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
raw_v6_match(struct net * net,const struct sock * sk,unsigned short num,const struct in6_addr * loc_addr,const struct in6_addr * rmt_addr,int dif,int sdif)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 */
icmpv6_filter(const struct sock * sk,const struct sk_buff * skb)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
rawv6_mh_filter_register(mh_filter_t filter)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
rawv6_mh_filter_unregister(mh_filter_t filter)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 */
ipv6_raw_deliver(struct sk_buff * skb,int nexthdr)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
raw6_local_deliver(struct sk_buff * skb,int nexthdr)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 */
rawv6_bind(struct sock * sk,struct sockaddr * uaddr,int addr_len)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
rawv6_err(struct sock * sk,struct sk_buff * skb,u8 type,u8 code,int offset,__be32 info)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
raw6_icmp_error(struct sk_buff * skb,int nexthdr,u8 type,u8 code,int inner_offset,__be32 info)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
rawv6_rcv_skb(struct sock * sk,struct sk_buff * skb)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 */
rawv6_rcv(struct sock * sk,struct sk_buff * skb)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
rawv6_recvmsg(struct sock * sk,struct msghdr * msg,size_t len,int flags,int * addr_len)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
455 copied = skb->len;
456 if (copied > len) {
457 copied = len;
458 msg->msg_flags |= MSG_TRUNC;
459 }
460
461 if (skb_csum_unnecessary(skb)) {
462 err = skb_copy_datagram_msg(skb, 0, msg, copied);
463 } else if (msg->msg_flags&MSG_TRUNC) {
464 if (__skb_checksum_complete(skb))
465 goto csum_copy_err;
466 err = skb_copy_datagram_msg(skb, 0, msg, copied);
467 } else {
468 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
469 if (err == -EINVAL)
470 goto csum_copy_err;
471 }
472 if (err)
473 goto out_free;
474
475 /* Copy the address. */
476 if (sin6) {
477 sin6->sin6_family = AF_INET6;
478 sin6->sin6_port = 0;
479 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
480 sin6->sin6_flowinfo = 0;
481 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
482 inet6_iif(skb));
483 *addr_len = sizeof(*sin6);
484 }
485
486 sock_recv_cmsgs(msg, sk, skb);
487
488 if (np->rxopt.all)
489 ip6_datagram_recv_ctl(sk, msg, skb);
490
491 err = copied;
492 if (flags & MSG_TRUNC)
493 err = skb->len;
494
495 out_free:
496 skb_free_datagram(sk, skb);
497 out:
498 return err;
499
500 csum_copy_err:
501 skb_kill_datagram(sk, skb, flags);
502
503 /* Error for blocking case is chosen to masquerade
504 as some normal condition.
505 */
506 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
507 goto out;
508 }
509
rawv6_push_pending_frames(struct sock * sk,struct flowi6 * fl6,struct raw6_sock * rp)510 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
511 struct raw6_sock *rp)
512 {
513 struct ipv6_txoptions *opt;
514 struct sk_buff *skb;
515 int err = 0;
516 int offset;
517 int len;
518 int total_len;
519 __wsum tmp_csum;
520 __sum16 csum;
521
522 if (!rp->checksum)
523 goto send;
524
525 skb = skb_peek(&sk->sk_write_queue);
526 if (!skb)
527 goto out;
528
529 offset = rp->offset;
530 total_len = inet_sk(sk)->cork.base.length;
531 opt = inet6_sk(sk)->cork.opt;
532 total_len -= opt ? opt->opt_flen : 0;
533
534 if (offset >= total_len - 1) {
535 err = -EINVAL;
536 ip6_flush_pending_frames(sk);
537 goto out;
538 }
539
540 /* should be check HW csum miyazawa */
541 if (skb_queue_len(&sk->sk_write_queue) == 1) {
542 /*
543 * Only one fragment on the socket.
544 */
545 tmp_csum = skb->csum;
546 } else {
547 struct sk_buff *csum_skb = NULL;
548 tmp_csum = 0;
549
550 skb_queue_walk(&sk->sk_write_queue, skb) {
551 tmp_csum = csum_add(tmp_csum, skb->csum);
552
553 if (csum_skb)
554 continue;
555
556 len = skb->len - skb_transport_offset(skb);
557 if (offset >= len) {
558 offset -= len;
559 continue;
560 }
561
562 csum_skb = skb;
563 }
564
565 skb = csum_skb;
566 }
567
568 offset += skb_transport_offset(skb);
569 err = skb_copy_bits(skb, offset, &csum, 2);
570 if (err < 0) {
571 ip6_flush_pending_frames(sk);
572 goto out;
573 }
574
575 /* in case cksum was not initialized */
576 if (unlikely(csum))
577 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
578
579 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
580 total_len, fl6->flowi6_proto, tmp_csum);
581
582 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
583 csum = CSUM_MANGLED_0;
584
585 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
586
587 send:
588 err = ip6_push_pending_frames(sk);
589 out:
590 return err;
591 }
592
rawv6_send_hdrinc(struct sock * sk,struct msghdr * msg,int length,struct flowi6 * fl6,struct dst_entry ** dstp,unsigned int flags,const struct sockcm_cookie * sockc)593 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
594 struct flowi6 *fl6, struct dst_entry **dstp,
595 unsigned int flags, const struct sockcm_cookie *sockc)
596 {
597 struct net *net = sock_net(sk);
598 struct ipv6hdr *iph;
599 struct sk_buff *skb;
600 int err;
601 struct rt6_info *rt = dst_rt6_info(*dstp);
602 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
603 int tlen = rt->dst.dev->needed_tailroom;
604
605 if (length > rt->dst.dev->mtu) {
606 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
607 return -EMSGSIZE;
608 }
609 if (length < sizeof(struct ipv6hdr))
610 return -EINVAL;
611 if (flags&MSG_PROBE)
612 goto out;
613
614 skb = sock_alloc_send_skb(sk,
615 length + hlen + tlen + 15,
616 flags & MSG_DONTWAIT, &err);
617 if (!skb)
618 goto error;
619 skb_reserve(skb, hlen);
620
621 skb->protocol = htons(ETH_P_IPV6);
622 skb->priority = READ_ONCE(sk->sk_priority);
623 skb->mark = sockc->mark;
624 skb_set_delivery_type_by_clockid(skb, sockc->transmit_time, sk->sk_clockid);
625
626 skb_put(skb, length);
627 skb_reset_network_header(skb);
628 iph = ipv6_hdr(skb);
629
630 skb->ip_summed = CHECKSUM_NONE;
631
632 skb_setup_tx_timestamp(skb, sockc->tsflags);
633
634 if (flags & MSG_CONFIRM)
635 skb_set_dst_pending_confirm(skb, 1);
636
637 skb->transport_header = skb->network_header;
638 err = memcpy_from_msg(iph, msg, length);
639 if (err) {
640 err = -EFAULT;
641 kfree_skb(skb);
642 goto error;
643 }
644
645 skb_dst_set(skb, &rt->dst);
646 *dstp = NULL;
647
648 /* if egress device is enslaved to an L3 master device pass the
649 * skb to its handler for processing
650 */
651 skb = l3mdev_ip6_out(sk, skb);
652 if (unlikely(!skb))
653 return 0;
654
655 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev
656 * in the error path. Since skb has been freed, the dst could
657 * have been queued for deletion.
658 */
659 rcu_read_lock();
660 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
661 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
662 NULL, rt->dst.dev, dst_output);
663 if (err > 0)
664 err = net_xmit_errno(err);
665 if (err) {
666 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
667 rcu_read_unlock();
668 goto error_check;
669 }
670 rcu_read_unlock();
671 out:
672 return 0;
673
674 error:
675 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
676 error_check:
677 if (err == -ENOBUFS && !inet6_test_bit(RECVERR6, sk))
678 err = 0;
679 return err;
680 }
681
682 struct raw6_frag_vec {
683 struct msghdr *msg;
684 int hlen;
685 char c[4];
686 };
687
rawv6_probe_proto_opt(struct raw6_frag_vec * rfv,struct flowi6 * fl6)688 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
689 {
690 int err = 0;
691 switch (fl6->flowi6_proto) {
692 case IPPROTO_ICMPV6:
693 rfv->hlen = 2;
694 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
695 if (!err) {
696 fl6->fl6_icmp_type = rfv->c[0];
697 fl6->fl6_icmp_code = rfv->c[1];
698 }
699 break;
700 case IPPROTO_MH:
701 rfv->hlen = 4;
702 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
703 if (!err)
704 fl6->fl6_mh_type = rfv->c[2];
705 }
706 return err;
707 }
708
raw6_getfrag(void * from,char * to,int offset,int len,int odd,struct sk_buff * skb)709 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
710 struct sk_buff *skb)
711 {
712 struct raw6_frag_vec *rfv = from;
713
714 if (offset < rfv->hlen) {
715 int copy = min(rfv->hlen - offset, len);
716
717 if (skb->ip_summed == CHECKSUM_PARTIAL)
718 memcpy(to, rfv->c + offset, copy);
719 else
720 skb->csum = csum_block_add(
721 skb->csum,
722 csum_partial_copy_nocheck(rfv->c + offset,
723 to, copy),
724 odd);
725
726 odd = 0;
727 offset += copy;
728 to += copy;
729 len -= copy;
730
731 if (!len)
732 return 0;
733 }
734
735 offset -= rfv->hlen;
736
737 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
738 }
739
rawv6_sendmsg(struct sock * sk,struct msghdr * msg,size_t len)740 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
741 {
742 struct ipv6_txoptions *opt_to_free = NULL;
743 struct ipv6_txoptions opt_space;
744 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
745 struct in6_addr *daddr, *final_p, final;
746 struct inet_sock *inet = inet_sk(sk);
747 struct ipv6_pinfo *np = inet6_sk(sk);
748 struct raw6_sock *rp = raw6_sk(sk);
749 struct ipv6_txoptions *opt = NULL;
750 struct ip6_flowlabel *flowlabel = NULL;
751 struct dst_entry *dst = NULL;
752 struct raw6_frag_vec rfv;
753 struct flowi6 fl6;
754 struct ipcm6_cookie ipc6;
755 int addr_len = msg->msg_namelen;
756 int hdrincl;
757 u16 proto;
758 int err;
759
760 /* Rough check on arithmetic overflow,
761 better check is made in ip6_append_data().
762 */
763 if (len > INT_MAX)
764 return -EMSGSIZE;
765
766 /* Mirror BSD error message compatibility */
767 if (msg->msg_flags & MSG_OOB)
768 return -EOPNOTSUPP;
769
770 hdrincl = inet_test_bit(HDRINCL, sk);
771
772 /*
773 * Get and verify the address.
774 */
775 memset(&fl6, 0, sizeof(fl6));
776
777 fl6.flowi6_mark = READ_ONCE(sk->sk_mark);
778 fl6.flowi6_uid = sk->sk_uid;
779
780 ipcm6_init_sk(&ipc6, sk);
781 ipc6.sockc.tsflags = READ_ONCE(sk->sk_tsflags);
782 ipc6.sockc.mark = fl6.flowi6_mark;
783
784 if (sin6) {
785 if (addr_len < SIN6_LEN_RFC2133)
786 return -EINVAL;
787
788 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
789 return -EAFNOSUPPORT;
790
791 /* port is the proto value [0..255] carried in nexthdr */
792 proto = ntohs(sin6->sin6_port);
793
794 if (!proto)
795 proto = inet->inet_num;
796 else if (proto != inet->inet_num &&
797 inet->inet_num != IPPROTO_RAW)
798 return -EINVAL;
799
800 if (proto > 255)
801 return -EINVAL;
802
803 daddr = &sin6->sin6_addr;
804 if (inet6_test_bit(SNDFLOW, sk)) {
805 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
806 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
807 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
808 if (IS_ERR(flowlabel))
809 return -EINVAL;
810 }
811 }
812
813 /*
814 * Otherwise it will be difficult to maintain
815 * sk->sk_dst_cache.
816 */
817 if (sk->sk_state == TCP_ESTABLISHED &&
818 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
819 daddr = &sk->sk_v6_daddr;
820
821 if (addr_len >= sizeof(struct sockaddr_in6) &&
822 sin6->sin6_scope_id &&
823 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
824 fl6.flowi6_oif = sin6->sin6_scope_id;
825 } else {
826 if (sk->sk_state != TCP_ESTABLISHED)
827 return -EDESTADDRREQ;
828
829 proto = inet->inet_num;
830 daddr = &sk->sk_v6_daddr;
831 fl6.flowlabel = np->flow_label;
832 }
833
834 if (fl6.flowi6_oif == 0)
835 fl6.flowi6_oif = sk->sk_bound_dev_if;
836
837 if (msg->msg_controllen) {
838 opt = &opt_space;
839 memset(opt, 0, sizeof(struct ipv6_txoptions));
840 opt->tot_len = sizeof(struct ipv6_txoptions);
841 ipc6.opt = opt;
842
843 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
844 if (err < 0) {
845 fl6_sock_release(flowlabel);
846 return err;
847 }
848 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
849 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
850 if (IS_ERR(flowlabel))
851 return -EINVAL;
852 }
853 if (!(opt->opt_nflen|opt->opt_flen))
854 opt = NULL;
855 }
856 if (!opt) {
857 opt = txopt_get(np);
858 opt_to_free = opt;
859 }
860 if (flowlabel)
861 opt = fl6_merge_options(&opt_space, flowlabel, opt);
862 opt = ipv6_fixup_options(&opt_space, opt);
863
864 fl6.flowi6_proto = proto;
865 fl6.flowi6_mark = ipc6.sockc.mark;
866
867 if (!hdrincl) {
868 rfv.msg = msg;
869 rfv.hlen = 0;
870 err = rawv6_probe_proto_opt(&rfv, &fl6);
871 if (err)
872 goto out;
873 }
874
875 if (!ipv6_addr_any(daddr))
876 fl6.daddr = *daddr;
877 else
878 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
879 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
880 fl6.saddr = np->saddr;
881
882 final_p = fl6_update_dst(&fl6, opt, &final);
883
884 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
885 fl6.flowi6_oif = READ_ONCE(np->mcast_oif);
886 else if (!fl6.flowi6_oif)
887 fl6.flowi6_oif = READ_ONCE(np->ucast_oif);
888 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
889
890 if (hdrincl)
891 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
892
893 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
894
895 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
896 if (IS_ERR(dst)) {
897 err = PTR_ERR(dst);
898 goto out;
899 }
900 if (ipc6.hlimit < 0)
901 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
902
903 if (msg->msg_flags&MSG_CONFIRM)
904 goto do_confirm;
905
906 back_from_confirm:
907 if (hdrincl)
908 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst,
909 msg->msg_flags, &ipc6.sockc);
910 else {
911 ipc6.opt = opt;
912 lock_sock(sk);
913 err = ip6_append_data(sk, raw6_getfrag, &rfv,
914 len, 0, &ipc6, &fl6, dst_rt6_info(dst),
915 msg->msg_flags);
916
917 if (err)
918 ip6_flush_pending_frames(sk);
919 else if (!(msg->msg_flags & MSG_MORE))
920 err = rawv6_push_pending_frames(sk, &fl6, rp);
921 release_sock(sk);
922 }
923 done:
924 dst_release(dst);
925 out:
926 fl6_sock_release(flowlabel);
927 txopt_put(opt_to_free);
928 return err < 0 ? err : len;
929 do_confirm:
930 if (msg->msg_flags & MSG_PROBE)
931 dst_confirm_neigh(dst, &fl6.daddr);
932 if (!(msg->msg_flags & MSG_PROBE) || len)
933 goto back_from_confirm;
934 err = 0;
935 goto done;
936 }
937
rawv6_seticmpfilter(struct sock * sk,int optname,sockptr_t optval,int optlen)938 static int rawv6_seticmpfilter(struct sock *sk, int optname,
939 sockptr_t optval, int optlen)
940 {
941 switch (optname) {
942 case ICMPV6_FILTER:
943 if (optlen > sizeof(struct icmp6_filter))
944 optlen = sizeof(struct icmp6_filter);
945 if (copy_from_sockptr(&raw6_sk(sk)->filter, optval, optlen))
946 return -EFAULT;
947 return 0;
948 default:
949 return -ENOPROTOOPT;
950 }
951
952 return 0;
953 }
954
rawv6_geticmpfilter(struct sock * sk,int optname,char __user * optval,int __user * optlen)955 static int rawv6_geticmpfilter(struct sock *sk, int optname,
956 char __user *optval, int __user *optlen)
957 {
958 int len;
959
960 switch (optname) {
961 case ICMPV6_FILTER:
962 if (get_user(len, optlen))
963 return -EFAULT;
964 if (len < 0)
965 return -EINVAL;
966 if (len > sizeof(struct icmp6_filter))
967 len = sizeof(struct icmp6_filter);
968 if (put_user(len, optlen))
969 return -EFAULT;
970 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
971 return -EFAULT;
972 return 0;
973 default:
974 return -ENOPROTOOPT;
975 }
976
977 return 0;
978 }
979
980
do_rawv6_setsockopt(struct sock * sk,int level,int optname,sockptr_t optval,unsigned int optlen)981 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
982 sockptr_t optval, unsigned int optlen)
983 {
984 struct raw6_sock *rp = raw6_sk(sk);
985 int val;
986
987 if (optlen < sizeof(val))
988 return -EINVAL;
989
990 if (copy_from_sockptr(&val, optval, sizeof(val)))
991 return -EFAULT;
992
993 switch (optname) {
994 case IPV6_HDRINCL:
995 if (sk->sk_type != SOCK_RAW)
996 return -EINVAL;
997 inet_assign_bit(HDRINCL, sk, val);
998 return 0;
999 case IPV6_CHECKSUM:
1000 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1001 level == IPPROTO_IPV6) {
1002 /*
1003 * RFC3542 tells that IPV6_CHECKSUM socket
1004 * option in the IPPROTO_IPV6 level is not
1005 * allowed on ICMPv6 sockets.
1006 * If you want to set it, use IPPROTO_RAW
1007 * level IPV6_CHECKSUM socket option
1008 * (Linux extension).
1009 */
1010 return -EINVAL;
1011 }
1012
1013 /* You may get strange result with a positive odd offset;
1014 RFC2292bis agrees with me. */
1015 if (val > 0 && (val&1))
1016 return -EINVAL;
1017 if (val < 0) {
1018 rp->checksum = 0;
1019 } else {
1020 rp->checksum = 1;
1021 rp->offset = val;
1022 }
1023
1024 return 0;
1025
1026 default:
1027 return -ENOPROTOOPT;
1028 }
1029 }
1030
rawv6_setsockopt(struct sock * sk,int level,int optname,sockptr_t optval,unsigned int optlen)1031 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1032 sockptr_t optval, unsigned int optlen)
1033 {
1034 switch (level) {
1035 case SOL_RAW:
1036 break;
1037
1038 case SOL_ICMPV6:
1039 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1040 return -EOPNOTSUPP;
1041 return rawv6_seticmpfilter(sk, optname, optval, optlen);
1042 case SOL_IPV6:
1043 if (optname == IPV6_CHECKSUM ||
1044 optname == IPV6_HDRINCL)
1045 break;
1046 fallthrough;
1047 default:
1048 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1049 }
1050
1051 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1052 }
1053
do_rawv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1054 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1055 char __user *optval, int __user *optlen)
1056 {
1057 struct raw6_sock *rp = raw6_sk(sk);
1058 int val, len;
1059
1060 if (get_user(len, optlen))
1061 return -EFAULT;
1062
1063 switch (optname) {
1064 case IPV6_HDRINCL:
1065 val = inet_test_bit(HDRINCL, sk);
1066 break;
1067 case IPV6_CHECKSUM:
1068 /*
1069 * We allow getsockopt() for IPPROTO_IPV6-level
1070 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1071 * since RFC3542 is silent about it.
1072 */
1073 if (rp->checksum == 0)
1074 val = -1;
1075 else
1076 val = rp->offset;
1077 break;
1078
1079 default:
1080 return -ENOPROTOOPT;
1081 }
1082
1083 len = min_t(unsigned int, sizeof(int), len);
1084
1085 if (put_user(len, optlen))
1086 return -EFAULT;
1087 if (copy_to_user(optval, &val, len))
1088 return -EFAULT;
1089 return 0;
1090 }
1091
rawv6_getsockopt(struct sock * sk,int level,int optname,char __user * optval,int __user * optlen)1092 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1093 char __user *optval, int __user *optlen)
1094 {
1095 switch (level) {
1096 case SOL_RAW:
1097 break;
1098
1099 case SOL_ICMPV6:
1100 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1101 return -EOPNOTSUPP;
1102 return rawv6_geticmpfilter(sk, optname, optval, optlen);
1103 case SOL_IPV6:
1104 if (optname == IPV6_CHECKSUM ||
1105 optname == IPV6_HDRINCL)
1106 break;
1107 fallthrough;
1108 default:
1109 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1110 }
1111
1112 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1113 }
1114
rawv6_ioctl(struct sock * sk,int cmd,int * karg)1115 static int rawv6_ioctl(struct sock *sk, int cmd, int *karg)
1116 {
1117 switch (cmd) {
1118 case SIOCOUTQ: {
1119 *karg = sk_wmem_alloc_get(sk);
1120 return 0;
1121 }
1122 case SIOCINQ: {
1123 struct sk_buff *skb;
1124
1125 spin_lock_bh(&sk->sk_receive_queue.lock);
1126 skb = skb_peek(&sk->sk_receive_queue);
1127 if (skb)
1128 *karg = skb->len;
1129 else
1130 *karg = 0;
1131 spin_unlock_bh(&sk->sk_receive_queue.lock);
1132 return 0;
1133 }
1134
1135 default:
1136 #ifdef CONFIG_IPV6_MROUTE
1137 return ip6mr_ioctl(sk, cmd, karg);
1138 #else
1139 return -ENOIOCTLCMD;
1140 #endif
1141 }
1142 }
1143
1144 #ifdef CONFIG_COMPAT
compat_rawv6_ioctl(struct sock * sk,unsigned int cmd,unsigned long arg)1145 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1146 {
1147 switch (cmd) {
1148 case SIOCOUTQ:
1149 case SIOCINQ:
1150 return -ENOIOCTLCMD;
1151 default:
1152 #ifdef CONFIG_IPV6_MROUTE
1153 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1154 #else
1155 return -ENOIOCTLCMD;
1156 #endif
1157 }
1158 }
1159 #endif
1160
rawv6_close(struct sock * sk,long timeout)1161 static void rawv6_close(struct sock *sk, long timeout)
1162 {
1163 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1164 ip6_ra_control(sk, -1);
1165 ip6mr_sk_done(sk);
1166 sk_common_release(sk);
1167 }
1168
raw6_destroy(struct sock * sk)1169 static void raw6_destroy(struct sock *sk)
1170 {
1171 lock_sock(sk);
1172 ip6_flush_pending_frames(sk);
1173 release_sock(sk);
1174 }
1175
rawv6_init_sk(struct sock * sk)1176 static int rawv6_init_sk(struct sock *sk)
1177 {
1178 struct raw6_sock *rp = raw6_sk(sk);
1179
1180 switch (inet_sk(sk)->inet_num) {
1181 case IPPROTO_ICMPV6:
1182 rp->checksum = 1;
1183 rp->offset = 2;
1184 break;
1185 case IPPROTO_MH:
1186 rp->checksum = 1;
1187 rp->offset = 4;
1188 break;
1189 default:
1190 break;
1191 }
1192 return 0;
1193 }
1194
1195 struct proto rawv6_prot = {
1196 .name = "RAWv6",
1197 .owner = THIS_MODULE,
1198 .close = rawv6_close,
1199 .destroy = raw6_destroy,
1200 .connect = ip6_datagram_connect_v6_only,
1201 .disconnect = __udp_disconnect,
1202 .ioctl = rawv6_ioctl,
1203 .init = rawv6_init_sk,
1204 .setsockopt = rawv6_setsockopt,
1205 .getsockopt = rawv6_getsockopt,
1206 .sendmsg = rawv6_sendmsg,
1207 .recvmsg = rawv6_recvmsg,
1208 .bind = rawv6_bind,
1209 .backlog_rcv = rawv6_rcv_skb,
1210 .hash = raw_hash_sk,
1211 .unhash = raw_unhash_sk,
1212 .obj_size = sizeof(struct raw6_sock),
1213 .ipv6_pinfo_offset = offsetof(struct raw6_sock, inet6),
1214 .useroffset = offsetof(struct raw6_sock, filter),
1215 .usersize = sizeof_field(struct raw6_sock, filter),
1216 .h.raw_hash = &raw_v6_hashinfo,
1217 #ifdef CONFIG_COMPAT
1218 .compat_ioctl = compat_rawv6_ioctl,
1219 #endif
1220 .diag_destroy = raw_abort,
1221 };
1222
1223 #ifdef CONFIG_PROC_FS
raw6_seq_show(struct seq_file * seq,void * v)1224 static int raw6_seq_show(struct seq_file *seq, void *v)
1225 {
1226 if (v == SEQ_START_TOKEN) {
1227 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1228 } else {
1229 struct sock *sp = v;
1230 __u16 srcp = inet_sk(sp)->inet_num;
1231 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1232 raw_seq_private(seq)->bucket);
1233 }
1234 return 0;
1235 }
1236
1237 static const struct seq_operations raw6_seq_ops = {
1238 .start = raw_seq_start,
1239 .next = raw_seq_next,
1240 .stop = raw_seq_stop,
1241 .show = raw6_seq_show,
1242 };
1243
raw6_init_net(struct net * net)1244 static int __net_init raw6_init_net(struct net *net)
1245 {
1246 if (!proc_create_net_data("raw6", 0444, net->proc_net, &raw6_seq_ops,
1247 sizeof(struct raw_iter_state), &raw_v6_hashinfo))
1248 return -ENOMEM;
1249
1250 return 0;
1251 }
1252
raw6_exit_net(struct net * net)1253 static void __net_exit raw6_exit_net(struct net *net)
1254 {
1255 remove_proc_entry("raw6", net->proc_net);
1256 }
1257
1258 static struct pernet_operations raw6_net_ops = {
1259 .init = raw6_init_net,
1260 .exit = raw6_exit_net,
1261 };
1262
raw6_proc_init(void)1263 int __init raw6_proc_init(void)
1264 {
1265 return register_pernet_subsys(&raw6_net_ops);
1266 }
1267
raw6_proc_exit(void)1268 void raw6_proc_exit(void)
1269 {
1270 unregister_pernet_subsys(&raw6_net_ops);
1271 }
1272 #endif /* CONFIG_PROC_FS */
1273
1274 /* Same as inet6_dgram_ops, sans udp_poll. */
1275 const struct proto_ops inet6_sockraw_ops = {
1276 .family = PF_INET6,
1277 .owner = THIS_MODULE,
1278 .release = inet6_release,
1279 .bind = inet6_bind,
1280 .connect = inet_dgram_connect, /* ok */
1281 .socketpair = sock_no_socketpair, /* a do nothing */
1282 .accept = sock_no_accept, /* a do nothing */
1283 .getname = inet6_getname,
1284 .poll = datagram_poll, /* ok */
1285 .ioctl = inet6_ioctl, /* must change */
1286 .gettstamp = sock_gettstamp,
1287 .listen = sock_no_listen, /* ok */
1288 .shutdown = inet_shutdown, /* ok */
1289 .setsockopt = sock_common_setsockopt, /* ok */
1290 .getsockopt = sock_common_getsockopt, /* ok */
1291 .sendmsg = inet_sendmsg, /* ok */
1292 .recvmsg = sock_common_recvmsg, /* ok */
1293 .mmap = sock_no_mmap,
1294 #ifdef CONFIG_COMPAT
1295 .compat_ioctl = inet6_compat_ioctl,
1296 #endif
1297 };
1298
1299 static struct inet_protosw rawv6_protosw = {
1300 .type = SOCK_RAW,
1301 .protocol = IPPROTO_IP, /* wild card */
1302 .prot = &rawv6_prot,
1303 .ops = &inet6_sockraw_ops,
1304 .flags = INET_PROTOSW_REUSE,
1305 };
1306
rawv6_init(void)1307 int __init rawv6_init(void)
1308 {
1309 return inet6_register_protosw(&rawv6_protosw);
1310 }
1311
rawv6_exit(void)1312 void rawv6_exit(void)
1313 {
1314 inet6_unregister_protosw(&rawv6_protosw);
1315 }
1316