?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6.git] / net / ipv6 / raw.c
bloba4ed2416399ed52622b9c60460d411b9507e6d85
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 #define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */
67
68 static struct raw_hashinfo raw_v6_hashinfo = {
69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
70 };
71
72 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
73 unsigned short num, const struct in6_addr *loc_addr,
74 const struct in6_addr *rmt_addr, int dif)
75 {
76 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
77
78 sk_for_each_from(sk)
79 if (inet_sk(sk)->inet_num == num) {
80 struct ipv6_pinfo *np = inet6_sk(sk);
81
82 if (!net_eq(sock_net(sk), net))
83 continue;
84
85 if (!ipv6_addr_any(&np->daddr) &&
86 !ipv6_addr_equal(&np->daddr, rmt_addr))
87 continue;
88
89 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
90 continue;
91
92 if (!ipv6_addr_any(&np->rcv_saddr)) {
93 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
94 goto found;
95 if (is_multicast &&
96 inet6_mc_check(sk, loc_addr, rmt_addr))
97 goto found;
98 continue;
99 }
100 goto found;
101 }
102 sk = NULL;
103 found:
104 return sk;
105 }
106
107 /*
108 * 0 - deliver
109 * 1 - block
110 */
111 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
112 {
113 struct icmp6hdr _hdr;
114 const struct icmp6hdr *hdr;
115
116 /* We require only the four bytes of the ICMPv6 header, not any
117 * additional bytes of message body in "struct icmp6hdr".
118 */
119 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
120 ICMPV6_HDRLEN, &_hdr);
121 if (hdr) {
122 const __u32 *data = &raw6_sk(sk)->filter.data[0];
123 unsigned int type = hdr->icmp6_type;
124
125 return (data[type >> 5] & (1U << (type & 31))) != 0;
126 }
127 return 1;
128 }
129
130 #if IS_ENABLED(CONFIG_IPV6_MIP6)
131 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
132
133 static mh_filter_t __rcu *mh_filter __read_mostly;
134
135 int rawv6_mh_filter_register(mh_filter_t filter)
136 {
137 rcu_assign_pointer(mh_filter, filter);
138 return 0;
139 }
140 EXPORT_SYMBOL(rawv6_mh_filter_register);
141
142 int rawv6_mh_filter_unregister(mh_filter_t filter)
143 {
144 RCU_INIT_POINTER(mh_filter, NULL);
145 synchronize_rcu();
146 return 0;
147 }
148 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
149
150 #endif
151
152 /*
153 * demultiplex raw sockets.
154 * (should consider queueing the skb in the sock receive_queue
155 * without calling rawv6.c)
156 *
157 * Caller owns SKB so we must make clones.
158 */
159 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
160 {
161 const struct in6_addr *saddr;
162 const struct in6_addr *daddr;
163 struct sock *sk;
164 bool delivered = false;
165 __u8 hash;
166 struct net *net;
167
168 saddr = &ipv6_hdr(skb)->saddr;
169 daddr = saddr + 1;
170
171 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
172
173 read_lock(&raw_v6_hashinfo.lock);
174 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
175
176 if (sk == NULL)
177 goto out;
178
179 net = dev_net(skb->dev);
180 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
181
182 while (sk) {
183 int filtered;
184
185 delivered = true;
186 switch (nexthdr) {
187 case IPPROTO_ICMPV6:
188 filtered = icmpv6_filter(sk, skb);
189 break;
190
191 #if IS_ENABLED(CONFIG_IPV6_MIP6)
192 case IPPROTO_MH:
193 {
194 /* XXX: To validate MH only once for each packet,
195 * this is placed here. It should be after checking
196 * xfrm policy, however it doesn't. The checking xfrm
197 * policy is placed in rawv6_rcv() because it is
198 * required for each socket.
199 */
200 mh_filter_t *filter;
201
202 filter = rcu_dereference(mh_filter);
203 filtered = filter ? (*filter)(sk, skb) : 0;
204 break;
205 }
206 #endif
207 default:
208 filtered = 0;
209 break;
210 }
211
212 if (filtered < 0)
213 break;
214 if (filtered == 0) {
215 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
216
217 /* Not releasing hash table! */
218 if (clone) {
219 nf_reset(clone);
220 rawv6_rcv(sk, clone);
221 }
222 }
223 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
224 IP6CB(skb)->iif);
225 }
226 out:
227 read_unlock(&raw_v6_hashinfo.lock);
228 return delivered;
229 }
230
231 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
232 {
233 struct sock *raw_sk;
234
235 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
236 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
237 raw_sk = NULL;
238
239 return raw_sk != NULL;
240 }
241
242 /* This cleans up af_inet6 a bit. -DaveM */
243 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
244 {
245 struct inet_sock *inet = inet_sk(sk);
246 struct ipv6_pinfo *np = inet6_sk(sk);
247 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
248 __be32 v4addr = 0;
249 int addr_type;
250 int err;
251
252 if (addr_len < SIN6_LEN_RFC2133)
253 return -EINVAL;
254 addr_type = ipv6_addr_type(&addr->sin6_addr);
255
256 /* Raw sockets are IPv6 only */
257 if (addr_type == IPV6_ADDR_MAPPED)
258 return -EADDRNOTAVAIL;
259
260 lock_sock(sk);
261
262 err = -EINVAL;
263 if (sk->sk_state != TCP_CLOSE)
264 goto out;
265
266 rcu_read_lock();
267 /* Check if the address belongs to the host. */
268 if (addr_type != IPV6_ADDR_ANY) {
269 struct net_device *dev = NULL;
270
271 if (__ipv6_addr_needs_scope_id(addr_type)) {
272 if (addr_len >= sizeof(struct sockaddr_in6) &&
273 addr->sin6_scope_id) {
274 /* Override any existing binding, if another
275 * one is supplied by user.
276 */
277 sk->sk_bound_dev_if = addr->sin6_scope_id;
278 }
279
280 /* Binding to link-local address requires an interface */
281 if (!sk->sk_bound_dev_if)
282 goto out_unlock;
283
284 err = -ENODEV;
285 dev = dev_get_by_index_rcu(sock_net(sk),
286 sk->sk_bound_dev_if);
287 if (!dev)
288 goto out_unlock;
289 }
290
291 /* ipv4 addr of the socket is invalid. Only the
292 * unspecified and mapped address have a v4 equivalent.
293 */
294 v4addr = LOOPBACK4_IPV6;
295 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
296 err = -EADDRNOTAVAIL;
297 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
298 dev, 0)) {
299 goto out_unlock;
300 }
301 }
302 }
303
304 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
305 np->rcv_saddr = addr->sin6_addr;
306 if (!(addr_type & IPV6_ADDR_MULTICAST))
307 np->saddr = addr->sin6_addr;
308 err = 0;
309 out_unlock:
310 rcu_read_unlock();
311 out:
312 release_sock(sk);
313 return err;
314 }
315
316 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
317 struct inet6_skb_parm *opt,
318 u8 type, u8 code, int offset, __be32 info)
319 {
320 struct inet_sock *inet = inet_sk(sk);
321 struct ipv6_pinfo *np = inet6_sk(sk);
322 int err;
323 int harderr;
324
325 /* Report error on raw socket, if:
326 1. User requested recverr.
327 2. Socket is connected (otherwise the error indication
328 is useless without recverr and error is hard.
329 */
330 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
331 return;
332
333 harderr = icmpv6_err_convert(type, code, &err);
334 if (type == ICMPV6_PKT_TOOBIG) {
335 ip6_sk_update_pmtu(skb, sk, info);
336 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
337 }
338 if (type == NDISC_REDIRECT) {
339 ip6_sk_redirect(skb, sk);
340 return;
341 }
342 if (np->recverr) {
343 u8 *payload = skb->data;
344 if (!inet->hdrincl)
345 payload += offset;
346 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
347 }
348
349 if (np->recverr || harderr) {
350 sk->sk_err = err;
351 sk->sk_error_report(sk);
352 }
353 }
354
355 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
356 u8 type, u8 code, int inner_offset, __be32 info)
357 {
358 struct sock *sk;
359 int hash;
360 const struct in6_addr *saddr, *daddr;
361 struct net *net;
362
363 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
364
365 read_lock(&raw_v6_hashinfo.lock);
366 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
367 if (sk != NULL) {
368 /* Note: ipv6_hdr(skb) != skb->data */
369 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
370 saddr = &ip6h->saddr;
371 daddr = &ip6h->daddr;
372 net = dev_net(skb->dev);
373
374 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
375 IP6CB(skb)->iif))) {
376 rawv6_err(sk, skb, NULL, type, code,
377 inner_offset, info);
378 sk = sk_next(sk);
379 }
380 }
381 read_unlock(&raw_v6_hashinfo.lock);
382 }
383
384 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
385 {
386 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
387 skb_checksum_complete(skb)) {
388 atomic_inc(&sk->sk_drops);
389 kfree_skb(skb);
390 return NET_RX_DROP;
391 }
392
393 /* Charge it to the socket. */
394 skb_dst_drop(skb);
395 if (sock_queue_rcv_skb(sk, skb) < 0) {
396 kfree_skb(skb);
397 return NET_RX_DROP;
398 }
399
400 return 0;
401 }
402
403 /*
404 * This is next to useless...
405 * if we demultiplex in network layer we don't need the extra call
406 * just to queue the skb...
407 * maybe we could have the network decide upon a hint if it
408 * should call raw_rcv for demultiplexing
409 */
410 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
411 {
412 struct inet_sock *inet = inet_sk(sk);
413 struct raw6_sock *rp = raw6_sk(sk);
414
415 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
416 atomic_inc(&sk->sk_drops);
417 kfree_skb(skb);
418 return NET_RX_DROP;
419 }
420
421 if (!rp->checksum)
422 skb->ip_summed = CHECKSUM_UNNECESSARY;
423
424 if (skb->ip_summed == CHECKSUM_COMPLETE) {
425 skb_postpull_rcsum(skb, skb_network_header(skb),
426 skb_network_header_len(skb));
427 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
428 &ipv6_hdr(skb)->daddr,
429 skb->len, inet->inet_num, skb->csum))
430 skb->ip_summed = CHECKSUM_UNNECESSARY;
431 }
432 if (!skb_csum_unnecessary(skb))
433 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
434 &ipv6_hdr(skb)->daddr,
435 skb->len,
436 inet->inet_num, 0));
437
438 if (inet->hdrincl) {
439 if (skb_checksum_complete(skb)) {
440 atomic_inc(&sk->sk_drops);
441 kfree_skb(skb);
442 return NET_RX_DROP;
443 }
444 }
445
446 rawv6_rcv_skb(sk, skb);
447 return 0;
448 }
449
450
451 /*
452 * This should be easy, if there is something there
453 * we return it, otherwise we block.
454 */
455
456 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
457 struct msghdr *msg, size_t len,
458 int noblock, int flags, int *addr_len)
459 {
460 struct ipv6_pinfo *np = inet6_sk(sk);
461 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
462 struct sk_buff *skb;
463 size_t copied;
464 int err;
465
466 if (flags & MSG_OOB)
467 return -EOPNOTSUPP;
468
469 if (addr_len)
470 *addr_len=sizeof(*sin6);
471
472 if (flags & MSG_ERRQUEUE)
473 return ipv6_recv_error(sk, msg, len);
474
475 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
476 return ipv6_recv_rxpmtu(sk, msg, len);
477
478 skb = skb_recv_datagram(sk, flags, noblock, &err);
479 if (!skb)
480 goto out;
481
482 copied = skb->len;
483 if (copied > len) {
484 copied = len;
485 msg->msg_flags |= MSG_TRUNC;
486 }
487
488 if (skb_csum_unnecessary(skb)) {
489 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
490 } else if (msg->msg_flags&MSG_TRUNC) {
491 if (__skb_checksum_complete(skb))
492 goto csum_copy_err;
493 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
494 } else {
495 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
496 if (err == -EINVAL)
497 goto csum_copy_err;
498 }
499 if (err)
500 goto out_free;
501
502 /* Copy the address. */
503 if (sin6) {
504 sin6->sin6_family = AF_INET6;
505 sin6->sin6_port = 0;
506 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
507 sin6->sin6_flowinfo = 0;
508 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
509 IP6CB(skb)->iif);
510 }
511
512 sock_recv_ts_and_drops(msg, sk, skb);
513
514 if (np->rxopt.all)
515 ip6_datagram_recv_ctl(sk, msg, skb);
516
517 err = copied;
518 if (flags & MSG_TRUNC)
519 err = skb->len;
520
521 out_free:
522 skb_free_datagram(sk, skb);
523 out:
524 return err;
525
526 csum_copy_err:
527 skb_kill_datagram(sk, skb, flags);
528
529 /* Error for blocking case is chosen to masquerade
530 as some normal condition.
531 */
532 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
533 goto out;
534 }
535
536 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
537 struct raw6_sock *rp)
538 {
539 struct sk_buff *skb;
540 int err = 0;
541 int offset;
542 int len;
543 int total_len;
544 __wsum tmp_csum;
545 __sum16 csum;
546
547 if (!rp->checksum)
548 goto send;
549
550 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
551 goto out;
552
553 offset = rp->offset;
554 total_len = inet_sk(sk)->cork.base.length;
555 if (offset >= total_len - 1) {
556 err = -EINVAL;
557 ip6_flush_pending_frames(sk);
558 goto out;
559 }
560
561 /* should be check HW csum miyazawa */
562 if (skb_queue_len(&sk->sk_write_queue) == 1) {
563 /*
564 * Only one fragment on the socket.
565 */
566 tmp_csum = skb->csum;
567 } else {
568 struct sk_buff *csum_skb = NULL;
569 tmp_csum = 0;
570
571 skb_queue_walk(&sk->sk_write_queue, skb) {
572 tmp_csum = csum_add(tmp_csum, skb->csum);
573
574 if (csum_skb)
575 continue;
576
577 len = skb->len - skb_transport_offset(skb);
578 if (offset >= len) {
579 offset -= len;
580 continue;
581 }
582
583 csum_skb = skb;
584 }
585
586 skb = csum_skb;
587 }
588
589 offset += skb_transport_offset(skb);
590 if (skb_copy_bits(skb, offset, &csum, 2))
591 BUG();
592
593 /* in case cksum was not initialized */
594 if (unlikely(csum))
595 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
596
597 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
598 total_len, fl6->flowi6_proto, tmp_csum);
599
600 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
601 csum = CSUM_MANGLED_0;
602
603 if (skb_store_bits(skb, offset, &csum, 2))
604 BUG();
605
606 send:
607 err = ip6_push_pending_frames(sk);
608 out:
609 return err;
610 }
611
612 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
613 struct flowi6 *fl6, struct dst_entry **dstp,
614 unsigned int flags)
615 {
616 struct ipv6_pinfo *np = inet6_sk(sk);
617 struct ipv6hdr *iph;
618 struct sk_buff *skb;
619 int err;
620 struct rt6_info *rt = (struct rt6_info *)*dstp;
621 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
622 int tlen = rt->dst.dev->needed_tailroom;
623
624 if (length > rt->dst.dev->mtu) {
625 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
626 return -EMSGSIZE;
627 }
628 if (flags&MSG_PROBE)
629 goto out;
630
631 skb = sock_alloc_send_skb(sk,
632 length + hlen + tlen + 15,
633 flags & MSG_DONTWAIT, &err);
634 if (skb == NULL)
635 goto error;
636 skb_reserve(skb, hlen);
637
638 skb->protocol = htons(ETH_P_IPV6);
639 skb->priority = sk->sk_priority;
640 skb->mark = sk->sk_mark;
641 skb_dst_set(skb, &rt->dst);
642 *dstp = NULL;
643
644 skb_put(skb, length);
645 skb_reset_network_header(skb);
646 iph = ipv6_hdr(skb);
647
648 skb->ip_summed = CHECKSUM_NONE;
649
650 skb->transport_header = skb->network_header;
651 err = memcpy_fromiovecend((void *)iph, from, 0, length);
652 if (err)
653 goto error_fault;
654
655 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
656 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
657 rt->dst.dev, dst_output);
658 if (err > 0)
659 err = net_xmit_errno(err);
660 if (err)
661 goto error;
662 out:
663 return 0;
664
665 error_fault:
666 err = -EFAULT;
667 kfree_skb(skb);
668 error:
669 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
670 if (err == -ENOBUFS && !np->recverr)
671 err = 0;
672 return err;
673 }
674
675 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg)
676 {
677 struct iovec *iov;
678 u8 __user *type = NULL;
679 u8 __user *code = NULL;
680 u8 len = 0;
681 int probed = 0;
682 int i;
683
684 if (!msg->msg_iov)
685 return 0;
686
687 for (i = 0; i < msg->msg_iovlen; i++) {
688 iov = &msg->msg_iov[i];
689 if (!iov)
690 continue;
691
692 switch (fl6->flowi6_proto) {
693 case IPPROTO_ICMPV6:
694 /* check if one-byte field is readable or not. */
695 if (iov->iov_base && iov->iov_len < 1)
696 break;
697
698 if (!type) {
699 type = iov->iov_base;
700 /* check if code field is readable or not. */
701 if (iov->iov_len > 1)
702 code = type + 1;
703 } else if (!code)
704 code = iov->iov_base;
705
706 if (type && code) {
707 if (get_user(fl6->fl6_icmp_type, type) ||
708 get_user(fl6->fl6_icmp_code, code))
709 return -EFAULT;
710 probed = 1;
711 }
712 break;
713 case IPPROTO_MH:
714 if (iov->iov_base && iov->iov_len < 1)
715 break;
716 /* check if type field is readable or not. */
717 if (iov->iov_len > 2 - len) {
718 u8 __user *p = iov->iov_base;
719 if (get_user(fl6->fl6_mh_type, &p[2 - len]))
720 return -EFAULT;
721 probed = 1;
722 } else
723 len += iov->iov_len;
724
725 break;
726 default:
727 probed = 1;
728 break;
729 }
730 if (probed)
731 break;
732 }
733 return 0;
734 }
735
736 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
737 struct msghdr *msg, size_t len)
738 {
739 struct ipv6_txoptions opt_space;
740 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
741 struct in6_addr *daddr, *final_p, final;
742 struct inet_sock *inet = inet_sk(sk);
743 struct ipv6_pinfo *np = inet6_sk(sk);
744 struct raw6_sock *rp = raw6_sk(sk);
745 struct ipv6_txoptions *opt = NULL;
746 struct ip6_flowlabel *flowlabel = NULL;
747 struct dst_entry *dst = NULL;
748 struct flowi6 fl6;
749 int addr_len = msg->msg_namelen;
750 int hlimit = -1;
751 int tclass = -1;
752 int dontfrag = -1;
753 u16 proto;
754 int err;
755
756 /* Rough check on arithmetic overflow,
757 better check is made in ip6_append_data().
758 */
759 if (len > INT_MAX)
760 return -EMSGSIZE;
761
762 /* Mirror BSD error message compatibility */
763 if (msg->msg_flags & MSG_OOB)
764 return -EOPNOTSUPP;
765
766 /*
767 * Get and verify the address.
768 */
769 memset(&fl6, 0, sizeof(fl6));
770
771 fl6.flowi6_mark = sk->sk_mark;
772
773 if (sin6) {
774 if (addr_len < SIN6_LEN_RFC2133)
775 return -EINVAL;
776
777 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
778 return -EAFNOSUPPORT;
779
780 /* port is the proto value [0..255] carried in nexthdr */
781 proto = ntohs(sin6->sin6_port);
782
783 if (!proto)
784 proto = inet->inet_num;
785 else if (proto != inet->inet_num)
786 return -EINVAL;
787
788 if (proto > 255)
789 return -EINVAL;
790
791 daddr = &sin6->sin6_addr;
792 if (np->sndflow) {
793 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
794 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
795 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
796 if (flowlabel == NULL)
797 return -EINVAL;
798 daddr = &flowlabel->dst;
799 }
800 }
801
802 /*
803 * Otherwise it will be difficult to maintain
804 * sk->sk_dst_cache.
805 */
806 if (sk->sk_state == TCP_ESTABLISHED &&
807 ipv6_addr_equal(daddr, &np->daddr))
808 daddr = &np->daddr;
809
810 if (addr_len >= sizeof(struct sockaddr_in6) &&
811 sin6->sin6_scope_id &&
812 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
813 fl6.flowi6_oif = sin6->sin6_scope_id;
814 } else {
815 if (sk->sk_state != TCP_ESTABLISHED)
816 return -EDESTADDRREQ;
817
818 proto = inet->inet_num;
819 daddr = &np->daddr;
820 fl6.flowlabel = np->flow_label;
821 }
822
823 if (fl6.flowi6_oif == 0)
824 fl6.flowi6_oif = sk->sk_bound_dev_if;
825
826 if (msg->msg_controllen) {
827 opt = &opt_space;
828 memset(opt, 0, sizeof(struct ipv6_txoptions));
829 opt->tot_len = sizeof(struct ipv6_txoptions);
830
831 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
832 &hlimit, &tclass, &dontfrag);
833 if (err < 0) {
834 fl6_sock_release(flowlabel);
835 return err;
836 }
837 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
838 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
839 if (flowlabel == NULL)
840 return -EINVAL;
841 }
842 if (!(opt->opt_nflen|opt->opt_flen))
843 opt = NULL;
844 }
845 if (opt == NULL)
846 opt = np->opt;
847 if (flowlabel)
848 opt = fl6_merge_options(&opt_space, flowlabel, opt);
849 opt = ipv6_fixup_options(&opt_space, opt);
850
851 fl6.flowi6_proto = proto;
852 err = rawv6_probe_proto_opt(&fl6, msg);
853 if (err)
854 goto out;
855
856 if (!ipv6_addr_any(daddr))
857 fl6.daddr = *daddr;
858 else
859 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
860 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
861 fl6.saddr = np->saddr;
862
863 final_p = fl6_update_dst(&fl6, opt, &final);
864
865 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
866 fl6.flowi6_oif = np->mcast_oif;
867 else if (!fl6.flowi6_oif)
868 fl6.flowi6_oif = np->ucast_oif;
869 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
870
871 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
872 if (IS_ERR(dst)) {
873 err = PTR_ERR(dst);
874 goto out;
875 }
876 if (hlimit < 0) {
877 if (ipv6_addr_is_multicast(&fl6.daddr))
878 hlimit = np->mcast_hops;
879 else
880 hlimit = np->hop_limit;
881 if (hlimit < 0)
882 hlimit = ip6_dst_hoplimit(dst);
883 }
884
885 if (tclass < 0)
886 tclass = np->tclass;
887
888 if (dontfrag < 0)
889 dontfrag = np->dontfrag;
890
891 if (msg->msg_flags&MSG_CONFIRM)
892 goto do_confirm;
893
894 back_from_confirm:
895 if (inet->hdrincl)
896 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags);
897 else {
898 lock_sock(sk);
899 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
900 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
901 msg->msg_flags, dontfrag);
902
903 if (err)
904 ip6_flush_pending_frames(sk);
905 else if (!(msg->msg_flags & MSG_MORE))
906 err = rawv6_push_pending_frames(sk, &fl6, rp);
907 release_sock(sk);
908 }
909 done:
910 dst_release(dst);
911 out:
912 fl6_sock_release(flowlabel);
913 return err<0?err:len;
914 do_confirm:
915 dst_confirm(dst);
916 if (!(msg->msg_flags & MSG_PROBE) || len)
917 goto back_from_confirm;
918 err = 0;
919 goto done;
920 }
921
922 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
923 char __user *optval, int optlen)
924 {
925 switch (optname) {
926 case ICMPV6_FILTER:
927 if (optlen > sizeof(struct icmp6_filter))
928 optlen = sizeof(struct icmp6_filter);
929 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
930 return -EFAULT;
931 return 0;
932 default:
933 return -ENOPROTOOPT;
934 }
935
936 return 0;
937 }
938
939 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
940 char __user *optval, int __user *optlen)
941 {
942 int len;
943
944 switch (optname) {
945 case ICMPV6_FILTER:
946 if (get_user(len, optlen))
947 return -EFAULT;
948 if (len < 0)
949 return -EINVAL;
950 if (len > sizeof(struct icmp6_filter))
951 len = sizeof(struct icmp6_filter);
952 if (put_user(len, optlen))
953 return -EFAULT;
954 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
955 return -EFAULT;
956 return 0;
957 default:
958 return -ENOPROTOOPT;
959 }
960
961 return 0;
962 }
963
964
965 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
966 char __user *optval, unsigned int optlen)
967 {
968 struct raw6_sock *rp = raw6_sk(sk);
969 int val;
970
971 if (get_user(val, (int __user *)optval))
972 return -EFAULT;
973
974 switch (optname) {
975 case IPV6_CHECKSUM:
976 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
977 level == IPPROTO_IPV6) {
978 /*
979 * RFC3542 tells that IPV6_CHECKSUM socket
980 * option in the IPPROTO_IPV6 level is not
981 * allowed on ICMPv6 sockets.
982 * If you want to set it, use IPPROTO_RAW
983 * level IPV6_CHECKSUM socket option
984 * (Linux extension).
985 */
986 return -EINVAL;
987 }
988
989 /* You may get strange result with a positive odd offset;
990 RFC2292bis agrees with me. */
991 if (val > 0 && (val&1))
992 return -EINVAL;
993 if (val < 0) {
994 rp->checksum = 0;
995 } else {
996 rp->checksum = 1;
997 rp->offset = val;
998 }
999
1000 return 0;
1001
1002 default:
1003 return -ENOPROTOOPT;
1004 }
1005 }
1006
1007 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1008 char __user *optval, unsigned int optlen)
1009 {
1010 switch (level) {
1011 case SOL_RAW:
1012 break;
1013
1014 case SOL_ICMPV6:
1015 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1016 return -EOPNOTSUPP;
1017 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1018 case SOL_IPV6:
1019 if (optname == IPV6_CHECKSUM)
1020 break;
1021 default:
1022 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1023 }
1024
1025 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1026 }
1027
1028 #ifdef CONFIG_COMPAT
1029 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1030 char __user *optval, unsigned int optlen)
1031 {
1032 switch (level) {
1033 case SOL_RAW:
1034 break;
1035 case SOL_ICMPV6:
1036 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1037 return -EOPNOTSUPP;
1038 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1039 case SOL_IPV6:
1040 if (optname == IPV6_CHECKSUM)
1041 break;
1042 default:
1043 return compat_ipv6_setsockopt(sk, level, optname,
1044 optval, optlen);
1045 }
1046 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1047 }
1048 #endif
1049
1050 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1051 char __user *optval, int __user *optlen)
1052 {
1053 struct raw6_sock *rp = raw6_sk(sk);
1054 int val, len;
1055
1056 if (get_user(len,optlen))
1057 return -EFAULT;
1058
1059 switch (optname) {
1060 case IPV6_CHECKSUM:
1061 /*
1062 * We allow getsockopt() for IPPROTO_IPV6-level
1063 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1064 * since RFC3542 is silent about it.
1065 */
1066 if (rp->checksum == 0)
1067 val = -1;
1068 else
1069 val = rp->offset;
1070 break;
1071
1072 default:
1073 return -ENOPROTOOPT;
1074 }
1075
1076 len = min_t(unsigned int, sizeof(int), len);
1077
1078 if (put_user(len, optlen))
1079 return -EFAULT;
1080 if (copy_to_user(optval,&val,len))
1081 return -EFAULT;
1082 return 0;
1083 }
1084
1085 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1086 char __user *optval, int __user *optlen)
1087 {
1088 switch (level) {
1089 case SOL_RAW:
1090 break;
1091
1092 case SOL_ICMPV6:
1093 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1094 return -EOPNOTSUPP;
1095 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1096 case SOL_IPV6:
1097 if (optname == IPV6_CHECKSUM)
1098 break;
1099 default:
1100 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1101 }
1102
1103 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1104 }
1105
1106 #ifdef CONFIG_COMPAT
1107 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1108 char __user *optval, int __user *optlen)
1109 {
1110 switch (level) {
1111 case SOL_RAW:
1112 break;
1113 case SOL_ICMPV6:
1114 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1115 return -EOPNOTSUPP;
1116 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1117 case SOL_IPV6:
1118 if (optname == IPV6_CHECKSUM)
1119 break;
1120 default:
1121 return compat_ipv6_getsockopt(sk, level, optname,
1122 optval, optlen);
1123 }
1124 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1125 }
1126 #endif
1127
1128 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1129 {
1130 switch (cmd) {
1131 case SIOCOUTQ: {
1132 int amount = sk_wmem_alloc_get(sk);
1133
1134 return put_user(amount, (int __user *)arg);
1135 }
1136 case SIOCINQ: {
1137 struct sk_buff *skb;
1138 int amount = 0;
1139
1140 spin_lock_bh(&sk->sk_receive_queue.lock);
1141 skb = skb_peek(&sk->sk_receive_queue);
1142 if (skb != NULL)
1143 amount = skb_tail_pointer(skb) -
1144 skb_transport_header(skb);
1145 spin_unlock_bh(&sk->sk_receive_queue.lock);
1146 return put_user(amount, (int __user *)arg);
1147 }
1148
1149 default:
1150 #ifdef CONFIG_IPV6_MROUTE
1151 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1152 #else
1153 return -ENOIOCTLCMD;
1154 #endif
1155 }
1156 }
1157
1158 #ifdef CONFIG_COMPAT
1159 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1160 {
1161 switch (cmd) {
1162 case SIOCOUTQ:
1163 case SIOCINQ:
1164 return -ENOIOCTLCMD;
1165 default:
1166 #ifdef CONFIG_IPV6_MROUTE
1167 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1168 #else
1169 return -ENOIOCTLCMD;
1170 #endif
1171 }
1172 }
1173 #endif
1174
1175 static void rawv6_close(struct sock *sk, long timeout)
1176 {
1177 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1178 ip6_ra_control(sk, -1);
1179 ip6mr_sk_done(sk);
1180 sk_common_release(sk);
1181 }
1182
1183 static void raw6_destroy(struct sock *sk)
1184 {
1185 lock_sock(sk);
1186 ip6_flush_pending_frames(sk);
1187 release_sock(sk);
1188
1189 inet6_destroy_sock(sk);
1190 }
1191
1192 static int rawv6_init_sk(struct sock *sk)
1193 {
1194 struct raw6_sock *rp = raw6_sk(sk);
1195
1196 switch (inet_sk(sk)->inet_num) {
1197 case IPPROTO_ICMPV6:
1198 rp->checksum = 1;
1199 rp->offset = 2;
1200 break;
1201 case IPPROTO_MH:
1202 rp->checksum = 1;
1203 rp->offset = 4;
1204 break;
1205 default:
1206 break;
1207 }
1208 return 0;
1209 }
1210
1211 struct proto rawv6_prot = {
1212 .name = "RAWv6",
1213 .owner = THIS_MODULE,
1214 .close = rawv6_close,
1215 .destroy = raw6_destroy,
1216 .connect = ip6_datagram_connect,
1217 .disconnect = udp_disconnect,
1218 .ioctl = rawv6_ioctl,
1219 .init = rawv6_init_sk,
1220 .setsockopt = rawv6_setsockopt,
1221 .getsockopt = rawv6_getsockopt,
1222 .sendmsg = rawv6_sendmsg,
1223 .recvmsg = rawv6_recvmsg,
1224 .bind = rawv6_bind,
1225 .backlog_rcv = rawv6_rcv_skb,
1226 .hash = raw_hash_sk,
1227 .unhash = raw_unhash_sk,
1228 .obj_size = sizeof(struct raw6_sock),
1229 .h.raw_hash = &raw_v6_hashinfo,
1230 #ifdef CONFIG_COMPAT
1231 .compat_setsockopt = compat_rawv6_setsockopt,
1232 .compat_getsockopt = compat_rawv6_getsockopt,
1233 .compat_ioctl = compat_rawv6_ioctl,
1234 #endif
1235 };
1236
1237 #ifdef CONFIG_PROC_FS
1238 static int raw6_seq_show(struct seq_file *seq, void *v)
1239 {
1240 if (v == SEQ_START_TOKEN) {
1241 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1242 } else {
1243 struct sock *sp = v;
1244 __u16 srcp = inet_sk(sp)->inet_num;
1245 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1246 raw_seq_private(seq)->bucket);
1247 }
1248 return 0;
1249 }
1250
1251 static const struct seq_operations raw6_seq_ops = {
1252 .start = raw_seq_start,
1253 .next = raw_seq_next,
1254 .stop = raw_seq_stop,
1255 .show = raw6_seq_show,
1256 };
1257
1258 static int raw6_seq_open(struct inode *inode, struct file *file)
1259 {
1260 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1261 }
1262
1263 static const struct file_operations raw6_seq_fops = {
1264 .owner = THIS_MODULE,
1265 .open = raw6_seq_open,
1266 .read = seq_read,
1267 .llseek = seq_lseek,
1268 .release = seq_release_net,
1269 };
1270
1271 static int __net_init raw6_init_net(struct net *net)
1272 {
1273 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops))
1274 return -ENOMEM;
1275
1276 return 0;
1277 }
1278
1279 static void __net_exit raw6_exit_net(struct net *net)
1280 {
1281 remove_proc_entry("raw6", net->proc_net);
1282 }
1283
1284 static struct pernet_operations raw6_net_ops = {
1285 .init = raw6_init_net,
1286 .exit = raw6_exit_net,
1287 };
1288
1289 int __init raw6_proc_init(void)
1290 {
1291 return register_pernet_subsys(&raw6_net_ops);
1292 }
1293
1294 void raw6_proc_exit(void)
1295 {
1296 unregister_pernet_subsys(&raw6_net_ops);
1297 }
1298 #endif /* CONFIG_PROC_FS */
1299
1300 /* Same as inet6_dgram_ops, sans udp_poll. */
1301 static const struct proto_ops inet6_sockraw_ops = {
1302 .family = PF_INET6,
1303 .owner = THIS_MODULE,
1304 .release = inet6_release,
1305 .bind = inet6_bind,
1306 .connect = inet_dgram_connect, /* ok */
1307 .socketpair = sock_no_socketpair, /* a do nothing */
1308 .accept = sock_no_accept, /* a do nothing */
1309 .getname = inet6_getname,
1310 .poll = datagram_poll, /* ok */
1311 .ioctl = inet6_ioctl, /* must change */
1312 .listen = sock_no_listen, /* ok */
1313 .shutdown = inet_shutdown, /* ok */
1314 .setsockopt = sock_common_setsockopt, /* ok */
1315 .getsockopt = sock_common_getsockopt, /* ok */
1316 .sendmsg = inet_sendmsg, /* ok */
1317 .recvmsg = sock_common_recvmsg, /* ok */
1318 .mmap = sock_no_mmap,
1319 .sendpage = sock_no_sendpage,
1320 #ifdef CONFIG_COMPAT
1321 .compat_setsockopt = compat_sock_common_setsockopt,
1322 .compat_getsockopt = compat_sock_common_getsockopt,
1323 #endif
1324 };
1325
1326 static struct inet_protosw rawv6_protosw = {
1327 .type = SOCK_RAW,
1328 .protocol = IPPROTO_IP, /* wild card */
1329 .prot = &rawv6_prot,
1330 .ops = &inet6_sockraw_ops,
1331 .no_check = UDP_CSUM_DEFAULT,
1332 .flags = INET_PROTOSW_REUSE,
1333 };
1334
1335 int __init rawv6_init(void)
1336 {
1337 int ret;
1338
1339 ret = inet6_register_protosw(&rawv6_protosw);
1340 if (ret)
1341 goto out;
1342 out:
1343 return ret;
1344 }
1345
1346 void rawv6_exit(void)
1347 {
1348 inet6_unregister_protosw(&rawv6_protosw);
1349 }
Linus' kernel tree