Merge branches 'timers/clocksource', 'timers/hrtimers', 'timers/nohz', 'timers/ntp...
[linux-2.6] / 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)->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         /* Check if the address belongs to the host. */
261         if (addr_type != IPV6_ADDR_ANY) {
262                 struct net_device *dev = NULL;
263
264                 if (addr_type & IPV6_ADDR_LINKLOCAL) {
265                         if (addr_len >= sizeof(struct sockaddr_in6) &&
266                             addr->sin6_scope_id) {
267                                 /* Override any existing binding, if another
268                                  * one is supplied by user.
269                                  */
270                                 sk->sk_bound_dev_if = addr->sin6_scope_id;
271                         }
272
273                         /* Binding to link-local address requires an interface */
274                         if (!sk->sk_bound_dev_if)
275                                 goto out;
276
277                         dev = dev_get_by_index(sock_net(sk), sk->sk_bound_dev_if);
278                         if (!dev) {
279                                 err = -ENODEV;
280                                 goto out;
281                         }
282                 }
283
284                 /* ipv4 addr of the socket is invalid.  Only the
285                  * unspecified and mapped address have a v4 equivalent.
286                  */
287                 v4addr = LOOPBACK4_IPV6;
288                 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
289                         err = -EADDRNOTAVAIL;
290                         if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
291                                            dev, 0)) {
292                                 if (dev)
293                                         dev_put(dev);
294                                 goto out;
295                         }
296                 }
297                 if (dev)
298                         dev_put(dev);
299         }
300
301         inet->rcv_saddr = inet->saddr = v4addr;
302         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
303         if (!(addr_type & IPV6_ADDR_MULTICAST))
304                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
305         err = 0;
306 out:
307         release_sock(sk);
308         return err;
309 }
310
311 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
312                struct inet6_skb_parm *opt,
313                int type, int code, int offset, __be32 info)
314 {
315         struct inet_sock *inet = inet_sk(sk);
316         struct ipv6_pinfo *np = inet6_sk(sk);
317         int err;
318         int harderr;
319
320         /* Report error on raw socket, if:
321            1. User requested recverr.
322            2. Socket is connected (otherwise the error indication
323               is useless without recverr and error is hard.
324          */
325         if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
326                 return;
327
328         harderr = icmpv6_err_convert(type, code, &err);
329         if (type == ICMPV6_PKT_TOOBIG)
330                 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
331
332         if (np->recverr) {
333                 u8 *payload = skb->data;
334                 if (!inet->hdrincl)
335                         payload += offset;
336                 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
337         }
338
339         if (np->recverr || harderr) {
340                 sk->sk_err = err;
341                 sk->sk_error_report(sk);
342         }
343 }
344
345 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
346                 int type, int code, int inner_offset, __be32 info)
347 {
348         struct sock *sk;
349         int hash;
350         struct in6_addr *saddr, *daddr;
351         struct net *net;
352
353         hash = nexthdr & (RAW_HTABLE_SIZE - 1);
354
355         read_lock(&raw_v6_hashinfo.lock);
356         sk = sk_head(&raw_v6_hashinfo.ht[hash]);
357         if (sk != NULL) {
358                 /* Note: ipv6_hdr(skb) != skb->data */
359                 struct ipv6hdr *ip6h = (struct ipv6hdr *)skb->data;
360                 saddr = &ip6h->saddr;
361                 daddr = &ip6h->daddr;
362                 net = dev_net(skb->dev);
363
364                 while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
365                                                 IP6CB(skb)->iif))) {
366                         rawv6_err(sk, skb, NULL, type, code,
367                                         inner_offset, info);
368                         sk = sk_next(sk);
369                 }
370         }
371         read_unlock(&raw_v6_hashinfo.lock);
372 }
373
374 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
375 {
376         if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
377             skb_checksum_complete(skb)) {
378                 atomic_inc(&sk->sk_drops);
379                 kfree_skb(skb);
380                 return NET_RX_DROP;
381         }
382
383         /* Charge it to the socket. */
384         if (sock_queue_rcv_skb(sk,skb)<0) {
385                 atomic_inc(&sk->sk_drops);
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->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->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         skb = skb_recv_datagram(sk, flags, noblock, &err);
466         if (!skb)
467                 goto out;
468
469         copied = skb->len;
470         if (copied > len) {
471                 copied = len;
472                 msg->msg_flags |= MSG_TRUNC;
473         }
474
475         if (skb_csum_unnecessary(skb)) {
476                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
477         } else if (msg->msg_flags&MSG_TRUNC) {
478                 if (__skb_checksum_complete(skb))
479                         goto csum_copy_err;
480                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
481         } else {
482                 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
483                 if (err == -EINVAL)
484                         goto csum_copy_err;
485         }
486         if (err)
487                 goto out_free;
488
489         /* Copy the address. */
490         if (sin6) {
491                 sin6->sin6_family = AF_INET6;
492                 sin6->sin6_port = 0;
493                 ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
494                 sin6->sin6_flowinfo = 0;
495                 sin6->sin6_scope_id = 0;
496                 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
497                         sin6->sin6_scope_id = IP6CB(skb)->iif;
498         }
499
500         sock_recv_timestamp(msg, sk, skb);
501
502         if (np->rxopt.all)
503                 datagram_recv_ctl(sk, msg, skb);
504
505         err = copied;
506         if (flags & MSG_TRUNC)
507                 err = skb->len;
508
509 out_free:
510         skb_free_datagram(sk, skb);
511 out:
512         return err;
513
514 csum_copy_err:
515         skb_kill_datagram(sk, skb, flags);
516
517         /* Error for blocking case is chosen to masquerade
518            as some normal condition.
519          */
520         err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
521         atomic_inc(&sk->sk_drops);
522         goto out;
523 }
524
525 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
526                                      struct raw6_sock *rp)
527 {
528         struct sk_buff *skb;
529         int err = 0;
530         int offset;
531         int len;
532         int total_len;
533         __wsum tmp_csum;
534         __sum16 csum;
535
536         if (!rp->checksum)
537                 goto send;
538
539         if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
540                 goto out;
541
542         offset = rp->offset;
543         total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
544                                                 skb->data);
545         if (offset >= total_len - 1) {
546                 err = -EINVAL;
547                 ip6_flush_pending_frames(sk);
548                 goto out;
549         }
550
551         /* should be check HW csum miyazawa */
552         if (skb_queue_len(&sk->sk_write_queue) == 1) {
553                 /*
554                  * Only one fragment on the socket.
555                  */
556                 tmp_csum = skb->csum;
557         } else {
558                 struct sk_buff *csum_skb = NULL;
559                 tmp_csum = 0;
560
561                 skb_queue_walk(&sk->sk_write_queue, skb) {
562                         tmp_csum = csum_add(tmp_csum, skb->csum);
563
564                         if (csum_skb)
565                                 continue;
566
567                         len = skb->len - skb_transport_offset(skb);
568                         if (offset >= len) {
569                                 offset -= len;
570                                 continue;
571                         }
572
573                         csum_skb = skb;
574                 }
575
576                 skb = csum_skb;
577         }
578
579         offset += skb_transport_offset(skb);
580         if (skb_copy_bits(skb, offset, &csum, 2))
581                 BUG();
582
583         /* in case cksum was not initialized */
584         if (unlikely(csum))
585                 tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
586
587         csum = csum_ipv6_magic(&fl->fl6_src,
588                                    &fl->fl6_dst,
589                                    total_len, fl->proto, tmp_csum);
590
591         if (csum == 0 && fl->proto == IPPROTO_UDP)
592                 csum = CSUM_MANGLED_0;
593
594         if (skb_store_bits(skb, offset, &csum, 2))
595                 BUG();
596
597 send:
598         err = ip6_push_pending_frames(sk);
599 out:
600         return err;
601 }
602
603 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
604                         struct flowi *fl, struct rt6_info *rt,
605                         unsigned int flags)
606 {
607         struct ipv6_pinfo *np = inet6_sk(sk);
608         struct ipv6hdr *iph;
609         struct sk_buff *skb;
610         int err;
611
612         if (length > rt->u.dst.dev->mtu) {
613                 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
614                 return -EMSGSIZE;
615         }
616         if (flags&MSG_PROBE)
617                 goto out;
618
619         skb = sock_alloc_send_skb(sk,
620                                   length + LL_ALLOCATED_SPACE(rt->u.dst.dev) + 15,
621                                   flags & MSG_DONTWAIT, &err);
622         if (skb == NULL)
623                 goto error;
624         skb_reserve(skb, LL_RESERVED_SPACE(rt->u.dst.dev));
625
626         skb->priority = sk->sk_priority;
627         skb->mark = sk->sk_mark;
628         skb->dst = dst_clone(&rt->u.dst);
629
630         skb_put(skb, length);
631         skb_reset_network_header(skb);
632         iph = ipv6_hdr(skb);
633
634         skb->ip_summed = CHECKSUM_NONE;
635
636         skb->transport_header = skb->network_header;
637         err = memcpy_fromiovecend((void *)iph, from, 0, length);
638         if (err)
639                 goto error_fault;
640
641         IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
642         err = NF_HOOK(PF_INET6, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
643                       dst_output);
644         if (err > 0)
645                 err = np->recverr ? net_xmit_errno(err) : 0;
646         if (err)
647                 goto error;
648 out:
649         return 0;
650
651 error_fault:
652         err = -EFAULT;
653         kfree_skb(skb);
654 error:
655         IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
656         return err;
657 }
658
659 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
660 {
661         struct iovec *iov;
662         u8 __user *type = NULL;
663         u8 __user *code = NULL;
664         u8 len = 0;
665         int probed = 0;
666         int i;
667
668         if (!msg->msg_iov)
669                 return 0;
670
671         for (i = 0; i < msg->msg_iovlen; i++) {
672                 iov = &msg->msg_iov[i];
673                 if (!iov)
674                         continue;
675
676                 switch (fl->proto) {
677                 case IPPROTO_ICMPV6:
678                         /* check if one-byte field is readable or not. */
679                         if (iov->iov_base && iov->iov_len < 1)
680                                 break;
681
682                         if (!type) {
683                                 type = iov->iov_base;
684                                 /* check if code field is readable or not. */
685                                 if (iov->iov_len > 1)
686                                         code = type + 1;
687                         } else if (!code)
688                                 code = iov->iov_base;
689
690                         if (type && code) {
691                                 if (get_user(fl->fl_icmp_type, type) ||
692                                     get_user(fl->fl_icmp_code, code))
693                                         return -EFAULT;
694                                 probed = 1;
695                         }
696                         break;
697                 case IPPROTO_MH:
698                         if (iov->iov_base && iov->iov_len < 1)
699                                 break;
700                         /* check if type field is readable or not. */
701                         if (iov->iov_len > 2 - len) {
702                                 u8 __user *p = iov->iov_base;
703                                 if (get_user(fl->fl_mh_type, &p[2 - len]))
704                                         return -EFAULT;
705                                 probed = 1;
706                         } else
707                                 len += iov->iov_len;
708
709                         break;
710                 default:
711                         probed = 1;
712                         break;
713                 }
714                 if (probed)
715                         break;
716         }
717         return 0;
718 }
719
720 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
721                    struct msghdr *msg, size_t len)
722 {
723         struct ipv6_txoptions opt_space;
724         struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
725         struct in6_addr *daddr, *final_p = NULL, final;
726         struct inet_sock *inet = inet_sk(sk);
727         struct ipv6_pinfo *np = inet6_sk(sk);
728         struct raw6_sock *rp = raw6_sk(sk);
729         struct ipv6_txoptions *opt = NULL;
730         struct ip6_flowlabel *flowlabel = NULL;
731         struct dst_entry *dst = NULL;
732         struct flowi fl;
733         int addr_len = msg->msg_namelen;
734         int hlimit = -1;
735         int tclass = -1;
736         u16 proto;
737         int err;
738
739         /* Rough check on arithmetic overflow,
740            better check is made in ip6_append_data().
741          */
742         if (len > INT_MAX)
743                 return -EMSGSIZE;
744
745         /* Mirror BSD error message compatibility */
746         if (msg->msg_flags & MSG_OOB)
747                 return -EOPNOTSUPP;
748
749         /*
750          *      Get and verify the address.
751          */
752         memset(&fl, 0, sizeof(fl));
753
754         fl.mark = sk->sk_mark;
755
756         if (sin6) {
757                 if (addr_len < SIN6_LEN_RFC2133)
758                         return -EINVAL;
759
760                 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
761                         return(-EAFNOSUPPORT);
762
763                 /* port is the proto value [0..255] carried in nexthdr */
764                 proto = ntohs(sin6->sin6_port);
765
766                 if (!proto)
767                         proto = inet->num;
768                 else if (proto != inet->num)
769                         return(-EINVAL);
770
771                 if (proto > 255)
772                         return(-EINVAL);
773
774                 daddr = &sin6->sin6_addr;
775                 if (np->sndflow) {
776                         fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
777                         if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
778                                 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
779                                 if (flowlabel == NULL)
780                                         return -EINVAL;
781                                 daddr = &flowlabel->dst;
782                         }
783                 }
784
785                 /*
786                  * Otherwise it will be difficult to maintain
787                  * sk->sk_dst_cache.
788                  */
789                 if (sk->sk_state == TCP_ESTABLISHED &&
790                     ipv6_addr_equal(daddr, &np->daddr))
791                         daddr = &np->daddr;
792
793                 if (addr_len >= sizeof(struct sockaddr_in6) &&
794                     sin6->sin6_scope_id &&
795                     ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
796                         fl.oif = sin6->sin6_scope_id;
797         } else {
798                 if (sk->sk_state != TCP_ESTABLISHED)
799                         return -EDESTADDRREQ;
800
801                 proto = inet->num;
802                 daddr = &np->daddr;
803                 fl.fl6_flowlabel = np->flow_label;
804         }
805
806         if (fl.oif == 0)
807                 fl.oif = sk->sk_bound_dev_if;
808
809         if (msg->msg_controllen) {
810                 opt = &opt_space;
811                 memset(opt, 0, sizeof(struct ipv6_txoptions));
812                 opt->tot_len = sizeof(struct ipv6_txoptions);
813
814                 err = datagram_send_ctl(sock_net(sk), msg, &fl, opt, &hlimit, &tclass);
815                 if (err < 0) {
816                         fl6_sock_release(flowlabel);
817                         return err;
818                 }
819                 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
820                         flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
821                         if (flowlabel == NULL)
822                                 return -EINVAL;
823                 }
824                 if (!(opt->opt_nflen|opt->opt_flen))
825                         opt = NULL;
826         }
827         if (opt == NULL)
828                 opt = np->opt;
829         if (flowlabel)
830                 opt = fl6_merge_options(&opt_space, flowlabel, opt);
831         opt = ipv6_fixup_options(&opt_space, opt);
832
833         fl.proto = proto;
834         err = rawv6_probe_proto_opt(&fl, msg);
835         if (err)
836                 goto out;
837
838         if (!ipv6_addr_any(daddr))
839                 ipv6_addr_copy(&fl.fl6_dst, daddr);
840         else
841                 fl.fl6_dst.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
842         if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
843                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
844
845         /* merge ip6_build_xmit from ip6_output */
846         if (opt && opt->srcrt) {
847                 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
848                 ipv6_addr_copy(&final, &fl.fl6_dst);
849                 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
850                 final_p = &final;
851         }
852
853         if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
854                 fl.oif = np->mcast_oif;
855         security_sk_classify_flow(sk, &fl);
856
857         err = ip6_dst_lookup(sk, &dst, &fl);
858         if (err)
859                 goto out;
860         if (final_p)
861                 ipv6_addr_copy(&fl.fl6_dst, final_p);
862
863         if ((err = __xfrm_lookup(&dst, &fl, sk, XFRM_LOOKUP_WAIT)) < 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                 if (tclass < 0)
882                         tclass = 0;
883         }
884
885         if (msg->msg_flags&MSG_CONFIRM)
886                 goto do_confirm;
887
888 back_from_confirm:
889         if (inet->hdrincl) {
890                 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
891         } else {
892                 lock_sock(sk);
893                 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
894                         len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
895                         msg->msg_flags);
896
897                 if (err)
898                         ip6_flush_pending_frames(sk);
899                 else if (!(msg->msg_flags & MSG_MORE))
900                         err = rawv6_push_pending_frames(sk, &fl, rp);
901                 release_sock(sk);
902         }
903 done:
904         dst_release(dst);
905 out:
906         fl6_sock_release(flowlabel);
907         return err<0?err:len;
908 do_confirm:
909         dst_confirm(dst);
910         if (!(msg->msg_flags & MSG_PROBE) || len)
911                 goto back_from_confirm;
912         err = 0;
913         goto done;
914 }
915
916 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
917                                char __user *optval, int optlen)
918 {
919         switch (optname) {
920         case ICMPV6_FILTER:
921                 if (optlen > sizeof(struct icmp6_filter))
922                         optlen = sizeof(struct icmp6_filter);
923                 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
924                         return -EFAULT;
925                 return 0;
926         default:
927                 return -ENOPROTOOPT;
928         }
929
930         return 0;
931 }
932
933 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
934                                char __user *optval, int __user *optlen)
935 {
936         int len;
937
938         switch (optname) {
939         case ICMPV6_FILTER:
940                 if (get_user(len, optlen))
941                         return -EFAULT;
942                 if (len < 0)
943                         return -EINVAL;
944                 if (len > sizeof(struct icmp6_filter))
945                         len = sizeof(struct icmp6_filter);
946                 if (put_user(len, optlen))
947                         return -EFAULT;
948                 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
949                         return -EFAULT;
950                 return 0;
951         default:
952                 return -ENOPROTOOPT;
953         }
954
955         return 0;
956 }
957
958
959 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
960                             char __user *optval, int optlen)
961 {
962         struct raw6_sock *rp = raw6_sk(sk);
963         int val;
964
965         if (get_user(val, (int __user *)optval))
966                 return -EFAULT;
967
968         switch (optname) {
969                 case IPV6_CHECKSUM:
970                         if (inet_sk(sk)->num == IPPROTO_ICMPV6 &&
971                             level == IPPROTO_IPV6) {
972                                 /*
973                                  * RFC3542 tells that IPV6_CHECKSUM socket
974                                  * option in the IPPROTO_IPV6 level is not
975                                  * allowed on ICMPv6 sockets.
976                                  * If you want to set it, use IPPROTO_RAW
977                                  * level IPV6_CHECKSUM socket option
978                                  * (Linux extension).
979                                  */
980                                 return -EINVAL;
981                         }
982
983                         /* You may get strange result with a positive odd offset;
984                            RFC2292bis agrees with me. */
985                         if (val > 0 && (val&1))
986                                 return(-EINVAL);
987                         if (val < 0) {
988                                 rp->checksum = 0;
989                         } else {
990                                 rp->checksum = 1;
991                                 rp->offset = val;
992                         }
993
994                         return 0;
995                         break;
996
997                 default:
998                         return(-ENOPROTOOPT);
999         }
1000 }
1001
1002 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1003                           char __user *optval, int optlen)
1004 {
1005         switch(level) {
1006                 case SOL_RAW:
1007                         break;
1008
1009                 case SOL_ICMPV6:
1010                         if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1011                                 return -EOPNOTSUPP;
1012                         return rawv6_seticmpfilter(sk, level, optname, optval,
1013                                                    optlen);
1014                 case SOL_IPV6:
1015                         if (optname == IPV6_CHECKSUM)
1016                                 break;
1017                 default:
1018                         return ipv6_setsockopt(sk, level, optname, optval,
1019                                                optlen);
1020         }
1021
1022         return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1023 }
1024
1025 #ifdef CONFIG_COMPAT
1026 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
1027                                    char __user *optval, int optlen)
1028 {
1029         switch (level) {
1030         case SOL_RAW:
1031                 break;
1032         case SOL_ICMPV6:
1033                 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1034                         return -EOPNOTSUPP;
1035                 return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1036         case SOL_IPV6:
1037                 if (optname == IPV6_CHECKSUM)
1038                         break;
1039         default:
1040                 return compat_ipv6_setsockopt(sk, level, optname,
1041                                               optval, optlen);
1042         }
1043         return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1044 }
1045 #endif
1046
1047 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1048                             char __user *optval, int __user *optlen)
1049 {
1050         struct raw6_sock *rp = raw6_sk(sk);
1051         int val, len;
1052
1053         if (get_user(len,optlen))
1054                 return -EFAULT;
1055
1056         switch (optname) {
1057         case IPV6_CHECKSUM:
1058                 /*
1059                  * We allow getsockopt() for IPPROTO_IPV6-level
1060                  * IPV6_CHECKSUM socket option on ICMPv6 sockets
1061                  * since RFC3542 is silent about it.
1062                  */
1063                 if (rp->checksum == 0)
1064                         val = -1;
1065                 else
1066                         val = rp->offset;
1067                 break;
1068
1069         default:
1070                 return -ENOPROTOOPT;
1071         }
1072
1073         len = min_t(unsigned int, sizeof(int), len);
1074
1075         if (put_user(len, optlen))
1076                 return -EFAULT;
1077         if (copy_to_user(optval,&val,len))
1078                 return -EFAULT;
1079         return 0;
1080 }
1081
1082 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1083                           char __user *optval, int __user *optlen)
1084 {
1085         switch(level) {
1086                 case SOL_RAW:
1087                         break;
1088
1089                 case SOL_ICMPV6:
1090                         if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1091                                 return -EOPNOTSUPP;
1092                         return rawv6_geticmpfilter(sk, level, optname, optval,
1093                                                    optlen);
1094                 case SOL_IPV6:
1095                         if (optname == IPV6_CHECKSUM)
1096                                 break;
1097                 default:
1098                         return ipv6_getsockopt(sk, level, optname, optval,
1099                                                optlen);
1100         }
1101
1102         return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1103 }
1104
1105 #ifdef CONFIG_COMPAT
1106 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1107                                    char __user *optval, int __user *optlen)
1108 {
1109         switch (level) {
1110         case SOL_RAW:
1111                 break;
1112         case SOL_ICMPV6:
1113                 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1114                         return -EOPNOTSUPP;
1115                 return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1116         case SOL_IPV6:
1117                 if (optname == IPV6_CHECKSUM)
1118                         break;
1119         default:
1120                 return compat_ipv6_getsockopt(sk, level, optname,
1121                                               optval, optlen);
1122         }
1123         return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1124 }
1125 #endif
1126
1127 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1128 {
1129         switch(cmd) {
1130                 case SIOCOUTQ:
1131                 {
1132                         int amount = atomic_read(&sk->sk_wmem_alloc);
1133                         return put_user(amount, (int __user *)arg);
1134                 }
1135                 case SIOCINQ:
1136                 {
1137                         struct sk_buff *skb;
1138                         int amount = 0;
1139
1140                         spin_lock_bh(&sk->sk_receive_queue.lock);
1141                         skb = skb_peek(&sk->sk_receive_queue);
1142                         if (skb != NULL)
1143                                 amount = skb->tail - skb->transport_header;
1144                         spin_unlock_bh(&sk->sk_receive_queue.lock);
1145                         return put_user(amount, (int __user *)arg);
1146                 }
1147
1148                 default:
1149 #ifdef CONFIG_IPV6_MROUTE
1150                         return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1151 #else
1152                         return -ENOIOCTLCMD;
1153 #endif
1154         }
1155 }
1156
1157 static void rawv6_close(struct sock *sk, long timeout)
1158 {
1159         if (inet_sk(sk)->num == IPPROTO_RAW)
1160                 ip6_ra_control(sk, -1);
1161         ip6mr_sk_done(sk);
1162         sk_common_release(sk);
1163 }
1164
1165 static void raw6_destroy(struct sock *sk)
1166 {
1167         lock_sock(sk);
1168         ip6_flush_pending_frames(sk);
1169         release_sock(sk);
1170
1171         inet6_destroy_sock(sk);
1172 }
1173
1174 static int rawv6_init_sk(struct sock *sk)
1175 {
1176         struct raw6_sock *rp = raw6_sk(sk);
1177
1178         switch (inet_sk(sk)->num) {
1179         case IPPROTO_ICMPV6:
1180                 rp->checksum = 1;
1181                 rp->offset   = 2;
1182                 break;
1183         case IPPROTO_MH:
1184                 rp->checksum = 1;
1185                 rp->offset   = 4;
1186                 break;
1187         default:
1188                 break;
1189         }
1190         return(0);
1191 }
1192
1193 struct proto rawv6_prot = {
1194         .name              = "RAWv6",
1195         .owner             = THIS_MODULE,
1196         .close             = rawv6_close,
1197         .destroy           = raw6_destroy,
1198         .connect           = ip6_datagram_connect,
1199         .disconnect        = udp_disconnect,
1200         .ioctl             = rawv6_ioctl,
1201         .init              = rawv6_init_sk,
1202         .setsockopt        = rawv6_setsockopt,
1203         .getsockopt        = rawv6_getsockopt,
1204         .sendmsg           = rawv6_sendmsg,
1205         .recvmsg           = rawv6_recvmsg,
1206         .bind              = rawv6_bind,
1207         .backlog_rcv       = rawv6_rcv_skb,
1208         .hash              = raw_hash_sk,
1209         .unhash            = raw_unhash_sk,
1210         .obj_size          = sizeof(struct raw6_sock),
1211         .h.raw_hash        = &raw_v6_hashinfo,
1212 #ifdef CONFIG_COMPAT
1213         .compat_setsockopt = compat_rawv6_setsockopt,
1214         .compat_getsockopt = compat_rawv6_getsockopt,
1215 #endif
1216 };
1217
1218 #ifdef CONFIG_PROC_FS
1219 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1220 {
1221         struct ipv6_pinfo *np = inet6_sk(sp);
1222         struct in6_addr *dest, *src;
1223         __u16 destp, srcp;
1224
1225         dest  = &np->daddr;
1226         src   = &np->rcv_saddr;
1227         destp = 0;
1228         srcp  = inet_sk(sp)->num;
1229         seq_printf(seq,
1230                    "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1231                    "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d\n",
1232                    i,
1233                    src->s6_addr32[0], src->s6_addr32[1],
1234                    src->s6_addr32[2], src->s6_addr32[3], srcp,
1235                    dest->s6_addr32[0], dest->s6_addr32[1],
1236                    dest->s6_addr32[2], dest->s6_addr32[3], destp,
1237                    sp->sk_state,
1238                    atomic_read(&sp->sk_wmem_alloc),
1239                    atomic_read(&sp->sk_rmem_alloc),
1240                    0, 0L, 0,
1241                    sock_i_uid(sp), 0,
1242                    sock_i_ino(sp),
1243                    atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1244 }
1245
1246 static int raw6_seq_show(struct seq_file *seq, void *v)
1247 {
1248         if (v == SEQ_START_TOKEN)
1249                 seq_printf(seq,
1250                            "  sl  "
1251                            "local_address                         "
1252                            "remote_address                        "
1253                            "st tx_queue rx_queue tr tm->when retrnsmt"
1254                            "   uid  timeout inode ref pointer drops\n");
1255         else
1256                 raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1257         return 0;
1258 }
1259
1260 static const struct seq_operations raw6_seq_ops = {
1261         .start =        raw_seq_start,
1262         .next =         raw_seq_next,
1263         .stop =         raw_seq_stop,
1264         .show =         raw6_seq_show,
1265 };
1266
1267 static int raw6_seq_open(struct inode *inode, struct file *file)
1268 {
1269         return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1270 }
1271
1272 static const struct file_operations raw6_seq_fops = {
1273         .owner =        THIS_MODULE,
1274         .open =         raw6_seq_open,
1275         .read =         seq_read,
1276         .llseek =       seq_lseek,
1277         .release =      seq_release_net,
1278 };
1279
1280 static int raw6_init_net(struct net *net)
1281 {
1282         if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1283                 return -ENOMEM;
1284
1285         return 0;
1286 }
1287
1288 static void raw6_exit_net(struct net *net)
1289 {
1290         proc_net_remove(net, "raw6");
1291 }
1292
1293 static struct pernet_operations raw6_net_ops = {
1294         .init = raw6_init_net,
1295         .exit = raw6_exit_net,
1296 };
1297
1298 int __init raw6_proc_init(void)
1299 {
1300         return register_pernet_subsys(&raw6_net_ops);
1301 }
1302
1303 void raw6_proc_exit(void)
1304 {
1305         unregister_pernet_subsys(&raw6_net_ops);
1306 }
1307 #endif  /* CONFIG_PROC_FS */
1308
1309 /* Same as inet6_dgram_ops, sans udp_poll.  */
1310 static const struct proto_ops inet6_sockraw_ops = {
1311         .family            = PF_INET6,
1312         .owner             = THIS_MODULE,
1313         .release           = inet6_release,
1314         .bind              = inet6_bind,
1315         .connect           = inet_dgram_connect,        /* ok           */
1316         .socketpair        = sock_no_socketpair,        /* a do nothing */
1317         .accept            = sock_no_accept,            /* a do nothing */
1318         .getname           = inet6_getname,
1319         .poll              = datagram_poll,             /* ok           */
1320         .ioctl             = inet6_ioctl,               /* must change  */
1321         .listen            = sock_no_listen,            /* ok           */
1322         .shutdown          = inet_shutdown,             /* ok           */
1323         .setsockopt        = sock_common_setsockopt,    /* ok           */
1324         .getsockopt        = sock_common_getsockopt,    /* ok           */
1325         .sendmsg           = inet_sendmsg,              /* ok           */
1326         .recvmsg           = sock_common_recvmsg,       /* ok           */
1327         .mmap              = sock_no_mmap,
1328         .sendpage          = sock_no_sendpage,
1329 #ifdef CONFIG_COMPAT
1330         .compat_setsockopt = compat_sock_common_setsockopt,
1331         .compat_getsockopt = compat_sock_common_getsockopt,
1332 #endif
1333 };
1334
1335 static struct inet_protosw rawv6_protosw = {
1336         .type           = SOCK_RAW,
1337         .protocol       = IPPROTO_IP,   /* wild card */
1338         .prot           = &rawv6_prot,
1339         .ops            = &inet6_sockraw_ops,
1340         .capability     = CAP_NET_RAW,
1341         .no_check       = UDP_CSUM_DEFAULT,
1342         .flags          = INET_PROTOSW_REUSE,
1343 };
1344
1345 int __init rawv6_init(void)
1346 {
1347         int ret;
1348
1349         ret = inet6_register_protosw(&rawv6_protosw);
1350         if (ret)
1351                 goto out;
1352 out:
1353         return ret;
1354 }
1355
1356 void rawv6_exit(void)
1357 {
1358         inet6_unregister_protosw(&rawv6_protosw);
1359 }