?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
perf bpf: Move perf_event_output() from stdio.h to bpf.h
[linux/fpc-iii.git] / net / ipv6 / raw.c
blob5e0efd3954e90ade89eb4da17cd5ecef1894a1a3
1 /*
2 * RAW sockets for IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Adapted from linux/net/ipv4/raw.c
9 *
10 * Fixes:
11 * Hideaki YOSHIFUJI : sin6_scope_id support
12 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
13 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
14 *
15 * This program is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU General Public License
17 * as published by the Free Software Foundation; either version
18 * 2 of the License, or (at your option) any later version.
19 */
20
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/slab.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/netfilter.h>
32 #include <linux/netfilter_ipv6.h>
33 #include <linux/skbuff.h>
34 #include <linux/compat.h>
35 #include <linux/uaccess.h>
36 #include <asm/ioctls.h>
37
38 #include <net/net_namespace.h>
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
42
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
50 #include <net/udp.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #if IS_ENABLED(CONFIG_IPV6_MIP6)
54 #include <net/mip6.h>
55 #endif
56 #include <linux/mroute6.h>
57
58 #include <net/raw.h>
59 #include <net/rawv6.h>
60 #include <net/xfrm.h>
61
62 #include <linux/proc_fs.h>
63 #include <linux/seq_file.h>
64 #include <linux/export.h>
65
66 #define ICMPV6_HDRLEN 4 /* ICMPv6 header, RFC 4443 Section 2.1 */
67
68 struct raw_hashinfo raw_v6_hashinfo = {
69 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
70 };
71 EXPORT_SYMBOL_GPL(raw_v6_hashinfo);
72
73 struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
74 unsigned short num, const struct in6_addr *loc_addr,
75 const struct in6_addr *rmt_addr, int dif, int sdif)
76 {
77 bool is_multicast = ipv6_addr_is_multicast(loc_addr);
78
79 sk_for_each_from(sk)
80 if (inet_sk(sk)->inet_num == num) {
81
82 if (!net_eq(sock_net(sk), net))
83 continue;
84
85 if (!ipv6_addr_any(&sk->sk_v6_daddr) &&
86 !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr))
87 continue;
88
89 if (sk->sk_bound_dev_if &&
90 sk->sk_bound_dev_if != dif &&
91 sk->sk_bound_dev_if != sdif)
92 continue;
93
94 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
95 if (ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr))
96 goto found;
97 if (is_multicast &&
98 inet6_mc_check(sk, loc_addr, rmt_addr))
99 goto found;
100 continue;
101 }
102 goto found;
103 }
104 sk = NULL;
105 found:
106 return sk;
107 }
108 EXPORT_SYMBOL_GPL(__raw_v6_lookup);
109
110 /*
111 * 0 - deliver
112 * 1 - block
113 */
114 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
115 {
116 struct icmp6hdr _hdr;
117 const struct icmp6hdr *hdr;
118
119 /* We require only the four bytes of the ICMPv6 header, not any
120 * additional bytes of message body in "struct icmp6hdr".
121 */
122 hdr = skb_header_pointer(skb, skb_transport_offset(skb),
123 ICMPV6_HDRLEN, &_hdr);
124 if (hdr) {
125 const __u32 *data = &raw6_sk(sk)->filter.data[0];
126 unsigned int type = hdr->icmp6_type;
127
128 return (data[type >> 5] & (1U << (type & 31))) != 0;
129 }
130 return 1;
131 }
132
133 #if IS_ENABLED(CONFIG_IPV6_MIP6)
134 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
135
136 static mh_filter_t __rcu *mh_filter __read_mostly;
137
138 int rawv6_mh_filter_register(mh_filter_t filter)
139 {
140 rcu_assign_pointer(mh_filter, filter);
141 return 0;
142 }
143 EXPORT_SYMBOL(rawv6_mh_filter_register);
144
145 int rawv6_mh_filter_unregister(mh_filter_t filter)
146 {
147 RCU_INIT_POINTER(mh_filter, NULL);
148 synchronize_rcu();
149 return 0;
150 }
151 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
152
153 #endif
154
155 /*
156 * demultiplex raw sockets.
157 * (should consider queueing the skb in the sock receive_queue
158 * without calling rawv6.c)
159 *
160 * Caller owns SKB so we must make clones.
161 */
162 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
163 {
164 const struct in6_addr *saddr;
165 const struct in6_addr *daddr;
166 struct sock *sk;
167 bool delivered = false;
168 __u8 hash;
169 struct net *net;
170
171 saddr = &ipv6_hdr(skb)->saddr;
172 daddr = saddr + 1;
173
174 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
175
176 read_lock(&raw_v6_hashinfo.lock);
177 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
178
179 if (!sk)
180 goto out;
181
182 net = dev_net(skb->dev);
183 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr,
184 inet6_iif(skb), inet6_sdif(skb));
185
186 while (sk) {
187 int filtered;
188
189 delivered = true;
190 switch (nexthdr) {
191 case IPPROTO_ICMPV6:
192 filtered = icmpv6_filter(sk, skb);
193 break;
194
195 #if IS_ENABLED(CONFIG_IPV6_MIP6)
196 case IPPROTO_MH:
197 {
198 /* XXX: To validate MH only once for each packet,
199 * this is placed here. It should be after checking
200 * xfrm policy, however it doesn't. The checking xfrm
201 * policy is placed in rawv6_rcv() because it is
202 * required for each socket.
203 */
204 mh_filter_t *filter;
205
206 filter = rcu_dereference(mh_filter);
207 filtered = filter ? (*filter)(sk, skb) : 0;
208 break;
209 }
210 #endif
211 default:
212 filtered = 0;
213 break;
214 }
215
216 if (filtered < 0)
217 break;
218 if (filtered == 0) {
219 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
220
221 /* Not releasing hash table! */
222 if (clone) {
223 nf_reset(clone);
224 rawv6_rcv(sk, clone);
225 }
226 }
227 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
228 inet6_iif(skb), inet6_sdif(skb));
229 }
230 out:
231 read_unlock(&raw_v6_hashinfo.lock);
232 return delivered;
233 }
234
235 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
236 {
237 struct sock *raw_sk;
238
239 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
240 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
241 raw_sk = NULL;
242
243 return raw_sk != NULL;
244 }
245
246 /* This cleans up af_inet6 a bit. -DaveM */
247 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
248 {
249 struct inet_sock *inet = inet_sk(sk);
250 struct ipv6_pinfo *np = inet6_sk(sk);
251 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
252 __be32 v4addr = 0;
253 int addr_type;
254 int err;
255
256 if (addr_len < SIN6_LEN_RFC2133)
257 return -EINVAL;
258
259 if (addr->sin6_family != AF_INET6)
260 return -EINVAL;
261
262 addr_type = ipv6_addr_type(&addr->sin6_addr);
263
264 /* Raw sockets are IPv6 only */
265 if (addr_type == IPV6_ADDR_MAPPED)
266 return -EADDRNOTAVAIL;
267
268 lock_sock(sk);
269
270 err = -EINVAL;
271 if (sk->sk_state != TCP_CLOSE)
272 goto out;
273
274 rcu_read_lock();
275 /* Check if the address belongs to the host. */
276 if (addr_type != IPV6_ADDR_ANY) {
277 struct net_device *dev = NULL;
278
279 if (__ipv6_addr_needs_scope_id(addr_type)) {
280 if (addr_len >= sizeof(struct sockaddr_in6) &&
281 addr->sin6_scope_id) {
282 /* Override any existing binding, if another
283 * one is supplied by user.
284 */
285 sk->sk_bound_dev_if = addr->sin6_scope_id;
286 }
287
288 /* Binding to link-local address requires an interface */
289 if (!sk->sk_bound_dev_if)
290 goto out_unlock;
291
292 err = -ENODEV;
293 dev = dev_get_by_index_rcu(sock_net(sk),
294 sk->sk_bound_dev_if);
295 if (!dev)
296 goto out_unlock;
297 }
298
299 /* ipv4 addr of the socket is invalid. Only the
300 * unspecified and mapped address have a v4 equivalent.
301 */
302 v4addr = LOOPBACK4_IPV6;
303 if (!(addr_type & IPV6_ADDR_MULTICAST) &&
304 !sock_net(sk)->ipv6.sysctl.ip_nonlocal_bind) {
305 err = -EADDRNOTAVAIL;
306 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
307 dev, 0)) {
308 goto out_unlock;
309 }
310 }
311 }
312
313 inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
314 sk->sk_v6_rcv_saddr = addr->sin6_addr;
315 if (!(addr_type & IPV6_ADDR_MULTICAST))
316 np->saddr = addr->sin6_addr;
317 err = 0;
318 out_unlock:
319 rcu_read_unlock();
320 out:
321 release_sock(sk);
322 return err;
323 }
324
325 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
326 struct inet6_skb_parm *opt,
327 u8 type, u8 code, int offset, __be32 info)
328 {
329 struct inet_sock *inet = inet_sk(sk);
330 struct ipv6_pinfo *np = inet6_sk(sk);
331 int err;
332 int harderr;
333
334 /* Report error on raw socket, if:
335 1. User requested recverr.
336 2. Socket is connected (otherwise the error indication
337 is useless without recverr and error is hard.
338 */
339 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
340 return;
341
342 harderr = icmpv6_err_convert(type, code, &err);
343 if (type == ICMPV6_PKT_TOOBIG) {
344 ip6_sk_update_pmtu(skb, sk, info);
345 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
346 }
347 if (type == NDISC_REDIRECT) {
348 ip6_sk_redirect(skb, sk);
349 return;
350 }
351 if (np->recverr) {
352 u8 *payload = skb->data;
353 if (!inet->hdrincl)
354 payload += offset;
355 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
356 }
357
358 if (np->recverr || harderr) {
359 sk->sk_err = err;
360 sk->sk_error_report(sk);
361 }
362 }
363
364 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
365 u8 type, u8 code, int inner_offset, __be32 info)
366 {
367 struct sock *sk;
368 int hash;
369 const struct in6_addr *saddr, *daddr;
370 struct net *net;
371
372 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
373
374 read_lock(&raw_v6_hashinfo.lock);
375 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
376 if (sk) {
377 /* Note: ipv6_hdr(skb) != skb->data */
378 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
379 saddr = &ip6h->saddr;
380 daddr = &ip6h->daddr;
381 net = dev_net(skb->dev);
382
383 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
384 inet6_iif(skb), inet6_iif(skb)))) {
385 rawv6_err(sk, skb, NULL, type, code,
386 inner_offset, info);
387 sk = sk_next(sk);
388 }
389 }
390 read_unlock(&raw_v6_hashinfo.lock);
391 }
392
393 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
394 {
395 if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
396 skb_checksum_complete(skb)) {
397 atomic_inc(&sk->sk_drops);
398 kfree_skb(skb);
399 return NET_RX_DROP;
400 }
401
402 /* Charge it to the socket. */
403 skb_dst_drop(skb);
404 if (sock_queue_rcv_skb(sk, skb) < 0) {
405 kfree_skb(skb);
406 return NET_RX_DROP;
407 }
408
409 return 0;
410 }
411
412 /*
413 * This is next to useless...
414 * if we demultiplex in network layer we don't need the extra call
415 * just to queue the skb...
416 * maybe we could have the network decide upon a hint if it
417 * should call raw_rcv for demultiplexing
418 */
419 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
420 {
421 struct inet_sock *inet = inet_sk(sk);
422 struct raw6_sock *rp = raw6_sk(sk);
423
424 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
425 atomic_inc(&sk->sk_drops);
426 kfree_skb(skb);
427 return NET_RX_DROP;
428 }
429
430 if (!rp->checksum)
431 skb->ip_summed = CHECKSUM_UNNECESSARY;
432
433 if (skb->ip_summed == CHECKSUM_COMPLETE) {
434 skb_postpull_rcsum(skb, skb_network_header(skb),
435 skb_network_header_len(skb));
436 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
437 &ipv6_hdr(skb)->daddr,
438 skb->len, inet->inet_num, skb->csum))
439 skb->ip_summed = CHECKSUM_UNNECESSARY;
440 }
441 if (!skb_csum_unnecessary(skb))
442 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
443 &ipv6_hdr(skb)->daddr,
444 skb->len,
445 inet->inet_num, 0));
446
447 if (inet->hdrincl) {
448 if (skb_checksum_complete(skb)) {
449 atomic_inc(&sk->sk_drops);
450 kfree_skb(skb);
451 return NET_RX_DROP;
452 }
453 }
454
455 rawv6_rcv_skb(sk, skb);
456 return 0;
457 }
458
459
460 /*
461 * This should be easy, if there is something there
462 * we return it, otherwise we block.
463 */
464
465 static int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
466 int noblock, int flags, int *addr_len)
467 {
468 struct ipv6_pinfo *np = inet6_sk(sk);
469 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
470 struct sk_buff *skb;
471 size_t copied;
472 int err;
473
474 if (flags & MSG_OOB)
475 return -EOPNOTSUPP;
476
477 if (flags & MSG_ERRQUEUE)
478 return ipv6_recv_error(sk, msg, len, addr_len);
479
480 if (np->rxpmtu && np->rxopt.bits.rxpmtu)
481 return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
482
483 skb = skb_recv_datagram(sk, flags, noblock, &err);
484 if (!skb)
485 goto out;
486
487 copied = skb->len;
488 if (copied > len) {
489 copied = len;
490 msg->msg_flags |= MSG_TRUNC;
491 }
492
493 if (skb_csum_unnecessary(skb)) {
494 err = skb_copy_datagram_msg(skb, 0, msg, copied);
495 } else if (msg->msg_flags&MSG_TRUNC) {
496 if (__skb_checksum_complete(skb))
497 goto csum_copy_err;
498 err = skb_copy_datagram_msg(skb, 0, msg, copied);
499 } else {
500 err = skb_copy_and_csum_datagram_msg(skb, 0, msg);
501 if (err == -EINVAL)
502 goto csum_copy_err;
503 }
504 if (err)
505 goto out_free;
506
507 /* Copy the address. */
508 if (sin6) {
509 sin6->sin6_family = AF_INET6;
510 sin6->sin6_port = 0;
511 sin6->sin6_addr = ipv6_hdr(skb)->saddr;
512 sin6->sin6_flowinfo = 0;
513 sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
514 inet6_iif(skb));
515 *addr_len = sizeof(*sin6);
516 }
517
518 sock_recv_ts_and_drops(msg, sk, skb);
519
520 if (np->rxopt.all)
521 ip6_datagram_recv_ctl(sk, msg, skb);
522
523 err = copied;
524 if (flags & MSG_TRUNC)
525 err = skb->len;
526
527 out_free:
528 skb_free_datagram(sk, skb);
529 out:
530 return err;
531
532 csum_copy_err:
533 skb_kill_datagram(sk, skb, flags);
534
535 /* Error for blocking case is chosen to masquerade
536 as some normal condition.
537 */
538 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
539 goto out;
540 }
541
542 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
543 struct raw6_sock *rp)
544 {
545 struct sk_buff *skb;
546 int err = 0;
547 int offset;
548 int len;
549 int total_len;
550 __wsum tmp_csum;
551 __sum16 csum;
552
553 if (!rp->checksum)
554 goto send;
555
556 skb = skb_peek(&sk->sk_write_queue);
557 if (!skb)
558 goto out;
559
560 offset = rp->offset;
561 total_len = inet_sk(sk)->cork.base.length;
562 if (offset >= total_len - 1) {
563 err = -EINVAL;
564 ip6_flush_pending_frames(sk);
565 goto out;
566 }
567
568 /* should be check HW csum miyazawa */
569 if (skb_queue_len(&sk->sk_write_queue) == 1) {
570 /*
571 * Only one fragment on the socket.
572 */
573 tmp_csum = skb->csum;
574 } else {
575 struct sk_buff *csum_skb = NULL;
576 tmp_csum = 0;
577
578 skb_queue_walk(&sk->sk_write_queue, skb) {
579 tmp_csum = csum_add(tmp_csum, skb->csum);
580
581 if (csum_skb)
582 continue;
583
584 len = skb->len - skb_transport_offset(skb);
585 if (offset >= len) {
586 offset -= len;
587 continue;
588 }
589
590 csum_skb = skb;
591 }
592
593 skb = csum_skb;
594 }
595
596 offset += skb_transport_offset(skb);
597 err = skb_copy_bits(skb, offset, &csum, 2);
598 if (err < 0) {
599 ip6_flush_pending_frames(sk);
600 goto out;
601 }
602
603 /* in case cksum was not initialized */
604 if (unlikely(csum))
605 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
606
607 csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
608 total_len, fl6->flowi6_proto, tmp_csum);
609
610 if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
611 csum = CSUM_MANGLED_0;
612
613 BUG_ON(skb_store_bits(skb, offset, &csum, 2));
614
615 send:
616 err = ip6_push_pending_frames(sk);
617 out:
618 return err;
619 }
620
621 static int rawv6_send_hdrinc(struct sock *sk, struct msghdr *msg, int length,
622 struct flowi6 *fl6, struct dst_entry **dstp,
623 unsigned int flags, const struct sockcm_cookie *sockc)
624 {
625 struct ipv6_pinfo *np = inet6_sk(sk);
626 struct net *net = sock_net(sk);
627 struct ipv6hdr *iph;
628 struct sk_buff *skb;
629 int err;
630 struct rt6_info *rt = (struct rt6_info *)*dstp;
631 int hlen = LL_RESERVED_SPACE(rt->dst.dev);
632 int tlen = rt->dst.dev->needed_tailroom;
633
634 if (length > rt->dst.dev->mtu) {
635 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
636 return -EMSGSIZE;
637 }
638 if (length < sizeof(struct ipv6hdr))
639 return -EINVAL;
640 if (flags&MSG_PROBE)
641 goto out;
642
643 skb = sock_alloc_send_skb(sk,
644 length + hlen + tlen + 15,
645 flags & MSG_DONTWAIT, &err);
646 if (!skb)
647 goto error;
648 skb_reserve(skb, hlen);
649
650 skb->protocol = htons(ETH_P_IPV6);
651 skb->priority = sk->sk_priority;
652 skb->mark = sk->sk_mark;
653 skb->tstamp = sockc->transmit_time;
654
655 skb_put(skb, length);
656 skb_reset_network_header(skb);
657 iph = ipv6_hdr(skb);
658
659 skb->ip_summed = CHECKSUM_NONE;
660
661 if (flags & MSG_CONFIRM)
662 skb_set_dst_pending_confirm(skb, 1);
663
664 skb->transport_header = skb->network_header;
665 err = memcpy_from_msg(iph, msg, length);
666 if (err) {
667 err = -EFAULT;
668 kfree_skb(skb);
669 goto error;
670 }
671
672 skb_dst_set(skb, &rt->dst);
673 *dstp = NULL;
674
675 /* if egress device is enslaved to an L3 master device pass the
676 * skb to its handler for processing
677 */
678 skb = l3mdev_ip6_out(sk, skb);
679 if (unlikely(!skb))
680 return 0;
681
682 /* Acquire rcu_read_lock() in case we need to use rt->rt6i_idev
683 * in the error path. Since skb has been freed, the dst could
684 * have been queued for deletion.
685 */
686 rcu_read_lock();
687 IP6_UPD_PO_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
688 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, net, sk, skb,
689 NULL, rt->dst.dev, dst_output);
690 if (err > 0)
691 err = net_xmit_errno(err);
692 if (err) {
693 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
694 rcu_read_unlock();
695 goto error_check;
696 }
697 rcu_read_unlock();
698 out:
699 return 0;
700
701 error:
702 IP6_INC_STATS(net, rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
703 error_check:
704 if (err == -ENOBUFS && !np->recverr)
705 err = 0;
706 return err;
707 }
708
709 struct raw6_frag_vec {
710 struct msghdr *msg;
711 int hlen;
712 char c[4];
713 };
714
715 static int rawv6_probe_proto_opt(struct raw6_frag_vec *rfv, struct flowi6 *fl6)
716 {
717 int err = 0;
718 switch (fl6->flowi6_proto) {
719 case IPPROTO_ICMPV6:
720 rfv->hlen = 2;
721 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
722 if (!err) {
723 fl6->fl6_icmp_type = rfv->c[0];
724 fl6->fl6_icmp_code = rfv->c[1];
725 }
726 break;
727 case IPPROTO_MH:
728 rfv->hlen = 4;
729 err = memcpy_from_msg(rfv->c, rfv->msg, rfv->hlen);
730 if (!err)
731 fl6->fl6_mh_type = rfv->c[2];
732 }
733 return err;
734 }
735
736 static int raw6_getfrag(void *from, char *to, int offset, int len, int odd,
737 struct sk_buff *skb)
738 {
739 struct raw6_frag_vec *rfv = from;
740
741 if (offset < rfv->hlen) {
742 int copy = min(rfv->hlen - offset, len);
743
744 if (skb->ip_summed == CHECKSUM_PARTIAL)
745 memcpy(to, rfv->c + offset, copy);
746 else
747 skb->csum = csum_block_add(
748 skb->csum,
749 csum_partial_copy_nocheck(rfv->c + offset,
750 to, copy, 0),
751 odd);
752
753 odd = 0;
754 offset += copy;
755 to += copy;
756 len -= copy;
757
758 if (!len)
759 return 0;
760 }
761
762 offset -= rfv->hlen;
763
764 return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
765 }
766
767 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
768 {
769 struct ipv6_txoptions *opt_to_free = NULL;
770 struct ipv6_txoptions opt_space;
771 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
772 struct in6_addr *daddr, *final_p, final;
773 struct inet_sock *inet = inet_sk(sk);
774 struct ipv6_pinfo *np = inet6_sk(sk);
775 struct raw6_sock *rp = raw6_sk(sk);
776 struct ipv6_txoptions *opt = NULL;
777 struct ip6_flowlabel *flowlabel = NULL;
778 struct dst_entry *dst = NULL;
779 struct raw6_frag_vec rfv;
780 struct flowi6 fl6;
781 struct ipcm6_cookie ipc6;
782 int addr_len = msg->msg_namelen;
783 u16 proto;
784 int err;
785
786 /* Rough check on arithmetic overflow,
787 better check is made in ip6_append_data().
788 */
789 if (len > INT_MAX)
790 return -EMSGSIZE;
791
792 /* Mirror BSD error message compatibility */
793 if (msg->msg_flags & MSG_OOB)
794 return -EOPNOTSUPP;
795
796 /*
797 * Get and verify the address.
798 */
799 memset(&fl6, 0, sizeof(fl6));
800
801 fl6.flowi6_mark = sk->sk_mark;
802 fl6.flowi6_uid = sk->sk_uid;
803
804 ipcm6_init(&ipc6);
805 ipc6.sockc.tsflags = sk->sk_tsflags;
806
807 if (sin6) {
808 if (addr_len < SIN6_LEN_RFC2133)
809 return -EINVAL;
810
811 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
812 return -EAFNOSUPPORT;
813
814 /* port is the proto value [0..255] carried in nexthdr */
815 proto = ntohs(sin6->sin6_port);
816
817 if (!proto)
818 proto = inet->inet_num;
819 else if (proto != inet->inet_num)
820 return -EINVAL;
821
822 if (proto > 255)
823 return -EINVAL;
824
825 daddr = &sin6->sin6_addr;
826 if (np->sndflow) {
827 fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
828 if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
829 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
830 if (!flowlabel)
831 return -EINVAL;
832 }
833 }
834
835 /*
836 * Otherwise it will be difficult to maintain
837 * sk->sk_dst_cache.
838 */
839 if (sk->sk_state == TCP_ESTABLISHED &&
840 ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
841 daddr = &sk->sk_v6_daddr;
842
843 if (addr_len >= sizeof(struct sockaddr_in6) &&
844 sin6->sin6_scope_id &&
845 __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
846 fl6.flowi6_oif = sin6->sin6_scope_id;
847 } else {
848 if (sk->sk_state != TCP_ESTABLISHED)
849 return -EDESTADDRREQ;
850
851 proto = inet->inet_num;
852 daddr = &sk->sk_v6_daddr;
853 fl6.flowlabel = np->flow_label;
854 }
855
856 if (fl6.flowi6_oif == 0)
857 fl6.flowi6_oif = sk->sk_bound_dev_if;
858
859 if (msg->msg_controllen) {
860 opt = &opt_space;
861 memset(opt, 0, sizeof(struct ipv6_txoptions));
862 opt->tot_len = sizeof(struct ipv6_txoptions);
863 ipc6.opt = opt;
864
865 err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6);
866 if (err < 0) {
867 fl6_sock_release(flowlabel);
868 return err;
869 }
870 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
871 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
872 if (!flowlabel)
873 return -EINVAL;
874 }
875 if (!(opt->opt_nflen|opt->opt_flen))
876 opt = NULL;
877 }
878 if (!opt) {
879 opt = txopt_get(np);
880 opt_to_free = opt;
881 }
882 if (flowlabel)
883 opt = fl6_merge_options(&opt_space, flowlabel, opt);
884 opt = ipv6_fixup_options(&opt_space, opt);
885
886 fl6.flowi6_proto = proto;
887 rfv.msg = msg;
888 rfv.hlen = 0;
889 err = rawv6_probe_proto_opt(&rfv, &fl6);
890 if (err)
891 goto out;
892
893 if (!ipv6_addr_any(daddr))
894 fl6.daddr = *daddr;
895 else
896 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
897 if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
898 fl6.saddr = np->saddr;
899
900 final_p = fl6_update_dst(&fl6, opt, &final);
901
902 if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
903 fl6.flowi6_oif = np->mcast_oif;
904 else if (!fl6.flowi6_oif)
905 fl6.flowi6_oif = np->ucast_oif;
906 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
907
908 if (inet->hdrincl)
909 fl6.flowi6_flags |= FLOWI_FLAG_KNOWN_NH;
910
911 if (ipc6.tclass < 0)
912 ipc6.tclass = np->tclass;
913
914 fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
915
916 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
917 if (IS_ERR(dst)) {
918 err = PTR_ERR(dst);
919 goto out;
920 }
921 if (ipc6.hlimit < 0)
922 ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
923
924 if (ipc6.dontfrag < 0)
925 ipc6.dontfrag = np->dontfrag;
926
927 if (msg->msg_flags&MSG_CONFIRM)
928 goto do_confirm;
929
930 back_from_confirm:
931 if (inet->hdrincl)
932 err = rawv6_send_hdrinc(sk, msg, len, &fl6, &dst,
933 msg->msg_flags, &ipc6.sockc);
934 else {
935 ipc6.opt = opt;
936 lock_sock(sk);
937 err = ip6_append_data(sk, raw6_getfrag, &rfv,
938 len, 0, &ipc6, &fl6, (struct rt6_info *)dst,
939 msg->msg_flags);
940
941 if (err)
942 ip6_flush_pending_frames(sk);
943 else if (!(msg->msg_flags & MSG_MORE))
944 err = rawv6_push_pending_frames(sk, &fl6, rp);
945 release_sock(sk);
946 }
947 done:
948 dst_release(dst);
949 out:
950 fl6_sock_release(flowlabel);
951 txopt_put(opt_to_free);
952 return err < 0 ? err : len;
953 do_confirm:
954 if (msg->msg_flags & MSG_PROBE)
955 dst_confirm_neigh(dst, &fl6.daddr);
956 if (!(msg->msg_flags & MSG_PROBE) || len)
957 goto back_from_confirm;
958 err = 0;
959 goto done;
960 }
961
962 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
963 char __user *optval, int optlen)
964 {
965 switch (optname) {
966 case ICMPV6_FILTER:
967 if (optlen > sizeof(struct icmp6_filter))
968 optlen = sizeof(struct icmp6_filter);
969 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
970 return -EFAULT;
971 return 0;
972 default:
973 return -ENOPROTOOPT;
974 }
975
976 return 0;
977 }
978
979 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
980 char __user *optval, int __user *optlen)
981 {
982 int len;
983
984 switch (optname) {
985 case ICMPV6_FILTER:
986 if (get_user(len, optlen))
987 return -EFAULT;
988 if (len < 0)
989 return -EINVAL;
990 if (len > sizeof(struct icmp6_filter))
991 len = sizeof(struct icmp6_filter);
992 if (put_user(len, optlen))
993 return -EFAULT;
994 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
995 return -EFAULT;
996 return 0;
997 default:
998 return -ENOPROTOOPT;
999 }
1000
1001 return 0;
1002 }
1003
1004
1005 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
1006 char __user *optval, unsigned int optlen)
1007 {
1008 struct raw6_sock *rp = raw6_sk(sk);
1009 int val;
1010
1011 if (get_user(val, (int __user *)optval))
1012 return -EFAULT;
1013
1014 switch (optname) {
1015 case IPV6_HDRINCL:
1016 if (sk->sk_type != SOCK_RAW)
1017 return -EINVAL;
1018 inet_sk(sk)->hdrincl = !!val;
1019 return 0;
1020 case IPV6_CHECKSUM:
1021 if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
1022 level == IPPROTO_IPV6) {
1023 /*
1024 * RFC3542 tells that IPV6_CHECKSUM socket
1025 * option in the IPPROTO_IPV6 level is not
1026 * allowed on ICMPv6 sockets.
1027 * If you want to set it, use IPPROTO_RAW
1028 * level IPV6_CHECKSUM socket option
1029 * (Linux extension).
1030 */
1031 return -EINVAL;
1032 }
1033
1034 /* You may get strange result with a positive odd offset;
1035 RFC2292bis agrees with me. */
1036 if (val > 0 && (val&1))
1037 return -EINVAL;
1038 if (val < 0) {
1039 rp->checksum = 0;
1040 } else {
1041 rp->checksum = 1;
1042 rp->offset = val;
1043 }
1044
1045 return 0;
1046
1047 default:
1048 return -ENOPROTOOPT;
1049 }
1050 }
1051
1052 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1053 char __user *optval, unsigned int optlen)
1054 {
1055 switch (level) {
1056 case SOL_RAW:
1057 break;
1058
1059 case SOL_ICMPV6:
1060 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1061 return -EOPNOTSUPP;
1062 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1063 case SOL_IPV6:
1064 if (optname == IPV6_CHECKSUM ||
1065 optname == IPV6_HDRINCL)
1066 break;
1067 /* fall through */
1068 default:
1069 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1070 }
1071
1072 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1073 }
1074
1075 #ifdef CONFIG_COMPAT
1076 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1077 char __user *optval, unsigned int optlen)
1078 {
1079 switch (level) {
1080 case SOL_RAW:
1081 break;
1082 case SOL_ICMPV6:
1083 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1084 return -EOPNOTSUPP;
1085 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1086 case SOL_IPV6:
1087 if (optname == IPV6_CHECKSUM ||
1088 optname == IPV6_HDRINCL)
1089 break;
1090 /* fall through */
1091 default:
1092 return compat_ipv6_setsockopt(sk, level, optname,
1093 optval, optlen);
1094 }
1095 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1096 }
1097 #endif
1098
1099 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1100 char __user *optval, int __user *optlen)
1101 {
1102 struct raw6_sock *rp = raw6_sk(sk);
1103 int val, len;
1104
1105 if (get_user(len, optlen))
1106 return -EFAULT;
1107
1108 switch (optname) {
1109 case IPV6_HDRINCL:
1110 val = inet_sk(sk)->hdrincl;
1111 break;
1112 case IPV6_CHECKSUM:
1113 /*
1114 * We allow getsockopt() for IPPROTO_IPV6-level
1115 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1116 * since RFC3542 is silent about it.
1117 */
1118 if (rp->checksum == 0)
1119 val = -1;
1120 else
1121 val = rp->offset;
1122 break;
1123
1124 default:
1125 return -ENOPROTOOPT;
1126 }
1127
1128 len = min_t(unsigned int, sizeof(int), len);
1129
1130 if (put_user(len, optlen))
1131 return -EFAULT;
1132 if (copy_to_user(optval, &val, len))
1133 return -EFAULT;
1134 return 0;
1135 }
1136
1137 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1138 char __user *optval, int __user *optlen)
1139 {
1140 switch (level) {
1141 case SOL_RAW:
1142 break;
1143
1144 case SOL_ICMPV6:
1145 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1146 return -EOPNOTSUPP;
1147 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1148 case SOL_IPV6:
1149 if (optname == IPV6_CHECKSUM ||
1150 optname == IPV6_HDRINCL)
1151 break;
1152 /* fall through */
1153 default:
1154 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1155 }
1156
1157 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1158 }
1159
1160 #ifdef CONFIG_COMPAT
1161 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1162 char __user *optval, int __user *optlen)
1163 {
1164 switch (level) {
1165 case SOL_RAW:
1166 break;
1167 case SOL_ICMPV6:
1168 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1169 return -EOPNOTSUPP;
1170 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1171 case SOL_IPV6:
1172 if (optname == IPV6_CHECKSUM ||
1173 optname == IPV6_HDRINCL)
1174 break;
1175 /* fall through */
1176 default:
1177 return compat_ipv6_getsockopt(sk, level, optname,
1178 optval, optlen);
1179 }
1180 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1181 }
1182 #endif
1183
1184 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1185 {
1186 switch (cmd) {
1187 case SIOCOUTQ: {
1188 int amount = sk_wmem_alloc_get(sk);
1189
1190 return put_user(amount, (int __user *)arg);
1191 }
1192 case SIOCINQ: {
1193 struct sk_buff *skb;
1194 int amount = 0;
1195
1196 spin_lock_bh(&sk->sk_receive_queue.lock);
1197 skb = skb_peek(&sk->sk_receive_queue);
1198 if (skb)
1199 amount = skb->len;
1200 spin_unlock_bh(&sk->sk_receive_queue.lock);
1201 return put_user(amount, (int __user *)arg);
1202 }
1203
1204 default:
1205 #ifdef CONFIG_IPV6_MROUTE
1206 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1207 #else
1208 return -ENOIOCTLCMD;
1209 #endif
1210 }
1211 }
1212
1213 #ifdef CONFIG_COMPAT
1214 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1215 {
1216 switch (cmd) {
1217 case SIOCOUTQ:
1218 case SIOCINQ:
1219 return -ENOIOCTLCMD;
1220 default:
1221 #ifdef CONFIG_IPV6_MROUTE
1222 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1223 #else
1224 return -ENOIOCTLCMD;
1225 #endif
1226 }
1227 }
1228 #endif
1229
1230 static void rawv6_close(struct sock *sk, long timeout)
1231 {
1232 if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1233 ip6_ra_control(sk, -1);
1234 ip6mr_sk_done(sk);
1235 sk_common_release(sk);
1236 }
1237
1238 static void raw6_destroy(struct sock *sk)
1239 {
1240 lock_sock(sk);
1241 ip6_flush_pending_frames(sk);
1242 release_sock(sk);
1243
1244 inet6_destroy_sock(sk);
1245 }
1246
1247 static int rawv6_init_sk(struct sock *sk)
1248 {
1249 struct raw6_sock *rp = raw6_sk(sk);
1250
1251 switch (inet_sk(sk)->inet_num) {
1252 case IPPROTO_ICMPV6:
1253 rp->checksum = 1;
1254 rp->offset = 2;
1255 break;
1256 case IPPROTO_MH:
1257 rp->checksum = 1;
1258 rp->offset = 4;
1259 break;
1260 default:
1261 break;
1262 }
1263 return 0;
1264 }
1265
1266 struct proto rawv6_prot = {
1267 .name = "RAWv6",
1268 .owner = THIS_MODULE,
1269 .close = rawv6_close,
1270 .destroy = raw6_destroy,
1271 .connect = ip6_datagram_connect_v6_only,
1272 .disconnect = __udp_disconnect,
1273 .ioctl = rawv6_ioctl,
1274 .init = rawv6_init_sk,
1275 .setsockopt = rawv6_setsockopt,
1276 .getsockopt = rawv6_getsockopt,
1277 .sendmsg = rawv6_sendmsg,
1278 .recvmsg = rawv6_recvmsg,
1279 .bind = rawv6_bind,
1280 .backlog_rcv = rawv6_rcv_skb,
1281 .hash = raw_hash_sk,
1282 .unhash = raw_unhash_sk,
1283 .obj_size = sizeof(struct raw6_sock),
1284 .useroffset = offsetof(struct raw6_sock, filter),
1285 .usersize = sizeof_field(struct raw6_sock, filter),
1286 .h.raw_hash = &raw_v6_hashinfo,
1287 #ifdef CONFIG_COMPAT
1288 .compat_setsockopt = compat_rawv6_setsockopt,
1289 .compat_getsockopt = compat_rawv6_getsockopt,
1290 .compat_ioctl = compat_rawv6_ioctl,
1291 #endif
1292 .diag_destroy = raw_abort,
1293 };
1294
1295 #ifdef CONFIG_PROC_FS
1296 static int raw6_seq_show(struct seq_file *seq, void *v)
1297 {
1298 if (v == SEQ_START_TOKEN) {
1299 seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1300 } else {
1301 struct sock *sp = v;
1302 __u16 srcp = inet_sk(sp)->inet_num;
1303 ip6_dgram_sock_seq_show(seq, v, srcp, 0,
1304 raw_seq_private(seq)->bucket);
1305 }
1306 return 0;
1307 }
1308
1309 static const struct seq_operations raw6_seq_ops = {
1310 .start = raw_seq_start,
1311 .next = raw_seq_next,
1312 .stop = raw_seq_stop,
1313 .show = raw6_seq_show,
1314 };
1315
1316 static int __net_init raw6_init_net(struct net *net)
1317 {
1318 if (!proc_create_net_data("raw6", 0444, net->proc_net, &raw6_seq_ops,
1319 sizeof(struct raw_iter_state), &raw_v6_hashinfo))
1320 return -ENOMEM;
1321
1322 return 0;
1323 }
1324
1325 static void __net_exit raw6_exit_net(struct net *net)
1326 {
1327 remove_proc_entry("raw6", net->proc_net);
1328 }
1329
1330 static struct pernet_operations raw6_net_ops = {
1331 .init = raw6_init_net,
1332 .exit = raw6_exit_net,
1333 };
1334
1335 int __init raw6_proc_init(void)
1336 {
1337 return register_pernet_subsys(&raw6_net_ops);
1338 }
1339
1340 void raw6_proc_exit(void)
1341 {
1342 unregister_pernet_subsys(&raw6_net_ops);
1343 }
1344 #endif /* CONFIG_PROC_FS */
1345
1346 /* Same as inet6_dgram_ops, sans udp_poll. */
1347 const struct proto_ops inet6_sockraw_ops = {
1348 .family = PF_INET6,
1349 .owner = THIS_MODULE,
1350 .release = inet6_release,
1351 .bind = inet6_bind,
1352 .connect = inet_dgram_connect, /* ok */
1353 .socketpair = sock_no_socketpair, /* a do nothing */
1354 .accept = sock_no_accept, /* a do nothing */
1355 .getname = inet6_getname,
1356 .poll = datagram_poll, /* ok */
1357 .ioctl = inet6_ioctl, /* must change */
1358 .listen = sock_no_listen, /* ok */
1359 .shutdown = inet_shutdown, /* ok */
1360 .setsockopt = sock_common_setsockopt, /* ok */
1361 .getsockopt = sock_common_getsockopt, /* ok */
1362 .sendmsg = inet_sendmsg, /* ok */
1363 .recvmsg = sock_common_recvmsg, /* ok */
1364 .mmap = sock_no_mmap,
1365 .sendpage = sock_no_sendpage,
1366 #ifdef CONFIG_COMPAT
1367 .compat_setsockopt = compat_sock_common_setsockopt,
1368 .compat_getsockopt = compat_sock_common_getsockopt,
1369 #endif
1370 };
1371
1372 static struct inet_protosw rawv6_protosw = {
1373 .type = SOCK_RAW,
1374 .protocol = IPPROTO_IP, /* wild card */
1375 .prot = &rawv6_prot,
1376 .ops = &inet6_sockraw_ops,
1377 .flags = INET_PROTOSW_REUSE,
1378 };
1379
1380 int __init rawv6_init(void)
1381 {
1382 return inet6_register_protosw(&rawv6_protosw);
1383 }
1384
1385 void rawv6_exit(void)
1386 {
1387 inet6_unregister_protosw(&rawv6_protosw);
1388 }
Linux with added FPC-III support