1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * RAW sockets for IPv6
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 *
9 * Adapted from linux/net/ipv4/raw.c
10 *
11 * Fixes:
12 * Hideaki YOSHIFUJI : sin6_scope_id support
13 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
14 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
15 */
16
17 #include <linux/errno.h>
18 #include <linux/types.h>
19 #include <linux/socket.h>
20 #include <linux/slab.h>
21 #include <linux/sockios.h>
22 #include <linux/net.h>
23 #include <linux/in6.h>
24 #include <linux/netdevice.h>
25 #include <linux/if_arp.h>
26 #include <linux/icmpv6.h>
27 #include <linux/netfilter.h>
28 #include <linux/netfilter_ipv6.h>
29 #include <linux/skbuff.h>
30 #include <linux/compat.h>
31 #include <linux/uaccess.h>
32 #include <asm/ioctls.h>
33
34 #include <net/net_namespace.h>
35 #include <net/ip.h>
36 #include <net/sock.h>
37 #include <net/snmp.h>
38
39 #include <net/ipv6.h>
40 #include <net/ndisc.h>
41 #include <net/protocol.h>
42 #include <net/ip6_route.h>
43 #include <net/ip6_checksum.h>
44 #include <net/addrconf.h>
45 #include <net/transp_v6.h>
46 #include <net/udp.h>
47 #include <net/inet_common.h>
48 #include <net/tcp_states.h>
49 #if IS_ENABLED(CONFIG_IPV6_MIP6)
50 #include <net/mip6.h>
51 #endif
52 #include <linux/mroute6.h>
53
54 #include <net/raw.h>
55 #include <net/rawv6.h>
56 #include <net/xfrm.h>
57
58 #include <linux/proc_fs.h>
59 #include <linux/seq_file.h>
60 #include <linux/export.h>
61
62 #define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */
63
64 struct raw_hashinfo raw_v6_hashinfo;
65 EXPORT_SYMBOL_GPL(raw_v6_hashinfo);
66
67 bool raw_v6_match(struct net *net, const struct sock *sk, unsigned short num,
68 const struct in6_addr *loc_addr,
69 const struct in6_addr *rmt_addr, int dif, int sdif)
70 {
71 if (inet_sk(sk)->inet_num != num ||
72 !net_eq(sock_net(sk), net) ||
73 (!ipv6_addr_any(&sk->sk_v6_daddr) &&
74 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
75 !raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if,
76 dif, sdif))
77 return false;
78
79 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr) ||
80 ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr) ||
81 (ipv6_addr_is_multicast(loc_addr) &&
82 inet6_mc_check(sk, loc_addr, rmt_addr)))
83 return true;
84
85 return false;
86 }
87 EXPORT_SYMBOL_GPL(raw_v6_match);
88
89 /*
90 * 0 - deliver
91 * 1 - block
92 */
93 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
94 {
95 struct icmp6hdr _hdr;
96 const struct icmp6hdr *hdr;
97
98 /* We require only the four bytes of the ICMPv6 header, not any
99 * additional bytes of message body in "struct icmp6hdr".
100 */
101 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
102 ICMPV6_HDRLEN, &_hdr);
103 if (hdr) {
104 const __u32 *data = &raw6_sk(sk)->filter.data[0];
105 unsigned int type = hdr->icmp6_type;
106
107 return (data[type >> 5] & (1U << (type & 31))) != 0;
108 }
109 return 1;
110 }
111
112 #if IS_ENABLED(CONFIG_IPV6_MIP6)
113 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
114
115 static mh_filter_t __rcu *mh_filter __read_mostly;
116
117 int rawv6_mh_filter_register(mh_filter_t filter)
118 {
119 rcu_assign_pointer(mh_filter, filter);
120 return 0;
121 }
122 EXPORT_SYMBOL(rawv6_mh_filter_register);
123
124 int rawv6_mh_filter_unregister(mh_filter_t filter)
125 {
126 RCU_INIT_POINTER(mh_filter, NULL);
127 synchronize_rcu();
128 return 0;
129 }
130 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
131
132 #endif
133
134 /*
135 * demultiplex raw sockets.
136 * (should consider queueing the skb in the sock receive_queue
137 * without calling rawv6.c)
138 *
139 * Caller owns SKB so we must make clones.
140 */
141 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
142 {
143 struct net *net = dev_net(skb->dev);
144 const struct in6_addr *saddr;
145 const struct in6_addr *daddr;
146 struct hlist_head *hlist;
147 struct sock *sk;
148 bool delivered = false;
149 __u8 hash;
150
151 saddr = &ipv6_hdr(skb)->saddr;
152 daddr = saddr + 1;
153
154 hash = raw_hashfunc(net, nexthdr);
155 hlist = &raw_v6_hashinfo.ht[hash];
156 rcu_read_lock();
157 sk_for_each_rcu(sk, hlist) {
158 int filtered;
159
160 if (!raw_v6_match(net, sk, nexthdr, daddr, saddr,
161 inet6_iif(skb), inet6_sdif(skb)))
162 continue;
163
164 if (atomic_read(&sk->sk_rmem_alloc) >=
165 READ_ONCE(sk->sk_rcvbuf)) {
166 atomic_inc(&sk->sk_drops);
167 continue;
168 }
169
170 delivered = true;
171 switch (nexthdr) {
172 case IPPROTO_ICMPV6:
173 filtered = icmpv6_filter(sk, skb);
174 break;
175
176 #if IS_ENABLED(CONFIG_IPV6_MIP6)
177 case IPPROTO_MH:
178 {
179 /* XXX: To validate MH only once for each packet,
180 * this is placed here. It should be after checking
181 * xfrm policy, however it doesn't. The checking xfrm
182 * policy is placed in rawv6_rcv() because it is
183 * required for each socket.
184 */
185 mh_filter_t *filter;
186
187 filter = rcu_dereference(mh_filter);
188 filtered = filter ? (*filter)(sk, skb) : 0;
189 break;
190 }
191 #endif
192 default:
193 filtered = 0;
194 break;
195 }
196
197 if (filtered < 0)
198 break;
199 if (filtered == 0) {
200 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
201
202 /* Not releasing hash table! */
203 if (clone)
204 rawv6_rcv(sk, clone);
205 }
206 }
207 rcu_read_unlock();
208 return delivered;
209 }
210
211 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
212 {
213 return ipv6_raw_deliver(skb, nexthdr);
214 }
215
216 /* This cleans up af_inet6 a bit. -DaveM */
217 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
218 {
219 struct inet_sock *inet = inet_sk(sk);
220 struct ipv6_pinfo *np = inet6_sk(sk);
221 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
222 __be32 v4addr = 0;
223 int addr_type;
224 int err;
225
226 if (addr_len < SIN6_LEN_RFC2133)
227 return -EINVAL;
228
229 if (addr->sin6_family != AF_INET6)
230 return -EINVAL;
231
232 addr_type = ipv6_addr_type(&addr->sin6_addr);
233
234 /* Raw sockets are IPv6 only */
235 if (addr_type == IPV6_ADDR_MAPPED)
236 return -EADDRNOTAVAIL;
237
238 lock_sock(sk);
239
240 err = -EINVAL;
241 if (sk->sk_state != TCP_CLOSE)
242 goto out;
243
244 rcu_read_lock();
245 /* Check if the address belongs to the host. */
246 if (addr_type != IPV6_ADDR_ANY) {
247 struct net_device *dev = NULL;
248
249 if (__ipv6_addr_needs_scope_id(addr_type)) {
250 if (addr_len >= sizeof(struct sockaddr_in6) &&
251 addr->sin6_scope_id) {
252 /* Override any existing binding, if another
253 * one is supplied by user.
254 */
255 sk->sk_bound_dev_if = addr->sin6_scope_id;
256 }
257
258 /* Binding to link-local address requires an interface */
259 if (!sk->sk_bound_dev_if)
260 goto out_unlock;
261 }
262
263 if (sk->sk_bound_dev_if) {
264 err = -ENODEV;
265 dev = dev_get_by_index_rcu(sock_net(sk),
266 sk->sk_bound_dev_if);
267 if (!dev)
268 goto out_unlock;
269 }
270
271 /* ipv4 addr of the socket is invalid. Only the
272 * unspecified and mapped address have a v4 equivalent.
273 */
274 v4addr = LOOPBACK4_IPV6;
275 if (!(addr_type & IPV6_ADDR_MULTICAST) &&
276 !ipv6_can_nonlocal_bind(sock_net(sk), inet)) {
277 err = -EADDRNOTAVAIL;
278 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
279 dev, 0)) {
280 goto out_unlock;
281 }
282 }
283 }
284
285 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
286 sk->sk_v6_rcv_saddr = addr->sin6_addr;
287 if (!(addr_type & IPV6_ADDR_MULTICAST))
288 np->saddr = addr->sin6_addr;
289 err = 0;
290 out_unlock:
291 rcu_read_unlock();
292 out:
293 release_sock(sk);
294 return err;
295 }
296
297 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
298 u8 type, u8 code, int offset, __be32 info)
299 {
300 bool recverr = inet6_test_bit(RECVERR6, sk);
301 struct ipv6_pinfo *np = inet6_sk(sk);
302 int err;
303 int harderr;
304
305 /* Report error on raw socket, if:
306 1. User requested recverr.
307 2. Socket is connected (otherwise the error indication
308 is useless without recverr and error is hard.
309 */
310 if (!recverr && sk->sk_state != TCP_ESTABLISHED)
311 return;
312
313 harderr = icmpv6_err_convert(type, code, &err);
314 if (type == ICMPV6_PKT_TOOBIG) {
315 ip6_sk_update_pmtu(skb, sk, info);
316 harderr = (READ_ONCE(np->pmtudisc) == IPV6_PMTUDISC_DO);
317 }
318 if (type == NDISC_REDIRECT) {
319 ip6_sk_redirect(skb, sk);
320 return;
321 }
322 if (recverr) {
323 u8 *payload = skb->data;
324 if (!inet_test_bit(HDRINCL, sk))
325 payload += offset;
326 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
327 }
328
329 if (recverr || harderr) {
330 sk->sk_err = err;
331 sk_error_report(sk);
332 }
333 }
334
335 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
336 u8 type, u8 code, int inner_offset, __be32 info)
337 {
338 struct net *net = dev_net(skb->dev);
339 struct hlist_head *hlist;
340 struct sock *sk;
341 int hash;
342
343 hash = raw_hashfunc(net, nexthdr);
344 hlist = &raw_v6_hashinfo.ht[hash];
345 rcu_read_lock();
346 sk_for_each_rcu(sk, hlist) {
347 /* Note: ipv6_hdr(skb) != skb->data */
348 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
349
350 if (!raw_v6_match(net, sk, nexthdr, &ip6h->saddr, &ip6h->daddr,
351 inet6_iif(skb), inet6_iif(skb)))
352 continue;
353 rawv6_err(sk, skb, type, code, inner_offset, info);
354 }
355 rcu_read_unlock();
356 }
357
358 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
359 {
360 enum skb_drop_reason reason;
361
362 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
363 skb_checksum_complete(skb)) {
364 atomic_inc(&sk->sk_drops);
365 sk_skb_reason_drop(sk, skb, SKB_DROP_REASON_SKB_CSUM);
366 return NET_RX_DROP;
367 }
368
369 /* Charge it to the socket. */
370 skb_dst_drop(skb);
371 if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) {
372 sk_skb_reason_drop(sk, skb, reason);
373 return NET_RX_DROP;
374 }
375
376 return 0;
377 }
378
379 /*
380 * This is next to useless...
381 * if we demultiplex in network layer we don't need the extra call
382 * just to queue the skb...
383 * maybe we could have the network decide upon a hint if it
384 * should call raw_rcv for demultiplexing
385 */
386 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
387 {
388 struct inet_sock *inet = inet_sk(sk);
389 struct raw6_sock *rp = raw6_sk(sk);
390
391 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
392 atomic_inc(&sk->sk_drops);
393 sk_skb_reason_drop(sk, skb, SKB_DROP_REASON_XFRM_POLICY);
394 return NET_RX_DROP;
395 }
396 nf_reset_ct(skb);
397
398 if (!rp->checksum)
399 skb->ip_summed = CHECKSUM_UNNECESSARY;
400
401 if (skb->ip_summed == CHECKSUM_COMPLETE) {
402 skb_postpull_rcsum(skb, skb_network_header(skb),
403 skb_network_header_len(skb));
404 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
405 &ipv6_hdr(skb)->daddr,
406 skb->len, inet->inet_num, skb->csum))
407 skb->ip_summed = CHECKSUM_UNNECESSARY;
408 }
409 if (!skb_csum_unnecessary(skb))
410 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
411 &ipv6_hdr(skb)->daddr,
412 skb->len,
413 inet->inet_num, 0));
414
415 if (inet_test_bit(HDRINCL, sk)) {
416 if (skb_checksum_complete(skb)) {
417 atomic_inc(&sk->sk_drops);
418 sk_skb_reason_drop(sk, skb, SKB_DROP_REASON_SKB_CSUM);
419 return NET_RX_DROP;
420 }
421 }
422
423 rawv6_rcv_skb(sk, skb);
424 return 0;
425 }
426
427
428 /*
429 * This should be easy, if there is something there
430 * we return it, otherwise we block.
431 */
432
433 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
434 int flags, int *addr_len)
435 {
436 struct ipv6_pinfo *np = inet6_sk(sk);
437 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
438 struct sk_buff *skb;
439 size_t copied;
440 int err;
441
442 if (flags & MSG_OOB)
443 return -EOPNOTSUPP;
444
445 if (flags & MSG_ERRQUEUE)
446 return ipv6_recv_error(sk, msg, len, addr_len);
447
448 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
449 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
450
451 skb = skb_recv_datagram(sk, flags, &err);
452 if (!skb)
453 goto out;
454 if (ccs_socket_post_recvmsg_permission(sk, skb, flags)) {
455 err = -EAGAIN; /* Hope less harmful than -EPERM. */
456 goto out;
457 }
458
459 copied = skb->len;
460 if (copied > len) {
461 copied = len;
462 msg->msg_flags |= MSG_TRUNC;
463 }
464
465 if (skb_csum_unnecessary(skb)) {
466 err = skb_copy_datagram_msg(skb, 0, msg, copied);
467 } else if (msg->msg_flags&MSG_TRUNC) {
468 if (__skb_checksum_complete(skb))
469 goto csum_copy_err;
470 err = skb_copy_datagram_msg(skb, 0, msg, copied);
471 } else {
472 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
473 if (err == -EINVAL)
474 goto csum_copy_err;
475 }
476 if (err)
477 goto out_free;
478
479 /* Copy the address. */
480 if (sin6) {
481 sin6->sin6_family = AF_INET6;
482 sin6->sin6_port = 0;
483 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
484 sin6->sin6_flowinfo = 0;
485 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
486 inet6_iif(skb));
487 *addr_len = sizeof(*sin6);
488 }
489
490 sock_recv_cmsgs(msg, sk, skb);
491
492 if (np->rxopt.all)
493 ip6_datagram_recv_ctl(sk, msg, skb);
494
495 err = copied;
496 if (flags & MSG_TRUNC)
497 err = skb->len;
498
499 out_free:
500 skb_free_datagram(sk, skb);
501 out:
502 return err;
503
504 csum_copy_err:
505 skb_kill_datagram(sk, skb, flags);
506
507 /* Error for blocking case is chosen to masquerade
508 as some normal condition.
509 */
510 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
511 goto out;
512 }
513
514 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
515 struct raw6_sock *rp)
516 {
517 struct ipv6_txoptions *opt;
518 struct sk_buff *skb;
519 int err = 0;
520 int offset;
521 int len;
522 int total_len;
523 __wsum tmp_csum;
524 __sum16 csum;
525
526 if (!rp->checksum)
527 goto send;
528
529 skb = skb_peek(&sk->sk_write_queue);
530 if (!skb)
531 goto out;
532
533 offset = rp->offset;
534 total_len = inet_sk(sk)->cork.base.length;
535 opt = inet6_sk(sk)->cork.opt;
536 total_len -= opt ? opt->opt_flen : 0;
537
538 if (offset >= total_len - 1) {
539 err = -EINVAL;
540 ip6_flush_pending_frames(sk);
541 goto out;
542 }
543
544 /* should be check HW csum miyazawa */
545 if (skb_queue_len(&sk->sk_write_queue) == 1) {
546 /*
547 * Only one fragment on the socket.
548 */
549 tmp_csum = skb->csum;
550 } else {
551 struct sk_buff *csum_skb = NULL;
552 tmp_csum = 0;
553
554 skb_queue_walk(&sk->sk_write_queue, skb) {
555 tmp_csum = csum_add(tmp_csum, skb->csum);
556
557 if (csum_skb)
558 continue;
559
560 len = skb->len - skb_transport_offset(skb);
561 if (offset >= len) {
562 offset -= len;
563 continue;
564 }
565
566 csum_skb = skb;
567 }
568
569 skb = csum_skb;
570 }
571
572 offset += skb_transport_offset(skb);
573 err = skb_copy_bits(skb, offset, &csum, 2);
574 if (err < 0) {
575 ip6_flush_pending_frames(sk);
576 goto out;
577 }
578
579 /* in case cksum was not initialized */
580 if (unlikely(csum))
581 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
582
583 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
584 total_len, fl6->flowi6_proto, tmp_csum);
585
586 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
587 csum = CSUM_MANGLED_0;
588
589 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
590
591 send:
592 err = ip6_push_pending_frames(sk);
593 out:
594 return err;
595 }
596
597 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
598 struct flowi6 *fl6, struct dst_entry **dstp,
599 unsigned int flags, const struct sockcm_cookie *sockc)
600 {
601 struct net *net = sock_net(sk);
602 struct ipv6hdr *iph;
603 struct sk_buff *skb;
604 int err;
605 struct rt6_info *rt = dst_rt6_info(*dstp);
606 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
607 int tlen = rt->dst.dev->needed_tailroom;
608
609 if (length > rt->dst.dev->mtu) {
610 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
611 return -EMSGSIZE;
612 }
613 if (length < sizeof(struct ipv6hdr))
614 return -EINVAL;
615 if (flags&MSG_PROBE)
616 goto out;
617
618 skb = sock_alloc_send_skb(sk,
619 length + hlen + tlen + 15,
620 flags & MSG_DONTWAIT, &err);
621 if (!skb)
622 goto error;
623 skb_reserve(skb, hlen);
624
625 skb->protocol = htons(ETH_P_IPV6);
626 skb->priority = READ_ONCE(sk->sk_priority);
627 skb->mark = sockc->mark;
628 skb_set_delivery_type_by_clockid(skb, sockc->transmit_time, sk->sk_clockid);
629
630 skb_put(skb, length);
631 skb_reset_network_header(skb);
632 iph = ipv6_hdr(skb);
633
634 skb->ip_summed = CHECKSUM_NONE;
635
636 skb_setup_tx_timestamp(skb, sockc->tsflags);
637
638 if (flags & MSG_CONFIRM)
639 skb_set_dst_pending_confirm(skb, 1);
640
641 skb->transport_header = skb->network_header;
642 err = memcpy_from_msg(iph, msg, length);
643 if (err) {
644 err = -EFAULT;
645 kfree_skb(skb);
646 goto error;
647 }
648
649 skb_dst_set(skb, &rt->dst);
650 *dstp = NULL;
651
652 /* if egress device is enslaved to an L3 master device pass the
653 * skb to its handler for processing
654 */
655 skb = l3mdev_ip6_out(sk, skb);
656 if (unlikely(!skb))
657 return 0;
658
659 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev
660 * in the error path. Since skb has been freed, the dst could
661 * have been queued for deletion.
662 */
663 rcu_read_lock();
664 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
665 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
666 NULL, rt->dst.dev, dst_output);
667 if (err > 0)
668 err = net_xmit_errno(err);
669 if (err) {
670 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
671 rcu_read_unlock();
672 goto error_check;
673 }
674 rcu_read_unlock();
675 out:
676 return 0;
677
678 error:
679 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
680 error_check:
681 if (err == -ENOBUFS && !inet6_test_bit(RECVERR6, sk))
682 err = 0;
683 return err;
684 }
685
686 struct raw6_frag_vec {
687 struct msghdr *msg;
688 int hlen;
689 char c[4];
690 };
691
692 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
693 {
694 int err = 0;
695 switch (fl6->flowi6_proto) {
696 case IPPROTO_ICMPV6:
697 rfv->hlen = 2;
698 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
699 if (!err) {
700 fl6->fl6_icmp_type = rfv->c[0];
701 fl6->fl6_icmp_code = rfv->c[1];
702 }
703 break;
704 case IPPROTO_MH:
705 rfv->hlen = 4;
706 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
707 if (!err)
708 fl6->fl6_mh_type = rfv->c[2];
709 }
710 return err;
711 }
712
713 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
714 struct sk_buff *skb)
715 {
716 struct raw6_frag_vec *rfv = from;
717
718 if (offset < rfv->hlen) {
719 int copy = min(rfv->hlen - offset, len);
720
721 if (skb->ip_summed == CHECKSUM_PARTIAL)
722 memcpy(to, rfv->c + offset, copy);
723 else
724 skb->csum = csum_block_add(
725 skb->csum,
726 csum_partial_copy_nocheck(rfv->c + offset,
727 to, copy),
728 odd);
729
730 odd = 0;
731 offset += copy;
732 to += copy;
733 len -= copy;
734
735 if (!len)
736 return 0;
737 }
738
739 offset -= rfv->hlen;
740
741 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
742 }
743
744 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
745 {
746 struct ipv6_txoptions *opt_to_free = NULL;
747 struct ipv6_txoptions opt_space;
748 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
749 struct in6_addr *daddr, *final_p, final;
750 struct inet_sock *inet = inet_sk(sk);
751 struct ipv6_pinfo *np = inet6_sk(sk);
752 struct raw6_sock *rp = raw6_sk(sk);
753 struct ipv6_txoptions *opt = NULL;
754 struct ip6_flowlabel *flowlabel = NULL;
755 struct dst_entry *dst = NULL;
756 struct raw6_frag_vec rfv;
757 struct flowi6 fl6;
758 struct ipcm6_cookie ipc6;
759 int addr_len = msg->msg_namelen;
760 int hdrincl;
761 u16 proto;
762 int err;
763
764 /* Rough check on arithmetic overflow,
765 better check is made in ip6_append_data().
766 */
767 if (len > INT_MAX)
768 return -EMSGSIZE;
769
770 /* Mirror BSD error message compatibility */
771 if (msg->msg_flags & MSG_OOB)
772 return -EOPNOTSUPP;
773
774 hdrincl = inet_test_bit(HDRINCL, sk);
775
776 /*
777 * Get and verify the address.
778 */
779 memset(&fl6, 0, sizeof(fl6));
780
781 fl6.flowi6_mark = READ_ONCE(sk->sk_mark);
782 fl6.flowi6_uid = sk->sk_uid;
783
784 ipcm6_init(&ipc6);
785 ipc6.sockc.tsflags = READ_ONCE(sk->sk_tsflags);
786 ipc6.sockc.mark = fl6.flowi6_mark;
787
788 if (sin6) {
789 if (addr_len < SIN6_LEN_RFC2133)
790 return -EINVAL;
791
792 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
793 return -EAFNOSUPPORT;
794
795 /* port is the proto value [0..255] carried in nexthdr */
796 proto = ntohs(sin6->sin6_port);
797
798 if (!proto)
799 proto = inet->inet_num;
800 else if (proto != inet->inet_num &&
801 inet->inet_num != IPPROTO_RAW)
802 return -EINVAL;
803
804 if (proto > 255)
805 return -EINVAL;
806
807 daddr = &sin6->sin6_addr;
808 if (inet6_test_bit(SNDFLOW, sk)) {
809 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
810 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
811 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
812 if (IS_ERR(flowlabel))
813 return -EINVAL;
814 }
815 }
816
817 /*
818 * Otherwise it will be difficult to maintain
819 * sk->sk_dst_cache.
820 */
821 if (sk->sk_state == TCP_ESTABLISHED &&
822 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
823 daddr = &sk->sk_v6_daddr;
824
825 if (addr_len >= sizeof(struct sockaddr_in6) &&
826 sin6->sin6_scope_id &&
827 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
828 fl6.flowi6_oif = sin6->sin6_scope_id;
829 } else {
830 if (sk->sk_state != TCP_ESTABLISHED)
831 return -EDESTADDRREQ;
832
833 proto = inet->inet_num;
834 daddr = &sk->sk_v6_daddr;
835 fl6.flowlabel = np->flow_label;
836 }
837
838 if (fl6.flowi6_oif == 0)
839 fl6.flowi6_oif = sk->sk_bound_dev_if;
840
841 if (msg->msg_controllen) {
842 opt = &opt_space;
843 memset(opt, 0, sizeof(struct ipv6_txoptions));
844 opt->tot_len = sizeof(struct ipv6_txoptions);
845 ipc6.opt = opt;
846
847 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
848 if (err < 0) {
849 fl6_sock_release(flowlabel);
850 return err;
851 }
852 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
853 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
854 if (IS_ERR(flowlabel))
855 return -EINVAL;
856 }
857 if (!(opt->opt_nflen|opt->opt_flen))
858 opt = NULL;
859 }
860 if (!opt) {
861 opt = txopt_get(np);
862 opt_to_free = opt;
863 }
864 if (flowlabel)
865 opt = fl6_merge_options(&opt_space, flowlabel, opt);
866 opt = ipv6_fixup_options(&opt_space, opt);
867
868 fl6.flowi6_proto = proto;
869 fl6.flowi6_mark = ipc6.sockc.mark;
870
871 if (!hdrincl) {
872 rfv.msg = msg;
873 rfv.hlen = 0;
874 err = rawv6_probe_proto_opt(&rfv, &fl6);
875 if (err)
876 goto out;
877 }
878
879 if (!ipv6_addr_any(daddr))
880 fl6.daddr = *daddr;
881 else
882 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
883 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
884 fl6.saddr = np->saddr;
885
886 final_p = fl6_update_dst(&fl6, opt, &final);
887
888 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
889 fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
890 else if (!fl6.flowi6_oif)
891 fl6.flowi6_oif = READ_ONCE(np->ucast_oif);
892 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
893
894 if (hdrincl)
895 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
896
897 if (ipc6.tclass < 0)
898 ipc6.tclass = np->tclass;
899
900 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
901
902 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
903 if (IS_ERR(dst)) {
904 err = PTR_ERR(dst);
905 goto out;
906 }
907 if (ipc6.hlimit < 0)
908 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
909
910 if (ipc6.dontfrag < 0)
911 ipc6.dontfrag = inet6_test_bit(DONTFRAG, sk);
912
913 if (msg->msg_flags&MSG_CONFIRM)
914 goto do_confirm;
915
916 back_from_confirm:
917 if (hdrincl)
918 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst,
919 msg->msg_flags, &ipc6.sockc);
920 else {
921 ipc6.opt = opt;
922 lock_sock(sk);
923 err = ip6_append_data(sk, raw6_getfrag, &rfv,
924 len, 0, &ipc6, &fl6, dst_rt6_info(dst),
925 msg->msg_flags);
926
927 if (err)
928 ip6_flush_pending_frames(sk);
929 else if (!(msg->msg_flags & MSG_MORE))
930 err = rawv6_push_pending_frames(sk, &fl6, rp);
931 release_sock(sk);
932 }
933 done:
934 dst_release(dst);
935 out:
936 fl6_sock_release(flowlabel);
937 txopt_put(opt_to_free);
938 return err < 0 ? err : len;
939 do_confirm:
940 if (msg->msg_flags & MSG_PROBE)
941 dst_confirm_neigh(dst, &fl6.daddr);
942 if (!(msg->msg_flags & MSG_PROBE) || len)
943 goto back_from_confirm;
944 err = 0;
945 goto done;
946 }
947
948 static int rawv6_seticmpfilter(struct sock *sk, int optname,
949 sockptr_t optval, int optlen)
950 {
951 switch (optname) {
952 case ICMPV6_FILTER:
953 if (optlen > sizeof(struct icmp6_filter))
954 optlen = sizeof(struct icmp6_filter);
955 if (copy_from_sockptr(&raw6_sk(sk)->filter, optval, optlen))
956 return -EFAULT;
957 return 0;
958 default:
959 return -ENOPROTOOPT;
960 }
961
962 return 0;
963 }
964
965 static int rawv6_geticmpfilter(struct sock *sk, int optname,
966 char __user *optval, int __user *optlen)
967 {
968 int len;
969
970 switch (optname) {
971 case ICMPV6_FILTER:
972 if (get_user(len, optlen))
973 return -EFAULT;
974 if (len < 0)
975 return -EINVAL;
976 if (len > sizeof(struct icmp6_filter))
977 len = sizeof(struct icmp6_filter);
978 if (put_user(len, optlen))
979 return -EFAULT;
980 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
981 return -EFAULT;
982 return 0;
983 default:
984 return -ENOPROTOOPT;
985 }
986
987 return 0;
988 }
989
990
991 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
992 sockptr_t optval, unsigned int optlen)
993 {
994 struct raw6_sock *rp = raw6_sk(sk);
995 int val;
996
997 if (optlen < sizeof(val))
998 return -EINVAL;
999
1000 if (copy_from_sockptr(&val, optval, sizeof(val)))
1001 return -EFAULT;
1002
1003 switch (optname) {
1004 case IPV6_HDRINCL:
1005 if (sk->sk_type != SOCK_RAW)
1006 return -EINVAL;
1007 inet_assign_bit(HDRINCL, sk, val);
1008 return 0;
1009 case IPV6_CHECKSUM:
1010 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1011 level == IPPROTO_IPV6) {
1012 /*
1013 * RFC3542 tells that IPV6_CHECKSUM socket
1014 * option in the IPPROTO_IPV6 level is not
1015 * allowed on ICMPv6 sockets.
1016 * If you want to set it, use IPPROTO_RAW
1017 * level IPV6_CHECKSUM socket option
1018 * (Linux extension).
1019 */
1020 return -EINVAL;
1021 }
1022
1023 /* You may get strange result with a positive odd offset;
1024 RFC2292bis agrees with me. */
1025 if (val > 0 && (val&1))
1026 return -EINVAL;
1027 if (val < 0) {
1028 rp->checksum = 0;
1029 } else {
1030 rp->checksum = 1;
1031 rp->offset = val;
1032 }
1033
1034 return 0;
1035
1036 default:
1037 return -ENOPROTOOPT;
1038 }
1039 }
1040
1041 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1042 sockptr_t optval, unsigned int optlen)
1043 {
1044 switch (level) {
1045 case SOL_RAW:
1046 break;
1047
1048 case SOL_ICMPV6:
1049 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1050 return -EOPNOTSUPP;
1051 return rawv6_seticmpfilter(sk, optname, optval, optlen);
1052 case SOL_IPV6:
1053 if (optname == IPV6_CHECKSUM ||
1054 optname == IPV6_HDRINCL)
1055 break;
1056 fallthrough;
1057 default:
1058 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1059 }
1060
1061 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1062 }
1063
1064 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1065 char __user *optval, int __user *optlen)
1066 {
1067 struct raw6_sock *rp = raw6_sk(sk);
1068 int val, len;
1069
1070 if (get_user(len, optlen))
1071 return -EFAULT;
1072
1073 switch (optname) {
1074 case IPV6_HDRINCL:
1075 val = inet_test_bit(HDRINCL, sk);
1076 break;
1077 case IPV6_CHECKSUM:
1078 /*
1079 * We allow getsockopt() for IPPROTO_IPV6-level
1080 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1081 * since RFC3542 is silent about it.
1082 */
1083 if (rp->checksum == 0)
1084 val = -1;
1085 else
1086 val = rp->offset;
1087 break;
1088
1089 default:
1090 return -ENOPROTOOPT;
1091 }
1092
1093 len = min_t(unsigned int, sizeof(int), len);
1094
1095 if (put_user(len, optlen))
1096 return -EFAULT;
1097 if (copy_to_user(optval, &val, len))
1098 return -EFAULT;
1099 return 0;
1100 }
1101
1102 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1103 char __user *optval, int __user *optlen)
1104 {
1105 switch (level) {
1106 case SOL_RAW:
1107 break;
1108
1109 case SOL_ICMPV6:
1110 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1111 return -EOPNOTSUPP;
1112 return rawv6_geticmpfilter(sk, optname, optval, optlen);
1113 case SOL_IPV6:
1114 if (optname == IPV6_CHECKSUM ||
1115 optname == IPV6_HDRINCL)
1116 break;
1117 fallthrough;
1118 default:
1119 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1120 }
1121
1122 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1123 }
1124
1125 static int rawv6_ioctl(struct sock *sk, int cmd, int *karg)
1126 {
1127 switch (cmd) {
1128 case SIOCOUTQ: {
1129 *karg = sk_wmem_alloc_get(sk);
1130 return 0;
1131 }
1132 case SIOCINQ: {
1133 struct sk_buff *skb;
1134
1135 spin_lock_bh(&sk->sk_receive_queue.lock);
1136 skb = skb_peek(&sk->sk_receive_queue);
1137 if (skb)
1138 *karg = skb->len;
1139 else
1140 *karg = 0;
1141 spin_unlock_bh(&sk->sk_receive_queue.lock);
1142 return 0;
1143 }
1144
1145 default:
1146 #ifdef CONFIG_IPV6_MROUTE
1147 return ip6mr_ioctl(sk, cmd, karg);
1148 #else
1149 return -ENOIOCTLCMD;
1150 #endif
1151 }
1152 }
1153
1154 #ifdef CONFIG_COMPAT
1155 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1156 {
1157 switch (cmd) {
1158 case SIOCOUTQ:
1159 case SIOCINQ:
1160 return -ENOIOCTLCMD;
1161 default:
1162 #ifdef CONFIG_IPV6_MROUTE
1163 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1164 #else
1165 return -ENOIOCTLCMD;
1166 #endif
1167 }
1168 }
1169 #endif
1170
1171 static void rawv6_close(struct sock *sk, long timeout)
1172 {
1173 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1174 ip6_ra_control(sk, -1);
1175 ip6mr_sk_done(sk);
1176 sk_common_release(sk);
1177 }
1178
1179 static void raw6_destroy(struct sock *sk)
1180 {
1181 lock_sock(sk);
1182 ip6_flush_pending_frames(sk);
1183 release_sock(sk);
1184 }
1185
1186 static int rawv6_init_sk(struct sock *sk)
1187 {
1188 struct raw6_sock *rp = raw6_sk(sk);
1189
1190 switch (inet_sk(sk)->inet_num) {
1191 case IPPROTO_ICMPV6:
1192 rp->checksum = 1;
1193 rp->offset = 2;
1194 break;
1195 case IPPROTO_MH:
1196 rp->checksum = 1;
1197 rp->offset = 4;
1198 break;
1199 default:
1200 break;
1201 }
1202 return 0;
1203 }
1204
1205 struct proto rawv6_prot = {
1206 .name = "RAWv6",
1207 .owner = THIS_MODULE,
1208 .close = rawv6_close,
1209 .destroy = raw6_destroy,
1210 .connect = ip6_datagram_connect_v6_only,
1211 .disconnect = __udp_disconnect,
1212 .ioctl = rawv6_ioctl,
1213 .init = rawv6_init_sk,
1214 .setsockopt = rawv6_setsockopt,
1215 .getsockopt = rawv6_getsockopt,
1216 .sendmsg = rawv6_sendmsg,
1217 .recvmsg = rawv6_recvmsg,
1218 .bind = rawv6_bind,
1219 .backlog_rcv = rawv6_rcv_skb,
1220 .hash = raw_hash_sk,
1221 .unhash = raw_unhash_sk,
1222 .obj_size = sizeof(struct raw6_sock),
1223 .ipv6_pinfo_offset = offsetof(struct raw6_sock, inet6),
1224 .useroffset = offsetof(struct raw6_sock, filter),
1225 .usersize = sizeof_field(struct raw6_sock, filter),
1226 .h.raw_hash = &raw_v6_hashinfo,
1227 #ifdef CONFIG_COMPAT
1228 .compat_ioctl = compat_rawv6_ioctl,
1229 #endif
1230 .diag_destroy = raw_abort,
1231 };
1232
1233 #ifdef CONFIG_PROC_FS
1234 static int raw6_seq_show(struct seq_file *seq, void *v)
1235 {
1236 if (v == SEQ_START_TOKEN) {
1237 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1238 } else {
1239 struct sock *sp = v;
1240 __u16 srcp = inet_sk(sp)->inet_num;
1241 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1242 raw_seq_private(seq)->bucket);
1243 }
1244 return 0;
1245 }
1246
1247 static const struct seq_operations raw6_seq_ops = {
1248 .start = raw_seq_start,
1249 .next = raw_seq_next,
1250 .stop = raw_seq_stop,
1251 .show = raw6_seq_show,
1252 };
1253
1254 static int __net_init raw6_init_net(struct net *net)
1255 {
1256 if (!proc_create_net_data("raw6", 0444, net->proc_net, &raw6_seq_ops,
1257 sizeof(struct raw_iter_state), &raw_v6_hashinfo))
1258 return -ENOMEM;
1259
1260 return 0;
1261 }
1262
1263 static void __net_exit raw6_exit_net(struct net *net)
1264 {
1265 remove_proc_entry("raw6", net->proc_net);
1266 }
1267
1268 static struct pernet_operations raw6_net_ops = {
1269 .init = raw6_init_net,
1270 .exit = raw6_exit_net,
1271 };
1272
1273 int __init raw6_proc_init(void)
1274 {
1275 return register_pernet_subsys(&raw6_net_ops);
1276 }
1277
1278 void raw6_proc_exit(void)
1279 {
1280 unregister_pernet_subsys(&raw6_net_ops);
1281 }
1282 #endif /* CONFIG_PROC_FS */
1283
1284 /* Same as inet6_dgram_ops, sans udp_poll. */
1285 const struct proto_ops inet6_sockraw_ops = {
1286 .family = PF_INET6,
1287 .owner = THIS_MODULE,
1288 .release = inet6_release,
1289 .bind = inet6_bind,
1290 .connect = inet_dgram_connect, /* ok */
1291 .socketpair = sock_no_socketpair, /* a do nothing */
1292 .accept = sock_no_accept, /* a do nothing */
1293 .getname = inet6_getname,
1294 .poll = datagram_poll, /* ok */
1295 .ioctl = inet6_ioctl, /* must change */
1296 .gettstamp = sock_gettstamp,
1297 .listen = sock_no_listen, /* ok */
1298 .shutdown = inet_shutdown, /* ok */
1299 .setsockopt = sock_common_setsockopt, /* ok */
1300 .getsockopt = sock_common_getsockopt, /* ok */
1301 .sendmsg = inet_sendmsg, /* ok */
1302 .recvmsg = sock_common_recvmsg, /* ok */
1303 .mmap = sock_no_mmap,
1304 #ifdef CONFIG_COMPAT
1305 .compat_ioctl = inet6_compat_ioctl,
1306 #endif
1307 };
1308
1309 static struct inet_protosw rawv6_protosw = {
1310 .type = SOCK_RAW,
1311 .protocol = IPPROTO_IP, /* wild card */
1312 .prot = &rawv6_prot,
1313 .ops = &inet6_sockraw_ops,
1314 .flags = INET_PROTOSW_REUSE,
1315 };
1316
1317 int __init rawv6_init(void)
1318 {
1319 return inet6_register_protosw(&rawv6_protosw);
1320 }
1321
1322 void rawv6_exit(void)
1323 {
1324 inet6_unregister_protosw(&rawv6_protosw);
1325 }
1326
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.