?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 4.13.16
[linux/fpc-iii.git] / net / ipv6 / raw.c
blob60be012fe7085cc7a199e84333cef5ee95ed1f04
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 struct raw_hashinfo raw_v6_hashinfo = {
69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
70 };
71 EXPORT_SYMBOL_GPL(raw_v6_hashinfo);
72
73 struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
74 unsigned short num, const struct in6_addr *loc_addr,
75 const struct in6_addr *rmt_addr, int dif)
76 {
77 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
78
79 sk_for_each_from(sk)
80 if (inet_sk(sk)->inet_num == num) {
81
82 if (!net_eq(sock_net(sk), net))
83 continue;
84
85 if (!ipv6_addr_any(&sk->sk_v6_daddr) &&
86 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr))
87 continue;
88
89 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
90 continue;
91
92 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
93 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))
94 goto found;
95 if (is_multicast &&
96 inet6_mc_check(sk, loc_addr, rmt_addr))
97 goto found;
98 continue;
99 }
100 goto found;
101 }
102 sk = NULL;
103 found:
104 return sk;
105 }
106 EXPORT_SYMBOL_GPL(__raw_v6_lookup);
107
108 /*
109 * 0 - deliver
110 * 1 - block
111 */
112 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
113 {
114 struct icmp6hdr _hdr;
115 const struct icmp6hdr *hdr;
116
117 /* We require only the four bytes of the ICMPv6 header, not any
118 * additional bytes of message body in "struct icmp6hdr".
119 */
120 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
121 ICMPV6_HDRLEN, &_hdr);
122 if (hdr) {
123 const __u32 *data = &raw6_sk(sk)->filter.data[0];
124 unsigned int type = hdr->icmp6_type;
125
126 return (data[type >> 5] & (1U << (type & 31))) != 0;
127 }
128 return 1;
129 }
130
131 #if IS_ENABLED(CONFIG_IPV6_MIP6)
132 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
133
134 static mh_filter_t __rcu *mh_filter __read_mostly;
135
136 int rawv6_mh_filter_register(mh_filter_t filter)
137 {
138 rcu_assign_pointer(mh_filter, filter);
139 return 0;
140 }
141 EXPORT_SYMBOL(rawv6_mh_filter_register);
142
143 int rawv6_mh_filter_unregister(mh_filter_t filter)
144 {
145 RCU_INIT_POINTER(mh_filter, NULL);
146 synchronize_rcu();
147 return 0;
148 }
149 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
150
151 #endif
152
153 /*
154 * demultiplex raw sockets.
155 * (should consider queueing the skb in the sock receive_queue
156 * without calling rawv6.c)
157 *
158 * Caller owns SKB so we must make clones.
159 */
160 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
161 {
162 const struct in6_addr *saddr;
163 const struct in6_addr *daddr;
164 struct sock *sk;
165 bool delivered = false;
166 __u8 hash;
167 struct net *net;
168
169 saddr = &ipv6_hdr(skb)->saddr;
170 daddr = saddr + 1;
171
172 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
173
174 read_lock(&raw_v6_hashinfo.lock);
175 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
176
177 if (!sk)
178 goto out;
179
180 net = dev_net(skb->dev);
181 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, inet6_iif(skb));
182
183 while (sk) {
184 int filtered;
185
186 delivered = true;
187 switch (nexthdr) {
188 case IPPROTO_ICMPV6:
189 filtered = icmpv6_filter(sk, skb);
190 break;
191
192 #if IS_ENABLED(CONFIG_IPV6_MIP6)
193 case IPPROTO_MH:
194 {
195 /* XXX: To validate MH only once for each packet,
196 * this is placed here. It should be after checking
197 * xfrm policy, however it doesn't. The checking xfrm
198 * policy is placed in rawv6_rcv() because it is
199 * required for each socket.
200 */
201 mh_filter_t *filter;
202
203 filter = rcu_dereference(mh_filter);
204 filtered = filter ? (*filter)(sk, skb) : 0;
205 break;
206 }
207 #endif
208 default:
209 filtered = 0;
210 break;
211 }
212
213 if (filtered < 0)
214 break;
215 if (filtered == 0) {
216 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
217
218 /* Not releasing hash table! */
219 if (clone) {
220 nf_reset(clone);
221 rawv6_rcv(sk, clone);
222 }
223 }
224 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
225 inet6_iif(skb));
226 }
227 out:
228 read_unlock(&raw_v6_hashinfo.lock);
229 return delivered;
230 }
231
232 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
233 {
234 struct sock *raw_sk;
235
236 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
237 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
238 raw_sk = NULL;
239
240 return raw_sk != NULL;
241 }
242
243 /* This cleans up af_inet6 a bit. -DaveM */
244 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
245 {
246 struct inet_sock *inet = inet_sk(sk);
247 struct ipv6_pinfo *np = inet6_sk(sk);
248 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
249 __be32 v4addr = 0;
250 int addr_type;
251 int err;
252
253 if (addr_len < SIN6_LEN_RFC2133)
254 return -EINVAL;
255
256 if (addr->sin6_family != AF_INET6)
257 return -EINVAL;
258
259 addr_type = ipv6_addr_type(&addr->sin6_addr);
260
261 /* Raw sockets are IPv6 only */
262 if (addr_type == IPV6_ADDR_MAPPED)
263 return -EADDRNOTAVAIL;
264
265 lock_sock(sk);
266
267 err = -EINVAL;
268 if (sk->sk_state != TCP_CLOSE)
269 goto out;
270
271 rcu_read_lock();
272 /* Check if the address belongs to the host. */
273 if (addr_type != IPV6_ADDR_ANY) {
274 struct net_device *dev = NULL;
275
276 if (__ipv6_addr_needs_scope_id(addr_type)) {
277 if (addr_len >= sizeof(struct sockaddr_in6) &&
278 addr->sin6_scope_id) {
279 /* Override any existing binding, if another
280 * one is supplied by user.
281 */
282 sk->sk_bound_dev_if = addr->sin6_scope_id;
283 }
284
285 /* Binding to link-local address requires an interface */
286 if (!sk->sk_bound_dev_if)
287 goto out_unlock;
288
289 err = -ENODEV;
290 dev = dev_get_by_index_rcu(sock_net(sk),
291 sk->sk_bound_dev_if);
292 if (!dev)
293 goto out_unlock;
294 }
295
296 /* ipv4 addr of the socket is invalid. Only the
297 * unspecified and mapped address have a v4 equivalent.
298 */
299 v4addr = LOOPBACK4_IPV6;
300 if (!(addr_type & IPV6_ADDR_MULTICAST) &&
301 !sock_net(sk)->ipv6.sysctl.ip_nonlocal_bind) {
302 err = -EADDRNOTAVAIL;
303 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
304 dev, 0)) {
305 goto out_unlock;
306 }
307 }
308 }
309
310 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
311 sk->sk_v6_rcv_saddr = addr->sin6_addr;
312 if (!(addr_type & IPV6_ADDR_MULTICAST))
313 np->saddr = addr->sin6_addr;
314 err = 0;
315 out_unlock:
316 rcu_read_unlock();
317 out:
318 release_sock(sk);
319 return err;
320 }
321
322 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
323 struct inet6_skb_parm *opt,
324 u8 type, u8 code, int offset, __be32 info)
325 {
326 struct inet_sock *inet = inet_sk(sk);
327 struct ipv6_pinfo *np = inet6_sk(sk);
328 int err;
329 int harderr;
330
331 /* Report error on raw socket, if:
332 1. User requested recverr.
333 2. Socket is connected (otherwise the error indication
334 is useless without recverr and error is hard.
335 */
336 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
337 return;
338
339 harderr = icmpv6_err_convert(type, code, &err);
340 if (type == ICMPV6_PKT_TOOBIG) {
341 ip6_sk_update_pmtu(skb, sk, info);
342 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
343 }
344 if (type == NDISC_REDIRECT) {
345 ip6_sk_redirect(skb, sk);
346 return;
347 }
348 if (np->recverr) {
349 u8 *payload = skb->data;
350 if (!inet->hdrincl)
351 payload += offset;
352 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
353 }
354
355 if (np->recverr || harderr) {
356 sk->sk_err = err;
357 sk->sk_error_report(sk);
358 }
359 }
360
361 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
362 u8 type, u8 code, int inner_offset, __be32 info)
363 {
364 struct sock *sk;
365 int hash;
366 const struct in6_addr *saddr, *daddr;
367 struct net *net;
368
369 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
370
371 read_lock(&raw_v6_hashinfo.lock);
372 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
373 if (sk) {
374 /* Note: ipv6_hdr(skb) != skb->data */
375 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
376 saddr = &ip6h->saddr;
377 daddr = &ip6h->daddr;
378 net = dev_net(skb->dev);
379
380 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
381 inet6_iif(skb)))) {
382 rawv6_err(sk, skb, NULL, type, code,
383 inner_offset, info);
384 sk = sk_next(sk);
385 }
386 }
387 read_unlock(&raw_v6_hashinfo.lock);
388 }
389
390 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
391 {
392 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
393 skb_checksum_complete(skb)) {
394 atomic_inc(&sk->sk_drops);
395 kfree_skb(skb);
396 return NET_RX_DROP;
397 }
398
399 /* Charge it to the socket. */
400 skb_dst_drop(skb);
401 if (sock_queue_rcv_skb(sk, skb) < 0) {
402 kfree_skb(skb);
403 return NET_RX_DROP;
404 }
405
406 return 0;
407 }
408
409 /*
410 * This is next to useless...
411 * if we demultiplex in network layer we don't need the extra call
412 * just to queue the skb...
413 * maybe we could have the network decide upon a hint if it
414 * should call raw_rcv for demultiplexing
415 */
416 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
417 {
418 struct inet_sock *inet = inet_sk(sk);
419 struct raw6_sock *rp = raw6_sk(sk);
420
421 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
422 atomic_inc(&sk->sk_drops);
423 kfree_skb(skb);
424 return NET_RX_DROP;
425 }
426
427 if (!rp->checksum)
428 skb->ip_summed = CHECKSUM_UNNECESSARY;
429
430 if (skb->ip_summed == CHECKSUM_COMPLETE) {
431 skb_postpull_rcsum(skb, skb_network_header(skb),
432 skb_network_header_len(skb));
433 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
434 &ipv6_hdr(skb)->daddr,
435 skb->len, inet->inet_num, skb->csum))
436 skb->ip_summed = CHECKSUM_UNNECESSARY;
437 }
438 if (!skb_csum_unnecessary(skb))
439 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
440 &ipv6_hdr(skb)->daddr,
441 skb->len,
442 inet->inet_num, 0));
443
444 if (inet->hdrincl) {
445 if (skb_checksum_complete(skb)) {
446 atomic_inc(&sk->sk_drops);
447 kfree_skb(skb);
448 return NET_RX_DROP;
449 }
450 }
451
452 rawv6_rcv_skb(sk, skb);
453 return 0;
454 }
455
456
457 /*
458 * This should be easy, if there is something there
459 * we return it, otherwise we block.
460 */
461
462 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
463 int noblock, int flags, int *addr_len)
464 {
465 struct ipv6_pinfo *np = inet6_sk(sk);
466 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
467 struct sk_buff *skb;
468 size_t copied;
469 int err;
470
471 if (flags & MSG_OOB)
472 return -EOPNOTSUPP;
473
474 if (flags & MSG_ERRQUEUE)
475 return ipv6_recv_error(sk, msg, len, addr_len);
476
477 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
478 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
479
480 skb = skb_recv_datagram(sk, flags, noblock, &err);
481 if (!skb)
482 goto out;
483
484 copied = skb->len;
485 if (copied > len) {
486 copied = len;
487 msg->msg_flags |= MSG_TRUNC;
488 }
489
490 if (skb_csum_unnecessary(skb)) {
491 err = skb_copy_datagram_msg(skb, 0, msg, copied);
492 } else if (msg->msg_flags&MSG_TRUNC) {
493 if (__skb_checksum_complete(skb))
494 goto csum_copy_err;
495 err = skb_copy_datagram_msg(skb, 0, msg, copied);
496 } else {
497 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
498 if (err == -EINVAL)
499 goto csum_copy_err;
500 }
501 if (err)
502 goto out_free;
503
504 /* Copy the address. */
505 if (sin6) {
506 sin6->sin6_family = AF_INET6;
507 sin6->sin6_port = 0;
508 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
509 sin6->sin6_flowinfo = 0;
510 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
511 inet6_iif(skb));
512 *addr_len = sizeof(*sin6);
513 }
514
515 sock_recv_ts_and_drops(msg, sk, skb);
516
517 if (np->rxopt.all)
518 ip6_datagram_recv_ctl(sk, msg, skb);
519
520 err = copied;
521 if (flags & MSG_TRUNC)
522 err = skb->len;
523
524 out_free:
525 skb_free_datagram(sk, skb);
526 out:
527 return err;
528
529 csum_copy_err:
530 skb_kill_datagram(sk, skb, flags);
531
532 /* Error for blocking case is chosen to masquerade
533 as some normal condition.
534 */
535 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
536 goto out;
537 }
538
539 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
540 struct raw6_sock *rp)
541 {
542 struct sk_buff *skb;
543 int err = 0;
544 int offset;
545 int len;
546 int total_len;
547 __wsum tmp_csum;
548 __sum16 csum;
549
550 if (!rp->checksum)
551 goto send;
552
553 skb = skb_peek(&sk->sk_write_queue);
554 if (!skb)
555 goto out;
556
557 offset = rp->offset;
558 total_len = inet_sk(sk)->cork.base.length;
559 if (offset >= total_len - 1) {
560 err = -EINVAL;
561 ip6_flush_pending_frames(sk);
562 goto out;
563 }
564
565 /* should be check HW csum miyazawa */
566 if (skb_queue_len(&sk->sk_write_queue) == 1) {
567 /*
568 * Only one fragment on the socket.
569 */
570 tmp_csum = skb->csum;
571 } else {
572 struct sk_buff *csum_skb = NULL;
573 tmp_csum = 0;
574
575 skb_queue_walk(&sk->sk_write_queue, skb) {
576 tmp_csum = csum_add(tmp_csum, skb->csum);
577
578 if (csum_skb)
579 continue;
580
581 len = skb->len - skb_transport_offset(skb);
582 if (offset >= len) {
583 offset -= len;
584 continue;
585 }
586
587 csum_skb = skb;
588 }
589
590 skb = csum_skb;
591 }
592
593 offset += skb_transport_offset(skb);
594 err = skb_copy_bits(skb, offset, &csum, 2);
595 if (err < 0) {
596 ip6_flush_pending_frames(sk);
597 goto out;
598 }
599
600 /* in case cksum was not initialized */
601 if (unlikely(csum))
602 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
603
604 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
605 total_len, fl6->flowi6_proto, tmp_csum);
606
607 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
608 csum = CSUM_MANGLED_0;
609
610 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
611
612 send:
613 err = ip6_push_pending_frames(sk);
614 out:
615 return err;
616 }
617
618 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
619 struct flowi6 *fl6, struct dst_entry **dstp,
620 unsigned int flags)
621 {
622 struct ipv6_pinfo *np = inet6_sk(sk);
623 struct net *net = sock_net(sk);
624 struct ipv6hdr *iph;
625 struct sk_buff *skb;
626 int err;
627 struct rt6_info *rt = (struct rt6_info *)*dstp;
628 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
629 int tlen = rt->dst.dev->needed_tailroom;
630
631 if (length > rt->dst.dev->mtu) {
632 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
633 return -EMSGSIZE;
634 }
635 if (length < sizeof(struct ipv6hdr))
636 return -EINVAL;
637 if (flags&MSG_PROBE)
638 goto out;
639
640 skb = sock_alloc_send_skb(sk,
641 length + hlen + tlen + 15,
642 flags & MSG_DONTWAIT, &err);
643 if (!skb)
644 goto error;
645 skb_reserve(skb, hlen);
646
647 skb->protocol = htons(ETH_P_IPV6);
648 skb->priority = sk->sk_priority;
649 skb->mark = sk->sk_mark;
650 skb_dst_set(skb, &rt->dst);
651 *dstp = NULL;
652
653 skb_put(skb, length);
654 skb_reset_network_header(skb);
655 iph = ipv6_hdr(skb);
656
657 skb->ip_summed = CHECKSUM_NONE;
658
659 if (flags & MSG_CONFIRM)
660 skb_set_dst_pending_confirm(skb, 1);
661
662 skb->transport_header = skb->network_header;
663 err = memcpy_from_msg(iph, msg, length);
664 if (err)
665 goto error_fault;
666
667 /* if egress device is enslaved to an L3 master device pass the
668 * skb to its handler for processing
669 */
670 skb = l3mdev_ip6_out(sk, skb);
671 if (unlikely(!skb))
672 return 0;
673
674 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
675 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
676 NULL, rt->dst.dev, dst_output);
677 if (err > 0)
678 err = net_xmit_errno(err);
679 if (err)
680 goto error;
681 out:
682 return 0;
683
684 error_fault:
685 err = -EFAULT;
686 kfree_skb(skb);
687 error:
688 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
689 if (err == -ENOBUFS && !np->recverr)
690 err = 0;
691 return err;
692 }
693
694 struct raw6_frag_vec {
695 struct msghdr *msg;
696 int hlen;
697 char c[4];
698 };
699
700 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
701 {
702 int err = 0;
703 switch (fl6->flowi6_proto) {
704 case IPPROTO_ICMPV6:
705 rfv->hlen = 2;
706 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
707 if (!err) {
708 fl6->fl6_icmp_type = rfv->c[0];
709 fl6->fl6_icmp_code = rfv->c[1];
710 }
711 break;
712 case IPPROTO_MH:
713 rfv->hlen = 4;
714 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
715 if (!err)
716 fl6->fl6_mh_type = rfv->c[2];
717 }
718 return err;
719 }
720
721 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
722 struct sk_buff *skb)
723 {
724 struct raw6_frag_vec *rfv = from;
725
726 if (offset < rfv->hlen) {
727 int copy = min(rfv->hlen - offset, len);
728
729 if (skb->ip_summed == CHECKSUM_PARTIAL)
730 memcpy(to, rfv->c + offset, copy);
731 else
732 skb->csum = csum_block_add(
733 skb->csum,
734 csum_partial_copy_nocheck(rfv->c + offset,
735 to, copy, 0),
736 odd);
737
738 odd = 0;
739 offset += copy;
740 to += copy;
741 len -= copy;
742
743 if (!len)
744 return 0;
745 }
746
747 offset -= rfv->hlen;
748
749 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
750 }
751
752 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
753 {
754 struct ipv6_txoptions *opt_to_free = NULL;
755 struct ipv6_txoptions opt_space;
756 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
757 struct in6_addr *daddr, *final_p, final;
758 struct inet_sock *inet = inet_sk(sk);
759 struct ipv6_pinfo *np = inet6_sk(sk);
760 struct raw6_sock *rp = raw6_sk(sk);
761 struct ipv6_txoptions *opt = NULL;
762 struct ip6_flowlabel *flowlabel = NULL;
763 struct dst_entry *dst = NULL;
764 struct raw6_frag_vec rfv;
765 struct flowi6 fl6;
766 struct sockcm_cookie sockc;
767 struct ipcm6_cookie ipc6;
768 int addr_len = msg->msg_namelen;
769 u16 proto;
770 int err;
771
772 /* Rough check on arithmetic overflow,
773 better check is made in ip6_append_data().
774 */
775 if (len > INT_MAX)
776 return -EMSGSIZE;
777
778 /* Mirror BSD error message compatibility */
779 if (msg->msg_flags & MSG_OOB)
780 return -EOPNOTSUPP;
781
782 /*
783 * Get and verify the address.
784 */
785 memset(&fl6, 0, sizeof(fl6));
786
787 fl6.flowi6_mark = sk->sk_mark;
788 fl6.flowi6_uid = sk->sk_uid;
789
790 ipc6.hlimit = -1;
791 ipc6.tclass = -1;
792 ipc6.dontfrag = -1;
793 ipc6.opt = NULL;
794
795 if (sin6) {
796 if (addr_len < SIN6_LEN_RFC2133)
797 return -EINVAL;
798
799 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
800 return -EAFNOSUPPORT;
801
802 /* port is the proto value [0..255] carried in nexthdr */
803 proto = ntohs(sin6->sin6_port);
804
805 if (!proto)
806 proto = inet->inet_num;
807 else if (proto != inet->inet_num)
808 return -EINVAL;
809
810 if (proto > 255)
811 return -EINVAL;
812
813 daddr = &sin6->sin6_addr;
814 if (np->sndflow) {
815 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
816 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
817 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
818 if (!flowlabel)
819 return -EINVAL;
820 }
821 }
822
823 /*
824 * Otherwise it will be difficult to maintain
825 * sk->sk_dst_cache.
826 */
827 if (sk->sk_state == TCP_ESTABLISHED &&
828 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
829 daddr = &sk->sk_v6_daddr;
830
831 if (addr_len >= sizeof(struct sockaddr_in6) &&
832 sin6->sin6_scope_id &&
833 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
834 fl6.flowi6_oif = sin6->sin6_scope_id;
835 } else {
836 if (sk->sk_state != TCP_ESTABLISHED)
837 return -EDESTADDRREQ;
838
839 proto = inet->inet_num;
840 daddr = &sk->sk_v6_daddr;
841 fl6.flowlabel = np->flow_label;
842 }
843
844 if (fl6.flowi6_oif == 0)
845 fl6.flowi6_oif = sk->sk_bound_dev_if;
846
847 sockc.tsflags = sk->sk_tsflags;
848 if (msg->msg_controllen) {
849 opt = &opt_space;
850 memset(opt, 0, sizeof(struct ipv6_txoptions));
851 opt->tot_len = sizeof(struct ipv6_txoptions);
852 ipc6.opt = opt;
853
854 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc);
855 if (err < 0) {
856 fl6_sock_release(flowlabel);
857 return err;
858 }
859 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
860 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
861 if (!flowlabel)
862 return -EINVAL;
863 }
864 if (!(opt->opt_nflen|opt->opt_flen))
865 opt = NULL;
866 }
867 if (!opt) {
868 opt = txopt_get(np);
869 opt_to_free = opt;
870 }
871 if (flowlabel)
872 opt = fl6_merge_options(&opt_space, flowlabel, opt);
873 opt = ipv6_fixup_options(&opt_space, opt);
874
875 fl6.flowi6_proto = proto;
876 rfv.msg = msg;
877 rfv.hlen = 0;
878 err = rawv6_probe_proto_opt(&rfv, &fl6);
879 if (err)
880 goto out;
881
882 if (!ipv6_addr_any(daddr))
883 fl6.daddr = *daddr;
884 else
885 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
886 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
887 fl6.saddr = np->saddr;
888
889 final_p = fl6_update_dst(&fl6, opt, &final);
890
891 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
892 fl6.flowi6_oif = np->mcast_oif;
893 else if (!fl6.flowi6_oif)
894 fl6.flowi6_oif = np->ucast_oif;
895 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
896
897 if (inet->hdrincl)
898 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
899
900 if (ipc6.tclass < 0)
901 ipc6.tclass = np->tclass;
902
903 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
904
905 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
906 if (IS_ERR(dst)) {
907 err = PTR_ERR(dst);
908 goto out;
909 }
910 if (ipc6.hlimit < 0)
911 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
912
913 if (ipc6.dontfrag < 0)
914 ipc6.dontfrag = np->dontfrag;
915
916 if (msg->msg_flags&MSG_CONFIRM)
917 goto do_confirm;
918
919 back_from_confirm:
920 if (inet->hdrincl)
921 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst, msg->msg_flags);
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, &sockc);
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 char __user *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_user(&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 char __user *optval, unsigned int optlen)
995 {
996 struct raw6_sock *rp = raw6_sk(sk);
997 int val;
998
999 if (get_user(val, (int __user *)optval))
1000 return -EFAULT;
1001
1002 switch (optname) {
1003 case IPV6_HDRINCL:
1004 if (sk->sk_type != SOCK_RAW)
1005 return -EINVAL;
1006 inet_sk(sk)->hdrincl = !!val;
1007 return 0;
1008 case IPV6_CHECKSUM:
1009 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1010 level == IPPROTO_IPV6) {
1011 /*
1012 * RFC3542 tells that IPV6_CHECKSUM socket
1013 * option in the IPPROTO_IPV6 level is not
1014 * allowed on ICMPv6 sockets.
1015 * If you want to set it, use IPPROTO_RAW
1016 * level IPV6_CHECKSUM socket option
1017 * (Linux extension).
1018 */
1019 return -EINVAL;
1020 }
1021
1022 /* You may get strange result with a positive odd offset;
1023 RFC2292bis agrees with me. */
1024 if (val > 0 && (val&1))
1025 return -EINVAL;
1026 if (val < 0) {
1027 rp->checksum = 0;
1028 } else {
1029 rp->checksum = 1;
1030 rp->offset = val;
1031 }
1032
1033 return 0;
1034
1035 default:
1036 return -ENOPROTOOPT;
1037 }
1038 }
1039
1040 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1041 char __user *optval, unsigned int optlen)
1042 {
1043 switch (level) {
1044 case SOL_RAW:
1045 break;
1046
1047 case SOL_ICMPV6:
1048 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1049 return -EOPNOTSUPP;
1050 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1051 case SOL_IPV6:
1052 if (optname == IPV6_CHECKSUM ||
1053 optname == IPV6_HDRINCL)
1054 break;
1055 default:
1056 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1057 }
1058
1059 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1060 }
1061
1062 #ifdef CONFIG_COMPAT
1063 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1064 char __user *optval, unsigned int optlen)
1065 {
1066 switch (level) {
1067 case SOL_RAW:
1068 break;
1069 case SOL_ICMPV6:
1070 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1071 return -EOPNOTSUPP;
1072 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1073 case SOL_IPV6:
1074 if (optname == IPV6_CHECKSUM ||
1075 optname == IPV6_HDRINCL)
1076 break;
1077 default:
1078 return compat_ipv6_setsockopt(sk, level, optname,
1079 optval, optlen);
1080 }
1081 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1082 }
1083 #endif
1084
1085 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1086 char __user *optval, int __user *optlen)
1087 {
1088 struct raw6_sock *rp = raw6_sk(sk);
1089 int val, len;
1090
1091 if (get_user(len, optlen))
1092 return -EFAULT;
1093
1094 switch (optname) {
1095 case IPV6_HDRINCL:
1096 val = inet_sk(sk)->hdrincl;
1097 break;
1098 case IPV6_CHECKSUM:
1099 /*
1100 * We allow getsockopt() for IPPROTO_IPV6-level
1101 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1102 * since RFC3542 is silent about it.
1103 */
1104 if (rp->checksum == 0)
1105 val = -1;
1106 else
1107 val = rp->offset;
1108 break;
1109
1110 default:
1111 return -ENOPROTOOPT;
1112 }
1113
1114 len = min_t(unsigned int, sizeof(int), len);
1115
1116 if (put_user(len, optlen))
1117 return -EFAULT;
1118 if (copy_to_user(optval, &val, len))
1119 return -EFAULT;
1120 return 0;
1121 }
1122
1123 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1124 char __user *optval, int __user *optlen)
1125 {
1126 switch (level) {
1127 case SOL_RAW:
1128 break;
1129
1130 case SOL_ICMPV6:
1131 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1132 return -EOPNOTSUPP;
1133 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1134 case SOL_IPV6:
1135 if (optname == IPV6_CHECKSUM ||
1136 optname == IPV6_HDRINCL)
1137 break;
1138 default:
1139 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1140 }
1141
1142 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1143 }
1144
1145 #ifdef CONFIG_COMPAT
1146 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1147 char __user *optval, int __user *optlen)
1148 {
1149 switch (level) {
1150 case SOL_RAW:
1151 break;
1152 case SOL_ICMPV6:
1153 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1154 return -EOPNOTSUPP;
1155 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1156 case SOL_IPV6:
1157 if (optname == IPV6_CHECKSUM ||
1158 optname == IPV6_HDRINCL)
1159 break;
1160 default:
1161 return compat_ipv6_getsockopt(sk, level, optname,
1162 optval, optlen);
1163 }
1164 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1165 }
1166 #endif
1167
1168 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1169 {
1170 switch (cmd) {
1171 case SIOCOUTQ: {
1172 int amount = sk_wmem_alloc_get(sk);
1173
1174 return put_user(amount, (int __user *)arg);
1175 }
1176 case SIOCINQ: {
1177 struct sk_buff *skb;
1178 int amount = 0;
1179
1180 spin_lock_bh(&sk->sk_receive_queue.lock);
1181 skb = skb_peek(&sk->sk_receive_queue);
1182 if (skb)
1183 amount = skb->len;
1184 spin_unlock_bh(&sk->sk_receive_queue.lock);
1185 return put_user(amount, (int __user *)arg);
1186 }
1187
1188 default:
1189 #ifdef CONFIG_IPV6_MROUTE
1190 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1191 #else
1192 return -ENOIOCTLCMD;
1193 #endif
1194 }
1195 }
1196
1197 #ifdef CONFIG_COMPAT
1198 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1199 {
1200 switch (cmd) {
1201 case SIOCOUTQ:
1202 case SIOCINQ:
1203 return -ENOIOCTLCMD;
1204 default:
1205 #ifdef CONFIG_IPV6_MROUTE
1206 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1207 #else
1208 return -ENOIOCTLCMD;
1209 #endif
1210 }
1211 }
1212 #endif
1213
1214 static void rawv6_close(struct sock *sk, long timeout)
1215 {
1216 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1217 ip6_ra_control(sk, -1);
1218 ip6mr_sk_done(sk);
1219 sk_common_release(sk);
1220 }
1221
1222 static void raw6_destroy(struct sock *sk)
1223 {
1224 lock_sock(sk);
1225 ip6_flush_pending_frames(sk);
1226 release_sock(sk);
1227
1228 inet6_destroy_sock(sk);
1229 }
1230
1231 static int rawv6_init_sk(struct sock *sk)
1232 {
1233 struct raw6_sock *rp = raw6_sk(sk);
1234
1235 switch (inet_sk(sk)->inet_num) {
1236 case IPPROTO_ICMPV6:
1237 rp->checksum = 1;
1238 rp->offset = 2;
1239 break;
1240 case IPPROTO_MH:
1241 rp->checksum = 1;
1242 rp->offset = 4;
1243 break;
1244 default:
1245 break;
1246 }
1247 return 0;
1248 }
1249
1250 struct proto rawv6_prot = {
1251 .name = "RAWv6",
1252 .owner = THIS_MODULE,
1253 .close = rawv6_close,
1254 .destroy = raw6_destroy,
1255 .connect = ip6_datagram_connect_v6_only,
1256 .disconnect = __udp_disconnect,
1257 .ioctl = rawv6_ioctl,
1258 .init = rawv6_init_sk,
1259 .setsockopt = rawv6_setsockopt,
1260 .getsockopt = rawv6_getsockopt,
1261 .sendmsg = rawv6_sendmsg,
1262 .recvmsg = rawv6_recvmsg,
1263 .bind = rawv6_bind,
1264 .backlog_rcv = rawv6_rcv_skb,
1265 .hash = raw_hash_sk,
1266 .unhash = raw_unhash_sk,
1267 .obj_size = sizeof(struct raw6_sock),
1268 .h.raw_hash = &raw_v6_hashinfo,
1269 #ifdef CONFIG_COMPAT
1270 .compat_setsockopt = compat_rawv6_setsockopt,
1271 .compat_getsockopt = compat_rawv6_getsockopt,
1272 .compat_ioctl = compat_rawv6_ioctl,
1273 #endif
1274 .diag_destroy = raw_abort,
1275 };
1276
1277 #ifdef CONFIG_PROC_FS
1278 static int raw6_seq_show(struct seq_file *seq, void *v)
1279 {
1280 if (v == SEQ_START_TOKEN) {
1281 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1282 } else {
1283 struct sock *sp = v;
1284 __u16 srcp = inet_sk(sp)->inet_num;
1285 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1286 raw_seq_private(seq)->bucket);
1287 }
1288 return 0;
1289 }
1290
1291 static const struct seq_operations raw6_seq_ops = {
1292 .start = raw_seq_start,
1293 .next = raw_seq_next,
1294 .stop = raw_seq_stop,
1295 .show = raw6_seq_show,
1296 };
1297
1298 static int raw6_seq_open(struct inode *inode, struct file *file)
1299 {
1300 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1301 }
1302
1303 static const struct file_operations raw6_seq_fops = {
1304 .owner = THIS_MODULE,
1305 .open = raw6_seq_open,
1306 .read = seq_read,
1307 .llseek = seq_lseek,
1308 .release = seq_release_net,
1309 };
1310
1311 static int __net_init raw6_init_net(struct net *net)
1312 {
1313 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops))
1314 return -ENOMEM;
1315
1316 return 0;
1317 }
1318
1319 static void __net_exit raw6_exit_net(struct net *net)
1320 {
1321 remove_proc_entry("raw6", net->proc_net);
1322 }
1323
1324 static struct pernet_operations raw6_net_ops = {
1325 .init = raw6_init_net,
1326 .exit = raw6_exit_net,
1327 };
1328
1329 int __init raw6_proc_init(void)
1330 {
1331 return register_pernet_subsys(&raw6_net_ops);
1332 }
1333
1334 void raw6_proc_exit(void)
1335 {
1336 unregister_pernet_subsys(&raw6_net_ops);
1337 }
1338 #endif /* CONFIG_PROC_FS */
1339
1340 /* Same as inet6_dgram_ops, sans udp_poll. */
1341 const struct proto_ops inet6_sockraw_ops = {
1342 .family = PF_INET6,
1343 .owner = THIS_MODULE,
1344 .release = inet6_release,
1345 .bind = inet6_bind,
1346 .connect = inet_dgram_connect, /* ok */
1347 .socketpair = sock_no_socketpair, /* a do nothing */
1348 .accept = sock_no_accept, /* a do nothing */
1349 .getname = inet6_getname,
1350 .poll = datagram_poll, /* ok */
1351 .ioctl = inet6_ioctl, /* must change */
1352 .listen = sock_no_listen, /* ok */
1353 .shutdown = inet_shutdown, /* ok */
1354 .setsockopt = sock_common_setsockopt, /* ok */
1355 .getsockopt = sock_common_getsockopt, /* ok */
1356 .sendmsg = inet_sendmsg, /* ok */
1357 .recvmsg = sock_common_recvmsg, /* ok */
1358 .mmap = sock_no_mmap,
1359 .sendpage = sock_no_sendpage,
1360 #ifdef CONFIG_COMPAT
1361 .compat_setsockopt = compat_sock_common_setsockopt,
1362 .compat_getsockopt = compat_sock_common_getsockopt,
1363 #endif
1364 };
1365
1366 static struct inet_protosw rawv6_protosw = {
1367 .type = SOCK_RAW,
1368 .protocol = IPPROTO_IP, /* wild card */
1369 .prot = &rawv6_prot,
1370 .ops = &inet6_sockraw_ops,
1371 .flags = INET_PROTOSW_REUSE,
1372 };
1373
1374 int __init rawv6_init(void)
1375 {
1376 return inet6_register_protosw(&rawv6_protosw);
1377 }
1378
1379 void rawv6_exit(void)
1380 {
1381 inet6_unregister_protosw(&rawv6_protosw);
1382 }
Linux with added FPC-III support