]> git.ipfire.org Git - people/arne_f/kernel.git/blob - net/ipv6/raw.c
projects / people / arne_f / kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge tag 'v4.9.32' into linux-4.9.x-grsecurity-3.1
[people/arne_f/kernel.git] / net / ipv6 / raw.c
1 /*
2 * RAW sockets for IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Adapted from linux/net/ipv4/raw.c
9 *
10 * Fixes:
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
13 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
14 *
15 * This program is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU General Public License
17 * as published by the Free Software Foundation; either version
18 * 2 of the License, or (at your option) any later version.
19 */
20
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/slab.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/netfilter.h>
32 #include <linux/netfilter_ipv6.h>
33 #include <linux/skbuff.h>
34 #include <linux/compat.h>
35 #include <linux/uaccess.h>
36 #include <asm/ioctls.h>
37
38 #include <net/net_namespace.h>
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
42
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
50 #include <net/udp.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #if IS_ENABLED(CONFIG_IPV6_MIP6)
54 #include <net/mip6.h>
55 #endif
56 #include <linux/mroute6.h>
57
58 #include <net/raw.h>
59 #include <net/rawv6.h>
60 #include <net/xfrm.h>
61
62 #include <linux/proc_fs.h>
63 #include <linux/seq_file.h>
64 #include <linux/export.h>
65
66 #define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */
67
68 static struct raw_hashinfo raw_v6_hashinfo = {
69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
70 };
71
72 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
73 unsigned short num, const struct in6_addr *loc_addr,
74 const struct in6_addr *rmt_addr, int dif)
75 {
76 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
77
78 sk_for_each_from(sk)
79 if (inet_sk(sk)->inet_num == num) {
80
81 if (!net_eq(sock_net(sk), net))
82 continue;
83
84 if (!ipv6_addr_any(&sk->sk_v6_daddr) &&
85 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr))
86 continue;
87
88 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
89 continue;
90
91 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
92 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))
93 goto found;
94 if (is_multicast &&
95 inet6_mc_check(sk, loc_addr, rmt_addr))
96 goto found;
97 continue;
98 }
99 goto found;
100 }
101 sk = NULL;
102 found:
103 return sk;
104 }
105
106 /*
107 * 0 - deliver
108 * 1 - block
109 */
110 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
111 {
112 struct icmp6hdr _hdr;
113 const struct icmp6hdr *hdr;
114
115 /* We require only the four bytes of the ICMPv6 header, not any
116 * additional bytes of message body in "struct icmp6hdr".
117 */
118 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
119 ICMPV6_HDRLEN, &_hdr);
120 if (hdr) {
121 const __u32 *data = &raw6_sk(sk)->filter.data[0];
122 unsigned int type = hdr->icmp6_type;
123
124 return (data[type >> 5] & (1U << (type & 31))) != 0;
125 }
126 return 1;
127 }
128
129 #if IS_ENABLED(CONFIG_IPV6_MIP6)
130 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
131
132 static mh_filter_t __rcu *mh_filter __read_mostly;
133
134 int rawv6_mh_filter_register(mh_filter_t filter)
135 {
136 rcu_assign_pointer(mh_filter, filter);
137 return 0;
138 }
139 EXPORT_SYMBOL(rawv6_mh_filter_register);
140
141 int rawv6_mh_filter_unregister(mh_filter_t filter)
142 {
143 RCU_INIT_POINTER(mh_filter, NULL);
144 synchronize_rcu();
145 return 0;
146 }
147 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
148
149 #endif
150
151 /*
152 * demultiplex raw sockets.
153 * (should consider queueing the skb in the sock receive_queue
154 * without calling rawv6.c)
155 *
156 * Caller owns SKB so we must make clones.
157 */
158 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
159 {
160 const struct in6_addr *saddr;
161 const struct in6_addr *daddr;
162 struct sock *sk;
163 bool delivered = false;
164 __u8 hash;
165 struct net *net;
166
167 saddr = &ipv6_hdr(skb)->saddr;
168 daddr = saddr + 1;
169
170 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
171
172 read_lock(&raw_v6_hashinfo.lock);
173 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
174
175 if (!sk)
176 goto out;
177
178 net = dev_net(skb->dev);
179 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, inet6_iif(skb));
180
181 while (sk) {
182 int filtered;
183
184 delivered = true;
185 switch (nexthdr) {
186 case IPPROTO_ICMPV6:
187 filtered = icmpv6_filter(sk, skb);
188 break;
189
190 #if IS_ENABLED(CONFIG_IPV6_MIP6)
191 case IPPROTO_MH:
192 {
193 /* XXX: To validate MH only once for each packet,
194 * this is placed here. It should be after checking
195 * xfrm policy, however it doesn't. The checking xfrm
196 * policy is placed in rawv6_rcv() because it is
197 * required for each socket.
198 */
199 mh_filter_t *filter;
200
201 filter = rcu_dereference(mh_filter);
202 filtered = filter ? (*filter)(sk, skb) : 0;
203 break;
204 }
205 #endif
206 default:
207 filtered = 0;
208 break;
209 }
210
211 if (filtered < 0)
212 break;
213 if (filtered == 0) {
214 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
215
216 /* Not releasing hash table! */
217 if (clone) {
218 nf_reset(clone);
219 rawv6_rcv(sk, clone);
220 }
221 }
222 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
223 inet6_iif(skb));
224 }
225 out:
226 read_unlock(&raw_v6_hashinfo.lock);
227 return delivered;
228 }
229
230 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
231 {
232 struct sock *raw_sk;
233
234 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
235 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
236 raw_sk = NULL;
237
238 return raw_sk != NULL;
239 }
240
241 /* This cleans up af_inet6 a bit. -DaveM */
242 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
243 {
244 struct inet_sock *inet = inet_sk(sk);
245 struct ipv6_pinfo *np = inet6_sk(sk);
246 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
247 __be32 v4addr = 0;
248 int addr_type;
249 int err;
250
251 if (addr_len < SIN6_LEN_RFC2133)
252 return -EINVAL;
253
254 if (addr->sin6_family != AF_INET6)
255 return -EINVAL;
256
257 addr_type = ipv6_addr_type(&addr->sin6_addr);
258
259 /* Raw sockets are IPv6 only */
260 if (addr_type == IPV6_ADDR_MAPPED)
261 return -EADDRNOTAVAIL;
262
263 lock_sock(sk);
264
265 err = -EINVAL;
266 if (sk->sk_state != TCP_CLOSE)
267 goto out;
268
269 rcu_read_lock();
270 /* Check if the address belongs to the host. */
271 if (addr_type != IPV6_ADDR_ANY) {
272 struct net_device *dev = NULL;
273
274 if (__ipv6_addr_needs_scope_id(addr_type)) {
275 if (addr_len >= sizeof(struct sockaddr_in6) &&
276 addr->sin6_scope_id) {
277 /* Override any existing binding, if another
278 * one is supplied by user.
279 */
280 sk->sk_bound_dev_if = addr->sin6_scope_id;
281 }
282
283 /* Binding to link-local address requires an interface */
284 if (!sk->sk_bound_dev_if)
285 goto out_unlock;
286
287 err = -ENODEV;
288 dev = dev_get_by_index_rcu(sock_net(sk),
289 sk->sk_bound_dev_if);
290 if (!dev)
291 goto out_unlock;
292 }
293
294 /* ipv4 addr of the socket is invalid. Only the
295 * unspecified and mapped address have a v4 equivalent.
296 */
297 v4addr = LOOPBACK4_IPV6;
298 if (!(addr_type & IPV6_ADDR_MULTICAST) &&
299 !sock_net(sk)->ipv6.sysctl.ip_nonlocal_bind) {
300 err = -EADDRNOTAVAIL;
301 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
302 dev, 0)) {
303 goto out_unlock;
304 }
305 }
306 }
307
308 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
309 sk->sk_v6_rcv_saddr = addr->sin6_addr;
310 if (!(addr_type & IPV6_ADDR_MULTICAST))
311 np->saddr = addr->sin6_addr;
312 err = 0;
313 out_unlock:
314 rcu_read_unlock();
315 out:
316 release_sock(sk);
317 return err;
318 }
319
320 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
321 struct inet6_skb_parm *opt,
322 u8 type, u8 code, int offset, __be32 info)
323 {
324 struct inet_sock *inet = inet_sk(sk);
325 struct ipv6_pinfo *np = inet6_sk(sk);
326 int err;
327 int harderr;
328
329 /* Report error on raw socket, if:
330 1. User requested recverr.
331 2. Socket is connected (otherwise the error indication
332 is useless without recverr and error is hard.
333 */
334 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
335 return;
336
337 harderr = icmpv6_err_convert(type, code, &err);
338 if (type == ICMPV6_PKT_TOOBIG) {
339 ip6_sk_update_pmtu(skb, sk, info);
340 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
341 }
342 if (type == NDISC_REDIRECT) {
343 ip6_sk_redirect(skb, sk);
344 return;
345 }
346 if (np->recverr) {
347 u8 *payload = skb->data;
348 if (!inet->hdrincl)
349 payload += offset;
350 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
351 }
352
353 if (np->recverr || harderr) {
354 sk->sk_err = err;
355 sk->sk_error_report(sk);
356 }
357 }
358
359 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
360 u8 type, u8 code, int inner_offset, __be32 info)
361 {
362 struct sock *sk;
363 int hash;
364 const struct in6_addr *saddr, *daddr;
365 struct net *net;
366
367 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
368
369 read_lock(&raw_v6_hashinfo.lock);
370 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
371 if (sk) {
372 /* Note: ipv6_hdr(skb) != skb->data */
373 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
374 saddr = &ip6h->saddr;
375 daddr = &ip6h->daddr;
376 net = dev_net(skb->dev);
377
378 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
379 inet6_iif(skb)))) {
380 rawv6_err(sk, skb, NULL, type, code,
381 inner_offset, info);
382 sk = sk_next(sk);
383 }
384 }
385 read_unlock(&raw_v6_hashinfo.lock);
386 }
387
388 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
389 {
390 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
391 skb_checksum_complete(skb)) {
392 atomic_inc_unchecked(&sk->sk_drops);
393 kfree_skb(skb);
394 return NET_RX_DROP;
395 }
396
397 /* Charge it to the socket. */
398 skb_dst_drop(skb);
399 if (sock_queue_rcv_skb(sk, skb) < 0) {
400 kfree_skb(skb);
401 return NET_RX_DROP;
402 }
403
404 return 0;
405 }
406
407 /*
408 * This is next to useless...
409 * if we demultiplex in network layer we don't need the extra call
410 * just to queue the skb...
411 * maybe we could have the network decide upon a hint if it
412 * should call raw_rcv for demultiplexing
413 */
414 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
415 {
416 struct inet_sock *inet = inet_sk(sk);
417 struct raw6_sock *rp = raw6_sk(sk);
418
419 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
420 atomic_inc_unchecked(&sk->sk_drops);
421 kfree_skb(skb);
422 return NET_RX_DROP;
423 }
424
425 if (!rp->checksum)
426 skb->ip_summed = CHECKSUM_UNNECESSARY;
427
428 if (skb->ip_summed == CHECKSUM_COMPLETE) {
429 skb_postpull_rcsum(skb, skb_network_header(skb),
430 skb_network_header_len(skb));
431 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
432 &ipv6_hdr(skb)->daddr,
433 skb->len, inet->inet_num, skb->csum))
434 skb->ip_summed = CHECKSUM_UNNECESSARY;
435 }
436 if (!skb_csum_unnecessary(skb))
437 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
438 &ipv6_hdr(skb)->daddr,
439 skb->len,
440 inet->inet_num, 0));
441
442 if (inet->hdrincl) {
443 if (skb_checksum_complete(skb)) {
444 atomic_inc_unchecked(&sk->sk_drops);
445 kfree_skb(skb);
446 return NET_RX_DROP;
447 }
448 }
449
450 rawv6_rcv_skb(sk, skb);
451 return 0;
452 }
453
454
455 /*
456 * This should be easy, if there is something there
457 * we return it, otherwise we block.
458 */
459
460 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
461 int noblock, int flags, int *addr_len)
462 {
463 struct ipv6_pinfo *np = inet6_sk(sk);
464 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
465 struct sk_buff *skb;
466 size_t copied;
467 int err;
468
469 if (flags & MSG_OOB)
470 return -EOPNOTSUPP;
471
472 if (flags & MSG_ERRQUEUE)
473 return ipv6_recv_error(sk, msg, len, addr_len);
474
475 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
476 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
477
478 skb = skb_recv_datagram(sk, flags, noblock, &err);
479 if (!skb)
480 goto out;
481
482 copied = skb->len;
483 if (copied > len) {
484 copied = len;
485 msg->msg_flags |= MSG_TRUNC;
486 }
487
488 if (skb_csum_unnecessary(skb)) {
489 err = skb_copy_datagram_msg(skb, 0, msg, copied);
490 } else if (msg->msg_flags&MSG_TRUNC) {
491 if (__skb_checksum_complete(skb))
492 goto csum_copy_err;
493 err = skb_copy_datagram_msg(skb, 0, msg, copied);
494 } else {
495 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
496 if (err == -EINVAL)
497 goto csum_copy_err;
498 }
499 if (err)
500 goto out_free;
501
502 /* Copy the address. */
503 if (sin6) {
504 sin6->sin6_family = AF_INET6;
505 sin6->sin6_port = 0;
506 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
507 sin6->sin6_flowinfo = 0;
508 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
509 inet6_iif(skb));
510 *addr_len = sizeof(*sin6);
511 }
512
513 sock_recv_ts_and_drops(msg, sk, skb);
514
515 if (np->rxopt.all)
516 ip6_datagram_recv_ctl(sk, msg, skb);
517
518 err = copied;
519 if (flags & MSG_TRUNC)
520 err = skb->len;
521
522 out_free:
523 skb_free_datagram(sk, skb);
524 out:
525 return err;
526
527 csum_copy_err:
528 skb_kill_datagram(sk, skb, flags);
529
530 /* Error for blocking case is chosen to masquerade
531 as some normal condition.
532 */
533 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
534 goto out;
535 }
536
537 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
538 struct raw6_sock *rp)
539 {
540 struct sk_buff *skb;
541 int err = 0;
542 int offset;
543 int len;
544 int total_len;
545 __wsum tmp_csum;
546 __sum16 csum;
547
548 if (!rp->checksum)
549 goto send;
550
551 skb = skb_peek(&sk->sk_write_queue);
552 if (!skb)
553 goto out;
554
555 offset = rp->offset;
556 total_len = inet_sk(sk)->cork.base.length;
557 if (offset >= total_len - 1) {
558 err = -EINVAL;
559 ip6_flush_pending_frames(sk);
560 goto out;
561 }
562
563 /* should be check HW csum miyazawa */
564 if (skb_queue_len(&sk->sk_write_queue) == 1) {
565 /*
566 * Only one fragment on the socket.
567 */
568 tmp_csum = skb->csum;
569 } else {
570 struct sk_buff *csum_skb = NULL;
571 tmp_csum = 0;
572
573 skb_queue_walk(&sk->sk_write_queue, skb) {
574 tmp_csum = csum_add(tmp_csum, skb->csum);
575
576 if (csum_skb)
577 continue;
578
579 len = skb->len - skb_transport_offset(skb);
580 if (offset >= len) {
581 offset -= len;
582 continue;
583 }
584
585 csum_skb = skb;
586 }
587
588 skb = csum_skb;
589 }
590
591 offset += skb_transport_offset(skb);
592 err = skb_copy_bits(skb, offset, &csum, 2);
593 if (err < 0) {
594 ip6_flush_pending_frames(sk);
595 goto out;
596 }
597
598 /* in case cksum was not initialized */
599 if (unlikely(csum))
600 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
601
602 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
603 total_len, fl6->flowi6_proto, tmp_csum);
604
605 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
606 csum = CSUM_MANGLED_0;
607
608 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
609
610 send:
611 err = ip6_push_pending_frames(sk);
612 out:
613 return err;
614 }
615
616 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, unsigned int length,
617 struct flowi6 *fl6, struct dst_entry **dstp,
618 unsigned int flags)
619 {
620 struct ipv6_pinfo *np = inet6_sk(sk);
621 struct net *net = sock_net(sk);
622 struct ipv6hdr *iph;
623 struct sk_buff *skb;
624 int err;
625 struct rt6_info *rt = (struct rt6_info *)*dstp;
626 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
627 int tlen = rt->dst.dev->needed_tailroom;
628
629 if (length > rt->dst.dev->mtu) {
630 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
631 return -EMSGSIZE;
632 }
633 if (length < sizeof(struct ipv6hdr))
634 return -EINVAL;
635 if (flags&MSG_PROBE)
636 goto out;
637
638 skb = sock_alloc_send_skb(sk,
639 length + hlen + tlen + 15,
640 flags & MSG_DONTWAIT, &err);
641 if (!skb)
642 goto error;
643 skb_reserve(skb, hlen);
644
645 skb->protocol = htons(ETH_P_IPV6);
646 skb->priority = sk->sk_priority;
647 skb->mark = sk->sk_mark;
648 skb_dst_set(skb, &rt->dst);
649 *dstp = NULL;
650
651 skb_put(skb, length);
652 skb_reset_network_header(skb);
653 iph = ipv6_hdr(skb);
654
655 skb->ip_summed = CHECKSUM_NONE;
656
657 skb->transport_header = skb->network_header;
658 err = memcpy_from_msg(iph, msg, length);
659 if (err)
660 goto error_fault;
661
662 /* if egress device is enslaved to an L3 master device pass the
663 * skb to its handler for processing
664 */
665 skb = l3mdev_ip6_out(sk, skb);
666 if (unlikely(!skb))
667 return 0;
668
669 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
670 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
671 NULL, rt->dst.dev, dst_output);
672 if (err > 0)
673 err = net_xmit_errno(err);
674 if (err)
675 goto error;
676 out:
677 return 0;
678
679 error_fault:
680 err = -EFAULT;
681 kfree_skb(skb);
682 error:
683 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
684 if (err == -ENOBUFS && !np->recverr)
685 err = 0;
686 return err;
687 }
688
689 struct raw6_frag_vec {
690 struct msghdr *msg;
691 int hlen;
692 char c[4];
693 };
694
695 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
696 {
697 int err = 0;
698 switch (fl6->flowi6_proto) {
699 case IPPROTO_ICMPV6:
700 rfv->hlen = 2;
701 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
702 if (!err) {
703 fl6->fl6_icmp_type = rfv->c[0];
704 fl6->fl6_icmp_code = rfv->c[1];
705 }
706 break;
707 case IPPROTO_MH:
708 rfv->hlen = 4;
709 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
710 if (!err)
711 fl6->fl6_mh_type = rfv->c[2];
712 }
713 return err;
714 }
715
716 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
717 struct sk_buff *skb)
718 {
719 struct raw6_frag_vec *rfv = from;
720
721 if (offset < rfv->hlen) {
722 int copy = min(rfv->hlen - offset, len);
723
724 if (skb->ip_summed == CHECKSUM_PARTIAL)
725 memcpy(to, rfv->c + offset, copy);
726 else
727 skb->csum = csum_block_add(
728 skb->csum,
729 csum_partial_copy_nocheck(rfv->c + offset,
730 to, copy, 0),
731 odd);
732
733 odd = 0;
734 offset += copy;
735 to += copy;
736 len -= copy;
737
738 if (!len)
739 return 0;
740 }
741
742 offset -= rfv->hlen;
743
744 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
745 }
746
747 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
748 {
749 struct ipv6_txoptions *opt_to_free = NULL;
750 struct ipv6_txoptions opt_space;
751 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
752 struct in6_addr *daddr, *final_p, final;
753 struct inet_sock *inet = inet_sk(sk);
754 struct ipv6_pinfo *np = inet6_sk(sk);
755 struct raw6_sock *rp = raw6_sk(sk);
756 struct ipv6_txoptions *opt = NULL;
757 struct ip6_flowlabel *flowlabel = NULL;
758 struct dst_entry *dst = NULL;
759 struct raw6_frag_vec rfv;
760 struct flowi6 fl6;
761 struct sockcm_cookie sockc;
762 struct ipcm6_cookie ipc6;
763 int addr_len = msg->msg_namelen;
764 u16 proto;
765 int err;
766
767 /* Rough check on arithmetic overflow,
768 better check is made in ip6_append_data().
769 */
770 if (len > INT_MAX)
771 return -EMSGSIZE;
772
773 /* Mirror BSD error message compatibility */
774 if (msg->msg_flags & MSG_OOB)
775 return -EOPNOTSUPP;
776
777 /*
778 * Get and verify the address.
779 */
780 memset(&fl6, 0, sizeof(fl6));
781
782 fl6.flowi6_mark = sk->sk_mark;
783
784 ipc6.hlimit = -1;
785 ipc6.tclass = -1;
786 ipc6.dontfrag = -1;
787 ipc6.opt = NULL;
788
789 if (sin6) {
790 if (addr_len < SIN6_LEN_RFC2133)
791 return -EINVAL;
792
793 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
794 return -EAFNOSUPPORT;
795
796 /* port is the proto value [0..255] carried in nexthdr */
797 proto = ntohs(sin6->sin6_port);
798
799 if (!proto)
800 proto = inet->inet_num;
801 else if (proto != inet->inet_num)
802 return -EINVAL;
803
804 if (proto > 255)
805 return -EINVAL;
806
807 daddr = &sin6->sin6_addr;
808 if (np->sndflow) {
809 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
810 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
811 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
812 if (!flowlabel)
813 return -EINVAL;
814 }
815 }
816
817 /*
818 * Otherwise it will be difficult to maintain
819 * sk->sk_dst_cache.
820 */
821 if (sk->sk_state == TCP_ESTABLISHED &&
822 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
823 daddr = &sk->sk_v6_daddr;
824
825 if (addr_len >= sizeof(struct sockaddr_in6) &&
826 sin6->sin6_scope_id &&
827 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
828 fl6.flowi6_oif = sin6->sin6_scope_id;
829 } else {
830 if (sk->sk_state != TCP_ESTABLISHED)
831 return -EDESTADDRREQ;
832
833 proto = inet->inet_num;
834 daddr = &sk->sk_v6_daddr;
835 fl6.flowlabel = np->flow_label;
836 }
837
838 if (fl6.flowi6_oif == 0)
839 fl6.flowi6_oif = sk->sk_bound_dev_if;
840
841 sockc.tsflags = sk->sk_tsflags;
842 if (msg->msg_controllen) {
843 opt = &opt_space;
844 memset(opt, 0, sizeof(struct ipv6_txoptions));
845 opt->tot_len = sizeof(struct ipv6_txoptions);
846 ipc6.opt = opt;
847
848 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc);
849 if (err < 0) {
850 fl6_sock_release(flowlabel);
851 return err;
852 }
853 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
854 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
855 if (!flowlabel)
856 return -EINVAL;
857 }
858 if (!(opt->opt_nflen|opt->opt_flen))
859 opt = NULL;
860 }
861 if (!opt) {
862 opt = txopt_get(np);
863 opt_to_free = opt;
864 }
865 if (flowlabel)
866 opt = fl6_merge_options(&opt_space, flowlabel, opt);
867 opt = ipv6_fixup_options(&opt_space, opt);
868
869 fl6.flowi6_proto = proto;
870 rfv.msg = msg;
871 rfv.hlen = 0;
872 err = rawv6_probe_proto_opt(&rfv, &fl6);
873 if (err)
874 goto out;
875
876 if (!ipv6_addr_any(daddr))
877 fl6.daddr = *daddr;
878 else
879 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
880 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
881 fl6.saddr = np->saddr;
882
883 final_p = fl6_update_dst(&fl6, opt, &final);
884
885 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
886 fl6.flowi6_oif = np->mcast_oif;
887 else if (!fl6.flowi6_oif)
888 fl6.flowi6_oif = np->ucast_oif;
889 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
890
891 if (inet->hdrincl)
892 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
893
894 if (ipc6.tclass < 0)
895 ipc6.tclass = np->tclass;
896
897 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
898
899 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
900 if (IS_ERR(dst)) {
901 err = PTR_ERR(dst);
902 goto out;
903 }
904 if (ipc6.hlimit < 0)
905 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
906
907 if (ipc6.dontfrag < 0)
908 ipc6.dontfrag = np->dontfrag;
909
910 if (msg->msg_flags&MSG_CONFIRM)
911 goto do_confirm;
912
913 back_from_confirm:
914 if (inet->hdrincl)
915 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst, msg->msg_flags);
916 else {
917 ipc6.opt = opt;
918 lock_sock(sk);
919 err = ip6_append_data(sk, raw6_getfrag, &rfv,
920 len, 0, &ipc6, &fl6, (struct rt6_info *)dst,
921 msg->msg_flags, &sockc);
922
923 if (err)
924 ip6_flush_pending_frames(sk);
925 else if (!(msg->msg_flags & MSG_MORE))
926 err = rawv6_push_pending_frames(sk, &fl6, rp);
927 release_sock(sk);
928 }
929 done:
930 dst_release(dst);
931 out:
932 fl6_sock_release(flowlabel);
933 txopt_put(opt_to_free);
934 return err < 0 ? err : len;
935 do_confirm:
936 dst_confirm(dst);
937 if (!(msg->msg_flags & MSG_PROBE) || len)
938 goto back_from_confirm;
939 err = 0;
940 goto done;
941 }
942
943 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
944 char __user *optval, int optlen)
945 {
946 switch (optname) {
947 case ICMPV6_FILTER:
948 if (optlen > sizeof(struct icmp6_filter))
949 optlen = sizeof(struct icmp6_filter);
950 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
951 return -EFAULT;
952 return 0;
953 default:
954 return -ENOPROTOOPT;
955 }
956
957 return 0;
958 }
959
960 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
961 char __user *optval, int __user *optlen)
962 {
963 int len;
964
965 switch (optname) {
966 case ICMPV6_FILTER:
967 if (get_user(len, optlen))
968 return -EFAULT;
969 if (len < 0)
970 return -EINVAL;
971 if (len > sizeof(struct icmp6_filter))
972 len = sizeof(struct icmp6_filter);
973 if (put_user(len, optlen))
974 return -EFAULT;
975 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
976 return -EFAULT;
977 return 0;
978 default:
979 return -ENOPROTOOPT;
980 }
981
982 return 0;
983 }
984
985
986 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
987 char __user *optval, unsigned int optlen)
988 {
989 struct raw6_sock *rp = raw6_sk(sk);
990 int val;
991
992 if (get_user(val, (int __user *)optval))
993 return -EFAULT;
994
995 switch (optname) {
996 case IPV6_HDRINCL:
997 if (sk->sk_type != SOCK_RAW)
998 return -EINVAL;
999 inet_sk(sk)->hdrincl = !!val;
1000 return 0;
1001 case IPV6_CHECKSUM:
1002 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1003 level == IPPROTO_IPV6) {
1004 /*
1005 * RFC3542 tells that IPV6_CHECKSUM socket
1006 * option in the IPPROTO_IPV6 level is not
1007 * allowed on ICMPv6 sockets.
1008 * If you want to set it, use IPPROTO_RAW
1009 * level IPV6_CHECKSUM socket option
1010 * (Linux extension).
1011 */
1012 return -EINVAL;
1013 }
1014
1015 /* You may get strange result with a positive odd offset;
1016 RFC2292bis agrees with me. */
1017 if (val > 0 && (val&1))
1018 return -EINVAL;
1019 if (val < 0) {
1020 rp->checksum = 0;
1021 } else {
1022 rp->checksum = 1;
1023 rp->offset = val;
1024 }
1025
1026 return 0;
1027
1028 default:
1029 return -ENOPROTOOPT;
1030 }
1031 }
1032
1033 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1034 char __user *optval, unsigned int optlen)
1035 {
1036 switch (level) {
1037 case SOL_RAW:
1038 break;
1039
1040 case SOL_ICMPV6:
1041 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1042 return -EOPNOTSUPP;
1043 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1044 case SOL_IPV6:
1045 if (optname == IPV6_CHECKSUM ||
1046 optname == IPV6_HDRINCL)
1047 break;
1048 default:
1049 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1050 }
1051
1052 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1053 }
1054
1055 #ifdef CONFIG_COMPAT
1056 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1057 char __user *optval, unsigned int optlen)
1058 {
1059 switch (level) {
1060 case SOL_RAW:
1061 break;
1062 case SOL_ICMPV6:
1063 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1064 return -EOPNOTSUPP;
1065 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1066 case SOL_IPV6:
1067 if (optname == IPV6_CHECKSUM ||
1068 optname == IPV6_HDRINCL)
1069 break;
1070 default:
1071 return compat_ipv6_setsockopt(sk, level, optname,
1072 optval, optlen);
1073 }
1074 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1075 }
1076 #endif
1077
1078 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1079 char __user *optval, int __user *optlen)
1080 {
1081 struct raw6_sock *rp = raw6_sk(sk);
1082 int val, len;
1083
1084 if (get_user(len, optlen))
1085 return -EFAULT;
1086
1087 switch (optname) {
1088 case IPV6_HDRINCL:
1089 val = inet_sk(sk)->hdrincl;
1090 break;
1091 case IPV6_CHECKSUM:
1092 /*
1093 * We allow getsockopt() for IPPROTO_IPV6-level
1094 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1095 * since RFC3542 is silent about it.
1096 */
1097 if (rp->checksum == 0)
1098 val = -1;
1099 else
1100 val = rp->offset;
1101 break;
1102
1103 default:
1104 return -ENOPROTOOPT;
1105 }
1106
1107 len = min_t(unsigned int, sizeof(int), len);
1108
1109 if (put_user(len, optlen))
1110 return -EFAULT;
1111 if (copy_to_user(optval, &val, len))
1112 return -EFAULT;
1113 return 0;
1114 }
1115
1116 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1117 char __user *optval, int __user *optlen)
1118 {
1119 switch (level) {
1120 case SOL_RAW:
1121 break;
1122
1123 case SOL_ICMPV6:
1124 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1125 return -EOPNOTSUPP;
1126 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1127 case SOL_IPV6:
1128 if (optname == IPV6_CHECKSUM ||
1129 optname == IPV6_HDRINCL)
1130 break;
1131 default:
1132 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1133 }
1134
1135 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1136 }
1137
1138 #ifdef CONFIG_COMPAT
1139 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1140 char __user *optval, int __user *optlen)
1141 {
1142 switch (level) {
1143 case SOL_RAW:
1144 break;
1145 case SOL_ICMPV6:
1146 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1147 return -EOPNOTSUPP;
1148 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1149 case SOL_IPV6:
1150 if (optname == IPV6_CHECKSUM ||
1151 optname == IPV6_HDRINCL)
1152 break;
1153 default:
1154 return compat_ipv6_getsockopt(sk, level, optname,
1155 optval, optlen);
1156 }
1157 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1158 }
1159 #endif
1160
1161 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1162 {
1163 switch (cmd) {
1164 case SIOCOUTQ: {
1165 int amount = sk_wmem_alloc_get(sk);
1166
1167 return put_user(amount, (int __user *)arg);
1168 }
1169 case SIOCINQ: {
1170 struct sk_buff *skb;
1171 int amount = 0;
1172
1173 spin_lock_bh(&sk->sk_receive_queue.lock);
1174 skb = skb_peek(&sk->sk_receive_queue);
1175 if (skb)
1176 amount = skb->len;
1177 spin_unlock_bh(&sk->sk_receive_queue.lock);
1178 return put_user(amount, (int __user *)arg);
1179 }
1180
1181 default:
1182 #ifdef CONFIG_IPV6_MROUTE
1183 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1184 #else
1185 return -ENOIOCTLCMD;
1186 #endif
1187 }
1188 }
1189
1190 #ifdef CONFIG_COMPAT
1191 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1192 {
1193 switch (cmd) {
1194 case SIOCOUTQ:
1195 case SIOCINQ:
1196 return -ENOIOCTLCMD;
1197 default:
1198 #ifdef CONFIG_IPV6_MROUTE
1199 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1200 #else
1201 return -ENOIOCTLCMD;
1202 #endif
1203 }
1204 }
1205 #endif
1206
1207 static void rawv6_close(struct sock *sk, long timeout)
1208 {
1209 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1210 ip6_ra_control(sk, -1);
1211 ip6mr_sk_done(sk);
1212 sk_common_release(sk);
1213 }
1214
1215 static void raw6_destroy(struct sock *sk)
1216 {
1217 lock_sock(sk);
1218 ip6_flush_pending_frames(sk);
1219 release_sock(sk);
1220
1221 inet6_destroy_sock(sk);
1222 }
1223
1224 static int rawv6_init_sk(struct sock *sk)
1225 {
1226 struct raw6_sock *rp = raw6_sk(sk);
1227
1228 switch (inet_sk(sk)->inet_num) {
1229 case IPPROTO_ICMPV6:
1230 rp->checksum = 1;
1231 rp->offset = 2;
1232 break;
1233 case IPPROTO_MH:
1234 rp->checksum = 1;
1235 rp->offset = 4;
1236 break;
1237 default:
1238 break;
1239 }
1240 return 0;
1241 }
1242
1243 struct proto rawv6_prot = {
1244 .name = "RAWv6",
1245 .owner = THIS_MODULE,
1246 .close = rawv6_close,
1247 .destroy = raw6_destroy,
1248 .connect = ip6_datagram_connect_v6_only,
1249 .disconnect = __udp_disconnect,
1250 .ioctl = rawv6_ioctl,
1251 .init = rawv6_init_sk,
1252 .setsockopt = rawv6_setsockopt,
1253 .getsockopt = rawv6_getsockopt,
1254 .sendmsg = rawv6_sendmsg,
1255 .recvmsg = rawv6_recvmsg,
1256 .bind = rawv6_bind,
1257 .backlog_rcv = rawv6_rcv_skb,
1258 .hash = raw_hash_sk,
1259 .unhash = raw_unhash_sk,
1260 .obj_size = sizeof(struct raw6_sock),
1261 .useroffset = offsetof(struct raw6_sock, filter),
1262 .usersize = sizeof(((struct raw6_sock *)0)->filter),
1263 .h.raw_hash = &raw_v6_hashinfo,
1264 #ifdef CONFIG_COMPAT
1265 .compat_setsockopt = compat_rawv6_setsockopt,
1266 .compat_getsockopt = compat_rawv6_getsockopt,
1267 .compat_ioctl = compat_rawv6_ioctl,
1268 #endif
1269 };
1270
1271 #ifdef CONFIG_PROC_FS
1272 static int raw6_seq_show(struct seq_file *seq, void *v)
1273 {
1274 if (v == SEQ_START_TOKEN) {
1275 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1276 } else {
1277 struct sock *sp = v;
1278 __u16 srcp = inet_sk(sp)->inet_num;
1279 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1280 raw_seq_private(seq)->bucket);
1281 }
1282 return 0;
1283 }
1284
1285 static const struct seq_operations raw6_seq_ops = {
1286 .start = raw_seq_start,
1287 .next = raw_seq_next,
1288 .stop = raw_seq_stop,
1289 .show = raw6_seq_show,
1290 };
1291
1292 static int raw6_seq_open(struct inode *inode, struct file *file)
1293 {
1294 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1295 }
1296
1297 static const struct file_operations raw6_seq_fops = {
1298 .owner = THIS_MODULE,
1299 .open = raw6_seq_open,
1300 .read = seq_read,
1301 .llseek = seq_lseek,
1302 .release = seq_release_net,
1303 };
1304
1305 static int __net_init raw6_init_net(struct net *net)
1306 {
1307 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops))
1308 return -ENOMEM;
1309
1310 return 0;
1311 }
1312
1313 static void __net_exit raw6_exit_net(struct net *net)
1314 {
1315 remove_proc_entry("raw6", net->proc_net);
1316 }
1317
1318 static struct pernet_operations raw6_net_ops = {
1319 .init = raw6_init_net,
1320 .exit = raw6_exit_net,
1321 };
1322
1323 int __init raw6_proc_init(void)
1324 {
1325 return register_pernet_subsys(&raw6_net_ops);
1326 }
1327
1328 void raw6_proc_exit(void)
1329 {
1330 unregister_pernet_subsys(&raw6_net_ops);
1331 }
1332 #endif /* CONFIG_PROC_FS */
1333
1334 /* Same as inet6_dgram_ops, sans udp_poll. */
1335 const struct proto_ops inet6_sockraw_ops = {
1336 .family = PF_INET6,
1337 .owner = THIS_MODULE,
1338 .release = inet6_release,
1339 .bind = inet6_bind,
1340 .connect = inet_dgram_connect, /* ok */
1341 .socketpair = sock_no_socketpair, /* a do nothing */
1342 .accept = sock_no_accept, /* a do nothing */
1343 .getname = inet6_getname,
1344 .poll = datagram_poll, /* ok */
1345 .ioctl = inet6_ioctl, /* must change */
1346 .listen = sock_no_listen, /* ok */
1347 .shutdown = inet_shutdown, /* ok */
1348 .setsockopt = sock_common_setsockopt, /* ok */
1349 .getsockopt = sock_common_getsockopt, /* ok */
1350 .sendmsg = inet_sendmsg, /* ok */
1351 .recvmsg = sock_common_recvmsg, /* ok */
1352 .mmap = sock_no_mmap,
1353 .sendpage = sock_no_sendpage,
1354 #ifdef CONFIG_COMPAT
1355 .compat_setsockopt = compat_sock_common_setsockopt,
1356 .compat_getsockopt = compat_sock_common_getsockopt,
1357 #endif
1358 };
1359
1360 static struct inet_protosw rawv6_protosw = {
1361 .type = SOCK_RAW,
1362 .protocol = IPPROTO_IP, /* wild card */
1363 .prot = &rawv6_prot,
1364 .ops = &inet6_sockraw_ops,
1365 .flags = INET_PROTOSW_REUSE,
1366 };
1367
1368 int __init rawv6_init(void)
1369 {
1370 return inet6_register_protosw(&rawv6_protosw);
1371 }
1372
1373 void rawv6_exit(void)
1374 {
1375 inet6_unregister_protosw(&rawv6_protosw);
1376 }
Arne's tree of linux kernel.