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 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
66 };
67 EXPORT_SYMBOL_GPL(raw_v6_hashinfo);
68
69 struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
70 unsigned short num, const struct in6_addr *loc_addr,
71 const struct in6_addr *rmt_addr, int dif, int sdif)
72 {
73 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
74
75 sk_for_each_from(sk)
76 if (inet_sk(sk)->inet_num == num) {
77
78 if (!net_eq(sock_net(sk), net))
79 continue;
80
81 if (!ipv6_addr_any(&sk->sk_v6_daddr) &&
82 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr))
83 continue;
84
85 if (!raw_sk_bound_dev_eq(net, sk->sk_bound_dev_if,
86 dif, sdif))
87 continue;
88
89 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
90 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))
91 goto found;
92 if (is_multicast &&
93 inet6_mc_check(sk, loc_addr, rmt_addr))
94 goto found;
95 continue;
96 }
97 goto found;
98 }
99 sk = NULL;
100 found:
101 return sk;
102 }
103 EXPORT_SYMBOL_GPL(__raw_v6_lookup);
104
105 /*
106 * 0 - deliver
107 * 1 - block
108 */
109 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
110 {
111 struct icmp6hdr _hdr;
112 const struct icmp6hdr *hdr;
113
114 /* We require only the four bytes of the ICMPv6 header, not any
115 * additional bytes of message body in "struct icmp6hdr".
116 */
117 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
118 ICMPV6_HDRLEN, &_hdr);
119 if (hdr) {
120 const __u32 *data = &raw6_sk(sk)->filter.data[0];
121 unsigned int type = hdr->icmp6_type;
122
123 return (data[type >> 5] & (1U << (type & 31))) != 0;
124 }
125 return 1;
126 }
127
128 #if IS_ENABLED(CONFIG_IPV6_MIP6)
129 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
130
131 static mh_filter_t __rcu *mh_filter __read_mostly;
132
133 int rawv6_mh_filter_register(mh_filter_t filter)
134 {
135 rcu_assign_pointer(mh_filter, filter);
136 return 0;
137 }
138 EXPORT_SYMBOL(rawv6_mh_filter_register);
139
140 int rawv6_mh_filter_unregister(mh_filter_t filter)
141 {
142 RCU_INIT_POINTER(mh_filter, NULL);
143 synchronize_rcu();
144 return 0;
145 }
146 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
147
148 #endif
149
150 /*
151 * demultiplex raw sockets.
152 * (should consider queueing the skb in the sock receive_queue
153 * without calling rawv6.c)
154 *
155 * Caller owns SKB so we must make clones.
156 */
157 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
158 {
159 const struct in6_addr *saddr;
160 const struct in6_addr *daddr;
161 struct sock *sk;
162 bool delivered = false;
163 __u8 hash;
164 struct net *net;
165
166 saddr = &ipv6_hdr(skb)->saddr;
167 daddr = saddr + 1;
168
169 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
170
171 read_lock(&raw_v6_hashinfo.lock);
172 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
173
174 if (!sk)
175 goto out;
176
177 net = dev_net(skb->dev);
178 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr,
179 inet6_iif(skb), inet6_sdif(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_ct(clone);
219 rawv6_rcv(sk, clone);
220 }
221 }
222 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
223 inet6_iif(skb), inet6_sdif(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
288 if (sk->sk_bound_dev_if) {
289 err = -ENODEV;
290 dev = dev_get_by_index_rcu(sock_net(sk),
291 sk->sk_bound_dev_if);
292 if (!dev)
293 goto out_unlock;
294 }
295
296 /* ipv4 addr of the socket is invalid. Only the
297 * unspecified and mapped address have a v4 equivalent.
298 */
299 v4addr = LOOPBACK4_IPV6;
300 if (!(addr_type & IPV6_ADDR_MULTICAST) &&
301 !ipv6_can_nonlocal_bind(sock_net(sk), inet)) {
302 err = -EADDRNOTAVAIL;
303 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
304 dev, 0)) {
305 goto out_unlock;
306 }
307 }
308 }
309
310 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
311 sk->sk_v6_rcv_saddr = addr->sin6_addr;
312 if (!(addr_type & IPV6_ADDR_MULTICAST))
313 np->saddr = addr->sin6_addr;
314 err = 0;
315 out_unlock:
316 rcu_read_unlock();
317 out:
318 release_sock(sk);
319 return err;
320 }
321
322 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
323 struct inet6_skb_parm *opt,
324 u8 type, u8 code, int offset, __be32 info)
325 {
326 struct inet_sock *inet = inet_sk(sk);
327 struct ipv6_pinfo *np = inet6_sk(sk);
328 int err;
329 int harderr;
330
331 /* Report error on raw socket, if:
332 1. User requested recverr.
333 2. Socket is connected (otherwise the error indication
334 is useless without recverr and error is hard.
335 */
336 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
337 return;
338
339 harderr = icmpv6_err_convert(type, code, &err);
340 if (type == ICMPV6_PKT_TOOBIG) {
341 ip6_sk_update_pmtu(skb, sk, info);
342 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
343 }
344 if (type == NDISC_REDIRECT) {
345 ip6_sk_redirect(skb, sk);
346 return;
347 }
348 if (np->recverr) {
349 u8 *payload = skb->data;
350 if (!inet->hdrincl)
351 payload += offset;
352 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
353 }
354
355 if (np->recverr || harderr) {
356 sk->sk_err = err;
357 sk->sk_error_report(sk);
358 }
359 }
360
361 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
362 u8 type, u8 code, int inner_offset, __be32 info)
363 {
364 struct sock *sk;
365 int hash;
366 const struct in6_addr *saddr, *daddr;
367 struct net *net;
368
369 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
370
371 read_lock(&raw_v6_hashinfo.lock);
372 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
373 if (sk) {
374 /* Note: ipv6_hdr(skb) != skb->data */
375 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
376 saddr = &ip6h->saddr;
377 daddr = &ip6h->daddr;
378 net = dev_net(skb->dev);
379
380 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
381 inet6_iif(skb), inet6_iif(skb)))) {
382 rawv6_err(sk, skb, NULL, type, code,
383 inner_offset, info);
384 sk = sk_next(sk);
385 }
386 }
387 read_unlock(&raw_v6_hashinfo.lock);
388 }
389
390 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
391 {
392 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
393 skb_checksum_complete(skb)) {
394 atomic_inc(&sk->sk_drops);
395 kfree_skb(skb);
396 return NET_RX_DROP;
397 }
398
399 /* Charge it to the socket. */
400 skb_dst_drop(skb);
401 if (sock_queue_rcv_skb(sk, skb) < 0) {
402 kfree_skb(skb);
403 return NET_RX_DROP;
404 }
405
406 return 0;
407 }
408
409 /*
410 * This is next to useless...
411 * if we demultiplex in network layer we don't need the extra call
412 * just to queue the skb...
413 * maybe we could have the network decide upon a hint if it
414 * should call raw_rcv for demultiplexing
415 */
416 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
417 {
418 struct inet_sock *inet = inet_sk(sk);
419 struct raw6_sock *rp = raw6_sk(sk);
420
421 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
422 atomic_inc(&sk->sk_drops);
423 kfree_skb(skb);
424 return NET_RX_DROP;
425 }
426
427 if (!rp->checksum)
428 skb->ip_summed = CHECKSUM_UNNECESSARY;
429
430 if (skb->ip_summed == CHECKSUM_COMPLETE) {
431 skb_postpull_rcsum(skb, skb_network_header(skb),
432 skb_network_header_len(skb));
433 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
434 &ipv6_hdr(skb)->daddr,
435 skb->len, inet->inet_num, skb->csum))
436 skb->ip_summed = CHECKSUM_UNNECESSARY;
437 }
438 if (!skb_csum_unnecessary(skb))
439 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
440 &ipv6_hdr(skb)->daddr,
441 skb->len,
442 inet->inet_num, 0));
443
444 if (inet->hdrincl) {
445 if (skb_checksum_complete(skb)) {
446 atomic_inc(&sk->sk_drops);
447 kfree_skb(skb);
448 return NET_RX_DROP;
449 }
450 }
451
452 rawv6_rcv_skb(sk, skb);
453 return 0;
454 }
455
456
457 /*
458 * This should be easy, if there is something there
459 * we return it, otherwise we block.
460 */
461
462 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
463 int noblock, int flags, int *addr_len)
464 {
465 struct ipv6_pinfo *np = inet6_sk(sk);
466 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
467 struct sk_buff *skb;
468 size_t copied;
469 int err;
470
471 if (flags & MSG_OOB)
472 return -EOPNOTSUPP;
473
474 if (flags & MSG_ERRQUEUE)
475 return ipv6_recv_error(sk, msg, len, addr_len);
476
477 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
478 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
479
480 skb = skb_recv_datagram(sk, flags, noblock, &err);
481 if (!skb)
482 goto out;
483 if (ccs_socket_post_recvmsg_permission(sk, skb, flags)) {
484 err = -EAGAIN; /* Hope less harmful than -EPERM. */
485 goto out;
486 }
487
488 copied = skb->len;
489 if (copied > len) {
490 copied = len;
491 msg->msg_flags |= MSG_TRUNC;
492 }
493
494 if (skb_csum_unnecessary(skb)) {
495 err = skb_copy_datagram_msg(skb, 0, msg, copied);
496 } else if (msg->msg_flags&MSG_TRUNC) {
497 if (__skb_checksum_complete(skb))
498 goto csum_copy_err;
499 err = skb_copy_datagram_msg(skb, 0, msg, copied);
500 } else {
501 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
502 if (err == -EINVAL)
503 goto csum_copy_err;
504 }
505 if (err)
506 goto out_free;
507
508 /* Copy the address. */
509 if (sin6) {
510 sin6->sin6_family = AF_INET6;
511 sin6->sin6_port = 0;
512 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
513 sin6->sin6_flowinfo = 0;
514 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
515 inet6_iif(skb));
516 *addr_len = sizeof(*sin6);
517 }
518
519 sock_recv_ts_and_drops(msg, sk, skb);
520
521 if (np->rxopt.all)
522 ip6_datagram_recv_ctl(sk, msg, skb);
523
524 err = copied;
525 if (flags & MSG_TRUNC)
526 err = skb->len;
527
528 out_free:
529 skb_free_datagram(sk, skb);
530 out:
531 return err;
532
533 csum_copy_err:
534 skb_kill_datagram(sk, skb, flags);
535
536 /* Error for blocking case is chosen to masquerade
537 as some normal condition.
538 */
539 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
540 goto out;
541 }
542
543 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
544 struct raw6_sock *rp)
545 {
546 struct sk_buff *skb;
547 int err = 0;
548 int offset;
549 int len;
550 int total_len;
551 __wsum tmp_csum;
552 __sum16 csum;
553
554 if (!rp->checksum)
555 goto send;
556
557 skb = skb_peek(&sk->sk_write_queue);
558 if (!skb)
559 goto out;
560
561 offset = rp->offset;
562 total_len = inet_sk(sk)->cork.base.length;
563 if (offset >= total_len - 1) {
564 err = -EINVAL;
565 ip6_flush_pending_frames(sk);
566 goto out;
567 }
568
569 /* should be check HW csum miyazawa */
570 if (skb_queue_len(&sk->sk_write_queue) == 1) {
571 /*
572 * Only one fragment on the socket.
573 */
574 tmp_csum = skb->csum;
575 } else {
576 struct sk_buff *csum_skb = NULL;
577 tmp_csum = 0;
578
579 skb_queue_walk(&sk->sk_write_queue, skb) {
580 tmp_csum = csum_add(tmp_csum, skb->csum);
581
582 if (csum_skb)
583 continue;
584
585 len = skb->len - skb_transport_offset(skb);
586 if (offset >= len) {
587 offset -= len;
588 continue;
589 }
590
591 csum_skb = skb;
592 }
593
594 skb = csum_skb;
595 }
596
597 offset += skb_transport_offset(skb);
598 err = skb_copy_bits(skb, offset, &csum, 2);
599 if (err < 0) {
600 ip6_flush_pending_frames(sk);
601 goto out;
602 }
603
604 /* in case cksum was not initialized */
605 if (unlikely(csum))
606 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
607
608 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
609 total_len, fl6->flowi6_proto, tmp_csum);
610
611 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
612 csum = CSUM_MANGLED_0;
613
614 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
615
616 send:
617 err = ip6_push_pending_frames(sk);
618 out:
619 return err;
620 }
621
622 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
623 struct flowi6 *fl6, struct dst_entry **dstp,
624 unsigned int flags, const struct sockcm_cookie *sockc)
625 {
626 struct ipv6_pinfo *np = inet6_sk(sk);
627 struct net *net = sock_net(sk);
628 struct ipv6hdr *iph;
629 struct sk_buff *skb;
630 int err;
631 struct rt6_info *rt = (struct rt6_info *)*dstp;
632 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
633 int tlen = rt->dst.dev->needed_tailroom;
634
635 if (length > rt->dst.dev->mtu) {
636 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
637 return -EMSGSIZE;
638 }
639 if (length < sizeof(struct ipv6hdr))
640 return -EINVAL;
641 if (flags&MSG_PROBE)
642 goto out;
643
644 skb = sock_alloc_send_skb(sk,
645 length + hlen + tlen + 15,
646 flags & MSG_DONTWAIT, &err);
647 if (!skb)
648 goto error;
649 skb_reserve(skb, hlen);
650
651 skb->protocol = htons(ETH_P_IPV6);
652 skb->priority = sk->sk_priority;
653 skb->mark = sockc->mark;
654 skb->tstamp = sockc->transmit_time;
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 skb_setup_tx_timestamp(skb, sockc->tsflags);
663
664 if (flags & MSG_CONFIRM)
665 skb_set_dst_pending_confirm(skb, 1);
666
667 skb->transport_header = skb->network_header;
668 err = memcpy_from_msg(iph, msg, length);
669 if (err) {
670 err = -EFAULT;
671 kfree_skb(skb);
672 goto error;
673 }
674
675 skb_dst_set(skb, &rt->dst);
676 *dstp = NULL;
677
678 /* if egress device is enslaved to an L3 master device pass the
679 * skb to its handler for processing
680 */
681 skb = l3mdev_ip6_out(sk, skb);
682 if (unlikely(!skb))
683 return 0;
684
685 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev
686 * in the error path. Since skb has been freed, the dst could
687 * have been queued for deletion.
688 */
689 rcu_read_lock();
690 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
691 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
692 NULL, rt->dst.dev, dst_output);
693 if (err > 0)
694 err = net_xmit_errno(err);
695 if (err) {
696 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
697 rcu_read_unlock();
698 goto error_check;
699 }
700 rcu_read_unlock();
701 out:
702 return 0;
703
704 error:
705 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
706 error_check:
707 if (err == -ENOBUFS && !np->recverr)
708 err = 0;
709 return err;
710 }
711
712 struct raw6_frag_vec {
713 struct msghdr *msg;
714 int hlen;
715 char c[4];
716 };
717
718 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
719 {
720 int err = 0;
721 switch (fl6->flowi6_proto) {
722 case IPPROTO_ICMPV6:
723 rfv->hlen = 2;
724 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
725 if (!err) {
726 fl6->fl6_icmp_type = rfv->c[0];
727 fl6->fl6_icmp_code = rfv->c[1];
728 }
729 break;
730 case IPPROTO_MH:
731 rfv->hlen = 4;
732 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
733 if (!err)
734 fl6->fl6_mh_type = rfv->c[2];
735 }
736 return err;
737 }
738
739 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
740 struct sk_buff *skb)
741 {
742 struct raw6_frag_vec *rfv = from;
743
744 if (offset < rfv->hlen) {
745 int copy = min(rfv->hlen - offset, len);
746
747 if (skb->ip_summed == CHECKSUM_PARTIAL)
748 memcpy(to, rfv->c + offset, copy);
749 else
750 skb->csum = csum_block_add(
751 skb->csum,
752 csum_partial_copy_nocheck(rfv->c + offset,
753 to, copy),
754 odd);
755
756 odd = 0;
757 offset += copy;
758 to += copy;
759 len -= copy;
760
761 if (!len)
762 return 0;
763 }
764
765 offset -= rfv->hlen;
766
767 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
768 }
769
770 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
771 {
772 struct ipv6_txoptions *opt_to_free = NULL;
773 struct ipv6_txoptions opt_space;
774 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
775 struct in6_addr *daddr, *final_p, final;
776 struct inet_sock *inet = inet_sk(sk);
777 struct ipv6_pinfo *np = inet6_sk(sk);
778 struct raw6_sock *rp = raw6_sk(sk);
779 struct ipv6_txoptions *opt = NULL;
780 struct ip6_flowlabel *flowlabel = NULL;
781 struct dst_entry *dst = NULL;
782 struct raw6_frag_vec rfv;
783 struct flowi6 fl6;
784 struct ipcm6_cookie ipc6;
785 int addr_len = msg->msg_namelen;
786 int hdrincl;
787 u16 proto;
788 int err;
789
790 /* Rough check on arithmetic overflow,
791 better check is made in ip6_append_data().
792 */
793 if (len > INT_MAX)
794 return -EMSGSIZE;
795
796 /* Mirror BSD error message compatibility */
797 if (msg->msg_flags & MSG_OOB)
798 return -EOPNOTSUPP;
799
800 /* hdrincl should be READ_ONCE(inet->hdrincl)
801 * but READ_ONCE() doesn't work with bit fields.
802 * Doing this indirectly yields the same result.
803 */
804 hdrincl = inet->hdrincl;
805 hdrincl = READ_ONCE(hdrincl);
806
807 /*
808 * Get and verify the address.
809 */
810 memset(&fl6, 0, sizeof(fl6));
811
812 fl6.flowi6_mark = sk->sk_mark;
813 fl6.flowi6_uid = sk->sk_uid;
814
815 ipcm6_init(&ipc6);
816 ipc6.sockc.tsflags = sk->sk_tsflags;
817 ipc6.sockc.mark = sk->sk_mark;
818
819 if (sin6) {
820 if (addr_len < SIN6_LEN_RFC2133)
821 return -EINVAL;
822
823 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
824 return -EAFNOSUPPORT;
825
826 /* port is the proto value [0..255] carried in nexthdr */
827 proto = ntohs(sin6->sin6_port);
828
829 if (!proto)
830 proto = inet->inet_num;
831 else if (proto != inet->inet_num)
832 return -EINVAL;
833
834 if (proto > 255)
835 return -EINVAL;
836
837 daddr = &sin6->sin6_addr;
838 if (np->sndflow) {
839 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
840 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
841 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
842 if (IS_ERR(flowlabel))
843 return -EINVAL;
844 }
845 }
846
847 /*
848 * Otherwise it will be difficult to maintain
849 * sk->sk_dst_cache.
850 */
851 if (sk->sk_state == TCP_ESTABLISHED &&
852 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
853 daddr = &sk->sk_v6_daddr;
854
855 if (addr_len >= sizeof(struct sockaddr_in6) &&
856 sin6->sin6_scope_id &&
857 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
858 fl6.flowi6_oif = sin6->sin6_scope_id;
859 } else {
860 if (sk->sk_state != TCP_ESTABLISHED)
861 return -EDESTADDRREQ;
862
863 proto = inet->inet_num;
864 daddr = &sk->sk_v6_daddr;
865 fl6.flowlabel = np->flow_label;
866 }
867
868 if (fl6.flowi6_oif == 0)
869 fl6.flowi6_oif = sk->sk_bound_dev_if;
870
871 if (msg->msg_controllen) {
872 opt = &opt_space;
873 memset(opt, 0, sizeof(struct ipv6_txoptions));
874 opt->tot_len = sizeof(struct ipv6_txoptions);
875 ipc6.opt = opt;
876
877 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
878 if (err < 0) {
879 fl6_sock_release(flowlabel);
880 return err;
881 }
882 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
883 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
884 if (IS_ERR(flowlabel))
885 return -EINVAL;
886 }
887 if (!(opt->opt_nflen|opt->opt_flen))
888 opt = NULL;
889 }
890 if (!opt) {
891 opt = txopt_get(np);
892 opt_to_free = opt;
893 }
894 if (flowlabel)
895 opt = fl6_merge_options(&opt_space, flowlabel, opt);
896 opt = ipv6_fixup_options(&opt_space, opt);
897
898 fl6.flowi6_proto = proto;
899 fl6.flowi6_mark = ipc6.sockc.mark;
900
901 if (!hdrincl) {
902 rfv.msg = msg;
903 rfv.hlen = 0;
904 err = rawv6_probe_proto_opt(&rfv, &fl6);
905 if (err)
906 goto out;
907 }
908
909 if (!ipv6_addr_any(daddr))
910 fl6.daddr = *daddr;
911 else
912 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
913 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
914 fl6.saddr = np->saddr;
915
916 final_p = fl6_update_dst(&fl6, opt, &final);
917
918 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
919 fl6.flowi6_oif = np->mcast_oif;
920 else if (!fl6.flowi6_oif)
921 fl6.flowi6_oif = np->ucast_oif;
922 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
923
924 if (hdrincl)
925 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
926
927 if (ipc6.tclass < 0)
928 ipc6.tclass = np->tclass;
929
930 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
931
932 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
933 if (IS_ERR(dst)) {
934 err = PTR_ERR(dst);
935 goto out;
936 }
937 if (ipc6.hlimit < 0)
938 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
939
940 if (ipc6.dontfrag < 0)
941 ipc6.dontfrag = np->dontfrag;
942
943 if (msg->msg_flags&MSG_CONFIRM)
944 goto do_confirm;
945
946 back_from_confirm:
947 if (hdrincl)
948 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst,
949 msg->msg_flags, &ipc6.sockc);
950 else {
951 ipc6.opt = opt;
952 lock_sock(sk);
953 err = ip6_append_data(sk, raw6_getfrag, &rfv,
954 len, 0, &ipc6, &fl6, (struct rt6_info *)dst,
955 msg->msg_flags);
956
957 if (err)
958 ip6_flush_pending_frames(sk);
959 else if (!(msg->msg_flags & MSG_MORE))
960 err = rawv6_push_pending_frames(sk, &fl6, rp);
961 release_sock(sk);
962 }
963 done:
964 dst_release(dst);
965 out:
966 fl6_sock_release(flowlabel);
967 txopt_put(opt_to_free);
968 return err < 0 ? err : len;
969 do_confirm:
970 if (msg->msg_flags & MSG_PROBE)
971 dst_confirm_neigh(dst, &fl6.daddr);
972 if (!(msg->msg_flags & MSG_PROBE) || len)
973 goto back_from_confirm;
974 err = 0;
975 goto done;
976 }
977
978 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
979 sockptr_t optval, int optlen)
980 {
981 switch (optname) {
982 case ICMPV6_FILTER:
983 if (optlen > sizeof(struct icmp6_filter))
984 optlen = sizeof(struct icmp6_filter);
985 if (copy_from_sockptr(&raw6_sk(sk)->filter, optval, optlen))
986 return -EFAULT;
987 return 0;
988 default:
989 return -ENOPROTOOPT;
990 }
991
992 return 0;
993 }
994
995 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
996 char __user *optval, int __user *optlen)
997 {
998 int len;
999
1000 switch (optname) {
1001 case ICMPV6_FILTER:
1002 if (get_user(len, optlen))
1003 return -EFAULT;
1004 if (len < 0)
1005 return -EINVAL;
1006 if (len > sizeof(struct icmp6_filter))
1007 len = sizeof(struct icmp6_filter);
1008 if (put_user(len, optlen))
1009 return -EFAULT;
1010 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
1011 return -EFAULT;
1012 return 0;
1013 default:
1014 return -ENOPROTOOPT;
1015 }
1016
1017 return 0;
1018 }
1019
1020
1021 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
1022 sockptr_t optval, unsigned int optlen)
1023 {
1024 struct raw6_sock *rp = raw6_sk(sk);
1025 int val;
1026
1027 if (copy_from_sockptr(&val, optval, sizeof(val)))
1028 return -EFAULT;
1029
1030 switch (optname) {
1031 case IPV6_HDRINCL:
1032 if (sk->sk_type != SOCK_RAW)
1033 return -EINVAL;
1034 inet_sk(sk)->hdrincl = !!val;
1035 return 0;
1036 case IPV6_CHECKSUM:
1037 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1038 level == IPPROTO_IPV6) {
1039 /*
1040 * RFC3542 tells that IPV6_CHECKSUM socket
1041 * option in the IPPROTO_IPV6 level is not
1042 * allowed on ICMPv6 sockets.
1043 * If you want to set it, use IPPROTO_RAW
1044 * level IPV6_CHECKSUM socket option
1045 * (Linux extension).
1046 */
1047 return -EINVAL;
1048 }
1049
1050 /* You may get strange result with a positive odd offset;
1051 RFC2292bis agrees with me. */
1052 if (val > 0 && (val&1))
1053 return -EINVAL;
1054 if (val < 0) {
1055 rp->checksum = 0;
1056 } else {
1057 rp->checksum = 1;
1058 rp->offset = val;
1059 }
1060
1061 return 0;
1062
1063 default:
1064 return -ENOPROTOOPT;
1065 }
1066 }
1067
1068 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1069 sockptr_t optval, unsigned int optlen)
1070 {
1071 switch (level) {
1072 case SOL_RAW:
1073 break;
1074
1075 case SOL_ICMPV6:
1076 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1077 return -EOPNOTSUPP;
1078 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1079 case SOL_IPV6:
1080 if (optname == IPV6_CHECKSUM ||
1081 optname == IPV6_HDRINCL)
1082 break;
1083 fallthrough;
1084 default:
1085 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1086 }
1087
1088 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1089 }
1090
1091 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1092 char __user *optval, int __user *optlen)
1093 {
1094 struct raw6_sock *rp = raw6_sk(sk);
1095 int val, len;
1096
1097 if (get_user(len, optlen))
1098 return -EFAULT;
1099
1100 switch (optname) {
1101 case IPV6_HDRINCL:
1102 val = inet_sk(sk)->hdrincl;
1103 break;
1104 case IPV6_CHECKSUM:
1105 /*
1106 * We allow getsockopt() for IPPROTO_IPV6-level
1107 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1108 * since RFC3542 is silent about it.
1109 */
1110 if (rp->checksum == 0)
1111 val = -1;
1112 else
1113 val = rp->offset;
1114 break;
1115
1116 default:
1117 return -ENOPROTOOPT;
1118 }
1119
1120 len = min_t(unsigned int, sizeof(int), len);
1121
1122 if (put_user(len, optlen))
1123 return -EFAULT;
1124 if (copy_to_user(optval, &val, len))
1125 return -EFAULT;
1126 return 0;
1127 }
1128
1129 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1130 char __user *optval, int __user *optlen)
1131 {
1132 switch (level) {
1133 case SOL_RAW:
1134 break;
1135
1136 case SOL_ICMPV6:
1137 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1138 return -EOPNOTSUPP;
1139 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1140 case SOL_IPV6:
1141 if (optname == IPV6_CHECKSUM ||
1142 optname == IPV6_HDRINCL)
1143 break;
1144 fallthrough;
1145 default:
1146 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1147 }
1148
1149 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1150 }
1151
1152 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1153 {
1154 switch (cmd) {
1155 case SIOCOUTQ: {
1156 int amount = sk_wmem_alloc_get(sk);
1157
1158 return put_user(amount, (int __user *)arg);
1159 }
1160 case SIOCINQ: {
1161 struct sk_buff *skb;
1162 int amount = 0;
1163
1164 spin_lock_bh(&sk->sk_receive_queue.lock);
1165 skb = skb_peek(&sk->sk_receive_queue);
1166 if (skb)
1167 amount = skb->len;
1168 spin_unlock_bh(&sk->sk_receive_queue.lock);
1169 return put_user(amount, (int __user *)arg);
1170 }
1171
1172 default:
1173 #ifdef CONFIG_IPV6_MROUTE
1174 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1175 #else
1176 return -ENOIOCTLCMD;
1177 #endif
1178 }
1179 }
1180
1181 #ifdef CONFIG_COMPAT
1182 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1183 {
1184 switch (cmd) {
1185 case SIOCOUTQ:
1186 case SIOCINQ:
1187 return -ENOIOCTLCMD;
1188 default:
1189 #ifdef CONFIG_IPV6_MROUTE
1190 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1191 #else
1192 return -ENOIOCTLCMD;
1193 #endif
1194 }
1195 }
1196 #endif
1197
1198 static void rawv6_close(struct sock *sk, long timeout)
1199 {
1200 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1201 ip6_ra_control(sk, -1);
1202 ip6mr_sk_done(sk);
1203 sk_common_release(sk);
1204 }
1205
1206 static void raw6_destroy(struct sock *sk)
1207 {
1208 lock_sock(sk);
1209 ip6_flush_pending_frames(sk);
1210 release_sock(sk);
1211
1212 inet6_destroy_sock(sk);
1213 }
1214
1215 static int rawv6_init_sk(struct sock *sk)
1216 {
1217 struct raw6_sock *rp = raw6_sk(sk);
1218
1219 switch (inet_sk(sk)->inet_num) {
1220 case IPPROTO_ICMPV6:
1221 rp->checksum = 1;
1222 rp->offset = 2;
1223 break;
1224 case IPPROTO_MH:
1225 rp->checksum = 1;
1226 rp->offset = 4;
1227 break;
1228 default:
1229 break;
1230 }
1231 return 0;
1232 }
1233
1234 struct proto rawv6_prot = {
1235 .name = "RAWv6",
1236 .owner = THIS_MODULE,
1237 .close = rawv6_close,
1238 .destroy = raw6_destroy,
1239 .connect = ip6_datagram_connect_v6_only,
1240 .disconnect = __udp_disconnect,
1241 .ioctl = rawv6_ioctl,
1242 .init = rawv6_init_sk,
1243 .setsockopt = rawv6_setsockopt,
1244 .getsockopt = rawv6_getsockopt,
1245 .sendmsg = rawv6_sendmsg,
1246 .recvmsg = rawv6_recvmsg,
1247 .bind = rawv6_bind,
1248 .backlog_rcv = rawv6_rcv_skb,
1249 .hash = raw_hash_sk,
1250 .unhash = raw_unhash_sk,
1251 .obj_size = sizeof(struct raw6_sock),
1252 .useroffset = offsetof(struct raw6_sock, filter),
1253 .usersize = sizeof_field(struct raw6_sock, filter),
1254 .h.raw_hash = &raw_v6_hashinfo,
1255 #ifdef CONFIG_COMPAT
1256 .compat_ioctl = compat_rawv6_ioctl,
1257 #endif
1258 .diag_destroy = raw_abort,
1259 };
1260
1261 #ifdef CONFIG_PROC_FS
1262 static int raw6_seq_show(struct seq_file *seq, void *v)
1263 {
1264 if (v == SEQ_START_TOKEN) {
1265 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1266 } else {
1267 struct sock *sp = v;
1268 __u16 srcp = inet_sk(sp)->inet_num;
1269 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1270 raw_seq_private(seq)->bucket);
1271 }
1272 return 0;
1273 }
1274
1275 static const struct seq_operations raw6_seq_ops = {
1276 .start = raw_seq_start,
1277 .next = raw_seq_next,
1278 .stop = raw_seq_stop,
1279 .show = raw6_seq_show,
1280 };
1281
1282 static int __net_init raw6_init_net(struct net *net)
1283 {
1284 if (!proc_create_net_data("raw6", 0444, net->proc_net, &raw6_seq_ops,
1285 sizeof(struct raw_iter_state), &raw_v6_hashinfo))
1286 return -ENOMEM;
1287
1288 return 0;
1289 }
1290
1291 static void __net_exit raw6_exit_net(struct net *net)
1292 {
1293 remove_proc_entry("raw6", net->proc_net);
1294 }
1295
1296 static struct pernet_operations raw6_net_ops = {
1297 .init = raw6_init_net,
1298 .exit = raw6_exit_net,
1299 };
1300
1301 int __init raw6_proc_init(void)
1302 {
1303 return register_pernet_subsys(&raw6_net_ops);
1304 }
1305
1306 void raw6_proc_exit(void)
1307 {
1308 unregister_pernet_subsys(&raw6_net_ops);
1309 }
1310 #endif /* CONFIG_PROC_FS */
1311
1312 /* Same as inet6_dgram_ops, sans udp_poll. */
1313 const struct proto_ops inet6_sockraw_ops = {
1314 .family = PF_INET6,
1315 .owner = THIS_MODULE,
1316 .release = inet6_release,
1317 .bind = inet6_bind,
1318 .connect = inet_dgram_connect, /* ok */
1319 .socketpair = sock_no_socketpair, /* a do nothing */
1320 .accept = sock_no_accept, /* a do nothing */
1321 .getname = inet6_getname,
1322 .poll = datagram_poll, /* ok */
1323 .ioctl = inet6_ioctl, /* must change */
1324 .gettstamp = sock_gettstamp,
1325 .listen = sock_no_listen, /* ok */
1326 .shutdown = inet_shutdown, /* ok */
1327 .setsockopt = sock_common_setsockopt, /* ok */
1328 .getsockopt = sock_common_getsockopt, /* ok */
1329 .sendmsg = inet_sendmsg, /* ok */
1330 .recvmsg = sock_common_recvmsg, /* ok */
1331 .mmap = sock_no_mmap,
1332 .sendpage = sock_no_sendpage,
1333 #ifdef CONFIG_COMPAT
1334 .compat_ioctl = inet6_compat_ioctl,
1335 #endif
1336 };
1337
1338 static struct inet_protosw rawv6_protosw = {
1339 .type = SOCK_RAW,
1340 .protocol = IPPROTO_IP, /* wild card */
1341 .prot = &rawv6_prot,
1342 .ops = &inet6_sockraw_ops,
1343 .flags = INET_PROTOSW_REUSE,
1344 };
1345
1346 int __init rawv6_init(void)
1347 {
1348 return inet6_register_protosw(&rawv6_protosw);
1349 }
1350
1351 void rawv6_exit(void)
1352 {
1353 inet6_unregister_protosw(&rawv6_protosw);
1354 }
1355
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.