?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ipv6: Consider sk_bound_dev_if when binding a raw socket to an address
[linux-stable.git] / net / ipv6 / raw.c
blob8f409031d24557e6b8d5b178f7276bd293604335
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 <linux/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 struct raw_hashinfo raw_v6_hashinfo = {
69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
70 };
71 EXPORT_SYMBOL_GPL(raw_v6_hashinfo);
72
73 struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
74 unsigned short num, const struct in6_addr *loc_addr,
75 const struct in6_addr *rmt_addr, int dif, int sdif)
76 {
77 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
78
79 sk_for_each_from(sk)
80 if (inet_sk(sk)->inet_num == num) {
81
82 if (!net_eq(sock_net(sk), net))
83 continue;
84
85 if (!ipv6_addr_any(&sk->sk_v6_daddr) &&
86 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr))
87 continue;
88
89 if (sk->sk_bound_dev_if &&
90 sk->sk_bound_dev_if != dif &&
91 sk->sk_bound_dev_if != sdif)
92 continue;
93
94 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
95 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))
96 goto found;
97 if (is_multicast &&
98 inet6_mc_check(sk, loc_addr, rmt_addr))
99 goto found;
100 continue;
101 }
102 goto found;
103 }
104 sk = NULL;
105 found:
106 return sk;
107 }
108 EXPORT_SYMBOL_GPL(__raw_v6_lookup);
109
110 /*
111 * 0 - deliver
112 * 1 - block
113 */
114 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
115 {
116 struct icmp6hdr _hdr;
117 const struct icmp6hdr *hdr;
118
119 /* We require only the four bytes of the ICMPv6 header, not any
120 * additional bytes of message body in "struct icmp6hdr".
121 */
122 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
123 ICMPV6_HDRLEN, &_hdr);
124 if (hdr) {
125 const __u32 *data = &raw6_sk(sk)->filter.data[0];
126 unsigned int type = hdr->icmp6_type;
127
128 return (data[type >> 5] & (1U << (type & 31))) != 0;
129 }
130 return 1;
131 }
132
133 #if IS_ENABLED(CONFIG_IPV6_MIP6)
134 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
135
136 static mh_filter_t __rcu *mh_filter __read_mostly;
137
138 int rawv6_mh_filter_register(mh_filter_t filter)
139 {
140 rcu_assign_pointer(mh_filter, filter);
141 return 0;
142 }
143 EXPORT_SYMBOL(rawv6_mh_filter_register);
144
145 int rawv6_mh_filter_unregister(mh_filter_t filter)
146 {
147 RCU_INIT_POINTER(mh_filter, NULL);
148 synchronize_rcu();
149 return 0;
150 }
151 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
152
153 #endif
154
155 /*
156 * demultiplex raw sockets.
157 * (should consider queueing the skb in the sock receive_queue
158 * without calling rawv6.c)
159 *
160 * Caller owns SKB so we must make clones.
161 */
162 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
163 {
164 const struct in6_addr *saddr;
165 const struct in6_addr *daddr;
166 struct sock *sk;
167 bool delivered = false;
168 __u8 hash;
169 struct net *net;
170
171 saddr = &ipv6_hdr(skb)->saddr;
172 daddr = saddr + 1;
173
174 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
175
176 read_lock(&raw_v6_hashinfo.lock);
177 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
178
179 if (!sk)
180 goto out;
181
182 net = dev_net(skb->dev);
183 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr,
184 inet6_iif(skb), inet6_sdif(skb));
185
186 while (sk) {
187 int filtered;
188
189 delivered = true;
190 switch (nexthdr) {
191 case IPPROTO_ICMPV6:
192 filtered = icmpv6_filter(sk, skb);
193 break;
194
195 #if IS_ENABLED(CONFIG_IPV6_MIP6)
196 case IPPROTO_MH:
197 {
198 /* XXX: To validate MH only once for each packet,
199 * this is placed here. It should be after checking
200 * xfrm policy, however it doesn't. The checking xfrm
201 * policy is placed in rawv6_rcv() because it is
202 * required for each socket.
203 */
204 mh_filter_t *filter;
205
206 filter = rcu_dereference(mh_filter);
207 filtered = filter ? (*filter)(sk, skb) : 0;
208 break;
209 }
210 #endif
211 default:
212 filtered = 0;
213 break;
214 }
215
216 if (filtered < 0)
217 break;
218 if (filtered == 0) {
219 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
220
221 /* Not releasing hash table! */
222 if (clone) {
223 nf_reset(clone);
224 rawv6_rcv(sk, clone);
225 }
226 }
227 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
228 inet6_iif(skb), inet6_sdif(skb));
229 }
230 out:
231 read_unlock(&raw_v6_hashinfo.lock);
232 return delivered;
233 }
234
235 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
236 {
237 struct sock *raw_sk;
238
239 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
240 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
241 raw_sk = NULL;
242
243 return raw_sk != NULL;
244 }
245
246 /* This cleans up af_inet6 a bit. -DaveM */
247 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
248 {
249 struct inet_sock *inet = inet_sk(sk);
250 struct ipv6_pinfo *np = inet6_sk(sk);
251 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
252 __be32 v4addr = 0;
253 int addr_type;
254 int err;
255
256 if (addr_len < SIN6_LEN_RFC2133)
257 return -EINVAL;
258
259 if (addr->sin6_family != AF_INET6)
260 return -EINVAL;
261
262 addr_type = ipv6_addr_type(&addr->sin6_addr);
263
264 /* Raw sockets are IPv6 only */
265 if (addr_type == IPV6_ADDR_MAPPED)
266 return -EADDRNOTAVAIL;
267
268 lock_sock(sk);
269
270 err = -EINVAL;
271 if (sk->sk_state != TCP_CLOSE)
272 goto out;
273
274 rcu_read_lock();
275 /* Check if the address belongs to the host. */
276 if (addr_type != IPV6_ADDR_ANY) {
277 struct net_device *dev = NULL;
278
279 if (__ipv6_addr_needs_scope_id(addr_type)) {
280 if (addr_len >= sizeof(struct sockaddr_in6) &&
281 addr->sin6_scope_id) {
282 /* Override any existing binding, if another
283 * one is supplied by user.
284 */
285 sk->sk_bound_dev_if = addr->sin6_scope_id;
286 }
287
288 /* Binding to link-local address requires an interface */
289 if (!sk->sk_bound_dev_if)
290 goto out_unlock;
291 }
292
293 if (sk->sk_bound_dev_if) {
294 err = -ENODEV;
295 dev = dev_get_by_index_rcu(sock_net(sk),
296 sk->sk_bound_dev_if);
297 if (!dev)
298 goto out_unlock;
299 }
300
301 /* ipv4 addr of the socket is invalid. Only the
302 * unspecified and mapped address have a v4 equivalent.
303 */
304 v4addr = LOOPBACK4_IPV6;
305 if (!(addr_type & IPV6_ADDR_MULTICAST) &&
306 !sock_net(sk)->ipv6.sysctl.ip_nonlocal_bind) {
307 err = -EADDRNOTAVAIL;
308 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
309 dev, 0)) {
310 goto out_unlock;
311 }
312 }
313 }
314
315 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
316 sk->sk_v6_rcv_saddr = addr->sin6_addr;
317 if (!(addr_type & IPV6_ADDR_MULTICAST))
318 np->saddr = addr->sin6_addr;
319 err = 0;
320 out_unlock:
321 rcu_read_unlock();
322 out:
323 release_sock(sk);
324 return err;
325 }
326
327 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
328 struct inet6_skb_parm *opt,
329 u8 type, u8 code, int offset, __be32 info)
330 {
331 struct inet_sock *inet = inet_sk(sk);
332 struct ipv6_pinfo *np = inet6_sk(sk);
333 int err;
334 int harderr;
335
336 /* Report error on raw socket, if:
337 1. User requested recverr.
338 2. Socket is connected (otherwise the error indication
339 is useless without recverr and error is hard.
340 */
341 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
342 return;
343
344 harderr = icmpv6_err_convert(type, code, &err);
345 if (type == ICMPV6_PKT_TOOBIG) {
346 ip6_sk_update_pmtu(skb, sk, info);
347 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
348 }
349 if (type == NDISC_REDIRECT) {
350 ip6_sk_redirect(skb, sk);
351 return;
352 }
353 if (np->recverr) {
354 u8 *payload = skb->data;
355 if (!inet->hdrincl)
356 payload += offset;
357 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
358 }
359
360 if (np->recverr || harderr) {
361 sk->sk_err = err;
362 sk->sk_error_report(sk);
363 }
364 }
365
366 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
367 u8 type, u8 code, int inner_offset, __be32 info)
368 {
369 struct sock *sk;
370 int hash;
371 const struct in6_addr *saddr, *daddr;
372 struct net *net;
373
374 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
375
376 read_lock(&raw_v6_hashinfo.lock);
377 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
378 if (sk) {
379 /* Note: ipv6_hdr(skb) != skb->data */
380 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
381 saddr = &ip6h->saddr;
382 daddr = &ip6h->daddr;
383 net = dev_net(skb->dev);
384
385 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
386 inet6_iif(skb), inet6_iif(skb)))) {
387 rawv6_err(sk, skb, NULL, type, code,
388 inner_offset, info);
389 sk = sk_next(sk);
390 }
391 }
392 read_unlock(&raw_v6_hashinfo.lock);
393 }
394
395 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
396 {
397 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
398 skb_checksum_complete(skb)) {
399 atomic_inc(&sk->sk_drops);
400 kfree_skb(skb);
401 return NET_RX_DROP;
402 }
403
404 /* Charge it to the socket. */
405 skb_dst_drop(skb);
406 if (sock_queue_rcv_skb(sk, skb) < 0) {
407 kfree_skb(skb);
408 return NET_RX_DROP;
409 }
410
411 return 0;
412 }
413
414 /*
415 * This is next to useless...
416 * if we demultiplex in network layer we don't need the extra call
417 * just to queue the skb...
418 * maybe we could have the network decide upon a hint if it
419 * should call raw_rcv for demultiplexing
420 */
421 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
422 {
423 struct inet_sock *inet = inet_sk(sk);
424 struct raw6_sock *rp = raw6_sk(sk);
425
426 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
427 atomic_inc(&sk->sk_drops);
428 kfree_skb(skb);
429 return NET_RX_DROP;
430 }
431
432 if (!rp->checksum)
433 skb->ip_summed = CHECKSUM_UNNECESSARY;
434
435 if (skb->ip_summed == CHECKSUM_COMPLETE) {
436 skb_postpull_rcsum(skb, skb_network_header(skb),
437 skb_network_header_len(skb));
438 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
439 &ipv6_hdr(skb)->daddr,
440 skb->len, inet->inet_num, skb->csum))
441 skb->ip_summed = CHECKSUM_UNNECESSARY;
442 }
443 if (!skb_csum_unnecessary(skb))
444 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
445 &ipv6_hdr(skb)->daddr,
446 skb->len,
447 inet->inet_num, 0));
448
449 if (inet->hdrincl) {
450 if (skb_checksum_complete(skb)) {
451 atomic_inc(&sk->sk_drops);
452 kfree_skb(skb);
453 return NET_RX_DROP;
454 }
455 }
456
457 rawv6_rcv_skb(sk, skb);
458 return 0;
459 }
460
461
462 /*
463 * This should be easy, if there is something there
464 * we return it, otherwise we block.
465 */
466
467 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
468 int noblock, int flags, int *addr_len)
469 {
470 struct ipv6_pinfo *np = inet6_sk(sk);
471 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
472 struct sk_buff *skb;
473 size_t copied;
474 int err;
475
476 if (flags & MSG_OOB)
477 return -EOPNOTSUPP;
478
479 if (flags & MSG_ERRQUEUE)
480 return ipv6_recv_error(sk, msg, len, addr_len);
481
482 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
483 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
484
485 skb = skb_recv_datagram(sk, flags, noblock, &err);
486 if (!skb)
487 goto out;
488
489 copied = skb->len;
490 if (copied > len) {
491 copied = len;
492 msg->msg_flags |= MSG_TRUNC;
493 }
494
495 if (skb_csum_unnecessary(skb)) {
496 err = skb_copy_datagram_msg(skb, 0, msg, copied);
497 } else if (msg->msg_flags&MSG_TRUNC) {
498 if (__skb_checksum_complete(skb))
499 goto csum_copy_err;
500 err = skb_copy_datagram_msg(skb, 0, msg, copied);
501 } else {
502 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
503 if (err == -EINVAL)
504 goto csum_copy_err;
505 }
506 if (err)
507 goto out_free;
508
509 /* Copy the address. */
510 if (sin6) {
511 sin6->sin6_family = AF_INET6;
512 sin6->sin6_port = 0;
513 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
514 sin6->sin6_flowinfo = 0;
515 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
516 inet6_iif(skb));
517 *addr_len = sizeof(*sin6);
518 }
519
520 sock_recv_ts_and_drops(msg, sk, skb);
521
522 if (np->rxopt.all)
523 ip6_datagram_recv_ctl(sk, msg, skb);
524
525 err = copied;
526 if (flags & MSG_TRUNC)
527 err = skb->len;
528
529 out_free:
530 skb_free_datagram(sk, skb);
531 out:
532 return err;
533
534 csum_copy_err:
535 skb_kill_datagram(sk, skb, flags);
536
537 /* Error for blocking case is chosen to masquerade
538 as some normal condition.
539 */
540 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
541 goto out;
542 }
543
544 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
545 struct raw6_sock *rp)
546 {
547 struct sk_buff *skb;
548 int err = 0;
549 int offset;
550 int len;
551 int total_len;
552 __wsum tmp_csum;
553 __sum16 csum;
554
555 if (!rp->checksum)
556 goto send;
557
558 skb = skb_peek(&sk->sk_write_queue);
559 if (!skb)
560 goto out;
561
562 offset = rp->offset;
563 total_len = inet_sk(sk)->cork.base.length;
564 if (offset >= total_len - 1) {
565 err = -EINVAL;
566 ip6_flush_pending_frames(sk);
567 goto out;
568 }
569
570 /* should be check HW csum miyazawa */
571 if (skb_queue_len(&sk->sk_write_queue) == 1) {
572 /*
573 * Only one fragment on the socket.
574 */
575 tmp_csum = skb->csum;
576 } else {
577 struct sk_buff *csum_skb = NULL;
578 tmp_csum = 0;
579
580 skb_queue_walk(&sk->sk_write_queue, skb) {
581 tmp_csum = csum_add(tmp_csum, skb->csum);
582
583 if (csum_skb)
584 continue;
585
586 len = skb->len - skb_transport_offset(skb);
587 if (offset >= len) {
588 offset -= len;
589 continue;
590 }
591
592 csum_skb = skb;
593 }
594
595 skb = csum_skb;
596 }
597
598 offset += skb_transport_offset(skb);
599 err = skb_copy_bits(skb, offset, &csum, 2);
600 if (err < 0) {
601 ip6_flush_pending_frames(sk);
602 goto out;
603 }
604
605 /* in case cksum was not initialized */
606 if (unlikely(csum))
607 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
608
609 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
610 total_len, fl6->flowi6_proto, tmp_csum);
611
612 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
613 csum = CSUM_MANGLED_0;
614
615 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
616
617 send:
618 err = ip6_push_pending_frames(sk);
619 out:
620 return err;
621 }
622
623 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
624 struct flowi6 *fl6, struct dst_entry **dstp,
625 unsigned int flags)
626 {
627 struct ipv6_pinfo *np = inet6_sk(sk);
628 struct net *net = sock_net(sk);
629 struct ipv6hdr *iph;
630 struct sk_buff *skb;
631 int err;
632 struct rt6_info *rt = (struct rt6_info *)*dstp;
633 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
634 int tlen = rt->dst.dev->needed_tailroom;
635
636 if (length > rt->dst.dev->mtu) {
637 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
638 return -EMSGSIZE;
639 }
640 if (length < sizeof(struct ipv6hdr))
641 return -EINVAL;
642 if (flags&MSG_PROBE)
643 goto out;
644
645 skb = sock_alloc_send_skb(sk,
646 length + hlen + tlen + 15,
647 flags & MSG_DONTWAIT, &err);
648 if (!skb)
649 goto error;
650 skb_reserve(skb, hlen);
651
652 skb->protocol = htons(ETH_P_IPV6);
653 skb->priority = sk->sk_priority;
654 skb->mark = sk->sk_mark;
655
656 skb_put(skb, length);
657 skb_reset_network_header(skb);
658 iph = ipv6_hdr(skb);
659
660 skb->ip_summed = CHECKSUM_NONE;
661
662 if (flags & MSG_CONFIRM)
663 skb_set_dst_pending_confirm(skb, 1);
664
665 skb->transport_header = skb->network_header;
666 err = memcpy_from_msg(iph, msg, length);
667 if (err) {
668 err = -EFAULT;
669 kfree_skb(skb);
670 goto error;
671 }
672
673 skb_dst_set(skb, &rt->dst);
674 *dstp = NULL;
675
676 /* if egress device is enslaved to an L3 master device pass the
677 * skb to its handler for processing
678 */
679 skb = l3mdev_ip6_out(sk, skb);
680 if (unlikely(!skb))
681 return 0;
682
683 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev
684 * in the error path. Since skb has been freed, the dst could
685 * have been queued for deletion.
686 */
687 rcu_read_lock();
688 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
689 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
690 NULL, rt->dst.dev, dst_output);
691 if (err > 0)
692 err = net_xmit_errno(err);
693 if (err) {
694 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
695 rcu_read_unlock();
696 goto error_check;
697 }
698 rcu_read_unlock();
699 out:
700 return 0;
701
702 error:
703 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
704 error_check:
705 if (err == -ENOBUFS && !np->recverr)
706 err = 0;
707 return err;
708 }
709
710 struct raw6_frag_vec {
711 struct msghdr *msg;
712 int hlen;
713 char c[4];
714 };
715
716 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
717 {
718 int err = 0;
719 switch (fl6->flowi6_proto) {
720 case IPPROTO_ICMPV6:
721 rfv->hlen = 2;
722 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
723 if (!err) {
724 fl6->fl6_icmp_type = rfv->c[0];
725 fl6->fl6_icmp_code = rfv->c[1];
726 }
727 break;
728 case IPPROTO_MH:
729 rfv->hlen = 4;
730 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
731 if (!err)
732 fl6->fl6_mh_type = rfv->c[2];
733 }
734 return err;
735 }
736
737 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
738 struct sk_buff *skb)
739 {
740 struct raw6_frag_vec *rfv = from;
741
742 if (offset < rfv->hlen) {
743 int copy = min(rfv->hlen - offset, len);
744
745 if (skb->ip_summed == CHECKSUM_PARTIAL)
746 memcpy(to, rfv->c + offset, copy);
747 else
748 skb->csum = csum_block_add(
749 skb->csum,
750 csum_partial_copy_nocheck(rfv->c + offset,
751 to, copy, 0),
752 odd);
753
754 odd = 0;
755 offset += copy;
756 to += copy;
757 len -= copy;
758
759 if (!len)
760 return 0;
761 }
762
763 offset -= rfv->hlen;
764
765 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
766 }
767
768 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
769 {
770 struct ipv6_txoptions *opt_to_free = NULL;
771 struct ipv6_txoptions opt_space;
772 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
773 struct in6_addr *daddr, *final_p, final;
774 struct inet_sock *inet = inet_sk(sk);
775 struct ipv6_pinfo *np = inet6_sk(sk);
776 struct raw6_sock *rp = raw6_sk(sk);
777 struct ipv6_txoptions *opt = NULL;
778 struct ip6_flowlabel *flowlabel = NULL;
779 struct dst_entry *dst = NULL;
780 struct raw6_frag_vec rfv;
781 struct flowi6 fl6;
782 struct sockcm_cookie sockc;
783 struct ipcm6_cookie ipc6;
784 int addr_len = msg->msg_namelen;
785 u16 proto;
786 int err;
787
788 /* Rough check on arithmetic overflow,
789 better check is made in ip6_append_data().
790 */
791 if (len > INT_MAX)
792 return -EMSGSIZE;
793
794 /* Mirror BSD error message compatibility */
795 if (msg->msg_flags & MSG_OOB)
796 return -EOPNOTSUPP;
797
798 /*
799 * Get and verify the address.
800 */
801 memset(&fl6, 0, sizeof(fl6));
802
803 fl6.flowi6_mark = sk->sk_mark;
804 fl6.flowi6_uid = sk->sk_uid;
805
806 ipc6.hlimit = -1;
807 ipc6.tclass = -1;
808 ipc6.dontfrag = -1;
809 ipc6.opt = NULL;
810
811 if (sin6) {
812 if (addr_len < SIN6_LEN_RFC2133)
813 return -EINVAL;
814
815 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
816 return -EAFNOSUPPORT;
817
818 /* port is the proto value [0..255] carried in nexthdr */
819 proto = ntohs(sin6->sin6_port);
820
821 if (!proto)
822 proto = inet->inet_num;
823 else if (proto != inet->inet_num)
824 return -EINVAL;
825
826 if (proto > 255)
827 return -EINVAL;
828
829 daddr = &sin6->sin6_addr;
830 if (np->sndflow) {
831 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
832 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
833 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
834 if (!flowlabel)
835 return -EINVAL;
836 }
837 }
838
839 /*
840 * Otherwise it will be difficult to maintain
841 * sk->sk_dst_cache.
842 */
843 if (sk->sk_state == TCP_ESTABLISHED &&
844 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
845 daddr = &sk->sk_v6_daddr;
846
847 if (addr_len >= sizeof(struct sockaddr_in6) &&
848 sin6->sin6_scope_id &&
849 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
850 fl6.flowi6_oif = sin6->sin6_scope_id;
851 } else {
852 if (sk->sk_state != TCP_ESTABLISHED)
853 return -EDESTADDRREQ;
854
855 proto = inet->inet_num;
856 daddr = &sk->sk_v6_daddr;
857 fl6.flowlabel = np->flow_label;
858 }
859
860 if (fl6.flowi6_oif == 0)
861 fl6.flowi6_oif = sk->sk_bound_dev_if;
862
863 sockc.tsflags = sk->sk_tsflags;
864 if (msg->msg_controllen) {
865 opt = &opt_space;
866 memset(opt, 0, sizeof(struct ipv6_txoptions));
867 opt->tot_len = sizeof(struct ipv6_txoptions);
868 ipc6.opt = opt;
869
870 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc);
871 if (err < 0) {
872 fl6_sock_release(flowlabel);
873 return err;
874 }
875 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
876 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
877 if (!flowlabel)
878 return -EINVAL;
879 }
880 if (!(opt->opt_nflen|opt->opt_flen))
881 opt = NULL;
882 }
883 if (!opt) {
884 opt = txopt_get(np);
885 opt_to_free = opt;
886 }
887 if (flowlabel)
888 opt = fl6_merge_options(&opt_space, flowlabel, opt);
889 opt = ipv6_fixup_options(&opt_space, opt);
890
891 fl6.flowi6_proto = proto;
892 rfv.msg = msg;
893 rfv.hlen = 0;
894 err = rawv6_probe_proto_opt(&rfv, &fl6);
895 if (err)
896 goto out;
897
898 if (!ipv6_addr_any(daddr))
899 fl6.daddr = *daddr;
900 else
901 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
902 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
903 fl6.saddr = np->saddr;
904
905 final_p = fl6_update_dst(&fl6, opt, &final);
906
907 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
908 fl6.flowi6_oif = np->mcast_oif;
909 else if (!fl6.flowi6_oif)
910 fl6.flowi6_oif = np->ucast_oif;
911 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
912
913 if (inet->hdrincl)
914 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
915
916 if (ipc6.tclass < 0)
917 ipc6.tclass = np->tclass;
918
919 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
920
921 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
922 if (IS_ERR(dst)) {
923 err = PTR_ERR(dst);
924 goto out;
925 }
926 if (ipc6.hlimit < 0)
927 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
928
929 if (ipc6.dontfrag < 0)
930 ipc6.dontfrag = np->dontfrag;
931
932 if (msg->msg_flags&MSG_CONFIRM)
933 goto do_confirm;
934
935 back_from_confirm:
936 if (inet->hdrincl)
937 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst, msg->msg_flags);
938 else {
939 ipc6.opt = opt;
940 lock_sock(sk);
941 err = ip6_append_data(sk, raw6_getfrag, &rfv,
942 len, 0, &ipc6, &fl6, (struct rt6_info *)dst,
943 msg->msg_flags, &sockc);
944
945 if (err)
946 ip6_flush_pending_frames(sk);
947 else if (!(msg->msg_flags & MSG_MORE))
948 err = rawv6_push_pending_frames(sk, &fl6, rp);
949 release_sock(sk);
950 }
951 done:
952 dst_release(dst);
953 out:
954 fl6_sock_release(flowlabel);
955 txopt_put(opt_to_free);
956 return err < 0 ? err : len;
957 do_confirm:
958 if (msg->msg_flags & MSG_PROBE)
959 dst_confirm_neigh(dst, &fl6.daddr);
960 if (!(msg->msg_flags & MSG_PROBE) || len)
961 goto back_from_confirm;
962 err = 0;
963 goto done;
964 }
965
966 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
967 char __user *optval, int optlen)
968 {
969 switch (optname) {
970 case ICMPV6_FILTER:
971 if (optlen > sizeof(struct icmp6_filter))
972 optlen = sizeof(struct icmp6_filter);
973 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
974 return -EFAULT;
975 return 0;
976 default:
977 return -ENOPROTOOPT;
978 }
979
980 return 0;
981 }
982
983 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
984 char __user *optval, int __user *optlen)
985 {
986 int len;
987
988 switch (optname) {
989 case ICMPV6_FILTER:
990 if (get_user(len, optlen))
991 return -EFAULT;
992 if (len < 0)
993 return -EINVAL;
994 if (len > sizeof(struct icmp6_filter))
995 len = sizeof(struct icmp6_filter);
996 if (put_user(len, optlen))
997 return -EFAULT;
998 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
999 return -EFAULT;
1000 return 0;
1001 default:
1002 return -ENOPROTOOPT;
1003 }
1004
1005 return 0;
1006 }
1007
1008
1009 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
1010 char __user *optval, unsigned int optlen)
1011 {
1012 struct raw6_sock *rp = raw6_sk(sk);
1013 int val;
1014
1015 if (get_user(val, (int __user *)optval))
1016 return -EFAULT;
1017
1018 switch (optname) {
1019 case IPV6_HDRINCL:
1020 if (sk->sk_type != SOCK_RAW)
1021 return -EINVAL;
1022 inet_sk(sk)->hdrincl = !!val;
1023 return 0;
1024 case IPV6_CHECKSUM:
1025 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1026 level == IPPROTO_IPV6) {
1027 /*
1028 * RFC3542 tells that IPV6_CHECKSUM socket
1029 * option in the IPPROTO_IPV6 level is not
1030 * allowed on ICMPv6 sockets.
1031 * If you want to set it, use IPPROTO_RAW
1032 * level IPV6_CHECKSUM socket option
1033 * (Linux extension).
1034 */
1035 return -EINVAL;
1036 }
1037
1038 /* You may get strange result with a positive odd offset;
1039 RFC2292bis agrees with me. */
1040 if (val > 0 && (val&1))
1041 return -EINVAL;
1042 if (val < 0) {
1043 rp->checksum = 0;
1044 } else {
1045 rp->checksum = 1;
1046 rp->offset = val;
1047 }
1048
1049 return 0;
1050
1051 default:
1052 return -ENOPROTOOPT;
1053 }
1054 }
1055
1056 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1057 char __user *optval, unsigned int optlen)
1058 {
1059 switch (level) {
1060 case SOL_RAW:
1061 break;
1062
1063 case SOL_ICMPV6:
1064 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1065 return -EOPNOTSUPP;
1066 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1067 case SOL_IPV6:
1068 if (optname == IPV6_CHECKSUM ||
1069 optname == IPV6_HDRINCL)
1070 break;
1071 default:
1072 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1073 }
1074
1075 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1076 }
1077
1078 #ifdef CONFIG_COMPAT
1079 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1080 char __user *optval, unsigned int optlen)
1081 {
1082 switch (level) {
1083 case SOL_RAW:
1084 break;
1085 case SOL_ICMPV6:
1086 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1087 return -EOPNOTSUPP;
1088 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1089 case SOL_IPV6:
1090 if (optname == IPV6_CHECKSUM ||
1091 optname == IPV6_HDRINCL)
1092 break;
1093 default:
1094 return compat_ipv6_setsockopt(sk, level, optname,
1095 optval, optlen);
1096 }
1097 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1098 }
1099 #endif
1100
1101 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1102 char __user *optval, int __user *optlen)
1103 {
1104 struct raw6_sock *rp = raw6_sk(sk);
1105 int val, len;
1106
1107 if (get_user(len, optlen))
1108 return -EFAULT;
1109
1110 switch (optname) {
1111 case IPV6_HDRINCL:
1112 val = inet_sk(sk)->hdrincl;
1113 break;
1114 case IPV6_CHECKSUM:
1115 /*
1116 * We allow getsockopt() for IPPROTO_IPV6-level
1117 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1118 * since RFC3542 is silent about it.
1119 */
1120 if (rp->checksum == 0)
1121 val = -1;
1122 else
1123 val = rp->offset;
1124 break;
1125
1126 default:
1127 return -ENOPROTOOPT;
1128 }
1129
1130 len = min_t(unsigned int, sizeof(int), len);
1131
1132 if (put_user(len, optlen))
1133 return -EFAULT;
1134 if (copy_to_user(optval, &val, len))
1135 return -EFAULT;
1136 return 0;
1137 }
1138
1139 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1140 char __user *optval, int __user *optlen)
1141 {
1142 switch (level) {
1143 case SOL_RAW:
1144 break;
1145
1146 case SOL_ICMPV6:
1147 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1148 return -EOPNOTSUPP;
1149 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1150 case SOL_IPV6:
1151 if (optname == IPV6_CHECKSUM ||
1152 optname == IPV6_HDRINCL)
1153 break;
1154 default:
1155 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1156 }
1157
1158 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1159 }
1160
1161 #ifdef CONFIG_COMPAT
1162 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1163 char __user *optval, int __user *optlen)
1164 {
1165 switch (level) {
1166 case SOL_RAW:
1167 break;
1168 case SOL_ICMPV6:
1169 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1170 return -EOPNOTSUPP;
1171 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1172 case SOL_IPV6:
1173 if (optname == IPV6_CHECKSUM ||
1174 optname == IPV6_HDRINCL)
1175 break;
1176 default:
1177 return compat_ipv6_getsockopt(sk, level, optname,
1178 optval, optlen);
1179 }
1180 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1181 }
1182 #endif
1183
1184 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1185 {
1186 switch (cmd) {
1187 case SIOCOUTQ: {
1188 int amount = sk_wmem_alloc_get(sk);
1189
1190 return put_user(amount, (int __user *)arg);
1191 }
1192 case SIOCINQ: {
1193 struct sk_buff *skb;
1194 int amount = 0;
1195
1196 spin_lock_bh(&sk->sk_receive_queue.lock);
1197 skb = skb_peek(&sk->sk_receive_queue);
1198 if (skb)
1199 amount = skb->len;
1200 spin_unlock_bh(&sk->sk_receive_queue.lock);
1201 return put_user(amount, (int __user *)arg);
1202 }
1203
1204 default:
1205 #ifdef CONFIG_IPV6_MROUTE
1206 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1207 #else
1208 return -ENOIOCTLCMD;
1209 #endif
1210 }
1211 }
1212
1213 #ifdef CONFIG_COMPAT
1214 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1215 {
1216 switch (cmd) {
1217 case SIOCOUTQ:
1218 case SIOCINQ:
1219 return -ENOIOCTLCMD;
1220 default:
1221 #ifdef CONFIG_IPV6_MROUTE
1222 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1223 #else
1224 return -ENOIOCTLCMD;
1225 #endif
1226 }
1227 }
1228 #endif
1229
1230 static void rawv6_close(struct sock *sk, long timeout)
1231 {
1232 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1233 ip6_ra_control(sk, -1);
1234 ip6mr_sk_done(sk);
1235 sk_common_release(sk);
1236 }
1237
1238 static void raw6_destroy(struct sock *sk)
1239 {
1240 lock_sock(sk);
1241 ip6_flush_pending_frames(sk);
1242 release_sock(sk);
1243
1244 inet6_destroy_sock(sk);
1245 }
1246
1247 static int rawv6_init_sk(struct sock *sk)
1248 {
1249 struct raw6_sock *rp = raw6_sk(sk);
1250
1251 switch (inet_sk(sk)->inet_num) {
1252 case IPPROTO_ICMPV6:
1253 rp->checksum = 1;
1254 rp->offset = 2;
1255 break;
1256 case IPPROTO_MH:
1257 rp->checksum = 1;
1258 rp->offset = 4;
1259 break;
1260 default:
1261 break;
1262 }
1263 return 0;
1264 }
1265
1266 struct proto rawv6_prot = {
1267 .name = "RAWv6",
1268 .owner = THIS_MODULE,
1269 .close = rawv6_close,
1270 .destroy = raw6_destroy,
1271 .connect = ip6_datagram_connect_v6_only,
1272 .disconnect = __udp_disconnect,
1273 .ioctl = rawv6_ioctl,
1274 .init = rawv6_init_sk,
1275 .setsockopt = rawv6_setsockopt,
1276 .getsockopt = rawv6_getsockopt,
1277 .sendmsg = rawv6_sendmsg,
1278 .recvmsg = rawv6_recvmsg,
1279 .bind = rawv6_bind,
1280 .backlog_rcv = rawv6_rcv_skb,
1281 .hash = raw_hash_sk,
1282 .unhash = raw_unhash_sk,
1283 .obj_size = sizeof(struct raw6_sock),
1284 .h.raw_hash = &raw_v6_hashinfo,
1285 #ifdef CONFIG_COMPAT
1286 .compat_setsockopt = compat_rawv6_setsockopt,
1287 .compat_getsockopt = compat_rawv6_getsockopt,
1288 .compat_ioctl = compat_rawv6_ioctl,
1289 #endif
1290 .diag_destroy = raw_abort,
1291 };
1292
1293 #ifdef CONFIG_PROC_FS
1294 static int raw6_seq_show(struct seq_file *seq, void *v)
1295 {
1296 if (v == SEQ_START_TOKEN) {
1297 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1298 } else {
1299 struct sock *sp = v;
1300 __u16 srcp = inet_sk(sp)->inet_num;
1301 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1302 raw_seq_private(seq)->bucket);
1303 }
1304 return 0;
1305 }
1306
1307 static const struct seq_operations raw6_seq_ops = {
1308 .start = raw_seq_start,
1309 .next = raw_seq_next,
1310 .stop = raw_seq_stop,
1311 .show = raw6_seq_show,
1312 };
1313
1314 static int raw6_seq_open(struct inode *inode, struct file *file)
1315 {
1316 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1317 }
1318
1319 static const struct file_operations raw6_seq_fops = {
1320 .owner = THIS_MODULE,
1321 .open = raw6_seq_open,
1322 .read = seq_read,
1323 .llseek = seq_lseek,
1324 .release = seq_release_net,
1325 };
1326
1327 static int __net_init raw6_init_net(struct net *net)
1328 {
1329 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops))
1330 return -ENOMEM;
1331
1332 return 0;
1333 }
1334
1335 static void __net_exit raw6_exit_net(struct net *net)
1336 {
1337 remove_proc_entry("raw6", net->proc_net);
1338 }
1339
1340 static struct pernet_operations raw6_net_ops = {
1341 .init = raw6_init_net,
1342 .exit = raw6_exit_net,
1343 };
1344
1345 int __init raw6_proc_init(void)
1346 {
1347 return register_pernet_subsys(&raw6_net_ops);
1348 }
1349
1350 void raw6_proc_exit(void)
1351 {
1352 unregister_pernet_subsys(&raw6_net_ops);
1353 }
1354 #endif /* CONFIG_PROC_FS */
1355
1356 /* Same as inet6_dgram_ops, sans udp_poll. */
1357 const struct proto_ops inet6_sockraw_ops = {
1358 .family = PF_INET6,
1359 .owner = THIS_MODULE,
1360 .release = inet6_release,
1361 .bind = inet6_bind,
1362 .connect = inet_dgram_connect, /* ok */
1363 .socketpair = sock_no_socketpair, /* a do nothing */
1364 .accept = sock_no_accept, /* a do nothing */
1365 .getname = inet6_getname,
1366 .poll = datagram_poll, /* ok */
1367 .ioctl = inet6_ioctl, /* must change */
1368 .listen = sock_no_listen, /* ok */
1369 .shutdown = inet_shutdown, /* ok */
1370 .setsockopt = sock_common_setsockopt, /* ok */
1371 .getsockopt = sock_common_getsockopt, /* ok */
1372 .sendmsg = inet_sendmsg, /* ok */
1373 .recvmsg = sock_common_recvmsg, /* ok */
1374 .mmap = sock_no_mmap,
1375 .sendpage = sock_no_sendpage,
1376 #ifdef CONFIG_COMPAT
1377 .compat_setsockopt = compat_sock_common_setsockopt,
1378 .compat_getsockopt = compat_sock_common_getsockopt,
1379 #endif
1380 };
1381
1382 static struct inet_protosw rawv6_protosw = {
1383 .type = SOCK_RAW,
1384 .protocol = IPPROTO_IP, /* wild card */
1385 .prot = &rawv6_prot,
1386 .ops = &inet6_sockraw_ops,
1387 .flags = INET_PROTOSW_REUSE,
1388 };
1389
1390 int __init rawv6_init(void)
1391 {
1392 return inet6_register_protosw(&rawv6_protosw);
1393 }
1394
1395 void rawv6_exit(void)
1396 {
1397 inet6_unregister_protosw(&rawv6_protosw);
1398 }
Linux kernel stable tree