?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 3.0.85
[linux/fpc-iii.git] / net / ipv6 / raw.c
blobfb812a638f8f87ffbec96690b91b1765745328f6
1 /*
2 * RAW sockets for IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Adapted from linux/net/ipv4/raw.c
9 *
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
14 *
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.
19 */
20
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>
37
38 #include <net/net_namespace.h>
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
42
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>
57
58 #include <net/raw.h>
59 #include <net/rawv6.h>
60 #include <net/xfrm.h>
61
62 #include <linux/proc_fs.h>
63 #include <linux/seq_file.h>
64
65 static struct raw_hashinfo raw_v6_hashinfo = {
66 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
67 };
68
69 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
70 unsigned short num, const struct in6_addr *loc_addr,
71 const struct in6_addr *rmt_addr, int dif)
72 {
73 struct hlist_node *node;
74 int is_multicast = ipv6_addr_is_multicast(loc_addr);
75
76 sk_for_each_from(sk, node)
77 if (inet_sk(sk)->inet_num == num) {
78 struct ipv6_pinfo *np = inet6_sk(sk);
79
80 if (!net_eq(sock_net(sk), net))
81 continue;
82
83 if (!ipv6_addr_any(&np->daddr) &&
84 !ipv6_addr_equal(&np->daddr, rmt_addr))
85 continue;
86
87 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
88 continue;
89
90 if (!ipv6_addr_any(&np->rcv_saddr)) {
91 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
92 goto found;
93 if (is_multicast &&
94 inet6_mc_check(sk, loc_addr, rmt_addr))
95 goto found;
96 continue;
97 }
98 goto found;
99 }
100 sk = NULL;
101 found:
102 return sk;
103 }
104
105 /*
106 * 0 - deliver
107 * 1 - block
108 */
109 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
110 {
111 struct icmp6hdr *_hdr;
112 const struct icmp6hdr *hdr;
113
114 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
115 sizeof(_hdr), &_hdr);
116 if (hdr) {
117 const __u32 *data = &raw6_sk(sk)->filter.data[0];
118 unsigned int type = hdr->icmp6_type;
119
120 return (data[type >> 5] & (1U << (type & 31))) != 0;
121 }
122 return 1;
123 }
124
125 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
126 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
127
128 static mh_filter_t __rcu *mh_filter __read_mostly;
129
130 int rawv6_mh_filter_register(mh_filter_t filter)
131 {
132 rcu_assign_pointer(mh_filter, filter);
133 return 0;
134 }
135 EXPORT_SYMBOL(rawv6_mh_filter_register);
136
137 int rawv6_mh_filter_unregister(mh_filter_t filter)
138 {
139 rcu_assign_pointer(mh_filter, NULL);
140 synchronize_rcu();
141 return 0;
142 }
143 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
144
145 #endif
146
147 /*
148 * demultiplex raw sockets.
149 * (should consider queueing the skb in the sock receive_queue
150 * without calling rawv6.c)
151 *
152 * Caller owns SKB so we must make clones.
153 */
154 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
155 {
156 const struct in6_addr *saddr;
157 const struct in6_addr *daddr;
158 struct sock *sk;
159 int delivered = 0;
160 __u8 hash;
161 struct net *net;
162
163 saddr = &ipv6_hdr(skb)->saddr;
164 daddr = saddr + 1;
165
166 hash = nexthdr & (MAX_INET_PROTOS - 1);
167
168 read_lock(&raw_v6_hashinfo.lock);
169 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
170
171 if (sk == NULL)
172 goto out;
173
174 net = dev_net(skb->dev);
175 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
176
177 while (sk) {
178 int filtered;
179
180 delivered = 1;
181 switch (nexthdr) {
182 case IPPROTO_ICMPV6:
183 filtered = icmpv6_filter(sk, skb);
184 break;
185
186 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
187 case IPPROTO_MH:
188 {
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.
194 */
195 mh_filter_t *filter;
196
197 filter = rcu_dereference(mh_filter);
198 filtered = filter ? (*filter)(sk, skb) : 0;
199 break;
200 }
201 #endif
202 default:
203 filtered = 0;
204 break;
205 }
206
207 if (filtered < 0)
208 break;
209 if (filtered == 0) {
210 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
211
212 /* Not releasing hash table! */
213 if (clone) {
214 nf_reset(clone);
215 rawv6_rcv(sk, clone);
216 }
217 }
218 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
219 IP6CB(skb)->iif);
220 }
221 out:
222 read_unlock(&raw_v6_hashinfo.lock);
223 return delivered;
224 }
225
226 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
227 {
228 struct sock *raw_sk;
229
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;
233
234 return raw_sk != NULL;
235 }
236
237 /* This cleans up af_inet6 a bit. -DaveM */
238 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
239 {
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;
246
247 if (addr_len < SIN6_LEN_RFC2133)
248 return -EINVAL;
249 addr_type = ipv6_addr_type(&addr->sin6_addr);
250
251 /* Raw sockets are IPv6 only */
252 if (addr_type == IPV6_ADDR_MAPPED)
253 return -EADDRNOTAVAIL;
254
255 lock_sock(sk);
256
257 err = -EINVAL;
258 if (sk->sk_state != TCP_CLOSE)
259 goto out;
260
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;
265
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.
271 */
272 sk->sk_bound_dev_if = addr->sin6_scope_id;
273 }
274
275 /* Binding to link-local address requires an interface */
276 if (!sk->sk_bound_dev_if)
277 goto out_unlock;
278
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;
284 }
285
286 /* ipv4 addr of the socket is invalid. Only the
287 * unspecified and mapped address have a v4 equivalent.
288 */
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;
295 }
296 }
297 }
298
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;
309 }
310
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)
314 {
315 struct inet_sock *inet = inet_sk(sk);
316 struct ipv6_pinfo *np = inet6_sk(sk);
317 int err;
318 int harderr;
319
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.
324 */
325 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
326 return;
327
328 harderr = icmpv6_err_convert(type, code, &err);
329 if (type == ICMPV6_PKT_TOOBIG)
330 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
331
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);
337 }
338
339 if (np->recverr || harderr) {
340 sk->sk_err = err;
341 sk->sk_error_report(sk);
342 }
343 }
344
345 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
346 u8 type, u8 code, int inner_offset, __be32 info)
347 {
348 struct sock *sk;
349 int hash;
350 const struct in6_addr *saddr, *daddr;
351 struct net *net;
352
353 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
354
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 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
360 saddr = &ip6h->saddr;
361 daddr = &ip6h->daddr;
362 net = dev_net(skb->dev);
363
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);
369 }
370 }
371 read_unlock(&raw_v6_hashinfo.lock);
372 }
373
374 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
375 {
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;
381 }
382
383 /* Charge it to the socket. */
384 if (ip_queue_rcv_skb(sk, skb) < 0) {
385 kfree_skb(skb);
386 return NET_RX_DROP;
387 }
388
389 return 0;
390 }
391
392 /*
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
398 */
399 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
400 {
401 struct inet_sock *inet = inet_sk(sk);
402 struct raw6_sock *rp = raw6_sk(sk);
403
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;
408 }
409
410 if (!rp->checksum)
411 skb->ip_summed = CHECKSUM_UNNECESSARY;
412
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;
420 }
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));
426
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;
432 }
433 }
434
435 rawv6_rcv_skb(sk, skb);
436 return 0;
437 }
438
439
440 /*
441 * This should be easy, if there is something there
442 * we return it, otherwise we block.
443 */
444
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)
448 {
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;
454
455 if (flags & MSG_OOB)
456 return -EOPNOTSUPP;
457
458 if (addr_len)
459 *addr_len=sizeof(*sin6);
460
461 if (flags & MSG_ERRQUEUE)
462 return ipv6_recv_error(sk, msg, len);
463
464 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
465 return ipv6_recv_rxpmtu(sk, msg, len);
466
467 skb = skb_recv_datagram(sk, flags, noblock, &err);
468 if (!skb)
469 goto out;
470
471 copied = skb->len;
472 if (copied > len) {
473 copied = len;
474 msg->msg_flags |= MSG_TRUNC;
475 }
476
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;
487 }
488 if (err)
489 goto out_free;
490
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;
500 }
501
502 sock_recv_ts_and_drops(msg, sk, skb);
503
504 if (np->rxopt.all)
505 datagram_recv_ctl(sk, msg, skb);
506
507 err = copied;
508 if (flags & MSG_TRUNC)
509 err = skb->len;
510
511 out_free:
512 skb_free_datagram(sk, skb);
513 out:
514 return err;
515
516 csum_copy_err:
517 skb_kill_datagram(sk, skb, flags);
518
519 /* Error for blocking case is chosen to masquerade
520 as some normal condition.
521 */
522 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
523 goto out;
524 }
525
526 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
527 struct raw6_sock *rp)
528 {
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;
536
537 if (!rp->checksum)
538 goto send;
539
540 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
541 goto out;
542
543 offset = rp->offset;
544 total_len = inet_sk(sk)->cork.base.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;
550 }
551
552 /* should be check HW csum miyazawa */
553 if (skb_queue_len(&sk->sk_write_queue) == 1) {
554 /*
555 * Only one fragment on the socket.
556 */
557 tmp_csum = skb->csum;
558 } else {
559 struct sk_buff *csum_skb = NULL;
560 tmp_csum = 0;
561
562 skb_queue_walk(&sk->sk_write_queue, skb) {
563 tmp_csum = csum_add(tmp_csum, skb->csum);
564
565 if (csum_skb)
566 continue;
567
568 len = skb->len - skb_transport_offset(skb);
569 if (offset >= len) {
570 offset -= len;
571 continue;
572 }
573
574 csum_skb = skb;
575 }
576
577 skb = csum_skb;
578 }
579
580 offset += skb_transport_offset(skb);
581 if (skb_copy_bits(skb, offset, &csum, 2))
582 BUG();
583
584 /* in case cksum was not initialized */
585 if (unlikely(csum))
586 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
587
588 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
589 total_len, fl6->flowi6_proto, tmp_csum);
590
591 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
592 csum = CSUM_MANGLED_0;
593
594 if (skb_store_bits(skb, offset, &csum, 2))
595 BUG();
596
597 send:
598 err = ip6_push_pending_frames(sk);
599 out:
600 return err;
601 }
602
603 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
604 struct flowi6 *fl6, struct dst_entry **dstp,
605 unsigned int flags)
606 {
607 struct ipv6_pinfo *np = inet6_sk(sk);
608 struct ipv6hdr *iph;
609 struct sk_buff *skb;
610 int err;
611 struct rt6_info *rt = (struct rt6_info *)*dstp;
612
613 if (length > rt->dst.dev->mtu) {
614 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
615 return -EMSGSIZE;
616 }
617 if (flags&MSG_PROBE)
618 goto out;
619
620 skb = sock_alloc_send_skb(sk,
621 length + LL_ALLOCATED_SPACE(rt->dst.dev) + 15,
622 flags & MSG_DONTWAIT, &err);
623 if (skb == NULL)
624 goto error;
625 skb_reserve(skb, LL_RESERVED_SPACE(rt->dst.dev));
626
627 skb->priority = sk->sk_priority;
628 skb->mark = sk->sk_mark;
629 skb_dst_set(skb, &rt->dst);
630 *dstp = NULL;
631
632 skb_put(skb, length);
633 skb_reset_network_header(skb);
634 iph = ipv6_hdr(skb);
635
636 skb->ip_summed = CHECKSUM_NONE;
637
638 skb->transport_header = skb->network_header;
639 err = memcpy_fromiovecend((void *)iph, from, 0, length);
640 if (err)
641 goto error_fault;
642
643 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
644 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
645 rt->dst.dev, dst_output);
646 if (err > 0)
647 err = net_xmit_errno(err);
648 if (err)
649 goto error;
650 out:
651 return 0;
652
653 error_fault:
654 err = -EFAULT;
655 kfree_skb(skb);
656 error:
657 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
658 if (err == -ENOBUFS && !np->recverr)
659 err = 0;
660 return err;
661 }
662
663 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg)
664 {
665 struct iovec *iov;
666 u8 __user *type = NULL;
667 u8 __user *code = NULL;
668 u8 len = 0;
669 int probed = 0;
670 int i;
671
672 if (!msg->msg_iov)
673 return 0;
674
675 for (i = 0; i < msg->msg_iovlen; i++) {
676 iov = &msg->msg_iov[i];
677 if (!iov)
678 continue;
679
680 switch (fl6->flowi6_proto) {
681 case IPPROTO_ICMPV6:
682 /* check if one-byte field is readable or not. */
683 if (iov->iov_base && iov->iov_len < 1)
684 break;
685
686 if (!type) {
687 type = iov->iov_base;
688 /* check if code field is readable or not. */
689 if (iov->iov_len > 1)
690 code = type + 1;
691 } else if (!code)
692 code = iov->iov_base;
693
694 if (type && code) {
695 if (get_user(fl6->fl6_icmp_type, type) ||
696 get_user(fl6->fl6_icmp_code, code))
697 return -EFAULT;
698 probed = 1;
699 }
700 break;
701 case IPPROTO_MH:
702 if (iov->iov_base && iov->iov_len < 1)
703 break;
704 /* check if type field is readable or not. */
705 if (iov->iov_len > 2 - len) {
706 u8 __user *p = iov->iov_base;
707 if (get_user(fl6->fl6_mh_type, &p[2 - len]))
708 return -EFAULT;
709 probed = 1;
710 } else
711 len += iov->iov_len;
712
713 break;
714 default:
715 probed = 1;
716 break;
717 }
718 if (probed)
719 break;
720 }
721 return 0;
722 }
723
724 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
725 struct msghdr *msg, size_t len)
726 {
727 struct ipv6_txoptions opt_space;
728 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
729 struct in6_addr *daddr, *final_p, final;
730 struct inet_sock *inet = inet_sk(sk);
731 struct ipv6_pinfo *np = inet6_sk(sk);
732 struct raw6_sock *rp = raw6_sk(sk);
733 struct ipv6_txoptions *opt = NULL;
734 struct ip6_flowlabel *flowlabel = NULL;
735 struct dst_entry *dst = NULL;
736 struct flowi6 fl6;
737 int addr_len = msg->msg_namelen;
738 int hlimit = -1;
739 int tclass = -1;
740 int dontfrag = -1;
741 u16 proto;
742 int err;
743
744 /* Rough check on arithmetic overflow,
745 better check is made in ip6_append_data().
746 */
747 if (len > INT_MAX)
748 return -EMSGSIZE;
749
750 /* Mirror BSD error message compatibility */
751 if (msg->msg_flags & MSG_OOB)
752 return -EOPNOTSUPP;
753
754 /*
755 * Get and verify the address.
756 */
757 memset(&fl6, 0, sizeof(fl6));
758
759 fl6.flowi6_mark = sk->sk_mark;
760
761 if (sin6) {
762 if (addr_len < SIN6_LEN_RFC2133)
763 return -EINVAL;
764
765 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
766 return -EAFNOSUPPORT;
767
768 /* port is the proto value [0..255] carried in nexthdr */
769 proto = ntohs(sin6->sin6_port);
770
771 if (!proto)
772 proto = inet->inet_num;
773 else if (proto != inet->inet_num)
774 return -EINVAL;
775
776 if (proto > 255)
777 return -EINVAL;
778
779 daddr = &sin6->sin6_addr;
780 if (np->sndflow) {
781 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
782 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
783 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
784 if (flowlabel == NULL)
785 return -EINVAL;
786 daddr = &flowlabel->dst;
787 }
788 }
789
790 /*
791 * Otherwise it will be difficult to maintain
792 * sk->sk_dst_cache.
793 */
794 if (sk->sk_state == TCP_ESTABLISHED &&
795 ipv6_addr_equal(daddr, &np->daddr))
796 daddr = &np->daddr;
797
798 if (addr_len >= sizeof(struct sockaddr_in6) &&
799 sin6->sin6_scope_id &&
800 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
801 fl6.flowi6_oif = sin6->sin6_scope_id;
802 } else {
803 if (sk->sk_state != TCP_ESTABLISHED)
804 return -EDESTADDRREQ;
805
806 proto = inet->inet_num;
807 daddr = &np->daddr;
808 fl6.flowlabel = np->flow_label;
809 }
810
811 if (fl6.flowi6_oif == 0)
812 fl6.flowi6_oif = sk->sk_bound_dev_if;
813
814 if (msg->msg_controllen) {
815 opt = &opt_space;
816 memset(opt, 0, sizeof(struct ipv6_txoptions));
817 opt->tot_len = sizeof(struct ipv6_txoptions);
818
819 err = datagram_send_ctl(sock_net(sk), msg, &fl6, opt, &hlimit,
820 &tclass, &dontfrag);
821 if (err < 0) {
822 fl6_sock_release(flowlabel);
823 return err;
824 }
825 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
826 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
827 if (flowlabel == NULL)
828 return -EINVAL;
829 }
830 if (!(opt->opt_nflen|opt->opt_flen))
831 opt = NULL;
832 }
833 if (opt == NULL)
834 opt = np->opt;
835 if (flowlabel)
836 opt = fl6_merge_options(&opt_space, flowlabel, opt);
837 opt = ipv6_fixup_options(&opt_space, opt);
838
839 fl6.flowi6_proto = proto;
840 err = rawv6_probe_proto_opt(&fl6, msg);
841 if (err)
842 goto out;
843
844 if (!ipv6_addr_any(daddr))
845 ipv6_addr_copy(&fl6.daddr, daddr);
846 else
847 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
848 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
849 ipv6_addr_copy(&fl6.saddr, &np->saddr);
850
851 final_p = fl6_update_dst(&fl6, opt, &final);
852
853 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
854 fl6.flowi6_oif = np->mcast_oif;
855 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
856
857 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
858 if (IS_ERR(dst)) {
859 err = PTR_ERR(dst);
860 goto out;
861 }
862 if (hlimit < 0) {
863 if (ipv6_addr_is_multicast(&fl6.daddr))
864 hlimit = np->mcast_hops;
865 else
866 hlimit = np->hop_limit;
867 if (hlimit < 0)
868 hlimit = ip6_dst_hoplimit(dst);
869 }
870
871 if (tclass < 0)
872 tclass = np->tclass;
873
874 if (dontfrag < 0)
875 dontfrag = np->dontfrag;
876
877 if (msg->msg_flags&MSG_CONFIRM)
878 goto do_confirm;
879
880 back_from_confirm:
881 if (inet->hdrincl)
882 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags);
883 else {
884 lock_sock(sk);
885 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
886 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
887 msg->msg_flags, dontfrag);
888
889 if (err)
890 ip6_flush_pending_frames(sk);
891 else if (!(msg->msg_flags & MSG_MORE))
892 err = rawv6_push_pending_frames(sk, &fl6, rp);
893 release_sock(sk);
894 }
895 done:
896 dst_release(dst);
897 out:
898 fl6_sock_release(flowlabel);
899 return err<0?err:len;
900 do_confirm:
901 dst_confirm(dst);
902 if (!(msg->msg_flags & MSG_PROBE) || len)
903 goto back_from_confirm;
904 err = 0;
905 goto done;
906 }
907
908 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
909 char __user *optval, int optlen)
910 {
911 switch (optname) {
912 case ICMPV6_FILTER:
913 if (optlen > sizeof(struct icmp6_filter))
914 optlen = sizeof(struct icmp6_filter);
915 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
916 return -EFAULT;
917 return 0;
918 default:
919 return -ENOPROTOOPT;
920 }
921
922 return 0;
923 }
924
925 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
926 char __user *optval, int __user *optlen)
927 {
928 int len;
929
930 switch (optname) {
931 case ICMPV6_FILTER:
932 if (get_user(len, optlen))
933 return -EFAULT;
934 if (len < 0)
935 return -EINVAL;
936 if (len > sizeof(struct icmp6_filter))
937 len = sizeof(struct icmp6_filter);
938 if (put_user(len, optlen))
939 return -EFAULT;
940 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
941 return -EFAULT;
942 return 0;
943 default:
944 return -ENOPROTOOPT;
945 }
946
947 return 0;
948 }
949
950
951 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
952 char __user *optval, unsigned int optlen)
953 {
954 struct raw6_sock *rp = raw6_sk(sk);
955 int val;
956
957 if (get_user(val, (int __user *)optval))
958 return -EFAULT;
959
960 switch (optname) {
961 case IPV6_CHECKSUM:
962 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
963 level == IPPROTO_IPV6) {
964 /*
965 * RFC3542 tells that IPV6_CHECKSUM socket
966 * option in the IPPROTO_IPV6 level is not
967 * allowed on ICMPv6 sockets.
968 * If you want to set it, use IPPROTO_RAW
969 * level IPV6_CHECKSUM socket option
970 * (Linux extension).
971 */
972 return -EINVAL;
973 }
974
975 /* You may get strange result with a positive odd offset;
976 RFC2292bis agrees with me. */
977 if (val > 0 && (val&1))
978 return -EINVAL;
979 if (val < 0) {
980 rp->checksum = 0;
981 } else {
982 rp->checksum = 1;
983 rp->offset = val;
984 }
985
986 return 0;
987 break;
988
989 default:
990 return -ENOPROTOOPT;
991 }
992 }
993
994 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
995 char __user *optval, unsigned int optlen)
996 {
997 switch(level) {
998 case SOL_RAW:
999 break;
1000
1001 case SOL_ICMPV6:
1002 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1003 return -EOPNOTSUPP;
1004 return rawv6_seticmpfilter(sk, level, optname, optval,
1005 optlen);
1006 case SOL_IPV6:
1007 if (optname == IPV6_CHECKSUM)
1008 break;
1009 default:
1010 return ipv6_setsockopt(sk, level, optname, optval,
1011 optlen);
1012 }
1013
1014 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1015 }
1016
1017 #ifdef CONFIG_COMPAT
1018 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1019 char __user *optval, unsigned int optlen)
1020 {
1021 switch (level) {
1022 case SOL_RAW:
1023 break;
1024 case SOL_ICMPV6:
1025 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1026 return -EOPNOTSUPP;
1027 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1028 case SOL_IPV6:
1029 if (optname == IPV6_CHECKSUM)
1030 break;
1031 default:
1032 return compat_ipv6_setsockopt(sk, level, optname,
1033 optval, optlen);
1034 }
1035 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1036 }
1037 #endif
1038
1039 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1040 char __user *optval, int __user *optlen)
1041 {
1042 struct raw6_sock *rp = raw6_sk(sk);
1043 int val, len;
1044
1045 if (get_user(len,optlen))
1046 return -EFAULT;
1047
1048 switch (optname) {
1049 case IPV6_CHECKSUM:
1050 /*
1051 * We allow getsockopt() for IPPROTO_IPV6-level
1052 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1053 * since RFC3542 is silent about it.
1054 */
1055 if (rp->checksum == 0)
1056 val = -1;
1057 else
1058 val = rp->offset;
1059 break;
1060
1061 default:
1062 return -ENOPROTOOPT;
1063 }
1064
1065 len = min_t(unsigned int, sizeof(int), len);
1066
1067 if (put_user(len, optlen))
1068 return -EFAULT;
1069 if (copy_to_user(optval,&val,len))
1070 return -EFAULT;
1071 return 0;
1072 }
1073
1074 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1075 char __user *optval, int __user *optlen)
1076 {
1077 switch(level) {
1078 case SOL_RAW:
1079 break;
1080
1081 case SOL_ICMPV6:
1082 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1083 return -EOPNOTSUPP;
1084 return rawv6_geticmpfilter(sk, level, optname, optval,
1085 optlen);
1086 case SOL_IPV6:
1087 if (optname == IPV6_CHECKSUM)
1088 break;
1089 default:
1090 return ipv6_getsockopt(sk, level, optname, optval,
1091 optlen);
1092 }
1093
1094 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1095 }
1096
1097 #ifdef CONFIG_COMPAT
1098 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1099 char __user *optval, int __user *optlen)
1100 {
1101 switch (level) {
1102 case SOL_RAW:
1103 break;
1104 case SOL_ICMPV6:
1105 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1106 return -EOPNOTSUPP;
1107 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1108 case SOL_IPV6:
1109 if (optname == IPV6_CHECKSUM)
1110 break;
1111 default:
1112 return compat_ipv6_getsockopt(sk, level, optname,
1113 optval, optlen);
1114 }
1115 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1116 }
1117 #endif
1118
1119 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1120 {
1121 switch(cmd) {
1122 case SIOCOUTQ:
1123 {
1124 int amount = sk_wmem_alloc_get(sk);
1125
1126 return put_user(amount, (int __user *)arg);
1127 }
1128 case SIOCINQ:
1129 {
1130 struct sk_buff *skb;
1131 int amount = 0;
1132
1133 spin_lock_bh(&sk->sk_receive_queue.lock);
1134 skb = skb_peek(&sk->sk_receive_queue);
1135 if (skb != NULL)
1136 amount = skb->tail - skb->transport_header;
1137 spin_unlock_bh(&sk->sk_receive_queue.lock);
1138 return put_user(amount, (int __user *)arg);
1139 }
1140
1141 default:
1142 #ifdef CONFIG_IPV6_MROUTE
1143 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1144 #else
1145 return -ENOIOCTLCMD;
1146 #endif
1147 }
1148 }
1149
1150 #ifdef CONFIG_COMPAT
1151 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1152 {
1153 switch (cmd) {
1154 case SIOCOUTQ:
1155 case SIOCINQ:
1156 return -ENOIOCTLCMD;
1157 default:
1158 #ifdef CONFIG_IPV6_MROUTE
1159 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1160 #else
1161 return -ENOIOCTLCMD;
1162 #endif
1163 }
1164 }
1165 #endif
1166
1167 static void rawv6_close(struct sock *sk, long timeout)
1168 {
1169 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1170 ip6_ra_control(sk, -1);
1171 ip6mr_sk_done(sk);
1172 sk_common_release(sk);
1173 }
1174
1175 static void raw6_destroy(struct sock *sk)
1176 {
1177 lock_sock(sk);
1178 ip6_flush_pending_frames(sk);
1179 release_sock(sk);
1180
1181 inet6_destroy_sock(sk);
1182 }
1183
1184 static int rawv6_init_sk(struct sock *sk)
1185 {
1186 struct raw6_sock *rp = raw6_sk(sk);
1187
1188 switch (inet_sk(sk)->inet_num) {
1189 case IPPROTO_ICMPV6:
1190 rp->checksum = 1;
1191 rp->offset = 2;
1192 break;
1193 case IPPROTO_MH:
1194 rp->checksum = 1;
1195 rp->offset = 4;
1196 break;
1197 default:
1198 break;
1199 }
1200 return 0;
1201 }
1202
1203 struct proto rawv6_prot = {
1204 .name = "RAWv6",
1205 .owner = THIS_MODULE,
1206 .close = rawv6_close,
1207 .destroy = raw6_destroy,
1208 .connect = ip6_datagram_connect,
1209 .disconnect = udp_disconnect,
1210 .ioctl = rawv6_ioctl,
1211 .init = rawv6_init_sk,
1212 .setsockopt = rawv6_setsockopt,
1213 .getsockopt = rawv6_getsockopt,
1214 .sendmsg = rawv6_sendmsg,
1215 .recvmsg = rawv6_recvmsg,
1216 .bind = rawv6_bind,
1217 .backlog_rcv = rawv6_rcv_skb,
1218 .hash = raw_hash_sk,
1219 .unhash = raw_unhash_sk,
1220 .obj_size = sizeof(struct raw6_sock),
1221 .h.raw_hash = &raw_v6_hashinfo,
1222 #ifdef CONFIG_COMPAT
1223 .compat_setsockopt = compat_rawv6_setsockopt,
1224 .compat_getsockopt = compat_rawv6_getsockopt,
1225 .compat_ioctl = compat_rawv6_ioctl,
1226 #endif
1227 };
1228
1229 #ifdef CONFIG_PROC_FS
1230 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1231 {
1232 struct ipv6_pinfo *np = inet6_sk(sp);
1233 const struct in6_addr *dest, *src;
1234 __u16 destp, srcp;
1235
1236 dest = &np->daddr;
1237 src = &np->rcv_saddr;
1238 destp = 0;
1239 srcp = inet_sk(sp)->inet_num;
1240 seq_printf(seq,
1241 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1242 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1243 i,
1244 src->s6_addr32[0], src->s6_addr32[1],
1245 src->s6_addr32[2], src->s6_addr32[3], srcp,
1246 dest->s6_addr32[0], dest->s6_addr32[1],
1247 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1248 sp->sk_state,
1249 sk_wmem_alloc_get(sp),
1250 sk_rmem_alloc_get(sp),
1251 0, 0L, 0,
1252 sock_i_uid(sp), 0,
1253 sock_i_ino(sp),
1254 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1255 }
1256
1257 static int raw6_seq_show(struct seq_file *seq, void *v)
1258 {
1259 if (v == SEQ_START_TOKEN)
1260 seq_printf(seq,
1261 " sl "
1262 "local_address "
1263 "remote_address "
1264 "st tx_queue rx_queue tr tm->when retrnsmt"
1265 " uid timeout inode ref pointer drops\n");
1266 else
1267 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1268 return 0;
1269 }
1270
1271 static const struct seq_operations raw6_seq_ops = {
1272 .start = raw_seq_start,
1273 .next = raw_seq_next,
1274 .stop = raw_seq_stop,
1275 .show = raw6_seq_show,
1276 };
1277
1278 static int raw6_seq_open(struct inode *inode, struct file *file)
1279 {
1280 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1281 }
1282
1283 static const struct file_operations raw6_seq_fops = {
1284 .owner = THIS_MODULE,
1285 .open = raw6_seq_open,
1286 .read = seq_read,
1287 .llseek = seq_lseek,
1288 .release = seq_release_net,
1289 };
1290
1291 static int __net_init raw6_init_net(struct net *net)
1292 {
1293 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1294 return -ENOMEM;
1295
1296 return 0;
1297 }
1298
1299 static void __net_exit raw6_exit_net(struct net *net)
1300 {
1301 proc_net_remove(net, "raw6");
1302 }
1303
1304 static struct pernet_operations raw6_net_ops = {
1305 .init = raw6_init_net,
1306 .exit = raw6_exit_net,
1307 };
1308
1309 int __init raw6_proc_init(void)
1310 {
1311 return register_pernet_subsys(&raw6_net_ops);
1312 }
1313
1314 void raw6_proc_exit(void)
1315 {
1316 unregister_pernet_subsys(&raw6_net_ops);
1317 }
1318 #endif /* CONFIG_PROC_FS */
1319
1320 /* Same as inet6_dgram_ops, sans udp_poll. */
1321 static const struct proto_ops inet6_sockraw_ops = {
1322 .family = PF_INET6,
1323 .owner = THIS_MODULE,
1324 .release = inet6_release,
1325 .bind = inet6_bind,
1326 .connect = inet_dgram_connect, /* ok */
1327 .socketpair = sock_no_socketpair, /* a do nothing */
1328 .accept = sock_no_accept, /* a do nothing */
1329 .getname = inet6_getname,
1330 .poll = datagram_poll, /* ok */
1331 .ioctl = inet6_ioctl, /* must change */
1332 .listen = sock_no_listen, /* ok */
1333 .shutdown = inet_shutdown, /* ok */
1334 .setsockopt = sock_common_setsockopt, /* ok */
1335 .getsockopt = sock_common_getsockopt, /* ok */
1336 .sendmsg = inet_sendmsg, /* ok */
1337 .recvmsg = sock_common_recvmsg, /* ok */
1338 .mmap = sock_no_mmap,
1339 .sendpage = sock_no_sendpage,
1340 #ifdef CONFIG_COMPAT
1341 .compat_setsockopt = compat_sock_common_setsockopt,
1342 .compat_getsockopt = compat_sock_common_getsockopt,
1343 #endif
1344 };
1345
1346 static struct inet_protosw rawv6_protosw = {
1347 .type = SOCK_RAW,
1348 .protocol = IPPROTO_IP, /* wild card */
1349 .prot = &rawv6_prot,
1350 .ops = &inet6_sockraw_ops,
1351 .no_check = UDP_CSUM_DEFAULT,
1352 .flags = INET_PROTOSW_REUSE,
1353 };
1354
1355 int __init rawv6_init(void)
1356 {
1357 int ret;
1358
1359 ret = inet6_register_protosw(&rawv6_protosw);
1360 if (ret)
1361 goto out;
1362 out:
1363 return ret;
1364 }
1365
1366 void rawv6_exit(void)
1367 {
1368 inet6_unregister_protosw(&rawv6_protosw);
1369 }
Linux with added FPC-III support