net: spread __net_init, __net_exit
[linux-3.10.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/sockios.h>
25 #include <linux/net.h>
26 #include <linux/in6.h>
27 #include <linux/netdevice.h>
28 #include <linux/if_arp.h>
29 #include <linux/icmpv6.h>
30 #include <linux/netfilter.h>
31 #include <linux/netfilter_ipv6.h>
32 #include <linux/skbuff.h>
33 #include <asm/uaccess.h>
34 #include <asm/ioctls.h>
35
36 #include <net/net_namespace.h>
37 #include <net/ip.h>
38 #include <net/sock.h>
39 #include <net/snmp.h>
40
41 #include <net/ipv6.h>
42 #include <net/ndisc.h>
43 #include <net/protocol.h>
44 #include <net/ip6_route.h>
45 #include <net/ip6_checksum.h>
46 #include <net/addrconf.h>
47 #include <net/transp_v6.h>
48 #include <net/udp.h>
49 #include <net/inet_common.h>
50 #include <net/tcp_states.h>
51 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
52 #include <net/mip6.h>
53 #endif
54 #include <linux/mroute6.h>
55
56 #include <net/raw.h>
57 #include <net/rawv6.h>
58 #include <net/xfrm.h>
59
60 #include <linux/proc_fs.h>
61 #include <linux/seq_file.h>
62
63 static struct raw_hashinfo raw_v6_hashinfo = {
64         .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
65 };
66
67 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
68                 unsigned short num, struct in6_addr *loc_addr,
69                 struct in6_addr *rmt_addr, int dif)
70 {
71         struct hlist_node *node;
72         int is_multicast = ipv6_addr_is_multicast(loc_addr);
73
74         sk_for_each_from(sk, node)
75                 if (inet_sk(sk)->inet_num == num) {
76                         struct ipv6_pinfo *np = inet6_sk(sk);
77
78                         if (!net_eq(sock_net(sk), net))
79                                 continue;
80
81                         if (!ipv6_addr_any(&np->daddr) &&
82                             !ipv6_addr_equal(&np->daddr, rmt_addr))
83                                 continue;
84
85                         if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
86                                 continue;
87
88                         if (!ipv6_addr_any(&np->rcv_saddr)) {
89                                 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
90                                         goto found;
91                                 if (is_multicast &&
92                                     inet6_mc_check(sk, loc_addr, rmt_addr))
93                                         goto found;
94                                 continue;
95                         }
96                         goto found;
97                 }
98         sk = NULL;
99 found:
100         return sk;
101 }
102
103 /*
104  *      0 - deliver
105  *      1 - block
106  */
107 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
108 {
109         struct icmp6hdr *icmph;
110         struct raw6_sock *rp = raw6_sk(sk);
111
112         if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
113                 __u32 *data = &rp->filter.data[0];
114                 int bit_nr;
115
116                 icmph = (struct icmp6hdr *) skb->data;
117                 bit_nr = icmph->icmp6_type;
118
119                 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
120         }
121         return 0;
122 }
123
124 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
125 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
126
127 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
128                                            struct sk_buff *skb))
129 {
130         rcu_assign_pointer(mh_filter, filter);
131         return 0;
132 }
133 EXPORT_SYMBOL(rawv6_mh_filter_register);
134
135 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
136                                              struct sk_buff *skb))
137 {
138         rcu_assign_pointer(mh_filter, NULL);
139         synchronize_rcu();
140         return 0;
141 }
142 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
143
144 #endif
145
146 /*
147  *      demultiplex raw sockets.
148  *      (should consider queueing the skb in the sock receive_queue
149  *      without calling rawv6.c)
150  *
151  *      Caller owns SKB so we must make clones.
152  */
153 static int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
154 {
155         struct in6_addr *saddr;
156         struct in6_addr *daddr;
157         struct sock *sk;
158         int delivered = 0;
159         __u8 hash;
160         struct net *net;
161
162         saddr = &ipv6_hdr(skb)->saddr;
163         daddr = saddr + 1;
164
165         hash = nexthdr & (MAX_INET_PROTOS - 1);
166
167         read_lock(&raw_v6_hashinfo.lock);
168         sk = sk_head(&raw_v6_hashinfo.ht[hash]);
169
170         if (sk == NULL)
171                 goto out;
172
173         net = dev_net(skb->dev);
174         sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
175
176         while (sk) {
177                 int filtered;
178
179                 delivered = 1;
180                 switch (nexthdr) {
181                 case IPPROTO_ICMPV6:
182                         filtered = icmpv6_filter(sk, skb);
183                         break;
184
185 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
186                 case IPPROTO_MH:
187                 {
188                         /* XXX: To validate MH only once for each packet,
189                          * this is placed here. It should be after checking
190                          * xfrm policy, however it doesn't. The checking xfrm
191                          * policy is placed in rawv6_rcv() because it is
192                          * required for each socket.
193                          */
194                         int (*filter)(struct sock *sock, struct sk_buff *skb);
195
196                         filter = rcu_dereference(mh_filter);
197                         filtered = filter ? filter(sk, skb) : 0;
198                         break;
199                 }
200 #endif
201                 default:
202                         filtered = 0;
203                         break;
204                 }
205
206                 if (filtered < 0)
207                         break;
208                 if (filtered == 0) {
209                         struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
210
211                         /* Not releasing hash table! */
212                         if (clone) {
213                                 nf_reset(clone);
214                                 rawv6_rcv(sk, clone);
215                         }
216                 }
217                 sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
218                                      IP6CB(skb)->iif);
219         }
220 out:
221         read_unlock(&raw_v6_hashinfo.lock);
222         return delivered;
223 }
224
225 int raw6_local_deliver(struct sk_buff *skb, int nexthdr)
226 {
227         struct sock *raw_sk;
228
229         raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (MAX_INET_PROTOS - 1)]);
230         if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
231                 raw_sk = NULL;
232
233         return raw_sk != NULL;
234 }
235
236 /* This cleans up af_inet6 a bit. -DaveM */
237 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
238 {
239         struct inet_sock *inet = inet_sk(sk);
240         struct ipv6_pinfo *np = inet6_sk(sk);
241         struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
242         __be32 v4addr = 0;
243         int addr_type;
244         int err;
245
246         if (addr_len < SIN6_LEN_RFC2133)
247                 return -EINVAL;
248         addr_type = ipv6_addr_type(&addr->sin6_addr);
249
250         /* Raw sockets are IPv6 only */
251         if (addr_type == IPV6_ADDR_MAPPED)
252                 return -EADDRNOTAVAIL;
253
254         lock_sock(sk);
255
256         err = -EINVAL;
257         if (sk->sk_state != TCP_CLOSE)
258                 goto out;
259
260         rcu_read_lock();
261         /* Check if the address belongs to the host. */
262         if (addr_type != IPV6_ADDR_ANY) {
263                 struct net_device *dev = NULL;
264
265                 if (addr_type & IPV6_ADDR_LINKLOCAL) {
266                         if (addr_len >= sizeof(struct sockaddr_in6) &&
267                             addr->sin6_scope_id) {
268                                 /* Override any existing binding, if another
269                                  * one is supplied by user.
270                                  */
271                                 sk->sk_bound_dev_if = addr->sin6_scope_id;
272                         }
273
274                         /* Binding to link-local address requires an interface */
275                         if (!sk->sk_bound_dev_if)
276                                 goto out_unlock;
277
278                         err = -ENODEV;
279                         dev = dev_get_by_index_rcu(sock_net(sk),
280                                                    sk->sk_bound_dev_if);
281                         if (!dev)
282                                 goto out_unlock;
283                 }
284
285                 /* ipv4 addr of the socket is invalid.  Only the
286                  * unspecified and mapped address have a v4 equivalent.
287                  */
288                 v4addr = LOOPBACK4_IPV6;
289                 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
290                         err = -EADDRNOTAVAIL;
291                         if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
292                                            dev, 0)) {
293                                 goto out_unlock;
294                         }
295                 }
296         }
297
298         inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
299         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
300         if (!(addr_type & IPV6_ADDR_MULTICAST))
301                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
302         err = 0;
303 out_unlock:
304         rcu_read_unlock();
305 out:
306         release_sock(sk);
307         return err;
308 }
309
310 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
311                struct inet6_skb_parm *opt,
312                u8 type, u8 code, int offset, __be32 info)
313 {
314         struct inet_sock *inet = inet_sk(sk);
315         struct ipv6_pinfo *np = inet6_sk(sk);
316         int err;
317         int harderr;
318
319         /* Report error on raw socket, if:
320            1. User requested recverr.
321            2. Socket is connected (otherwise the error indication
322               is useless without recverr and error is hard.
323          */
324         if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
325                 return;
326
327         harderr = icmpv6_err_convert(type, code, &err);
328         if (type == ICMPV6_PKT_TOOBIG)
329                 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
330
331         if (np->recverr) {
332                 u8 *payload = skb->data;
333                 if (!inet->hdrincl)
334                         payload += offset;
335                 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
336         }
337
338         if (np->recverr || harderr) {
339                 sk->sk_err = err;
340                 sk->sk_error_report(sk);
341         }
342 }
343
344 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
345                 u8 type, u8 code, int inner_offset, __be32 info)
346 {
347         struct sock *sk;
348         int hash;
349         struct in6_addr *saddr, *daddr;
350         struct net *net;
351
352         hash = nexthdr & (RAW_HTABLE_SIZE - 1);
353
354         read_lock(&raw_v6_hashinfo.lock);
355         sk = sk_head(&raw_v6_hashinfo.ht[hash]);
356         if (sk != NULL) {
357                 /* Note: ipv6_hdr(skb) != skb->data */
358                 struct ipv6hdr *ip6h = (struct ipv6hdr *)skb->data;
359                 saddr = &ip6h->saddr;
360                 daddr = &ip6h->daddr;
361                 net = dev_net(skb->dev);
362
363                 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
364                                                 IP6CB(skb)->iif))) {
365                         rawv6_err(sk, skb, NULL, type, code,
366                                         inner_offset, info);
367                         sk = sk_next(sk);
368                 }
369         }
370         read_unlock(&raw_v6_hashinfo.lock);
371 }
372
373 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
374 {
375         if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
376             skb_checksum_complete(skb)) {
377                 atomic_inc(&sk->sk_drops);
378                 kfree_skb(skb);
379                 return NET_RX_DROP;
380         }
381
382         /* Charge it to the socket. */
383         if (sock_queue_rcv_skb(sk, skb) < 0) {
384                 kfree_skb(skb);
385                 return NET_RX_DROP;
386         }
387
388         return 0;
389 }
390
391 /*
392  *      This is next to useless...
393  *      if we demultiplex in network layer we don't need the extra call
394  *      just to queue the skb...
395  *      maybe we could have the network decide upon a hint if it
396  *      should call raw_rcv for demultiplexing
397  */
398 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
399 {
400         struct inet_sock *inet = inet_sk(sk);
401         struct raw6_sock *rp = raw6_sk(sk);
402
403         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
404                 atomic_inc(&sk->sk_drops);
405                 kfree_skb(skb);
406                 return NET_RX_DROP;
407         }
408
409         if (!rp->checksum)
410                 skb->ip_summed = CHECKSUM_UNNECESSARY;
411
412         if (skb->ip_summed == CHECKSUM_COMPLETE) {
413                 skb_postpull_rcsum(skb, skb_network_header(skb),
414                                    skb_network_header_len(skb));
415                 if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
416                                      &ipv6_hdr(skb)->daddr,
417                                      skb->len, inet->inet_num, skb->csum))
418                         skb->ip_summed = CHECKSUM_UNNECESSARY;
419         }
420         if (!skb_csum_unnecessary(skb))
421                 skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
422                                                          &ipv6_hdr(skb)->daddr,
423                                                          skb->len,
424                                                          inet->inet_num, 0));
425
426         if (inet->hdrincl) {
427                 if (skb_checksum_complete(skb)) {
428                         atomic_inc(&sk->sk_drops);
429                         kfree_skb(skb);
430                         return NET_RX_DROP;
431                 }
432         }
433
434         rawv6_rcv_skb(sk, skb);
435         return 0;
436 }
437
438
439 /*
440  *      This should be easy, if there is something there
441  *      we return it, otherwise we block.
442  */
443
444 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
445                   struct msghdr *msg, size_t len,
446                   int noblock, int flags, int *addr_len)
447 {
448         struct ipv6_pinfo *np = inet6_sk(sk);
449         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
450         struct sk_buff *skb;
451         size_t copied;
452         int err;
453
454         if (flags & MSG_OOB)
455                 return -EOPNOTSUPP;
456
457         if (addr_len)
458                 *addr_len=sizeof(*sin6);
459
460         if (flags & MSG_ERRQUEUE)
461                 return ipv6_recv_error(sk, msg, len);
462
463         skb = skb_recv_datagram(sk, flags, noblock, &err);
464         if (!skb)
465                 goto out;
466
467         copied = skb->len;
468         if (copied > len) {
469                 copied = len;
470                 msg->msg_flags |= MSG_TRUNC;
471         }
472
473         if (skb_csum_unnecessary(skb)) {
474                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
475         } else if (msg->msg_flags&MSG_TRUNC) {
476                 if (__skb_checksum_complete(skb))
477                         goto csum_copy_err;
478                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
479         } else {
480                 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
481                 if (err == -EINVAL)
482                         goto csum_copy_err;
483         }
484         if (err)
485                 goto out_free;
486
487         /* Copy the address. */
488         if (sin6) {
489                 sin6->sin6_family = AF_INET6;
490                 sin6->sin6_port = 0;
491                 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
492                 sin6->sin6_flowinfo = 0;
493                 sin6->sin6_scope_id = 0;
494                 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
495                         sin6->sin6_scope_id = IP6CB(skb)->iif;
496         }
497
498         sock_recv_ts_and_drops(msg, sk, skb);
499
500         if (np->rxopt.all)
501                 datagram_recv_ctl(sk, msg, skb);
502
503         err = copied;
504         if (flags & MSG_TRUNC)
505                 err = skb->len;
506
507 out_free:
508         skb_free_datagram(sk, skb);
509 out:
510         return err;
511
512 csum_copy_err:
513         skb_kill_datagram(sk, skb, flags);
514
515         /* Error for blocking case is chosen to masquerade
516            as some normal condition.
517          */
518         err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
519         goto out;
520 }
521
522 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
523                                      struct raw6_sock *rp)
524 {
525         struct sk_buff *skb;
526         int err = 0;
527         int offset;
528         int len;
529         int total_len;
530         __wsum tmp_csum;
531         __sum16 csum;
532
533         if (!rp->checksum)
534                 goto send;
535
536         if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
537                 goto out;
538
539         offset = rp->offset;
540         total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
541                                                 skb->data);
542         if (offset >= total_len - 1) {
543                 err = -EINVAL;
544                 ip6_flush_pending_frames(sk);
545                 goto out;
546         }
547
548         /* should be check HW csum miyazawa */
549         if (skb_queue_len(&sk->sk_write_queue) == 1) {
550                 /*
551                  * Only one fragment on the socket.
552                  */
553                 tmp_csum = skb->csum;
554         } else {
555                 struct sk_buff *csum_skb = NULL;
556                 tmp_csum = 0;
557
558                 skb_queue_walk(&sk->sk_write_queue, skb) {
559                         tmp_csum = csum_add(tmp_csum, skb->csum);
560
561                         if (csum_skb)
562                                 continue;
563
564                         len = skb->len - skb_transport_offset(skb);
565                         if (offset >= len) {
566                                 offset -= len;
567                                 continue;
568                         }
569
570                         csum_skb = skb;
571                 }
572
573                 skb = csum_skb;
574         }
575
576         offset += skb_transport_offset(skb);
577         if (skb_copy_bits(skb, offset, &csum, 2))
578                 BUG();
579
580         /* in case cksum was not initialized */
581         if (unlikely(csum))
582                 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
583
584         csum = csum_ipv6_magic(&fl->fl6_src,
585                                    &fl->fl6_dst,
586                                    total_len, fl->proto, tmp_csum);
587
588         if (csum == 0 && fl->proto == IPPROTO_UDP)
589                 csum = CSUM_MANGLED_0;
590
591         if (skb_store_bits(skb, offset, &csum, 2))
592                 BUG();
593
594 send:
595         err = ip6_push_pending_frames(sk);
596 out:
597         return err;
598 }
599
600 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
601                         struct flowi *fl, struct rt6_info *rt,
602                         unsigned int flags)
603 {
604         struct ipv6_pinfo *np = inet6_sk(sk);
605         struct ipv6hdr *iph;
606         struct sk_buff *skb;
607         int err;
608
609         if (length > rt->u.dst.dev->mtu) {
610                 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
611                 return -EMSGSIZE;
612         }
613         if (flags&MSG_PROBE)
614                 goto out;
615
616         skb = sock_alloc_send_skb(sk,
617                                   length + LL_ALLOCATED_SPACE(rt->u.dst.dev) + 15,
618                                   flags & MSG_DONTWAIT, &err);
619         if (skb == NULL)
620                 goto error;
621         skb_reserve(skb, LL_RESERVED_SPACE(rt->u.dst.dev));
622
623         skb->priority = sk->sk_priority;
624         skb->mark = sk->sk_mark;
625         skb_dst_set(skb, dst_clone(&rt->u.dst));
626
627         skb_put(skb, length);
628         skb_reset_network_header(skb);
629         iph = ipv6_hdr(skb);
630
631         skb->ip_summed = CHECKSUM_NONE;
632
633         skb->transport_header = skb->network_header;
634         err = memcpy_fromiovecend((void *)iph, from, 0, length);
635         if (err)
636                 goto error_fault;
637
638         IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
639         err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
640                       dst_output);
641         if (err > 0)
642                 err = net_xmit_errno(err);
643         if (err)
644                 goto error;
645 out:
646         return 0;
647
648 error_fault:
649         err = -EFAULT;
650         kfree_skb(skb);
651 error:
652         IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
653         if (err == -ENOBUFS && !np->recverr)
654                 err = 0;
655         return err;
656 }
657
658 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
659 {
660         struct iovec *iov;
661         u8 __user *type = NULL;
662         u8 __user *code = NULL;
663         u8 len = 0;
664         int probed = 0;
665         int i;
666
667         if (!msg->msg_iov)
668                 return 0;
669
670         for (i = 0; i < msg->msg_iovlen; i++) {
671                 iov = &msg->msg_iov[i];
672                 if (!iov)
673                         continue;
674
675                 switch (fl->proto) {
676                 case IPPROTO_ICMPV6:
677                         /* check if one-byte field is readable or not. */
678                         if (iov->iov_base && iov->iov_len < 1)
679                                 break;
680
681                         if (!type) {
682                                 type = iov->iov_base;
683                                 /* check if code field is readable or not. */
684                                 if (iov->iov_len > 1)
685                                         code = type + 1;
686                         } else if (!code)
687                                 code = iov->iov_base;
688
689                         if (type && code) {
690                                 if (get_user(fl->fl_icmp_type, type) ||
691                                     get_user(fl->fl_icmp_code, code))
692                                         return -EFAULT;
693                                 probed = 1;
694                         }
695                         break;
696                 case IPPROTO_MH:
697                         if (iov->iov_base && iov->iov_len < 1)
698                                 break;
699                         /* check if type field is readable or not. */
700                         if (iov->iov_len > 2 - len) {
701                                 u8 __user *p = iov->iov_base;
702                                 if (get_user(fl->fl_mh_type, &p[2 - len]))
703                                         return -EFAULT;
704                                 probed = 1;
705                         } else
706                                 len += iov->iov_len;
707
708                         break;
709                 default:
710                         probed = 1;
711                         break;
712                 }
713                 if (probed)
714                         break;
715         }
716         return 0;
717 }
718
719 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
720                    struct msghdr *msg, size_t len)
721 {
722         struct ipv6_txoptions opt_space;
723         struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
724         struct in6_addr *daddr, *final_p = NULL, final;
725         struct inet_sock *inet = inet_sk(sk);
726         struct ipv6_pinfo *np = inet6_sk(sk);
727         struct raw6_sock *rp = raw6_sk(sk);
728         struct ipv6_txoptions *opt = NULL;
729         struct ip6_flowlabel *flowlabel = NULL;
730         struct dst_entry *dst = NULL;
731         struct flowi fl;
732         int addr_len = msg->msg_namelen;
733         int hlimit = -1;
734         int tclass = -1;
735         u16 proto;
736         int err;
737
738         /* Rough check on arithmetic overflow,
739            better check is made in ip6_append_data().
740          */
741         if (len > INT_MAX)
742                 return -EMSGSIZE;
743
744         /* Mirror BSD error message compatibility */
745         if (msg->msg_flags & MSG_OOB)
746                 return -EOPNOTSUPP;
747
748         /*
749          *      Get and verify the address.
750          */
751         memset(&fl, 0, sizeof(fl));
752
753         fl.mark = sk->sk_mark;
754
755         if (sin6) {
756                 if (addr_len < SIN6_LEN_RFC2133)
757                         return -EINVAL;
758
759                 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
760                         return(-EAFNOSUPPORT);
761
762                 /* port is the proto value [0..255] carried in nexthdr */
763                 proto = ntohs(sin6->sin6_port);
764
765                 if (!proto)
766                         proto = inet->inet_num;
767                 else if (proto != inet->inet_num)
768                         return(-EINVAL);
769
770                 if (proto > 255)
771                         return(-EINVAL);
772
773                 daddr = &sin6->sin6_addr;
774                 if (np->sndflow) {
775                         fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
776                         if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
777                                 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
778                                 if (flowlabel == NULL)
779                                         return -EINVAL;
780                                 daddr = &flowlabel->dst;
781                         }
782                 }
783
784                 /*
785                  * Otherwise it will be difficult to maintain
786                  * sk->sk_dst_cache.
787                  */
788                 if (sk->sk_state == TCP_ESTABLISHED &&
789                     ipv6_addr_equal(daddr, &np->daddr))
790                         daddr = &np->daddr;
791
792                 if (addr_len >= sizeof(struct sockaddr_in6) &&
793                     sin6->sin6_scope_id &&
794                     ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
795                         fl.oif = sin6->sin6_scope_id;
796         } else {
797                 if (sk->sk_state != TCP_ESTABLISHED)
798                         return -EDESTADDRREQ;
799
800                 proto = inet->inet_num;
801                 daddr = &np->daddr;
802                 fl.fl6_flowlabel = np->flow_label;
803         }
804
805         if (fl.oif == 0)
806                 fl.oif = sk->sk_bound_dev_if;
807
808         if (msg->msg_controllen) {
809                 opt = &opt_space;
810                 memset(opt, 0, sizeof(struct ipv6_txoptions));
811                 opt->tot_len = sizeof(struct ipv6_txoptions);
812
813                 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit, &tclass);
814                 if (err < 0) {
815                         fl6_sock_release(flowlabel);
816                         return err;
817                 }
818                 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
819                         flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
820                         if (flowlabel == NULL)
821                                 return -EINVAL;
822                 }
823                 if (!(opt->opt_nflen|opt->opt_flen))
824                         opt = NULL;
825         }
826         if (opt == NULL)
827                 opt = np->opt;
828         if (flowlabel)
829                 opt = fl6_merge_options(&opt_space, flowlabel, opt);
830         opt = ipv6_fixup_options(&opt_space, opt);
831
832         fl.proto = proto;
833         err = rawv6_probe_proto_opt(&fl, msg);
834         if (err)
835                 goto out;
836
837         if (!ipv6_addr_any(daddr))
838                 ipv6_addr_copy(&fl.fl6_dst, daddr);
839         else
840                 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
841         if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
842                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
843
844         /* merge ip6_build_xmit from ip6_output */
845         if (opt && opt->srcrt) {
846                 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
847                 ipv6_addr_copy(&final, &fl.fl6_dst);
848                 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
849                 final_p = &final;
850         }
851
852         if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
853                 fl.oif = np->mcast_oif;
854         security_sk_classify_flow(sk, &fl);
855
856         err = ip6_dst_lookup(sk, &dst, &fl);
857         if (err)
858                 goto out;
859         if (final_p)
860                 ipv6_addr_copy(&fl.fl6_dst, final_p);
861
862         err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT);
863         if (err < 0) {
864                 if (err == -EREMOTE)
865                         err = ip6_dst_blackhole(sk, &dst, &fl);
866                 if (err < 0)
867                         goto out;
868         }
869
870         if (hlimit < 0) {
871                 if (ipv6_addr_is_multicast(&fl.fl6_dst))
872                         hlimit = np->mcast_hops;
873                 else
874                         hlimit = np->hop_limit;
875                 if (hlimit < 0)
876                         hlimit = ip6_dst_hoplimit(dst);
877         }
878
879         if (tclass < 0)
880                 tclass = np->tclass;
881
882         if (msg->msg_flags&MSG_CONFIRM)
883                 goto do_confirm;
884
885 back_from_confirm:
886         if (inet->hdrincl) {
887                 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
888         } else {
889                 lock_sock(sk);
890                 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
891                         len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
892                         msg->msg_flags);
893
894                 if (err)
895                         ip6_flush_pending_frames(sk);
896                 else if (!(msg->msg_flags & MSG_MORE))
897                         err = rawv6_push_pending_frames(sk, &fl, rp);
898                 release_sock(sk);
899         }
900 done:
901         dst_release(dst);
902 out:
903         fl6_sock_release(flowlabel);
904         return err<0?err:len;
905 do_confirm:
906         dst_confirm(dst);
907         if (!(msg->msg_flags & MSG_PROBE) || len)
908                 goto back_from_confirm;
909         err = 0;
910         goto done;
911 }
912
913 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
914                                char __user *optval, int optlen)
915 {
916         switch (optname) {
917         case ICMPV6_FILTER:
918                 if (optlen > sizeof(struct icmp6_filter))
919                         optlen = sizeof(struct icmp6_filter);
920                 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
921                         return -EFAULT;
922                 return 0;
923         default:
924                 return -ENOPROTOOPT;
925         }
926
927         return 0;
928 }
929
930 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
931                                char __user *optval, int __user *optlen)
932 {
933         int len;
934
935         switch (optname) {
936         case ICMPV6_FILTER:
937                 if (get_user(len, optlen))
938                         return -EFAULT;
939                 if (len < 0)
940                         return -EINVAL;
941                 if (len > sizeof(struct icmp6_filter))
942                         len = sizeof(struct icmp6_filter);
943                 if (put_user(len, optlen))
944                         return -EFAULT;
945                 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
946                         return -EFAULT;
947                 return 0;
948         default:
949                 return -ENOPROTOOPT;
950         }
951
952         return 0;
953 }
954
955
956 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
957                             char __user *optval, unsigned int optlen)
958 {
959         struct raw6_sock *rp = raw6_sk(sk);
960         int val;
961
962         if (get_user(val, (int __user *)optval))
963                 return -EFAULT;
964
965         switch (optname) {
966                 case IPV6_CHECKSUM:
967                         if (inet_sk(sk)->inet_num == IPPROTO_ICMPV6 &&
968                             level == IPPROTO_IPV6) {
969                                 /*
970                                  * RFC3542 tells that IPV6_CHECKSUM socket
971                                  * option in the IPPROTO_IPV6 level is not
972                                  * allowed on ICMPv6 sockets.
973                                  * If you want to set it, use IPPROTO_RAW
974                                  * level IPV6_CHECKSUM socket option
975                                  * (Linux extension).
976                                  */
977                                 return -EINVAL;
978                         }
979
980                         /* You may get strange result with a positive odd offset;
981                            RFC2292bis agrees with me. */
982                         if (val > 0 && (val&1))
983                                 return(-EINVAL);
984                         if (val < 0) {
985                                 rp->checksum = 0;
986                         } else {
987                                 rp->checksum = 1;
988                                 rp->offset = val;
989                         }
990
991                         return 0;
992                         break;
993
994                 default:
995                         return(-ENOPROTOOPT);
996         }
997 }
998
999 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1000                           char __user *optval, unsigned int optlen)
1001 {
1002         switch(level) {
1003                 case SOL_RAW:
1004                         break;
1005
1006                 case SOL_ICMPV6:
1007                         if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1008                                 return -EOPNOTSUPP;
1009                         return rawv6_seticmpfilter(sk, level, optname, optval,
1010                                                    optlen);
1011                 case SOL_IPV6:
1012                         if (optname == IPV6_CHECKSUM)
1013                                 break;
1014                 default:
1015                         return ipv6_setsockopt(sk, level, optname, optval,
1016                                                optlen);
1017         }
1018
1019         return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1020 }
1021
1022 #ifdef CONFIG_COMPAT
1023 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1024                                    char __user *optval, unsigned int optlen)
1025 {
1026         switch (level) {
1027         case SOL_RAW:
1028                 break;
1029         case SOL_ICMPV6:
1030                 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1031                         return -EOPNOTSUPP;
1032                 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1033         case SOL_IPV6:
1034                 if (optname == IPV6_CHECKSUM)
1035                         break;
1036         default:
1037                 return compat_ipv6_setsockopt(sk, level, optname,
1038                                               optval, optlen);
1039         }
1040         return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1041 }
1042 #endif
1043
1044 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1045                             char __user *optval, int __user *optlen)
1046 {
1047         struct raw6_sock *rp = raw6_sk(sk);
1048         int val, len;
1049
1050         if (get_user(len,optlen))
1051                 return -EFAULT;
1052
1053         switch (optname) {
1054         case IPV6_CHECKSUM:
1055                 /*
1056                  * We allow getsockopt() for IPPROTO_IPV6-level
1057                  * IPV6_CHECKSUM socket option on ICMPv6 sockets
1058                  * since RFC3542 is silent about it.
1059                  */
1060                 if (rp->checksum == 0)
1061                         val = -1;
1062                 else
1063                         val = rp->offset;
1064                 break;
1065
1066         default:
1067                 return -ENOPROTOOPT;
1068         }
1069
1070         len = min_t(unsigned int, sizeof(int), len);
1071
1072         if (put_user(len, optlen))
1073                 return -EFAULT;
1074         if (copy_to_user(optval,&val,len))
1075                 return -EFAULT;
1076         return 0;
1077 }
1078
1079 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1080                           char __user *optval, int __user *optlen)
1081 {
1082         switch(level) {
1083                 case SOL_RAW:
1084                         break;
1085
1086                 case SOL_ICMPV6:
1087                         if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1088                                 return -EOPNOTSUPP;
1089                         return rawv6_geticmpfilter(sk, level, optname, optval,
1090                                                    optlen);
1091                 case SOL_IPV6:
1092                         if (optname == IPV6_CHECKSUM)
1093                                 break;
1094                 default:
1095                         return ipv6_getsockopt(sk, level, optname, optval,
1096                                                optlen);
1097         }
1098
1099         return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1100 }
1101
1102 #ifdef CONFIG_COMPAT
1103 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1104                                    char __user *optval, int __user *optlen)
1105 {
1106         switch (level) {
1107         case SOL_RAW:
1108                 break;
1109         case SOL_ICMPV6:
1110                 if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1111                         return -EOPNOTSUPP;
1112                 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1113         case SOL_IPV6:
1114                 if (optname == IPV6_CHECKSUM)
1115                         break;
1116         default:
1117                 return compat_ipv6_getsockopt(sk, level, optname,
1118                                               optval, optlen);
1119         }
1120         return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1121 }
1122 #endif
1123
1124 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1125 {
1126         switch(cmd) {
1127                 case SIOCOUTQ:
1128                 {
1129                         int amount = sk_wmem_alloc_get(sk);
1130
1131                         return put_user(amount, (int __user *)arg);
1132                 }
1133                 case SIOCINQ:
1134                 {
1135                         struct sk_buff *skb;
1136                         int amount = 0;
1137
1138                         spin_lock_bh(&sk->sk_receive_queue.lock);
1139                         skb = skb_peek(&sk->sk_receive_queue);
1140                         if (skb != NULL)
1141                                 amount = skb->tail - skb->transport_header;
1142                         spin_unlock_bh(&sk->sk_receive_queue.lock);
1143                         return put_user(amount, (int __user *)arg);
1144                 }
1145
1146                 default:
1147 #ifdef CONFIG_IPV6_MROUTE
1148                         return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1149 #else
1150                         return -ENOIOCTLCMD;
1151 #endif
1152         }
1153 }
1154
1155 static void rawv6_close(struct sock *sk, long timeout)
1156 {
1157         if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1158                 ip6_ra_control(sk, -1);
1159         ip6mr_sk_done(sk);
1160         sk_common_release(sk);
1161 }
1162
1163 static void raw6_destroy(struct sock *sk)
1164 {
1165         lock_sock(sk);
1166         ip6_flush_pending_frames(sk);
1167         release_sock(sk);
1168
1169         inet6_destroy_sock(sk);
1170 }
1171
1172 static int rawv6_init_sk(struct sock *sk)
1173 {
1174         struct raw6_sock *rp = raw6_sk(sk);
1175
1176         switch (inet_sk(sk)->inet_num) {
1177         case IPPROTO_ICMPV6:
1178                 rp->checksum = 1;
1179                 rp->offset   = 2;
1180                 break;
1181         case IPPROTO_MH:
1182                 rp->checksum = 1;
1183                 rp->offset   = 4;
1184                 break;
1185         default:
1186                 break;
1187         }
1188         return(0);
1189 }
1190
1191 struct proto rawv6_prot = {
1192         .name              = "RAWv6",
1193         .owner             = THIS_MODULE,
1194         .close             = rawv6_close,
1195         .destroy           = raw6_destroy,
1196         .connect           = ip6_datagram_connect,
1197         .disconnect        = udp_disconnect,
1198         .ioctl             = rawv6_ioctl,
1199         .init              = rawv6_init_sk,
1200         .setsockopt        = rawv6_setsockopt,
1201         .getsockopt        = rawv6_getsockopt,
1202         .sendmsg           = rawv6_sendmsg,
1203         .recvmsg           = rawv6_recvmsg,
1204         .bind              = rawv6_bind,
1205         .backlog_rcv       = rawv6_rcv_skb,
1206         .hash              = raw_hash_sk,
1207         .unhash            = raw_unhash_sk,
1208         .obj_size          = sizeof(struct raw6_sock),
1209         .h.raw_hash        = &raw_v6_hashinfo,
1210 #ifdef CONFIG_COMPAT
1211         .compat_setsockopt = compat_rawv6_setsockopt,
1212         .compat_getsockopt = compat_rawv6_getsockopt,
1213 #endif
1214 };
1215
1216 #ifdef CONFIG_PROC_FS
1217 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1218 {
1219         struct ipv6_pinfo *np = inet6_sk(sp);
1220         struct in6_addr *dest, *src;
1221         __u16 destp, srcp;
1222
1223         dest  = &np->daddr;
1224         src   = &np->rcv_saddr;
1225         destp = 0;
1226         srcp  = inet_sk(sp)->inet_num;
1227         seq_printf(seq,
1228                    "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1229                    "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1230                    i,
1231                    src->s6_addr32[0], src->s6_addr32[1],
1232                    src->s6_addr32[2], src->s6_addr32[3], srcp,
1233                    dest->s6_addr32[0], dest->s6_addr32[1],
1234                    dest->s6_addr32[2], dest->s6_addr32[3], destp,
1235                    sp->sk_state,
1236                    sk_wmem_alloc_get(sp),
1237                    sk_rmem_alloc_get(sp),
1238                    0, 0L, 0,
1239                    sock_i_uid(sp), 0,
1240                    sock_i_ino(sp),
1241                    atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1242 }
1243
1244 static int raw6_seq_show(struct seq_file *seq, void *v)
1245 {
1246         if (v == SEQ_START_TOKEN)
1247                 seq_printf(seq,
1248                            "  sl  "
1249                            "local_address                         "
1250                            "remote_address                        "
1251                            "st tx_queue rx_queue tr tm->when retrnsmt"
1252                            "   uid  timeout inode ref pointer drops\n");
1253         else
1254                 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1255         return 0;
1256 }
1257
1258 static const struct seq_operations raw6_seq_ops = {
1259         .start =        raw_seq_start,
1260         .next =         raw_seq_next,
1261         .stop =         raw_seq_stop,
1262         .show =         raw6_seq_show,
1263 };
1264
1265 static int raw6_seq_open(struct inode *inode, struct file *file)
1266 {
1267         return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1268 }
1269
1270 static const struct file_operations raw6_seq_fops = {
1271         .owner =        THIS_MODULE,
1272         .open =         raw6_seq_open,
1273         .read =         seq_read,
1274         .llseek =       seq_lseek,
1275         .release =      seq_release_net,
1276 };
1277
1278 static int __net_init raw6_init_net(struct net *net)
1279 {
1280         if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1281                 return -ENOMEM;
1282
1283         return 0;
1284 }
1285
1286 static void __net_exit raw6_exit_net(struct net *net)
1287 {
1288         proc_net_remove(net, "raw6");
1289 }
1290
1291 static struct pernet_operations raw6_net_ops = {
1292         .init = raw6_init_net,
1293         .exit = raw6_exit_net,
1294 };
1295
1296 int __init raw6_proc_init(void)
1297 {
1298         return register_pernet_subsys(&raw6_net_ops);
1299 }
1300
1301 void raw6_proc_exit(void)
1302 {
1303         unregister_pernet_subsys(&raw6_net_ops);
1304 }
1305 #endif  /* CONFIG_PROC_FS */
1306
1307 /* Same as inet6_dgram_ops, sans udp_poll.  */
1308 static const struct proto_ops inet6_sockraw_ops = {
1309         .family            = PF_INET6,
1310         .owner             = THIS_MODULE,
1311         .release           = inet6_release,
1312         .bind              = inet6_bind,
1313         .connect           = inet_dgram_connect,        /* ok           */
1314         .socketpair        = sock_no_socketpair,        /* a do nothing */
1315         .accept            = sock_no_accept,            /* a do nothing */
1316         .getname           = inet6_getname,
1317         .poll              = datagram_poll,             /* ok           */
1318         .ioctl             = inet6_ioctl,               /* must change  */
1319         .listen            = sock_no_listen,            /* ok           */
1320         .shutdown          = inet_shutdown,             /* ok           */
1321         .setsockopt        = sock_common_setsockopt,    /* ok           */
1322         .getsockopt        = sock_common_getsockopt,    /* ok           */
1323         .sendmsg           = inet_sendmsg,              /* ok           */
1324         .recvmsg           = sock_common_recvmsg,       /* ok           */
1325         .mmap              = sock_no_mmap,
1326         .sendpage          = sock_no_sendpage,
1327 #ifdef CONFIG_COMPAT
1328         .compat_setsockopt = compat_sock_common_setsockopt,
1329         .compat_getsockopt = compat_sock_common_getsockopt,
1330 #endif
1331 };
1332
1333 static struct inet_protosw rawv6_protosw = {
1334         .type           = SOCK_RAW,
1335         .protocol       = IPPROTO_IP,   /* wild card */
1336         .prot           = &rawv6_prot,
1337         .ops            = &inet6_sockraw_ops,
1338         .no_check       = UDP_CSUM_DEFAULT,
1339         .flags          = INET_PROTOSW_REUSE,
1340 };
1341
1342 int __init rawv6_init(void)
1343 {
1344         int ret;
1345
1346         ret = inet6_register_protosw(&rawv6_protosw);
1347         if (ret)
1348                 goto out;
1349 out:
1350         return ret;
1351 }
1352
1353 void rawv6_exit(void)
1354 {
1355         inet6_unregister_protosw(&rawv6_protosw);
1356 }