?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
[SCSI] tgt: use scsi_init_io instead of scsi_alloc_sgtable
[linux-2.6/kvm.git] / net / ipv6 / raw.c
blob4d880551fe6ab327260a09401d2a5a4f3c9e7315
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
57 #include <net/raw.h>
58 #include <net/rawv6.h>
59 #include <net/xfrm.h>
60
61 #include <linux/proc_fs.h>
62 #include <linux/seq_file.h>
63
64 static struct raw_hashinfo raw_v6_hashinfo = {
65 .lock = __RW_LOCK_UNLOCKED(),
66 };
67
68 static void raw_v6_hash(struct sock *sk)
69 {
70 raw_hash_sk(sk, &raw_v6_hashinfo);
71 }
72
73 static void raw_v6_unhash(struct sock *sk)
74 {
75 raw_unhash_sk(sk, &raw_v6_hashinfo);
76 }
77
78
79 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
80 unsigned short num, struct in6_addr *loc_addr,
81 struct in6_addr *rmt_addr, int dif)
82 {
83 struct hlist_node *node;
84 int is_multicast = ipv6_addr_is_multicast(loc_addr);
85
86 sk_for_each_from(sk, node)
87 if (inet_sk(sk)->num == num) {
88 struct ipv6_pinfo *np = inet6_sk(sk);
89
90 if (sk->sk_net != net)
91 continue;
92
93 if (!ipv6_addr_any(&np->daddr) &&
94 !ipv6_addr_equal(&np->daddr, rmt_addr))
95 continue;
96
97 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
98 continue;
99
100 if (!ipv6_addr_any(&np->rcv_saddr)) {
101 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
102 goto found;
103 if (is_multicast &&
104 inet6_mc_check(sk, loc_addr, rmt_addr))
105 goto found;
106 continue;
107 }
108 goto found;
109 }
110 sk = NULL;
111 found:
112 return sk;
113 }
114
115 /*
116 * 0 - deliver
117 * 1 - block
118 */
119 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
120 {
121 struct icmp6hdr *icmph;
122 struct raw6_sock *rp = raw6_sk(sk);
123
124 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
125 __u32 *data = &rp->filter.data[0];
126 int bit_nr;
127
128 icmph = (struct icmp6hdr *) skb->data;
129 bit_nr = icmph->icmp6_type;
130
131 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
132 }
133 return 0;
134 }
135
136 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
137 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
138
139 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
140 struct sk_buff *skb))
141 {
142 rcu_assign_pointer(mh_filter, filter);
143 return 0;
144 }
145 EXPORT_SYMBOL(rawv6_mh_filter_register);
146
147 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
148 struct sk_buff *skb))
149 {
150 rcu_assign_pointer(mh_filter, NULL);
151 synchronize_rcu();
152 return 0;
153 }
154 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
155
156 #endif
157
158 /*
159 * demultiplex raw sockets.
160 * (should consider queueing the skb in the sock receive_queue
161 * without calling rawv6.c)
162 *
163 * Caller owns SKB so we must make clones.
164 */
165 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
166 {
167 struct in6_addr *saddr;
168 struct in6_addr *daddr;
169 struct sock *sk;
170 int delivered = 0;
171 __u8 hash;
172 struct net *net;
173
174 saddr = &ipv6_hdr(skb)->saddr;
175 daddr = saddr + 1;
176
177 hash = nexthdr & (MAX_INET_PROTOS - 1);
178
179 read_lock(&raw_v6_hashinfo.lock);
180 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
181
182 /*
183 * The first socket found will be delivered after
184 * delivery to transport protocols.
185 */
186
187 if (sk == NULL)
188 goto out;
189
190 net = skb->dev->nd_net;
191 sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
192
193 while (sk) {
194 int filtered;
195
196 delivered = 1;
197 switch (nexthdr) {
198 case IPPROTO_ICMPV6:
199 filtered = icmpv6_filter(sk, skb);
200 break;
201
202 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
203 case IPPROTO_MH:
204 {
205 /* XXX: To validate MH only once for each packet,
206 * this is placed here. It should be after checking
207 * xfrm policy, however it doesn't. The checking xfrm
208 * policy is placed in rawv6_rcv() because it is
209 * required for each socket.
210 */
211 int (*filter)(struct sock *sock, struct sk_buff *skb);
212
213 filter = rcu_dereference(mh_filter);
214 filtered = filter ? filter(sk, skb) : 0;
215 break;
216 }
217 #endif
218 default:
219 filtered = 0;
220 break;
221 }
222
223 if (filtered < 0)
224 break;
225 if (filtered == 0) {
226 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
227
228 /* Not releasing hash table! */
229 if (clone) {
230 nf_reset(clone);
231 rawv6_rcv(sk, clone);
232 }
233 }
234 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
235 IP6CB(skb)->iif);
236 }
237 out:
238 read_unlock(&raw_v6_hashinfo.lock);
239 return delivered;
240 }
241
242 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
243 {
244 struct sock *raw_sk;
245
246 raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
247 if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
248 raw_sk = NULL;
249
250 return raw_sk != NULL;
251 }
252
253 /* This cleans up af_inet6 a bit. -DaveM */
254 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
255 {
256 struct inet_sock *inet = inet_sk(sk);
257 struct ipv6_pinfo *np = inet6_sk(sk);
258 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
259 __be32 v4addr = 0;
260 int addr_type;
261 int err;
262
263 if (addr_len < SIN6_LEN_RFC2133)
264 return -EINVAL;
265 addr_type = ipv6_addr_type(&addr->sin6_addr);
266
267 /* Raw sockets are IPv6 only */
268 if (addr_type == IPV6_ADDR_MAPPED)
269 return(-EADDRNOTAVAIL);
270
271 lock_sock(sk);
272
273 err = -EINVAL;
274 if (sk->sk_state != TCP_CLOSE)
275 goto out;
276
277 /* Check if the address belongs to the host. */
278 if (addr_type != IPV6_ADDR_ANY) {
279 struct net_device *dev = NULL;
280
281 if (addr_type & IPV6_ADDR_LINKLOCAL) {
282 if (addr_len >= sizeof(struct sockaddr_in6) &&
283 addr->sin6_scope_id) {
284 /* Override any existing binding, if another
285 * one is supplied by user.
286 */
287 sk->sk_bound_dev_if = addr->sin6_scope_id;
288 }
289
290 /* Binding to link-local address requires an interface */
291 if (!sk->sk_bound_dev_if)
292 goto out;
293
294 dev = dev_get_by_index(sk->sk_net, sk->sk_bound_dev_if);
295 if (!dev) {
296 err = -ENODEV;
297 goto out;
298 }
299 }
300
301 /* ipv4 addr of the socket is invalid. Only the
302 * unspecified and mapped address have a v4 equivalent.
303 */
304 v4addr = LOOPBACK4_IPV6;
305 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
306 err = -EADDRNOTAVAIL;
307 if (!ipv6_chk_addr(sk->sk_net, &addr->sin6_addr,
308 dev, 0)) {
309 if (dev)
310 dev_put(dev);
311 goto out;
312 }
313 }
314 if (dev)
315 dev_put(dev);
316 }
317
318 inet->rcv_saddr = inet->saddr = v4addr;
319 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
320 if (!(addr_type & IPV6_ADDR_MULTICAST))
321 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
322 err = 0;
323 out:
324 release_sock(sk);
325 return err;
326 }
327
328 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
329 struct inet6_skb_parm *opt,
330 int type, int code, int offset, __be32 info)
331 {
332 struct inet_sock *inet = inet_sk(sk);
333 struct ipv6_pinfo *np = inet6_sk(sk);
334 int err;
335 int harderr;
336
337 /* Report error on raw socket, if:
338 1. User requested recverr.
339 2. Socket is connected (otherwise the error indication
340 is useless without recverr and error is hard.
341 */
342 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
343 return;
344
345 harderr = icmpv6_err_convert(type, code, &err);
346 if (type == ICMPV6_PKT_TOOBIG)
347 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
348
349 if (np->recverr) {
350 u8 *payload = skb->data;
351 if (!inet->hdrincl)
352 payload += offset;
353 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
354 }
355
356 if (np->recverr || harderr) {
357 sk->sk_err = err;
358 sk->sk_error_report(sk);
359 }
360 }
361
362 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
363 int type, int code, int inner_offset, __be32 info)
364 {
365 struct sock *sk;
366 int hash;
367 struct in6_addr *saddr, *daddr;
368 struct net *net;
369
370 hash = nexthdr & (RAW_HTABLE_SIZE - 1);
371
372 read_lock(&raw_v6_hashinfo.lock);
373 sk = sk_head(&raw_v6_hashinfo.ht[hash]);
374 if (sk != NULL) {
375 saddr = &ipv6_hdr(skb)->saddr;
376 daddr = &ipv6_hdr(skb)->daddr;
377 net = skb->dev->nd_net;
378
379 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
380 IP6CB(skb)->iif))) {
381 rawv6_err(sk, skb, NULL, type, code,
382 inner_offset, info);
383 sk = sk_next(sk);
384 }
385 }
386 read_unlock(&raw_v6_hashinfo.lock);
387 }
388
389 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
390 {
391 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
392 skb_checksum_complete(skb)) {
393 atomic_inc(&sk->sk_drops);
394 kfree_skb(skb);
395 return 0;
396 }
397
398 /* Charge it to the socket. */
399 if (sock_queue_rcv_skb(sk,skb)<0) {
400 atomic_inc(&sk->sk_drops);
401 kfree_skb(skb);
402 return 0;
403 }
404
405 return 0;
406 }
407
408 /*
409 * This is next to useless...
410 * if we demultiplex in network layer we don't need the extra call
411 * just to queue the skb...
412 * maybe we could have the network decide upon a hint if it
413 * should call raw_rcv for demultiplexing
414 */
415 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
416 {
417 struct inet_sock *inet = inet_sk(sk);
418 struct raw6_sock *rp = raw6_sk(sk);
419
420 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
421 atomic_inc(&sk->sk_drops);
422 kfree_skb(skb);
423 return NET_RX_DROP;
424 }
425
426 if (!rp->checksum)
427 skb->ip_summed = CHECKSUM_UNNECESSARY;
428
429 if (skb->ip_summed == CHECKSUM_COMPLETE) {
430 skb_postpull_rcsum(skb, skb_network_header(skb),
431 skb_network_header_len(skb));
432 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
433 &ipv6_hdr(skb)->daddr,
434 skb->len, inet->num, skb->csum))
435 skb->ip_summed = CHECKSUM_UNNECESSARY;
436 }
437 if (!skb_csum_unnecessary(skb))
438 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
439 &ipv6_hdr(skb)->daddr,
440 skb->len,
441 inet->num, 0));
442
443 if (inet->hdrincl) {
444 if (skb_checksum_complete(skb)) {
445 atomic_inc(&sk->sk_drops);
446 kfree_skb(skb);
447 return 0;
448 }
449 }
450
451 rawv6_rcv_skb(sk, skb);
452 return 0;
453 }
454
455
456 /*
457 * This should be easy, if there is something there
458 * we return it, otherwise we block.
459 */
460
461 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
462 struct msghdr *msg, size_t len,
463 int noblock, int flags, int *addr_len)
464 {
465 struct ipv6_pinfo *np = inet6_sk(sk);
466 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
467 struct sk_buff *skb;
468 size_t copied;
469 int err;
470
471 if (flags & MSG_OOB)
472 return -EOPNOTSUPP;
473
474 if (addr_len)
475 *addr_len=sizeof(*sin6);
476
477 if (flags & MSG_ERRQUEUE)
478 return ipv6_recv_error(sk, msg, len);
479
480 skb = skb_recv_datagram(sk, flags, noblock, &err);
481 if (!skb)
482 goto out;
483
484 copied = skb->len;
485 if (copied > len) {
486 copied = len;
487 msg->msg_flags |= MSG_TRUNC;
488 }
489
490 if (skb_csum_unnecessary(skb)) {
491 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
492 } else if (msg->msg_flags&MSG_TRUNC) {
493 if (__skb_checksum_complete(skb))
494 goto csum_copy_err;
495 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
496 } else {
497 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
498 if (err == -EINVAL)
499 goto csum_copy_err;
500 }
501 if (err)
502 goto out_free;
503
504 /* Copy the address. */
505 if (sin6) {
506 sin6->sin6_family = AF_INET6;
507 sin6->sin6_port = 0;
508 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
509 sin6->sin6_flowinfo = 0;
510 sin6->sin6_scope_id = 0;
511 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
512 sin6->sin6_scope_id = IP6CB(skb)->iif;
513 }
514
515 sock_recv_timestamp(msg, sk, skb);
516
517 if (np->rxopt.all)
518 datagram_recv_ctl(sk, msg, skb);
519
520 err = copied;
521 if (flags & MSG_TRUNC)
522 err = skb->len;
523
524 out_free:
525 skb_free_datagram(sk, skb);
526 out:
527 return err;
528
529 csum_copy_err:
530 skb_kill_datagram(sk, skb, flags);
531
532 /* Error for blocking case is chosen to masquerade
533 as some normal condition.
534 */
535 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
536 atomic_inc(&sk->sk_drops);
537 goto out;
538 }
539
540 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
541 struct raw6_sock *rp)
542 {
543 struct sk_buff *skb;
544 int err = 0;
545 int offset;
546 int len;
547 int total_len;
548 __wsum tmp_csum;
549 __sum16 csum;
550
551 if (!rp->checksum)
552 goto send;
553
554 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
555 goto out;
556
557 offset = rp->offset;
558 total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
559 skb->data);
560 if (offset >= total_len - 1) {
561 err = -EINVAL;
562 ip6_flush_pending_frames(sk);
563 goto out;
564 }
565
566 /* should be check HW csum miyazawa */
567 if (skb_queue_len(&sk->sk_write_queue) == 1) {
568 /*
569 * Only one fragment on the socket.
570 */
571 tmp_csum = skb->csum;
572 } else {
573 struct sk_buff *csum_skb = NULL;
574 tmp_csum = 0;
575
576 skb_queue_walk(&sk->sk_write_queue, skb) {
577 tmp_csum = csum_add(tmp_csum, skb->csum);
578
579 if (csum_skb)
580 continue;
581
582 len = skb->len - skb_transport_offset(skb);
583 if (offset >= len) {
584 offset -= len;
585 continue;
586 }
587
588 csum_skb = skb;
589 }
590
591 skb = csum_skb;
592 }
593
594 offset += skb_transport_offset(skb);
595 if (skb_copy_bits(skb, offset, &csum, 2))
596 BUG();
597
598 /* in case cksum was not initialized */
599 if (unlikely(csum))
600 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
601
602 csum = csum_ipv6_magic(&fl->fl6_src,
603 &fl->fl6_dst,
604 total_len, fl->proto, tmp_csum);
605
606 if (csum == 0 && fl->proto == IPPROTO_UDP)
607 csum = CSUM_MANGLED_0;
608
609 if (skb_store_bits(skb, offset, &csum, 2))
610 BUG();
611
612 send:
613 err = ip6_push_pending_frames(sk);
614 out:
615 return err;
616 }
617
618 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
619 struct flowi *fl, struct rt6_info *rt,
620 unsigned int flags)
621 {
622 struct ipv6_pinfo *np = inet6_sk(sk);
623 struct ipv6hdr *iph;
624 struct sk_buff *skb;
625 unsigned int hh_len;
626 int err;
627
628 if (length > rt->u.dst.dev->mtu) {
629 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
630 return -EMSGSIZE;
631 }
632 if (flags&MSG_PROBE)
633 goto out;
634
635 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
636
637 skb = sock_alloc_send_skb(sk, length+hh_len+15,
638 flags&MSG_DONTWAIT, &err);
639 if (skb == NULL)
640 goto error;
641 skb_reserve(skb, hh_len);
642
643 skb->priority = sk->sk_priority;
644 skb->dst = dst_clone(&rt->u.dst);
645
646 skb_put(skb, length);
647 skb_reset_network_header(skb);
648 iph = ipv6_hdr(skb);
649
650 skb->ip_summed = CHECKSUM_NONE;
651
652 skb->transport_header = skb->network_header;
653 err = memcpy_fromiovecend((void *)iph, from, 0, length);
654 if (err)
655 goto error_fault;
656
657 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
658 err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
659 dst_output);
660 if (err > 0)
661 err = np->recverr ? net_xmit_errno(err) : 0;
662 if (err)
663 goto error;
664 out:
665 return 0;
666
667 error_fault:
668 err = -EFAULT;
669 kfree_skb(skb);
670 error:
671 IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
672 return err;
673 }
674
675 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
676 {
677 struct iovec *iov;
678 u8 __user *type = NULL;
679 u8 __user *code = NULL;
680 u8 len = 0;
681 int probed = 0;
682 int i;
683
684 if (!msg->msg_iov)
685 return 0;
686
687 for (i = 0; i < msg->msg_iovlen; i++) {
688 iov = &msg->msg_iov[i];
689 if (!iov)
690 continue;
691
692 switch (fl->proto) {
693 case IPPROTO_ICMPV6:
694 /* check if one-byte field is readable or not. */
695 if (iov->iov_base && iov->iov_len < 1)
696 break;
697
698 if (!type) {
699 type = iov->iov_base;
700 /* check if code field is readable or not. */
701 if (iov->iov_len > 1)
702 code = type + 1;
703 } else if (!code)
704 code = iov->iov_base;
705
706 if (type && code) {
707 if (get_user(fl->fl_icmp_type, type) ||
708 get_user(fl->fl_icmp_code, code))
709 return -EFAULT;
710 probed = 1;
711 }
712 break;
713 case IPPROTO_MH:
714 if (iov->iov_base && iov->iov_len < 1)
715 break;
716 /* check if type field is readable or not. */
717 if (iov->iov_len > 2 - len) {
718 u8 __user *p = iov->iov_base;
719 if (get_user(fl->fl_mh_type, &p[2 - len]))
720 return -EFAULT;
721 probed = 1;
722 } else
723 len += iov->iov_len;
724
725 break;
726 default:
727 probed = 1;
728 break;
729 }
730 if (probed)
731 break;
732 }
733 return 0;
734 }
735
736 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
737 struct msghdr *msg, size_t len)
738 {
739 struct ipv6_txoptions opt_space;
740 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
741 struct in6_addr *daddr, *final_p = NULL, final;
742 struct inet_sock *inet = inet_sk(sk);
743 struct ipv6_pinfo *np = inet6_sk(sk);
744 struct raw6_sock *rp = raw6_sk(sk);
745 struct ipv6_txoptions *opt = NULL;
746 struct ip6_flowlabel *flowlabel = NULL;
747 struct dst_entry *dst = NULL;
748 struct flowi fl;
749 int addr_len = msg->msg_namelen;
750 int hlimit = -1;
751 int tclass = -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(&fl, 0, sizeof(fl));
769
770 if (sin6) {
771 if (addr_len < SIN6_LEN_RFC2133)
772 return -EINVAL;
773
774 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
775 return(-EAFNOSUPPORT);
776
777 /* port is the proto value [0..255] carried in nexthdr */
778 proto = ntohs(sin6->sin6_port);
779
780 if (!proto)
781 proto = inet->num;
782 else if (proto != inet->num)
783 return(-EINVAL);
784
785 if (proto > 255)
786 return(-EINVAL);
787
788 daddr = &sin6->sin6_addr;
789 if (np->sndflow) {
790 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
791 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
792 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
793 if (flowlabel == NULL)
794 return -EINVAL;
795 daddr = &flowlabel->dst;
796 }
797 }
798
799 /*
800 * Otherwise it will be difficult to maintain
801 * sk->sk_dst_cache.
802 */
803 if (sk->sk_state == TCP_ESTABLISHED &&
804 ipv6_addr_equal(daddr, &np->daddr))
805 daddr = &np->daddr;
806
807 if (addr_len >= sizeof(struct sockaddr_in6) &&
808 sin6->sin6_scope_id &&
809 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
810 fl.oif = sin6->sin6_scope_id;
811 } else {
812 if (sk->sk_state != TCP_ESTABLISHED)
813 return -EDESTADDRREQ;
814
815 proto = inet->num;
816 daddr = &np->daddr;
817 fl.fl6_flowlabel = np->flow_label;
818 }
819
820 if (ipv6_addr_any(daddr)) {
821 /*
822 * unspecified destination address
823 * treated as error... is this correct ?
824 */
825 fl6_sock_release(flowlabel);
826 return(-EINVAL);
827 }
828
829 if (fl.oif == 0)
830 fl.oif = sk->sk_bound_dev_if;
831
832 if (msg->msg_controllen) {
833 opt = &opt_space;
834 memset(opt, 0, sizeof(struct ipv6_txoptions));
835 opt->tot_len = sizeof(struct ipv6_txoptions);
836
837 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
838 if (err < 0) {
839 fl6_sock_release(flowlabel);
840 return err;
841 }
842 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
843 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
844 if (flowlabel == NULL)
845 return -EINVAL;
846 }
847 if (!(opt->opt_nflen|opt->opt_flen))
848 opt = NULL;
849 }
850 if (opt == NULL)
851 opt = np->opt;
852 if (flowlabel)
853 opt = fl6_merge_options(&opt_space, flowlabel, opt);
854 opt = ipv6_fixup_options(&opt_space, opt);
855
856 fl.proto = proto;
857 err = rawv6_probe_proto_opt(&fl, msg);
858 if (err)
859 goto out;
860
861 ipv6_addr_copy(&fl.fl6_dst, daddr);
862 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
863 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
864
865 /* merge ip6_build_xmit from ip6_output */
866 if (opt && opt->srcrt) {
867 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
868 ipv6_addr_copy(&final, &fl.fl6_dst);
869 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
870 final_p = &final;
871 }
872
873 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
874 fl.oif = np->mcast_oif;
875 security_sk_classify_flow(sk, &fl);
876
877 err = ip6_dst_lookup(sk, &dst, &fl);
878 if (err)
879 goto out;
880 if (final_p)
881 ipv6_addr_copy(&fl.fl6_dst, final_p);
882
883 if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 0) {
884 if (err == -EREMOTE)
885 err = ip6_dst_blackhole(sk, &dst, &fl);
886 if (err < 0)
887 goto out;
888 }
889
890 if (hlimit < 0) {
891 if (ipv6_addr_is_multicast(&fl.fl6_dst))
892 hlimit = np->mcast_hops;
893 else
894 hlimit = np->hop_limit;
895 if (hlimit < 0)
896 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
897 if (hlimit < 0)
898 hlimit = ipv6_get_hoplimit(dst->dev);
899 }
900
901 if (tclass < 0) {
902 tclass = np->tclass;
903 if (tclass < 0)
904 tclass = 0;
905 }
906
907 if (msg->msg_flags&MSG_CONFIRM)
908 goto do_confirm;
909
910 back_from_confirm:
911 if (inet->hdrincl) {
912 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
913 } else {
914 lock_sock(sk);
915 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
916 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
917 msg->msg_flags);
918
919 if (err)
920 ip6_flush_pending_frames(sk);
921 else if (!(msg->msg_flags & MSG_MORE))
922 err = rawv6_push_pending_frames(sk, &fl, rp);
923 release_sock(sk);
924 }
925 done:
926 dst_release(dst);
927 out:
928 fl6_sock_release(flowlabel);
929 return err<0?err:len;
930 do_confirm:
931 dst_confirm(dst);
932 if (!(msg->msg_flags & MSG_PROBE) || len)
933 goto back_from_confirm;
934 err = 0;
935 goto done;
936 }
937
938 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
939 char __user *optval, int optlen)
940 {
941 switch (optname) {
942 case ICMPV6_FILTER:
943 if (optlen > sizeof(struct icmp6_filter))
944 optlen = sizeof(struct icmp6_filter);
945 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
946 return -EFAULT;
947 return 0;
948 default:
949 return -ENOPROTOOPT;
950 }
951
952 return 0;
953 }
954
955 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
956 char __user *optval, int __user *optlen)
957 {
958 int len;
959
960 switch (optname) {
961 case ICMPV6_FILTER:
962 if (get_user(len, optlen))
963 return -EFAULT;
964 if (len < 0)
965 return -EINVAL;
966 if (len > sizeof(struct icmp6_filter))
967 len = sizeof(struct icmp6_filter);
968 if (put_user(len, optlen))
969 return -EFAULT;
970 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
971 return -EFAULT;
972 return 0;
973 default:
974 return -ENOPROTOOPT;
975 }
976
977 return 0;
978 }
979
980
981 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
982 char __user *optval, int optlen)
983 {
984 struct raw6_sock *rp = raw6_sk(sk);
985 int val;
986
987 if (get_user(val, (int __user *)optval))
988 return -EFAULT;
989
990 switch (optname) {
991 case IPV6_CHECKSUM:
992 /* You may get strange result with a positive odd offset;
993 RFC2292bis agrees with me. */
994 if (val > 0 && (val&1))
995 return(-EINVAL);
996 if (val < 0) {
997 rp->checksum = 0;
998 } else {
999 rp->checksum = 1;
1000 rp->offset = val;
1001 }
1002
1003 return 0;
1004 break;
1005
1006 default:
1007 return(-ENOPROTOOPT);
1008 }
1009 }
1010
1011 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1012 char __user *optval, int optlen)
1013 {
1014 switch(level) {
1015 case SOL_RAW:
1016 break;
1017
1018 case SOL_ICMPV6:
1019 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1020 return -EOPNOTSUPP;
1021 return rawv6_seticmpfilter(sk, level, optname, optval,
1022 optlen);
1023 case SOL_IPV6:
1024 if (optname == IPV6_CHECKSUM)
1025 break;
1026 default:
1027 return ipv6_setsockopt(sk, level, optname, optval,
1028 optlen);
1029 }
1030
1031 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1032 }
1033
1034 #ifdef CONFIG_COMPAT
1035 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1036 char __user *optval, int optlen)
1037 {
1038 switch (level) {
1039 case SOL_RAW:
1040 break;
1041 case SOL_ICMPV6:
1042 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1043 return -EOPNOTSUPP;
1044 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1045 case SOL_IPV6:
1046 if (optname == IPV6_CHECKSUM)
1047 break;
1048 default:
1049 return compat_ipv6_setsockopt(sk, level, optname,
1050 optval, optlen);
1051 }
1052 return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1053 }
1054 #endif
1055
1056 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1057 char __user *optval, int __user *optlen)
1058 {
1059 struct raw6_sock *rp = raw6_sk(sk);
1060 int val, len;
1061
1062 if (get_user(len,optlen))
1063 return -EFAULT;
1064
1065 switch (optname) {
1066 case IPV6_CHECKSUM:
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 return -ENOIOCTLCMD;
1154 }
1155 }
1156
1157 static void rawv6_close(struct sock *sk, long timeout)
1158 {
1159 if (inet_sk(sk)->num == IPPROTO_RAW)
1160 ip6_ra_control(sk, -1, NULL);
1161
1162 sk_common_release(sk);
1163 }
1164
1165 static int rawv6_init_sk(struct sock *sk)
1166 {
1167 struct raw6_sock *rp = raw6_sk(sk);
1168
1169 switch (inet_sk(sk)->num) {
1170 case IPPROTO_ICMPV6:
1171 rp->checksum = 1;
1172 rp->offset = 2;
1173 break;
1174 case IPPROTO_MH:
1175 rp->checksum = 1;
1176 rp->offset = 4;
1177 break;
1178 default:
1179 break;
1180 }
1181 return(0);
1182 }
1183
1184 DEFINE_PROTO_INUSE(rawv6)
1185
1186 struct proto rawv6_prot = {
1187 .name = "RAWv6",
1188 .owner = THIS_MODULE,
1189 .close = rawv6_close,
1190 .connect = ip6_datagram_connect,
1191 .disconnect = udp_disconnect,
1192 .ioctl = rawv6_ioctl,
1193 .init = rawv6_init_sk,
1194 .destroy = inet6_destroy_sock,
1195 .setsockopt = rawv6_setsockopt,
1196 .getsockopt = rawv6_getsockopt,
1197 .sendmsg = rawv6_sendmsg,
1198 .recvmsg = rawv6_recvmsg,
1199 .bind = rawv6_bind,
1200 .backlog_rcv = rawv6_rcv_skb,
1201 .hash = raw_v6_hash,
1202 .unhash = raw_v6_unhash,
1203 .obj_size = sizeof(struct raw6_sock),
1204 #ifdef CONFIG_COMPAT
1205 .compat_setsockopt = compat_rawv6_setsockopt,
1206 .compat_getsockopt = compat_rawv6_getsockopt,
1207 #endif
1208 REF_PROTO_INUSE(rawv6)
1209 };
1210
1211 #ifdef CONFIG_PROC_FS
1212 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1213 {
1214 struct ipv6_pinfo *np = inet6_sk(sp);
1215 struct in6_addr *dest, *src;
1216 __u16 destp, srcp;
1217
1218 dest = &np->daddr;
1219 src = &np->rcv_saddr;
1220 destp = 0;
1221 srcp = inet_sk(sp)->num;
1222 seq_printf(seq,
1223 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1224 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1225 i,
1226 src->s6_addr32[0], src->s6_addr32[1],
1227 src->s6_addr32[2], src->s6_addr32[3], srcp,
1228 dest->s6_addr32[0], dest->s6_addr32[1],
1229 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1230 sp->sk_state,
1231 atomic_read(&sp->sk_wmem_alloc),
1232 atomic_read(&sp->sk_rmem_alloc),
1233 0, 0L, 0,
1234 sock_i_uid(sp), 0,
1235 sock_i_ino(sp),
1236 atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1237 }
1238
1239 static int raw6_seq_show(struct seq_file *seq, void *v)
1240 {
1241 if (v == SEQ_START_TOKEN)
1242 seq_printf(seq,
1243 " sl "
1244 "local_address "
1245 "remote_address "
1246 "st tx_queue rx_queue tr tm->when retrnsmt"
1247 " uid timeout inode drops\n");
1248 else
1249 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1250 return 0;
1251 }
1252
1253 static const struct seq_operations raw6_seq_ops = {
1254 .start = raw_seq_start,
1255 .next = raw_seq_next,
1256 .stop = raw_seq_stop,
1257 .show = raw6_seq_show,
1258 };
1259
1260 static int raw6_seq_open(struct inode *inode, struct file *file)
1261 {
1262 return raw_seq_open(inode, file, &raw_v6_hashinfo, PF_INET6);
1263 }
1264
1265 static const struct file_operations raw6_seq_fops = {
1266 .owner = THIS_MODULE,
1267 .open = raw6_seq_open,
1268 .read = seq_read,
1269 .llseek = seq_lseek,
1270 .release = seq_release_net,
1271 };
1272
1273 static int raw6_init_net(struct net *net)
1274 {
1275 if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1276 return -ENOMEM;
1277
1278 return 0;
1279 }
1280
1281 static void raw6_exit_net(struct net *net)
1282 {
1283 proc_net_remove(net, "raw6");
1284 }
1285
1286 static struct pernet_operations raw6_net_ops = {
1287 .init = raw6_init_net,
1288 .exit = raw6_exit_net,
1289 };
1290
1291 int __init raw6_proc_init(void)
1292 {
1293 return register_pernet_subsys(&raw6_net_ops);
1294 }
1295
1296 void raw6_proc_exit(void)
1297 {
1298 unregister_pernet_subsys(&raw6_net_ops);
1299 }
1300 #endif /* CONFIG_PROC_FS */
1301
1302 /* Same as inet6_dgram_ops, sans udp_poll. */
1303 static const struct proto_ops inet6_sockraw_ops = {
1304 .family = PF_INET6,
1305 .owner = THIS_MODULE,
1306 .release = inet6_release,
1307 .bind = inet6_bind,
1308 .connect = inet_dgram_connect, /* ok */
1309 .socketpair = sock_no_socketpair, /* a do nothing */
1310 .accept = sock_no_accept, /* a do nothing */
1311 .getname = inet6_getname,
1312 .poll = datagram_poll, /* ok */
1313 .ioctl = inet6_ioctl, /* must change */
1314 .listen = sock_no_listen, /* ok */
1315 .shutdown = inet_shutdown, /* ok */
1316 .setsockopt = sock_common_setsockopt, /* ok */
1317 .getsockopt = sock_common_getsockopt, /* ok */
1318 .sendmsg = inet_sendmsg, /* ok */
1319 .recvmsg = sock_common_recvmsg, /* ok */
1320 .mmap = sock_no_mmap,
1321 .sendpage = sock_no_sendpage,
1322 #ifdef CONFIG_COMPAT
1323 .compat_setsockopt = compat_sock_common_setsockopt,
1324 .compat_getsockopt = compat_sock_common_getsockopt,
1325 #endif
1326 };
1327
1328 static struct inet_protosw rawv6_protosw = {
1329 .type = SOCK_RAW,
1330 .protocol = IPPROTO_IP, /* wild card */
1331 .prot = &rawv6_prot,
1332 .ops = &inet6_sockraw_ops,
1333 .capability = CAP_NET_RAW,
1334 .no_check = UDP_CSUM_DEFAULT,
1335 .flags = INET_PROTOSW_REUSE,
1336 };
1337
1338 int __init rawv6_init(void)
1339 {
1340 int ret;
1341
1342 ret = inet6_register_protosw(&rawv6_protosw);
1343 if (ret)
1344 goto out;
1345 out:
1346 return ret;
1347 }
1348
1349 void rawv6_exit(void)
1350 {
1351 inet6_unregister_protosw(&rawv6_protosw);
1352 }
kernel-based virtual machine