blob: d77d3352c96754c5b06e00f5d168762fd2dd1ddd [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>
Arnaldo Carvalho de Meloc752f072005-08-09 20:08:28 -070052#include <net/tcp_states.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070053
54#include <net/rawv6.h>
55#include <net/xfrm.h>
56
57#include <linux/proc_fs.h>
58#include <linux/seq_file.h>
59
60struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
61DEFINE_RWLOCK(raw_v6_lock);
62
63static void raw_v6_hash(struct sock *sk)
64{
65 struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
66 (RAWV6_HTABLE_SIZE - 1)];
67
68 write_lock_bh(&raw_v6_lock);
69 sk_add_node(sk, list);
70 sock_prot_inc_use(sk->sk_prot);
71 write_unlock_bh(&raw_v6_lock);
72}
73
74static void raw_v6_unhash(struct sock *sk)
75{
76 write_lock_bh(&raw_v6_lock);
77 if (sk_del_node_init(sk))
78 sock_prot_dec_use(sk->sk_prot);
79 write_unlock_bh(&raw_v6_lock);
80}
81
82
83/* Grumble... icmp and ip_input want to get at this... */
84struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
Andrew McDonald0bd1b592005-08-09 19:44:42 -070085 struct in6_addr *loc_addr, struct in6_addr *rmt_addr,
86 int dif)
Linus Torvalds1da177e2005-04-16 15:20:36 -070087{
88 struct hlist_node *node;
89 int is_multicast = ipv6_addr_is_multicast(loc_addr);
90
91 sk_for_each_from(sk, node)
92 if (inet_sk(sk)->num == num) {
93 struct ipv6_pinfo *np = inet6_sk(sk);
94
95 if (!ipv6_addr_any(&np->daddr) &&
96 !ipv6_addr_equal(&np->daddr, rmt_addr))
97 continue;
98
Andrew McDonald0bd1b592005-08-09 19:44:42 -070099 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
100 continue;
101
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102 if (!ipv6_addr_any(&np->rcv_saddr)) {
103 if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
104 goto found;
105 if (is_multicast &&
106 inet6_mc_check(sk, loc_addr, rmt_addr))
107 goto found;
108 continue;
109 }
110 goto found;
111 }
112 sk = NULL;
113found:
114 return sk;
115}
116
117/*
118 * 0 - deliver
119 * 1 - block
120 */
121static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
122{
123 struct icmp6hdr *icmph;
124 struct raw6_sock *rp = raw6_sk(sk);
125
126 if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
127 __u32 *data = &rp->filter.data[0];
128 int bit_nr;
129
130 icmph = (struct icmp6hdr *) skb->data;
131 bit_nr = icmph->icmp6_type;
132
133 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
134 }
135 return 0;
136}
137
138/*
139 * demultiplex raw sockets.
140 * (should consider queueing the skb in the sock receive_queue
141 * without calling rawv6.c)
142 *
143 * Caller owns SKB so we must make clones.
144 */
Patrick McHardyd13964f2005-08-09 19:45:02 -0700145int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146{
147 struct in6_addr *saddr;
148 struct in6_addr *daddr;
149 struct sock *sk;
Patrick McHardyd13964f2005-08-09 19:45:02 -0700150 int delivered = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 __u8 hash;
152
153 saddr = &skb->nh.ipv6h->saddr;
154 daddr = saddr + 1;
155
156 hash = nexthdr & (MAX_INET_PROTOS - 1);
157
158 read_lock(&raw_v6_lock);
159 sk = sk_head(&raw_v6_htable[hash]);
160
161 /*
162 * The first socket found will be delivered after
163 * delivery to transport protocols.
164 */
165
166 if (sk == NULL)
167 goto out;
168
YOSHIFUJI Hideaki2dac4b92005-09-01 17:44:49 -0700169 sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170
171 while (sk) {
Patrick McHardyd13964f2005-08-09 19:45:02 -0700172 delivered = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 if (nexthdr != IPPROTO_ICMPV6 || !icmpv6_filter(sk, skb)) {
174 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
175
176 /* Not releasing hash table! */
177 if (clone)
178 rawv6_rcv(sk, clone);
179 }
Andrew McDonald0bd1b592005-08-09 19:44:42 -0700180 sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
YOSHIFUJI Hideaki2dac4b92005-09-01 17:44:49 -0700181 IP6CB(skb)->iif);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 }
183out:
184 read_unlock(&raw_v6_lock);
Patrick McHardyd13964f2005-08-09 19:45:02 -0700185 return delivered;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186}
187
188/* This cleans up af_inet6 a bit. -DaveM */
189static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
190{
191 struct inet_sock *inet = inet_sk(sk);
192 struct ipv6_pinfo *np = inet6_sk(sk);
193 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
194 __u32 v4addr = 0;
195 int addr_type;
196 int err;
197
198 if (addr_len < SIN6_LEN_RFC2133)
199 return -EINVAL;
200 addr_type = ipv6_addr_type(&addr->sin6_addr);
201
202 /* Raw sockets are IPv6 only */
203 if (addr_type == IPV6_ADDR_MAPPED)
204 return(-EADDRNOTAVAIL);
205
206 lock_sock(sk);
207
208 err = -EINVAL;
209 if (sk->sk_state != TCP_CLOSE)
210 goto out;
211
212 /* Check if the address belongs to the host. */
213 if (addr_type != IPV6_ADDR_ANY) {
214 struct net_device *dev = NULL;
215
216 if (addr_type & IPV6_ADDR_LINKLOCAL) {
217 if (addr_len >= sizeof(struct sockaddr_in6) &&
218 addr->sin6_scope_id) {
219 /* Override any existing binding, if another
220 * one is supplied by user.
221 */
222 sk->sk_bound_dev_if = addr->sin6_scope_id;
223 }
224
225 /* Binding to link-local address requires an interface */
226 if (!sk->sk_bound_dev_if)
227 goto out;
228
229 dev = dev_get_by_index(sk->sk_bound_dev_if);
230 if (!dev) {
231 err = -ENODEV;
232 goto out;
233 }
234 }
235
236 /* ipv4 addr of the socket is invalid. Only the
237 * unspecified and mapped address have a v4 equivalent.
238 */
239 v4addr = LOOPBACK4_IPV6;
240 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
241 err = -EADDRNOTAVAIL;
242 if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
243 if (dev)
244 dev_put(dev);
245 goto out;
246 }
247 }
248 if (dev)
249 dev_put(dev);
250 }
251
252 inet->rcv_saddr = inet->saddr = v4addr;
253 ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
254 if (!(addr_type & IPV6_ADDR_MULTICAST))
255 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
256 err = 0;
257out:
258 release_sock(sk);
259 return err;
260}
261
262void rawv6_err(struct sock *sk, struct sk_buff *skb,
263 struct inet6_skb_parm *opt,
264 int type, int code, int offset, u32 info)
265{
266 struct inet_sock *inet = inet_sk(sk);
267 struct ipv6_pinfo *np = inet6_sk(sk);
268 int err;
269 int harderr;
270
271 /* Report error on raw socket, if:
272 1. User requested recverr.
273 2. Socket is connected (otherwise the error indication
274 is useless without recverr and error is hard.
275 */
276 if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
277 return;
278
279 harderr = icmpv6_err_convert(type, code, &err);
280 if (type == ICMPV6_PKT_TOOBIG)
281 harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
282
283 if (np->recverr) {
284 u8 *payload = skb->data;
285 if (!inet->hdrincl)
286 payload += offset;
287 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
288 }
289
290 if (np->recverr || harderr) {
291 sk->sk_err = err;
292 sk->sk_error_report(sk);
293 }
294}
295
296static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
297{
298 if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
299 skb->ip_summed != CHECKSUM_UNNECESSARY) {
300 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
301 /* FIXME: increment a raw6 drops counter here */
302 kfree_skb(skb);
303 return 0;
304 }
305 skb->ip_summed = CHECKSUM_UNNECESSARY;
306 }
307
308 /* Charge it to the socket. */
309 if (sock_queue_rcv_skb(sk,skb)<0) {
310 /* FIXME: increment a raw6 drops counter here */
311 kfree_skb(skb);
312 return 0;
313 }
314
315 return 0;
316}
317
318/*
319 * This is next to useless...
320 * if we demultiplex in network layer we don't need the extra call
321 * just to queue the skb...
322 * maybe we could have the network decide upon a hint if it
323 * should call raw_rcv for demultiplexing
324 */
325int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
326{
327 struct inet_sock *inet = inet_sk(sk);
328 struct raw6_sock *rp = raw6_sk(sk);
329
330 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
331 kfree_skb(skb);
332 return NET_RX_DROP;
333 }
334
335 if (!rp->checksum)
336 skb->ip_summed = CHECKSUM_UNNECESSARY;
337
338 if (skb->ip_summed != CHECKSUM_UNNECESSARY) {
339 if (skb->ip_summed == CHECKSUM_HW) {
Patrick McHardy793245e2005-08-16 20:39:38 -0700340 skb_postpull_rcsum(skb, skb->nh.raw,
341 skb->h.raw - skb->nh.raw);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 skb->ip_summed = CHECKSUM_UNNECESSARY;
343 if (csum_ipv6_magic(&skb->nh.ipv6h->saddr,
344 &skb->nh.ipv6h->daddr,
345 skb->len, inet->num, skb->csum)) {
Patrick McHardy64ce2072005-08-09 20:50:53 -0700346 LIMIT_NETDEBUG(KERN_DEBUG "raw v6 hw csum failure.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 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);
Mark J Cox6d1cfe32005-09-19 17:55:30 -0700630 get_user(fl->fl_icmp_code, code);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 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;
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900658 int tclass = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 u16 proto;
660 int err;
661
662 /* Rough check on arithmetic overflow,
663 better check is made in ip6_build_xmit
664 */
665 if (len < 0)
666 return -EMSGSIZE;
667
668 /* Mirror BSD error message compatibility */
669 if (msg->msg_flags & MSG_OOB)
670 return -EOPNOTSUPP;
671
672 /*
673 * Get and verify the address.
674 */
675 memset(&fl, 0, sizeof(fl));
676
677 if (sin6) {
678 if (addr_len < SIN6_LEN_RFC2133)
679 return -EINVAL;
680
681 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
682 return(-EAFNOSUPPORT);
683
684 /* port is the proto value [0..255] carried in nexthdr */
685 proto = ntohs(sin6->sin6_port);
686
687 if (!proto)
688 proto = inet->num;
689 else if (proto != inet->num)
690 return(-EINVAL);
691
692 if (proto > 255)
693 return(-EINVAL);
694
695 daddr = &sin6->sin6_addr;
696 if (np->sndflow) {
697 fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
698 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
699 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
700 if (flowlabel == NULL)
701 return -EINVAL;
702 daddr = &flowlabel->dst;
703 }
704 }
705
706 /*
707 * Otherwise it will be difficult to maintain
708 * sk->sk_dst_cache.
709 */
710 if (sk->sk_state == TCP_ESTABLISHED &&
711 ipv6_addr_equal(daddr, &np->daddr))
712 daddr = &np->daddr;
713
714 if (addr_len >= sizeof(struct sockaddr_in6) &&
715 sin6->sin6_scope_id &&
716 ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
717 fl.oif = sin6->sin6_scope_id;
718 } else {
719 if (sk->sk_state != TCP_ESTABLISHED)
720 return -EDESTADDRREQ;
721
722 proto = inet->num;
723 daddr = &np->daddr;
724 fl.fl6_flowlabel = np->flow_label;
725 }
726
727 if (ipv6_addr_any(daddr)) {
728 /*
729 * unspecified destination address
730 * treated as error... is this correct ?
731 */
732 fl6_sock_release(flowlabel);
733 return(-EINVAL);
734 }
735
736 if (fl.oif == 0)
737 fl.oif = sk->sk_bound_dev_if;
738
739 if (msg->msg_controllen) {
740 opt = &opt_space;
741 memset(opt, 0, sizeof(struct ipv6_txoptions));
742 opt->tot_len = sizeof(struct ipv6_txoptions);
743
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900744 err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 if (err < 0) {
746 fl6_sock_release(flowlabel);
747 return err;
748 }
749 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
750 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
751 if (flowlabel == NULL)
752 return -EINVAL;
753 }
754 if (!(opt->opt_nflen|opt->opt_flen))
755 opt = NULL;
756 }
757 if (opt == NULL)
758 opt = np->opt;
YOSHIFUJI Hideakidf9890c2005-11-20 12:23:18 +0900759 if (flowlabel)
760 opt = fl6_merge_options(&opt_space, flowlabel, opt);
761 opt = ipv6_fixup_options(&opt_space, opt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
763 fl.proto = proto;
764 rawv6_probe_proto_opt(&fl, msg);
765
766 ipv6_addr_copy(&fl.fl6_dst, daddr);
767 if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
768 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
769
770 /* merge ip6_build_xmit from ip6_output */
771 if (opt && opt->srcrt) {
772 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
773 ipv6_addr_copy(&final, &fl.fl6_dst);
774 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
775 final_p = &final;
776 }
777
778 if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
779 fl.oif = np->mcast_oif;
780
781 err = ip6_dst_lookup(sk, &dst, &fl);
782 if (err)
783 goto out;
784 if (final_p)
785 ipv6_addr_copy(&fl.fl6_dst, final_p);
786
Patrick McHardye104411b2005-09-08 15:11:55 -0700787 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789
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
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900801 if (tclass < 0) {
802 tclass = np->cork.tclass;
803 if (tclass < 0)
804 tclass = 0;
805 }
806
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 if (msg->msg_flags&MSG_CONFIRM)
808 goto do_confirm;
809
810back_from_confirm:
811 if (inet->hdrincl) {
812 err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
813 } else {
814 lock_sock(sk);
YOSHIFUJI Hideaki41a1f8e2005-09-08 10:19:03 +0900815 err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
816 len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
817 msg->msg_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818
819 if (err)
820 ip6_flush_pending_frames(sk);
821 else if (!(msg->msg_flags & MSG_MORE))
Herbert Xu357b40a2005-04-19 22:30:14 -0700822 err = rawv6_push_pending_frames(sk, &fl, rp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 }
824done:
825 ip6_dst_store(sk, dst,
826 ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
827 &np->daddr : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828
829 release_sock(sk);
830out:
831 fl6_sock_release(flowlabel);
832 return err<0?err:len;
833do_confirm:
834 dst_confirm(dst);
835 if (!(msg->msg_flags & MSG_PROBE) || len)
836 goto back_from_confirm;
837 err = 0;
838 goto done;
839}
840
841static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
842 char __user *optval, int optlen)
843{
844 switch (optname) {
845 case ICMPV6_FILTER:
846 if (optlen > sizeof(struct icmp6_filter))
847 optlen = sizeof(struct icmp6_filter);
848 if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
849 return -EFAULT;
850 return 0;
851 default:
852 return -ENOPROTOOPT;
853 };
854
855 return 0;
856}
857
858static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
859 char __user *optval, int __user *optlen)
860{
861 int len;
862
863 switch (optname) {
864 case ICMPV6_FILTER:
865 if (get_user(len, optlen))
866 return -EFAULT;
867 if (len < 0)
868 return -EINVAL;
869 if (len > sizeof(struct icmp6_filter))
870 len = sizeof(struct icmp6_filter);
871 if (put_user(len, optlen))
872 return -EFAULT;
873 if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
874 return -EFAULT;
875 return 0;
876 default:
877 return -ENOPROTOOPT;
878 };
879
880 return 0;
881}
882
883
884static int rawv6_setsockopt(struct sock *sk, int level, int optname,
885 char __user *optval, int optlen)
886{
887 struct raw6_sock *rp = raw6_sk(sk);
888 int val;
889
890 switch(level) {
891 case SOL_RAW:
892 break;
893
894 case SOL_ICMPV6:
895 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
896 return -EOPNOTSUPP;
897 return rawv6_seticmpfilter(sk, level, optname, optval,
898 optlen);
899 case SOL_IPV6:
900 if (optname == IPV6_CHECKSUM)
901 break;
902 default:
903 return ipv6_setsockopt(sk, level, optname, optval,
904 optlen);
905 };
906
907 if (get_user(val, (int __user *)optval))
908 return -EFAULT;
909
910 switch (optname) {
911 case IPV6_CHECKSUM:
912 /* You may get strange result with a positive odd offset;
913 RFC2292bis agrees with me. */
914 if (val > 0 && (val&1))
915 return(-EINVAL);
916 if (val < 0) {
917 rp->checksum = 0;
918 } else {
919 rp->checksum = 1;
920 rp->offset = val;
921 }
922
923 return 0;
924 break;
925
926 default:
927 return(-ENOPROTOOPT);
928 }
929}
930
931static int rawv6_getsockopt(struct sock *sk, int level, int optname,
932 char __user *optval, int __user *optlen)
933{
934 struct raw6_sock *rp = raw6_sk(sk);
935 int val, len;
936
937 switch(level) {
938 case SOL_RAW:
939 break;
940
941 case SOL_ICMPV6:
942 if (inet_sk(sk)->num != IPPROTO_ICMPV6)
943 return -EOPNOTSUPP;
944 return rawv6_geticmpfilter(sk, level, optname, optval,
945 optlen);
946 case SOL_IPV6:
947 if (optname == IPV6_CHECKSUM)
948 break;
949 default:
950 return ipv6_getsockopt(sk, level, optname, optval,
951 optlen);
952 };
953
954 if (get_user(len,optlen))
955 return -EFAULT;
956
957 switch (optname) {
958 case IPV6_CHECKSUM:
959 if (rp->checksum == 0)
960 val = -1;
961 else
962 val = rp->offset;
963 break;
964
965 default:
966 return -ENOPROTOOPT;
967 }
968
969 len = min_t(unsigned int, sizeof(int), len);
970
971 if (put_user(len, optlen))
972 return -EFAULT;
973 if (copy_to_user(optval,&val,len))
974 return -EFAULT;
975 return 0;
976}
977
978static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
979{
980 switch(cmd) {
981 case SIOCOUTQ:
982 {
983 int amount = atomic_read(&sk->sk_wmem_alloc);
984 return put_user(amount, (int __user *)arg);
985 }
986 case SIOCINQ:
987 {
988 struct sk_buff *skb;
989 int amount = 0;
990
Herbert Xue0f9f852005-06-18 22:56:18 -0700991 spin_lock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 skb = skb_peek(&sk->sk_receive_queue);
993 if (skb != NULL)
994 amount = skb->tail - skb->h.raw;
Herbert Xue0f9f852005-06-18 22:56:18 -0700995 spin_unlock_bh(&sk->sk_receive_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 return put_user(amount, (int __user *)arg);
997 }
998
999 default:
1000 return -ENOIOCTLCMD;
1001 }
1002}
1003
1004static void rawv6_close(struct sock *sk, long timeout)
1005{
1006 if (inet_sk(sk)->num == IPPROTO_RAW)
1007 ip6_ra_control(sk, -1, NULL);
1008
1009 sk_common_release(sk);
1010}
1011
1012static int rawv6_init_sk(struct sock *sk)
1013{
1014 if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
1015 struct raw6_sock *rp = raw6_sk(sk);
1016 rp->checksum = 1;
1017 rp->offset = 2;
1018 }
1019 return(0);
1020}
1021
1022struct proto rawv6_prot = {
1023 .name = "RAWv6",
1024 .owner = THIS_MODULE,
1025 .close = rawv6_close,
1026 .connect = ip6_datagram_connect,
1027 .disconnect = udp_disconnect,
1028 .ioctl = rawv6_ioctl,
1029 .init = rawv6_init_sk,
1030 .destroy = inet6_destroy_sock,
1031 .setsockopt = rawv6_setsockopt,
1032 .getsockopt = rawv6_getsockopt,
1033 .sendmsg = rawv6_sendmsg,
1034 .recvmsg = rawv6_recvmsg,
1035 .bind = rawv6_bind,
1036 .backlog_rcv = rawv6_rcv_skb,
1037 .hash = raw_v6_hash,
1038 .unhash = raw_v6_unhash,
1039 .obj_size = sizeof(struct raw6_sock),
1040};
1041
1042#ifdef CONFIG_PROC_FS
1043struct raw6_iter_state {
1044 int bucket;
1045};
1046
1047#define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1048
1049static struct sock *raw6_get_first(struct seq_file *seq)
1050{
1051 struct sock *sk;
1052 struct hlist_node *node;
1053 struct raw6_iter_state* state = raw6_seq_private(seq);
1054
1055 for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1056 sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1057 if (sk->sk_family == PF_INET6)
1058 goto out;
1059 sk = NULL;
1060out:
1061 return sk;
1062}
1063
1064static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1065{
1066 struct raw6_iter_state* state = raw6_seq_private(seq);
1067
1068 do {
1069 sk = sk_next(sk);
1070try_again:
1071 ;
1072 } while (sk && sk->sk_family != PF_INET6);
1073
1074 if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1075 sk = sk_head(&raw_v6_htable[state->bucket]);
1076 goto try_again;
1077 }
1078 return sk;
1079}
1080
1081static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1082{
1083 struct sock *sk = raw6_get_first(seq);
1084 if (sk)
1085 while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1086 --pos;
1087 return pos ? NULL : sk;
1088}
1089
1090static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1091{
1092 read_lock(&raw_v6_lock);
1093 return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1094}
1095
1096static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1097{
1098 struct sock *sk;
1099
1100 if (v == SEQ_START_TOKEN)
1101 sk = raw6_get_first(seq);
1102 else
1103 sk = raw6_get_next(seq, v);
1104 ++*pos;
1105 return sk;
1106}
1107
1108static void raw6_seq_stop(struct seq_file *seq, void *v)
1109{
1110 read_unlock(&raw_v6_lock);
1111}
1112
1113static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1114{
1115 struct ipv6_pinfo *np = inet6_sk(sp);
1116 struct in6_addr *dest, *src;
1117 __u16 destp, srcp;
1118
1119 dest = &np->daddr;
1120 src = &np->rcv_saddr;
1121 destp = 0;
1122 srcp = inet_sk(sp)->num;
1123 seq_printf(seq,
1124 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1125 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1126 i,
1127 src->s6_addr32[0], src->s6_addr32[1],
1128 src->s6_addr32[2], src->s6_addr32[3], srcp,
1129 dest->s6_addr32[0], dest->s6_addr32[1],
1130 dest->s6_addr32[2], dest->s6_addr32[3], destp,
1131 sp->sk_state,
1132 atomic_read(&sp->sk_wmem_alloc),
1133 atomic_read(&sp->sk_rmem_alloc),
1134 0, 0L, 0,
1135 sock_i_uid(sp), 0,
1136 sock_i_ino(sp),
1137 atomic_read(&sp->sk_refcnt), sp);
1138}
1139
1140static int raw6_seq_show(struct seq_file *seq, void *v)
1141{
1142 if (v == SEQ_START_TOKEN)
1143 seq_printf(seq,
1144 " sl "
1145 "local_address "
1146 "remote_address "
1147 "st tx_queue rx_queue tr tm->when retrnsmt"
1148 " uid timeout inode\n");
1149 else
1150 raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1151 return 0;
1152}
1153
1154static struct seq_operations raw6_seq_ops = {
1155 .start = raw6_seq_start,
1156 .next = raw6_seq_next,
1157 .stop = raw6_seq_stop,
1158 .show = raw6_seq_show,
1159};
1160
1161static int raw6_seq_open(struct inode *inode, struct file *file)
1162{
1163 struct seq_file *seq;
1164 int rc = -ENOMEM;
1165 struct raw6_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1166 if (!s)
1167 goto out;
1168 rc = seq_open(file, &raw6_seq_ops);
1169 if (rc)
1170 goto out_kfree;
1171 seq = file->private_data;
1172 seq->private = s;
1173 memset(s, 0, sizeof(*s));
1174out:
1175 return rc;
1176out_kfree:
1177 kfree(s);
1178 goto out;
1179}
1180
1181static struct file_operations raw6_seq_fops = {
1182 .owner = THIS_MODULE,
1183 .open = raw6_seq_open,
1184 .read = seq_read,
1185 .llseek = seq_lseek,
1186 .release = seq_release_private,
1187};
1188
1189int __init raw6_proc_init(void)
1190{
1191 if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1192 return -ENOMEM;
1193 return 0;
1194}
1195
1196void raw6_proc_exit(void)
1197{
1198 proc_net_remove("raw6");
1199}
1200#endif /* CONFIG_PROC_FS */