?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge remote-tracking branch 'upstream/master' into kvm-devel
[linux-2.6/kvm.git] / net / ipv6 / raw.c
blob6f7824e1cea48ea16b4a3860475e4fd367860834
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 __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
110 {
111 struct icmp6hdr *icmph;
112 struct raw6_sock *rp = raw6_sk(sk);
113
114 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
115 __u32 *data = &rp->filter.data[0];
116 int bit_nr;
117
118 icmph = (struct icmp6hdr *) skb->data;
119 bit_nr = icmph->icmp6_type;
120
121 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
122 }
123 return 0;
124 }
125
126 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
127 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
128
129 static mh_filter_t __rcu *mh_filter __read_mostly;
130
131 int rawv6_mh_filter_register(mh_filter_t filter)
132 {
133 RCU_INIT_POINTER(mh_filter, filter);
134 return 0;
135 }
136 EXPORT_SYMBOL(rawv6_mh_filter_register);
137
138 int rawv6_mh_filter_unregister(mh_filter_t filter)
139 {
140 RCU_INIT_POINTER(mh_filter, NULL);
141 synchronize_rcu();
142 return 0;
143 }
144 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
145
146 #endif
147
148 /*
149 * demultiplex raw sockets.
150 * (should consider queueing the skb in the sock receive_queue
151 * without calling rawv6.c)
152 *
153 * Caller owns SKB so we must make clones.
154 */
155 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
156 {
157 const struct in6_addr *saddr;
158 const struct in6_addr *daddr;
159 struct sock *sk;
160 int delivered = 0;
161 __u8 hash;
162 struct net *net;
163
164 saddr = &ipv6_hdr(skb)->saddr;
165 daddr = saddr + 1;
166
167 hash = nexthdr & (MAX_INET_PROTOS - 1);
168
169 read_lock(&raw_v6_hashinfo.lock);
170 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
171
172 if (sk == NULL)
173 goto out;
174
175 net = dev_net(skb->dev);
176 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
177
178 while (sk) {
179 int filtered;
180
181 delivered = 1;
182 switch (nexthdr) {
183 case IPPROTO_ICMPV6:
184 filtered = icmpv6_filter(sk, skb);
185 break;
186
187 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
188 case IPPROTO_MH:
189 {
190 /* XXX: To validate MH only once for each packet,
191 * this is placed here. It should be after checking
192 * xfrm policy, however it doesn't. The checking xfrm
193 * policy is placed in rawv6_rcv() because it is
194 * required for each socket.
195 */
196 mh_filter_t *filter;
197
198 filter = rcu_dereference(mh_filter);
199 filtered = filter ? (*filter)(sk, skb) : 0;
200 break;
201 }
202 #endif
203 default:
204 filtered = 0;
205 break;
206 }
207
208 if (filtered < 0)
209 break;
210 if (filtered == 0) {
211 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
212
213 /* Not releasing hash table! */
214 if (clone) {
215 nf_reset(clone);
216 rawv6_rcv(sk, clone);
217 }
218 }
219 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
220 IP6CB(skb)->iif);
221 }
222 out:
223 read_unlock(&raw_v6_hashinfo.lock);
224 return delivered;
225 }
226
227 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
228 {
229 struct sock *raw_sk;
230
231 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
232 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
233 raw_sk = NULL;
234
235 return raw_sk != NULL;
236 }
237
238 /* This cleans up af_inet6 a bit. -DaveM */
239 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
240 {
241 struct inet_sock *inet = inet_sk(sk);
242 struct ipv6_pinfo *np = inet6_sk(sk);
243 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
244 __be32 v4addr = 0;
245 int addr_type;
246 int err;
247
248 if (addr_len < SIN6_LEN_RFC2133)
249 return -EINVAL;
250 addr_type = ipv6_addr_type(&addr->sin6_addr);
251
252 /* Raw sockets are IPv6 only */
253 if (addr_type == IPV6_ADDR_MAPPED)
254 return -EADDRNOTAVAIL;
255
256 lock_sock(sk);
257
258 err = -EINVAL;
259 if (sk->sk_state != TCP_CLOSE)
260 goto out;
261
262 rcu_read_lock();
263 /* Check if the address belongs to the host. */
264 if (addr_type != IPV6_ADDR_ANY) {
265 struct net_device *dev = NULL;
266
267 if (addr_type & IPV6_ADDR_LINKLOCAL) {
268 if (addr_len >= sizeof(struct sockaddr_in6) &&
269 addr->sin6_scope_id) {
270 /* Override any existing binding, if another
271 * one is supplied by user.
272 */
273 sk->sk_bound_dev_if = addr->sin6_scope_id;
274 }
275
276 /* Binding to link-local address requires an interface */
277 if (!sk->sk_bound_dev_if)
278 goto out_unlock;
279
280 err = -ENODEV;
281 dev = dev_get_by_index_rcu(sock_net(sk),
282 sk->sk_bound_dev_if);
283 if (!dev)
284 goto out_unlock;
285 }
286
287 /* ipv4 addr of the socket is invalid. Only the
288 * unspecified and mapped address have a v4 equivalent.
289 */
290 v4addr = LOOPBACK4_IPV6;
291 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
292 err = -EADDRNOTAVAIL;
293 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
294 dev, 0)) {
295 goto out_unlock;
296 }
297 }
298 }
299
300 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
301 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
302 if (!(addr_type & IPV6_ADDR_MULTICAST))
303 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
304 err = 0;
305 out_unlock:
306 rcu_read_unlock();
307 out:
308 release_sock(sk);
309 return err;
310 }
311
312 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
313 struct inet6_skb_parm *opt,
314 u8 type, u8 code, int offset, __be32 info)
315 {
316 struct inet_sock *inet = inet_sk(sk);
317 struct ipv6_pinfo *np = inet6_sk(sk);
318 int err;
319 int harderr;
320
321 /* Report error on raw socket, if:
322 1. User requested recverr.
323 2. Socket is connected (otherwise the error indication
324 is useless without recverr and error is hard.
325 */
326 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
327 return;
328
329 harderr = icmpv6_err_convert(type, code, &err);
330 if (type == ICMPV6_PKT_TOOBIG)
331 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
332
333 if (np->recverr) {
334 u8 *payload = skb->data;
335 if (!inet->hdrincl)
336 payload += offset;
337 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
338 }
339
340 if (np->recverr || harderr) {
341 sk->sk_err = err;
342 sk->sk_error_report(sk);
343 }
344 }
345
346 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
347 u8 type, u8 code, int inner_offset, __be32 info)
348 {
349 struct sock *sk;
350 int hash;
351 const struct in6_addr *saddr, *daddr;
352 struct net *net;
353
354 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
355
356 read_lock(&raw_v6_hashinfo.lock);
357 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
358 if (sk != NULL) {
359 /* Note: ipv6_hdr(skb) != skb->data */
360 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
361 saddr = &ip6h->saddr;
362 daddr = &ip6h->daddr;
363 net = dev_net(skb->dev);
364
365 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
366 IP6CB(skb)->iif))) {
367 rawv6_err(sk, skb, NULL, type, code,
368 inner_offset, info);
369 sk = sk_next(sk);
370 }
371 }
372 read_unlock(&raw_v6_hashinfo.lock);
373 }
374
375 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
376 {
377 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
378 skb_checksum_complete(skb)) {
379 atomic_inc(&sk->sk_drops);
380 kfree_skb(skb);
381 return NET_RX_DROP;
382 }
383
384 /* Charge it to the socket. */
385 if (ip_queue_rcv_skb(sk, skb) < 0) {
386 kfree_skb(skb);
387 return NET_RX_DROP;
388 }
389
390 return 0;
391 }
392
393 /*
394 * This is next to useless...
395 * if we demultiplex in network layer we don't need the extra call
396 * just to queue the skb...
397 * maybe we could have the network decide upon a hint if it
398 * should call raw_rcv for demultiplexing
399 */
400 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
401 {
402 struct inet_sock *inet = inet_sk(sk);
403 struct raw6_sock *rp = raw6_sk(sk);
404
405 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
406 atomic_inc(&sk->sk_drops);
407 kfree_skb(skb);
408 return NET_RX_DROP;
409 }
410
411 if (!rp->checksum)
412 skb->ip_summed = CHECKSUM_UNNECESSARY;
413
414 if (skb->ip_summed == CHECKSUM_COMPLETE) {
415 skb_postpull_rcsum(skb, skb_network_header(skb),
416 skb_network_header_len(skb));
417 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
418 &ipv6_hdr(skb)->daddr,
419 skb->len, inet->inet_num, skb->csum))
420 skb->ip_summed = CHECKSUM_UNNECESSARY;
421 }
422 if (!skb_csum_unnecessary(skb))
423 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
424 &ipv6_hdr(skb)->daddr,
425 skb->len,
426 inet->inet_num, 0));
427
428 if (inet->hdrincl) {
429 if (skb_checksum_complete(skb)) {
430 atomic_inc(&sk->sk_drops);
431 kfree_skb(skb);
432 return NET_RX_DROP;
433 }
434 }
435
436 rawv6_rcv_skb(sk, skb);
437 return 0;
438 }
439
440
441 /*
442 * This should be easy, if there is something there
443 * we return it, otherwise we block.
444 */
445
446 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
447 struct msghdr *msg, size_t len,
448 int noblock, int flags, int *addr_len)
449 {
450 struct ipv6_pinfo *np = inet6_sk(sk);
451 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
452 struct sk_buff *skb;
453 size_t copied;
454 int err;
455
456 if (flags & MSG_OOB)
457 return -EOPNOTSUPP;
458
459 if (addr_len)
460 *addr_len=sizeof(*sin6);
461
462 if (flags & MSG_ERRQUEUE)
463 return ipv6_recv_error(sk, msg, len);
464
465 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
466 return ipv6_recv_rxpmtu(sk, msg, len);
467
468 skb = skb_recv_datagram(sk, flags, noblock, &err);
469 if (!skb)
470 goto out;
471
472 copied = skb->len;
473 if (copied > len) {
474 copied = len;
475 msg->msg_flags |= MSG_TRUNC;
476 }
477
478 if (skb_csum_unnecessary(skb)) {
479 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
480 } else if (msg->msg_flags&MSG_TRUNC) {
481 if (__skb_checksum_complete(skb))
482 goto csum_copy_err;
483 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
484 } else {
485 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
486 if (err == -EINVAL)
487 goto csum_copy_err;
488 }
489 if (err)
490 goto out_free;
491
492 /* Copy the address. */
493 if (sin6) {
494 sin6->sin6_family = AF_INET6;
495 sin6->sin6_port = 0;
496 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
497 sin6->sin6_flowinfo = 0;
498 sin6->sin6_scope_id = 0;
499 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
500 sin6->sin6_scope_id = IP6CB(skb)->iif;
501 }
502
503 sock_recv_ts_and_drops(msg, sk, skb);
504
505 if (np->rxopt.all)
506 datagram_recv_ctl(sk, msg, skb);
507
508 err = copied;
509 if (flags & MSG_TRUNC)
510 err = skb->len;
511
512 out_free:
513 skb_free_datagram(sk, skb);
514 out:
515 return err;
516
517 csum_copy_err:
518 skb_kill_datagram(sk, skb, flags);
519
520 /* Error for blocking case is chosen to masquerade
521 as some normal condition.
522 */
523 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
524 goto out;
525 }
526
527 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
528 struct raw6_sock *rp)
529 {
530 struct sk_buff *skb;
531 int err = 0;
532 int offset;
533 int len;
534 int total_len;
535 __wsum tmp_csum;
536 __sum16 csum;
537
538 if (!rp->checksum)
539 goto send;
540
541 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
542 goto out;
543
544 offset = rp->offset;
545 total_len = inet_sk(sk)->cork.base.length;
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), sk, msg, &fl6, opt,
820 &hlimit, &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
988 default:
989 return -ENOPROTOOPT;
990 }
991 }
992
993 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
994 char __user *optval, unsigned int optlen)
995 {
996 switch (level) {
997 case SOL_RAW:
998 break;
999
1000 case SOL_ICMPV6:
1001 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1002 return -EOPNOTSUPP;
1003 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1004 case SOL_IPV6:
1005 if (optname == IPV6_CHECKSUM)
1006 break;
1007 default:
1008 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1009 }
1010
1011 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1012 }
1013
1014 #ifdef CONFIG_COMPAT
1015 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1016 char __user *optval, unsigned int optlen)
1017 {
1018 switch (level) {
1019 case SOL_RAW:
1020 break;
1021 case SOL_ICMPV6:
1022 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1023 return -EOPNOTSUPP;
1024 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1025 case SOL_IPV6:
1026 if (optname == IPV6_CHECKSUM)
1027 break;
1028 default:
1029 return compat_ipv6_setsockopt(sk, level, optname,
1030 optval, optlen);
1031 }
1032 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1033 }
1034 #endif
1035
1036 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1037 char __user *optval, int __user *optlen)
1038 {
1039 struct raw6_sock *rp = raw6_sk(sk);
1040 int val, len;
1041
1042 if (get_user(len,optlen))
1043 return -EFAULT;
1044
1045 switch (optname) {
1046 case IPV6_CHECKSUM:
1047 /*
1048 * We allow getsockopt() for IPPROTO_IPV6-level
1049 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1050 * since RFC3542 is silent about it.
1051 */
1052 if (rp->checksum == 0)
1053 val = -1;
1054 else
1055 val = rp->offset;
1056 break;
1057
1058 default:
1059 return -ENOPROTOOPT;
1060 }
1061
1062 len = min_t(unsigned int, sizeof(int), len);
1063
1064 if (put_user(len, optlen))
1065 return -EFAULT;
1066 if (copy_to_user(optval,&val,len))
1067 return -EFAULT;
1068 return 0;
1069 }
1070
1071 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1072 char __user *optval, int __user *optlen)
1073 {
1074 switch (level) {
1075 case SOL_RAW:
1076 break;
1077
1078 case SOL_ICMPV6:
1079 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1080 return -EOPNOTSUPP;
1081 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1082 case SOL_IPV6:
1083 if (optname == IPV6_CHECKSUM)
1084 break;
1085 default:
1086 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1087 }
1088
1089 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1090 }
1091
1092 #ifdef CONFIG_COMPAT
1093 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1094 char __user *optval, int __user *optlen)
1095 {
1096 switch (level) {
1097 case SOL_RAW:
1098 break;
1099 case SOL_ICMPV6:
1100 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1101 return -EOPNOTSUPP;
1102 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1103 case SOL_IPV6:
1104 if (optname == IPV6_CHECKSUM)
1105 break;
1106 default:
1107 return compat_ipv6_getsockopt(sk, level, optname,
1108 optval, optlen);
1109 }
1110 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1111 }
1112 #endif
1113
1114 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1115 {
1116 switch (cmd) {
1117 case SIOCOUTQ: {
1118 int amount = sk_wmem_alloc_get(sk);
1119
1120 return put_user(amount, (int __user *)arg);
1121 }
1122 case SIOCINQ: {
1123 struct sk_buff *skb;
1124 int amount = 0;
1125
1126 spin_lock_bh(&sk->sk_receive_queue.lock);
1127 skb = skb_peek(&sk->sk_receive_queue);
1128 if (skb != NULL)
1129 amount = skb->tail - skb->transport_header;
1130 spin_unlock_bh(&sk->sk_receive_queue.lock);
1131 return put_user(amount, (int __user *)arg);
1132 }
1133
1134 default:
1135 #ifdef CONFIG_IPV6_MROUTE
1136 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1137 #else
1138 return -ENOIOCTLCMD;
1139 #endif
1140 }
1141 }
1142
1143 #ifdef CONFIG_COMPAT
1144 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1145 {
1146 switch (cmd) {
1147 case SIOCOUTQ:
1148 case SIOCINQ:
1149 return -ENOIOCTLCMD;
1150 default:
1151 #ifdef CONFIG_IPV6_MROUTE
1152 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1153 #else
1154 return -ENOIOCTLCMD;
1155 #endif
1156 }
1157 }
1158 #endif
1159
1160 static void rawv6_close(struct sock *sk, long timeout)
1161 {
1162 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1163 ip6_ra_control(sk, -1);
1164 ip6mr_sk_done(sk);
1165 sk_common_release(sk);
1166 }
1167
1168 static void raw6_destroy(struct sock *sk)
1169 {
1170 lock_sock(sk);
1171 ip6_flush_pending_frames(sk);
1172 release_sock(sk);
1173
1174 inet6_destroy_sock(sk);
1175 }
1176
1177 static int rawv6_init_sk(struct sock *sk)
1178 {
1179 struct raw6_sock *rp = raw6_sk(sk);
1180
1181 switch (inet_sk(sk)->inet_num) {
1182 case IPPROTO_ICMPV6:
1183 rp->checksum = 1;
1184 rp->offset = 2;
1185 break;
1186 case IPPROTO_MH:
1187 rp->checksum = 1;
1188 rp->offset = 4;
1189 break;
1190 default:
1191 break;
1192 }
1193 return 0;
1194 }
1195
1196 struct proto rawv6_prot = {
1197 .name = "RAWv6",
1198 .owner = THIS_MODULE,
1199 .close = rawv6_close,
1200 .destroy = raw6_destroy,
1201 .connect = ip6_datagram_connect,
1202 .disconnect = udp_disconnect,
1203 .ioctl = rawv6_ioctl,
1204 .init = rawv6_init_sk,
1205 .setsockopt = rawv6_setsockopt,
1206 .getsockopt = rawv6_getsockopt,
1207 .sendmsg = rawv6_sendmsg,
1208 .recvmsg = rawv6_recvmsg,
1209 .bind = rawv6_bind,
1210 .backlog_rcv = rawv6_rcv_skb,
1211 .hash = raw_hash_sk,
1212 .unhash = raw_unhash_sk,
1213 .obj_size = sizeof(struct raw6_sock),
1214 .h.raw_hash = &raw_v6_hashinfo,
1215 #ifdef CONFIG_COMPAT
1216 .compat_setsockopt = compat_rawv6_setsockopt,
1217 .compat_getsockopt = compat_rawv6_getsockopt,
1218 .compat_ioctl = compat_rawv6_ioctl,
1219 #endif
1220 };
1221
1222 #ifdef CONFIG_PROC_FS
1223 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1224 {
1225 struct ipv6_pinfo *np = inet6_sk(sp);
1226 const struct in6_addr *dest, *src;
1227 __u16 destp, srcp;
1228
1229 dest = &np->daddr;
1230 src = &np->rcv_saddr;
1231 destp = 0;
1232 srcp = inet_sk(sp)->inet_num;
1233 seq_printf(seq,
1234 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1235 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1236 i,
1237 src->s6_addr32[0], src->s6_addr32[1],
1238 src->s6_addr32[2], src->s6_addr32[3], srcp,
1239 dest->s6_addr32[0], dest->s6_addr32[1],
1240 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1241 sp->sk_state,
1242 sk_wmem_alloc_get(sp),
1243 sk_rmem_alloc_get(sp),
1244 0, 0L, 0,
1245 sock_i_uid(sp), 0,
1246 sock_i_ino(sp),
1247 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1248 }
1249
1250 static int raw6_seq_show(struct seq_file *seq, void *v)
1251 {
1252 if (v == SEQ_START_TOKEN)
1253 seq_printf(seq,
1254 " sl "
1255 "local_address "
1256 "remote_address "
1257 "st tx_queue rx_queue tr tm->when retrnsmt"
1258 " uid timeout inode ref pointer drops\n");
1259 else
1260 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1261 return 0;
1262 }
1263
1264 static const struct seq_operations raw6_seq_ops = {
1265 .start = raw_seq_start,
1266 .next = raw_seq_next,
1267 .stop = raw_seq_stop,
1268 .show = raw6_seq_show,
1269 };
1270
1271 static int raw6_seq_open(struct inode *inode, struct file *file)
1272 {
1273 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1274 }
1275
1276 static const struct file_operations raw6_seq_fops = {
1277 .owner = THIS_MODULE,
1278 .open = raw6_seq_open,
1279 .read = seq_read,
1280 .llseek = seq_lseek,
1281 .release = seq_release_net,
1282 };
1283
1284 static int __net_init raw6_init_net(struct net *net)
1285 {
1286 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1287 return -ENOMEM;
1288
1289 return 0;
1290 }
1291
1292 static void __net_exit raw6_exit_net(struct net *net)
1293 {
1294 proc_net_remove(net, "raw6");
1295 }
1296
1297 static struct pernet_operations raw6_net_ops = {
1298 .init = raw6_init_net,
1299 .exit = raw6_exit_net,
1300 };
1301
1302 int __init raw6_proc_init(void)
1303 {
1304 return register_pernet_subsys(&raw6_net_ops);
1305 }
1306
1307 void raw6_proc_exit(void)
1308 {
1309 unregister_pernet_subsys(&raw6_net_ops);
1310 }
1311 #endif /* CONFIG_PROC_FS */
1312
1313 /* Same as inet6_dgram_ops, sans udp_poll. */
1314 static const struct proto_ops inet6_sockraw_ops = {
1315 .family = PF_INET6,
1316 .owner = THIS_MODULE,
1317 .release = inet6_release,
1318 .bind = inet6_bind,
1319 .connect = inet_dgram_connect, /* ok */
1320 .socketpair = sock_no_socketpair, /* a do nothing */
1321 .accept = sock_no_accept, /* a do nothing */
1322 .getname = inet6_getname,
1323 .poll = datagram_poll, /* ok */
1324 .ioctl = inet6_ioctl, /* must change */
1325 .listen = sock_no_listen, /* ok */
1326 .shutdown = inet_shutdown, /* ok */
1327 .setsockopt = sock_common_setsockopt, /* ok */
1328 .getsockopt = sock_common_getsockopt, /* ok */
1329 .sendmsg = inet_sendmsg, /* ok */
1330 .recvmsg = sock_common_recvmsg, /* ok */
1331 .mmap = sock_no_mmap,
1332 .sendpage = sock_no_sendpage,
1333 #ifdef CONFIG_COMPAT
1334 .compat_setsockopt = compat_sock_common_setsockopt,
1335 .compat_getsockopt = compat_sock_common_getsockopt,
1336 #endif
1337 };
1338
1339 static struct inet_protosw rawv6_protosw = {
1340 .type = SOCK_RAW,
1341 .protocol = IPPROTO_IP, /* wild card */
1342 .prot = &rawv6_prot,
1343 .ops = &inet6_sockraw_ops,
1344 .no_check = UDP_CSUM_DEFAULT,
1345 .flags = INET_PROTOSW_REUSE,
1346 };
1347
1348 int __init rawv6_init(void)
1349 {
1350 int ret;
1351
1352 ret = inet6_register_protosw(&rawv6_protosw);
1353 if (ret)
1354 goto out;
1355 out:
1356 return ret;
1357 }
1358
1359 void rawv6_exit(void)
1360 {
1361 inet6_unregister_protosw(&rawv6_protosw);
1362 }
kernel-based virtual machine