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