
1188 rp->checksum = 1;
1189 rp->offset = 4;
1190 break;
1191 default:
1192 break;
1194 return 0;
1197 struct proto rawv6_prot = {
1198 .name = "RAWv6",
1199 .owner = THIS_MODULE,
1200 .close = rawv6_close,
1201 .destroy = raw6_destroy,
1202 .connect = ip6_datagram_connect,
1203 .disconnect = udp_disconnect,
1204 .ioctl = rawv6_ioctl,
1205 .init = rawv6_init_sk,
1206 .setsockopt = rawv6_setsockopt,
1207 .getsockopt = rawv6_getsockopt,
1208 .sendmsg = rawv6_sendmsg,
1209 .recvmsg = rawv6_recvmsg,
1210 .bind = rawv6_bind,
1211 .backlog_rcv = rawv6_rcv_skb,
1212 .hash = raw_hash_sk,
1213 .unhash = raw_unhash_sk,
1214 .obj_size = sizeof(struct raw6_sock),
1215 .h.raw_hash = &raw_v6_hashinfo,
1216 #ifdef CONFIG_COMPAT
1217 .compat_setsockopt = compat_rawv6_setsockopt,
1218 .compat_getsockopt = compat_rawv6_getsockopt,
1219 .compat_ioctl = compat_rawv6_ioctl,
1220 #endif
1223 #ifdef CONFIG_PROC_FS
1224 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1226 struct ipv6_pinfo *np = inet6_sk(sp);
1227 const struct in6_addr *dest, *src;
1228 __u16 destp, srcp;
1230 dest = &np->daddr;
1231 src = &np->rcv_saddr;
1232 destp = 0;
1233 srcp = inet_sk(sp)->inet_num;
1234 seq_printf(seq,
1235 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1236 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1238 src->s6_addr32[0], src->s6_addr32[1],
1239 src->s6_addr32[2], src->s6_addr32[3], srcp,
1240 dest->s6_addr32[0], dest->s6_addr32[1],
1241 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1242 sp->sk_state,
1243 sk_wmem_alloc_get(sp),
1244 sk_rmem_alloc_get(sp),
1245 0, 0L, 0,
1246 sock_i_uid(sp), 0,
1247 sock_i_ino(sp),
1248 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1251 static int raw6_seq_show(struct seq_file *seq, void *v)
1253 if (v == SEQ_START_TOKEN)
1254 seq_printf(seq,
1255 " sl "
1256 "local_address "
1257 "remote_address "
1258 "st tx_queue rx_queue tr tm->when retrnsmt"
1259 " uid timeout inode ref pointer drops\n");
1260 else
1261 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1262 return 0;
1265 static const struct seq_operations raw6_seq_ops = {
1266 .start = raw_seq_start,
1267 .next = raw_seq_next,
1268 .stop = raw_seq_stop,
1269 .show = raw6_seq_show,
1272 static int raw6_seq_open(struct inode *inode, struct file *file)
1274 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1277 static const struct file_operations raw6_seq_fops = {
1278 .owner = THIS_MODULE,
1279 .open = raw6_seq_open,
1280 .read = seq_read,
1281 .llseek = seq_lseek,
1282 .release = seq_release_net,
1285 static int __net_init raw6_init_net(struct net *net)
1287 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1288 return -ENOMEM;
1290 return 0;
1293 static void __net_exit raw6_exit_net(struct net *net)
1295 proc_net_remove(net, "raw6");
1298 static struct pernet_operations raw6_net_ops = {
1299 .init = raw6_init_net,
1300 .exit = raw6_exit_net,
1303 int __init raw6_proc_init(void)
1305 return register_pernet_subsys(&raw6_net_ops);
1308 void raw6_proc_exit(void)
1310 unregister_pernet_subsys(&raw6_net_ops);
1312 #endif /* CONFIG_PROC_FS */
1314 /* Same as inet6_dgram_ops, sans udp_poll. */
1315 static const struct proto_ops inet6_sockraw_ops = {
1316 .family = PF_INET6,
1317 .owner = THIS_MODULE,
1318 .release = inet6_release,
1319 .bind = inet6_bind,
1320 .connect = inet_dgram_connect, /* ok */
1321 .socketpair = sock_no_socketpair, /* a do nothing */
1322 .accept = sock_no_accept, /* a do nothing */
1323 .getname = inet6_getname,
1324 .poll = datagram_poll, /* ok */
1325 .ioctl = inet6_ioctl, /* must change */
1326 .listen = sock_no_listen, /* ok */
1327 .shutdown = inet_shutdown, /* ok */
1328 .setsockopt = sock_common_setsockopt, /* ok */
1329 .getsockopt = sock_common_getsockopt, /* ok */
1330 .sendmsg = inet_sendmsg, /* ok */
1331 .recvmsg = sock_common_recvmsg, /* ok */
1332 .mmap = sock_no_mmap,
1333 .sendpage = sock_no_sendpage,
1334 #ifdef CONFIG_COMPAT
1335 .compat_setsockopt = compat_sock_common_setsockopt,
1336 .compat_getsockopt = compat_sock_common_getsockopt,
1337 #endif
1340 static struct inet_protosw rawv6_protosw = {
1341 .type = SOCK_RAW,
1342 .protocol = IPPROTO_IP, /* wild card */
1343 .prot = &rawv6_prot,
1344 .ops = &inet6_sockraw_ops,
1345 .no_check = UDP_CSUM_DEFAULT,
1346 .flags = INET_PROTOSW_REUSE,
1349 int __init rawv6_init(void)
1351 int ret;
1353 ret = inet6_register_protosw(&rawv6_protosw);
1354 if (ret)
1355 goto out;
1356 out:
1357 return ret;
1360 void rawv6_exit(void)
1362 inet6_unregister_protosw(&rawv6_protosw);