Linux Kernel  3.7.1
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
raw.c
Go to the documentation of this file.
1 /*
2  * RAW sockets for IPv6
3  * Linux INET6 implementation
4  *
5  * Authors:
6  * Pedro Roque <[email protected]>
7  *
8  * Adapted from linux/net/ipv4/raw.c
9  *
10  * Fixes:
11  * Hideaki YOSHIFUJI : sin6_scope_id support
12  * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
13  * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
14  *
15  * This program is free software; you can redistribute it and/or
16  * modify it under the terms of the GNU General Public License
17  * as published by the Free Software Foundation; either version
18  * 2 of the License, or (at your option) any later version.
19  */
20 
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/slab.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmpv6.h>
31 #include <linux/netfilter.h>
32 #include <linux/netfilter_ipv6.h>
33 #include <linux/skbuff.h>
34 #include <linux/compat.h>
35 #include <asm/uaccess.h>
36 #include <asm/ioctls.h>
37 
38 #include <net/net_namespace.h>
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
42 
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
50 #include <net/udp.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
54 #include <net/mip6.h>
55 #endif
56 #include <linux/mroute6.h>
57 
58 #include <net/raw.h>
59 #include <net/rawv6.h>
60 #include <net/xfrm.h>
61 
62 #include <linux/proc_fs.h>
63 #include <linux/seq_file.h>
64 #include <linux/export.h>
65 
66 static struct raw_hashinfo raw_v6_hashinfo = {
67  .lock = __RW_LOCK_UNLOCKED(raw_v6_hashinfo.lock),
68 };
69 
70 static struct sock *__raw_v6_lookup(struct net *net, struct sock *sk,
71  unsigned short num, const struct in6_addr *loc_addr,
72  const struct in6_addr *rmt_addr, int dif)
73 {
74  struct hlist_node *node;
75  bool is_multicast = ipv6_addr_is_multicast(loc_addr);
76 
77  sk_for_each_from(sk, node)
78  if (inet_sk(sk)->inet_num == num) {
79  struct ipv6_pinfo *np = inet6_sk(sk);
80 
81  if (!net_eq(sock_net(sk), net))
82  continue;
83 
84  if (!ipv6_addr_any(&np->daddr) &&
85  !ipv6_addr_equal(&np->daddr, rmt_addr))
86  continue;
87 
88  if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
89  continue;
90 
91  if (!ipv6_addr_any(&np->rcv_saddr)) {
92  if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
93  goto found;
94  if (is_multicast &&
95  inet6_mc_check(sk, loc_addr, rmt_addr))
96  goto found;
97  continue;
98  }
99  goto found;
100  }
101  sk = NULL;
102 found:
103  return sk;
104 }
105 
106 /*
107  * 0 - deliver
108  * 1 - block
109  */
110 static int icmpv6_filter(const struct sock *sk, const struct sk_buff *skb)
111 {
112  struct icmp6hdr *_hdr;
113  const struct icmp6hdr *hdr;
114 
115  hdr = skb_header_pointer(skb, skb_transport_offset(skb),
116  sizeof(_hdr), &_hdr);
117  if (hdr) {
118  const __u32 *data = &raw6_sk(sk)->filter.data[0];
119  unsigned int type = hdr->icmp6_type;
120 
121  return (data[type >> 5] & (1U << (type & 31))) != 0;
122  }
123  return 1;
124 }
125 
126 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
127 typedef int mh_filter_t(struct sock *sock, struct sk_buff *skb);
128 
129 static mh_filter_t __rcu *mh_filter __read_mostly;
130 
131 int rawv6_mh_filter_register(mh_filter_t filter)
132 {
133  rcu_assign_pointer(mh_filter, filter);
134  return 0;
135 }
136 EXPORT_SYMBOL(rawv6_mh_filter_register);
137 
138 int rawv6_mh_filter_unregister(mh_filter_t filter)
139 {
140  RCU_INIT_POINTER(mh_filter, NULL);
141  synchronize_rcu();
142  return 0;
143 }
144 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
145 
146 #endif
147 
148 /*
149  * demultiplex raw sockets.
150  * (should consider queueing the skb in the sock receive_queue
151  * without calling rawv6.c)
152  *
153  * Caller owns SKB so we must make clones.
154  */
155 static bool ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
156 {
157  const struct in6_addr *saddr;
158  const struct in6_addr *daddr;
159  struct sock *sk;
160  bool delivered = false;
161  __u8 hash;
162  struct net *net;
163 
164  saddr = &ipv6_hdr(skb)->saddr;
165  daddr = saddr + 1;
166 
167  hash = nexthdr & (RAW_HTABLE_SIZE - 1);
168 
169  read_lock(&raw_v6_hashinfo.lock);
170  sk = sk_head(&raw_v6_hashinfo.ht[hash]);
171 
172  if (sk == NULL)
173  goto out;
174 
175  net = dev_net(skb->dev);
176  sk = __raw_v6_lookup(net, sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
177 
178  while (sk) {
179  int filtered;
180 
181  delivered = true;
182  switch (nexthdr) {
183  case IPPROTO_ICMPV6:
184  filtered = icmpv6_filter(sk, skb);
185  break;
186 
187 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
188  case IPPROTO_MH:
189  {
190  /* XXX: To validate MH only once for each packet,
191  * this is placed here. It should be after checking
192  * xfrm policy, however it doesn't. The checking xfrm
193  * policy is placed in rawv6_rcv() because it is
194  * required for each socket.
195  */
196  mh_filter_t *filter;
197 
198  filter = rcu_dereference(mh_filter);
199  filtered = filter ? (*filter)(sk, skb) : 0;
200  break;
201  }
202 #endif
203  default:
204  filtered = 0;
205  break;
206  }
207 
208  if (filtered < 0)
209  break;
210  if (filtered == 0) {
211  struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
212 
213  /* Not releasing hash table! */
214  if (clone) {
215  nf_reset(clone);
216  rawv6_rcv(sk, clone);
217  }
218  }
219  sk = __raw_v6_lookup(net, sk_next(sk), nexthdr, daddr, saddr,
220  IP6CB(skb)->iif);
221  }
222 out:
223  read_unlock(&raw_v6_hashinfo.lock);
224  return delivered;
225 }
226 
227 bool raw6_local_deliver(struct sk_buff *skb, int nexthdr)
228 {
229  struct sock *raw_sk;
230 
231  raw_sk = sk_head(&raw_v6_hashinfo.ht[nexthdr & (RAW_HTABLE_SIZE - 1)]);
232  if (raw_sk && !ipv6_raw_deliver(skb, nexthdr))
233  raw_sk = NULL;
234 
235  return raw_sk != NULL;
236 }
237 
238 /* This cleans up af_inet6 a bit. -DaveM */
239 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
240 {
241  struct inet_sock *inet = inet_sk(sk);
242  struct ipv6_pinfo *np = inet6_sk(sk);
243  struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
244  __be32 v4addr = 0;
245  int addr_type;
246  int err;
247 
248  if (addr_len < SIN6_LEN_RFC2133)
249  return -EINVAL;
250  addr_type = ipv6_addr_type(&addr->sin6_addr);
251 
252  /* Raw sockets are IPv6 only */
253  if (addr_type == IPV6_ADDR_MAPPED)
254  return -EADDRNOTAVAIL;
255 
256  lock_sock(sk);
257 
258  err = -EINVAL;
259  if (sk->sk_state != TCP_CLOSE)
260  goto out;
261 
262  rcu_read_lock();
263  /* Check if the address belongs to the host. */
264  if (addr_type != IPV6_ADDR_ANY) {
265  struct net_device *dev = NULL;
266 
267  if (addr_type & IPV6_ADDR_LINKLOCAL) {
268  if (addr_len >= sizeof(struct sockaddr_in6) &&
269  addr->sin6_scope_id) {
270  /* Override any existing binding, if another
271  * one is supplied by user.
272  */
273  sk->sk_bound_dev_if = addr->sin6_scope_id;
274  }
275 
276  /* Binding to link-local address requires an interface */
277  if (!sk->sk_bound_dev_if)
278  goto out_unlock;
279 
280  err = -ENODEV;
281  dev = dev_get_by_index_rcu(sock_net(sk),
282  sk->sk_bound_dev_if);
283  if (!dev)
284  goto out_unlock;
285  }
286 
287  /* ipv4 addr of the socket is invalid. Only the
288  * unspecified and mapped address have a v4 equivalent.
289  */
290  v4addr = LOOPBACK4_IPV6;
291  if (!(addr_type & IPV6_ADDR_MULTICAST)) {
292  err = -EADDRNOTAVAIL;
293  if (!ipv6_chk_addr(sock_net(sk), &addr->sin6_addr,
294  dev, 0)) {
295  goto out_unlock;
296  }
297  }
298  }
299 
300  inet->inet_rcv_saddr = inet->inet_saddr = v4addr;
301  np->rcv_saddr = addr->sin6_addr;
302  if (!(addr_type & IPV6_ADDR_MULTICAST))
303  np->saddr = addr->sin6_addr;
304  err = 0;
305 out_unlock:
306  rcu_read_unlock();
307 out:
308  release_sock(sk);
309  return err;
310 }
311 
312 static void rawv6_err(struct sock *sk, struct sk_buff *skb,
313  struct inet6_skb_parm *opt,
314  u8 type, u8 code, int offset, __be32 info)
315 {
316  struct inet_sock *inet = inet_sk(sk);
317  struct ipv6_pinfo *np = inet6_sk(sk);
318  int err;
319  int harderr;
320 
321  /* Report error on raw socket, if:
322  1. User requested recverr.
323  2. Socket is connected (otherwise the error indication
324  is useless without recverr and error is hard.
325  */
326  if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
327  return;
328 
329  harderr = icmpv6_err_convert(type, code, &err);
330  if (type == ICMPV6_PKT_TOOBIG) {
331  ip6_sk_update_pmtu(skb, sk, info);
332  harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
333  }
334  if (type == NDISC_REDIRECT)
335  ip6_sk_redirect(skb, sk);
336  if (np->recverr) {
337  u8 *payload = skb->data;
338  if (!inet->hdrincl)
339  payload += offset;
340  ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
341  }
342 
343  if (np->recverr || harderr) {
344  sk->sk_err = err;
345  sk->sk_error_report(sk);
346  }
347 }
348 
349 void raw6_icmp_error(struct sk_buff *skb, int nexthdr,
350  u8 type, u8 code, int inner_offset, __be32 info)
351 {
352  struct sock *sk;
353  int hash;
354  const struct in6_addr *saddr, *daddr;
355  struct net *net;
356 
357  hash = nexthdr & (RAW_HTABLE_SIZE - 1);
358 
359  read_lock(&raw_v6_hashinfo.lock);
360  sk = sk_head(&raw_v6_hashinfo.ht[hash]);
361  if (sk != NULL) {
362  /* Note: ipv6_hdr(skb) != skb->data */
363  const struct ipv6hdr *ip6h = (const struct ipv6hdr *)skb->data;
364  saddr = &ip6h->saddr;
365  daddr = &ip6h->daddr;
366  net = dev_net(skb->dev);
367 
368  while ((sk = __raw_v6_lookup(net, sk, nexthdr, saddr, daddr,
369  IP6CB(skb)->iif))) {
370  rawv6_err(sk, skb, NULL, type, code,
371  inner_offset, info);
372  sk = sk_next(sk);
373  }
374  }
375  read_unlock(&raw_v6_hashinfo.lock);
376 }
377 
378 static inline int rawv6_rcv_skb(struct sock *sk, struct sk_buff *skb)
379 {
380  if ((raw6_sk(sk)->checksum || rcu_access_pointer(sk->sk_filter)) &&
381  skb_checksum_complete(skb)) {
382  atomic_inc(&sk->sk_drops);
383  kfree_skb(skb);
384  return NET_RX_DROP;
385  }
386 
387  /* Charge it to the socket. */
388  skb_dst_drop(skb);
389  if (sock_queue_rcv_skb(sk, skb) < 0) {
390  kfree_skb(skb);
391  return NET_RX_DROP;
392  }
393 
394  return 0;
395 }
396 
397 /*
398  * This is next to useless...
399  * if we demultiplex in network layer we don't need the extra call
400  * just to queue the skb...
401  * maybe we could have the network decide upon a hint if it
402  * should call raw_rcv for demultiplexing
403  */
404 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
405 {
406  struct inet_sock *inet = inet_sk(sk);
407  struct raw6_sock *rp = raw6_sk(sk);
408 
409  if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
410  atomic_inc(&sk->sk_drops);
411  kfree_skb(skb);
412  return NET_RX_DROP;
413  }
414 
415  if (!rp->checksum)
417 
418  if (skb->ip_summed == CHECKSUM_COMPLETE) {
419  skb_postpull_rcsum(skb, skb_network_header(skb),
420  skb_network_header_len(skb));
421  if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
422  &ipv6_hdr(skb)->daddr,
423  skb->len, inet->inet_num, skb->csum))
425  }
426  if (!skb_csum_unnecessary(skb))
427  skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
428  &ipv6_hdr(skb)->daddr,
429  skb->len,
430  inet->inet_num, 0));
431 
432  if (inet->hdrincl) {
433  if (skb_checksum_complete(skb)) {
434  atomic_inc(&sk->sk_drops);
435  kfree_skb(skb);
436  return NET_RX_DROP;
437  }
438  }
439 
440  rawv6_rcv_skb(sk, skb);
441  return 0;
442 }
443 
444 
445 /*
446  * This should be easy, if there is something there
447  * we return it, otherwise we block.
448  */
449 
450 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
451  struct msghdr *msg, size_t len,
452  int noblock, int flags, int *addr_len)
453 {
454  struct ipv6_pinfo *np = inet6_sk(sk);
455  struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
456  struct sk_buff *skb;
457  size_t copied;
458  int err;
459 
460  if (flags & MSG_OOB)
461  return -EOPNOTSUPP;
462 
463  if (addr_len)
464  *addr_len=sizeof(*sin6);
465 
466  if (flags & MSG_ERRQUEUE)
467  return ipv6_recv_error(sk, msg, len);
468 
469  if (np->rxpmtu && np->rxopt.bits.rxpmtu)
470  return ipv6_recv_rxpmtu(sk, msg, len);
471 
472  skb = skb_recv_datagram(sk, flags, noblock, &err);
473  if (!skb)
474  goto out;
475 
476  copied = skb->len;
477  if (copied > len) {
478  copied = len;
479  msg->msg_flags |= MSG_TRUNC;
480  }
481 
482  if (skb_csum_unnecessary(skb)) {
483  err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
484  } else if (msg->msg_flags&MSG_TRUNC) {
485  if (__skb_checksum_complete(skb))
486  goto csum_copy_err;
487  err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
488  } else {
489  err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
490  if (err == -EINVAL)
491  goto csum_copy_err;
492  }
493  if (err)
494  goto out_free;
495 
496  /* Copy the address. */
497  if (sin6) {
498  sin6->sin6_family = AF_INET6;
499  sin6->sin6_port = 0;
500  sin6->sin6_addr = ipv6_hdr(skb)->saddr;
501  sin6->sin6_flowinfo = 0;
502  sin6->sin6_scope_id = 0;
503  if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
504  sin6->sin6_scope_id = IP6CB(skb)->iif;
505  }
506 
507  sock_recv_ts_and_drops(msg, sk, skb);
508 
509  if (np->rxopt.all)
510  datagram_recv_ctl(sk, msg, skb);
511 
512  err = copied;
513  if (flags & MSG_TRUNC)
514  err = skb->len;
515 
516 out_free:
517  skb_free_datagram(sk, skb);
518 out:
519  return err;
520 
521 csum_copy_err:
522  skb_kill_datagram(sk, skb, flags);
523 
524  /* Error for blocking case is chosen to masquerade
525  as some normal condition.
526  */
527  err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
528  goto out;
529 }
530 
531 static int rawv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
532  struct raw6_sock *rp)
533 {
534  struct sk_buff *skb;
535  int err = 0;
536  int offset;
537  int len;
538  int total_len;
539  __wsum tmp_csum;
540  __sum16 csum;
541 
542  if (!rp->checksum)
543  goto send;
544 
545  if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
546  goto out;
547 
548  offset = rp->offset;
549  total_len = inet_sk(sk)->cork.base.length;
550  if (offset >= total_len - 1) {
551  err = -EINVAL;
553  goto out;
554  }
555 
556  /* should be check HW csum miyazawa */
557  if (skb_queue_len(&sk->sk_write_queue) == 1) {
558  /*
559  * Only one fragment on the socket.
560  */
561  tmp_csum = skb->csum;
562  } else {
563  struct sk_buff *csum_skb = NULL;
564  tmp_csum = 0;
565 
566  skb_queue_walk(&sk->sk_write_queue, skb) {
567  tmp_csum = csum_add(tmp_csum, skb->csum);
568 
569  if (csum_skb)
570  continue;
571 
572  len = skb->len - skb_transport_offset(skb);
573  if (offset >= len) {
574  offset -= len;
575  continue;
576  }
577 
578  csum_skb = skb;
579  }
580 
581  skb = csum_skb;
582  }
583 
584  offset += skb_transport_offset(skb);
585  if (skb_copy_bits(skb, offset, &csum, 2))
586  BUG();
587 
588  /* in case cksum was not initialized */
589  if (unlikely(csum))
590  tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
591 
592  csum = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
593  total_len, fl6->flowi6_proto, tmp_csum);
594 
595  if (csum == 0 && fl6->flowi6_proto == IPPROTO_UDP)
596  csum = CSUM_MANGLED_0;
597 
598  if (skb_store_bits(skb, offset, &csum, 2))
599  BUG();
600 
601 send:
602  err = ip6_push_pending_frames(sk);
603 out:
604  return err;
605 }
606 
607 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
608  struct flowi6 *fl6, struct dst_entry **dstp,
609  unsigned int flags)
610 {
611  struct ipv6_pinfo *np = inet6_sk(sk);
612  struct ipv6hdr *iph;
613  struct sk_buff *skb;
614  int err;
615  struct rt6_info *rt = (struct rt6_info *)*dstp;
616  int hlen = LL_RESERVED_SPACE(rt->dst.dev);
617  int tlen = rt->dst.dev->needed_tailroom;
618 
619  if (length > rt->dst.dev->mtu) {
620  ipv6_local_error(sk, EMSGSIZE, fl6, rt->dst.dev->mtu);
621  return -EMSGSIZE;
622  }
623  if (flags&MSG_PROBE)
624  goto out;
625 
626  skb = sock_alloc_send_skb(sk,
627  length + hlen + tlen + 15,
628  flags & MSG_DONTWAIT, &err);
629  if (skb == NULL)
630  goto error;
631  skb_reserve(skb, hlen);
632 
633  skb->priority = sk->sk_priority;
634  skb->mark = sk->sk_mark;
635  skb_dst_set(skb, &rt->dst);
636  *dstp = NULL;
637 
638  skb_put(skb, length);
639  skb_reset_network_header(skb);
640  iph = ipv6_hdr(skb);
641 
642  skb->ip_summed = CHECKSUM_NONE;
643 
644  skb->transport_header = skb->network_header;
645  err = memcpy_fromiovecend((void *)iph, from, 0, length);
646  if (err)
647  goto error_fault;
648 
649  IP6_UPD_PO_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUT, skb->len);
651  rt->dst.dev, dst_output);
652  if (err > 0)
653  err = net_xmit_errno(err);
654  if (err)
655  goto error;
656 out:
657  return 0;
658 
659 error_fault:
660  err = -EFAULT;
661  kfree_skb(skb);
662 error:
663  IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
664  if (err == -ENOBUFS && !np->recverr)
665  err = 0;
666  return err;
667 }
668 
669 static int rawv6_probe_proto_opt(struct flowi6 *fl6, struct msghdr *msg)
670 {
671  struct iovec *iov;
672  u8 __user *type = NULL;
673  u8 __user *code = NULL;
674  u8 len = 0;
675  int probed = 0;
676  int i;
677 
678  if (!msg->msg_iov)
679  return 0;
680 
681  for (i = 0; i < msg->msg_iovlen; i++) {
682  iov = &msg->msg_iov[i];
683  if (!iov)
684  continue;
685 
686  switch (fl6->flowi6_proto) {
687  case IPPROTO_ICMPV6:
688  /* check if one-byte field is readable or not. */
689  if (iov->iov_base && iov->iov_len < 1)
690  break;
691 
692  if (!type) {
693  type = iov->iov_base;
694  /* check if code field is readable or not. */
695  if (iov->iov_len > 1)
696  code = type + 1;
697  } else if (!code)
698  code = iov->iov_base;
699 
700  if (type && code) {
701  if (get_user(fl6->fl6_icmp_type, type) ||
702  get_user(fl6->fl6_icmp_code, code))
703  return -EFAULT;
704  probed = 1;
705  }
706  break;
707  case IPPROTO_MH:
708  if (iov->iov_base && iov->iov_len < 1)
709  break;
710  /* check if type field is readable or not. */
711  if (iov->iov_len > 2 - len) {
712  u8 __user *p = iov->iov_base;
713  if (get_user(fl6->fl6_mh_type, &p[2 - len]))
714  return -EFAULT;
715  probed = 1;
716  } else
717  len += iov->iov_len;
718 
719  break;
720  default:
721  probed = 1;
722  break;
723  }
724  if (probed)
725  break;
726  }
727  return 0;
728 }
729 
730 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
731  struct msghdr *msg, size_t len)
732 {
733  struct ipv6_txoptions opt_space;
734  struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
735  struct in6_addr *daddr, *final_p, final;
736  struct inet_sock *inet = inet_sk(sk);
737  struct ipv6_pinfo *np = inet6_sk(sk);
738  struct raw6_sock *rp = raw6_sk(sk);
739  struct ipv6_txoptions *opt = NULL;
740  struct ip6_flowlabel *flowlabel = NULL;
741  struct dst_entry *dst = NULL;
742  struct flowi6 fl6;
743  int addr_len = msg->msg_namelen;
744  int hlimit = -1;
745  int tclass = -1;
746  int dontfrag = -1;
747  u16 proto;
748  int err;
749 
750  /* Rough check on arithmetic overflow,
751  better check is made in ip6_append_data().
752  */
753  if (len > INT_MAX)
754  return -EMSGSIZE;
755 
756  /* Mirror BSD error message compatibility */
757  if (msg->msg_flags & MSG_OOB)
758  return -EOPNOTSUPP;
759 
760  /*
761  * Get and verify the address.
762  */
763  memset(&fl6, 0, sizeof(fl6));
764 
765  fl6.flowi6_mark = sk->sk_mark;
766 
767  if (sin6) {
768  if (addr_len < SIN6_LEN_RFC2133)
769  return -EINVAL;
770 
771  if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
772  return -EAFNOSUPPORT;
773 
774  /* port is the proto value [0..255] carried in nexthdr */
775  proto = ntohs(sin6->sin6_port);
776 
777  if (!proto)
778  proto = inet->inet_num;
779  else if (proto != inet->inet_num)
780  return -EINVAL;
781 
782  if (proto > 255)
783  return -EINVAL;
784 
785  daddr = &sin6->sin6_addr;
786  if (np->sndflow) {
788  if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
789  flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
790  if (flowlabel == NULL)
791  return -EINVAL;
792  daddr = &flowlabel->dst;
793  }
794  }
795 
796  /*
797  * Otherwise it will be difficult to maintain
798  * sk->sk_dst_cache.
799  */
800  if (sk->sk_state == TCP_ESTABLISHED &&
801  ipv6_addr_equal(daddr, &np->daddr))
802  daddr = &np->daddr;
803 
804  if (addr_len >= sizeof(struct sockaddr_in6) &&
805  sin6->sin6_scope_id &&
806  ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
807  fl6.flowi6_oif = sin6->sin6_scope_id;
808  } else {
809  if (sk->sk_state != TCP_ESTABLISHED)
810  return -EDESTADDRREQ;
811 
812  proto = inet->inet_num;
813  daddr = &np->daddr;
814  fl6.flowlabel = np->flow_label;
815  }
816 
817  if (fl6.flowi6_oif == 0)
818  fl6.flowi6_oif = sk->sk_bound_dev_if;
819 
820  if (msg->msg_controllen) {
821  opt = &opt_space;
822  memset(opt, 0, sizeof(struct ipv6_txoptions));
823  opt->tot_len = sizeof(struct ipv6_txoptions);
824 
825  err = datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
826  &hlimit, &tclass, &dontfrag);
827  if (err < 0) {
828  fl6_sock_release(flowlabel);
829  return err;
830  }
831  if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
832  flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
833  if (flowlabel == NULL)
834  return -EINVAL;
835  }
836  if (!(opt->opt_nflen|opt->opt_flen))
837  opt = NULL;
838  }
839  if (opt == NULL)
840  opt = np->opt;
841  if (flowlabel)
842  opt = fl6_merge_options(&opt_space, flowlabel, opt);
843  opt = ipv6_fixup_options(&opt_space, opt);
844 
845  fl6.flowi6_proto = proto;
846  err = rawv6_probe_proto_opt(&fl6, msg);
847  if (err)
848  goto out;
849 
850  if (!ipv6_addr_any(daddr))
851  fl6.daddr = *daddr;
852  else
853  fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
854  if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
855  fl6.saddr = np->saddr;
856 
857  final_p = fl6_update_dst(&fl6, opt, &final);
858 
859  if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr))
860  fl6.flowi6_oif = np->mcast_oif;
861  else if (!fl6.flowi6_oif)
862  fl6.flowi6_oif = np->ucast_oif;
863  security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
864 
865  dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true);
866  if (IS_ERR(dst)) {
867  err = PTR_ERR(dst);
868  goto out;
869  }
870  if (hlimit < 0) {
871  if (ipv6_addr_is_multicast(&fl6.daddr))
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 (dontfrag < 0)
883  dontfrag = np->dontfrag;
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, &fl6, &dst, msg->msg_flags);
891  else {
892  lock_sock(sk);
894  len, 0, hlimit, tclass, opt, &fl6, (struct rt6_info*)dst,
895  msg->msg_flags, dontfrag);
896 
897  if (err)
899  else if (!(msg->msg_flags & MSG_MORE))
900  err = rawv6_push_pending_frames(sk, &fl6, 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, unsigned 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)->inet_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 
996  default:
997  return -ENOPROTOOPT;
998  }
999 }
1000 
1001 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
1002  char __user *optval, unsigned int optlen)
1003 {
1004  switch (level) {
1005  case SOL_RAW:
1006  break;
1007 
1008  case SOL_ICMPV6:
1009  if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1010  return -EOPNOTSUPP;
1011  return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1012  case SOL_IPV6:
1013  if (optname == IPV6_CHECKSUM)
1014  break;
1015  default:
1016  return ipv6_setsockopt(sk, level, optname, optval, 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, optlen);
1090  case SOL_IPV6:
1091  if (optname == IPV6_CHECKSUM)
1092  break;
1093  default:
1094  return ipv6_getsockopt(sk, level, optname, optval, optlen);
1095  }
1096 
1097  return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1098 }
1099 
1100 #ifdef CONFIG_COMPAT
1101 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1102  char __user *optval, int __user *optlen)
1103 {
1104  switch (level) {
1105  case SOL_RAW:
1106  break;
1107  case SOL_ICMPV6:
1108  if (inet_sk(sk)->inet_num != IPPROTO_ICMPV6)
1109  return -EOPNOTSUPP;
1110  return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1111  case SOL_IPV6:
1112  if (optname == IPV6_CHECKSUM)
1113  break;
1114  default:
1115  return compat_ipv6_getsockopt(sk, level, optname,
1116  optval, optlen);
1117  }
1118  return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1119 }
1120 #endif
1121 
1122 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1123 {
1124  switch (cmd) {
1125  case SIOCOUTQ: {
1126  int amount = sk_wmem_alloc_get(sk);
1127 
1128  return put_user(amount, (int __user *)arg);
1129  }
1130  case SIOCINQ: {
1131  struct sk_buff *skb;
1132  int amount = 0;
1133 
1134  spin_lock_bh(&sk->sk_receive_queue.lock);
1135  skb = skb_peek(&sk->sk_receive_queue);
1136  if (skb != NULL)
1137  amount = skb->tail - skb->transport_header;
1138  spin_unlock_bh(&sk->sk_receive_queue.lock);
1139  return put_user(amount, (int __user *)arg);
1140  }
1141 
1142  default:
1143 #ifdef CONFIG_IPV6_MROUTE
1144  return ip6mr_ioctl(sk, cmd, (void __user *)arg);
1145 #else
1146  return -ENOIOCTLCMD;
1147 #endif
1148  }
1149 }
1150 
1151 #ifdef CONFIG_COMPAT
1152 static int compat_rawv6_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
1153 {
1154  switch (cmd) {
1155  case SIOCOUTQ:
1156  case SIOCINQ:
1157  return -ENOIOCTLCMD;
1158  default:
1159 #ifdef CONFIG_IPV6_MROUTE
1160  return ip6mr_compat_ioctl(sk, cmd, compat_ptr(arg));
1161 #else
1162  return -ENOIOCTLCMD;
1163 #endif
1164  }
1165 }
1166 #endif
1167 
1168 static void rawv6_close(struct sock *sk, long timeout)
1169 {
1170  if (inet_sk(sk)->inet_num == IPPROTO_RAW)
1171  ip6_ra_control(sk, -1);
1172  ip6mr_sk_done(sk);
1173  sk_common_release(sk);
1174 }
1175 
1176 static void raw6_destroy(struct sock *sk)
1177 {
1178  lock_sock(sk);
1180  release_sock(sk);
1181 
1182  inet6_destroy_sock(sk);
1183 }
1184 
1185 static int rawv6_init_sk(struct sock *sk)
1186 {
1187  struct raw6_sock *rp = raw6_sk(sk);
1188 
1189  switch (inet_sk(sk)->inet_num) {
1190  case IPPROTO_ICMPV6:
1191  rp->checksum = 1;
1192  rp->offset = 2;
1193  break;
1194  case IPPROTO_MH:
1195  rp->checksum = 1;
1196  rp->offset = 4;
1197  break;
1198  default:
1199  break;
1200  }
1201  return 0;
1202 }
1203 
1204 struct proto rawv6_prot = {
1205  .name = "RAWv6",
1206  .owner = THIS_MODULE,
1207  .close = rawv6_close,
1208  .destroy = raw6_destroy,
1209  .connect = ip6_datagram_connect,
1210  .disconnect = udp_disconnect,
1211  .ioctl = rawv6_ioctl,
1212  .init = rawv6_init_sk,
1213  .setsockopt = rawv6_setsockopt,
1214  .getsockopt = rawv6_getsockopt,
1215  .sendmsg = rawv6_sendmsg,
1216  .recvmsg = rawv6_recvmsg,
1217  .bind = rawv6_bind,
1218  .backlog_rcv = rawv6_rcv_skb,
1219  .hash = raw_hash_sk,
1220  .unhash = raw_unhash_sk,
1221  .obj_size = sizeof(struct raw6_sock),
1222  .h.raw_hash = &raw_v6_hashinfo,
1223 #ifdef CONFIG_COMPAT
1224  .compat_setsockopt = compat_rawv6_setsockopt,
1225  .compat_getsockopt = compat_rawv6_getsockopt,
1226  .compat_ioctl = compat_rawv6_ioctl,
1227 #endif
1228 };
1229 
1230 #ifdef CONFIG_PROC_FS
1231 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1232 {
1233  struct ipv6_pinfo *np = inet6_sk(sp);
1234  const struct in6_addr *dest, *src;
1235  __u16 destp, srcp;
1236 
1237  dest = &np->daddr;
1238  src = &np->rcv_saddr;
1239  destp = 0;
1240  srcp = inet_sk(sp)->inet_num;
1241  seq_printf(seq,
1242  "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1243  "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1244  i,
1245  src->s6_addr32[0], src->s6_addr32[1],
1246  src->s6_addr32[2], src->s6_addr32[3], srcp,
1247  dest->s6_addr32[0], dest->s6_addr32[1],
1248  dest->s6_addr32[2], dest->s6_addr32[3], destp,
1249  sp->sk_state,
1250  sk_wmem_alloc_get(sp),
1251  sk_rmem_alloc_get(sp),
1252  0, 0L, 0,
1253  from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1254  0,
1255  sock_i_ino(sp),
1256  atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1257 }
1258 
1259 static int raw6_seq_show(struct seq_file *seq, void *v)
1260 {
1261  if (v == SEQ_START_TOKEN)
1262  seq_printf(seq,
1263  " sl "
1264  "local_address "
1265  "remote_address "
1266  "st tx_queue rx_queue tr tm->when retrnsmt"
1267  " uid timeout inode ref pointer drops\n");
1268  else
1269  raw6_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1270  return 0;
1271 }
1272 
1273 static const struct seq_operations raw6_seq_ops = {
1274  .start = raw_seq_start,
1275  .next = raw_seq_next,
1276  .stop = raw_seq_stop,
1277  .show = raw6_seq_show,
1278 };
1279 
1280 static int raw6_seq_open(struct inode *inode, struct file *file)
1281 {
1282  return raw_seq_open(inode, file, &raw_v6_hashinfo, &raw6_seq_ops);
1283 }
1284 
1285 static const struct file_operations raw6_seq_fops = {
1286  .owner = THIS_MODULE,
1287  .open = raw6_seq_open,
1288  .read = seq_read,
1289  .llseek = seq_lseek,
1290  .release = seq_release_net,
1291 };
1292 
1293 static int __net_init raw6_init_net(struct net *net)
1294 {
1295  if (!proc_net_fops_create(net, "raw6", S_IRUGO, &raw6_seq_fops))
1296  return -ENOMEM;
1297 
1298  return 0;
1299 }
1300 
1301 static void __net_exit raw6_exit_net(struct net *net)
1302 {
1303  proc_net_remove(net, "raw6");
1304 }
1305 
1306 static struct pernet_operations raw6_net_ops = {
1307  .init = raw6_init_net,
1308  .exit = raw6_exit_net,
1309 };
1310 
1311 int __init raw6_proc_init(void)
1312 {
1313  return register_pernet_subsys(&raw6_net_ops);
1314 }
1315 
1316 void raw6_proc_exit(void)
1317 {
1318  unregister_pernet_subsys(&raw6_net_ops);
1319 }
1320 #endif /* CONFIG_PROC_FS */
1321 
1322 /* Same as inet6_dgram_ops, sans udp_poll. */
1323 static const struct proto_ops inet6_sockraw_ops = {
1324  .family = PF_INET6,
1325  .owner = THIS_MODULE,
1326  .release = inet6_release,
1327  .bind = inet6_bind,
1328  .connect = inet_dgram_connect, /* ok */
1329  .socketpair = sock_no_socketpair, /* a do nothing */
1330  .accept = sock_no_accept, /* a do nothing */
1331  .getname = inet6_getname,
1332  .poll = datagram_poll, /* ok */
1333  .ioctl = inet6_ioctl, /* must change */
1334  .listen = sock_no_listen, /* ok */
1335  .shutdown = inet_shutdown, /* ok */
1336  .setsockopt = sock_common_setsockopt, /* ok */
1337  .getsockopt = sock_common_getsockopt, /* ok */
1338  .sendmsg = inet_sendmsg, /* ok */
1339  .recvmsg = sock_common_recvmsg, /* ok */
1340  .mmap = sock_no_mmap,
1341  .sendpage = sock_no_sendpage,
1342 #ifdef CONFIG_COMPAT
1343  .compat_setsockopt = compat_sock_common_setsockopt,
1344  .compat_getsockopt = compat_sock_common_getsockopt,
1345 #endif
1346 };
1347 
1348 static struct inet_protosw rawv6_protosw = {
1349  .type = SOCK_RAW,
1350  .protocol = IPPROTO_IP, /* wild card */
1351  .prot = &rawv6_prot,
1352  .ops = &inet6_sockraw_ops,
1353  .no_check = UDP_CSUM_DEFAULT,
1354  .flags = INET_PROTOSW_REUSE,
1355 };
1356 
1358 {
1359  int ret;
1360 
1361  ret = inet6_register_protosw(&rawv6_protosw);
1362  if (ret)
1363  goto out;
1364 out:
1365  return ret;
1366 }
1367 
1368 void rawv6_exit(void)
1369 {
1370  inet6_unregister_protosw(&rawv6_protosw);
1371 }