blob: a082646e6f165ec24256aa0eb33e044ec74fa53f [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * RAW sockets for IPv6
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 *
8 * Adapted from linux/net/ipv4/raw.c
9 *
10 * $Id: raw.c,v 1.51 2002/02/01 22:01:04 davem Exp $
11 *
12 * Fixes:
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * YOSHIFUJI,H.@USAGI : raw checksum (RFC2292(bis) compliance)
15 * Kazunori MIYAZAWA @USAGI: change process style to use ip6_append_data
16 *
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
21 */
22
23#include <linux/errno.h>
24#include <linux/types.h>
25#include <linux/socket.h>
26#include <linux/sockios.h>
27#include <linux/sched.h>
28#include <linux/net.h>
29#include <linux/in6.h>
30#include <linux/netdevice.h>
31#include <linux/if_arp.h>
32#include <linux/icmpv6.h>
33#include <linux/netfilter.h>
34#include <linux/netfilter_ipv6.h>
35#include <asm/uaccess.h>
36#include <asm/ioctls.h>
Herbert Xu357b40a2005-04-19 22:30:14 -070037#include <asm/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
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
53#include <net/rawv6.h>
54#include <net/xfrm.h>
55
56#include <linux/proc_fs.h>
57#include <linux/seq_file.h>
58
59struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
60DEFINE_RWLOCK(raw_v6_lock);
61
62static void raw_v6_hash(struct sock *sk)
63{
64 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
65 (RAWV6_HTABLE_SIZE - 1)];
66
67 write_lock_bh(&raw_v6_lock);
68 sk_add_node(sk, list);
69 sock_prot_inc_use(sk->sk_prot);
70 write_unlock_bh(&raw_v6_lock);
71}
72
73static void raw_v6_unhash(struct sock *sk)
74{
75 write_lock_bh(&raw_v6_lock);
76 if (sk_del_node_init(sk))
77 sock_prot_dec_use(sk->sk_prot);
78 write_unlock_bh(&raw_v6_lock);
79}
80
81
82/* Grumble... icmp and ip_input want to get at this... */
83struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
Andrew McDonald0bd1b592005-08-09 19:44:42 -070084 struct in6_addr *loc_addr, struct in6_addr *rmt_addr,
85 int dif)
Linus Torvalds1da177e2005-04-16 15:20:36 -070086{
87 struct hlist_node *node;
88 int is_multicast = ipv6_addr_is_multicast(loc_addr);
89
90 sk_for_each_from(sk, node)
91 if (inet_sk(sk)->num == num) {
92 struct ipv6_pinfo *np = inet6_sk(sk);
93
94 if (!ipv6_addr_any(&np->daddr) &&
95 !ipv6_addr_equal(&np->daddr, rmt_addr))
96 continue;
97
Andrew McDonald0bd1b592005-08-09 19:44:42 -070098 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
99 continue;
100
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 if (!ipv6_addr_any(&np->rcv_saddr)) {
102 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
103 goto found;
104 if (is_multicast &&
105 inet6_mc_check(sk, loc_addr, rmt_addr))
106 goto found;
107 continue;
108 }
109 goto found;
110 }
111 sk = NULL;
112found:
113 return sk;
114}
115
116/*
117 * 0 - deliver
118 * 1 - block
119 */
120static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
121{
122 struct icmp6hdr *icmph;
123 struct raw6_sock *rp = raw6_sk(sk);
124
125 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
126 __u32 *data = &rp->filter.data[0];
127 int bit_nr;
128
129 icmph = (struct icmp6hdr *) skb->data;
130 bit_nr = icmph->icmp6_type;
131
132 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
133 }
134 return 0;
135}
136
137/*
138 * demultiplex raw sockets.
139 * (should consider queueing the skb in the sock receive_queue
140 * without calling rawv6.c)
141 *
142 * Caller owns SKB so we must make clones.
143 */
Patrick McHardyd13964f2005-08-09 19:45:02 -0700144int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145{
146 struct in6_addr *saddr;
147 struct in6_addr *daddr;
148 struct sock *sk;
Patrick McHardyd13964f2005-08-09 19:45:02 -0700149 int delivered = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 __u8 hash;
151
152 saddr = &skb->nh.ipv6h->saddr;
153 daddr = saddr + 1;
154
155 hash = nexthdr & (MAX_INET_PROTOS - 1);
156
157 read_lock(&raw_v6_lock);
158 sk = sk_head(&raw_v6_htable[hash]);
159
160 /*
161 * The first socket found will be delivered after
162 * delivery to transport protocols.
163 */
164
165 if (sk == NULL)
166 goto out;
167
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700168 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, skb->dev->ifindex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169
170 while (sk) {
Patrick McHardyd13964f2005-08-09 19:45:02 -0700171 delivered = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 if (nexthdr != IPPROTO_ICMPV6 || !icmpv6_filter(sk, skb)) {
173 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
174
175 /* Not releasing hash table! */
176 if (clone)
177 rawv6_rcv(sk, clone);
178 }
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700179 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
180 skb->dev->ifindex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 }
182out:
183 read_unlock(&raw_v6_lock);
Patrick McHardyd13964f2005-08-09 19:45:02 -0700184 return delivered;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185}
186
187/* This cleans up af_inet6 a bit. -DaveM */
188static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
189{
190 struct inet_sock *inet = inet_sk(sk);
191 struct ipv6_pinfo *np = inet6_sk(sk);
192 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
193 __u32 v4addr = 0;
194 int addr_type;
195 int err;
196
197 if (addr_len < SIN6_LEN_RFC2133)
198 return -EINVAL;
199 addr_type = ipv6_addr_type(&addr->sin6_addr);
200
201 /* Raw sockets are IPv6 only */
202 if (addr_type == IPV6_ADDR_MAPPED)
203 return(-EADDRNOTAVAIL);
204
205 lock_sock(sk);
206
207 err = -EINVAL;
208 if (sk->sk_state != TCP_CLOSE)
209 goto out;
210
211 /* Check if the address belongs to the host. */
212 if (addr_type != IPV6_ADDR_ANY) {
213 struct net_device *dev = NULL;
214
215 if (addr_type & IPV6_ADDR_LINKLOCAL) {
216 if (addr_len >= sizeof(struct sockaddr_in6) &&
217 addr->sin6_scope_id) {
218 /* Override any existing binding, if another
219 * one is supplied by user.
220 */
221 sk->sk_bound_dev_if = addr->sin6_scope_id;
222 }
223
224 /* Binding to link-local address requires an interface */
225 if (!sk->sk_bound_dev_if)
226 goto out;
227
228 dev = dev_get_by_index(sk->sk_bound_dev_if);
229 if (!dev) {
230 err = -ENODEV;
231 goto out;
232 }
233 }
234
235 /* ipv4 addr of the socket is invalid. Only the
236 * unspecified and mapped address have a v4 equivalent.
237 */
238 v4addr = LOOPBACK4_IPV6;
239 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
240 err = -EADDRNOTAVAIL;
241 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
242 if (dev)
243 dev_put(dev);
244 goto out;
245 }
246 }
247 if (dev)
248 dev_put(dev);
249 }
250
251 inet->rcv_saddr = inet->saddr = v4addr;
252 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
253 if (!(addr_type & IPV6_ADDR_MULTICAST))
254 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
255 err = 0;
256out:
257 release_sock(sk);
258 return err;
259}
260
261void rawv6_err(struct sock *sk, struct sk_buff *skb,
262 struct inet6_skb_parm *opt,
263 int type, int code, int offset, u32 info)
264{
265 struct inet_sock *inet = inet_sk(sk);
266 struct ipv6_pinfo *np = inet6_sk(sk);
267 int err;
268 int harderr;
269
270 /* Report error on raw socket, if:
271 1. User requested recverr.
272 2. Socket is connected (otherwise the error indication
273 is useless without recverr and error is hard.
274 */
275 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
276 return;
277
278 harderr = icmpv6_err_convert(type, code, &err);
279 if (type == ICMPV6_PKT_TOOBIG)
280 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
281
282 if (np->recverr) {
283 u8 *payload = skb->data;
284 if (!inet->hdrincl)
285 payload += offset;
286 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
287 }
288
289 if (np->recverr || harderr) {
290 sk->sk_err = err;
291 sk->sk_error_report(sk);
292 }
293}
294
295static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
296{
297 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
298 skb->ip_summed != CHECKSUM_UNNECESSARY) {
299 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
300 /* FIXME: increment a raw6 drops counter here */
301 kfree_skb(skb);
302 return 0;
303 }
304 skb->ip_summed = CHECKSUM_UNNECESSARY;
305 }
306
307 /* Charge it to the socket. */
308 if (sock_queue_rcv_skb(sk,skb)<0) {
309 /* FIXME: increment a raw6 drops counter here */
310 kfree_skb(skb);
311 return 0;
312 }
313
314 return 0;
315}
316
317/*
318 * This is next to useless...
319 * if we demultiplex in network layer we don't need the extra call
320 * just to queue the skb...
321 * maybe we could have the network decide upon a hint if it
322 * should call raw_rcv for demultiplexing
323 */
324int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
325{
326 struct inet_sock *inet = inet_sk(sk);
327 struct raw6_sock *rp = raw6_sk(sk);
328
329 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
330 kfree_skb(skb);
331 return NET_RX_DROP;
332 }
333
334 if (!rp->checksum)
335 skb->ip_summed = CHECKSUM_UNNECESSARY;
336
337 if (skb->ip_summed != CHECKSUM_UNNECESSARY) {
338 if (skb->ip_summed == CHECKSUM_HW) {
Patrick McHardy793245e2005-08-16 20:39:38 -0700339 skb_postpull_rcsum(skb, skb->nh.raw,
340 skb->h.raw - skb->nh.raw);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 skb->ip_summed = CHECKSUM_UNNECESSARY;
342 if (csum_ipv6_magic(&skb->nh.ipv6h->saddr,
343 &skb->nh.ipv6h->daddr,
344 skb->len, inet->num, skb->csum)) {
345 LIMIT_NETDEBUG(
346 printk(KERN_DEBUG "raw v6 hw csum failure.\n"));
347 skb->ip_summed = CHECKSUM_NONE;
348 }
349 }
350 if (skb->ip_summed == CHECKSUM_NONE)
351 skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
352 &skb->nh.ipv6h->daddr,
353 skb->len, inet->num, 0);
354 }
355
356 if (inet->hdrincl) {
357 if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
358 (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
359 /* FIXME: increment a raw6 drops counter here */
360 kfree_skb(skb);
361 return 0;
362 }
363 skb->ip_summed = CHECKSUM_UNNECESSARY;
364 }
365
366 rawv6_rcv_skb(sk, skb);
367 return 0;
368}
369
370
371/*
372 * This should be easy, if there is something there
373 * we return it, otherwise we block.
374 */
375
376static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
377 struct msghdr *msg, size_t len,
378 int noblock, int flags, int *addr_len)
379{
380 struct ipv6_pinfo *np = inet6_sk(sk);
381 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
382 struct sk_buff *skb;
383 size_t copied;
384 int err;
385
386 if (flags & MSG_OOB)
387 return -EOPNOTSUPP;
388
389 if (addr_len)
390 *addr_len=sizeof(*sin6);
391
392 if (flags & MSG_ERRQUEUE)
393 return ipv6_recv_error(sk, msg, len);
394
395 skb = skb_recv_datagram(sk, flags, noblock, &err);
396 if (!skb)
397 goto out;
398
399 copied = skb->len;
400 if (copied > len) {
401 copied = len;
402 msg->msg_flags |= MSG_TRUNC;
403 }
404
405 if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
406 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
407 } else if (msg->msg_flags&MSG_TRUNC) {
408 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum)))
409 goto csum_copy_err;
410 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
411 } else {
412 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
413 if (err == -EINVAL)
414 goto csum_copy_err;
415 }
416 if (err)
417 goto out_free;
418
419 /* Copy the address. */
420 if (sin6) {
421 sin6->sin6_family = AF_INET6;
422 ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
423 sin6->sin6_flowinfo = 0;
424 sin6->sin6_scope_id = 0;
425 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
426 sin6->sin6_scope_id = IP6CB(skb)->iif;
427 }
428
429 sock_recv_timestamp(msg, sk, skb);
430
431 if (np->rxopt.all)
432 datagram_recv_ctl(sk, msg, skb);
433
434 err = copied;
435 if (flags & MSG_TRUNC)
436 err = skb->len;
437
438out_free:
439 skb_free_datagram(sk, skb);
440out:
441 return err;
442
443csum_copy_err:
444 /* Clear queue. */
445 if (flags&MSG_PEEK) {
446 int clear = 0;
Herbert Xue0f9f852005-06-18 22:56:18 -0700447 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 if (skb == skb_peek(&sk->sk_receive_queue)) {
449 __skb_unlink(skb, &sk->sk_receive_queue);
450 clear = 1;
451 }
Herbert Xue0f9f852005-06-18 22:56:18 -0700452 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 if (clear)
454 kfree_skb(skb);
455 }
456
457 /* Error for blocking case is chosen to masquerade
458 as some normal condition.
459 */
460 err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
461 /* FIXME: increment a raw6 drops counter here */
462 goto out_free;
463}
464
465static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
Herbert Xu357b40a2005-04-19 22:30:14 -0700466 struct raw6_sock *rp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467{
468 struct sk_buff *skb;
469 int err = 0;
Herbert Xu357b40a2005-04-19 22:30:14 -0700470 int offset;
471 int len;
Herbert Xu679a8732005-05-03 14:24:36 -0700472 int total_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 u32 tmp_csum;
Herbert Xu357b40a2005-04-19 22:30:14 -0700474 u16 csum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475
476 if (!rp->checksum)
477 goto send;
478
479 if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
480 goto out;
481
Herbert Xu357b40a2005-04-19 22:30:14 -0700482 offset = rp->offset;
Herbert Xu679a8732005-05-03 14:24:36 -0700483 total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data);
484 if (offset >= total_len - 1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 err = -EINVAL;
Herbert Xu357b40a2005-04-19 22:30:14 -0700486 ip6_flush_pending_frames(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 goto out;
488 }
489
490 /* should be check HW csum miyazawa */
491 if (skb_queue_len(&sk->sk_write_queue) == 1) {
492 /*
493 * Only one fragment on the socket.
494 */
495 tmp_csum = skb->csum;
496 } else {
Herbert Xu357b40a2005-04-19 22:30:14 -0700497 struct sk_buff *csum_skb = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 tmp_csum = 0;
499
500 skb_queue_walk(&sk->sk_write_queue, skb) {
501 tmp_csum = csum_add(tmp_csum, skb->csum);
Herbert Xu357b40a2005-04-19 22:30:14 -0700502
503 if (csum_skb)
504 continue;
505
506 len = skb->len - (skb->h.raw - skb->data);
507 if (offset >= len) {
508 offset -= len;
509 continue;
510 }
511
512 csum_skb = skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 }
Herbert Xu357b40a2005-04-19 22:30:14 -0700514
515 skb = csum_skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 }
517
Herbert Xu357b40a2005-04-19 22:30:14 -0700518 offset += skb->h.raw - skb->data;
519 if (skb_copy_bits(skb, offset, &csum, 2))
520 BUG();
521
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 /* in case cksum was not initialized */
Herbert Xu357b40a2005-04-19 22:30:14 -0700523 if (unlikely(csum))
524 tmp_csum = csum_sub(tmp_csum, csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525
Herbert Xu357b40a2005-04-19 22:30:14 -0700526 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
527 &fl->fl6_dst,
Herbert Xu679a8732005-05-03 14:24:36 -0700528 total_len, fl->proto, tmp_csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
Herbert Xu357b40a2005-04-19 22:30:14 -0700530 if (tmp_csum == 0)
531 tmp_csum = -1;
532
533 csum = tmp_csum;
534 if (skb_store_bits(skb, offset, &csum, 2))
535 BUG();
536
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537send:
538 err = ip6_push_pending_frames(sk);
539out:
540 return err;
541}
542
543static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
544 struct flowi *fl, struct rt6_info *rt,
545 unsigned int flags)
546{
Herbert Xu3320da82005-04-19 22:32:22 -0700547 struct ipv6_pinfo *np = inet6_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 struct ipv6hdr *iph;
549 struct sk_buff *skb;
550 unsigned int hh_len;
551 int err;
552
553 if (length > rt->u.dst.dev->mtu) {
554 ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
555 return -EMSGSIZE;
556 }
557 if (flags&MSG_PROBE)
558 goto out;
559
560 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
561
562 skb = sock_alloc_send_skb(sk, length+hh_len+15,
563 flags&MSG_DONTWAIT, &err);
564 if (skb == NULL)
565 goto error;
566 skb_reserve(skb, hh_len);
567
568 skb->priority = sk->sk_priority;
569 skb->dst = dst_clone(&rt->u.dst);
570
571 skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
572
573 skb->ip_summed = CHECKSUM_NONE;
574
575 skb->h.raw = skb->nh.raw;
576 err = memcpy_fromiovecend((void *)iph, from, 0, length);
577 if (err)
578 goto error_fault;
579
580 IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
581 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
582 dst_output);
583 if (err > 0)
Herbert Xu3320da82005-04-19 22:32:22 -0700584 err = np->recverr ? net_xmit_errno(err) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 if (err)
586 goto error;
587out:
588 return 0;
589
590error_fault:
591 err = -EFAULT;
592 kfree_skb(skb);
593error:
594 IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
595 return err;
596}
597
598static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
599{
600 struct iovec *iov;
601 u8 __user *type = NULL;
602 u8 __user *code = NULL;
603 int probed = 0;
604 int i;
605
606 if (!msg->msg_iov)
607 return;
608
609 for (i = 0; i < msg->msg_iovlen; i++) {
610 iov = &msg->msg_iov[i];
611 if (!iov)
612 continue;
613
614 switch (fl->proto) {
615 case IPPROTO_ICMPV6:
616 /* check if one-byte field is readable or not. */
617 if (iov->iov_base && iov->iov_len < 1)
618 break;
619
620 if (!type) {
621 type = iov->iov_base;
622 /* check if code field is readable or not. */
623 if (iov->iov_len > 1)
624 code = type + 1;
625 } else if (!code)
626 code = iov->iov_base;
627
628 if (type && code) {
629 get_user(fl->fl_icmp_type, type);
630 __get_user(fl->fl_icmp_code, code);
631 probed = 1;
632 }
633 break;
634 default:
635 probed = 1;
636 break;
637 }
638 if (probed)
639 break;
640 }
641}
642
643static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
644 struct msghdr *msg, size_t len)
645{
646 struct ipv6_txoptions opt_space;
647 struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
648 struct in6_addr *daddr, *final_p = NULL, final;
649 struct inet_sock *inet = inet_sk(sk);
650 struct ipv6_pinfo *np = inet6_sk(sk);
651 struct raw6_sock *rp = raw6_sk(sk);
652 struct ipv6_txoptions *opt = NULL;
653 struct ip6_flowlabel *flowlabel = NULL;
654 struct dst_entry *dst = NULL;
655 struct flowi fl;
656 int addr_len = msg->msg_namelen;
657 int hlimit = -1;
658 u16 proto;
659 int err;
660
661 /* Rough check on arithmetic overflow,
662 better check is made in ip6_build_xmit
663 */
664 if (len < 0)
665 return -EMSGSIZE;
666
667 /* Mirror BSD error message compatibility */
668 if (msg->msg_flags & MSG_OOB)
669 return -EOPNOTSUPP;
670
671 /*
672 * Get and verify the address.
673 */
674 memset(&fl, 0, sizeof(fl));
675
676 if (sin6) {
677 if (addr_len < SIN6_LEN_RFC2133)
678 return -EINVAL;
679
680 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
681 return(-EAFNOSUPPORT);
682
683 /* port is the proto value [0..255] carried in nexthdr */
684 proto = ntohs(sin6->sin6_port);
685
686 if (!proto)
687 proto = inet->num;
688 else if (proto != inet->num)
689 return(-EINVAL);
690
691 if (proto > 255)
692 return(-EINVAL);
693
694 daddr = &sin6->sin6_addr;
695 if (np->sndflow) {
696 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
697 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
698 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
699 if (flowlabel == NULL)
700 return -EINVAL;
701 daddr = &flowlabel->dst;
702 }
703 }
704
705 /*
706 * Otherwise it will be difficult to maintain
707 * sk->sk_dst_cache.
708 */
709 if (sk->sk_state == TCP_ESTABLISHED &&
710 ipv6_addr_equal(daddr, &np->daddr))
711 daddr = &np->daddr;
712
713 if (addr_len >= sizeof(struct sockaddr_in6) &&
714 sin6->sin6_scope_id &&
715 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
716 fl.oif = sin6->sin6_scope_id;
717 } else {
718 if (sk->sk_state != TCP_ESTABLISHED)
719 return -EDESTADDRREQ;
720
721 proto = inet->num;
722 daddr = &np->daddr;
723 fl.fl6_flowlabel = np->flow_label;
724 }
725
726 if (ipv6_addr_any(daddr)) {
727 /*
728 * unspecified destination address
729 * treated as error... is this correct ?
730 */
731 fl6_sock_release(flowlabel);
732 return(-EINVAL);
733 }
734
735 if (fl.oif == 0)
736 fl.oif = sk->sk_bound_dev_if;
737
738 if (msg->msg_controllen) {
739 opt = &opt_space;
740 memset(opt, 0, sizeof(struct ipv6_txoptions));
741 opt->tot_len = sizeof(struct ipv6_txoptions);
742
743 err = datagram_send_ctl(msg, &fl, opt, &hlimit);
744 if (err < 0) {
745 fl6_sock_release(flowlabel);
746 return err;
747 }
748 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
749 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
750 if (flowlabel == NULL)
751 return -EINVAL;
752 }
753 if (!(opt->opt_nflen|opt->opt_flen))
754 opt = NULL;
755 }
756 if (opt == NULL)
757 opt = np->opt;
758 if (flowlabel)
759 opt = fl6_merge_options(&opt_space, flowlabel, opt);
760
761 fl.proto = proto;
762 rawv6_probe_proto_opt(&fl, msg);
763
764 ipv6_addr_copy(&fl.fl6_dst, daddr);
765 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
766 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
767
768 /* merge ip6_build_xmit from ip6_output */
769 if (opt && opt->srcrt) {
770 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
771 ipv6_addr_copy(&final, &fl.fl6_dst);
772 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
773 final_p = &final;
774 }
775
776 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
777 fl.oif = np->mcast_oif;
778
779 err = ip6_dst_lookup(sk, &dst, &fl);
780 if (err)
781 goto out;
782 if (final_p)
783 ipv6_addr_copy(&fl.fl6_dst, final_p);
784
785 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
786 dst_release(dst);
787 goto out;
788 }
789
790 if (hlimit < 0) {
791 if (ipv6_addr_is_multicast(&fl.fl6_dst))
792 hlimit = np->mcast_hops;
793 else
794 hlimit = np->hop_limit;
795 if (hlimit < 0)
796 hlimit = dst_metric(dst, RTAX_HOPLIMIT);
797 if (hlimit < 0)
798 hlimit = ipv6_get_hoplimit(dst->dev);
799 }
800
801 if (msg->msg_flags&MSG_CONFIRM)
802 goto do_confirm;
803
804back_from_confirm:
805 if (inet->hdrincl) {
806 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
807 } else {
808 lock_sock(sk);
809 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
810 hlimit, opt, &fl, (struct rt6_info*)dst, msg->msg_flags);
811
812 if (err)
813 ip6_flush_pending_frames(sk);
814 else if (!(msg->msg_flags & MSG_MORE))
Herbert Xu357b40a2005-04-19 22:30:14 -0700815 err = rawv6_push_pending_frames(sk, &fl, rp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 }
817done:
818 ip6_dst_store(sk, dst,
819 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
820 &np->daddr : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821
822 release_sock(sk);
823out:
824 fl6_sock_release(flowlabel);
825 return err<0?err:len;
826do_confirm:
827 dst_confirm(dst);
828 if (!(msg->msg_flags & MSG_PROBE) || len)
829 goto back_from_confirm;
830 err = 0;
831 goto done;
832}
833
834static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
835 char __user *optval, int optlen)
836{
837 switch (optname) {
838 case ICMPV6_FILTER:
839 if (optlen > sizeof(struct icmp6_filter))
840 optlen = sizeof(struct icmp6_filter);
841 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
842 return -EFAULT;
843 return 0;
844 default:
845 return -ENOPROTOOPT;
846 };
847
848 return 0;
849}
850
851static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
852 char __user *optval, int __user *optlen)
853{
854 int len;
855
856 switch (optname) {
857 case ICMPV6_FILTER:
858 if (get_user(len, optlen))
859 return -EFAULT;
860 if (len < 0)
861 return -EINVAL;
862 if (len > sizeof(struct icmp6_filter))
863 len = sizeof(struct icmp6_filter);
864 if (put_user(len, optlen))
865 return -EFAULT;
866 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
867 return -EFAULT;
868 return 0;
869 default:
870 return -ENOPROTOOPT;
871 };
872
873 return 0;
874}
875
876
877static int rawv6_setsockopt(struct sock *sk, int level, int optname,
878 char __user *optval, int optlen)
879{
880 struct raw6_sock *rp = raw6_sk(sk);
881 int val;
882
883 switch(level) {
884 case SOL_RAW:
885 break;
886
887 case SOL_ICMPV6:
888 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
889 return -EOPNOTSUPP;
890 return rawv6_seticmpfilter(sk, level, optname, optval,
891 optlen);
892 case SOL_IPV6:
893 if (optname == IPV6_CHECKSUM)
894 break;
895 default:
896 return ipv6_setsockopt(sk, level, optname, optval,
897 optlen);
898 };
899
900 if (get_user(val, (int __user *)optval))
901 return -EFAULT;
902
903 switch (optname) {
904 case IPV6_CHECKSUM:
905 /* You may get strange result with a positive odd offset;
906 RFC2292bis agrees with me. */
907 if (val > 0 && (val&1))
908 return(-EINVAL);
909 if (val < 0) {
910 rp->checksum = 0;
911 } else {
912 rp->checksum = 1;
913 rp->offset = val;
914 }
915
916 return 0;
917 break;
918
919 default:
920 return(-ENOPROTOOPT);
921 }
922}
923
924static int rawv6_getsockopt(struct sock *sk, int level, int optname,
925 char __user *optval, int __user *optlen)
926{
927 struct raw6_sock *rp = raw6_sk(sk);
928 int val, len;
929
930 switch(level) {
931 case SOL_RAW:
932 break;
933
934 case SOL_ICMPV6:
935 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
936 return -EOPNOTSUPP;
937 return rawv6_geticmpfilter(sk, level, optname, optval,
938 optlen);
939 case SOL_IPV6:
940 if (optname == IPV6_CHECKSUM)
941 break;
942 default:
943 return ipv6_getsockopt(sk, level, optname, optval,
944 optlen);
945 };
946
947 if (get_user(len,optlen))
948 return -EFAULT;
949
950 switch (optname) {
951 case IPV6_CHECKSUM:
952 if (rp->checksum == 0)
953 val = -1;
954 else
955 val = rp->offset;
956 break;
957
958 default:
959 return -ENOPROTOOPT;
960 }
961
962 len = min_t(unsigned int, sizeof(int), len);
963
964 if (put_user(len, optlen))
965 return -EFAULT;
966 if (copy_to_user(optval,&val,len))
967 return -EFAULT;
968 return 0;
969}
970
971static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
972{
973 switch(cmd) {
974 case SIOCOUTQ:
975 {
976 int amount = atomic_read(&sk->sk_wmem_alloc);
977 return put_user(amount, (int __user *)arg);
978 }
979 case SIOCINQ:
980 {
981 struct sk_buff *skb;
982 int amount = 0;
983
Herbert Xue0f9f852005-06-18 22:56:18 -0700984 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 skb = skb_peek(&sk->sk_receive_queue);
986 if (skb != NULL)
987 amount = skb->tail - skb->h.raw;
Herbert Xue0f9f852005-06-18 22:56:18 -0700988 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 return put_user(amount, (int __user *)arg);
990 }
991
992 default:
993 return -ENOIOCTLCMD;
994 }
995}
996
997static void rawv6_close(struct sock *sk, long timeout)
998{
999 if (inet_sk(sk)->num == IPPROTO_RAW)
1000 ip6_ra_control(sk, -1, NULL);
1001
1002 sk_common_release(sk);
1003}
1004
1005static int rawv6_init_sk(struct sock *sk)
1006{
1007 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
1008 struct raw6_sock *rp = raw6_sk(sk);
1009 rp->checksum = 1;
1010 rp->offset = 2;
1011 }
1012 return(0);
1013}
1014
1015struct proto rawv6_prot = {
1016 .name = "RAWv6",
1017 .owner = THIS_MODULE,
1018 .close = rawv6_close,
1019 .connect = ip6_datagram_connect,
1020 .disconnect = udp_disconnect,
1021 .ioctl = rawv6_ioctl,
1022 .init = rawv6_init_sk,
1023 .destroy = inet6_destroy_sock,
1024 .setsockopt = rawv6_setsockopt,
1025 .getsockopt = rawv6_getsockopt,
1026 .sendmsg = rawv6_sendmsg,
1027 .recvmsg = rawv6_recvmsg,
1028 .bind = rawv6_bind,
1029 .backlog_rcv = rawv6_rcv_skb,
1030 .hash = raw_v6_hash,
1031 .unhash = raw_v6_unhash,
1032 .obj_size = sizeof(struct raw6_sock),
1033};
1034
1035#ifdef CONFIG_PROC_FS
1036struct raw6_iter_state {
1037 int bucket;
1038};
1039
1040#define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1041
1042static struct sock *raw6_get_first(struct seq_file *seq)
1043{
1044 struct sock *sk;
1045 struct hlist_node *node;
1046 struct raw6_iter_state* state = raw6_seq_private(seq);
1047
1048 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1049 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1050 if (sk->sk_family == PF_INET6)
1051 goto out;
1052 sk = NULL;
1053out:
1054 return sk;
1055}
1056
1057static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1058{
1059 struct raw6_iter_state* state = raw6_seq_private(seq);
1060
1061 do {
1062 sk = sk_next(sk);
1063try_again:
1064 ;
1065 } while (sk && sk->sk_family != PF_INET6);
1066
1067 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1068 sk = sk_head(&raw_v6_htable[state->bucket]);
1069 goto try_again;
1070 }
1071 return sk;
1072}
1073
1074static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1075{
1076 struct sock *sk = raw6_get_first(seq);
1077 if (sk)
1078 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1079 --pos;
1080 return pos ? NULL : sk;
1081}
1082
1083static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1084{
1085 read_lock(&raw_v6_lock);
1086 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1087}
1088
1089static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1090{
1091 struct sock *sk;
1092
1093 if (v == SEQ_START_TOKEN)
1094 sk = raw6_get_first(seq);
1095 else
1096 sk = raw6_get_next(seq, v);
1097 ++*pos;
1098 return sk;
1099}
1100
1101static void raw6_seq_stop(struct seq_file *seq, void *v)
1102{
1103 read_unlock(&raw_v6_lock);
1104}
1105
1106static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1107{
1108 struct ipv6_pinfo *np = inet6_sk(sp);
1109 struct in6_addr *dest, *src;
1110 __u16 destp, srcp;
1111
1112 dest = &np->daddr;
1113 src = &np->rcv_saddr;
1114 destp = 0;
1115 srcp = inet_sk(sp)->num;
1116 seq_printf(seq,
1117 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1118 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1119 i,
1120 src->s6_addr32[0], src->s6_addr32[1],
1121 src->s6_addr32[2], src->s6_addr32[3], srcp,
1122 dest->s6_addr32[0], dest->s6_addr32[1],
1123 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1124 sp->sk_state,
1125 atomic_read(&sp->sk_wmem_alloc),
1126 atomic_read(&sp->sk_rmem_alloc),
1127 0, 0L, 0,
1128 sock_i_uid(sp), 0,
1129 sock_i_ino(sp),
1130 atomic_read(&sp->sk_refcnt), sp);
1131}
1132
1133static int raw6_seq_show(struct seq_file *seq, void *v)
1134{
1135 if (v == SEQ_START_TOKEN)
1136 seq_printf(seq,
1137 " sl "
1138 "local_address "
1139 "remote_address "
1140 "st tx_queue rx_queue tr tm->when retrnsmt"
1141 " uid timeout inode\n");
1142 else
1143 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1144 return 0;
1145}
1146
1147static struct seq_operations raw6_seq_ops = {
1148 .start = raw6_seq_start,
1149 .next = raw6_seq_next,
1150 .stop = raw6_seq_stop,
1151 .show = raw6_seq_show,
1152};
1153
1154static int raw6_seq_open(struct inode *inode, struct file *file)
1155{
1156 struct seq_file *seq;
1157 int rc = -ENOMEM;
1158 struct raw6_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1159 if (!s)
1160 goto out;
1161 rc = seq_open(file, &raw6_seq_ops);
1162 if (rc)
1163 goto out_kfree;
1164 seq = file->private_data;
1165 seq->private = s;
1166 memset(s, 0, sizeof(*s));
1167out:
1168 return rc;
1169out_kfree:
1170 kfree(s);
1171 goto out;
1172}
1173
1174static struct file_operations raw6_seq_fops = {
1175 .owner = THIS_MODULE,
1176 .open = raw6_seq_open,
1177 .read = seq_read,
1178 .llseek = seq_lseek,
1179 .release = seq_release_private,
1180};
1181
1182int __init raw6_proc_init(void)
1183{
1184 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1185 return -ENOMEM;
1186 return 0;
1187}
1188
1189void raw6_proc_exit(void)
1190{
1191 proc_net_remove("raw6");
1192}
1193#endif /* CONFIG_PROC_FS */