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, 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 struct hlist_nulls_head *hlist;
145 struct hlist_nulls_node *hnode;
146 const struct in6_addr *saddr;
147 const struct in6_addr *daddr;
148 struct sock *sk;
149 bool delivered = false;
150 __u8 hash;
151
152 saddr = &ipv6_hdr(skb)->saddr;
153 daddr = saddr + 1;
154
155 hash = raw_hashfunc(net, nexthdr);
156 hlist = &raw_v6_hashinfo.ht[hash];
157 rcu_read_lock();
158 sk_nulls_for_each(sk, hnode, hlist) {
159 int filtered;
160
161 if (!raw_v6_match(net, sk, nexthdr, daddr, saddr,
162 inet6_iif(skb), inet6_sdif(skb)))
163 continue;
164 delivered = true;
165 switch (nexthdr) {
166 case IPPROTO_ICMPV6:
167 filtered = icmpv6_filter(sk, skb);
168 break;
169
170 #if IS_ENABLED(CONFIG_IPV6_MIP6)
171 case IPPROTO_MH:
172 {
173 /* XXX: To validate MH only once for each packet,
174 * this is placed here. It should be after checking
175 * xfrm policy, however it doesn't. The checking xfrm
176 * policy is placed in rawv6_rcv() because it is
177 * required for each socket.
178 */
179 mh_filter_t *filter;
180
181 filter = rcu_dereference(mh_filter);
182 filtered = filter ? (*filter)(sk, skb) : 0;
183 break;
184 }
185 #endif
186 default:
187 filtered = 0;
188 break;
189 }
190
191 if (filtered < 0)
192 break;
193 if (filtered == 0) {
194 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
195
196 /* Not releasing hash table! */
197 if (clone) {
198 nf_reset_ct(clone);
199 rawv6_rcv(sk, clone);
200 }
201 }
202 }
203 rcu_read_unlock();
204 return delivered;
205 }
206
207 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
208 {
209 return ipv6_raw_deliver(skb, nexthdr);
210 }
211
212 /* This cleans up af_inet6 a bit. -DaveM */
213 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
214 {
215 struct inet_sock *inet = inet_sk(sk);
216 struct ipv6_pinfo *np = inet6_sk(sk);
217 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
218 __be32 v4addr = 0;
219 int addr_type;
220 int err;
221
222 if (addr_len < SIN6_LEN_RFC2133)
223 return -EINVAL;
224
225 if (addr->sin6_family != AF_INET6)
226 return -EINVAL;
227
228 addr_type = ipv6_addr_type(&addr->sin6_addr);
229
230 /* Raw sockets are IPv6 only */
231 if (addr_type == IPV6_ADDR_MAPPED)
232 return -EADDRNOTAVAIL;
233
234 lock_sock(sk);
235
236 err = -EINVAL;
237 if (sk->sk_state != TCP_CLOSE)
238 goto out;
239
240 rcu_read_lock();
241 /* Check if the address belongs to the host. */
242 if (addr_type != IPV6_ADDR_ANY) {
243 struct net_device *dev = NULL;
244
245 if (__ipv6_addr_needs_scope_id(addr_type)) {
246 if (addr_len >= sizeof(struct sockaddr_in6) &&
247 addr->sin6_scope_id) {
248 /* Override any existing binding, if another
249 * one is supplied by user.
250 */
251 sk->sk_bound_dev_if = addr->sin6_scope_id;
252 }
253
254 /* Binding to link-local address requires an interface */
255 if (!sk->sk_bound_dev_if)
256 goto out_unlock;
257 }
258
259 if (sk->sk_bound_dev_if) {
260 err = -ENODEV;
261 dev = dev_get_by_index_rcu(sock_net(sk),
262 sk->sk_bound_dev_if);
263 if (!dev)
264 goto out_unlock;
265 }
266
267 /* ipv4 addr of the socket is invalid. Only the
268 * unspecified and mapped address have a v4 equivalent.
269 */
270 v4addr = LOOPBACK4_IPV6;
271 if (!(addr_type & IPV6_ADDR_MULTICAST) &&
272 !ipv6_can_nonlocal_bind(sock_net(sk), inet)) {
273 err = -EADDRNOTAVAIL;
274 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
275 dev, 0)) {
276 goto out_unlock;
277 }
278 }
279 }
280
281 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
282 sk->sk_v6_rcv_saddr = addr->sin6_addr;
283 if (!(addr_type & IPV6_ADDR_MULTICAST))
284 np->saddr = addr->sin6_addr;
285 err = 0;
286 out_unlock:
287 rcu_read_unlock();
288 out:
289 release_sock(sk);
290 return err;
291 }
292
293 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
294 struct inet6_skb_parm *opt,
295 u8 type, u8 code, int offset, __be32 info)
296 {
297 struct inet_sock *inet = inet_sk(sk);
298 struct ipv6_pinfo *np = inet6_sk(sk);
299 int err;
300 int harderr;
301
302 /* Report error on raw socket, if:
303 1. User requested recverr.
304 2. Socket is connected (otherwise the error indication
305 is useless without recverr and error is hard.
306 */
307 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
308 return;
309
310 harderr = icmpv6_err_convert(type, code, &err);
311 if (type == ICMPV6_PKT_TOOBIG) {
312 ip6_sk_update_pmtu(skb, sk, info);
313 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
314 }
315 if (type == NDISC_REDIRECT) {
316 ip6_sk_redirect(skb, sk);
317 return;
318 }
319 if (np->recverr) {
320 u8 *payload = skb->data;
321 if (!inet->hdrincl)
322 payload += offset;
323 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
324 }
325
326 if (np->recverr || harderr) {
327 sk->sk_err = err;
328 sk_error_report(sk);
329 }
330 }
331
332 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
333 u8 type, u8 code, int inner_offset, __be32 info)
334 {
335 struct net *net = dev_net(skb->dev);
336 struct hlist_nulls_head *hlist;
337 struct hlist_nulls_node *hnode;
338 struct sock *sk;
339 int hash;
340
341 hash = raw_hashfunc(net, nexthdr);
342 hlist = &raw_v6_hashinfo.ht[hash];
343 rcu_read_lock();
344 sk_nulls_for_each(sk, hnode, hlist) {
345 /* Note: ipv6_hdr(skb) != skb->data */
346 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
347
348 if (!raw_v6_match(net, sk, nexthdr, &ip6h->saddr, &ip6h->daddr,
349 inet6_iif(skb), inet6_iif(skb)))
350 continue;
351 rawv6_err(sk, skb, NULL, type, code, inner_offset, info);
352 }
353 rcu_read_unlock();
354 }
355
356 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
357 {
358 enum skb_drop_reason reason;
359
360 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
361 skb_checksum_complete(skb)) {
362 atomic_inc(&sk->sk_drops);
363 kfree_skb_reason(skb, SKB_DROP_REASON_SKB_CSUM);
364 return NET_RX_DROP;
365 }
366
367 /* Charge it to the socket. */
368 skb_dst_drop(skb);
369 if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) {
370 kfree_skb_reason(skb, reason);
371 return NET_RX_DROP;
372 }
373
374 return 0;
375 }
376
377 /*
378 * This is next to useless...
379 * if we demultiplex in network layer we don't need the extra call
380 * just to queue the skb...
381 * maybe we could have the network decide upon a hint if it
382 * should call raw_rcv for demultiplexing
383 */
384 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
385 {
386 struct inet_sock *inet = inet_sk(sk);
387 struct raw6_sock *rp = raw6_sk(sk);
388
389 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
390 atomic_inc(&sk->sk_drops);
391 kfree_skb_reason(skb, SKB_DROP_REASON_XFRM_POLICY);
392 return NET_RX_DROP;
393 }
394
395 if (!rp->checksum)
396 skb->ip_summed = CHECKSUM_UNNECESSARY;
397
398 if (skb->ip_summed == CHECKSUM_COMPLETE) {
399 skb_postpull_rcsum(skb, skb_network_header(skb),
400 skb_network_header_len(skb));
401 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
402 &ipv6_hdr(skb)->daddr,
403 skb->len, inet->inet_num, skb->csum))
404 skb->ip_summed = CHECKSUM_UNNECESSARY;
405 }
406 if (!skb_csum_unnecessary(skb))
407 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
408 &ipv6_hdr(skb)->daddr,
409 skb->len,
410 inet->inet_num, 0));
411
412 if (inet->hdrincl) {
413 if (skb_checksum_complete(skb)) {
414 atomic_inc(&sk->sk_drops);
415 kfree_skb_reason(skb, SKB_DROP_REASON_SKB_CSUM);
416 return NET_RX_DROP;
417 }
418 }
419
420 rawv6_rcv_skb(sk, skb);
421 return 0;
422 }
423
424
425 /*
426 * This should be easy, if there is something there
427 * we return it, otherwise we block.
428 */
429
430 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
431 int flags, int *addr_len)
432 {
433 struct ipv6_pinfo *np = inet6_sk(sk);
434 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
435 struct sk_buff *skb;
436 size_t copied;
437 int err;
438
439 if (flags & MSG_OOB)
440 return -EOPNOTSUPP;
441
442 if (flags & MSG_ERRQUEUE)
443 return ipv6_recv_error(sk, msg, len, addr_len);
444
445 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
446 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
447
448 skb = skb_recv_datagram(sk, flags, &err);
449 if (!skb)
450 goto out;
451 if (ccs_socket_post_recvmsg_permission(sk, skb, flags)) {
452 err = -EAGAIN; /* Hope less harmful than -EPERM. */
453 goto out;
454 }
455
456 copied = skb->len;
457 if (copied > len) {
458 copied = len;
459 msg->msg_flags |= MSG_TRUNC;
460 }
461
462 if (skb_csum_unnecessary(skb)) {
463 err = skb_copy_datagram_msg(skb, 0, msg, copied);
464 } else if (msg->msg_flags&MSG_TRUNC) {
465 if (__skb_checksum_complete(skb))
466 goto csum_copy_err;
467 err = skb_copy_datagram_msg(skb, 0, msg, copied);
468 } else {
469 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
470 if (err == -EINVAL)
471 goto csum_copy_err;
472 }
473 if (err)
474 goto out_free;
475
476 /* Copy the address. */
477 if (sin6) {
478 sin6->sin6_family = AF_INET6;
479 sin6->sin6_port = 0;
480 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
481 sin6->sin6_flowinfo = 0;
482 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
483 inet6_iif(skb));
484 *addr_len = sizeof(*sin6);
485 }
486
487 sock_recv_cmsgs(msg, sk, skb);
488
489 if (np->rxopt.all)
490 ip6_datagram_recv_ctl(sk, msg, skb);
491
492 err = copied;
493 if (flags & MSG_TRUNC)
494 err = skb->len;
495
496 out_free:
497 skb_free_datagram(sk, skb);
498 out:
499 return err;
500
501 csum_copy_err:
502 skb_kill_datagram(sk, skb, flags);
503
504 /* Error for blocking case is chosen to masquerade
505 as some normal condition.
506 */
507 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
508 goto out;
509 }
510
511 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
512 struct raw6_sock *rp)
513 {
514 struct ipv6_txoptions *opt;
515 struct sk_buff *skb;
516 int err = 0;
517 int offset;
518 int len;
519 int total_len;
520 __wsum tmp_csum;
521 __sum16 csum;
522
523 if (!rp->checksum)
524 goto send;
525
526 skb = skb_peek(&sk->sk_write_queue);
527 if (!skb)
528 goto out;
529
530 offset = rp->offset;
531 total_len = inet_sk(sk)->cork.base.length;
532 opt = inet6_sk(sk)->cork.opt;
533 total_len -= opt ? opt->opt_flen : 0;
534
535 if (offset >= total_len - 1) {
536 err = -EINVAL;
537 ip6_flush_pending_frames(sk);
538 goto out;
539 }
540
541 /* should be check HW csum miyazawa */
542 if (skb_queue_len(&sk->sk_write_queue) == 1) {
543 /*
544 * Only one fragment on the socket.
545 */
546 tmp_csum = skb->csum;
547 } else {
548 struct sk_buff *csum_skb = NULL;
549 tmp_csum = 0;
550
551 skb_queue_walk(&sk->sk_write_queue, skb) {
552 tmp_csum = csum_add(tmp_csum, skb->csum);
553
554 if (csum_skb)
555 continue;
556
557 len = skb->len - skb_transport_offset(skb);
558 if (offset >= len) {
559 offset -= len;
560 continue;
561 }
562
563 csum_skb = skb;
564 }
565
566 skb = csum_skb;
567 }
568
569 offset += skb_transport_offset(skb);
570 err = skb_copy_bits(skb, offset, &csum, 2);
571 if (err < 0) {
572 ip6_flush_pending_frames(sk);
573 goto out;
574 }
575
576 /* in case cksum was not initialized */
577 if (unlikely(csum))
578 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
579
580 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
581 total_len, fl6->flowi6_proto, tmp_csum);
582
583 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
584 csum = CSUM_MANGLED_0;
585
586 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
587
588 send:
589 err = ip6_push_pending_frames(sk);
590 out:
591 return err;
592 }
593
594 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
595 struct flowi6 *fl6, struct dst_entry **dstp,
596 unsigned int flags, const struct sockcm_cookie *sockc)
597 {
598 struct ipv6_pinfo *np = inet6_sk(sk);
599 struct net *net = sock_net(sk);
600 struct ipv6hdr *iph;
601 struct sk_buff *skb;
602 int err;
603 struct rt6_info *rt = (struct rt6_info *)*dstp;
604 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
605 int tlen = rt->dst.dev->needed_tailroom;
606
607 if (length > rt->dst.dev->mtu) {
608 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
609 return -EMSGSIZE;
610 }
611 if (length < sizeof(struct ipv6hdr))
612 return -EINVAL;
613 if (flags&MSG_PROBE)
614 goto out;
615
616 skb = sock_alloc_send_skb(sk,
617 length + hlen + tlen + 15,
618 flags & MSG_DONTWAIT, &err);
619 if (!skb)
620 goto error;
621 skb_reserve(skb, hlen);
622
623 skb->protocol = htons(ETH_P_IPV6);
624 skb->priority = sk->sk_priority;
625 skb->mark = sockc->mark;
626 skb->tstamp = sockc->transmit_time;
627
628 skb_put(skb, length);
629 skb_reset_network_header(skb);
630 iph = ipv6_hdr(skb);
631
632 skb->ip_summed = CHECKSUM_NONE;
633
634 skb_setup_tx_timestamp(skb, sockc->tsflags);
635
636 if (flags & MSG_CONFIRM)
637 skb_set_dst_pending_confirm(skb, 1);
638
639 skb->transport_header = skb->network_header;
640 err = memcpy_from_msg(iph, msg, length);
641 if (err) {
642 err = -EFAULT;
643 kfree_skb(skb);
644 goto error;
645 }
646
647 skb_dst_set(skb, &rt->dst);
648 *dstp = NULL;
649
650 /* if egress device is enslaved to an L3 master device pass the
651 * skb to its handler for processing
652 */
653 skb = l3mdev_ip6_out(sk, skb);
654 if (unlikely(!skb))
655 return 0;
656
657 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev
658 * in the error path. Since skb has been freed, the dst could
659 * have been queued for deletion.
660 */
661 rcu_read_lock();
662 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
663 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
664 NULL, rt->dst.dev, dst_output);
665 if (err > 0)
666 err = net_xmit_errno(err);
667 if (err) {
668 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
669 rcu_read_unlock();
670 goto error_check;
671 }
672 rcu_read_unlock();
673 out:
674 return 0;
675
676 error:
677 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
678 error_check:
679 if (err == -ENOBUFS && !np->recverr)
680 err = 0;
681 return err;
682 }
683
684 struct raw6_frag_vec {
685 struct msghdr *msg;
686 int hlen;
687 char c[4];
688 };
689
690 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
691 {
692 int err = 0;
693 switch (fl6->flowi6_proto) {
694 case IPPROTO_ICMPV6:
695 rfv->hlen = 2;
696 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
697 if (!err) {
698 fl6->fl6_icmp_type = rfv->c[0];
699 fl6->fl6_icmp_code = rfv->c[1];
700 }
701 break;
702 case IPPROTO_MH:
703 rfv->hlen = 4;
704 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
705 if (!err)
706 fl6->fl6_mh_type = rfv->c[2];
707 }
708 return err;
709 }
710
711 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
712 struct sk_buff *skb)
713 {
714 struct raw6_frag_vec *rfv = from;
715
716 if (offset < rfv->hlen) {
717 int copy = min(rfv->hlen - offset, len);
718
719 if (skb->ip_summed == CHECKSUM_PARTIAL)
720 memcpy(to, rfv->c + offset, copy);
721 else
722 skb->csum = csum_block_add(
723 skb->csum,
724 csum_partial_copy_nocheck(rfv->c + offset,
725 to, copy),
726 odd);
727
728 odd = 0;
729 offset += copy;
730 to += copy;
731 len -= copy;
732
733 if (!len)
734 return 0;
735 }
736
737 offset -= rfv->hlen;
738
739 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
740 }
741
742 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
743 {
744 struct ipv6_txoptions *opt_to_free = NULL;
745 struct ipv6_txoptions opt_space;
746 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
747 struct in6_addr *daddr, *final_p, final;
748 struct inet_sock *inet = inet_sk(sk);
749 struct ipv6_pinfo *np = inet6_sk(sk);
750 struct raw6_sock *rp = raw6_sk(sk);
751 struct ipv6_txoptions *opt = NULL;
752 struct ip6_flowlabel *flowlabel = NULL;
753 struct dst_entry *dst = NULL;
754 struct raw6_frag_vec rfv;
755 struct flowi6 fl6;
756 struct ipcm6_cookie ipc6;
757 int addr_len = msg->msg_namelen;
758 int hdrincl;
759 u16 proto;
760 int err;
761
762 /* Rough check on arithmetic overflow,
763 better check is made in ip6_append_data().
764 */
765 if (len > INT_MAX)
766 return -EMSGSIZE;
767
768 /* Mirror BSD error message compatibility */
769 if (msg->msg_flags & MSG_OOB)
770 return -EOPNOTSUPP;
771
772 /* hdrincl should be READ_ONCE(inet->hdrincl)
773 * but READ_ONCE() doesn't work with bit fields.
774 * Doing this indirectly yields the same result.
775 */
776 hdrincl = inet->hdrincl;
777 hdrincl = READ_ONCE(hdrincl);
778
779 /*
780 * Get and verify the address.
781 */
782 memset(&fl6, 0, sizeof(fl6));
783
784 fl6.flowi6_mark = sk->sk_mark;
785 fl6.flowi6_uid = sk->sk_uid;
786
787 ipcm6_init(&ipc6);
788 ipc6.sockc.tsflags = sk->sk_tsflags;
789 ipc6.sockc.mark = sk->sk_mark;
790
791 if (sin6) {
792 if (addr_len < SIN6_LEN_RFC2133)
793 return -EINVAL;
794
795 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
796 return -EAFNOSUPPORT;
797
798 /* port is the proto value [0..255] carried in nexthdr */
799 proto = ntohs(sin6->sin6_port);
800
801 if (!proto)
802 proto = inet->inet_num;
803 else if (proto != inet->inet_num)
804 return -EINVAL;
805
806 if (proto > 255)
807 return -EINVAL;
808
809 daddr = &sin6->sin6_addr;
810 if (np->sndflow) {
811 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
812 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
813 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
814 if (IS_ERR(flowlabel))
815 return -EINVAL;
816 }
817 }
818
819 /*
820 * Otherwise it will be difficult to maintain
821 * sk->sk_dst_cache.
822 */
823 if (sk->sk_state == TCP_ESTABLISHED &&
824 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
825 daddr = &sk->sk_v6_daddr;
826
827 if (addr_len >= sizeof(struct sockaddr_in6) &&
828 sin6->sin6_scope_id &&
829 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
830 fl6.flowi6_oif = sin6->sin6_scope_id;
831 } else {
832 if (sk->sk_state != TCP_ESTABLISHED)
833 return -EDESTADDRREQ;
834
835 proto = inet->inet_num;
836 daddr = &sk->sk_v6_daddr;
837 fl6.flowlabel = np->flow_label;
838 }
839
840 if (fl6.flowi6_oif == 0)
841 fl6.flowi6_oif = sk->sk_bound_dev_if;
842
843 if (msg->msg_controllen) {
844 opt = &opt_space;
845 memset(opt, 0, sizeof(struct ipv6_txoptions));
846 opt->tot_len = sizeof(struct ipv6_txoptions);
847 ipc6.opt = opt;
848
849 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
850 if (err < 0) {
851 fl6_sock_release(flowlabel);
852 return err;
853 }
854 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
855 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
856 if (IS_ERR(flowlabel))
857 return -EINVAL;
858 }
859 if (!(opt->opt_nflen|opt->opt_flen))
860 opt = NULL;
861 }
862 if (!opt) {
863 opt = txopt_get(np);
864 opt_to_free = opt;
865 }
866 if (flowlabel)
867 opt = fl6_merge_options(&opt_space, flowlabel, opt);
868 opt = ipv6_fixup_options(&opt_space, opt);
869
870 fl6.flowi6_proto = proto;
871 fl6.flowi6_mark = ipc6.sockc.mark;
872
873 if (!hdrincl) {
874 rfv.msg = msg;
875 rfv.hlen = 0;
876 err = rawv6_probe_proto_opt(&rfv, &fl6);
877 if (err)
878 goto out;
879 }
880
881 if (!ipv6_addr_any(daddr))
882 fl6.daddr = *daddr;
883 else
884 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
885 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
886 fl6.saddr = np->saddr;
887
888 final_p = fl6_update_dst(&fl6, opt, &final);
889
890 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
891 fl6.flowi6_oif = np->mcast_oif;
892 else if (!fl6.flowi6_oif)
893 fl6.flowi6_oif = np->ucast_oif;
894 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
895
896 if (hdrincl)
897 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
898
899 if (ipc6.tclass < 0)
900 ipc6.tclass = np->tclass;
901
902 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
903
904 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
905 if (IS_ERR(dst)) {
906 err = PTR_ERR(dst);
907 goto out;
908 }
909 if (ipc6.hlimit < 0)
910 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
911
912 if (ipc6.dontfrag < 0)
913 ipc6.dontfrag = np->dontfrag;
914
915 if (msg->msg_flags&MSG_CONFIRM)
916 goto do_confirm;
917
918 back_from_confirm:
919 if (hdrincl)
920 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst,
921 msg->msg_flags, &ipc6.sockc);
922 else {
923 ipc6.opt = opt;
924 lock_sock(sk);
925 err = ip6_append_data(sk, raw6_getfrag, &rfv,
926 len, 0, &ipc6, &fl6, (struct rt6_info *)dst,
927 msg->msg_flags);
928
929 if (err)
930 ip6_flush_pending_frames(sk);
931 else if (!(msg->msg_flags & MSG_MORE))
932 err = rawv6_push_pending_frames(sk, &fl6, rp);
933 release_sock(sk);
934 }
935 done:
936 dst_release(dst);
937 out:
938 fl6_sock_release(flowlabel);
939 txopt_put(opt_to_free);
940 return err < 0 ? err : len;
941 do_confirm:
942 if (msg->msg_flags & MSG_PROBE)
943 dst_confirm_neigh(dst, &fl6.daddr);
944 if (!(msg->msg_flags & MSG_PROBE) || len)
945 goto back_from_confirm;
946 err = 0;
947 goto done;
948 }
949
950 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
951 sockptr_t optval, int optlen)
952 {
953 switch (optname) {
954 case ICMPV6_FILTER:
955 if (optlen > sizeof(struct icmp6_filter))
956 optlen = sizeof(struct icmp6_filter);
957 if (copy_from_sockptr(&raw6_sk(sk)->filter, optval, optlen))
958 return -EFAULT;
959 return 0;
960 default:
961 return -ENOPROTOOPT;
962 }
963
964 return 0;
965 }
966
967 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
968 char __user *optval, int __user *optlen)
969 {
970 int len;
971
972 switch (optname) {
973 case ICMPV6_FILTER:
974 if (get_user(len, optlen))
975 return -EFAULT;
976 if (len < 0)
977 return -EINVAL;
978 if (len > sizeof(struct icmp6_filter))
979 len = sizeof(struct icmp6_filter);
980 if (put_user(len, optlen))
981 return -EFAULT;
982 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
983 return -EFAULT;
984 return 0;
985 default:
986 return -ENOPROTOOPT;
987 }
988
989 return 0;
990 }
991
992
993 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
994 sockptr_t optval, unsigned int optlen)
995 {
996 struct raw6_sock *rp = raw6_sk(sk);
997 int val;
998
999 if (optlen < sizeof(val))
1000 return -EINVAL;
1001
1002 if (copy_from_sockptr(&val, optval, sizeof(val)))
1003 return -EFAULT;
1004
1005 switch (optname) {
1006 case IPV6_HDRINCL:
1007 if (sk->sk_type != SOCK_RAW)
1008 return -EINVAL;
1009 inet_sk(sk)->hdrincl = !!val;
1010 return 0;
1011 case IPV6_CHECKSUM:
1012 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1013 level == IPPROTO_IPV6) {
1014 /*
1015 * RFC3542 tells that IPV6_CHECKSUM socket
1016 * option in the IPPROTO_IPV6 level is not
1017 * allowed on ICMPv6 sockets.
1018 * If you want to set it, use IPPROTO_RAW
1019 * level IPV6_CHECKSUM socket option
1020 * (Linux extension).
1021 */
1022 return -EINVAL;
1023 }
1024
1025 /* You may get strange result with a positive odd offset;
1026 RFC2292bis agrees with me. */
1027 if (val > 0 && (val&1))
1028 return -EINVAL;
1029 if (val < 0) {
1030 rp->checksum = 0;
1031 } else {
1032 rp->checksum = 1;
1033 rp->offset = val;
1034 }
1035
1036 return 0;
1037
1038 default:
1039 return -ENOPROTOOPT;
1040 }
1041 }
1042
1043 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1044 sockptr_t optval, unsigned int optlen)
1045 {
1046 switch (level) {
1047 case SOL_RAW:
1048 break;
1049
1050 case SOL_ICMPV6:
1051 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1052 return -EOPNOTSUPP;
1053 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1054 case SOL_IPV6:
1055 if (optname == IPV6_CHECKSUM ||
1056 optname == IPV6_HDRINCL)
1057 break;
1058 fallthrough;
1059 default:
1060 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1061 }
1062
1063 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1064 }
1065
1066 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1067 char __user *optval, int __user *optlen)
1068 {
1069 struct raw6_sock *rp = raw6_sk(sk);
1070 int val, len;
1071
1072 if (get_user(len, optlen))
1073 return -EFAULT;
1074
1075 switch (optname) {
1076 case IPV6_HDRINCL:
1077 val = inet_sk(sk)->hdrincl;
1078 break;
1079 case IPV6_CHECKSUM:
1080 /*
1081 * We allow getsockopt() for IPPROTO_IPV6-level
1082 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1083 * since RFC3542 is silent about it.
1084 */
1085 if (rp->checksum == 0)
1086 val = -1;
1087 else
1088 val = rp->offset;
1089 break;
1090
1091 default:
1092 return -ENOPROTOOPT;
1093 }
1094
1095 len = min_t(unsigned int, sizeof(int), len);
1096
1097 if (put_user(len, optlen))
1098 return -EFAULT;
1099 if (copy_to_user(optval, &val, len))
1100 return -EFAULT;
1101 return 0;
1102 }
1103
1104 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1105 char __user *optval, int __user *optlen)
1106 {
1107 switch (level) {
1108 case SOL_RAW:
1109 break;
1110
1111 case SOL_ICMPV6:
1112 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1113 return -EOPNOTSUPP;
1114 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1115 case SOL_IPV6:
1116 if (optname == IPV6_CHECKSUM ||
1117 optname == IPV6_HDRINCL)
1118 break;
1119 fallthrough;
1120 default:
1121 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1122 }
1123
1124 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1125 }
1126
1127 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1128 {
1129 switch (cmd) {
1130 case SIOCOUTQ: {
1131 int amount = sk_wmem_alloc_get(sk);
1132
1133 return put_user(amount, (int __user *)arg);
1134 }
1135 case SIOCINQ: {
1136 struct sk_buff *skb;
1137 int amount = 0;
1138
1139 spin_lock_bh(&sk->sk_receive_queue.lock);
1140 skb = skb_peek(&sk->sk_receive_queue);
1141 if (skb)
1142 amount = skb->len;
1143 spin_unlock_bh(&sk->sk_receive_queue.lock);
1144 return put_user(amount, (int __user *)arg);
1145 }
1146
1147 default:
1148 #ifdef CONFIG_IPV6_MROUTE
1149 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1150 #else
1151 return -ENOIOCTLCMD;
1152 #endif
1153 }
1154 }
1155
1156 #ifdef CONFIG_COMPAT
1157 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1158 {
1159 switch (cmd) {
1160 case SIOCOUTQ:
1161 case SIOCINQ:
1162 return -ENOIOCTLCMD;
1163 default:
1164 #ifdef CONFIG_IPV6_MROUTE
1165 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1166 #else
1167 return -ENOIOCTLCMD;
1168 #endif
1169 }
1170 }
1171 #endif
1172
1173 static void rawv6_close(struct sock *sk, long timeout)
1174 {
1175 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1176 ip6_ra_control(sk, -1);
1177 ip6mr_sk_done(sk);
1178 sk_common_release(sk);
1179 }
1180
1181 static void raw6_destroy(struct sock *sk)
1182 {
1183 lock_sock(sk);
1184 ip6_flush_pending_frames(sk);
1185 release_sock(sk);
1186 }
1187
1188 static int rawv6_init_sk(struct sock *sk)
1189 {
1190 struct raw6_sock *rp = raw6_sk(sk);
1191
1192 switch (inet_sk(sk)->inet_num) {
1193 case IPPROTO_ICMPV6:
1194 rp->checksum = 1;
1195 rp->offset = 2;
1196 break;
1197 case IPPROTO_MH:
1198 rp->checksum = 1;
1199 rp->offset = 4;
1200 break;
1201 default:
1202 break;
1203 }
1204 return 0;
1205 }
1206
1207 struct proto rawv6_prot = {
1208 .name = "RAWv6",
1209 .owner = THIS_MODULE,
1210 .close = rawv6_close,
1211 .destroy = raw6_destroy,
1212 .connect = ip6_datagram_connect_v6_only,
1213 .disconnect = __udp_disconnect,
1214 .ioctl = rawv6_ioctl,
1215 .init = rawv6_init_sk,
1216 .setsockopt = rawv6_setsockopt,
1217 .getsockopt = rawv6_getsockopt,
1218 .sendmsg = rawv6_sendmsg,
1219 .recvmsg = rawv6_recvmsg,
1220 .bind = rawv6_bind,
1221 .backlog_rcv = rawv6_rcv_skb,
1222 .hash = raw_hash_sk,
1223 .unhash = raw_unhash_sk,
1224 .obj_size = sizeof(struct raw6_sock),
1225 .useroffset = offsetof(struct raw6_sock, filter),
1226 .usersize = sizeof_field(struct raw6_sock, filter),
1227 .h.raw_hash = &raw_v6_hashinfo,
1228 #ifdef CONFIG_COMPAT
1229 .compat_ioctl = compat_rawv6_ioctl,
1230 #endif
1231 .diag_destroy = raw_abort,
1232 };
1233
1234 #ifdef CONFIG_PROC_FS
1235 static int raw6_seq_show(struct seq_file *seq, void *v)
1236 {
1237 if (v == SEQ_START_TOKEN) {
1238 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1239 } else {
1240 struct sock *sp = v;
1241 __u16 srcp = inet_sk(sp)->inet_num;
1242 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1243 raw_seq_private(seq)->bucket);
1244 }
1245 return 0;
1246 }
1247
1248 static const struct seq_operations raw6_seq_ops = {
1249 .start = raw_seq_start,
1250 .next = raw_seq_next,
1251 .stop = raw_seq_stop,
1252 .show = raw6_seq_show,
1253 };
1254
1255 static int __net_init raw6_init_net(struct net *net)
1256 {
1257 if (!proc_create_net_data("raw6", 0444, net->proc_net, &raw6_seq_ops,
1258 sizeof(struct raw_iter_state), &raw_v6_hashinfo))
1259 return -ENOMEM;
1260
1261 return 0;
1262 }
1263
1264 static void __net_exit raw6_exit_net(struct net *net)
1265 {
1266 remove_proc_entry("raw6", net->proc_net);
1267 }
1268
1269 static struct pernet_operations raw6_net_ops = {
1270 .init = raw6_init_net,
1271 .exit = raw6_exit_net,
1272 };
1273
1274 int __init raw6_proc_init(void)
1275 {
1276 return register_pernet_subsys(&raw6_net_ops);
1277 }
1278
1279 void raw6_proc_exit(void)
1280 {
1281 unregister_pernet_subsys(&raw6_net_ops);
1282 }
1283 #endif /* CONFIG_PROC_FS */
1284
1285 /* Same as inet6_dgram_ops, sans udp_poll. */
1286 const struct proto_ops inet6_sockraw_ops = {
1287 .family = PF_INET6,
1288 .owner = THIS_MODULE,
1289 .release = inet6_release,
1290 .bind = inet6_bind,
1291 .connect = inet_dgram_connect, /* ok */
1292 .socketpair = sock_no_socketpair, /* a do nothing */
1293 .accept = sock_no_accept, /* a do nothing */
1294 .getname = inet6_getname,
1295 .poll = datagram_poll, /* ok */
1296 .ioctl = inet6_ioctl, /* must change */
1297 .gettstamp = sock_gettstamp,
1298 .listen = sock_no_listen, /* ok */
1299 .shutdown = inet_shutdown, /* ok */
1300 .setsockopt = sock_common_setsockopt, /* ok */
1301 .getsockopt = sock_common_getsockopt, /* ok */
1302 .sendmsg = inet_sendmsg, /* ok */
1303 .recvmsg = sock_common_recvmsg, /* ok */
1304 .mmap = sock_no_mmap,
1305 .sendpage = sock_no_sendpage,
1306 #ifdef CONFIG_COMPAT
1307 .compat_ioctl = inet6_compat_ioctl,
1308 #endif
1309 };
1310
1311 static struct inet_protosw rawv6_protosw = {
1312 .type = SOCK_RAW,
1313 .protocol = IPPROTO_IP, /* wild card */
1314 .prot = &rawv6_prot,
1315 .ops = &inet6_sockraw_ops,
1316 .flags = INET_PROTOSW_REUSE,
1317 };
1318
1319 int __init rawv6_init(void)
1320 {
1321 return inet6_register_protosw(&rawv6_protosw);
1322 }
1323
1324 void rawv6_exit(void)
1325 {
1326 inet6_unregister_protosw(&rawv6_protosw);
1327 }
1328
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.