?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
pidns: make pid->level and pid_ns->level unsigned
[linux-2.6/kvm.git] / net / ipv6 / raw.c
blob396f0ea11090a98d6a07898d0e79d58fb7652313
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 * $Id: raw.c,v 1.51 2002/02/01 22:01:04 davem Exp $
11 *
12 * Fixes:
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
16 *
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
21 */
22
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/netfilter.h>
33 #include <linux/netfilter_ipv6.h>
34 #include <linux/skbuff.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
65 static struct raw_hashinfo raw_v6_hashinfo = {
66 .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
67 };
68
69 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
70 unsigned short num, struct in6_addr *loc_addr,
71 struct in6_addr *rmt_addr, int dif)
72 {
73 struct hlist_node *node;
74 int is_multicast = ipv6_addr_is_multicast(loc_addr);
75
76 sk_for_each_from(sk, node)
77 if (inet_sk(sk)->num == num) {
78 struct ipv6_pinfo *np = inet6_sk(sk);
79
80 if (!net_eq(sock_net(sk), net))
81 continue;
82
83 if (!ipv6_addr_any(&np->daddr) &&
84 !ipv6_addr_equal(&np->daddr, rmt_addr))
85 continue;
86
87 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
88 continue;
89
90 if (!ipv6_addr_any(&np->rcv_saddr)) {
91 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
92 goto found;
93 if (is_multicast &&
94 inet6_mc_check(sk, loc_addr, rmt_addr))
95 goto found;
96 continue;
97 }
98 goto found;
99 }
100 sk = NULL;
101 found:
102 return sk;
103 }
104
105 /*
106 * 0 - deliver
107 * 1 - block
108 */
109 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
110 {
111 struct icmp6hdr *icmph;
112 struct raw6_sock *rp = raw6_sk(sk);
113
114 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
115 __u32 *data = &rp->filter.data[0];
116 int bit_nr;
117
118 icmph = (struct icmp6hdr *) skb->data;
119 bit_nr = icmph->icmp6_type;
120
121 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
122 }
123 return 0;
124 }
125
126 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
127 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
128
129 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
130 struct sk_buff *skb))
131 {
132 rcu_assign_pointer(mh_filter, filter);
133 return 0;
134 }
135 EXPORT_SYMBOL(rawv6_mh_filter_register);
136
137 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
138 struct sk_buff *skb))
139 {
140 rcu_assign_pointer(mh_filter, NULL);
141 synchronize_rcu();
142 return 0;
143 }
144 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
145
146 #endif
147
148 /*
149 * demultiplex raw sockets.
150 * (should consider queueing the skb in the sock receive_queue
151 * without calling rawv6.c)
152 *
153 * Caller owns SKB so we must make clones.
154 */
155 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
156 {
157 struct in6_addr *saddr;
158 struct in6_addr *daddr;
159 struct sock *sk;
160 int delivered = 0;
161 __u8 hash;
162 struct net *net;
163
164 saddr = &ipv6_hdr(skb)->saddr;
165 daddr = saddr + 1;
166
167 hash = nexthdr & (MAX_INET_PROTOS - 1);
168
169 read_lock(&raw_v6_hashinfo.lock);
170 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
171
172 if (sk == NULL)
173 goto out;
174
175 net = dev_net(skb->dev);
176 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
177
178 while (sk) {
179 int filtered;
180
181 delivered = 1;
182 switch (nexthdr) {
183 case IPPROTO_ICMPV6:
184 filtered = icmpv6_filter(sk, skb);
185 break;
186
187 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
188 case IPPROTO_MH:
189 {
190 /* XXX: To validate MH only once for each packet,
191 * this is placed here. It should be after checking
192 * xfrm policy, however it doesn't. The checking xfrm
193 * policy is placed in rawv6_rcv() because it is
194 * required for each socket.
195 */
196 int (*filter)(struct sock *sock, struct sk_buff *skb);
197
198 filter = rcu_dereference(mh_filter);
199 filtered = filter ? filter(sk, skb) : 0;
200 break;
201 }
202 #endif
203 default:
204 filtered = 0;
205 break;
206 }
207
208 if (filtered < 0)
209 break;
210 if (filtered == 0) {
211 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
212
213 /* Not releasing hash table! */
214 if (clone) {
215 nf_reset(clone);
216 rawv6_rcv(sk, clone);
217 }
218 }
219 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
220 IP6CB(skb)->iif);
221 }
222 out:
223 read_unlock(&raw_v6_hashinfo.lock);
224 return delivered;
225 }
226
227 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
228 {
229 struct sock *raw_sk;
230
231 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
232 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
233 raw_sk = NULL;
234
235 return raw_sk != NULL;
236 }
237
238 /* This cleans up af_inet6 a bit. -DaveM */
239 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
240 {
241 struct inet_sock *inet = inet_sk(sk);
242 struct ipv6_pinfo *np = inet6_sk(sk);
243 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
244 __be32 v4addr = 0;
245 int addr_type;
246 int err;
247
248 if (addr_len < SIN6_LEN_RFC2133)
249 return -EINVAL;
250 addr_type = ipv6_addr_type(&addr->sin6_addr);
251
252 /* Raw sockets are IPv6 only */
253 if (addr_type == IPV6_ADDR_MAPPED)
254 return(-EADDRNOTAVAIL);
255
256 lock_sock(sk);
257
258 err = -EINVAL;
259 if (sk->sk_state != TCP_CLOSE)
260 goto out;
261
262 /* Check if the address belongs to the host. */
263 if (addr_type != IPV6_ADDR_ANY) {
264 struct net_device *dev = NULL;
265
266 if (addr_type & IPV6_ADDR_LINKLOCAL) {
267 if (addr_len >= sizeof(struct sockaddr_in6) &&
268 addr->sin6_scope_id) {
269 /* Override any existing binding, if another
270 * one is supplied by user.
271 */
272 sk->sk_bound_dev_if = addr->sin6_scope_id;
273 }
274
275 /* Binding to link-local address requires an interface */
276 if (!sk->sk_bound_dev_if)
277 goto out;
278
279 dev = dev_get_by_index(sock_net(sk), sk->sk_bound_dev_if);
280 if (!dev) {
281 err = -ENODEV;
282 goto out;
283 }
284 }
285
286 /* ipv4 addr of the socket is invalid. Only the
287 * unspecified and mapped address have a v4 equivalent.
288 */
289 v4addr = LOOPBACK4_IPV6;
290 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
291 err = -EADDRNOTAVAIL;
292 if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
293 dev, 0)) {
294 if (dev)
295 dev_put(dev);
296 goto out;
297 }
298 }
299 if (dev)
300 dev_put(dev);
301 }
302
303 inet->rcv_saddr = inet->saddr = v4addr;
304 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
305 if (!(addr_type & IPV6_ADDR_MULTICAST))
306 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
307 err = 0;
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 int type, int 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 int type, int code, int inner_offset, __be32 info)
349 {
350 struct sock *sk;
351 int hash;
352 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 struct ipv6hdr *ip6h = (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 || sk->sk_filter) &&
379 skb_checksum_complete(skb)) {
380 atomic_inc(&sk->sk_drops);
381 kfree_skb(skb);
382 return 0;
383 }
384
385 /* Charge it to the socket. */
386 if (sock_queue_rcv_skb(sk,skb)<0) {
387 atomic_inc(&sk->sk_drops);
388 kfree_skb(skb);
389 return 0;
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->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->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 0;
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 skb = skb_recv_datagram(sk, flags, noblock, &err);
468 if (!skb)
469 goto out;
470
471 copied = skb->len;
472 if (copied > len) {
473 copied = len;
474 msg->msg_flags |= MSG_TRUNC;
475 }
476
477 if (skb_csum_unnecessary(skb)) {
478 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
479 } else if (msg->msg_flags&MSG_TRUNC) {
480 if (__skb_checksum_complete(skb))
481 goto csum_copy_err;
482 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
483 } else {
484 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
485 if (err == -EINVAL)
486 goto csum_copy_err;
487 }
488 if (err)
489 goto out_free;
490
491 /* Copy the address. */
492 if (sin6) {
493 sin6->sin6_family = AF_INET6;
494 sin6->sin6_port = 0;
495 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
496 sin6->sin6_flowinfo = 0;
497 sin6->sin6_scope_id = 0;
498 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
499 sin6->sin6_scope_id = IP6CB(skb)->iif;
500 }
501
502 sock_recv_timestamp(msg, sk, skb);
503
504 if (np->rxopt.all)
505 datagram_recv_ctl(sk, msg, skb);
506
507 err = copied;
508 if (flags & MSG_TRUNC)
509 err = skb->len;
510
511 out_free:
512 skb_free_datagram(sk, skb);
513 out:
514 return err;
515
516 csum_copy_err:
517 skb_kill_datagram(sk, skb, flags);
518
519 /* Error for blocking case is chosen to masquerade
520 as some normal condition.
521 */
522 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
523 atomic_inc(&sk->sk_drops);
524 goto out;
525 }
526
527 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
528 struct raw6_sock *rp)
529 {
530 struct sk_buff *skb;
531 int err = 0;
532 int offset;
533 int len;
534 int total_len;
535 __wsum tmp_csum;
536 __sum16 csum;
537
538 if (!rp->checksum)
539 goto send;
540
541 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
542 goto out;
543
544 offset = rp->offset;
545 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
546 skb->data);
547 if (offset >= total_len - 1) {
548 err = -EINVAL;
549 ip6_flush_pending_frames(sk);
550 goto out;
551 }
552
553 /* should be check HW csum miyazawa */
554 if (skb_queue_len(&sk->sk_write_queue) == 1) {
555 /*
556 * Only one fragment on the socket.
557 */
558 tmp_csum = skb->csum;
559 } else {
560 struct sk_buff *csum_skb = NULL;
561 tmp_csum = 0;
562
563 skb_queue_walk(&sk->sk_write_queue, skb) {
564 tmp_csum = csum_add(tmp_csum, skb->csum);
565
566 if (csum_skb)
567 continue;
568
569 len = skb->len - skb_transport_offset(skb);
570 if (offset >= len) {
571 offset -= len;
572 continue;
573 }
574
575 csum_skb = skb;
576 }
577
578 skb = csum_skb;
579 }
580
581 offset += skb_transport_offset(skb);
582 if (skb_copy_bits(skb, offset, &csum, 2))
583 BUG();
584
585 /* in case cksum was not initialized */
586 if (unlikely(csum))
587 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
588
589 csum = csum_ipv6_magic(&fl->fl6_src,
590 &fl->fl6_dst,
591 total_len, fl->proto, tmp_csum);
592
593 if (csum == 0 && fl->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 flowi *fl, struct rt6_info *rt,
607 unsigned int flags)
608 {
609 struct ipv6_pinfo *np = inet6_sk(sk);
610 struct ipv6hdr *iph;
611 struct sk_buff *skb;
612 unsigned int hh_len;
613 int err;
614
615 if (length > rt->u.dst.dev->mtu) {
616 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
617 return -EMSGSIZE;
618 }
619 if (flags&MSG_PROBE)
620 goto out;
621
622 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
623
624 skb = sock_alloc_send_skb(sk, length+hh_len+15,
625 flags&MSG_DONTWAIT, &err);
626 if (skb == NULL)
627 goto error;
628 skb_reserve(skb, hh_len);
629
630 skb->priority = sk->sk_priority;
631 skb->mark = sk->sk_mark;
632 skb->dst = dst_clone(&rt->u.dst);
633
634 skb_put(skb, length);
635 skb_reset_network_header(skb);
636 iph = ipv6_hdr(skb);
637
638 skb->ip_summed = CHECKSUM_NONE;
639
640 skb->transport_header = skb->network_header;
641 err = memcpy_fromiovecend((void *)iph, from, 0, length);
642 if (err)
643 goto error_fault;
644
645 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
646 err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
647 dst_output);
648 if (err > 0)
649 err = np->recverr ? net_xmit_errno(err) : 0;
650 if (err)
651 goto error;
652 out:
653 return 0;
654
655 error_fault:
656 err = -EFAULT;
657 kfree_skb(skb);
658 error:
659 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
660 return err;
661 }
662
663 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
664 {
665 struct iovec *iov;
666 u8 __user *type = NULL;
667 u8 __user *code = NULL;
668 u8 len = 0;
669 int probed = 0;
670 int i;
671
672 if (!msg->msg_iov)
673 return 0;
674
675 for (i = 0; i < msg->msg_iovlen; i++) {
676 iov = &msg->msg_iov[i];
677 if (!iov)
678 continue;
679
680 switch (fl->proto) {
681 case IPPROTO_ICMPV6:
682 /* check if one-byte field is readable or not. */
683 if (iov->iov_base && iov->iov_len < 1)
684 break;
685
686 if (!type) {
687 type = iov->iov_base;
688 /* check if code field is readable or not. */
689 if (iov->iov_len > 1)
690 code = type + 1;
691 } else if (!code)
692 code = iov->iov_base;
693
694 if (type && code) {
695 if (get_user(fl->fl_icmp_type, type) ||
696 get_user(fl->fl_icmp_code, code))
697 return -EFAULT;
698 probed = 1;
699 }
700 break;
701 case IPPROTO_MH:
702 if (iov->iov_base && iov->iov_len < 1)
703 break;
704 /* check if type field is readable or not. */
705 if (iov->iov_len > 2 - len) {
706 u8 __user *p = iov->iov_base;
707 if (get_user(fl->fl_mh_type, &p[2 - len]))
708 return -EFAULT;
709 probed = 1;
710 } else
711 len += iov->iov_len;
712
713 break;
714 default:
715 probed = 1;
716 break;
717 }
718 if (probed)
719 break;
720 }
721 return 0;
722 }
723
724 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
725 struct msghdr *msg, size_t len)
726 {
727 struct ipv6_txoptions opt_space;
728 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
729 struct in6_addr *daddr, *final_p = NULL, final;
730 struct inet_sock *inet = inet_sk(sk);
731 struct ipv6_pinfo *np = inet6_sk(sk);
732 struct raw6_sock *rp = raw6_sk(sk);
733 struct ipv6_txoptions *opt = NULL;
734 struct ip6_flowlabel *flowlabel = NULL;
735 struct dst_entry *dst = NULL;
736 struct flowi fl;
737 int addr_len = msg->msg_namelen;
738 int hlimit = -1;
739 int tclass = -1;
740 u16 proto;
741 int err;
742
743 /* Rough check on arithmetic overflow,
744 better check is made in ip6_append_data().
745 */
746 if (len > INT_MAX)
747 return -EMSGSIZE;
748
749 /* Mirror BSD error message compatibility */
750 if (msg->msg_flags & MSG_OOB)
751 return -EOPNOTSUPP;
752
753 /*
754 * Get and verify the address.
755 */
756 memset(&fl, 0, sizeof(fl));
757
758 fl.mark = sk->sk_mark;
759
760 if (sin6) {
761 if (addr_len < SIN6_LEN_RFC2133)
762 return -EINVAL;
763
764 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
765 return(-EAFNOSUPPORT);
766
767 /* port is the proto value [0..255] carried in nexthdr */
768 proto = ntohs(sin6->sin6_port);
769
770 if (!proto)
771 proto = inet->num;
772 else if (proto != inet->num)
773 return(-EINVAL);
774
775 if (proto > 255)
776 return(-EINVAL);
777
778 daddr = &sin6->sin6_addr;
779 if (np->sndflow) {
780 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
781 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
782 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
783 if (flowlabel == NULL)
784 return -EINVAL;
785 daddr = &flowlabel->dst;
786 }
787 }
788
789 /*
790 * Otherwise it will be difficult to maintain
791 * sk->sk_dst_cache.
792 */
793 if (sk->sk_state == TCP_ESTABLISHED &&
794 ipv6_addr_equal(daddr, &np->daddr))
795 daddr = &np->daddr;
796
797 if (addr_len >= sizeof(struct sockaddr_in6) &&
798 sin6->sin6_scope_id &&
799 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
800 fl.oif = sin6->sin6_scope_id;
801 } else {
802 if (sk->sk_state != TCP_ESTABLISHED)
803 return -EDESTADDRREQ;
804
805 proto = inet->num;
806 daddr = &np->daddr;
807 fl.fl6_flowlabel = np->flow_label;
808 }
809
810 if (fl.oif == 0)
811 fl.oif = sk->sk_bound_dev_if;
812
813 if (msg->msg_controllen) {
814 opt = &opt_space;
815 memset(opt, 0, sizeof(struct ipv6_txoptions));
816 opt->tot_len = sizeof(struct ipv6_txoptions);
817
818 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
819 if (err < 0) {
820 fl6_sock_release(flowlabel);
821 return err;
822 }
823 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
824 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
825 if (flowlabel == NULL)
826 return -EINVAL;
827 }
828 if (!(opt->opt_nflen|opt->opt_flen))
829 opt = NULL;
830 }
831 if (opt == NULL)
832 opt = np->opt;
833 if (flowlabel)
834 opt = fl6_merge_options(&opt_space, flowlabel, opt);
835 opt = ipv6_fixup_options(&opt_space, opt);
836
837 fl.proto = proto;
838 err = rawv6_probe_proto_opt(&fl, msg);
839 if (err)
840 goto out;
841
842 if (!ipv6_addr_any(daddr))
843 ipv6_addr_copy(&fl.fl6_dst, daddr);
844 else
845 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
846 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
847 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
848
849 /* merge ip6_build_xmit from ip6_output */
850 if (opt && opt->srcrt) {
851 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
852 ipv6_addr_copy(&final, &fl.fl6_dst);
853 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
854 final_p = &final;
855 }
856
857 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
858 fl.oif = np->mcast_oif;
859 security_sk_classify_flow(sk, &fl);
860
861 err = ip6_dst_lookup(sk, &dst, &fl);
862 if (err)
863 goto out;
864 if (final_p)
865 ipv6_addr_copy(&fl.fl6_dst, final_p);
866
867 if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 0) {
868 if (err == -EREMOTE)
869 err = ip6_dst_blackhole(sk, &dst, &fl);
870 if (err < 0)
871 goto out;
872 }
873
874 if (hlimit < 0) {
875 if (ipv6_addr_is_multicast(&fl.fl6_dst))
876 hlimit = np->mcast_hops;
877 else
878 hlimit = np->hop_limit;
879 if (hlimit < 0)
880 hlimit = ip6_dst_hoplimit(dst);
881 }
882
883 if (tclass < 0) {
884 tclass = np->tclass;
885 if (tclass < 0)
886 tclass = 0;
887 }
888
889 if (msg->msg_flags&MSG_CONFIRM)
890 goto do_confirm;
891
892 back_from_confirm:
893 if (inet->hdrincl) {
894 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
895 } else {
896 lock_sock(sk);
897 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
898 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
899 msg->msg_flags);
900
901 if (err)
902 ip6_flush_pending_frames(sk);
903 else if (!(msg->msg_flags & MSG_MORE))
904 err = rawv6_push_pending_frames(sk, &fl, rp);
905 release_sock(sk);
906 }
907 done:
908 dst_release(dst);
909 out:
910 fl6_sock_release(flowlabel);
911 return err<0?err:len;
912 do_confirm:
913 dst_confirm(dst);
914 if (!(msg->msg_flags & MSG_PROBE) || len)
915 goto back_from_confirm;
916 err = 0;
917 goto done;
918 }
919
920 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
921 char __user *optval, int optlen)
922 {
923 switch (optname) {
924 case ICMPV6_FILTER:
925 if (optlen > sizeof(struct icmp6_filter))
926 optlen = sizeof(struct icmp6_filter);
927 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
928 return -EFAULT;
929 return 0;
930 default:
931 return -ENOPROTOOPT;
932 }
933
934 return 0;
935 }
936
937 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
938 char __user *optval, int __user *optlen)
939 {
940 int len;
941
942 switch (optname) {
943 case ICMPV6_FILTER:
944 if (get_user(len, optlen))
945 return -EFAULT;
946 if (len < 0)
947 return -EINVAL;
948 if (len > sizeof(struct icmp6_filter))
949 len = sizeof(struct icmp6_filter);
950 if (put_user(len, optlen))
951 return -EFAULT;
952 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
953 return -EFAULT;
954 return 0;
955 default:
956 return -ENOPROTOOPT;
957 }
958
959 return 0;
960 }
961
962
963 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
964 char __user *optval, int optlen)
965 {
966 struct raw6_sock *rp = raw6_sk(sk);
967 int val;
968
969 if (get_user(val, (int __user *)optval))
970 return -EFAULT;
971
972 switch (optname) {
973 case IPV6_CHECKSUM:
974 if (inet_sk(sk)->num == IPPROTO_ICMPV6 &&
975 level == IPPROTO_IPV6) {
976 /*
977 * RFC3542 tells that IPV6_CHECKSUM socket
978 * option in the IPPROTO_IPV6 level is not
979 * allowed on ICMPv6 sockets.
980 * If you want to set it, use IPPROTO_RAW
981 * level IPV6_CHECKSUM socket option
982 * (Linux extension).
983 */
984 return -EINVAL;
985 }
986
987 /* You may get strange result with a positive odd offset;
988 RFC2292bis agrees with me. */
989 if (val > 0 && (val&1))
990 return(-EINVAL);
991 if (val < 0) {
992 rp->checksum = 0;
993 } else {
994 rp->checksum = 1;
995 rp->offset = val;
996 }
997
998 return 0;
999 break;
1000
1001 default:
1002 return(-ENOPROTOOPT);
1003 }
1004 }
1005
1006 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1007 char __user *optval, int optlen)
1008 {
1009 switch(level) {
1010 case SOL_RAW:
1011 break;
1012
1013 case SOL_ICMPV6:
1014 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1015 return -EOPNOTSUPP;
1016 return rawv6_seticmpfilter(sk, level, optname, optval,
1017 optlen);
1018 case SOL_IPV6:
1019 if (optname == IPV6_CHECKSUM)
1020 break;
1021 default:
1022 return ipv6_setsockopt(sk, level, optname, optval,
1023 optlen);
1024 }
1025
1026 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1027 }
1028
1029 #ifdef CONFIG_COMPAT
1030 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1031 char __user *optval, int optlen)
1032 {
1033 switch (level) {
1034 case SOL_RAW:
1035 break;
1036 case SOL_ICMPV6:
1037 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1038 return -EOPNOTSUPP;
1039 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1040 case SOL_IPV6:
1041 if (optname == IPV6_CHECKSUM)
1042 break;
1043 default:
1044 return compat_ipv6_setsockopt(sk, level, optname,
1045 optval, optlen);
1046 }
1047 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1048 }
1049 #endif
1050
1051 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1052 char __user *optval, int __user *optlen)
1053 {
1054 struct raw6_sock *rp = raw6_sk(sk);
1055 int val, len;
1056
1057 if (get_user(len,optlen))
1058 return -EFAULT;
1059
1060 switch (optname) {
1061 case IPV6_CHECKSUM:
1062 /*
1063 * We allow getsockopt() for IPPROTO_IPV6-level
1064 * IPV6_CHECKSUM socket option on ICMPv6 sockets
1065 * since RFC3542 is silent about it.
1066 */
1067 if (rp->checksum == 0)
1068 val = -1;
1069 else
1070 val = rp->offset;
1071 break;
1072
1073 default:
1074 return -ENOPROTOOPT;
1075 }
1076
1077 len = min_t(unsigned int, sizeof(int), len);
1078
1079 if (put_user(len, optlen))
1080 return -EFAULT;
1081 if (copy_to_user(optval,&val,len))
1082 return -EFAULT;
1083 return 0;
1084 }
1085
1086 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1087 char __user *optval, int __user *optlen)
1088 {
1089 switch(level) {
1090 case SOL_RAW:
1091 break;
1092
1093 case SOL_ICMPV6:
1094 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1095 return -EOPNOTSUPP;
1096 return rawv6_geticmpfilter(sk, level, optname, optval,
1097 optlen);
1098 case SOL_IPV6:
1099 if (optname == IPV6_CHECKSUM)
1100 break;
1101 default:
1102 return ipv6_getsockopt(sk, level, optname, optval,
1103 optlen);
1104 }
1105
1106 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1107 }
1108
1109 #ifdef CONFIG_COMPAT
1110 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1111 char __user *optval, int __user *optlen)
1112 {
1113 switch (level) {
1114 case SOL_RAW:
1115 break;
1116 case SOL_ICMPV6:
1117 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1118 return -EOPNOTSUPP;
1119 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1120 case SOL_IPV6:
1121 if (optname == IPV6_CHECKSUM)
1122 break;
1123 default:
1124 return compat_ipv6_getsockopt(sk, level, optname,
1125 optval, optlen);
1126 }
1127 return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1128 }
1129 #endif
1130
1131 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1132 {
1133 switch(cmd) {
1134 case SIOCOUTQ:
1135 {
1136 int amount = atomic_read(&sk->sk_wmem_alloc);
1137 return put_user(amount, (int __user *)arg);
1138 }
1139 case SIOCINQ:
1140 {
1141 struct sk_buff *skb;
1142 int amount = 0;
1143
1144 spin_lock_bh(&sk->sk_receive_queue.lock);
1145 skb = skb_peek(&sk->sk_receive_queue);
1146 if (skb != NULL)
1147 amount = skb->tail - skb->transport_header;
1148 spin_unlock_bh(&sk->sk_receive_queue.lock);
1149 return put_user(amount, (int __user *)arg);
1150 }
1151
1152 default:
1153 #ifdef CONFIG_IPV6_MROUTE
1154 return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1155 #else
1156 return -ENOIOCTLCMD;
1157 #endif
1158 }
1159 }
1160
1161 static void rawv6_close(struct sock *sk, long timeout)
1162 {
1163 if (inet_sk(sk)->num == IPPROTO_RAW)
1164 ip6_ra_control(sk, -1, NULL);
1165 ip6mr_sk_done(sk);
1166 sk_common_release(sk);
1167 }
1168
1169 static int rawv6_init_sk(struct sock *sk)
1170 {
1171 struct raw6_sock *rp = raw6_sk(sk);
1172
1173 switch (inet_sk(sk)->num) {
1174 case IPPROTO_ICMPV6:
1175 rp->checksum = 1;
1176 rp->offset = 2;
1177 break;
1178 case IPPROTO_MH:
1179 rp->checksum = 1;
1180 rp->offset = 4;
1181 break;
1182 default:
1183 break;
1184 }
1185 return(0);
1186 }
1187
1188 struct proto rawv6_prot = {
1189 .name = "RAWv6",
1190 .owner = THIS_MODULE,
1191 .close = rawv6_close,
1192 .connect = ip6_datagram_connect,
1193 .disconnect = udp_disconnect,
1194 .ioctl = rawv6_ioctl,
1195 .init = rawv6_init_sk,
1196 .destroy = inet6_destroy_sock,
1197 .setsockopt = rawv6_setsockopt,
1198 .getsockopt = rawv6_getsockopt,
1199 .sendmsg = rawv6_sendmsg,
1200 .recvmsg = rawv6_recvmsg,
1201 .bind = rawv6_bind,
1202 .backlog_rcv = rawv6_rcv_skb,
1203 .hash = raw_hash_sk,
1204 .unhash = raw_unhash_sk,
1205 .obj_size = sizeof(struct raw6_sock),
1206 .h.raw_hash = &raw_v6_hashinfo,
1207 #ifdef CONFIG_COMPAT
1208 .compat_setsockopt = compat_rawv6_setsockopt,
1209 .compat_getsockopt = compat_rawv6_getsockopt,
1210 #endif
1211 };
1212
1213 #ifdef CONFIG_PROC_FS
1214 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1215 {
1216 struct ipv6_pinfo *np = inet6_sk(sp);
1217 struct in6_addr *dest, *src;
1218 __u16 destp, srcp;
1219
1220 dest = &np->daddr;
1221 src = &np->rcv_saddr;
1222 destp = 0;
1223 srcp = inet_sk(sp)->num;
1224 seq_printf(seq,
1225 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1226 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1227 i,
1228 src->s6_addr32[0], src->s6_addr32[1],
1229 src->s6_addr32[2], src->s6_addr32[3], srcp,
1230 dest->s6_addr32[0], dest->s6_addr32[1],
1231 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1232 sp->sk_state,
1233 atomic_read(&sp->sk_wmem_alloc),
1234 atomic_read(&sp->sk_rmem_alloc),
1235 0, 0L, 0,
1236 sock_i_uid(sp), 0,
1237 sock_i_ino(sp),
1238 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1239 }
1240
1241 static int raw6_seq_show(struct seq_file *seq, void *v)
1242 {
1243 if (v == SEQ_START_TOKEN)
1244 seq_printf(seq,
1245 " sl "
1246 "local_address "
1247 "remote_address "
1248 "st tx_queue rx_queue tr tm->when retrnsmt"
1249 " uid timeout inode drops\n");
1250 else
1251 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1252 return 0;
1253 }
1254
1255 static const struct seq_operations raw6_seq_ops = {
1256 .start = raw_seq_start,
1257 .next = raw_seq_next,
1258 .stop = raw_seq_stop,
1259 .show = raw6_seq_show,
1260 };
1261
1262 static int raw6_seq_open(struct inode *inode, struct file *file)
1263 {
1264 return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1265 }
1266
1267 static const struct file_operations raw6_seq_fops = {
1268 .owner = THIS_MODULE,
1269 .open = raw6_seq_open,
1270 .read = seq_read,
1271 .llseek = seq_lseek,
1272 .release = seq_release_net,
1273 };
1274
1275 static int raw6_init_net(struct net *net)
1276 {
1277 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1278 return -ENOMEM;
1279
1280 return 0;
1281 }
1282
1283 static void raw6_exit_net(struct net *net)
1284 {
1285 proc_net_remove(net, "raw6");
1286 }
1287
1288 static struct pernet_operations raw6_net_ops = {
1289 .init = raw6_init_net,
1290 .exit = raw6_exit_net,
1291 };
1292
1293 int __init raw6_proc_init(void)
1294 {
1295 return register_pernet_subsys(&raw6_net_ops);
1296 }
1297
1298 void raw6_proc_exit(void)
1299 {
1300 unregister_pernet_subsys(&raw6_net_ops);
1301 }
1302 #endif /* CONFIG_PROC_FS */
1303
1304 /* Same as inet6_dgram_ops, sans udp_poll. */
1305 static const struct proto_ops inet6_sockraw_ops = {
1306 .family = PF_INET6,
1307 .owner = THIS_MODULE,
1308 .release = inet6_release,
1309 .bind = inet6_bind,
1310 .connect = inet_dgram_connect, /* ok */
1311 .socketpair = sock_no_socketpair, /* a do nothing */
1312 .accept = sock_no_accept, /* a do nothing */
1313 .getname = inet6_getname,
1314 .poll = datagram_poll, /* ok */
1315 .ioctl = inet6_ioctl, /* must change */
1316 .listen = sock_no_listen, /* ok */
1317 .shutdown = inet_shutdown, /* ok */
1318 .setsockopt = sock_common_setsockopt, /* ok */
1319 .getsockopt = sock_common_getsockopt, /* ok */
1320 .sendmsg = inet_sendmsg, /* ok */
1321 .recvmsg = sock_common_recvmsg, /* ok */
1322 .mmap = sock_no_mmap,
1323 .sendpage = sock_no_sendpage,
1324 #ifdef CONFIG_COMPAT
1325 .compat_setsockopt = compat_sock_common_setsockopt,
1326 .compat_getsockopt = compat_sock_common_getsockopt,
1327 #endif
1328 };
1329
1330 static struct inet_protosw rawv6_protosw = {
1331 .type = SOCK_RAW,
1332 .protocol = IPPROTO_IP, /* wild card */
1333 .prot = &rawv6_prot,
1334 .ops = &inet6_sockraw_ops,
1335 .capability = CAP_NET_RAW,
1336 .no_check = UDP_CSUM_DEFAULT,
1337 .flags = INET_PROTOSW_REUSE,
1338 };
1339
1340 int __init rawv6_init(void)
1341 {
1342 int ret;
1343
1344 ret = inet6_register_protosw(&rawv6_protosw);
1345 if (ret)
1346 goto out;
1347 out:
1348 return ret;
1349 }
1350
1351 void rawv6_exit(void)
1352 {
1353 inet6_unregister_protosw(&rawv6_protosw);
1354 }
kernel-based virtual machine