?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
spi-topcliff-pch: Fix issue for transmitting over 4KByte
[zen-stable.git] / net / ipv6 / raw.c
blobd02f7e4dd6113c8c23cb8a3f7ee9be9cbae37685
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 defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
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 static struct raw_hashinfo raw_v6_hashinfo = {
67 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
68 };
69
70 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
71 unsigned short num, const struct in6_addr *loc_addr,
72 const struct in6_addr *rmt_addr, int dif)
73 {
74 struct hlist_node *node;
75 int is_multicast = ipv6_addr_is_multicast(loc_addr);
76
77 sk_for_each_from(sk, node)
78 if (inet_sk(sk)->inet_num == num) {
79 struct ipv6_pinfo *np = inet6_sk(sk);
80
81 if (!net_eq(sock_net(sk), net))
82 continue;
83
84 if (!ipv6_addr_any(&np->daddr) &&
85 !ipv6_addr_equal(&np->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(&np->rcv_saddr)) {
92 if (ipv6_addr_equal(&np->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 __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
111 {
112 struct icmp6hdr *icmph;
113 struct raw6_sock *rp = raw6_sk(sk);
114
115 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
116 __u32 *data = &rp->filter.data[0];
117 int bit_nr;
118
119 icmph = (struct icmp6hdr *) skb->data;
120 bit_nr = icmph->icmp6_type;
121
122 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
123 }
124 return 0;
125 }
126
127 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
128 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
129
130 static mh_filter_t __rcu *mh_filter __read_mostly;
131
132 int rawv6_mh_filter_register(mh_filter_t filter)
133 {
134 rcu_assign_pointer(mh_filter, filter);
135 return 0;
136 }
137 EXPORT_SYMBOL(rawv6_mh_filter_register);
138
139 int rawv6_mh_filter_unregister(mh_filter_t filter)
140 {
141 RCU_INIT_POINTER(mh_filter, NULL);
142 synchronize_rcu();
143 return 0;
144 }
145 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
146
147 #endif
148
149 /*
150 * demultiplex raw sockets.
151 * (should consider queueing the skb in the sock receive_queue
152 * without calling rawv6.c)
153 *
154 * Caller owns SKB so we must make clones.
155 */
156 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
157 {
158 const struct in6_addr *saddr;
159 const struct in6_addr *daddr;
160 struct sock *sk;
161 int delivered = 0;
162 __u8 hash;
163 struct net *net;
164
165 saddr = &ipv6_hdr(skb)->saddr;
166 daddr = saddr + 1;
167
168 hash = nexthdr & (MAX_INET_PROTOS - 1);
169
170 read_lock(&raw_v6_hashinfo.lock);
171 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
172
173 if (sk == NULL)
174 goto out;
175
176 net = dev_net(skb->dev);
177 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
178
179 while (sk) {
180 int filtered;
181
182 delivered = 1;
183 switch (nexthdr) {
184 case IPPROTO_ICMPV6:
185 filtered = icmpv6_filter(sk, skb);
186 break;
187
188 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
189 case IPPROTO_MH:
190 {
191 /* XXX: To validate MH only once for each packet,
192 * this is placed here. It should be after checking
193 * xfrm policy, however it doesn't. The checking xfrm
194 * policy is placed in rawv6_rcv() because it is
195 * required for each socket.
196 */
197 mh_filter_t *filter;
198
199 filter = rcu_dereference(mh_filter);
200 filtered = filter ? (*filter)(sk, skb) : 0;
201 break;
202 }
203 #endif
204 default:
205 filtered = 0;
206 break;
207 }
208
209 if (filtered < 0)
210 break;
211 if (filtered == 0) {
212 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
213
214 /* Not releasing hash table! */
215 if (clone) {
216 nf_reset(clone);
217 rawv6_rcv(sk, clone);
218 }
219 }
220 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
221 IP6CB(skb)->iif);
222 }
223 out:
224 read_unlock(&raw_v6_hashinfo.lock);
225 return delivered;
226 }
227
228 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
229 {
230 struct sock *raw_sk;
231
232 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
233 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
234 raw_sk = NULL;
235
236 return raw_sk != NULL;
237 }
238
239 /* This cleans up af_inet6 a bit. -DaveM */
240 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
241 {
242 struct inet_sock *inet = inet_sk(sk);
243 struct ipv6_pinfo *np = inet6_sk(sk);
244 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
245 __be32 v4addr = 0;
246 int addr_type;
247 int err;
248
249 if (addr_len < SIN6_LEN_RFC2133)
250 return -EINVAL;
251 addr_type = ipv6_addr_type(&addr->sin6_addr);
252
253 /* Raw sockets are IPv6 only */
254 if (addr_type == IPV6_ADDR_MAPPED)
255 return -EADDRNOTAVAIL;
256
257 lock_sock(sk);
258
259 err = -EINVAL;
260 if (sk->sk_state != TCP_CLOSE)
261 goto out;
262
263 rcu_read_lock();
264 /* Check if the address belongs to the host. */
265 if (addr_type != IPV6_ADDR_ANY) {
266 struct net_device *dev = NULL;
267
268 if (addr_type & IPV6_ADDR_LINKLOCAL) {
269 if (addr_len >= sizeof(struct sockaddr_in6) &&
270 addr->sin6_scope_id) {
271 /* Override any existing binding, if another
272 * one is supplied by user.
273 */
274 sk->sk_bound_dev_if = addr->sin6_scope_id;
275 }
276
277 /* Binding to link-local address requires an interface */
278 if (!sk->sk_bound_dev_if)
279 goto out_unlock;
280
281 err = -ENODEV;
282 dev = dev_get_by_index_rcu(sock_net(sk),
283 sk->sk_bound_dev_if);
284 if (!dev)
285 goto out_unlock;
286 }
287
288 /* ipv4 addr of the socket is invalid. Only the
289 * unspecified and mapped address have a v4 equivalent.
290 */
291 v4addr = LOOPBACK4_IPV6;
292 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
293 err = -EADDRNOTAVAIL;
294 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
295 dev, 0)) {
296 goto out_unlock;
297 }
298 }
299 }
300
301 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
302 np->rcv_saddr = addr->sin6_addr;
303 if (!(addr_type & IPV6_ADDR_MULTICAST))
304 np->saddr = addr->sin6_addr;
305 err = 0;
306 out_unlock:
307 rcu_read_unlock();
308 out:
309 release_sock(sk);
310 return err;
311 }
312
313 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
314 struct inet6_skb_parm *opt,
315 u8 type, u8 code, int offset, __be32 info)
316 {
317 struct inet_sock *inet = inet_sk(sk);
318 struct ipv6_pinfo *np = inet6_sk(sk);
319 int err;
320 int harderr;
321
322 /* Report error on raw socket, if:
323 1. User requested recverr.
324 2. Socket is connected (otherwise the error indication
325 is useless without recverr and error is hard.
326 */
327 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
328 return;
329
330 harderr = icmpv6_err_convert(type, code, &err);
331 if (type == ICMPV6_PKT_TOOBIG)
332 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
333
334 if (np->recverr) {
335 u8 *payload = skb->data;
336 if (!inet->hdrincl)
337 payload += offset;
338 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
339 }
340
341 if (np->recverr || harderr) {
342 sk->sk_err = err;
343 sk->sk_error_report(sk);
344 }
345 }
346
347 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
348 u8 type, u8 code, int inner_offset, __be32 info)
349 {
350 struct sock *sk;
351 int hash;
352 const struct in6_addr *saddr, *daddr;
353 struct net *net;
354
355 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
356
357 read_lock(&raw_v6_hashinfo.lock);
358 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
359 if (sk != NULL) {
360 /* Note: ipv6_hdr(skb) != skb->data */
361 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
362 saddr = &ip6h->saddr;
363 daddr = &ip6h->daddr;
364 net = dev_net(skb->dev);
365
366 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
367 IP6CB(skb)->iif))) {
368 rawv6_err(sk, skb, NULL, type, code,
369 inner_offset, info);
370 sk = sk_next(sk);
371 }
372 }
373 read_unlock(&raw_v6_hashinfo.lock);
374 }
375
376 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
377 {
378 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
379 skb_checksum_complete(skb)) {
380 atomic_inc(&sk->sk_drops);
381 kfree_skb(skb);
382 return NET_RX_DROP;
383 }
384
385 /* Charge it to the socket. */
386 skb_dst_drop(skb);
387 if (sock_queue_rcv_skb(sk, skb) < 0) {
388 kfree_skb(skb);
389 return NET_RX_DROP;
390 }
391
392 return 0;
393 }
394
395 /*
396 * This is next to useless...
397 * if we demultiplex in network layer we don't need the extra call
398 * just to queue the skb...
399 * maybe we could have the network decide upon a hint if it
400 * should call raw_rcv for demultiplexing
401 */
402 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
403 {
404 struct inet_sock *inet = inet_sk(sk);
405 struct raw6_sock *rp = raw6_sk(sk);
406
407 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
408 atomic_inc(&sk->sk_drops);
409 kfree_skb(skb);
410 return NET_RX_DROP;
411 }
412
413 if (!rp->checksum)
414 skb->ip_summed = CHECKSUM_UNNECESSARY;
415
416 if (skb->ip_summed == CHECKSUM_COMPLETE) {
417 skb_postpull_rcsum(skb, skb_network_header(skb),
418 skb_network_header_len(skb));
419 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
420 &ipv6_hdr(skb)->daddr,
421 skb->len, inet->inet_num, skb->csum))
422 skb->ip_summed = CHECKSUM_UNNECESSARY;
423 }
424 if (!skb_csum_unnecessary(skb))
425 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
426 &ipv6_hdr(skb)->daddr,
427 skb->len,
428 inet->inet_num, 0));
429
430 if (inet->hdrincl) {
431 if (skb_checksum_complete(skb)) {
432 atomic_inc(&sk->sk_drops);
433 kfree_skb(skb);
434 return NET_RX_DROP;
435 }
436 }
437
438 rawv6_rcv_skb(sk, skb);
439 return 0;
440 }
441
442
443 /*
444 * This should be easy, if there is something there
445 * we return it, otherwise we block.
446 */
447
448 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
449 struct msghdr *msg, size_t len,
450 int noblock, int flags, int *addr_len)
451 {
452 struct ipv6_pinfo *np = inet6_sk(sk);
453 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
454 struct sk_buff *skb;
455 size_t copied;
456 int err;
457
458 if (flags & MSG_OOB)
459 return -EOPNOTSUPP;
460
461 if (addr_len)
462 *addr_len=sizeof(*sin6);
463
464 if (flags & MSG_ERRQUEUE)
465 return ipv6_recv_error(sk, msg, len);
466
467 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
468 return ipv6_recv_rxpmtu(sk, msg, len);
469
470 skb = skb_recv_datagram(sk, flags, noblock, &err);
471 if (!skb)
472 goto out;
473
474 copied = skb->len;
475 if (copied > len) {
476 copied = len;
477 msg->msg_flags |= MSG_TRUNC;
478 }
479
480 if (skb_csum_unnecessary(skb)) {
481 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
482 } else if (msg->msg_flags&MSG_TRUNC) {
483 if (__skb_checksum_complete(skb))
484 goto csum_copy_err;
485 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
486 } else {
487 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
488 if (err == -EINVAL)
489 goto csum_copy_err;
490 }
491 if (err)
492 goto out_free;
493
494 /* Copy the address. */
495 if (sin6) {
496 sin6->sin6_family = AF_INET6;
497 sin6->sin6_port = 0;
498 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
499 sin6->sin6_flowinfo = 0;
500 sin6->sin6_scope_id = 0;
501 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
502 sin6->sin6_scope_id = IP6CB(skb)->iif;
503 }
504
505 sock_recv_ts_and_drops(msg, sk, skb);
506
507 if (np->rxopt.all)
508 datagram_recv_ctl(sk, msg, skb);
509
510 err = copied;
511 if (flags & MSG_TRUNC)
512 err = skb->len;
513
514 out_free:
515 skb_free_datagram(sk, skb);
516 out:
517 return err;
518
519 csum_copy_err:
520 skb_kill_datagram(sk, skb, flags);
521
522 /* Error for blocking case is chosen to masquerade
523 as some normal condition.
524 */
525 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
526 goto out;
527 }
528
529 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
530 struct raw6_sock *rp)
531 {
532 struct sk_buff *skb;
533 int err = 0;
534 int offset;
535 int len;
536 int total_len;
537 __wsum tmp_csum;
538 __sum16 csum;
539
540 if (!rp->checksum)
541 goto send;
542
543 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
544 goto out;
545
546 offset = rp->offset;
547 total_len = inet_sk(sk)->cork.base.length;
548 if (offset >= total_len - 1) {
549 err = -EINVAL;
550 ip6_flush_pending_frames(sk);
551 goto out;
552 }
553
554 /* should be check HW csum miyazawa */
555 if (skb_queue_len(&sk->sk_write_queue) == 1) {
556 /*
557 * Only one fragment on the socket.
558 */
559 tmp_csum = skb->csum;
560 } else {
561 struct sk_buff *csum_skb = NULL;
562 tmp_csum = 0;
563
564 skb_queue_walk(&sk->sk_write_queue, skb) {
565 tmp_csum = csum_add(tmp_csum, skb->csum);
566
567 if (csum_skb)
568 continue;
569
570 len = skb->len - skb_transport_offset(skb);
571 if (offset >= len) {
572 offset -= len;
573 continue;
574 }
575
576 csum_skb = skb;
577 }
578
579 skb = csum_skb;
580 }
581
582 offset += skb_transport_offset(skb);
583 if (skb_copy_bits(skb, offset, &csum, 2))
584 BUG();
585
586 /* in case cksum was not initialized */
587 if (unlikely(csum))
588 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
589
590 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
591 total_len, fl6->flowi6_proto, tmp_csum);
592
593 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
594 csum = CSUM_MANGLED_0;
595
596 if (skb_store_bits(skb, offset, &csum, 2))
597 BUG();
598
599 send:
600 err = ip6_push_pending_frames(sk);
601 out:
602 return err;
603 }
604
605 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
606 struct flowi6 *fl6, struct dst_entry **dstp,
607 unsigned int flags)
608 {
609 struct ipv6_pinfo *np = inet6_sk(sk);
610 struct ipv6hdr *iph;
611 struct sk_buff *skb;
612 int err;
613 struct rt6_info *rt = (struct rt6_info *)*dstp;
614 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
615 int tlen = rt->dst.dev->needed_tailroom;
616
617 if (length > rt->dst.dev->mtu) {
618 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
619 return -EMSGSIZE;
620 }
621 if (flags&MSG_PROBE)
622 goto out;
623
624 skb = sock_alloc_send_skb(sk,
625 length + hlen + tlen + 15,
626 flags & MSG_DONTWAIT, &err);
627 if (skb == NULL)
628 goto error;
629 skb_reserve(skb, hlen);
630
631 skb->priority = sk->sk_priority;
632 skb->mark = sk->sk_mark;
633 skb_dst_set(skb, &rt->dst);
634 *dstp = NULL;
635
636 skb_put(skb, length);
637 skb_reset_network_header(skb);
638 iph = ipv6_hdr(skb);
639
640 skb->ip_summed = CHECKSUM_NONE;
641
642 skb->transport_header = skb->network_header;
643 err = memcpy_fromiovecend((void *)iph, from, 0, length);
644 if (err)
645 goto error_fault;
646
647 IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
648 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
649 rt->dst.dev, dst_output);
650 if (err > 0)
651 err = net_xmit_errno(err);
652 if (err)
653 goto error;
654 out:
655 return 0;
656
657 error_fault:
658 err = -EFAULT;
659 kfree_skb(skb);
660 error:
661 IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
662 if (err == -ENOBUFS && !np->recverr)
663 err = 0;
664 return err;
665 }
666
667 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg)
668 {
669 struct iovec *iov;
670 u8 __user *type = NULL;
671 u8 __user *code = NULL;
672 u8 len = 0;
673 int probed = 0;
674 int i;
675
676 if (!msg->msg_iov)
677 return 0;
678
679 for (i = 0; i < msg->msg_iovlen; i++) {
680 iov = &msg->msg_iov[i];
681 if (!iov)
682 continue;
683
684 switch (fl6->flowi6_proto) {
685 case IPPROTO_ICMPV6:
686 /* check if one-byte field is readable or not. */
687 if (iov->iov_base && iov->iov_len < 1)
688 break;
689
690 if (!type) {
691 type = iov->iov_base;
692 /* check if code field is readable or not. */
693 if (iov->iov_len > 1)
694 code = type + 1;
695 } else if (!code)
696 code = iov->iov_base;
697
698 if (type && code) {
699 if (get_user(fl6->fl6_icmp_type, type) ||
700 get_user(fl6->fl6_icmp_code, code))
701 return -EFAULT;
702 probed = 1;
703 }
704 break;
705 case IPPROTO_MH:
706 if (iov->iov_base && iov->iov_len < 1)
707 break;
708 /* check if type field is readable or not. */
709 if (iov->iov_len > 2 - len) {
710 u8 __user *p = iov->iov_base;
711 if (get_user(fl6->fl6_mh_type, &p[2 - len]))
712 return -EFAULT;
713 probed = 1;
714 } else
715 len += iov->iov_len;
716
717 break;
718 default:
719 probed = 1;
720 break;
721 }
722 if (probed)
723 break;
724 }
725 return 0;
726 }
727
728 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
729 struct msghdr *msg, size_t len)
730 {
731 struct ipv6_txoptions opt_space;
732 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
733 struct in6_addr *daddr, *final_p, final;
734 struct inet_sock *inet = inet_sk(sk);
735 struct ipv6_pinfo *np = inet6_sk(sk);
736 struct raw6_sock *rp = raw6_sk(sk);
737 struct ipv6_txoptions *opt = NULL;
738 struct ip6_flowlabel *flowlabel = NULL;
739 struct dst_entry *dst = NULL;
740 struct flowi6 fl6;
741 int addr_len = msg->msg_namelen;
742 int hlimit = -1;
743 int tclass = -1;
744 int dontfrag = -1;
745 u16 proto;
746 int err;
747
748 /* Rough check on arithmetic overflow,
749 better check is made in ip6_append_data().
750 */
751 if (len > INT_MAX)
752 return -EMSGSIZE;
753
754 /* Mirror BSD error message compatibility */
755 if (msg->msg_flags & MSG_OOB)
756 return -EOPNOTSUPP;
757
758 /*
759 * Get and verify the address.
760 */
761 memset(&fl6, 0, sizeof(fl6));
762
763 fl6.flowi6_mark = sk->sk_mark;
764
765 if (sin6) {
766 if (addr_len < SIN6_LEN_RFC2133)
767 return -EINVAL;
768
769 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
770 return -EAFNOSUPPORT;
771
772 /* port is the proto value [0..255] carried in nexthdr */
773 proto = ntohs(sin6->sin6_port);
774
775 if (!proto)
776 proto = inet->inet_num;
777 else if (proto != inet->inet_num)
778 return -EINVAL;
779
780 if (proto > 255)
781 return -EINVAL;
782
783 daddr = &sin6->sin6_addr;
784 if (np->sndflow) {
785 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
786 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
787 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
788 if (flowlabel == NULL)
789 return -EINVAL;
790 daddr = &flowlabel->dst;
791 }
792 }
793
794 /*
795 * Otherwise it will be difficult to maintain
796 * sk->sk_dst_cache.
797 */
798 if (sk->sk_state == TCP_ESTABLISHED &&
799 ipv6_addr_equal(daddr, &np->daddr))
800 daddr = &np->daddr;
801
802 if (addr_len >= sizeof(struct sockaddr_in6) &&
803 sin6->sin6_scope_id &&
804 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
805 fl6.flowi6_oif = sin6->sin6_scope_id;
806 } else {
807 if (sk->sk_state != TCP_ESTABLISHED)
808 return -EDESTADDRREQ;
809
810 proto = inet->inet_num;
811 daddr = &np->daddr;
812 fl6.flowlabel = np->flow_label;
813 }
814
815 if (fl6.flowi6_oif == 0)
816 fl6.flowi6_oif = sk->sk_bound_dev_if;
817
818 if (msg->msg_controllen) {
819 opt = &opt_space;
820 memset(opt, 0, sizeof(struct ipv6_txoptions));
821 opt->tot_len = sizeof(struct ipv6_txoptions);
822
823 err = datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
824 &hlimit, &tclass, &dontfrag);
825 if (err < 0) {
826 fl6_sock_release(flowlabel);
827 return err;
828 }
829 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
830 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
831 if (flowlabel == NULL)
832 return -EINVAL;
833 }
834 if (!(opt->opt_nflen|opt->opt_flen))
835 opt = NULL;
836 }
837 if (opt == NULL)
838 opt = np->opt;
839 if (flowlabel)
840 opt = fl6_merge_options(&opt_space, flowlabel, opt);
841 opt = ipv6_fixup_options(&opt_space, opt);
842
843 fl6.flowi6_proto = proto;
844 err = rawv6_probe_proto_opt(&fl6, msg);
845 if (err)
846 goto out;
847
848 if (!ipv6_addr_any(daddr))
849 fl6.daddr = *daddr;
850 else
851 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
852 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
853 fl6.saddr = np->saddr;
854
855 final_p = fl6_update_dst(&fl6, opt, &final);
856
857 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
858 fl6.flowi6_oif = np->mcast_oif;
859 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
860
861 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
862 if (IS_ERR(dst)) {
863 err = PTR_ERR(dst);
864 goto out;
865 }
866 if (hlimit < 0) {
867 if (ipv6_addr_is_multicast(&fl6.daddr))
868 hlimit = np->mcast_hops;
869 else
870 hlimit = np->hop_limit;
871 if (hlimit < 0)
872 hlimit = ip6_dst_hoplimit(dst);
873 }
874
875 if (tclass < 0)
876 tclass = np->tclass;
877
878 if (dontfrag < 0)
879 dontfrag = np->dontfrag;
880
881 if (msg->msg_flags&MSG_CONFIRM)
882 goto do_confirm;
883
884 back_from_confirm:
885 if (inet->hdrincl)
886 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags);
887 else {
888 lock_sock(sk);
889 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
890 len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
891 msg->msg_flags, dontfrag);
892
893 if (err)
894 ip6_flush_pending_frames(sk);
895 else if (!(msg->msg_flags & MSG_MORE))
896 err = rawv6_push_pending_frames(sk, &fl6, rp);
897 release_sock(sk);
898 }
899 done:
900 dst_release(dst);
901 out:
902 fl6_sock_release(flowlabel);
903 return err<0?err:len;
904 do_confirm:
905 dst_confirm(dst);
906 if (!(msg->msg_flags & MSG_PROBE) || len)
907 goto back_from_confirm;
908 err = 0;
909 goto done;
910 }
911
912 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
913 char __user *optval, int optlen)
914 {
915 switch (optname) {
916 case ICMPV6_FILTER:
917 if (optlen > sizeof(struct icmp6_filter))
918 optlen = sizeof(struct icmp6_filter);
919 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
920 return -EFAULT;
921 return 0;
922 default:
923 return -ENOPROTOOPT;
924 }
925
926 return 0;
927 }
928
929 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
930 char __user *optval, int __user *optlen)
931 {
932 int len;
933
934 switch (optname) {
935 case ICMPV6_FILTER:
936 if (get_user(len, optlen))
937 return -EFAULT;
938 if (len < 0)
939 return -EINVAL;
940 if (len > sizeof(struct icmp6_filter))
941 len = sizeof(struct icmp6_filter);
942 if (put_user(len, optlen))
943 return -EFAULT;
944 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
945 return -EFAULT;
946 return 0;
947 default:
948 return -ENOPROTOOPT;
949 }
950
951 return 0;
952 }
953
954
955 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
956 char __user *optval, unsigned int optlen)
957 {
958 struct raw6_sock *rp = raw6_sk(sk);
959 int val;
960
961 if (get_user(val, (int __user *)optval))
962 return -EFAULT;
963
964 switch (optname) {
965 case IPV6_CHECKSUM:
966 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
967 level == IPPROTO_IPV6) {
968 /*
969 * RFC3542 tells that IPV6_CHECKSUM socket
970 * option in the IPPROTO_IPV6 level is not
971 * allowed on ICMPv6 sockets.
972 * If you want to set it, use IPPROTO_RAW
973 * level IPV6_CHECKSUM socket option
974 * (Linux extension).
975 */
976 return -EINVAL;
977 }
978
979 /* You may get strange result with a positive odd offset;
980 RFC2292bis agrees with me. */
981 if (val > 0 && (val&1))
982 return -EINVAL;
983 if (val < 0) {
984 rp->checksum = 0;
985 } else {
986 rp->checksum = 1;
987 rp->offset = val;
988 }
989
990 return 0;
991
992 default:
993 return -ENOPROTOOPT;
994 }
995 }
996
997 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
998 char __user *optval, unsigned int optlen)
999 {
1000 switch (level) {
1001 case SOL_RAW:
1002 break;
1003
1004 case SOL_ICMPV6:
1005 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1006 return -EOPNOTSUPP;
1007 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1008 case SOL_IPV6:
1009 if (optname == IPV6_CHECKSUM)
1010 break;
1011 default:
1012 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1013 }
1014
1015 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1016 }
1017
1018 #ifdef CONFIG_COMPAT
1019 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1020 char __user *optval, unsigned int optlen)
1021 {
1022 switch (level) {
1023 case SOL_RAW:
1024 break;
1025 case SOL_ICMPV6:
1026 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1027 return -EOPNOTSUPP;
1028 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1029 case SOL_IPV6:
1030 if (optname == IPV6_CHECKSUM)
1031 break;
1032 default:
1033 return compat_ipv6_setsockopt(sk, level, optname,
1034 optval, optlen);
1035 }
1036 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1037 }
1038 #endif
1039
1040 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1041 char __user *optval, int __user *optlen)
1042 {
1043 struct raw6_sock *rp = raw6_sk(sk);
1044 int val, len;
1045
1046 if (get_user(len,optlen))
1047 return -EFAULT;
1048
1049 switch (optname) {
1050 case IPV6_CHECKSUM:
1051 /*
1052 * We allow getsockopt() for IPPROTO_IPV6-level
1053 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1054 * since RFC3542 is silent about it.
1055 */
1056 if (rp->checksum == 0)
1057 val = -1;
1058 else
1059 val = rp->offset;
1060 break;
1061
1062 default:
1063 return -ENOPROTOOPT;
1064 }
1065
1066 len = min_t(unsigned int, sizeof(int), len);
1067
1068 if (put_user(len, optlen))
1069 return -EFAULT;
1070 if (copy_to_user(optval,&val,len))
1071 return -EFAULT;
1072 return 0;
1073 }
1074
1075 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1076 char __user *optval, int __user *optlen)
1077 {
1078 switch (level) {
1079 case SOL_RAW:
1080 break;
1081
1082 case SOL_ICMPV6:
1083 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1084 return -EOPNOTSUPP;
1085 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1086 case SOL_IPV6:
1087 if (optname == IPV6_CHECKSUM)
1088 break;
1089 default:
1090 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1091 }
1092
1093 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1094 }
1095
1096 #ifdef CONFIG_COMPAT
1097 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1098 char __user *optval, int __user *optlen)
1099 {
1100 switch (level) {
1101 case SOL_RAW:
1102 break;
1103 case SOL_ICMPV6:
1104 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1105 return -EOPNOTSUPP;
1106 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1107 case SOL_IPV6:
1108 if (optname == IPV6_CHECKSUM)
1109 break;
1110 default:
1111 return compat_ipv6_getsockopt(sk, level, optname,
1112 optval, optlen);
1113 }
1114 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1115 }
1116 #endif
1117
1118 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1119 {
1120 switch (cmd) {
1121 case SIOCOUTQ: {
1122 int amount = sk_wmem_alloc_get(sk);
1123
1124 return put_user(amount, (int __user *)arg);
1125 }
1126 case SIOCINQ: {
1127 struct sk_buff *skb;
1128 int amount = 0;
1129
1130 spin_lock_bh(&sk->sk_receive_queue.lock);
1131 skb = skb_peek(&sk->sk_receive_queue);
1132 if (skb != NULL)
1133 amount = skb->tail - skb->transport_header;
1134 spin_unlock_bh(&sk->sk_receive_queue.lock);
1135 return put_user(amount, (int __user *)arg);
1136 }
1137
1138 default:
1139 #ifdef CONFIG_IPV6_MROUTE
1140 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1141 #else
1142 return -ENOIOCTLCMD;
1143 #endif
1144 }
1145 }
1146
1147 #ifdef CONFIG_COMPAT
1148 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1149 {
1150 switch (cmd) {
1151 case SIOCOUTQ:
1152 case SIOCINQ:
1153 return -ENOIOCTLCMD;
1154 default:
1155 #ifdef CONFIG_IPV6_MROUTE
1156 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1157 #else
1158 return -ENOIOCTLCMD;
1159 #endif
1160 }
1161 }
1162 #endif
1163
1164 static void rawv6_close(struct sock *sk, long timeout)
1165 {
1166 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1167 ip6_ra_control(sk, -1);
1168 ip6mr_sk_done(sk);
1169 sk_common_release(sk);
1170 }
1171
1172 static void raw6_destroy(struct sock *sk)
1173 {
1174 lock_sock(sk);
1175 ip6_flush_pending_frames(sk);
1176 release_sock(sk);
1177
1178 inet6_destroy_sock(sk);
1179 }
1180
1181 static int rawv6_init_sk(struct sock *sk)
1182 {
1183 struct raw6_sock *rp = raw6_sk(sk);
1184
1185 switch (inet_sk(sk)->inet_num) {
1186 case IPPROTO_ICMPV6:
1187 rp->checksum = 1;
1188 rp->offset = 2;
1189 break;
1190 case IPPROTO_MH:
1191 rp->checksum = 1;
1192 rp->offset = 4;
1193 break;
1194 default:
1195 break;
1196 }
1197 return 0;
1198 }
1199
1200 struct proto rawv6_prot = {
1201 .name = "RAWv6",
1202 .owner = THIS_MODULE,
1203 .close = rawv6_close,
1204 .destroy = raw6_destroy,
1205 .connect = ip6_datagram_connect,
1206 .disconnect = udp_disconnect,
1207 .ioctl = rawv6_ioctl,
1208 .init = rawv6_init_sk,
1209 .setsockopt = rawv6_setsockopt,
1210 .getsockopt = rawv6_getsockopt,
1211 .sendmsg = rawv6_sendmsg,
1212 .recvmsg = rawv6_recvmsg,
1213 .bind = rawv6_bind,
1214 .backlog_rcv = rawv6_rcv_skb,
1215 .hash = raw_hash_sk,
1216 .unhash = raw_unhash_sk,
1217 .obj_size = sizeof(struct raw6_sock),
1218 .h.raw_hash = &raw_v6_hashinfo,
1219 #ifdef CONFIG_COMPAT
1220 .compat_setsockopt = compat_rawv6_setsockopt,
1221 .compat_getsockopt = compat_rawv6_getsockopt,
1222 .compat_ioctl = compat_rawv6_ioctl,
1223 #endif
1224 };
1225
1226 #ifdef CONFIG_PROC_FS
1227 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1228 {
1229 struct ipv6_pinfo *np = inet6_sk(sp);
1230 const struct in6_addr *dest, *src;
1231 __u16 destp, srcp;
1232
1233 dest = &np->daddr;
1234 src = &np->rcv_saddr;
1235 destp = 0;
1236 srcp = inet_sk(sp)->inet_num;
1237 seq_printf(seq,
1238 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1239 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1240 i,
1241 src->s6_addr32[0], src->s6_addr32[1],
1242 src->s6_addr32[2], src->s6_addr32[3], srcp,
1243 dest->s6_addr32[0], dest->s6_addr32[1],
1244 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1245 sp->sk_state,
1246 sk_wmem_alloc_get(sp),
1247 sk_rmem_alloc_get(sp),
1248 0, 0L, 0,
1249 sock_i_uid(sp), 0,
1250 sock_i_ino(sp),
1251 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1252 }
1253
1254 static int raw6_seq_show(struct seq_file *seq, void *v)
1255 {
1256 if (v == SEQ_START_TOKEN)
1257 seq_printf(seq,
1258 " sl "
1259 "local_address "
1260 "remote_address "
1261 "st tx_queue rx_queue tr tm->when retrnsmt"
1262 " uid timeout inode ref pointer drops\n");
1263 else
1264 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1265 return 0;
1266 }
1267
1268 static const struct seq_operations raw6_seq_ops = {
1269 .start = raw_seq_start,
1270 .next = raw_seq_next,
1271 .stop = raw_seq_stop,
1272 .show = raw6_seq_show,
1273 };
1274
1275 static int raw6_seq_open(struct inode *inode, struct file *file)
1276 {
1277 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1278 }
1279
1280 static const struct file_operations raw6_seq_fops = {
1281 .owner = THIS_MODULE,
1282 .open = raw6_seq_open,
1283 .read = seq_read,
1284 .llseek = seq_lseek,
1285 .release = seq_release_net,
1286 };
1287
1288 static int __net_init raw6_init_net(struct net *net)
1289 {
1290 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1291 return -ENOMEM;
1292
1293 return 0;
1294 }
1295
1296 static void __net_exit raw6_exit_net(struct net *net)
1297 {
1298 proc_net_remove(net, "raw6");
1299 }
1300
1301 static struct pernet_operations raw6_net_ops = {
1302 .init = raw6_init_net,
1303 .exit = raw6_exit_net,
1304 };
1305
1306 int __init raw6_proc_init(void)
1307 {
1308 return register_pernet_subsys(&raw6_net_ops);
1309 }
1310
1311 void raw6_proc_exit(void)
1312 {
1313 unregister_pernet_subsys(&raw6_net_ops);
1314 }
1315 #endif /* CONFIG_PROC_FS */
1316
1317 /* Same as inet6_dgram_ops, sans udp_poll. */
1318 static const struct proto_ops inet6_sockraw_ops = {
1319 .family = PF_INET6,
1320 .owner = THIS_MODULE,
1321 .release = inet6_release,
1322 .bind = inet6_bind,
1323 .connect = inet_dgram_connect, /* ok */
1324 .socketpair = sock_no_socketpair, /* a do nothing */
1325 .accept = sock_no_accept, /* a do nothing */
1326 .getname = inet6_getname,
1327 .poll = datagram_poll, /* ok */
1328 .ioctl = inet6_ioctl, /* must change */
1329 .listen = sock_no_listen, /* ok */
1330 .shutdown = inet_shutdown, /* ok */
1331 .setsockopt = sock_common_setsockopt, /* ok */
1332 .getsockopt = sock_common_getsockopt, /* ok */
1333 .sendmsg = inet_sendmsg, /* ok */
1334 .recvmsg = sock_common_recvmsg, /* ok */
1335 .mmap = sock_no_mmap,
1336 .sendpage = sock_no_sendpage,
1337 #ifdef CONFIG_COMPAT
1338 .compat_setsockopt = compat_sock_common_setsockopt,
1339 .compat_getsockopt = compat_sock_common_getsockopt,
1340 #endif
1341 };
1342
1343 static struct inet_protosw rawv6_protosw = {
1344 .type = SOCK_RAW,
1345 .protocol = IPPROTO_IP, /* wild card */
1346 .prot = &rawv6_prot,
1347 .ops = &inet6_sockraw_ops,
1348 .no_check = UDP_CSUM_DEFAULT,
1349 .flags = INET_PROTOSW_REUSE,
1350 };
1351
1352 int __init rawv6_init(void)
1353 {
1354 int ret;
1355
1356 ret = inet6_register_protosw(&rawv6_protosw);
1357 if (ret)
1358 goto out;
1359 out:
1360 return ret;
1361 }
1362
1363 void rawv6_exit(void)
1364 {
1365 inet6_unregister_protosw(&rawv6_protosw);
1366 }