?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'dm-3.11-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/agk...
[linux-2.6.git] / net / ipv6 / raw.c
blobc45f7a5c36e96f98487ca194c5f08e5d0f931852
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 IS_ENABLED(CONFIG_IPV6_MIP6)
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 #include <linux/export.h>
65
66 static struct raw_hashinfo raw_v6_hashinfo = {
67 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
68 };
69
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)
73 {
74 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
75
76 sk_for_each_from(sk)
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 IS_ENABLED(CONFIG_IPV6_MIP6)
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_INIT_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 bool 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 bool delivered = false;
160 __u8 hash;
161 struct net *net;
162
163 saddr = &ipv6_hdr(skb)->saddr;
164 daddr = saddr + 1;
165
166 hash = nexthdr & (RAW_HTABLE_SIZE - 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 = true;
181 switch (nexthdr) {
182 case IPPROTO_ICMPV6:
183 filtered = icmpv6_filter(sk, skb);
184 break;
185
186 #if IS_ENABLED(CONFIG_IPV6_MIP6)
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 bool 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 & (RAW_HTABLE_SIZE - 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 (__ipv6_addr_needs_scope_id(addr_type)) {
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 np->rcv_saddr = addr->sin6_addr;
301 if (!(addr_type & IPV6_ADDR_MULTICAST))
302 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 ip6_sk_update_pmtu(skb, sk, info);
331 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
332 }
333 if (type == NDISC_REDIRECT)
334 ip6_sk_redirect(skb, sk);
335 if (np->recverr) {
336 u8 *payload = skb->data;
337 if (!inet->hdrincl)
338 payload += offset;
339 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
340 }
341
342 if (np->recverr || harderr) {
343 sk->sk_err = err;
344 sk->sk_error_report(sk);
345 }
346 }
347
348 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
349 u8 type, u8 code, int inner_offset, __be32 info)
350 {
351 struct sock *sk;
352 int hash;
353 const struct in6_addr *saddr, *daddr;
354 struct net *net;
355
356 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
357
358 read_lock(&raw_v6_hashinfo.lock);
359 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
360 if (sk != NULL) {
361 /* Note: ipv6_hdr(skb) != skb->data */
362 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
363 saddr = &ip6h->saddr;
364 daddr = &ip6h->daddr;
365 net = dev_net(skb->dev);
366
367 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
368 IP6CB(skb)->iif))) {
369 rawv6_err(sk, skb, NULL, type, code,
370 inner_offset, info);
371 sk = sk_next(sk);
372 }
373 }
374 read_unlock(&raw_v6_hashinfo.lock);
375 }
376
377 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
378 {
379 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
380 skb_checksum_complete(skb)) {
381 atomic_inc(&sk->sk_drops);
382 kfree_skb(skb);
383 return NET_RX_DROP;
384 }
385
386 /* Charge it to the socket. */
387 skb_dst_drop(skb);
388 if (sock_queue_rcv_skb(sk, skb) < 0) {
389 kfree_skb(skb);
390 return NET_RX_DROP;
391 }
392
393 return 0;
394 }
395
396 /*
397 * This is next to useless...
398 * if we demultiplex in network layer we don't need the extra call
399 * just to queue the skb...
400 * maybe we could have the network decide upon a hint if it
401 * should call raw_rcv for demultiplexing
402 */
403 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
404 {
405 struct inet_sock *inet = inet_sk(sk);
406 struct raw6_sock *rp = raw6_sk(sk);
407
408 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
409 atomic_inc(&sk->sk_drops);
410 kfree_skb(skb);
411 return NET_RX_DROP;
412 }
413
414 if (!rp->checksum)
415 skb->ip_summed = CHECKSUM_UNNECESSARY;
416
417 if (skb->ip_summed == CHECKSUM_COMPLETE) {
418 skb_postpull_rcsum(skb, skb_network_header(skb),
419 skb_network_header_len(skb));
420 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
421 &ipv6_hdr(skb)->daddr,
422 skb->len, inet->inet_num, skb->csum))
423 skb->ip_summed = CHECKSUM_UNNECESSARY;
424 }
425 if (!skb_csum_unnecessary(skb))
426 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
427 &ipv6_hdr(skb)->daddr,
428 skb->len,
429 inet->inet_num, 0));
430
431 if (inet->hdrincl) {
432 if (skb_checksum_complete(skb)) {
433 atomic_inc(&sk->sk_drops);
434 kfree_skb(skb);
435 return NET_RX_DROP;
436 }
437 }
438
439 rawv6_rcv_skb(sk, skb);
440 return 0;
441 }
442
443
444 /*
445 * This should be easy, if there is something there
446 * we return it, otherwise we block.
447 */
448
449 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
450 struct msghdr *msg, size_t len,
451 int noblock, int flags, int *addr_len)
452 {
453 struct ipv6_pinfo *np = inet6_sk(sk);
454 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
455 struct sk_buff *skb;
456 size_t copied;
457 int err;
458
459 if (flags & MSG_OOB)
460 return -EOPNOTSUPP;
461
462 if (addr_len)
463 *addr_len=sizeof(*sin6);
464
465 if (flags & MSG_ERRQUEUE)
466 return ipv6_recv_error(sk, msg, len);
467
468 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
469 return ipv6_recv_rxpmtu(sk, msg, len);
470
471 skb = skb_recv_datagram(sk, flags, noblock, &err);
472 if (!skb)
473 goto out;
474
475 copied = skb->len;
476 if (copied > len) {
477 copied = len;
478 msg->msg_flags |= MSG_TRUNC;
479 }
480
481 if (skb_csum_unnecessary(skb)) {
482 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
483 } else if (msg->msg_flags&MSG_TRUNC) {
484 if (__skb_checksum_complete(skb))
485 goto csum_copy_err;
486 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
487 } else {
488 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
489 if (err == -EINVAL)
490 goto csum_copy_err;
491 }
492 if (err)
493 goto out_free;
494
495 /* Copy the address. */
496 if (sin6) {
497 sin6->sin6_family = AF_INET6;
498 sin6->sin6_port = 0;
499 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
500 sin6->sin6_flowinfo = 0;
501 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
502 IP6CB(skb)->iif);
503 }
504
505 sock_recv_ts_and_drops(msg, sk, skb);
506
507 if (np->rxopt.all)
508 ip6_datagram_recv_ctl(sk, msg, skb);
509
510 err = copied;
511 if (flags & MSG_TRUNC)
512 err = skb->len;
513
514 out_free:
515 skb_free_datagram(sk, skb);
516 out:
517 return err;
518
519 csum_copy_err:
520 skb_kill_datagram(sk, skb, flags);
521
522 /* Error for blocking case is chosen to masquerade
523 as some normal condition.
524 */
525 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
526 goto out;
527 }
528
529 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
530 struct raw6_sock *rp)
531 {
532 struct sk_buff *skb;
533 int err = 0;
534 int offset;
535 int len;
536 int total_len;
537 __wsum tmp_csum;
538 __sum16 csum;
539
540 if (!rp->checksum)
541 goto send;
542
543 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
544 goto out;
545
546 offset = rp->offset;
547 total_len = inet_sk(sk)->cork.base.length;
548 if (offset >= total_len - 1) {
549 err = -EINVAL;
550 ip6_flush_pending_frames(sk);
551 goto out;
552 }
553
554 /* should be check HW csum miyazawa */
555 if (skb_queue_len(&sk->sk_write_queue) == 1) {
556 /*
557 * Only one fragment on the socket.
558 */
559 tmp_csum = skb->csum;
560 } else {
561 struct sk_buff *csum_skb = NULL;
562 tmp_csum = 0;
563
564 skb_queue_walk(&sk->sk_write_queue, skb) {
565 tmp_csum = csum_add(tmp_csum, skb->csum);
566
567 if (csum_skb)
568 continue;
569
570 len = skb->len - skb_transport_offset(skb);
571 if (offset >= len) {
572 offset -= len;
573 continue;
574 }
575
576 csum_skb = skb;
577 }
578
579 skb = csum_skb;
580 }
581
582 offset += skb_transport_offset(skb);
583 if (skb_copy_bits(skb, offset, &csum, 2))
584 BUG();
585
586 /* in case cksum was not initialized */
587 if (unlikely(csum))
588 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
589
590 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
591 total_len, fl6->flowi6_proto, tmp_csum);
592
593 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
594 csum = CSUM_MANGLED_0;
595
596 if (skb_store_bits(skb, offset, &csum, 2))
597 BUG();
598
599 send:
600 err = ip6_push_pending_frames(sk);
601 out:
602 return err;
603 }
604
605 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
606 struct flowi6 *fl6, struct dst_entry **dstp,
607 unsigned int flags)
608 {
609 struct ipv6_pinfo *np = inet6_sk(sk);
610 struct ipv6hdr *iph;
611 struct sk_buff *skb;
612 int err;
613 struct rt6_info *rt = (struct rt6_info *)*dstp;
614 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
615 int tlen = rt->dst.dev->needed_tailroom;
616
617 if (length > rt->dst.dev->mtu) {
618 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
619 return -EMSGSIZE;
620 }
621 if (flags&MSG_PROBE)
622 goto out;
623
624 skb = sock_alloc_send_skb(sk,
625 length + hlen + tlen + 15,
626 flags & MSG_DONTWAIT, &err);
627 if (skb == NULL)
628 goto error;
629 skb_reserve(skb, hlen);
630
631 skb->priority = sk->sk_priority;
632 skb->mark = sk->sk_mark;
633 skb_dst_set(skb, &rt->dst);
634 *dstp = NULL;
635
636 skb_put(skb, length);
637 skb_reset_network_header(skb);
638 iph = ipv6_hdr(skb);
639
640 skb->ip_summed = CHECKSUM_NONE;
641
642 skb->transport_header = skb->network_header;
643 err = memcpy_fromiovecend((void *)iph, from, 0, length);
644 if (err)
645 goto error_fault;
646
647 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
648 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
649 rt->dst.dev, dst_output);
650 if (err > 0)
651 err = net_xmit_errno(err);
652 if (err)
653 goto error;
654 out:
655 return 0;
656
657 error_fault:
658 err = -EFAULT;
659 kfree_skb(skb);
660 error:
661 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
662 if (err == -ENOBUFS && !np->recverr)
663 err = 0;
664 return err;
665 }
666
667 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg)
668 {
669 struct iovec *iov;
670 u8 __user *type = NULL;
671 u8 __user *code = NULL;
672 u8 len = 0;
673 int probed = 0;
674 int i;
675
676 if (!msg->msg_iov)
677 return 0;
678
679 for (i = 0; i < msg->msg_iovlen; i++) {
680 iov = &msg->msg_iov[i];
681 if (!iov)
682 continue;
683
684 switch (fl6->flowi6_proto) {
685 case IPPROTO_ICMPV6:
686 /* check if one-byte field is readable or not. */
687 if (iov->iov_base && iov->iov_len < 1)
688 break;
689
690 if (!type) {
691 type = iov->iov_base;
692 /* check if code field is readable or not. */
693 if (iov->iov_len > 1)
694 code = type + 1;
695 } else if (!code)
696 code = iov->iov_base;
697
698 if (type && code) {
699 if (get_user(fl6->fl6_icmp_type, type) ||
700 get_user(fl6->fl6_icmp_code, code))
701 return -EFAULT;
702 probed = 1;
703 }
704 break;
705 case IPPROTO_MH:
706 if (iov->iov_base && iov->iov_len < 1)
707 break;
708 /* check if type field is readable or not. */
709 if (iov->iov_len > 2 - len) {
710 u8 __user *p = iov->iov_base;
711 if (get_user(fl6->fl6_mh_type, &p[2 - len]))
712 return -EFAULT;
713 probed = 1;
714 } else
715 len += iov->iov_len;
716
717 break;
718 default:
719 probed = 1;
720 break;
721 }
722 if (probed)
723 break;
724 }
725 return 0;
726 }
727
728 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
729 struct msghdr *msg, size_t len)
730 {
731 struct ipv6_txoptions opt_space;
732 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
733 struct in6_addr *daddr, *final_p, final;
734 struct inet_sock *inet = inet_sk(sk);
735 struct ipv6_pinfo *np = inet6_sk(sk);
736 struct raw6_sock *rp = raw6_sk(sk);
737 struct ipv6_txoptions *opt = NULL;
738 struct ip6_flowlabel *flowlabel = NULL;
739 struct dst_entry *dst = NULL;
740 struct flowi6 fl6;
741 int addr_len = msg->msg_namelen;
742 int hlimit = -1;
743 int tclass = -1;
744 int dontfrag = -1;
745 u16 proto;
746 int err;
747
748 /* Rough check on arithmetic overflow,
749 better check is made in ip6_append_data().
750 */
751 if (len > INT_MAX)
752 return -EMSGSIZE;
753
754 /* Mirror BSD error message compatibility */
755 if (msg->msg_flags & MSG_OOB)
756 return -EOPNOTSUPP;
757
758 /*
759 * Get and verify the address.
760 */
761 memset(&fl6, 0, sizeof(fl6));
762
763 fl6.flowi6_mark = sk->sk_mark;
764
765 if (sin6) {
766 if (addr_len < SIN6_LEN_RFC2133)
767 return -EINVAL;
768
769 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
770 return -EAFNOSUPPORT;
771
772 /* port is the proto value [0..255] carried in nexthdr */
773 proto = ntohs(sin6->sin6_port);
774
775 if (!proto)
776 proto = inet->inet_num;
777 else if (proto != inet->inet_num)
778 return -EINVAL;
779
780 if (proto > 255)
781 return -EINVAL;
782
783 daddr = &sin6->sin6_addr;
784 if (np->sndflow) {
785 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
786 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
787 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
788 if (flowlabel == NULL)
789 return -EINVAL;
790 daddr = &flowlabel->dst;
791 }
792 }
793
794 /*
795 * Otherwise it will be difficult to maintain
796 * sk->sk_dst_cache.
797 */
798 if (sk->sk_state == TCP_ESTABLISHED &&
799 ipv6_addr_equal(daddr, &np->daddr))
800 daddr = &np->daddr;
801
802 if (addr_len >= sizeof(struct sockaddr_in6) &&
803 sin6->sin6_scope_id &&
804 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
805 fl6.flowi6_oif = sin6->sin6_scope_id;
806 } else {
807 if (sk->sk_state != TCP_ESTABLISHED)
808 return -EDESTADDRREQ;
809
810 proto = inet->inet_num;
811 daddr = &np->daddr;
812 fl6.flowlabel = np->flow_label;
813 }
814
815 if (fl6.flowi6_oif == 0)
816 fl6.flowi6_oif = sk->sk_bound_dev_if;
817
818 if (msg->msg_controllen) {
819 opt = &opt_space;
820 memset(opt, 0, sizeof(struct ipv6_txoptions));
821 opt->tot_len = sizeof(struct ipv6_txoptions);
822
823 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
824 &hlimit, &tclass, &dontfrag);
825 if (err < 0) {
826 fl6_sock_release(flowlabel);
827 return err;
828 }
829 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
830 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
831 if (flowlabel == NULL)
832 return -EINVAL;
833 }
834 if (!(opt->opt_nflen|opt->opt_flen))
835 opt = NULL;
836 }
837 if (opt == NULL)
838 opt = np->opt;
839 if (flowlabel)
840 opt = fl6_merge_options(&opt_space, flowlabel, opt);
841 opt = ipv6_fixup_options(&opt_space, opt);
842
843 fl6.flowi6_proto = proto;
844 err = rawv6_probe_proto_opt(&fl6, msg);
845 if (err)
846 goto out;
847
848 if (!ipv6_addr_any(daddr))
849 fl6.daddr = *daddr;
850 else
851 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
852 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
853 fl6.saddr = np->saddr;
854
855 final_p = fl6_update_dst(&fl6, opt, &final);
856
857 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
858 fl6.flowi6_oif = np->mcast_oif;
859 else if (!fl6.flowi6_oif)
860 fl6.flowi6_oif = np->ucast_oif;
861 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
862
863 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
864 if (IS_ERR(dst)) {
865 err = PTR_ERR(dst);
866 goto out;
867 }
868 if (hlimit < 0) {
869 if (ipv6_addr_is_multicast(&fl6.daddr))
870 hlimit = np->mcast_hops;
871 else
872 hlimit = np->hop_limit;
873 if (hlimit < 0)
874 hlimit = ip6_dst_hoplimit(dst);
875 }
876
877 if (tclass < 0)
878 tclass = np->tclass;
879
880 if (dontfrag < 0)
881 dontfrag = np->dontfrag;
882
883 if (msg->msg_flags&MSG_CONFIRM)
884 goto do_confirm;
885
886 back_from_confirm:
887 if (inet->hdrincl)
888 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags);
889 else {
890 lock_sock(sk);
891 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
892 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
893 msg->msg_flags, dontfrag);
894
895 if (err)
896 ip6_flush_pending_frames(sk);
897 else if (!(msg->msg_flags & MSG_MORE))
898 err = rawv6_push_pending_frames(sk, &fl6, rp);
899 release_sock(sk);
900 }
901 done:
902 dst_release(dst);
903 out:
904 fl6_sock_release(flowlabel);
905 return err<0?err:len;
906 do_confirm:
907 dst_confirm(dst);
908 if (!(msg->msg_flags & MSG_PROBE) || len)
909 goto back_from_confirm;
910 err = 0;
911 goto done;
912 }
913
914 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
915 char __user *optval, int optlen)
916 {
917 switch (optname) {
918 case ICMPV6_FILTER:
919 if (optlen > sizeof(struct icmp6_filter))
920 optlen = sizeof(struct icmp6_filter);
921 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
922 return -EFAULT;
923 return 0;
924 default:
925 return -ENOPROTOOPT;
926 }
927
928 return 0;
929 }
930
931 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
932 char __user *optval, int __user *optlen)
933 {
934 int len;
935
936 switch (optname) {
937 case ICMPV6_FILTER:
938 if (get_user(len, optlen))
939 return -EFAULT;
940 if (len < 0)
941 return -EINVAL;
942 if (len > sizeof(struct icmp6_filter))
943 len = sizeof(struct icmp6_filter);
944 if (put_user(len, optlen))
945 return -EFAULT;
946 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
947 return -EFAULT;
948 return 0;
949 default:
950 return -ENOPROTOOPT;
951 }
952
953 return 0;
954 }
955
956
957 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
958 char __user *optval, unsigned int optlen)
959 {
960 struct raw6_sock *rp = raw6_sk(sk);
961 int val;
962
963 if (get_user(val, (int __user *)optval))
964 return -EFAULT;
965
966 switch (optname) {
967 case IPV6_CHECKSUM:
968 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
969 level == IPPROTO_IPV6) {
970 /*
971 * RFC3542 tells that IPV6_CHECKSUM socket
972 * option in the IPPROTO_IPV6 level is not
973 * allowed on ICMPv6 sockets.
974 * If you want to set it, use IPPROTO_RAW
975 * level IPV6_CHECKSUM socket option
976 * (Linux extension).
977 */
978 return -EINVAL;
979 }
980
981 /* You may get strange result with a positive odd offset;
982 RFC2292bis agrees with me. */
983 if (val > 0 && (val&1))
984 return -EINVAL;
985 if (val < 0) {
986 rp->checksum = 0;
987 } else {
988 rp->checksum = 1;
989 rp->offset = val;
990 }
991
992 return 0;
993
994 default:
995 return -ENOPROTOOPT;
996 }
997 }
998
999 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1000 char __user *optval, unsigned int optlen)
1001 {
1002 switch (level) {
1003 case SOL_RAW:
1004 break;
1005
1006 case SOL_ICMPV6:
1007 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1008 return -EOPNOTSUPP;
1009 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1010 case SOL_IPV6:
1011 if (optname == IPV6_CHECKSUM)
1012 break;
1013 default:
1014 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1015 }
1016
1017 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1018 }
1019
1020 #ifdef CONFIG_COMPAT
1021 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1022 char __user *optval, unsigned int optlen)
1023 {
1024 switch (level) {
1025 case SOL_RAW:
1026 break;
1027 case SOL_ICMPV6:
1028 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1029 return -EOPNOTSUPP;
1030 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1031 case SOL_IPV6:
1032 if (optname == IPV6_CHECKSUM)
1033 break;
1034 default:
1035 return compat_ipv6_setsockopt(sk, level, optname,
1036 optval, optlen);
1037 }
1038 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1039 }
1040 #endif
1041
1042 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1043 char __user *optval, int __user *optlen)
1044 {
1045 struct raw6_sock *rp = raw6_sk(sk);
1046 int val, len;
1047
1048 if (get_user(len,optlen))
1049 return -EFAULT;
1050
1051 switch (optname) {
1052 case IPV6_CHECKSUM:
1053 /*
1054 * We allow getsockopt() for IPPROTO_IPV6-level
1055 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1056 * since RFC3542 is silent about it.
1057 */
1058 if (rp->checksum == 0)
1059 val = -1;
1060 else
1061 val = rp->offset;
1062 break;
1063
1064 default:
1065 return -ENOPROTOOPT;
1066 }
1067
1068 len = min_t(unsigned int, sizeof(int), len);
1069
1070 if (put_user(len, optlen))
1071 return -EFAULT;
1072 if (copy_to_user(optval,&val,len))
1073 return -EFAULT;
1074 return 0;
1075 }
1076
1077 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1078 char __user *optval, int __user *optlen)
1079 {
1080 switch (level) {
1081 case SOL_RAW:
1082 break;
1083
1084 case SOL_ICMPV6:
1085 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1086 return -EOPNOTSUPP;
1087 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1088 case SOL_IPV6:
1089 if (optname == IPV6_CHECKSUM)
1090 break;
1091 default:
1092 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1093 }
1094
1095 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1096 }
1097
1098 #ifdef CONFIG_COMPAT
1099 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1100 char __user *optval, int __user *optlen)
1101 {
1102 switch (level) {
1103 case SOL_RAW:
1104 break;
1105 case SOL_ICMPV6:
1106 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1107 return -EOPNOTSUPP;
1108 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1109 case SOL_IPV6:
1110 if (optname == IPV6_CHECKSUM)
1111 break;
1112 default:
1113 return compat_ipv6_getsockopt(sk, level, optname,
1114 optval, optlen);
1115 }
1116 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1117 }
1118 #endif
1119
1120 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1121 {
1122 switch (cmd) {
1123 case SIOCOUTQ: {
1124 int amount = sk_wmem_alloc_get(sk);
1125
1126 return put_user(amount, (int __user *)arg);
1127 }
1128 case SIOCINQ: {
1129 struct sk_buff *skb;
1130 int amount = 0;
1131
1132 spin_lock_bh(&sk->sk_receive_queue.lock);
1133 skb = skb_peek(&sk->sk_receive_queue);
1134 if (skb != NULL)
1135 amount = skb_tail_pointer(skb) -
1136 skb_transport_header(skb);
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 int raw6_seq_show(struct seq_file *seq, void *v)
1231 {
1232 if (v == SEQ_START_TOKEN) {
1233 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1234 } else {
1235 struct sock *sp = v;
1236 __u16 srcp = inet_sk(sp)->inet_num;
1237 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1238 raw_seq_private(seq)->bucket);
1239 }
1240 return 0;
1241 }
1242
1243 static const struct seq_operations raw6_seq_ops = {
1244 .start = raw_seq_start,
1245 .next = raw_seq_next,
1246 .stop = raw_seq_stop,
1247 .show = raw6_seq_show,
1248 };
1249
1250 static int raw6_seq_open(struct inode *inode, struct file *file)
1251 {
1252 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1253 }
1254
1255 static const struct file_operations raw6_seq_fops = {
1256 .owner = THIS_MODULE,
1257 .open = raw6_seq_open,
1258 .read = seq_read,
1259 .llseek = seq_lseek,
1260 .release = seq_release_net,
1261 };
1262
1263 static int __net_init raw6_init_net(struct net *net)
1264 {
1265 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops))
1266 return -ENOMEM;
1267
1268 return 0;
1269 }
1270
1271 static void __net_exit raw6_exit_net(struct net *net)
1272 {
1273 remove_proc_entry("raw6", net->proc_net);
1274 }
1275
1276 static struct pernet_operations raw6_net_ops = {
1277 .init = raw6_init_net,
1278 .exit = raw6_exit_net,
1279 };
1280
1281 int __init raw6_proc_init(void)
1282 {
1283 return register_pernet_subsys(&raw6_net_ops);
1284 }
1285
1286 void raw6_proc_exit(void)
1287 {
1288 unregister_pernet_subsys(&raw6_net_ops);
1289 }
1290 #endif /* CONFIG_PROC_FS */
1291
1292 /* Same as inet6_dgram_ops, sans udp_poll. */
1293 static const struct proto_ops inet6_sockraw_ops = {
1294 .family = PF_INET6,
1295 .owner = THIS_MODULE,
1296 .release = inet6_release,
1297 .bind = inet6_bind,
1298 .connect = inet_dgram_connect, /* ok */
1299 .socketpair = sock_no_socketpair, /* a do nothing */
1300 .accept = sock_no_accept, /* a do nothing */
1301 .getname = inet6_getname,
1302 .poll = datagram_poll, /* ok */
1303 .ioctl = inet6_ioctl, /* must change */
1304 .listen = sock_no_listen, /* ok */
1305 .shutdown = inet_shutdown, /* ok */
1306 .setsockopt = sock_common_setsockopt, /* ok */
1307 .getsockopt = sock_common_getsockopt, /* ok */
1308 .sendmsg = inet_sendmsg, /* ok */
1309 .recvmsg = sock_common_recvmsg, /* ok */
1310 .mmap = sock_no_mmap,
1311 .sendpage = sock_no_sendpage,
1312 #ifdef CONFIG_COMPAT
1313 .compat_setsockopt = compat_sock_common_setsockopt,
1314 .compat_getsockopt = compat_sock_common_getsockopt,
1315 #endif
1316 };
1317
1318 static struct inet_protosw rawv6_protosw = {
1319 .type = SOCK_RAW,
1320 .protocol = IPPROTO_IP, /* wild card */
1321 .prot = &rawv6_prot,
1322 .ops = &inet6_sockraw_ops,
1323 .no_check = UDP_CSUM_DEFAULT,
1324 .flags = INET_PROTOSW_REUSE,
1325 };
1326
1327 int __init rawv6_init(void)
1328 {
1329 int ret;
1330
1331 ret = inet6_register_protosw(&rawv6_protosw);
1332 if (ret)
1333 goto out;
1334 out:
1335 return ret;
1336 }
1337
1338 void rawv6_exit(void)
1339 {
1340 inet6_unregister_protosw(&rawv6_protosw);
1341 }
Linus' kernel tree