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