inet: constify ip headers and in6_addr
[linux-2.6.git] / net / ipv6 / raw.c
1 /*
2  *      RAW sockets for IPv6
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *
8  *      Adapted from linux/net/ipv4/raw.c
9  *
10  *      Fixes:
11  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
12  *      YOSHIFUJI,H.@USAGI      :       raw checksum (RFC2292(bis) compliance)
13  *      Kazunori MIYAZAWA @USAGI:       change process style to use ip6_append_data
14  *
15  *      This program is free software; you can redistribute it and/or
16  *      modify it under the terms of the GNU General Public License
17  *      as published by the Free Software Foundation; either version
18  *      2 of the License, or (at your option) any later version.
19  */
20
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/slab.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/netfilter.h>
32 #include <linux/netfilter_ipv6.h>
33 #include <linux/skbuff.h>
34 #include <linux/compat.h>
35 #include <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, const struct in6_addr *loc_addr,
71                 const 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)->inet_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 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
128
129 static mh_filter_t __rcu *mh_filter __read_mostly;
130
131 int rawv6_mh_filter_register(mh_filter_t filter)
132 {
133         rcu_assign_pointer(mh_filter, filter);
134         return 0;
135 }
136 EXPORT_SYMBOL(rawv6_mh_filter_register);
137
138 int rawv6_mh_filter_unregister(mh_filter_t filter)
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         const struct in6_addr *saddr;
158         const 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                         mh_filter_t *filter;
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         rcu_read_lock();
263         /* Check if the address belongs to the host. */
264         if (addr_type != IPV6_ADDR_ANY) {
265                 struct net_device *dev = NULL;
266
267                 if (addr_type & IPV6_ADDR_LINKLOCAL) {
268                         if (addr_len >= sizeof(struct sockaddr_in6) &&
269                             addr->sin6_scope_id) {
270                                 /* Override any existing binding, if another
271                                  * one is supplied by user.
272                                  */
273                                 sk->sk_bound_dev_if = addr->sin6_scope_id;
274                         }
275
276                         /* Binding to link-local address requires an interface */
277                         if (!sk->sk_bound_dev_if)
278                                 goto out_unlock;
279
280                         err = -ENODEV;
281                         dev = dev_get_by_index_rcu(sock_net(sk),
282                                                    sk->sk_bound_dev_if);
283                         if (!dev)
284                                 goto out_unlock;
285                 }
286
287                 /* ipv4 addr of the socket is invalid.  Only the
288                  * unspecified and mapped address have a v4 equivalent.
289                  */
290                 v4addr = LOOPBACK4_IPV6;
291                 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
292                         err = -EADDRNOTAVAIL;
293                         if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
294                                            dev, 0)) {
295                                 goto out_unlock;
296                         }
297                 }
298         }
299
300         inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
301         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
302         if (!(addr_type & IPV6_ADDR_MULTICAST))
303                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
304         err = 0;
305 out_unlock:
306         rcu_read_unlock();
307 out:
308         release_sock(sk);
309         return err;
310 }
311
312 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
313                struct inet6_skb_parm *opt,
314                u8 type, u8 code, int offset, __be32 info)
315 {
316         struct inet_sock *inet = inet_sk(sk);
317         struct ipv6_pinfo *np = inet6_sk(sk);
318         int err;
319         int harderr;
320
321         /* Report error on raw socket, if:
322            1. User requested recverr.
323            2. Socket is connected (otherwise the error indication
324               is useless without recverr and error is hard.
325          */
326         if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
327                 return;
328
329         harderr = icmpv6_err_convert(type, code, &err);
330         if (type == ICMPV6_PKT_TOOBIG)
331                 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
332
333         if (np->recverr) {
334                 u8 *payload = skb->data;
335                 if (!inet->hdrincl)
336                         payload += offset;
337                 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
338         }
339
340         if (np->recverr || harderr) {
341                 sk->sk_err = err;
342                 sk->sk_error_report(sk);
343         }
344 }
345
346 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
347                 u8 type, u8 code, int inner_offset, __be32 info)
348 {
349         struct sock *sk;
350         int hash;
351         const struct in6_addr *saddr, *daddr;
352         struct net *net;
353
354         hash = nexthdr & (RAW_HTABLE_SIZE - 1);
355
356         read_lock(&raw_v6_hashinfo.lock);
357         sk = sk_head(&raw_v6_hashinfo.ht[hash]);
358         if (sk != NULL) {
359                 /* Note: ipv6_hdr(skb) != skb->data */
360                 const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
361                 saddr = &ip6h->saddr;
362                 daddr = &ip6h->daddr;
363                 net = dev_net(skb->dev);
364
365                 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
366                                                 IP6CB(skb)->iif))) {
367                         rawv6_err(sk, skb, NULL, type, code,
368                                         inner_offset, info);
369                         sk = sk_next(sk);
370                 }
371         }
372         read_unlock(&raw_v6_hashinfo.lock);
373 }
374
375 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
376 {
377         if ((raw6_sk(sk)->checksum || rcu_dereference_raw(sk->sk_filter)) &&
378             skb_checksum_complete(skb)) {
379                 atomic_inc(&sk->sk_drops);
380                 kfree_skb(skb);
381                 return NET_RX_DROP;
382         }
383
384         /* Charge it to the socket. */
385         if (ip_queue_rcv_skb(sk, skb) < 0) {
386                 kfree_skb(skb);
387                 return NET_RX_DROP;
388         }
389
390         return 0;
391 }
392
393 /*
394  *      This is next to useless...
395  *      if we demultiplex in network layer we don't need the extra call
396  *      just to queue the skb...
397  *      maybe we could have the network decide upon a hint if it
398  *      should call raw_rcv for demultiplexing
399  */
400 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
401 {
402         struct inet_sock *inet = inet_sk(sk);
403         struct raw6_sock *rp = raw6_sk(sk);
404
405         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
406                 atomic_inc(&sk->sk_drops);
407                 kfree_skb(skb);
408                 return NET_RX_DROP;
409         }
410
411         if (!rp->checksum)
412                 skb->ip_summed = CHECKSUM_UNNECESSARY;
413
414         if (skb->ip_summed == CHECKSUM_COMPLETE) {
415                 skb_postpull_rcsum(skb, skb_network_header(skb),
416                                    skb_network_header_len(skb));
417                 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
418                                      &ipv6_hdr(skb)->daddr,
419                                      skb->len, inet->inet_num, skb->csum))
420                         skb->ip_summed = CHECKSUM_UNNECESSARY;
421         }
422         if (!skb_csum_unnecessary(skb))
423                 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
424                                                          &ipv6_hdr(skb)->daddr,
425                                                          skb->len,
426                                                          inet->inet_num, 0));
427
428         if (inet->hdrincl) {
429                 if (skb_checksum_complete(skb)) {
430                         atomic_inc(&sk->sk_drops);
431                         kfree_skb(skb);
432                         return NET_RX_DROP;
433                 }
434         }
435
436         rawv6_rcv_skb(sk, skb);
437         return 0;
438 }
439
440
441 /*
442  *      This should be easy, if there is something there
443  *      we return it, otherwise we block.
444  */
445
446 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
447                   struct msghdr *msg, size_t len,
448                   int noblock, int flags, int *addr_len)
449 {
450         struct ipv6_pinfo *np = inet6_sk(sk);
451         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
452         struct sk_buff *skb;
453         size_t copied;
454         int err;
455
456         if (flags & MSG_OOB)
457                 return -EOPNOTSUPP;
458
459         if (addr_len)
460                 *addr_len=sizeof(*sin6);
461
462         if (flags & MSG_ERRQUEUE)
463                 return ipv6_recv_error(sk, msg, len);
464
465         if (np->rxpmtu && np->rxopt.bits.rxpmtu)
466                 return ipv6_recv_rxpmtu(sk, msg, len);
467
468         skb = skb_recv_datagram(sk, flags, noblock, &err);
469         if (!skb)
470                 goto out;
471
472         copied = skb->len;
473         if (copied > len) {
474                 copied = len;
475                 msg->msg_flags |= MSG_TRUNC;
476         }
477
478         if (skb_csum_unnecessary(skb)) {
479                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
480         } else if (msg->msg_flags&MSG_TRUNC) {
481                 if (__skb_checksum_complete(skb))
482                         goto csum_copy_err;
483                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
484         } else {
485                 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
486                 if (err == -EINVAL)
487                         goto csum_copy_err;
488         }
489         if (err)
490                 goto out_free;
491
492         /* Copy the address. */
493         if (sin6) {
494                 sin6->sin6_family = AF_INET6;
495                 sin6->sin6_port = 0;
496                 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
497                 sin6->sin6_flowinfo = 0;
498                 sin6->sin6_scope_id = 0;
499                 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
500                         sin6->sin6_scope_id = IP6CB(skb)->iif;
501         }
502
503         sock_recv_ts_and_drops(msg, sk, skb);
504
505         if (np->rxopt.all)
506                 datagram_recv_ctl(sk, msg, skb);
507
508         err = copied;
509         if (flags & MSG_TRUNC)
510                 err = skb->len;
511
512 out_free:
513         skb_free_datagram(sk, skb);
514 out:
515         return err;
516
517 csum_copy_err:
518         skb_kill_datagram(sk, skb, flags);
519
520         /* Error for blocking case is chosen to masquerade
521            as some normal condition.
522          */
523         err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
524         goto out;
525 }
526
527 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
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(&fl6->saddr, &fl6->daddr,
590                                total_len, fl6->flowi6_proto, tmp_csum);
591
592         if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
593                 csum = CSUM_MANGLED_0;
594
595         if (skb_store_bits(skb, offset, &csum, 2))
596                 BUG();
597
598 send:
599         err = ip6_push_pending_frames(sk);
600 out:
601         return err;
602 }
603
604 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
605                         struct flowi6 *fl6, struct dst_entry **dstp,
606                         unsigned int flags)
607 {
608         struct ipv6_pinfo *np = inet6_sk(sk);
609         struct ipv6hdr *iph;
610         struct sk_buff *skb;
611         int err;
612         struct rt6_info *rt = (struct rt6_info *)*dstp;
613
614         if (length > rt->dst.dev->mtu) {
615                 ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
616                 return -EMSGSIZE;
617         }
618         if (flags&MSG_PROBE)
619                 goto out;
620
621         skb = sock_alloc_send_skb(sk,
622                                   length + LL_ALLOCATED_SPACE(rt->dst.dev) + 15,
623                                   flags & MSG_DONTWAIT, &err);
624         if (skb == NULL)
625                 goto error;
626         skb_reserve(skb, LL_RESERVED_SPACE(rt->dst.dev));
627
628         skb->priority = sk->sk_priority;
629         skb->mark = sk->sk_mark;
630         skb_dst_set(skb, &rt->dst);
631         *dstp = NULL;
632
633         skb_put(skb, length);
634         skb_reset_network_header(skb);
635         iph = ipv6_hdr(skb);
636
637         skb->ip_summed = CHECKSUM_NONE;
638
639         skb->transport_header = skb->network_header;
640         err = memcpy_fromiovecend((void *)iph, from, 0, length);
641         if (err)
642                 goto error_fault;
643
644         IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
645         err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, skb, NULL,
646                       rt->dst.dev, dst_output);
647         if (err > 0)
648                 err = net_xmit_errno(err);
649         if (err)
650                 goto error;
651 out:
652         return 0;
653
654 error_fault:
655         err = -EFAULT;
656         kfree_skb(skb);
657 error:
658         IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
659         if (err == -ENOBUFS && !np->recverr)
660                 err = 0;
661         return err;
662 }
663
664 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg)
665 {
666         struct iovec *iov;
667         u8 __user *type = NULL;
668         u8 __user *code = NULL;
669         u8 len = 0;
670         int probed = 0;
671         int i;
672
673         if (!msg->msg_iov)
674                 return 0;
675
676         for (i = 0; i < msg->msg_iovlen; i++) {
677                 iov = &msg->msg_iov[i];
678                 if (!iov)
679                         continue;
680
681                 switch (fl6->flowi6_proto) {
682                 case IPPROTO_ICMPV6:
683                         /* check if one-byte field is readable or not. */
684                         if (iov->iov_base && iov->iov_len < 1)
685                                 break;
686
687                         if (!type) {
688                                 type = iov->iov_base;
689                                 /* check if code field is readable or not. */
690                                 if (iov->iov_len > 1)
691                                         code = type + 1;
692                         } else if (!code)
693                                 code = iov->iov_base;
694
695                         if (type && code) {
696                                 if (get_user(fl6->fl6_icmp_type, type) ||
697                                     get_user(fl6->fl6_icmp_code, code))
698                                         return -EFAULT;
699                                 probed = 1;
700                         }
701                         break;
702                 case IPPROTO_MH:
703                         if (iov->iov_base && iov->iov_len < 1)
704                                 break;
705                         /* check if type field is readable or not. */
706                         if (iov->iov_len > 2 - len) {
707                                 u8 __user *p = iov->iov_base;
708                                 if (get_user(fl6->fl6_mh_type, &p[2 - len]))
709                                         return -EFAULT;
710                                 probed = 1;
711                         } else
712                                 len += iov->iov_len;
713
714                         break;
715                 default:
716                         probed = 1;
717                         break;
718                 }
719                 if (probed)
720                         break;
721         }
722         return 0;
723 }
724
725 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
726                    struct msghdr *msg, size_t len)
727 {
728         struct ipv6_txoptions opt_space;
729         struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
730         struct in6_addr *daddr, *final_p, final;
731         struct inet_sock *inet = inet_sk(sk);
732         struct ipv6_pinfo *np = inet6_sk(sk);
733         struct raw6_sock *rp = raw6_sk(sk);
734         struct ipv6_txoptions *opt = NULL;
735         struct ip6_flowlabel *flowlabel = NULL;
736         struct dst_entry *dst = NULL;
737         struct flowi6 fl6;
738         int addr_len = msg->msg_namelen;
739         int hlimit = -1;
740         int tclass = -1;
741         int dontfrag = -1;
742         u16 proto;
743         int err;
744
745         /* Rough check on arithmetic overflow,
746            better check is made in ip6_append_data().
747          */
748         if (len > INT_MAX)
749                 return -EMSGSIZE;
750
751         /* Mirror BSD error message compatibility */
752         if (msg->msg_flags & MSG_OOB)
753                 return -EOPNOTSUPP;
754
755         /*
756          *      Get and verify the address.
757          */
758         memset(&fl6, 0, sizeof(fl6));
759
760         fl6.flowi6_mark = sk->sk_mark;
761
762         if (sin6) {
763                 if (addr_len < SIN6_LEN_RFC2133)
764                         return -EINVAL;
765
766                 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
767                         return -EAFNOSUPPORT;
768
769                 /* port is the proto value [0..255] carried in nexthdr */
770                 proto = ntohs(sin6->sin6_port);
771
772                 if (!proto)
773                         proto = inet->inet_num;
774                 else if (proto != inet->inet_num)
775                         return -EINVAL;
776
777                 if (proto > 255)
778                         return -EINVAL;
779
780                 daddr = &sin6->sin6_addr;
781                 if (np->sndflow) {
782                         fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
783                         if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
784                                 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
785                                 if (flowlabel == NULL)
786                                         return -EINVAL;
787                                 daddr = &flowlabel->dst;
788                         }
789                 }
790
791                 /*
792                  * Otherwise it will be difficult to maintain
793                  * sk->sk_dst_cache.
794                  */
795                 if (sk->sk_state == TCP_ESTABLISHED &&
796                     ipv6_addr_equal(daddr, &np->daddr))
797                         daddr = &np->daddr;
798
799                 if (addr_len >= sizeof(struct sockaddr_in6) &&
800                     sin6->sin6_scope_id &&
801                     ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
802                         fl6.flowi6_oif = sin6->sin6_scope_id;
803         } else {
804                 if (sk->sk_state != TCP_ESTABLISHED)
805                         return -EDESTADDRREQ;
806
807                 proto = inet->inet_num;
808                 daddr = &np->daddr;
809                 fl6.flowlabel = np->flow_label;
810         }
811
812         if (fl6.flowi6_oif == 0)
813                 fl6.flowi6_oif = sk->sk_bound_dev_if;
814
815         if (msg->msg_controllen) {
816                 opt = &opt_space;
817                 memset(opt, 0, sizeof(struct ipv6_txoptions));
818                 opt->tot_len = sizeof(struct ipv6_txoptions);
819
820                 err = datagram_send_ctl(sock_net(sk), msg, &fl6, opt, &hlimit,
821                                         &tclass, &dontfrag);
822                 if (err < 0) {
823                         fl6_sock_release(flowlabel);
824                         return err;
825                 }
826                 if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
827                         flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
828                         if (flowlabel == NULL)
829                                 return -EINVAL;
830                 }
831                 if (!(opt->opt_nflen|opt->opt_flen))
832                         opt = NULL;
833         }
834         if (opt == NULL)
835                 opt = np->opt;
836         if (flowlabel)
837                 opt = fl6_merge_options(&opt_space, flowlabel, opt);
838         opt = ipv6_fixup_options(&opt_space, opt);
839
840         fl6.flowi6_proto = proto;
841         err = rawv6_probe_proto_opt(&fl6, msg);
842         if (err)
843                 goto out;
844
845         if (!ipv6_addr_any(daddr))
846                 ipv6_addr_copy(&fl6.daddr, daddr);
847         else
848                 fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
849         if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
850                 ipv6_addr_copy(&fl6.saddr, &np->saddr);
851
852         final_p = fl6_update_dst(&fl6, opt, &final);
853
854         if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
855                 fl6.flowi6_oif = np->mcast_oif;
856         security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
857
858         dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
859         if (IS_ERR(dst)) {
860                 err = PTR_ERR(dst);
861                 goto out;
862         }
863         if (hlimit < 0) {
864                 if (ipv6_addr_is_multicast(&fl6.daddr))
865                         hlimit = np->mcast_hops;
866                 else
867                         hlimit = np->hop_limit;
868                 if (hlimit < 0)
869                         hlimit = ip6_dst_hoplimit(dst);
870         }
871
872         if (tclass < 0)
873                 tclass = np->tclass;
874
875         if (dontfrag < 0)
876                 dontfrag = np->dontfrag;
877
878         if (msg->msg_flags&MSG_CONFIRM)
879                 goto do_confirm;
880
881 back_from_confirm:
882         if (inet->hdrincl)
883                 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl6, &dst, msg->msg_flags);
884         else {
885                 lock_sock(sk);
886                 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
887                         len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
888                         msg->msg_flags, dontfrag);
889
890                 if (err)
891                         ip6_flush_pending_frames(sk);
892                 else if (!(msg->msg_flags & MSG_MORE))
893                         err = rawv6_push_pending_frames(sk, &fl6, rp);
894                 release_sock(sk);
895         }
896 done:
897         dst_release(dst);
898 out:
899         fl6_sock_release(flowlabel);
900         return err<0?err:len;
901 do_confirm:
902         dst_confirm(dst);
903         if (!(msg->msg_flags & MSG_PROBE) || len)
904                 goto back_from_confirm;
905         err = 0;
906         goto done;
907 }
908
909 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
910                                char __user *optval, int optlen)
911 {
912         switch (optname) {
913         case ICMPV6_FILTER:
914                 if (optlen > sizeof(struct icmp6_filter))
915                         optlen = sizeof(struct icmp6_filter);
916                 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
917                         return -EFAULT;
918                 return 0;
919         default:
920                 return -ENOPROTOOPT;
921         }
922
923         return 0;
924 }
925
926 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
927                                char __user *optval, int __user *optlen)
928 {
929         int len;
930
931         switch (optname) {
932         case ICMPV6_FILTER:
933                 if (get_user(len, optlen))
934                         return -EFAULT;
935                 if (len < 0)
936                         return -EINVAL;
937                 if (len > sizeof(struct icmp6_filter))
938                         len = sizeof(struct icmp6_filter);
939                 if (put_user(len, optlen))
940                         return -EFAULT;
941                 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
942                         return -EFAULT;
943                 return 0;
944         default:
945                 return -ENOPROTOOPT;
946         }
947
948         return 0;
949 }
950
951
952 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
953                             char __user *optval, unsigned int optlen)
954 {
955         struct raw6_sock *rp = raw6_sk(sk);
956         int val;
957
958         if (get_user(val, (int __user *)optval))
959                 return -EFAULT;
960
961         switch (optname) {
962                 case IPV6_CHECKSUM:
963                         if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
964                             level == IPPROTO_IPV6) {
965                                 /*
966                                  * RFC3542 tells that IPV6_CHECKSUM socket
967                                  * option in the IPPROTO_IPV6 level is not
968                                  * allowed on ICMPv6 sockets.
969                                  * If you want to set it, use IPPROTO_RAW
970                                  * level IPV6_CHECKSUM socket option
971                                  * (Linux extension).
972                                  */
973                                 return -EINVAL;
974                         }
975
976                         /* You may get strange result with a positive odd offset;
977                            RFC2292bis agrees with me. */
978                         if (val > 0 && (val&1))
979                                 return -EINVAL;
980                         if (val < 0) {
981                                 rp->checksum = 0;
982                         } else {
983                                 rp->checksum = 1;
984                                 rp->offset = val;
985                         }
986
987                         return 0;
988                         break;
989
990                 default:
991                         return -ENOPROTOOPT;
992         }
993 }
994
995 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
996                           char __user *optval, unsigned int optlen)
997 {
998         switch(level) {
999                 case SOL_RAW:
1000                         break;
1001
1002                 case SOL_ICMPV6:
1003                         if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1004                                 return -EOPNOTSUPP;
1005                         return rawv6_seticmpfilter(sk, level, optname, optval,
1006                                                    optlen);
1007                 case SOL_IPV6:
1008                         if (optname == IPV6_CHECKSUM)
1009                                 break;
1010                 default:
1011                         return ipv6_setsockopt(sk, level, optname, optval,
1012                                                optlen);
1013         }
1014
1015         return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1016 }
1017
1018 #ifdef CONFIG_COMPAT
1019 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1020                                    char __user *optval, unsigned int optlen)
1021 {
1022         switch (level) {
1023         case SOL_RAW:
1024                 break;
1025         case SOL_ICMPV6:
1026                 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1027                         return -EOPNOTSUPP;
1028                 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1029         case SOL_IPV6:
1030                 if (optname == IPV6_CHECKSUM)
1031                         break;
1032         default:
1033                 return compat_ipv6_setsockopt(sk, level, optname,
1034                                               optval, optlen);
1035         }
1036         return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1037 }
1038 #endif
1039
1040 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1041                             char __user *optval, int __user *optlen)
1042 {
1043         struct raw6_sock *rp = raw6_sk(sk);
1044         int val, len;
1045
1046         if (get_user(len,optlen))
1047                 return -EFAULT;
1048
1049         switch (optname) {
1050         case IPV6_CHECKSUM:
1051                 /*
1052                  * We allow getsockopt() for IPPROTO_IPV6-level
1053                  * IPV6_CHECKSUM socket option on ICMPv6 sockets
1054                  * since RFC3542 is silent about it.
1055                  */
1056                 if (rp->checksum == 0)
1057                         val = -1;
1058                 else
1059                         val = rp->offset;
1060                 break;
1061
1062         default:
1063                 return -ENOPROTOOPT;
1064         }
1065
1066         len = min_t(unsigned int, sizeof(int), len);
1067
1068         if (put_user(len, optlen))
1069                 return -EFAULT;
1070         if (copy_to_user(optval,&val,len))
1071                 return -EFAULT;
1072         return 0;
1073 }
1074
1075 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1076                           char __user *optval, int __user *optlen)
1077 {
1078         switch(level) {
1079                 case SOL_RAW:
1080                         break;
1081
1082                 case SOL_ICMPV6:
1083                         if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1084                                 return -EOPNOTSUPP;
1085                         return rawv6_geticmpfilter(sk, level, optname, optval,
1086                                                    optlen);
1087                 case SOL_IPV6:
1088                         if (optname == IPV6_CHECKSUM)
1089                                 break;
1090                 default:
1091                         return ipv6_getsockopt(sk, level, optname, optval,
1092                                                optlen);
1093         }
1094
1095         return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1096 }
1097
1098 #ifdef CONFIG_COMPAT
1099 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1100                                    char __user *optval, int __user *optlen)
1101 {
1102         switch (level) {
1103         case SOL_RAW:
1104                 break;
1105         case SOL_ICMPV6:
1106                 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1107                         return -EOPNOTSUPP;
1108                 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1109         case SOL_IPV6:
1110                 if (optname == IPV6_CHECKSUM)
1111                         break;
1112         default:
1113                 return compat_ipv6_getsockopt(sk, level, optname,
1114                                               optval, optlen);
1115         }
1116         return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1117 }
1118 #endif
1119
1120 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1121 {
1122         switch(cmd) {
1123                 case SIOCOUTQ:
1124                 {
1125                         int amount = sk_wmem_alloc_get(sk);
1126
1127                         return put_user(amount, (int __user *)arg);
1128                 }
1129                 case SIOCINQ:
1130                 {
1131                         struct sk_buff *skb;
1132                         int amount = 0;
1133
1134                         spin_lock_bh(&sk->sk_receive_queue.lock);
1135                         skb = skb_peek(&sk->sk_receive_queue);
1136                         if (skb != NULL)
1137                                 amount = skb->tail - skb->transport_header;
1138                         spin_unlock_bh(&sk->sk_receive_queue.lock);
1139                         return put_user(amount, (int __user *)arg);
1140                 }
1141
1142                 default:
1143 #ifdef CONFIG_IPV6_MROUTE
1144                         return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1145 #else
1146                         return -ENOIOCTLCMD;
1147 #endif
1148         }
1149 }
1150
1151 #ifdef CONFIG_COMPAT
1152 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1153 {
1154         switch (cmd) {
1155         case SIOCOUTQ:
1156         case SIOCINQ:
1157                 return -ENOIOCTLCMD;
1158         default:
1159 #ifdef CONFIG_IPV6_MROUTE
1160                 return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1161 #else
1162                 return -ENOIOCTLCMD;
1163 #endif
1164         }
1165 }
1166 #endif
1167
1168 static void rawv6_close(struct sock *sk, long timeout)
1169 {
1170         if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1171                 ip6_ra_control(sk, -1);
1172         ip6mr_sk_done(sk);
1173         sk_common_release(sk);
1174 }
1175
1176 static void raw6_destroy(struct sock *sk)
1177 {
1178         lock_sock(sk);
1179         ip6_flush_pending_frames(sk);
1180         release_sock(sk);
1181
1182         inet6_destroy_sock(sk);
1183 }
1184
1185 static int rawv6_init_sk(struct sock *sk)
1186 {
1187         struct raw6_sock *rp = raw6_sk(sk);
1188
1189         switch (inet_sk(sk)->inet_num) {
1190         case IPPROTO_ICMPV6:
1191                 rp->checksum = 1;
1192                 rp->offset   = 2;
1193                 break;
1194         case IPPROTO_MH:
1195                 rp->checksum = 1;
1196                 rp->offset   = 4;
1197                 break;
1198         default:
1199                 break;
1200         }
1201         return 0;
1202 }
1203
1204 struct proto rawv6_prot = {
1205         .name              = "RAWv6",
1206         .owner             = THIS_MODULE,
1207         .close             = rawv6_close,
1208         .destroy           = raw6_destroy,
1209         .connect           = ip6_datagram_connect,
1210         .disconnect        = udp_disconnect,
1211         .ioctl             = rawv6_ioctl,
1212         .init              = rawv6_init_sk,
1213         .setsockopt        = rawv6_setsockopt,
1214         .getsockopt        = rawv6_getsockopt,
1215         .sendmsg           = rawv6_sendmsg,
1216         .recvmsg           = rawv6_recvmsg,
1217         .bind              = rawv6_bind,
1218         .backlog_rcv       = rawv6_rcv_skb,
1219         .hash              = raw_hash_sk,
1220         .unhash            = raw_unhash_sk,
1221         .obj_size          = sizeof(struct raw6_sock),
1222         .h.raw_hash        = &raw_v6_hashinfo,
1223 #ifdef CONFIG_COMPAT
1224         .compat_setsockopt = compat_rawv6_setsockopt,
1225         .compat_getsockopt = compat_rawv6_getsockopt,
1226         .compat_ioctl      = compat_rawv6_ioctl,
1227 #endif
1228 };
1229
1230 #ifdef CONFIG_PROC_FS
1231 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1232 {
1233         struct ipv6_pinfo *np = inet6_sk(sp);
1234         const struct in6_addr *dest, *src;
1235         __u16 destp, srcp;
1236
1237         dest  = &np->daddr;
1238         src   = &np->rcv_saddr;
1239         destp = 0;
1240         srcp  = inet_sk(sp)->inet_num;
1241         seq_printf(seq,
1242                    "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1243                    "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1244                    i,
1245                    src->s6_addr32[0], src->s6_addr32[1],
1246                    src->s6_addr32[2], src->s6_addr32[3], srcp,
1247                    dest->s6_addr32[0], dest->s6_addr32[1],
1248                    dest->s6_addr32[2], dest->s6_addr32[3], destp,
1249                    sp->sk_state,
1250                    sk_wmem_alloc_get(sp),
1251                    sk_rmem_alloc_get(sp),
1252                    0, 0L, 0,
1253                    sock_i_uid(sp), 0,
1254                    sock_i_ino(sp),
1255                    atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1256 }
1257
1258 static int raw6_seq_show(struct seq_file *seq, void *v)
1259 {
1260         if (v == SEQ_START_TOKEN)
1261                 seq_printf(seq,
1262                            "  sl  "
1263                            "local_address                         "
1264                            "remote_address                        "
1265                            "st tx_queue rx_queue tr tm->when retrnsmt"
1266                            "   uid  timeout inode ref pointer drops\n");
1267         else
1268                 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1269         return 0;
1270 }
1271
1272 static const struct seq_operations raw6_seq_ops = {
1273         .start =        raw_seq_start,
1274         .next =         raw_seq_next,
1275         .stop =         raw_seq_stop,
1276         .show =         raw6_seq_show,
1277 };
1278
1279 static int raw6_seq_open(struct inode *inode, struct file *file)
1280 {
1281         return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1282 }
1283
1284 static const struct file_operations raw6_seq_fops = {
1285         .owner =        THIS_MODULE,
1286         .open =         raw6_seq_open,
1287         .read =         seq_read,
1288         .llseek =       seq_lseek,
1289         .release =      seq_release_net,
1290 };
1291
1292 static int __net_init raw6_init_net(struct net *net)
1293 {
1294         if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1295                 return -ENOMEM;
1296
1297         return 0;
1298 }
1299
1300 static void __net_exit raw6_exit_net(struct net *net)
1301 {
1302         proc_net_remove(net, "raw6");
1303 }
1304
1305 static struct pernet_operations raw6_net_ops = {
1306         .init = raw6_init_net,
1307         .exit = raw6_exit_net,
1308 };
1309
1310 int __init raw6_proc_init(void)
1311 {
1312         return register_pernet_subsys(&raw6_net_ops);
1313 }
1314
1315 void raw6_proc_exit(void)
1316 {
1317         unregister_pernet_subsys(&raw6_net_ops);
1318 }
1319 #endif  /* CONFIG_PROC_FS */
1320
1321 /* Same as inet6_dgram_ops, sans udp_poll.  */
1322 static const struct proto_ops inet6_sockraw_ops = {
1323         .family            = PF_INET6,
1324         .owner             = THIS_MODULE,
1325         .release           = inet6_release,
1326         .bind              = inet6_bind,
1327         .connect           = inet_dgram_connect,        /* ok           */
1328         .socketpair        = sock_no_socketpair,        /* a do nothing */
1329         .accept            = sock_no_accept,            /* a do nothing */
1330         .getname           = inet6_getname,
1331         .poll              = datagram_poll,             /* ok           */
1332         .ioctl             = inet6_ioctl,               /* must change  */
1333         .listen            = sock_no_listen,            /* ok           */
1334         .shutdown          = inet_shutdown,             /* ok           */
1335         .setsockopt        = sock_common_setsockopt,    /* ok           */
1336         .getsockopt        = sock_common_getsockopt,    /* ok           */
1337         .sendmsg           = inet_sendmsg,              /* ok           */
1338         .recvmsg           = sock_common_recvmsg,       /* ok           */
1339         .mmap              = sock_no_mmap,
1340         .sendpage          = sock_no_sendpage,
1341 #ifdef CONFIG_COMPAT
1342         .compat_setsockopt = compat_sock_common_setsockopt,
1343         .compat_getsockopt = compat_sock_common_getsockopt,
1344 #endif
1345 };
1346
1347 static struct inet_protosw rawv6_protosw = {
1348         .type           = SOCK_RAW,
1349         .protocol       = IPPROTO_IP,   /* wild card */
1350         .prot           = &rawv6_prot,
1351         .ops            = &inet6_sockraw_ops,
1352         .no_check       = UDP_CSUM_DEFAULT,
1353         .flags          = INET_PROTOSW_REUSE,
1354 };
1355
1356 int __init rawv6_init(void)
1357 {
1358         int ret;
1359
1360         ret = inet6_register_protosw(&rawv6_protosw);
1361         if (ret)
1362                 goto out;
1363 out:
1364         return ret;
1365 }
1366
1367 void rawv6_exit(void)
1368 {
1369         inet6_unregister_protosw(&rawv6_protosw);
1370 }