?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 2.6.25.20
[linux/fpc-iii.git] / net / ipv6 / raw.c
blob0a6fbc1d1a502746eadc80345e1027de95100b03
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 * $Id: raw.c,v 1.51 2002/02/01 22:01:04 davem Exp $
11 *
12 * Fixes:
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
16 *
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
21 */
22
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/netfilter.h>
33 #include <linux/netfilter_ipv6.h>
34 #include <linux/skbuff.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
57 #include <net/raw.h>
58 #include <net/rawv6.h>
59 #include <net/xfrm.h>
60
61 #include <linux/proc_fs.h>
62 #include <linux/seq_file.h>
63
64 static struct raw_hashinfo raw_v6_hashinfo = {
65 .lock = __RW_LOCK_UNLOCKED(),
66 };
67
68 static void raw_v6_hash(struct sock *sk)
69 {
70 raw_hash_sk(sk, &raw_v6_hashinfo);
71 }
72
73 static void raw_v6_unhash(struct sock *sk)
74 {
75 raw_unhash_sk(sk, &raw_v6_hashinfo);
76 }
77
78
79 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
80 unsigned short num, struct in6_addr *loc_addr,
81 struct in6_addr *rmt_addr, int dif)
82 {
83 struct hlist_node *node;
84 int is_multicast = ipv6_addr_is_multicast(loc_addr);
85
86 sk_for_each_from(sk, node)
87 if (inet_sk(sk)->num == num) {
88 struct ipv6_pinfo *np = inet6_sk(sk);
89
90 if (sk->sk_net != net)
91 continue;
92
93 if (!ipv6_addr_any(&np->daddr) &&
94 !ipv6_addr_equal(&np->daddr, rmt_addr))
95 continue;
96
97 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
98 continue;
99
100 if (!ipv6_addr_any(&np->rcv_saddr)) {
101 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
102 goto found;
103 if (is_multicast &&
104 inet6_mc_check(sk, loc_addr, rmt_addr))
105 goto found;
106 continue;
107 }
108 goto found;
109 }
110 sk = NULL;
111 found:
112 return sk;
113 }
114
115 /*
116 * 0 - deliver
117 * 1 - block
118 */
119 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
120 {
121 struct icmp6hdr *icmph;
122 struct raw6_sock *rp = raw6_sk(sk);
123
124 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
125 __u32 *data = &rp->filter.data[0];
126 int bit_nr;
127
128 icmph = (struct icmp6hdr *) skb->data;
129 bit_nr = icmph->icmp6_type;
130
131 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
132 }
133 return 0;
134 }
135
136 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
137 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
138
139 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
140 struct sk_buff *skb))
141 {
142 rcu_assign_pointer(mh_filter, filter);
143 return 0;
144 }
145 EXPORT_SYMBOL(rawv6_mh_filter_register);
146
147 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
148 struct sk_buff *skb))
149 {
150 rcu_assign_pointer(mh_filter, NULL);
151 synchronize_rcu();
152 return 0;
153 }
154 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
155
156 #endif
157
158 /*
159 * demultiplex raw sockets.
160 * (should consider queueing the skb in the sock receive_queue
161 * without calling rawv6.c)
162 *
163 * Caller owns SKB so we must make clones.
164 */
165 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
166 {
167 struct in6_addr *saddr;
168 struct in6_addr *daddr;
169 struct sock *sk;
170 int delivered = 0;
171 __u8 hash;
172 struct net *net;
173
174 saddr = &ipv6_hdr(skb)->saddr;
175 daddr = saddr + 1;
176
177 hash = nexthdr & (MAX_INET_PROTOS - 1);
178
179 read_lock(&raw_v6_hashinfo.lock);
180 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
181
182 /*
183 * The first socket found will be delivered after
184 * delivery to transport protocols.
185 */
186
187 if (sk == NULL)
188 goto out;
189
190 net = skb->dev->nd_net;
191 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
192
193 while (sk) {
194 int filtered;
195
196 delivered = 1;
197 switch (nexthdr) {
198 case IPPROTO_ICMPV6:
199 filtered = icmpv6_filter(sk, skb);
200 break;
201
202 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
203 case IPPROTO_MH:
204 {
205 /* XXX: To validate MH only once for each packet,
206 * this is placed here. It should be after checking
207 * xfrm policy, however it doesn't. The checking xfrm
208 * policy is placed in rawv6_rcv() because it is
209 * required for each socket.
210 */
211 int (*filter)(struct sock *sock, struct sk_buff *skb);
212
213 filter = rcu_dereference(mh_filter);
214 filtered = filter ? filter(sk, skb) : 0;
215 break;
216 }
217 #endif
218 default:
219 filtered = 0;
220 break;
221 }
222
223 if (filtered < 0)
224 break;
225 if (filtered == 0) {
226 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
227
228 /* Not releasing hash table! */
229 if (clone) {
230 nf_reset(clone);
231 rawv6_rcv(sk, clone);
232 }
233 }
234 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
235 IP6CB(skb)->iif);
236 }
237 out:
238 read_unlock(&raw_v6_hashinfo.lock);
239 return delivered;
240 }
241
242 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
243 {
244 struct sock *raw_sk;
245
246 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
247 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
248 raw_sk = NULL;
249
250 return raw_sk != NULL;
251 }
252
253 /* This cleans up af_inet6 a bit. -DaveM */
254 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
255 {
256 struct inet_sock *inet = inet_sk(sk);
257 struct ipv6_pinfo *np = inet6_sk(sk);
258 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
259 __be32 v4addr = 0;
260 int addr_type;
261 int err;
262
263 if (addr_len < SIN6_LEN_RFC2133)
264 return -EINVAL;
265 addr_type = ipv6_addr_type(&addr->sin6_addr);
266
267 /* Raw sockets are IPv6 only */
268 if (addr_type == IPV6_ADDR_MAPPED)
269 return(-EADDRNOTAVAIL);
270
271 lock_sock(sk);
272
273 err = -EINVAL;
274 if (sk->sk_state != TCP_CLOSE)
275 goto out;
276
277 /* Check if the address belongs to the host. */
278 if (addr_type != IPV6_ADDR_ANY) {
279 struct net_device *dev = NULL;
280
281 if (addr_type & IPV6_ADDR_LINKLOCAL) {
282 if (addr_len >= sizeof(struct sockaddr_in6) &&
283 addr->sin6_scope_id) {
284 /* Override any existing binding, if another
285 * one is supplied by user.
286 */
287 sk->sk_bound_dev_if = addr->sin6_scope_id;
288 }
289
290 /* Binding to link-local address requires an interface */
291 if (!sk->sk_bound_dev_if)
292 goto out;
293
294 dev = dev_get_by_index(sk->sk_net, sk->sk_bound_dev_if);
295 if (!dev) {
296 err = -ENODEV;
297 goto out;
298 }
299 }
300
301 /* ipv4 addr of the socket is invalid. Only the
302 * unspecified and mapped address have a v4 equivalent.
303 */
304 v4addr = LOOPBACK4_IPV6;
305 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
306 err = -EADDRNOTAVAIL;
307 if (!ipv6_chk_addr(sk->sk_net, &addr->sin6_addr,
308 dev, 0)) {
309 if (dev)
310 dev_put(dev);
311 goto out;
312 }
313 }
314 if (dev)
315 dev_put(dev);
316 }
317
318 inet->rcv_saddr = inet->saddr = v4addr;
319 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
320 if (!(addr_type & IPV6_ADDR_MULTICAST))
321 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
322 err = 0;
323 out:
324 release_sock(sk);
325 return err;
326 }
327
328 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
329 struct inet6_skb_parm *opt,
330 int type, int code, int offset, __be32 info)
331 {
332 struct inet_sock *inet = inet_sk(sk);
333 struct ipv6_pinfo *np = inet6_sk(sk);
334 int err;
335 int harderr;
336
337 /* Report error on raw socket, if:
338 1. User requested recverr.
339 2. Socket is connected (otherwise the error indication
340 is useless without recverr and error is hard.
341 */
342 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
343 return;
344
345 harderr = icmpv6_err_convert(type, code, &err);
346 if (type == ICMPV6_PKT_TOOBIG)
347 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
348
349 if (np->recverr) {
350 u8 *payload = skb->data;
351 if (!inet->hdrincl)
352 payload += offset;
353 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
354 }
355
356 if (np->recverr || harderr) {
357 sk->sk_err = err;
358 sk->sk_error_report(sk);
359 }
360 }
361
362 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
363 int type, int code, int inner_offset, __be32 info)
364 {
365 struct sock *sk;
366 int hash;
367 struct in6_addr *saddr, *daddr;
368 struct net *net;
369
370 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
371
372 read_lock(&raw_v6_hashinfo.lock);
373 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
374 if (sk != NULL) {
375 struct ipv6hdr *hdr = (struct ipv6hdr *) skb->data;
376
377 saddr = &hdr->saddr;
378 daddr = &hdr->daddr;
379 net = skb->dev->nd_net;
380
381 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
382 IP6CB(skb)->iif))) {
383 rawv6_err(sk, skb, NULL, type, code,
384 inner_offset, info);
385 sk = sk_next(sk);
386 }
387 }
388 read_unlock(&raw_v6_hashinfo.lock);
389 }
390
391 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
392 {
393 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
394 skb_checksum_complete(skb)) {
395 atomic_inc(&sk->sk_drops);
396 kfree_skb(skb);
397 return 0;
398 }
399
400 /* Charge it to the socket. */
401 if (sock_queue_rcv_skb(sk,skb)<0) {
402 atomic_inc(&sk->sk_drops);
403 kfree_skb(skb);
404 return 0;
405 }
406
407 return 0;
408 }
409
410 /*
411 * This is next to useless...
412 * if we demultiplex in network layer we don't need the extra call
413 * just to queue the skb...
414 * maybe we could have the network decide upon a hint if it
415 * should call raw_rcv for demultiplexing
416 */
417 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
418 {
419 struct inet_sock *inet = inet_sk(sk);
420 struct raw6_sock *rp = raw6_sk(sk);
421
422 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
423 atomic_inc(&sk->sk_drops);
424 kfree_skb(skb);
425 return NET_RX_DROP;
426 }
427
428 if (!rp->checksum)
429 skb->ip_summed = CHECKSUM_UNNECESSARY;
430
431 if (skb->ip_summed == CHECKSUM_COMPLETE) {
432 skb_postpull_rcsum(skb, skb_network_header(skb),
433 skb_network_header_len(skb));
434 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
435 &ipv6_hdr(skb)->daddr,
436 skb->len, inet->num, skb->csum))
437 skb->ip_summed = CHECKSUM_UNNECESSARY;
438 }
439 if (!skb_csum_unnecessary(skb))
440 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
441 &ipv6_hdr(skb)->daddr,
442 skb->len,
443 inet->num, 0));
444
445 if (inet->hdrincl) {
446 if (skb_checksum_complete(skb)) {
447 atomic_inc(&sk->sk_drops);
448 kfree_skb(skb);
449 return 0;
450 }
451 }
452
453 rawv6_rcv_skb(sk, skb);
454 return 0;
455 }
456
457
458 /*
459 * This should be easy, if there is something there
460 * we return it, otherwise we block.
461 */
462
463 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
464 struct msghdr *msg, size_t len,
465 int noblock, int flags, int *addr_len)
466 {
467 struct ipv6_pinfo *np = inet6_sk(sk);
468 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
469 struct sk_buff *skb;
470 size_t copied;
471 int err;
472
473 if (flags & MSG_OOB)
474 return -EOPNOTSUPP;
475
476 if (addr_len)
477 *addr_len=sizeof(*sin6);
478
479 if (flags & MSG_ERRQUEUE)
480 return ipv6_recv_error(sk, msg, len);
481
482 skb = skb_recv_datagram(sk, flags, noblock, &err);
483 if (!skb)
484 goto out;
485
486 copied = skb->len;
487 if (copied > len) {
488 copied = len;
489 msg->msg_flags |= MSG_TRUNC;
490 }
491
492 if (skb_csum_unnecessary(skb)) {
493 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
494 } else if (msg->msg_flags&MSG_TRUNC) {
495 if (__skb_checksum_complete(skb))
496 goto csum_copy_err;
497 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
498 } else {
499 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
500 if (err == -EINVAL)
501 goto csum_copy_err;
502 }
503 if (err)
504 goto out_free;
505
506 /* Copy the address. */
507 if (sin6) {
508 sin6->sin6_family = AF_INET6;
509 sin6->sin6_port = 0;
510 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
511 sin6->sin6_flowinfo = 0;
512 sin6->sin6_scope_id = 0;
513 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
514 sin6->sin6_scope_id = IP6CB(skb)->iif;
515 }
516
517 sock_recv_timestamp(msg, sk, skb);
518
519 if (np->rxopt.all)
520 datagram_recv_ctl(sk, msg, skb);
521
522 err = copied;
523 if (flags & MSG_TRUNC)
524 err = skb->len;
525
526 out_free:
527 skb_free_datagram(sk, skb);
528 out:
529 return err;
530
531 csum_copy_err:
532 skb_kill_datagram(sk, skb, flags);
533
534 /* Error for blocking case is chosen to masquerade
535 as some normal condition.
536 */
537 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
538 atomic_inc(&sk->sk_drops);
539 goto out;
540 }
541
542 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
543 struct raw6_sock *rp)
544 {
545 struct sk_buff *skb;
546 int err = 0;
547 int offset;
548 int len;
549 int total_len;
550 __wsum tmp_csum;
551 __sum16 csum;
552
553 if (!rp->checksum)
554 goto send;
555
556 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
557 goto out;
558
559 offset = rp->offset;
560 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
561 skb->data);
562 if (offset >= total_len - 1) {
563 err = -EINVAL;
564 ip6_flush_pending_frames(sk);
565 goto out;
566 }
567
568 /* should be check HW csum miyazawa */
569 if (skb_queue_len(&sk->sk_write_queue) == 1) {
570 /*
571 * Only one fragment on the socket.
572 */
573 tmp_csum = skb->csum;
574 } else {
575 struct sk_buff *csum_skb = NULL;
576 tmp_csum = 0;
577
578 skb_queue_walk(&sk->sk_write_queue, skb) {
579 tmp_csum = csum_add(tmp_csum, skb->csum);
580
581 if (csum_skb)
582 continue;
583
584 len = skb->len - skb_transport_offset(skb);
585 if (offset >= len) {
586 offset -= len;
587 continue;
588 }
589
590 csum_skb = skb;
591 }
592
593 skb = csum_skb;
594 }
595
596 offset += skb_transport_offset(skb);
597 if (skb_copy_bits(skb, offset, &csum, 2))
598 BUG();
599
600 /* in case cksum was not initialized */
601 if (unlikely(csum))
602 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
603
604 csum = csum_ipv6_magic(&fl->fl6_src,
605 &fl->fl6_dst,
606 total_len, fl->proto, tmp_csum);
607
608 if (csum == 0 && fl->proto == IPPROTO_UDP)
609 csum = CSUM_MANGLED_0;
610
611 if (skb_store_bits(skb, offset, &csum, 2))
612 BUG();
613
614 send:
615 err = ip6_push_pending_frames(sk);
616 out:
617 return err;
618 }
619
620 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
621 struct flowi *fl, struct rt6_info *rt,
622 unsigned int flags)
623 {
624 struct ipv6_pinfo *np = inet6_sk(sk);
625 struct ipv6hdr *iph;
626 struct sk_buff *skb;
627 unsigned int hh_len;
628 int err;
629
630 if (length > rt->u.dst.dev->mtu) {
631 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
632 return -EMSGSIZE;
633 }
634 if (flags&MSG_PROBE)
635 goto out;
636
637 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
638
639 skb = sock_alloc_send_skb(sk, length+hh_len+15,
640 flags&MSG_DONTWAIT, &err);
641 if (skb == NULL)
642 goto error;
643 skb_reserve(skb, hh_len);
644
645 skb->priority = sk->sk_priority;
646 skb->mark = sk->sk_mark;
647 skb->dst = dst_clone(&rt->u.dst);
648
649 skb_put(skb, length);
650 skb_reset_network_header(skb);
651 iph = ipv6_hdr(skb);
652
653 skb->ip_summed = CHECKSUM_NONE;
654
655 skb->transport_header = skb->network_header;
656 err = memcpy_fromiovecend((void *)iph, from, 0, length);
657 if (err)
658 goto error_fault;
659
660 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
661 err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
662 dst_output);
663 if (err > 0)
664 err = np->recverr ? net_xmit_errno(err) : 0;
665 if (err)
666 goto error;
667 out:
668 return 0;
669
670 error_fault:
671 err = -EFAULT;
672 kfree_skb(skb);
673 error:
674 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
675 return err;
676 }
677
678 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
679 {
680 struct iovec *iov;
681 u8 __user *type = NULL;
682 u8 __user *code = NULL;
683 u8 len = 0;
684 int probed = 0;
685 int i;
686
687 if (!msg->msg_iov)
688 return 0;
689
690 for (i = 0; i < msg->msg_iovlen; i++) {
691 iov = &msg->msg_iov[i];
692 if (!iov)
693 continue;
694
695 switch (fl->proto) {
696 case IPPROTO_ICMPV6:
697 /* check if one-byte field is readable or not. */
698 if (iov->iov_base && iov->iov_len < 1)
699 break;
700
701 if (!type) {
702 type = iov->iov_base;
703 /* check if code field is readable or not. */
704 if (iov->iov_len > 1)
705 code = type + 1;
706 } else if (!code)
707 code = iov->iov_base;
708
709 if (type && code) {
710 if (get_user(fl->fl_icmp_type, type) ||
711 get_user(fl->fl_icmp_code, code))
712 return -EFAULT;
713 probed = 1;
714 }
715 break;
716 case IPPROTO_MH:
717 if (iov->iov_base && iov->iov_len < 1)
718 break;
719 /* check if type field is readable or not. */
720 if (iov->iov_len > 2 - len) {
721 u8 __user *p = iov->iov_base;
722 if (get_user(fl->fl_mh_type, &p[2 - len]))
723 return -EFAULT;
724 probed = 1;
725 } else
726 len += iov->iov_len;
727
728 break;
729 default:
730 probed = 1;
731 break;
732 }
733 if (probed)
734 break;
735 }
736 return 0;
737 }
738
739 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
740 struct msghdr *msg, size_t len)
741 {
742 struct ipv6_txoptions opt_space;
743 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
744 struct in6_addr *daddr, *final_p = NULL, final;
745 struct inet_sock *inet = inet_sk(sk);
746 struct ipv6_pinfo *np = inet6_sk(sk);
747 struct raw6_sock *rp = raw6_sk(sk);
748 struct ipv6_txoptions *opt = NULL;
749 struct ip6_flowlabel *flowlabel = NULL;
750 struct dst_entry *dst = NULL;
751 struct flowi fl;
752 int addr_len = msg->msg_namelen;
753 int hlimit = -1;
754 int tclass = -1;
755 u16 proto;
756 int err;
757
758 /* Rough check on arithmetic overflow,
759 better check is made in ip6_append_data().
760 */
761 if (len > INT_MAX)
762 return -EMSGSIZE;
763
764 /* Mirror BSD error message compatibility */
765 if (msg->msg_flags & MSG_OOB)
766 return -EOPNOTSUPP;
767
768 /*
769 * Get and verify the address.
770 */
771 memset(&fl, 0, sizeof(fl));
772
773 fl.mark = sk->sk_mark;
774
775 if (sin6) {
776 if (addr_len < SIN6_LEN_RFC2133)
777 return -EINVAL;
778
779 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
780 return(-EAFNOSUPPORT);
781
782 /* port is the proto value [0..255] carried in nexthdr */
783 proto = ntohs(sin6->sin6_port);
784
785 if (!proto)
786 proto = inet->num;
787 else if (proto != inet->num)
788 return(-EINVAL);
789
790 if (proto > 255)
791 return(-EINVAL);
792
793 daddr = &sin6->sin6_addr;
794 if (np->sndflow) {
795 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
796 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
797 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
798 if (flowlabel == NULL)
799 return -EINVAL;
800 daddr = &flowlabel->dst;
801 }
802 }
803
804 /*
805 * Otherwise it will be difficult to maintain
806 * sk->sk_dst_cache.
807 */
808 if (sk->sk_state == TCP_ESTABLISHED &&
809 ipv6_addr_equal(daddr, &np->daddr))
810 daddr = &np->daddr;
811
812 if (addr_len >= sizeof(struct sockaddr_in6) &&
813 sin6->sin6_scope_id &&
814 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
815 fl.oif = sin6->sin6_scope_id;
816 } else {
817 if (sk->sk_state != TCP_ESTABLISHED)
818 return -EDESTADDRREQ;
819
820 proto = inet->num;
821 daddr = &np->daddr;
822 fl.fl6_flowlabel = np->flow_label;
823 }
824
825 if (ipv6_addr_any(daddr)) {
826 /*
827 * unspecified destination address
828 * treated as error... is this correct ?
829 */
830 fl6_sock_release(flowlabel);
831 return(-EINVAL);
832 }
833
834 if (fl.oif == 0)
835 fl.oif = sk->sk_bound_dev_if;
836
837 if (msg->msg_controllen) {
838 opt = &opt_space;
839 memset(opt, 0, sizeof(struct ipv6_txoptions));
840 opt->tot_len = sizeof(struct ipv6_txoptions);
841
842 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
843 if (err < 0) {
844 fl6_sock_release(flowlabel);
845 return err;
846 }
847 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
848 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
849 if (flowlabel == NULL)
850 return -EINVAL;
851 }
852 if (!(opt->opt_nflen|opt->opt_flen))
853 opt = NULL;
854 }
855 if (opt == NULL)
856 opt = np->opt;
857 if (flowlabel)
858 opt = fl6_merge_options(&opt_space, flowlabel, opt);
859 opt = ipv6_fixup_options(&opt_space, opt);
860
861 fl.proto = proto;
862 err = rawv6_probe_proto_opt(&fl, msg);
863 if (err)
864 goto out;
865
866 ipv6_addr_copy(&fl.fl6_dst, daddr);
867 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
868 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
869
870 /* merge ip6_build_xmit from ip6_output */
871 if (opt && opt->srcrt) {
872 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
873 ipv6_addr_copy(&final, &fl.fl6_dst);
874 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
875 final_p = &final;
876 }
877
878 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
879 fl.oif = np->mcast_oif;
880 security_sk_classify_flow(sk, &fl);
881
882 err = ip6_dst_lookup(sk, &dst, &fl);
883 if (err)
884 goto out;
885 if (final_p)
886 ipv6_addr_copy(&fl.fl6_dst, final_p);
887
888 if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 0) {
889 if (err == -EREMOTE)
890 err = ip6_dst_blackhole(sk, &dst, &fl);
891 if (err < 0)
892 goto out;
893 }
894
895 if (hlimit < 0) {
896 if (ipv6_addr_is_multicast(&fl.fl6_dst))
897 hlimit = np->mcast_hops;
898 else
899 hlimit = np->hop_limit;
900 if (hlimit < 0)
901 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
902 if (hlimit < 0)
903 hlimit = ipv6_get_hoplimit(dst->dev);
904 }
905
906 if (tclass < 0) {
907 tclass = np->tclass;
908 if (tclass < 0)
909 tclass = 0;
910 }
911
912 if (msg->msg_flags&MSG_CONFIRM)
913 goto do_confirm;
914
915 back_from_confirm:
916 if (inet->hdrincl) {
917 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
918 } else {
919 lock_sock(sk);
920 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
921 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
922 msg->msg_flags);
923
924 if (err)
925 ip6_flush_pending_frames(sk);
926 else if (!(msg->msg_flags & MSG_MORE))
927 err = rawv6_push_pending_frames(sk, &fl, rp);
928 release_sock(sk);
929 }
930 done:
931 dst_release(dst);
932 out:
933 fl6_sock_release(flowlabel);
934 return err<0?err:len;
935 do_confirm:
936 dst_confirm(dst);
937 if (!(msg->msg_flags & MSG_PROBE) || len)
938 goto back_from_confirm;
939 err = 0;
940 goto done;
941 }
942
943 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
944 char __user *optval, int optlen)
945 {
946 switch (optname) {
947 case ICMPV6_FILTER:
948 if (optlen > sizeof(struct icmp6_filter))
949 optlen = sizeof(struct icmp6_filter);
950 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
951 return -EFAULT;
952 return 0;
953 default:
954 return -ENOPROTOOPT;
955 }
956
957 return 0;
958 }
959
960 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
961 char __user *optval, int __user *optlen)
962 {
963 int len;
964
965 switch (optname) {
966 case ICMPV6_FILTER:
967 if (get_user(len, optlen))
968 return -EFAULT;
969 if (len < 0)
970 return -EINVAL;
971 if (len > sizeof(struct icmp6_filter))
972 len = sizeof(struct icmp6_filter);
973 if (put_user(len, optlen))
974 return -EFAULT;
975 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
976 return -EFAULT;
977 return 0;
978 default:
979 return -ENOPROTOOPT;
980 }
981
982 return 0;
983 }
984
985
986 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
987 char __user *optval, int optlen)
988 {
989 struct raw6_sock *rp = raw6_sk(sk);
990 int val;
991
992 if (get_user(val, (int __user *)optval))
993 return -EFAULT;
994
995 switch (optname) {
996 case IPV6_CHECKSUM:
997 /* You may get strange result with a positive odd offset;
998 RFC2292bis agrees with me. */
999 if (val > 0 && (val&1))
1000 return(-EINVAL);
1001 if (val < 0) {
1002 rp->checksum = 0;
1003 } else {
1004 rp->checksum = 1;
1005 rp->offset = val;
1006 }
1007
1008 return 0;
1009 break;
1010
1011 default:
1012 return(-ENOPROTOOPT);
1013 }
1014 }
1015
1016 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1017 char __user *optval, int optlen)
1018 {
1019 switch(level) {
1020 case SOL_RAW:
1021 break;
1022
1023 case SOL_ICMPV6:
1024 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1025 return -EOPNOTSUPP;
1026 return rawv6_seticmpfilter(sk, level, optname, optval,
1027 optlen);
1028 case SOL_IPV6:
1029 if (optname == IPV6_CHECKSUM)
1030 break;
1031 default:
1032 return ipv6_setsockopt(sk, level, optname, optval,
1033 optlen);
1034 }
1035
1036 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1037 }
1038
1039 #ifdef CONFIG_COMPAT
1040 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1041 char __user *optval, int optlen)
1042 {
1043 switch (level) {
1044 case SOL_RAW:
1045 break;
1046 case SOL_ICMPV6:
1047 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1048 return -EOPNOTSUPP;
1049 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1050 case SOL_IPV6:
1051 if (optname == IPV6_CHECKSUM)
1052 break;
1053 default:
1054 return compat_ipv6_setsockopt(sk, level, optname,
1055 optval, optlen);
1056 }
1057 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1058 }
1059 #endif
1060
1061 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1062 char __user *optval, int __user *optlen)
1063 {
1064 struct raw6_sock *rp = raw6_sk(sk);
1065 int val, len;
1066
1067 if (get_user(len,optlen))
1068 return -EFAULT;
1069
1070 switch (optname) {
1071 case IPV6_CHECKSUM:
1072 if (rp->checksum == 0)
1073 val = -1;
1074 else
1075 val = rp->offset;
1076 break;
1077
1078 default:
1079 return -ENOPROTOOPT;
1080 }
1081
1082 len = min_t(unsigned int, sizeof(int), len);
1083
1084 if (put_user(len, optlen))
1085 return -EFAULT;
1086 if (copy_to_user(optval,&val,len))
1087 return -EFAULT;
1088 return 0;
1089 }
1090
1091 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1092 char __user *optval, int __user *optlen)
1093 {
1094 switch(level) {
1095 case SOL_RAW:
1096 break;
1097
1098 case SOL_ICMPV6:
1099 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1100 return -EOPNOTSUPP;
1101 return rawv6_geticmpfilter(sk, level, optname, optval,
1102 optlen);
1103 case SOL_IPV6:
1104 if (optname == IPV6_CHECKSUM)
1105 break;
1106 default:
1107 return ipv6_getsockopt(sk, level, optname, optval,
1108 optlen);
1109 }
1110
1111 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1112 }
1113
1114 #ifdef CONFIG_COMPAT
1115 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1116 char __user *optval, int __user *optlen)
1117 {
1118 switch (level) {
1119 case SOL_RAW:
1120 break;
1121 case SOL_ICMPV6:
1122 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1123 return -EOPNOTSUPP;
1124 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1125 case SOL_IPV6:
1126 if (optname == IPV6_CHECKSUM)
1127 break;
1128 default:
1129 return compat_ipv6_getsockopt(sk, level, optname,
1130 optval, optlen);
1131 }
1132 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1133 }
1134 #endif
1135
1136 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1137 {
1138 switch(cmd) {
1139 case SIOCOUTQ:
1140 {
1141 int amount = atomic_read(&sk->sk_wmem_alloc);
1142 return put_user(amount, (int __user *)arg);
1143 }
1144 case SIOCINQ:
1145 {
1146 struct sk_buff *skb;
1147 int amount = 0;
1148
1149 spin_lock_bh(&sk->sk_receive_queue.lock);
1150 skb = skb_peek(&sk->sk_receive_queue);
1151 if (skb != NULL)
1152 amount = skb->tail - skb->transport_header;
1153 spin_unlock_bh(&sk->sk_receive_queue.lock);
1154 return put_user(amount, (int __user *)arg);
1155 }
1156
1157 default:
1158 return -ENOIOCTLCMD;
1159 }
1160 }
1161
1162 static void rawv6_close(struct sock *sk, long timeout)
1163 {
1164 if (inet_sk(sk)->num == IPPROTO_RAW)
1165 ip6_ra_control(sk, -1, NULL);
1166
1167 sk_common_release(sk);
1168 }
1169
1170 static int rawv6_init_sk(struct sock *sk)
1171 {
1172 struct raw6_sock *rp = raw6_sk(sk);
1173
1174 switch (inet_sk(sk)->num) {
1175 case IPPROTO_ICMPV6:
1176 rp->checksum = 1;
1177 rp->offset = 2;
1178 break;
1179 case IPPROTO_MH:
1180 rp->checksum = 1;
1181 rp->offset = 4;
1182 break;
1183 default:
1184 break;
1185 }
1186 return(0);
1187 }
1188
1189 DEFINE_PROTO_INUSE(rawv6)
1190
1191 struct proto rawv6_prot = {
1192 .name = "RAWv6",
1193 .owner = THIS_MODULE,
1194 .close = rawv6_close,
1195 .connect = ip6_datagram_connect,
1196 .disconnect = udp_disconnect,
1197 .ioctl = rawv6_ioctl,
1198 .init = rawv6_init_sk,
1199 .destroy = inet6_destroy_sock,
1200 .setsockopt = rawv6_setsockopt,
1201 .getsockopt = rawv6_getsockopt,
1202 .sendmsg = rawv6_sendmsg,
1203 .recvmsg = rawv6_recvmsg,
1204 .bind = rawv6_bind,
1205 .backlog_rcv = rawv6_rcv_skb,
1206 .hash = raw_v6_hash,
1207 .unhash = raw_v6_unhash,
1208 .obj_size = sizeof(struct raw6_sock),
1209 #ifdef CONFIG_COMPAT
1210 .compat_setsockopt = compat_rawv6_setsockopt,
1211 .compat_getsockopt = compat_rawv6_getsockopt,
1212 #endif
1213 REF_PROTO_INUSE(rawv6)
1214 };
1215
1216 #ifdef CONFIG_PROC_FS
1217 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1218 {
1219 struct ipv6_pinfo *np = inet6_sk(sp);
1220 struct in6_addr *dest, *src;
1221 __u16 destp, srcp;
1222
1223 dest = &np->daddr;
1224 src = &np->rcv_saddr;
1225 destp = 0;
1226 srcp = inet_sk(sp)->num;
1227 seq_printf(seq,
1228 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1229 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1230 i,
1231 src->s6_addr32[0], src->s6_addr32[1],
1232 src->s6_addr32[2], src->s6_addr32[3], srcp,
1233 dest->s6_addr32[0], dest->s6_addr32[1],
1234 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1235 sp->sk_state,
1236 atomic_read(&sp->sk_wmem_alloc),
1237 atomic_read(&sp->sk_rmem_alloc),
1238 0, 0L, 0,
1239 sock_i_uid(sp), 0,
1240 sock_i_ino(sp),
1241 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1242 }
1243
1244 static int raw6_seq_show(struct seq_file *seq, void *v)
1245 {
1246 if (v == SEQ_START_TOKEN)
1247 seq_printf(seq,
1248 " sl "
1249 "local_address "
1250 "remote_address "
1251 "st tx_queue rx_queue tr tm->when retrnsmt"
1252 " uid timeout inode drops\n");
1253 else
1254 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1255 return 0;
1256 }
1257
1258 static const struct seq_operations raw6_seq_ops = {
1259 .start = raw_seq_start,
1260 .next = raw_seq_next,
1261 .stop = raw_seq_stop,
1262 .show = raw6_seq_show,
1263 };
1264
1265 static int raw6_seq_open(struct inode *inode, struct file *file)
1266 {
1267 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1268 }
1269
1270 static const struct file_operations raw6_seq_fops = {
1271 .owner = THIS_MODULE,
1272 .open = raw6_seq_open,
1273 .read = seq_read,
1274 .llseek = seq_lseek,
1275 .release = seq_release_net,
1276 };
1277
1278 static int raw6_init_net(struct net *net)
1279 {
1280 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1281 return -ENOMEM;
1282
1283 return 0;
1284 }
1285
1286 static void raw6_exit_net(struct net *net)
1287 {
1288 proc_net_remove(net, "raw6");
1289 }
1290
1291 static struct pernet_operations raw6_net_ops = {
1292 .init = raw6_init_net,
1293 .exit = raw6_exit_net,
1294 };
1295
1296 int __init raw6_proc_init(void)
1297 {
1298 return register_pernet_subsys(&raw6_net_ops);
1299 }
1300
1301 void raw6_proc_exit(void)
1302 {
1303 unregister_pernet_subsys(&raw6_net_ops);
1304 }
1305 #endif /* CONFIG_PROC_FS */
1306
1307 /* Same as inet6_dgram_ops, sans udp_poll. */
1308 static const struct proto_ops inet6_sockraw_ops = {
1309 .family = PF_INET6,
1310 .owner = THIS_MODULE,
1311 .release = inet6_release,
1312 .bind = inet6_bind,
1313 .connect = inet_dgram_connect, /* ok */
1314 .socketpair = sock_no_socketpair, /* a do nothing */
1315 .accept = sock_no_accept, /* a do nothing */
1316 .getname = inet6_getname,
1317 .poll = datagram_poll, /* ok */
1318 .ioctl = inet6_ioctl, /* must change */
1319 .listen = sock_no_listen, /* ok */
1320 .shutdown = inet_shutdown, /* ok */
1321 .setsockopt = sock_common_setsockopt, /* ok */
1322 .getsockopt = sock_common_getsockopt, /* ok */
1323 .sendmsg = inet_sendmsg, /* ok */
1324 .recvmsg = sock_common_recvmsg, /* ok */
1325 .mmap = sock_no_mmap,
1326 .sendpage = sock_no_sendpage,
1327 #ifdef CONFIG_COMPAT
1328 .compat_setsockopt = compat_sock_common_setsockopt,
1329 .compat_getsockopt = compat_sock_common_getsockopt,
1330 #endif
1331 };
1332
1333 static struct inet_protosw rawv6_protosw = {
1334 .type = SOCK_RAW,
1335 .protocol = IPPROTO_IP, /* wild card */
1336 .prot = &rawv6_prot,
1337 .ops = &inet6_sockraw_ops,
1338 .capability = CAP_NET_RAW,
1339 .no_check = UDP_CSUM_DEFAULT,
1340 .flags = INET_PROTOSW_REUSE,
1341 };
1342
1343 int __init rawv6_init(void)
1344 {
1345 int ret;
1346
1347 ret = inet6_register_protosw(&rawv6_protosw);
1348 if (ret)
1349 goto out;
1350 out:
1351 return ret;
1352 }
1353
1354 void rawv6_exit(void)
1355 {
1356 inet6_unregister_protosw(&rawv6_protosw);
1357 }
Linux with added FPC-III support