?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ARM: dts: MSM8974: Add restart node
[linux-2.6/btrfs-unstable.git] / net / ipv6 / raw.c
blob7fb4e14c467f60b4a65236b04fda13c6b5b9107d
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 <asm/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 == NULL)
176 goto out;
177
178 net = dev_net(skb->dev);
179 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
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 IP6CB(skb)->iif);
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 addr_type = ipv6_addr_type(&addr->sin6_addr);
254
255 /* Raw sockets are IPv6 only */
256 if (addr_type == IPV6_ADDR_MAPPED)
257 return -EADDRNOTAVAIL;
258
259 lock_sock(sk);
260
261 err = -EINVAL;
262 if (sk->sk_state != TCP_CLOSE)
263 goto out;
264
265 rcu_read_lock();
266 /* Check if the address belongs to the host. */
267 if (addr_type != IPV6_ADDR_ANY) {
268 struct net_device *dev = NULL;
269
270 if (__ipv6_addr_needs_scope_id(addr_type)) {
271 if (addr_len >= sizeof(struct sockaddr_in6) &&
272 addr->sin6_scope_id) {
273 /* Override any existing binding, if another
274 * one is supplied by user.
275 */
276 sk->sk_bound_dev_if = addr->sin6_scope_id;
277 }
278
279 /* Binding to link-local address requires an interface */
280 if (!sk->sk_bound_dev_if)
281 goto out_unlock;
282
283 err = -ENODEV;
284 dev = dev_get_by_index_rcu(sock_net(sk),
285 sk->sk_bound_dev_if);
286 if (!dev)
287 goto out_unlock;
288 }
289
290 /* ipv4 addr of the socket is invalid. Only the
291 * unspecified and mapped address have a v4 equivalent.
292 */
293 v4addr = LOOPBACK4_IPV6;
294 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
295 err = -EADDRNOTAVAIL;
296 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
297 dev, 0)) {
298 goto out_unlock;
299 }
300 }
301 }
302
303 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
304 sk->sk_v6_rcv_saddr = addr->sin6_addr;
305 if (!(addr_type & IPV6_ADDR_MULTICAST))
306 np->saddr = addr->sin6_addr;
307 err = 0;
308 out_unlock:
309 rcu_read_unlock();
310 out:
311 release_sock(sk);
312 return err;
313 }
314
315 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
316 struct inet6_skb_parm *opt,
317 u8 type, u8 code, int offset, __be32 info)
318 {
319 struct inet_sock *inet = inet_sk(sk);
320 struct ipv6_pinfo *np = inet6_sk(sk);
321 int err;
322 int harderr;
323
324 /* Report error on raw socket, if:
325 1. User requested recverr.
326 2. Socket is connected (otherwise the error indication
327 is useless without recverr and error is hard.
328 */
329 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
330 return;
331
332 harderr = icmpv6_err_convert(type, code, &err);
333 if (type == ICMPV6_PKT_TOOBIG) {
334 ip6_sk_update_pmtu(skb, sk, info);
335 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
336 }
337 if (type == NDISC_REDIRECT) {
338 ip6_sk_redirect(skb, sk);
339 return;
340 }
341 if (np->recverr) {
342 u8 *payload = skb->data;
343 if (!inet->hdrincl)
344 payload += offset;
345 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
346 }
347
348 if (np->recverr || harderr) {
349 sk->sk_err = err;
350 sk->sk_error_report(sk);
351 }
352 }
353
354 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
355 u8 type, u8 code, int inner_offset, __be32 info)
356 {
357 struct sock *sk;
358 int hash;
359 const struct in6_addr *saddr, *daddr;
360 struct net *net;
361
362 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
363
364 read_lock(&raw_v6_hashinfo.lock);
365 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
366 if (sk != NULL) {
367 /* Note: ipv6_hdr(skb) != skb->data */
368 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
369 saddr = &ip6h->saddr;
370 daddr = &ip6h->daddr;
371 net = dev_net(skb->dev);
372
373 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
374 IP6CB(skb)->iif))) {
375 rawv6_err(sk, skb, NULL, type, code,
376 inner_offset, info);
377 sk = sk_next(sk);
378 }
379 }
380 read_unlock(&raw_v6_hashinfo.lock);
381 }
382
383 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
384 {
385 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
386 skb_checksum_complete(skb)) {
387 atomic_inc(&sk->sk_drops);
388 kfree_skb(skb);
389 return NET_RX_DROP;
390 }
391
392 /* Charge it to the socket. */
393 skb_dst_drop(skb);
394 if (sock_queue_rcv_skb(sk, skb) < 0) {
395 kfree_skb(skb);
396 return NET_RX_DROP;
397 }
398
399 return 0;
400 }
401
402 /*
403 * This is next to useless...
404 * if we demultiplex in network layer we don't need the extra call
405 * just to queue the skb...
406 * maybe we could have the network decide upon a hint if it
407 * should call raw_rcv for demultiplexing
408 */
409 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
410 {
411 struct inet_sock *inet = inet_sk(sk);
412 struct raw6_sock *rp = raw6_sk(sk);
413
414 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
415 atomic_inc(&sk->sk_drops);
416 kfree_skb(skb);
417 return NET_RX_DROP;
418 }
419
420 if (!rp->checksum)
421 skb->ip_summed = CHECKSUM_UNNECESSARY;
422
423 if (skb->ip_summed == CHECKSUM_COMPLETE) {
424 skb_postpull_rcsum(skb, skb_network_header(skb),
425 skb_network_header_len(skb));
426 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
427 &ipv6_hdr(skb)->daddr,
428 skb->len, inet->inet_num, skb->csum))
429 skb->ip_summed = CHECKSUM_UNNECESSARY;
430 }
431 if (!skb_csum_unnecessary(skb))
432 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
433 &ipv6_hdr(skb)->daddr,
434 skb->len,
435 inet->inet_num, 0));
436
437 if (inet->hdrincl) {
438 if (skb_checksum_complete(skb)) {
439 atomic_inc(&sk->sk_drops);
440 kfree_skb(skb);
441 return NET_RX_DROP;
442 }
443 }
444
445 rawv6_rcv_skb(sk, skb);
446 return 0;
447 }
448
449
450 /*
451 * This should be easy, if there is something there
452 * we return it, otherwise we block.
453 */
454
455 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
456 struct msghdr *msg, size_t len,
457 int noblock, int flags, int *addr_len)
458 {
459 struct ipv6_pinfo *np = inet6_sk(sk);
460 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
461 struct sk_buff *skb;
462 size_t copied;
463 int err;
464
465 if (flags & MSG_OOB)
466 return -EOPNOTSUPP;
467
468 if (flags & MSG_ERRQUEUE)
469 return ipv6_recv_error(sk, msg, len, addr_len);
470
471 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
472 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
473
474 skb = skb_recv_datagram(sk, flags, noblock, &err);
475 if (!skb)
476 goto out;
477
478 copied = skb->len;
479 if (copied > len) {
480 copied = len;
481 msg->msg_flags |= MSG_TRUNC;
482 }
483
484 if (skb_csum_unnecessary(skb)) {
485 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
486 } else if (msg->msg_flags&MSG_TRUNC) {
487 if (__skb_checksum_complete(skb))
488 goto csum_copy_err;
489 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
490 } else {
491 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
492 if (err == -EINVAL)
493 goto csum_copy_err;
494 }
495 if (err)
496 goto out_free;
497
498 /* Copy the address. */
499 if (sin6) {
500 sin6->sin6_family = AF_INET6;
501 sin6->sin6_port = 0;
502 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
503 sin6->sin6_flowinfo = 0;
504 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
505 IP6CB(skb)->iif);
506 *addr_len = sizeof(*sin6);
507 }
508
509 sock_recv_ts_and_drops(msg, sk, skb);
510
511 if (np->rxopt.all)
512 ip6_datagram_recv_ctl(sk, msg, skb);
513
514 err = copied;
515 if (flags & MSG_TRUNC)
516 err = skb->len;
517
518 out_free:
519 skb_free_datagram(sk, skb);
520 out:
521 return err;
522
523 csum_copy_err:
524 skb_kill_datagram(sk, skb, flags);
525
526 /* Error for blocking case is chosen to masquerade
527 as some normal condition.
528 */
529 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
530 goto out;
531 }
532
533 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
534 struct raw6_sock *rp)
535 {
536 struct sk_buff *skb;
537 int err = 0;
538 int offset;
539 int len;
540 int total_len;
541 __wsum tmp_csum;
542 __sum16 csum;
543
544 if (!rp->checksum)
545 goto send;
546
547 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
548 goto out;
549
550 offset = rp->offset;
551 total_len = inet_sk(sk)->cork.base.length;
552 if (offset >= total_len - 1) {
553 err = -EINVAL;
554 ip6_flush_pending_frames(sk);
555 goto out;
556 }
557
558 /* should be check HW csum miyazawa */
559 if (skb_queue_len(&sk->sk_write_queue) == 1) {
560 /*
561 * Only one fragment on the socket.
562 */
563 tmp_csum = skb->csum;
564 } else {
565 struct sk_buff *csum_skb = NULL;
566 tmp_csum = 0;
567
568 skb_queue_walk(&sk->sk_write_queue, skb) {
569 tmp_csum = csum_add(tmp_csum, skb->csum);
570
571 if (csum_skb)
572 continue;
573
574 len = skb->len - skb_transport_offset(skb);
575 if (offset >= len) {
576 offset -= len;
577 continue;
578 }
579
580 csum_skb = skb;
581 }
582
583 skb = csum_skb;
584 }
585
586 offset += skb_transport_offset(skb);
587 if (skb_copy_bits(skb, offset, &csum, 2))
588 BUG();
589
590 /* in case cksum was not initialized */
591 if (unlikely(csum))
592 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
593
594 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
595 total_len, fl6->flowi6_proto, tmp_csum);
596
597 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
598 csum = CSUM_MANGLED_0;
599
600 if (skb_store_bits(skb, offset, &csum, 2))
601 BUG();
602
603 send:
604 err = ip6_push_pending_frames(sk);
605 out:
606 return err;
607 }
608
609 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
610 struct flowi6 *fl6, struct dst_entry **dstp,
611 unsigned int flags)
612 {
613 struct ipv6_pinfo *np = inet6_sk(sk);
614 struct ipv6hdr *iph;
615 struct sk_buff *skb;
616 int err;
617 struct rt6_info *rt = (struct rt6_info *)*dstp;
618 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
619 int tlen = rt->dst.dev->needed_tailroom;
620
621 if (length > rt->dst.dev->mtu) {
622 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
623 return -EMSGSIZE;
624 }
625 if (flags&MSG_PROBE)
626 goto out;
627
628 skb = sock_alloc_send_skb(sk,
629 length + hlen + tlen + 15,
630 flags & MSG_DONTWAIT, &err);
631 if (skb == NULL)
632 goto error;
633 skb_reserve(skb, hlen);
634
635 skb->protocol = htons(ETH_P_IPV6);
636 skb->priority = sk->sk_priority;
637 skb->mark = sk->sk_mark;
638 skb_dst_set(skb, &rt->dst);
639 *dstp = NULL;
640
641 skb_put(skb, length);
642 skb_reset_network_header(skb);
643 iph = ipv6_hdr(skb);
644
645 skb->ip_summed = CHECKSUM_NONE;
646
647 skb->transport_header = skb->network_header;
648 err = memcpy_fromiovecend((void *)iph, from, 0, length);
649 if (err)
650 goto error_fault;
651
652 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
653 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
654 rt->dst.dev, dst_output);
655 if (err > 0)
656 err = net_xmit_errno(err);
657 if (err)
658 goto error;
659 out:
660 return 0;
661
662 error_fault:
663 err = -EFAULT;
664 kfree_skb(skb);
665 error:
666 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
667 if (err == -ENOBUFS && !np->recverr)
668 err = 0;
669 return err;
670 }
671
672 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg)
673 {
674 struct iovec *iov;
675 u8 __user *type = NULL;
676 u8 __user *code = NULL;
677 u8 len = 0;
678 int probed = 0;
679 int i;
680
681 if (!msg->msg_iov)
682 return 0;
683
684 for (i = 0; i < msg->msg_iovlen; i++) {
685 iov = &msg->msg_iov[i];
686 if (!iov)
687 continue;
688
689 switch (fl6->flowi6_proto) {
690 case IPPROTO_ICMPV6:
691 /* check if one-byte field is readable or not. */
692 if (iov->iov_base && iov->iov_len < 1)
693 break;
694
695 if (!type) {
696 type = iov->iov_base;
697 /* check if code field is readable or not. */
698 if (iov->iov_len > 1)
699 code = type + 1;
700 } else if (!code)
701 code = iov->iov_base;
702
703 if (type && code) {
704 if (get_user(fl6->fl6_icmp_type, type) ||
705 get_user(fl6->fl6_icmp_code, code))
706 return -EFAULT;
707 probed = 1;
708 }
709 break;
710 case IPPROTO_MH:
711 if (iov->iov_base && iov->iov_len < 1)
712 break;
713 /* check if type field is readable or not. */
714 if (iov->iov_len > 2 - len) {
715 u8 __user *p = iov->iov_base;
716 if (get_user(fl6->fl6_mh_type, &p[2 - len]))
717 return -EFAULT;
718 probed = 1;
719 } else
720 len += iov->iov_len;
721
722 break;
723 default:
724 probed = 1;
725 break;
726 }
727 if (probed)
728 break;
729 }
730 return 0;
731 }
732
733 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
734 struct msghdr *msg, size_t len)
735 {
736 struct ipv6_txoptions opt_space;
737 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
738 struct in6_addr *daddr, *final_p, final;
739 struct inet_sock *inet = inet_sk(sk);
740 struct ipv6_pinfo *np = inet6_sk(sk);
741 struct raw6_sock *rp = raw6_sk(sk);
742 struct ipv6_txoptions *opt = NULL;
743 struct ip6_flowlabel *flowlabel = NULL;
744 struct dst_entry *dst = NULL;
745 struct flowi6 fl6;
746 int addr_len = msg->msg_namelen;
747 int hlimit = -1;
748 int tclass = -1;
749 int dontfrag = -1;
750 u16 proto;
751 int err;
752
753 /* Rough check on arithmetic overflow,
754 better check is made in ip6_append_data().
755 */
756 if (len > INT_MAX)
757 return -EMSGSIZE;
758
759 /* Mirror BSD error message compatibility */
760 if (msg->msg_flags & MSG_OOB)
761 return -EOPNOTSUPP;
762
763 /*
764 * Get and verify the address.
765 */
766 memset(&fl6, 0, sizeof(fl6));
767
768 fl6.flowi6_mark = sk->sk_mark;
769
770 if (sin6) {
771 if (addr_len < SIN6_LEN_RFC2133)
772 return -EINVAL;
773
774 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
775 return -EAFNOSUPPORT;
776
777 /* port is the proto value [0..255] carried in nexthdr */
778 proto = ntohs(sin6->sin6_port);
779
780 if (!proto)
781 proto = inet->inet_num;
782 else if (proto != inet->inet_num)
783 return -EINVAL;
784
785 if (proto > 255)
786 return -EINVAL;
787
788 daddr = &sin6->sin6_addr;
789 if (np->sndflow) {
790 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
791 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
792 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
793 if (flowlabel == NULL)
794 return -EINVAL;
795 daddr = &flowlabel->dst;
796 }
797 }
798
799 /*
800 * Otherwise it will be difficult to maintain
801 * sk->sk_dst_cache.
802 */
803 if (sk->sk_state == TCP_ESTABLISHED &&
804 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
805 daddr = &sk->sk_v6_daddr;
806
807 if (addr_len >= sizeof(struct sockaddr_in6) &&
808 sin6->sin6_scope_id &&
809 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
810 fl6.flowi6_oif = sin6->sin6_scope_id;
811 } else {
812 if (sk->sk_state != TCP_ESTABLISHED)
813 return -EDESTADDRREQ;
814
815 proto = inet->inet_num;
816 daddr = &sk->sk_v6_daddr;
817 fl6.flowlabel = np->flow_label;
818 }
819
820 if (fl6.flowi6_oif == 0)
821 fl6.flowi6_oif = sk->sk_bound_dev_if;
822
823 if (msg->msg_controllen) {
824 opt = &opt_space;
825 memset(opt, 0, sizeof(struct ipv6_txoptions));
826 opt->tot_len = sizeof(struct ipv6_txoptions);
827
828 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
829 &hlimit, &tclass, &dontfrag);
830 if (err < 0) {
831 fl6_sock_release(flowlabel);
832 return err;
833 }
834 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
835 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
836 if (flowlabel == NULL)
837 return -EINVAL;
838 }
839 if (!(opt->opt_nflen|opt->opt_flen))
840 opt = NULL;
841 }
842 if (opt == NULL)
843 opt = np->opt;
844 if (flowlabel)
845 opt = fl6_merge_options(&opt_space, flowlabel, opt);
846 opt = ipv6_fixup_options(&opt_space, opt);
847
848 fl6.flowi6_proto = proto;
849 err = rawv6_probe_proto_opt(&fl6, msg);
850 if (err)
851 goto out;
852
853 if (!ipv6_addr_any(daddr))
854 fl6.daddr = *daddr;
855 else
856 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
857 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
858 fl6.saddr = np->saddr;
859
860 final_p = fl6_update_dst(&fl6, opt, &final);
861
862 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
863 fl6.flowi6_oif = np->mcast_oif;
864 else if (!fl6.flowi6_oif)
865 fl6.flowi6_oif = np->ucast_oif;
866 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
867
868 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
869 if (IS_ERR(dst)) {
870 err = PTR_ERR(dst);
871 goto out;
872 }
873 if (hlimit < 0) {
874 if (ipv6_addr_is_multicast(&fl6.daddr))
875 hlimit = np->mcast_hops;
876 else
877 hlimit = np->hop_limit;
878 if (hlimit < 0)
879 hlimit = ip6_dst_hoplimit(dst);
880 }
881
882 if (tclass < 0)
883 tclass = np->tclass;
884
885 if (dontfrag < 0)
886 dontfrag = np->dontfrag;
887
888 if (msg->msg_flags&MSG_CONFIRM)
889 goto do_confirm;
890
891 back_from_confirm:
892 if (inet->hdrincl)
893 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags);
894 else {
895 lock_sock(sk);
896 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
897 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
898 msg->msg_flags, dontfrag);
899
900 if (err)
901 ip6_flush_pending_frames(sk);
902 else if (!(msg->msg_flags & MSG_MORE))
903 err = rawv6_push_pending_frames(sk, &fl6, rp);
904 release_sock(sk);
905 }
906 done:
907 dst_release(dst);
908 out:
909 fl6_sock_release(flowlabel);
910 return err<0?err:len;
911 do_confirm:
912 dst_confirm(dst);
913 if (!(msg->msg_flags & MSG_PROBE) || len)
914 goto back_from_confirm;
915 err = 0;
916 goto done;
917 }
918
919 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
920 char __user *optval, int optlen)
921 {
922 switch (optname) {
923 case ICMPV6_FILTER:
924 if (optlen > sizeof(struct icmp6_filter))
925 optlen = sizeof(struct icmp6_filter);
926 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
927 return -EFAULT;
928 return 0;
929 default:
930 return -ENOPROTOOPT;
931 }
932
933 return 0;
934 }
935
936 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
937 char __user *optval, int __user *optlen)
938 {
939 int len;
940
941 switch (optname) {
942 case ICMPV6_FILTER:
943 if (get_user(len, optlen))
944 return -EFAULT;
945 if (len < 0)
946 return -EINVAL;
947 if (len > sizeof(struct icmp6_filter))
948 len = sizeof(struct icmp6_filter);
949 if (put_user(len, optlen))
950 return -EFAULT;
951 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
952 return -EFAULT;
953 return 0;
954 default:
955 return -ENOPROTOOPT;
956 }
957
958 return 0;
959 }
960
961
962 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
963 char __user *optval, unsigned int optlen)
964 {
965 struct raw6_sock *rp = raw6_sk(sk);
966 int val;
967
968 if (get_user(val, (int __user *)optval))
969 return -EFAULT;
970
971 switch (optname) {
972 case IPV6_CHECKSUM:
973 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
974 level == IPPROTO_IPV6) {
975 /*
976 * RFC3542 tells that IPV6_CHECKSUM socket
977 * option in the IPPROTO_IPV6 level is not
978 * allowed on ICMPv6 sockets.
979 * If you want to set it, use IPPROTO_RAW
980 * level IPV6_CHECKSUM socket option
981 * (Linux extension).
982 */
983 return -EINVAL;
984 }
985
986 /* You may get strange result with a positive odd offset;
987 RFC2292bis agrees with me. */
988 if (val > 0 && (val&1))
989 return -EINVAL;
990 if (val < 0) {
991 rp->checksum = 0;
992 } else {
993 rp->checksum = 1;
994 rp->offset = val;
995 }
996
997 return 0;
998
999 default:
1000 return -ENOPROTOOPT;
1001 }
1002 }
1003
1004 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1005 char __user *optval, unsigned int optlen)
1006 {
1007 switch (level) {
1008 case SOL_RAW:
1009 break;
1010
1011 case SOL_ICMPV6:
1012 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1013 return -EOPNOTSUPP;
1014 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1015 case SOL_IPV6:
1016 if (optname == IPV6_CHECKSUM)
1017 break;
1018 default:
1019 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1020 }
1021
1022 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1023 }
1024
1025 #ifdef CONFIG_COMPAT
1026 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1027 char __user *optval, unsigned int optlen)
1028 {
1029 switch (level) {
1030 case SOL_RAW:
1031 break;
1032 case SOL_ICMPV6:
1033 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1034 return -EOPNOTSUPP;
1035 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1036 case SOL_IPV6:
1037 if (optname == IPV6_CHECKSUM)
1038 break;
1039 default:
1040 return compat_ipv6_setsockopt(sk, level, optname,
1041 optval, optlen);
1042 }
1043 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1044 }
1045 #endif
1046
1047 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1048 char __user *optval, int __user *optlen)
1049 {
1050 struct raw6_sock *rp = raw6_sk(sk);
1051 int val, len;
1052
1053 if (get_user(len,optlen))
1054 return -EFAULT;
1055
1056 switch (optname) {
1057 case IPV6_CHECKSUM:
1058 /*
1059 * We allow getsockopt() for IPPROTO_IPV6-level
1060 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1061 * since RFC3542 is silent about it.
1062 */
1063 if (rp->checksum == 0)
1064 val = -1;
1065 else
1066 val = rp->offset;
1067 break;
1068
1069 default:
1070 return -ENOPROTOOPT;
1071 }
1072
1073 len = min_t(unsigned int, sizeof(int), len);
1074
1075 if (put_user(len, optlen))
1076 return -EFAULT;
1077 if (copy_to_user(optval,&val,len))
1078 return -EFAULT;
1079 return 0;
1080 }
1081
1082 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1083 char __user *optval, int __user *optlen)
1084 {
1085 switch (level) {
1086 case SOL_RAW:
1087 break;
1088
1089 case SOL_ICMPV6:
1090 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1091 return -EOPNOTSUPP;
1092 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1093 case SOL_IPV6:
1094 if (optname == IPV6_CHECKSUM)
1095 break;
1096 default:
1097 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1098 }
1099
1100 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1101 }
1102
1103 #ifdef CONFIG_COMPAT
1104 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1105 char __user *optval, int __user *optlen)
1106 {
1107 switch (level) {
1108 case SOL_RAW:
1109 break;
1110 case SOL_ICMPV6:
1111 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1112 return -EOPNOTSUPP;
1113 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1114 case SOL_IPV6:
1115 if (optname == IPV6_CHECKSUM)
1116 break;
1117 default:
1118 return compat_ipv6_getsockopt(sk, level, optname,
1119 optval, optlen);
1120 }
1121 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1122 }
1123 #endif
1124
1125 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1126 {
1127 switch (cmd) {
1128 case SIOCOUTQ: {
1129 int amount = sk_wmem_alloc_get(sk);
1130
1131 return put_user(amount, (int __user *)arg);
1132 }
1133 case SIOCINQ: {
1134 struct sk_buff *skb;
1135 int amount = 0;
1136
1137 spin_lock_bh(&sk->sk_receive_queue.lock);
1138 skb = skb_peek(&sk->sk_receive_queue);
1139 if (skb != NULL)
1140 amount = skb_tail_pointer(skb) -
1141 skb_transport_header(skb);
1142 spin_unlock_bh(&sk->sk_receive_queue.lock);
1143 return put_user(amount, (int __user *)arg);
1144 }
1145
1146 default:
1147 #ifdef CONFIG_IPV6_MROUTE
1148 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1149 #else
1150 return -ENOIOCTLCMD;
1151 #endif
1152 }
1153 }
1154
1155 #ifdef CONFIG_COMPAT
1156 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1157 {
1158 switch (cmd) {
1159 case SIOCOUTQ:
1160 case SIOCINQ:
1161 return -ENOIOCTLCMD;
1162 default:
1163 #ifdef CONFIG_IPV6_MROUTE
1164 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1165 #else
1166 return -ENOIOCTLCMD;
1167 #endif
1168 }
1169 }
1170 #endif
1171
1172 static void rawv6_close(struct sock *sk, long timeout)
1173 {
1174 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1175 ip6_ra_control(sk, -1);
1176 ip6mr_sk_done(sk);
1177 sk_common_release(sk);
1178 }
1179
1180 static void raw6_destroy(struct sock *sk)
1181 {
1182 lock_sock(sk);
1183 ip6_flush_pending_frames(sk);
1184 release_sock(sk);
1185
1186 inet6_destroy_sock(sk);
1187 }
1188
1189 static int rawv6_init_sk(struct sock *sk)
1190 {
1191 struct raw6_sock *rp = raw6_sk(sk);
1192
1193 switch (inet_sk(sk)->inet_num) {
1194 case IPPROTO_ICMPV6:
1195 rp->checksum = 1;
1196 rp->offset = 2;
1197 break;
1198 case IPPROTO_MH:
1199 rp->checksum = 1;
1200 rp->offset = 4;
1201 break;
1202 default:
1203 break;
1204 }
1205 return 0;
1206 }
1207
1208 struct proto rawv6_prot = {
1209 .name = "RAWv6",
1210 .owner = THIS_MODULE,
1211 .close = rawv6_close,
1212 .destroy = raw6_destroy,
1213 .connect = ip6_datagram_connect,
1214 .disconnect = udp_disconnect,
1215 .ioctl = rawv6_ioctl,
1216 .init = rawv6_init_sk,
1217 .setsockopt = rawv6_setsockopt,
1218 .getsockopt = rawv6_getsockopt,
1219 .sendmsg = rawv6_sendmsg,
1220 .recvmsg = rawv6_recvmsg,
1221 .bind = rawv6_bind,
1222 .backlog_rcv = rawv6_rcv_skb,
1223 .hash = raw_hash_sk,
1224 .unhash = raw_unhash_sk,
1225 .obj_size = sizeof(struct raw6_sock),
1226 .h.raw_hash = &raw_v6_hashinfo,
1227 #ifdef CONFIG_COMPAT
1228 .compat_setsockopt = compat_rawv6_setsockopt,
1229 .compat_getsockopt = compat_rawv6_getsockopt,
1230 .compat_ioctl = compat_rawv6_ioctl,
1231 #endif
1232 };
1233
1234 #ifdef CONFIG_PROC_FS
1235 static int raw6_seq_show(struct seq_file *seq, void *v)
1236 {
1237 if (v == SEQ_START_TOKEN) {
1238 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1239 } else {
1240 struct sock *sp = v;
1241 __u16 srcp = inet_sk(sp)->inet_num;
1242 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1243 raw_seq_private(seq)->bucket);
1244 }
1245 return 0;
1246 }
1247
1248 static const struct seq_operations raw6_seq_ops = {
1249 .start = raw_seq_start,
1250 .next = raw_seq_next,
1251 .stop = raw_seq_stop,
1252 .show = raw6_seq_show,
1253 };
1254
1255 static int raw6_seq_open(struct inode *inode, struct file *file)
1256 {
1257 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1258 }
1259
1260 static const struct file_operations raw6_seq_fops = {
1261 .owner = THIS_MODULE,
1262 .open = raw6_seq_open,
1263 .read = seq_read,
1264 .llseek = seq_lseek,
1265 .release = seq_release_net,
1266 };
1267
1268 static int __net_init raw6_init_net(struct net *net)
1269 {
1270 if (!proc_create("raw6", S_IRUGO, net->proc_net, &raw6_seq_fops))
1271 return -ENOMEM;
1272
1273 return 0;
1274 }
1275
1276 static void __net_exit raw6_exit_net(struct net *net)
1277 {
1278 remove_proc_entry("raw6", net->proc_net);
1279 }
1280
1281 static struct pernet_operations raw6_net_ops = {
1282 .init = raw6_init_net,
1283 .exit = raw6_exit_net,
1284 };
1285
1286 int __init raw6_proc_init(void)
1287 {
1288 return register_pernet_subsys(&raw6_net_ops);
1289 }
1290
1291 void raw6_proc_exit(void)
1292 {
1293 unregister_pernet_subsys(&raw6_net_ops);
1294 }
1295 #endif /* CONFIG_PROC_FS */
1296
1297 /* Same as inet6_dgram_ops, sans udp_poll. */
1298 static const struct proto_ops inet6_sockraw_ops = {
1299 .family = PF_INET6,
1300 .owner = THIS_MODULE,
1301 .release = inet6_release,
1302 .bind = inet6_bind,
1303 .connect = inet_dgram_connect, /* ok */
1304 .socketpair = sock_no_socketpair, /* a do nothing */
1305 .accept = sock_no_accept, /* a do nothing */
1306 .getname = inet6_getname,
1307 .poll = datagram_poll, /* ok */
1308 .ioctl = inet6_ioctl, /* must change */
1309 .listen = sock_no_listen, /* ok */
1310 .shutdown = inet_shutdown, /* ok */
1311 .setsockopt = sock_common_setsockopt, /* ok */
1312 .getsockopt = sock_common_getsockopt, /* ok */
1313 .sendmsg = inet_sendmsg, /* ok */
1314 .recvmsg = sock_common_recvmsg, /* ok */
1315 .mmap = sock_no_mmap,
1316 .sendpage = sock_no_sendpage,
1317 #ifdef CONFIG_COMPAT
1318 .compat_setsockopt = compat_sock_common_setsockopt,
1319 .compat_getsockopt = compat_sock_common_getsockopt,
1320 #endif
1321 };
1322
1323 static struct inet_protosw rawv6_protosw = {
1324 .type = SOCK_RAW,
1325 .protocol = IPPROTO_IP, /* wild card */
1326 .prot = &rawv6_prot,
1327 .ops = &inet6_sockraw_ops,
1328 .no_check = UDP_CSUM_DEFAULT,
1329 .flags = INET_PROTOSW_REUSE,
1330 };
1331
1332 int __init rawv6_init(void)
1333 {
1334 int ret;
1335
1336 ret = inet6_register_protosw(&rawv6_protosw);
1337 if (ret)
1338 goto out;
1339 out:
1340 return ret;
1341 }
1342
1343 void rawv6_exit(void)
1344 {
1345 inet6_unregister_protosw(&rawv6_protosw);
1346 }
(deprecated) Btrfs devel tree