xref: /linux/net/ipv6/raw.c (revision 8b4a40809e5330c9da5d20107d693d92d73b31dc)
1 /*
2  *	RAW sockets for IPv6
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *
8  *	Adapted from linux/net/ipv4/raw.c
9  *
10  *	$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/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/netfilter.h>
33 #include <linux/netfilter_ipv6.h>
34 #include <linux/skbuff.h>
35 #include <asm/uaccess.h>
36 #include <asm/ioctls.h>
37 
38 #include <net/ip.h>
39 #include <net/sock.h>
40 #include <net/snmp.h>
41 
42 #include <net/ipv6.h>
43 #include <net/ndisc.h>
44 #include <net/protocol.h>
45 #include <net/ip6_route.h>
46 #include <net/ip6_checksum.h>
47 #include <net/addrconf.h>
48 #include <net/transp_v6.h>
49 #include <net/udp.h>
50 #include <net/inet_common.h>
51 #include <net/tcp_states.h>
52 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
53 #include <net/mip6.h>
54 #endif
55 
56 #include <net/rawv6.h>
57 #include <net/xfrm.h>
58 
59 #include <linux/proc_fs.h>
60 #include <linux/seq_file.h>
61 
62 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
63 DEFINE_RWLOCK(raw_v6_lock);
64 
65 static void raw_v6_hash(struct sock *sk)
66 {
67 	struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
68 						 (RAWV6_HTABLE_SIZE - 1)];
69 
70 	write_lock_bh(&raw_v6_lock);
71 	sk_add_node(sk, list);
72 	sock_prot_inc_use(sk->sk_prot);
73 	write_unlock_bh(&raw_v6_lock);
74 }
75 
76 static void raw_v6_unhash(struct sock *sk)
77 {
78 	write_lock_bh(&raw_v6_lock);
79 	if (sk_del_node_init(sk))
80 		sock_prot_dec_use(sk->sk_prot);
81 	write_unlock_bh(&raw_v6_lock);
82 }
83 
84 
85 /* Grumble... icmp and ip_input want to get at this... */
86 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
87 			     struct in6_addr *loc_addr, struct in6_addr *rmt_addr,
88 			     int dif)
89 {
90 	struct hlist_node *node;
91 	int is_multicast = ipv6_addr_is_multicast(loc_addr);
92 
93 	sk_for_each_from(sk, node)
94 		if (inet_sk(sk)->num == num) {
95 			struct ipv6_pinfo *np = inet6_sk(sk);
96 
97 			if (!ipv6_addr_any(&np->daddr) &&
98 			    !ipv6_addr_equal(&np->daddr, rmt_addr))
99 				continue;
100 
101 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
102 				continue;
103 
104 			if (!ipv6_addr_any(&np->rcv_saddr)) {
105 				if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
106 					goto found;
107 				if (is_multicast &&
108 				    inet6_mc_check(sk, loc_addr, rmt_addr))
109 					goto found;
110 				continue;
111 			}
112 			goto found;
113 		}
114 	sk = NULL;
115 found:
116 	return sk;
117 }
118 
119 /*
120  *	0 - deliver
121  *	1 - block
122  */
123 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
124 {
125 	struct icmp6hdr *icmph;
126 	struct raw6_sock *rp = raw6_sk(sk);
127 
128 	if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
129 		__u32 *data = &rp->filter.data[0];
130 		int bit_nr;
131 
132 		icmph = (struct icmp6hdr *) skb->data;
133 		bit_nr = icmph->icmp6_type;
134 
135 		return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
136 	}
137 	return 0;
138 }
139 
140 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
141 static int (*mh_filter)(struct sock *sock, struct sk_buff *skb);
142 
143 int rawv6_mh_filter_register(int (*filter)(struct sock *sock,
144 					   struct sk_buff *skb))
145 {
146 	rcu_assign_pointer(mh_filter, filter);
147 	return 0;
148 }
149 EXPORT_SYMBOL(rawv6_mh_filter_register);
150 
151 int rawv6_mh_filter_unregister(int (*filter)(struct sock *sock,
152 					     struct sk_buff *skb))
153 {
154 	rcu_assign_pointer(mh_filter, NULL);
155 	synchronize_rcu();
156 	return 0;
157 }
158 EXPORT_SYMBOL(rawv6_mh_filter_unregister);
159 
160 #endif
161 
162 /*
163  *	demultiplex raw sockets.
164  *	(should consider queueing the skb in the sock receive_queue
165  *	without calling rawv6.c)
166  *
167  *	Caller owns SKB so we must make clones.
168  */
169 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
170 {
171 	struct in6_addr *saddr;
172 	struct in6_addr *daddr;
173 	struct sock *sk;
174 	int delivered = 0;
175 	__u8 hash;
176 
177 	saddr = &ipv6_hdr(skb)->saddr;
178 	daddr = saddr + 1;
179 
180 	hash = nexthdr & (MAX_INET_PROTOS - 1);
181 
182 	read_lock(&raw_v6_lock);
183 	sk = sk_head(&raw_v6_htable[hash]);
184 
185 	/*
186 	 *	The first socket found will be delivered after
187 	 *	delivery to transport protocols.
188 	 */
189 
190 	if (sk == NULL)
191 		goto out;
192 
193 	sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
194 
195 	while (sk) {
196 		int filtered;
197 
198 		delivered = 1;
199 		switch (nexthdr) {
200 		case IPPROTO_ICMPV6:
201 			filtered = icmpv6_filter(sk, skb);
202 			break;
203 
204 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
205 		case IPPROTO_MH:
206 		{
207 			/* XXX: To validate MH only once for each packet,
208 			 * this is placed here. It should be after checking
209 			 * xfrm policy, however it doesn't. The checking xfrm
210 			 * policy is placed in rawv6_rcv() because it is
211 			 * required for each socket.
212 			 */
213 			int (*filter)(struct sock *sock, struct sk_buff *skb);
214 
215 			filter = rcu_dereference(mh_filter);
216 			filtered = filter ? filter(sk, skb) : 0;
217 			break;
218 		}
219 #endif
220 		default:
221 			filtered = 0;
222 			break;
223 		}
224 
225 		if (filtered < 0)
226 			break;
227 		if (filtered == 0) {
228 			struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
229 
230 			/* Not releasing hash table! */
231 			if (clone) {
232 				nf_reset(clone);
233 				rawv6_rcv(sk, clone);
234 			}
235 		}
236 		sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
237 				     IP6CB(skb)->iif);
238 	}
239 out:
240 	read_unlock(&raw_v6_lock);
241 	return delivered;
242 }
243 
244 /* This cleans up af_inet6 a bit. -DaveM */
245 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
246 {
247 	struct inet_sock *inet = inet_sk(sk);
248 	struct ipv6_pinfo *np = inet6_sk(sk);
249 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
250 	__be32 v4addr = 0;
251 	int addr_type;
252 	int err;
253 
254 	if (addr_len < SIN6_LEN_RFC2133)
255 		return -EINVAL;
256 	addr_type = ipv6_addr_type(&addr->sin6_addr);
257 
258 	/* Raw sockets are IPv6 only */
259 	if (addr_type == IPV6_ADDR_MAPPED)
260 		return(-EADDRNOTAVAIL);
261 
262 	lock_sock(sk);
263 
264 	err = -EINVAL;
265 	if (sk->sk_state != TCP_CLOSE)
266 		goto out;
267 
268 	/* Check if the address belongs to the host. */
269 	if (addr_type != IPV6_ADDR_ANY) {
270 		struct net_device *dev = NULL;
271 
272 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
273 			if (addr_len >= sizeof(struct sockaddr_in6) &&
274 			    addr->sin6_scope_id) {
275 				/* Override any existing binding, if another
276 				 * one is supplied by user.
277 				 */
278 				sk->sk_bound_dev_if = addr->sin6_scope_id;
279 			}
280 
281 			/* Binding to link-local address requires an interface */
282 			if (!sk->sk_bound_dev_if)
283 				goto out;
284 
285 			dev = dev_get_by_index(sk->sk_bound_dev_if);
286 			if (!dev) {
287 				err = -ENODEV;
288 				goto out;
289 			}
290 		}
291 
292 		/* ipv4 addr of the socket is invalid.  Only the
293 		 * unspecified and mapped address have a v4 equivalent.
294 		 */
295 		v4addr = LOOPBACK4_IPV6;
296 		if (!(addr_type & IPV6_ADDR_MULTICAST))	{
297 			err = -EADDRNOTAVAIL;
298 			if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
299 				if (dev)
300 					dev_put(dev);
301 				goto out;
302 			}
303 		}
304 		if (dev)
305 			dev_put(dev);
306 	}
307 
308 	inet->rcv_saddr = inet->saddr = v4addr;
309 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
310 	if (!(addr_type & IPV6_ADDR_MULTICAST))
311 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
312 	err = 0;
313 out:
314 	release_sock(sk);
315 	return err;
316 }
317 
318 void rawv6_err(struct sock *sk, struct sk_buff *skb,
319 	       struct inet6_skb_parm *opt,
320 	       int type, int code, int offset, __be32 info)
321 {
322 	struct inet_sock *inet = inet_sk(sk);
323 	struct ipv6_pinfo *np = inet6_sk(sk);
324 	int err;
325 	int harderr;
326 
327 	/* Report error on raw socket, if:
328 	   1. User requested recverr.
329 	   2. Socket is connected (otherwise the error indication
330 	      is useless without recverr and error is hard.
331 	 */
332 	if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
333 		return;
334 
335 	harderr = icmpv6_err_convert(type, code, &err);
336 	if (type == ICMPV6_PKT_TOOBIG)
337 		harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
338 
339 	if (np->recverr) {
340 		u8 *payload = skb->data;
341 		if (!inet->hdrincl)
342 			payload += offset;
343 		ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
344 	}
345 
346 	if (np->recverr || harderr) {
347 		sk->sk_err = err;
348 		sk->sk_error_report(sk);
349 	}
350 }
351 
352 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
353 {
354 	if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
355 	    skb_checksum_complete(skb)) {
356 		/* FIXME: increment a raw6 drops counter here */
357 		kfree_skb(skb);
358 		return 0;
359 	}
360 
361 	/* Charge it to the socket. */
362 	if (sock_queue_rcv_skb(sk,skb)<0) {
363 		/* FIXME: increment a raw6 drops counter here */
364 		kfree_skb(skb);
365 		return 0;
366 	}
367 
368 	return 0;
369 }
370 
371 /*
372  *	This is next to useless...
373  *	if we demultiplex in network layer we don't need the extra call
374  *	just to queue the skb...
375  *	maybe we could have the network decide upon a hint if it
376  *	should call raw_rcv for demultiplexing
377  */
378 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
379 {
380 	struct inet_sock *inet = inet_sk(sk);
381 	struct raw6_sock *rp = raw6_sk(sk);
382 
383 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
384 		kfree_skb(skb);
385 		return NET_RX_DROP;
386 	}
387 
388 	if (!rp->checksum)
389 		skb->ip_summed = CHECKSUM_UNNECESSARY;
390 
391 	if (skb->ip_summed == CHECKSUM_COMPLETE) {
392 		skb_postpull_rcsum(skb, skb_network_header(skb),
393 				   skb_network_header_len(skb));
394 		if (!csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
395 				     &ipv6_hdr(skb)->daddr,
396 				     skb->len, inet->num, skb->csum))
397 			skb->ip_summed = CHECKSUM_UNNECESSARY;
398 	}
399 	if (!skb_csum_unnecessary(skb))
400 		skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
401 							 &ipv6_hdr(skb)->daddr,
402 							 skb->len,
403 							 inet->num, 0));
404 
405 	if (inet->hdrincl) {
406 		if (skb_checksum_complete(skb)) {
407 			/* FIXME: increment a raw6 drops counter here */
408 			kfree_skb(skb);
409 			return 0;
410 		}
411 	}
412 
413 	rawv6_rcv_skb(sk, skb);
414 	return 0;
415 }
416 
417 
418 /*
419  *	This should be easy, if there is something there
420  *	we return it, otherwise we block.
421  */
422 
423 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
424 		  struct msghdr *msg, size_t len,
425 		  int noblock, int flags, int *addr_len)
426 {
427 	struct ipv6_pinfo *np = inet6_sk(sk);
428 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
429 	struct sk_buff *skb;
430 	size_t copied;
431 	int err;
432 
433 	if (flags & MSG_OOB)
434 		return -EOPNOTSUPP;
435 
436 	if (addr_len)
437 		*addr_len=sizeof(*sin6);
438 
439 	if (flags & MSG_ERRQUEUE)
440 		return ipv6_recv_error(sk, msg, len);
441 
442 	skb = skb_recv_datagram(sk, flags, noblock, &err);
443 	if (!skb)
444 		goto out;
445 
446 	copied = skb->len;
447 	if (copied > len) {
448 		copied = len;
449 		msg->msg_flags |= MSG_TRUNC;
450 	}
451 
452 	if (skb_csum_unnecessary(skb)) {
453 		err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
454 	} else if (msg->msg_flags&MSG_TRUNC) {
455 		if (__skb_checksum_complete(skb))
456 			goto csum_copy_err;
457 		err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
458 	} else {
459 		err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
460 		if (err == -EINVAL)
461 			goto csum_copy_err;
462 	}
463 	if (err)
464 		goto out_free;
465 
466 	/* Copy the address. */
467 	if (sin6) {
468 		sin6->sin6_family = AF_INET6;
469 		sin6->sin6_port = 0;
470 		ipv6_addr_copy(&sin6->sin6_addr, &ipv6_hdr(skb)->saddr);
471 		sin6->sin6_flowinfo = 0;
472 		sin6->sin6_scope_id = 0;
473 		if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
474 			sin6->sin6_scope_id = IP6CB(skb)->iif;
475 	}
476 
477 	sock_recv_timestamp(msg, sk, skb);
478 
479 	if (np->rxopt.all)
480 		datagram_recv_ctl(sk, msg, skb);
481 
482 	err = copied;
483 	if (flags & MSG_TRUNC)
484 		err = skb->len;
485 
486 out_free:
487 	skb_free_datagram(sk, skb);
488 out:
489 	return err;
490 
491 csum_copy_err:
492 	skb_kill_datagram(sk, skb, flags);
493 
494 	/* Error for blocking case is chosen to masquerade
495 	   as some normal condition.
496 	 */
497 	err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
498 	/* FIXME: increment a raw6 drops counter here */
499 	goto out;
500 }
501 
502 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
503 				     struct raw6_sock *rp)
504 {
505 	struct sk_buff *skb;
506 	int err = 0;
507 	int offset;
508 	int len;
509 	int total_len;
510 	__wsum tmp_csum;
511 	__sum16 csum;
512 
513 	if (!rp->checksum)
514 		goto send;
515 
516 	if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
517 		goto out;
518 
519 	offset = rp->offset;
520 	total_len = inet_sk(sk)->cork.length - (skb_network_header(skb) -
521 						skb->data);
522 	if (offset >= total_len - 1) {
523 		err = -EINVAL;
524 		ip6_flush_pending_frames(sk);
525 		goto out;
526 	}
527 
528 	/* should be check HW csum miyazawa */
529 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
530 		/*
531 		 * Only one fragment on the socket.
532 		 */
533 		tmp_csum = skb->csum;
534 	} else {
535 		struct sk_buff *csum_skb = NULL;
536 		tmp_csum = 0;
537 
538 		skb_queue_walk(&sk->sk_write_queue, skb) {
539 			tmp_csum = csum_add(tmp_csum, skb->csum);
540 
541 			if (csum_skb)
542 				continue;
543 
544 			len = skb->len - skb_transport_offset(skb);
545 			if (offset >= len) {
546 				offset -= len;
547 				continue;
548 			}
549 
550 			csum_skb = skb;
551 		}
552 
553 		skb = csum_skb;
554 	}
555 
556 	offset += skb_transport_offset(skb);
557 	if (skb_copy_bits(skb, offset, &csum, 2))
558 		BUG();
559 
560 	/* in case cksum was not initialized */
561 	if (unlikely(csum))
562 		tmp_csum = csum_sub(tmp_csum, csum_unfold(csum));
563 
564 	csum = csum_ipv6_magic(&fl->fl6_src,
565 				   &fl->fl6_dst,
566 				   total_len, fl->proto, tmp_csum);
567 
568 	if (csum == 0 && fl->proto == IPPROTO_UDP)
569 		csum = CSUM_MANGLED_0;
570 
571 	if (skb_store_bits(skb, offset, &csum, 2))
572 		BUG();
573 
574 send:
575 	err = ip6_push_pending_frames(sk);
576 out:
577 	return err;
578 }
579 
580 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
581 			struct flowi *fl, struct rt6_info *rt,
582 			unsigned int flags)
583 {
584 	struct ipv6_pinfo *np = inet6_sk(sk);
585 	struct ipv6hdr *iph;
586 	struct sk_buff *skb;
587 	unsigned int hh_len;
588 	int err;
589 
590 	if (length > rt->u.dst.dev->mtu) {
591 		ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
592 		return -EMSGSIZE;
593 	}
594 	if (flags&MSG_PROBE)
595 		goto out;
596 
597 	hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
598 
599 	skb = sock_alloc_send_skb(sk, length+hh_len+15,
600 				  flags&MSG_DONTWAIT, &err);
601 	if (skb == NULL)
602 		goto error;
603 	skb_reserve(skb, hh_len);
604 
605 	skb->priority = sk->sk_priority;
606 	skb->dst = dst_clone(&rt->u.dst);
607 
608 	skb_put(skb, length);
609 	skb_reset_network_header(skb);
610 	iph = ipv6_hdr(skb);
611 
612 	skb->ip_summed = CHECKSUM_NONE;
613 
614 	skb->transport_header = skb->network_header;
615 	err = memcpy_fromiovecend((void *)iph, from, 0, length);
616 	if (err)
617 		goto error_fault;
618 
619 	IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTREQUESTS);
620 	err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
621 		      dst_output);
622 	if (err > 0)
623 		err = np->recverr ? net_xmit_errno(err) : 0;
624 	if (err)
625 		goto error;
626 out:
627 	return 0;
628 
629 error_fault:
630 	err = -EFAULT;
631 	kfree_skb(skb);
632 error:
633 	IP6_INC_STATS(rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
634 	return err;
635 }
636 
637 static int rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
638 {
639 	struct iovec *iov;
640 	u8 __user *type = NULL;
641 	u8 __user *code = NULL;
642 	u8 len = 0;
643 	int probed = 0;
644 	int i;
645 
646 	if (!msg->msg_iov)
647 		return 0;
648 
649 	for (i = 0; i < msg->msg_iovlen; i++) {
650 		iov = &msg->msg_iov[i];
651 		if (!iov)
652 			continue;
653 
654 		switch (fl->proto) {
655 		case IPPROTO_ICMPV6:
656 			/* check if one-byte field is readable or not. */
657 			if (iov->iov_base && iov->iov_len < 1)
658 				break;
659 
660 			if (!type) {
661 				type = iov->iov_base;
662 				/* check if code field is readable or not. */
663 				if (iov->iov_len > 1)
664 					code = type + 1;
665 			} else if (!code)
666 				code = iov->iov_base;
667 
668 			if (type && code) {
669 				if (get_user(fl->fl_icmp_type, type) ||
670 				    get_user(fl->fl_icmp_code, code))
671 					return -EFAULT;
672 				probed = 1;
673 			}
674 			break;
675 		case IPPROTO_MH:
676 			if (iov->iov_base && iov->iov_len < 1)
677 				break;
678 			/* check if type field is readable or not. */
679 			if (iov->iov_len > 2 - len) {
680 				u8 __user *p = iov->iov_base;
681 				if (get_user(fl->fl_mh_type, &p[2 - len]))
682 					return -EFAULT;
683 				probed = 1;
684 			} else
685 				len += iov->iov_len;
686 
687 			break;
688 		default:
689 			probed = 1;
690 			break;
691 		}
692 		if (probed)
693 			break;
694 	}
695 	return 0;
696 }
697 
698 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
699 		   struct msghdr *msg, size_t len)
700 {
701 	struct ipv6_txoptions opt_space;
702 	struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
703 	struct in6_addr *daddr, *final_p = NULL, final;
704 	struct inet_sock *inet = inet_sk(sk);
705 	struct ipv6_pinfo *np = inet6_sk(sk);
706 	struct raw6_sock *rp = raw6_sk(sk);
707 	struct ipv6_txoptions *opt = NULL;
708 	struct ip6_flowlabel *flowlabel = NULL;
709 	struct dst_entry *dst = NULL;
710 	struct flowi fl;
711 	int addr_len = msg->msg_namelen;
712 	int hlimit = -1;
713 	int tclass = -1;
714 	u16 proto;
715 	int err;
716 
717 	/* Rough check on arithmetic overflow,
718 	   better check is made in ip6_append_data().
719 	 */
720 	if (len > INT_MAX)
721 		return -EMSGSIZE;
722 
723 	/* Mirror BSD error message compatibility */
724 	if (msg->msg_flags & MSG_OOB)
725 		return -EOPNOTSUPP;
726 
727 	/*
728 	 *	Get and verify the address.
729 	 */
730 	memset(&fl, 0, sizeof(fl));
731 
732 	if (sin6) {
733 		if (addr_len < SIN6_LEN_RFC2133)
734 			return -EINVAL;
735 
736 		if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
737 			return(-EAFNOSUPPORT);
738 
739 		/* port is the proto value [0..255] carried in nexthdr */
740 		proto = ntohs(sin6->sin6_port);
741 
742 		if (!proto)
743 			proto = inet->num;
744 		else if (proto != inet->num)
745 			return(-EINVAL);
746 
747 		if (proto > 255)
748 			return(-EINVAL);
749 
750 		daddr = &sin6->sin6_addr;
751 		if (np->sndflow) {
752 			fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
753 			if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
754 				flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
755 				if (flowlabel == NULL)
756 					return -EINVAL;
757 				daddr = &flowlabel->dst;
758 			}
759 		}
760 
761 		/*
762 		 * Otherwise it will be difficult to maintain
763 		 * sk->sk_dst_cache.
764 		 */
765 		if (sk->sk_state == TCP_ESTABLISHED &&
766 		    ipv6_addr_equal(daddr, &np->daddr))
767 			daddr = &np->daddr;
768 
769 		if (addr_len >= sizeof(struct sockaddr_in6) &&
770 		    sin6->sin6_scope_id &&
771 		    ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
772 			fl.oif = sin6->sin6_scope_id;
773 	} else {
774 		if (sk->sk_state != TCP_ESTABLISHED)
775 			return -EDESTADDRREQ;
776 
777 		proto = inet->num;
778 		daddr = &np->daddr;
779 		fl.fl6_flowlabel = np->flow_label;
780 	}
781 
782 	if (ipv6_addr_any(daddr)) {
783 		/*
784 		 * unspecified destination address
785 		 * treated as error... is this correct ?
786 		 */
787 		fl6_sock_release(flowlabel);
788 		return(-EINVAL);
789 	}
790 
791 	if (fl.oif == 0)
792 		fl.oif = sk->sk_bound_dev_if;
793 
794 	if (msg->msg_controllen) {
795 		opt = &opt_space;
796 		memset(opt, 0, sizeof(struct ipv6_txoptions));
797 		opt->tot_len = sizeof(struct ipv6_txoptions);
798 
799 		err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
800 		if (err < 0) {
801 			fl6_sock_release(flowlabel);
802 			return err;
803 		}
804 		if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
805 			flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
806 			if (flowlabel == NULL)
807 				return -EINVAL;
808 		}
809 		if (!(opt->opt_nflen|opt->opt_flen))
810 			opt = NULL;
811 	}
812 	if (opt == NULL)
813 		opt = np->opt;
814 	if (flowlabel)
815 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
816 	opt = ipv6_fixup_options(&opt_space, opt);
817 
818 	fl.proto = proto;
819 	err = rawv6_probe_proto_opt(&fl, msg);
820 	if (err)
821 		goto out;
822 
823 	ipv6_addr_copy(&fl.fl6_dst, daddr);
824 	if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
825 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
826 
827 	/* merge ip6_build_xmit from ip6_output */
828 	if (opt && opt->srcrt) {
829 		struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
830 		ipv6_addr_copy(&final, &fl.fl6_dst);
831 		ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
832 		final_p = &final;
833 	}
834 
835 	if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
836 		fl.oif = np->mcast_oif;
837 	security_sk_classify_flow(sk, &fl);
838 
839 	err = ip6_dst_lookup(sk, &dst, &fl);
840 	if (err)
841 		goto out;
842 	if (final_p)
843 		ipv6_addr_copy(&fl.fl6_dst, final_p);
844 
845 	if ((err = __xfrm_lookup(&dst, &fl, sk, 1)) < 0) {
846 		if (err == -EREMOTE)
847 			err = ip6_dst_blackhole(sk, &dst, &fl);
848 		if (err < 0)
849 			goto out;
850 	}
851 
852 	if (hlimit < 0) {
853 		if (ipv6_addr_is_multicast(&fl.fl6_dst))
854 			hlimit = np->mcast_hops;
855 		else
856 			hlimit = np->hop_limit;
857 		if (hlimit < 0)
858 			hlimit = dst_metric(dst, RTAX_HOPLIMIT);
859 		if (hlimit < 0)
860 			hlimit = ipv6_get_hoplimit(dst->dev);
861 	}
862 
863 	if (tclass < 0) {
864 		tclass = np->tclass;
865 		if (tclass < 0)
866 			tclass = 0;
867 	}
868 
869 	if (msg->msg_flags&MSG_CONFIRM)
870 		goto do_confirm;
871 
872 back_from_confirm:
873 	if (inet->hdrincl) {
874 		err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
875 	} else {
876 		lock_sock(sk);
877 		err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
878 			len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
879 			msg->msg_flags);
880 
881 		if (err)
882 			ip6_flush_pending_frames(sk);
883 		else if (!(msg->msg_flags & MSG_MORE))
884 			err = rawv6_push_pending_frames(sk, &fl, rp);
885 	}
886 done:
887 	dst_release(dst);
888 	if (!inet->hdrincl)
889 		release_sock(sk);
890 out:
891 	fl6_sock_release(flowlabel);
892 	return err<0?err:len;
893 do_confirm:
894 	dst_confirm(dst);
895 	if (!(msg->msg_flags & MSG_PROBE) || len)
896 		goto back_from_confirm;
897 	err = 0;
898 	goto done;
899 }
900 
901 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
902 			       char __user *optval, int optlen)
903 {
904 	switch (optname) {
905 	case ICMPV6_FILTER:
906 		if (optlen > sizeof(struct icmp6_filter))
907 			optlen = sizeof(struct icmp6_filter);
908 		if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
909 			return -EFAULT;
910 		return 0;
911 	default:
912 		return -ENOPROTOOPT;
913 	}
914 
915 	return 0;
916 }
917 
918 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
919 			       char __user *optval, int __user *optlen)
920 {
921 	int len;
922 
923 	switch (optname) {
924 	case ICMPV6_FILTER:
925 		if (get_user(len, optlen))
926 			return -EFAULT;
927 		if (len < 0)
928 			return -EINVAL;
929 		if (len > sizeof(struct icmp6_filter))
930 			len = sizeof(struct icmp6_filter);
931 		if (put_user(len, optlen))
932 			return -EFAULT;
933 		if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
934 			return -EFAULT;
935 		return 0;
936 	default:
937 		return -ENOPROTOOPT;
938 	}
939 
940 	return 0;
941 }
942 
943 
944 static int do_rawv6_setsockopt(struct sock *sk, int level, int optname,
945 			    char __user *optval, int optlen)
946 {
947 	struct raw6_sock *rp = raw6_sk(sk);
948 	int val;
949 
950 	if (get_user(val, (int __user *)optval))
951 		return -EFAULT;
952 
953 	switch (optname) {
954 		case IPV6_CHECKSUM:
955 			/* You may get strange result with a positive odd offset;
956 			   RFC2292bis agrees with me. */
957 			if (val > 0 && (val&1))
958 				return(-EINVAL);
959 			if (val < 0) {
960 				rp->checksum = 0;
961 			} else {
962 				rp->checksum = 1;
963 				rp->offset = val;
964 			}
965 
966 			return 0;
967 			break;
968 
969 		default:
970 			return(-ENOPROTOOPT);
971 	}
972 }
973 
974 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
975 			  char __user *optval, int optlen)
976 {
977 	switch(level) {
978 		case SOL_RAW:
979 			break;
980 
981 		case SOL_ICMPV6:
982 			if (inet_sk(sk)->num != IPPROTO_ICMPV6)
983 				return -EOPNOTSUPP;
984 			return rawv6_seticmpfilter(sk, level, optname, optval,
985 						   optlen);
986 		case SOL_IPV6:
987 			if (optname == IPV6_CHECKSUM)
988 				break;
989 		default:
990 			return ipv6_setsockopt(sk, level, optname, optval,
991 					       optlen);
992 	}
993 
994 	return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
995 }
996 
997 #ifdef CONFIG_COMPAT
998 static int compat_rawv6_setsockopt(struct sock *sk, int level, int optname,
999 				   char __user *optval, int optlen)
1000 {
1001 	switch (level) {
1002 	case SOL_RAW:
1003 		break;
1004 	case SOL_ICMPV6:
1005 		if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1006 			return -EOPNOTSUPP;
1007 		return rawv6_seticmpfilter(sk, level, optname, optval, optlen);
1008 	case SOL_IPV6:
1009 		if (optname == IPV6_CHECKSUM)
1010 			break;
1011 	default:
1012 		return compat_ipv6_setsockopt(sk, level, optname,
1013 					      optval, optlen);
1014 	}
1015 	return do_rawv6_setsockopt(sk, level, optname, optval, optlen);
1016 }
1017 #endif
1018 
1019 static int do_rawv6_getsockopt(struct sock *sk, int level, int optname,
1020 			    char __user *optval, int __user *optlen)
1021 {
1022 	struct raw6_sock *rp = raw6_sk(sk);
1023 	int val, len;
1024 
1025 	if (get_user(len,optlen))
1026 		return -EFAULT;
1027 
1028 	switch (optname) {
1029 	case IPV6_CHECKSUM:
1030 		if (rp->checksum == 0)
1031 			val = -1;
1032 		else
1033 			val = rp->offset;
1034 		break;
1035 
1036 	default:
1037 		return -ENOPROTOOPT;
1038 	}
1039 
1040 	len = min_t(unsigned int, sizeof(int), len);
1041 
1042 	if (put_user(len, optlen))
1043 		return -EFAULT;
1044 	if (copy_to_user(optval,&val,len))
1045 		return -EFAULT;
1046 	return 0;
1047 }
1048 
1049 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
1050 			  char __user *optval, int __user *optlen)
1051 {
1052 	switch(level) {
1053 		case SOL_RAW:
1054 			break;
1055 
1056 		case SOL_ICMPV6:
1057 			if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1058 				return -EOPNOTSUPP;
1059 			return rawv6_geticmpfilter(sk, level, optname, optval,
1060 						   optlen);
1061 		case SOL_IPV6:
1062 			if (optname == IPV6_CHECKSUM)
1063 				break;
1064 		default:
1065 			return ipv6_getsockopt(sk, level, optname, optval,
1066 					       optlen);
1067 	}
1068 
1069 	return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1070 }
1071 
1072 #ifdef CONFIG_COMPAT
1073 static int compat_rawv6_getsockopt(struct sock *sk, int level, int optname,
1074 				   char __user *optval, int __user *optlen)
1075 {
1076 	switch (level) {
1077 	case SOL_RAW:
1078 		break;
1079 	case SOL_ICMPV6:
1080 		if (inet_sk(sk)->num != IPPROTO_ICMPV6)
1081 			return -EOPNOTSUPP;
1082 		return rawv6_geticmpfilter(sk, level, optname, optval, optlen);
1083 	case SOL_IPV6:
1084 		if (optname == IPV6_CHECKSUM)
1085 			break;
1086 	default:
1087 		return compat_ipv6_getsockopt(sk, level, optname,
1088 					      optval, optlen);
1089 	}
1090 	return do_rawv6_getsockopt(sk, level, optname, optval, optlen);
1091 }
1092 #endif
1093 
1094 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
1095 {
1096 	switch(cmd) {
1097 		case SIOCOUTQ:
1098 		{
1099 			int amount = atomic_read(&sk->sk_wmem_alloc);
1100 			return put_user(amount, (int __user *)arg);
1101 		}
1102 		case SIOCINQ:
1103 		{
1104 			struct sk_buff *skb;
1105 			int amount = 0;
1106 
1107 			spin_lock_bh(&sk->sk_receive_queue.lock);
1108 			skb = skb_peek(&sk->sk_receive_queue);
1109 			if (skb != NULL)
1110 				amount = skb->tail - skb->transport_header;
1111 			spin_unlock_bh(&sk->sk_receive_queue.lock);
1112 			return put_user(amount, (int __user *)arg);
1113 		}
1114 
1115 		default:
1116 			return -ENOIOCTLCMD;
1117 	}
1118 }
1119 
1120 static void rawv6_close(struct sock *sk, long timeout)
1121 {
1122 	if (inet_sk(sk)->num == IPPROTO_RAW)
1123 		ip6_ra_control(sk, -1, NULL);
1124 
1125 	sk_common_release(sk);
1126 }
1127 
1128 static int rawv6_init_sk(struct sock *sk)
1129 {
1130 	struct raw6_sock *rp = raw6_sk(sk);
1131 
1132 	switch (inet_sk(sk)->num) {
1133 	case IPPROTO_ICMPV6:
1134 		rp->checksum = 1;
1135 		rp->offset   = 2;
1136 		break;
1137 	case IPPROTO_MH:
1138 		rp->checksum = 1;
1139 		rp->offset   = 4;
1140 		break;
1141 	default:
1142 		break;
1143 	}
1144 	return(0);
1145 }
1146 
1147 struct proto rawv6_prot = {
1148 	.name		   = "RAWv6",
1149 	.owner		   = THIS_MODULE,
1150 	.close		   = rawv6_close,
1151 	.connect	   = ip6_datagram_connect,
1152 	.disconnect	   = udp_disconnect,
1153 	.ioctl		   = rawv6_ioctl,
1154 	.init		   = rawv6_init_sk,
1155 	.destroy	   = inet6_destroy_sock,
1156 	.setsockopt	   = rawv6_setsockopt,
1157 	.getsockopt	   = rawv6_getsockopt,
1158 	.sendmsg	   = rawv6_sendmsg,
1159 	.recvmsg	   = rawv6_recvmsg,
1160 	.bind		   = rawv6_bind,
1161 	.backlog_rcv	   = rawv6_rcv_skb,
1162 	.hash		   = raw_v6_hash,
1163 	.unhash		   = raw_v6_unhash,
1164 	.obj_size	   = sizeof(struct raw6_sock),
1165 #ifdef CONFIG_COMPAT
1166 	.compat_setsockopt = compat_rawv6_setsockopt,
1167 	.compat_getsockopt = compat_rawv6_getsockopt,
1168 #endif
1169 };
1170 
1171 #ifdef CONFIG_PROC_FS
1172 struct raw6_iter_state {
1173 	int bucket;
1174 };
1175 
1176 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1177 
1178 static struct sock *raw6_get_first(struct seq_file *seq)
1179 {
1180 	struct sock *sk;
1181 	struct hlist_node *node;
1182 	struct raw6_iter_state* state = raw6_seq_private(seq);
1183 
1184 	for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1185 		sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1186 			if (sk->sk_family == PF_INET6)
1187 				goto out;
1188 	sk = NULL;
1189 out:
1190 	return sk;
1191 }
1192 
1193 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1194 {
1195 	struct raw6_iter_state* state = raw6_seq_private(seq);
1196 
1197 	do {
1198 		sk = sk_next(sk);
1199 try_again:
1200 		;
1201 	} while (sk && sk->sk_family != PF_INET6);
1202 
1203 	if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1204 		sk = sk_head(&raw_v6_htable[state->bucket]);
1205 		goto try_again;
1206 	}
1207 	return sk;
1208 }
1209 
1210 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1211 {
1212 	struct sock *sk = raw6_get_first(seq);
1213 	if (sk)
1214 		while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1215 			--pos;
1216 	return pos ? NULL : sk;
1217 }
1218 
1219 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1220 {
1221 	read_lock(&raw_v6_lock);
1222 	return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1223 }
1224 
1225 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1226 {
1227 	struct sock *sk;
1228 
1229 	if (v == SEQ_START_TOKEN)
1230 		sk = raw6_get_first(seq);
1231 	else
1232 		sk = raw6_get_next(seq, v);
1233 	++*pos;
1234 	return sk;
1235 }
1236 
1237 static void raw6_seq_stop(struct seq_file *seq, void *v)
1238 {
1239 	read_unlock(&raw_v6_lock);
1240 }
1241 
1242 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1243 {
1244 	struct ipv6_pinfo *np = inet6_sk(sp);
1245 	struct in6_addr *dest, *src;
1246 	__u16 destp, srcp;
1247 
1248 	dest  = &np->daddr;
1249 	src   = &np->rcv_saddr;
1250 	destp = 0;
1251 	srcp  = inet_sk(sp)->num;
1252 	seq_printf(seq,
1253 		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1254 		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1255 		   i,
1256 		   src->s6_addr32[0], src->s6_addr32[1],
1257 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1258 		   dest->s6_addr32[0], dest->s6_addr32[1],
1259 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1260 		   sp->sk_state,
1261 		   atomic_read(&sp->sk_wmem_alloc),
1262 		   atomic_read(&sp->sk_rmem_alloc),
1263 		   0, 0L, 0,
1264 		   sock_i_uid(sp), 0,
1265 		   sock_i_ino(sp),
1266 		   atomic_read(&sp->sk_refcnt), sp);
1267 }
1268 
1269 static int raw6_seq_show(struct seq_file *seq, void *v)
1270 {
1271 	if (v == SEQ_START_TOKEN)
1272 		seq_printf(seq,
1273 			   "  sl  "
1274 			   "local_address                         "
1275 			   "remote_address                        "
1276 			   "st tx_queue rx_queue tr tm->when retrnsmt"
1277 			   "   uid  timeout inode\n");
1278 	else
1279 		raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1280 	return 0;
1281 }
1282 
1283 static const struct seq_operations raw6_seq_ops = {
1284 	.start =	raw6_seq_start,
1285 	.next =		raw6_seq_next,
1286 	.stop =		raw6_seq_stop,
1287 	.show =		raw6_seq_show,
1288 };
1289 
1290 static int raw6_seq_open(struct inode *inode, struct file *file)
1291 {
1292 	struct seq_file *seq;
1293 	int rc = -ENOMEM;
1294 	struct raw6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
1295 	if (!s)
1296 		goto out;
1297 	rc = seq_open(file, &raw6_seq_ops);
1298 	if (rc)
1299 		goto out_kfree;
1300 	seq = file->private_data;
1301 	seq->private = s;
1302 out:
1303 	return rc;
1304 out_kfree:
1305 	kfree(s);
1306 	goto out;
1307 }
1308 
1309 static const struct file_operations raw6_seq_fops = {
1310 	.owner =	THIS_MODULE,
1311 	.open =		raw6_seq_open,
1312 	.read =		seq_read,
1313 	.llseek =	seq_lseek,
1314 	.release =	seq_release_private,
1315 };
1316 
1317 int __init raw6_proc_init(void)
1318 {
1319 	if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1320 		return -ENOMEM;
1321 	return 0;
1322 }
1323 
1324 void raw6_proc_exit(void)
1325 {
1326 	proc_net_remove("raw6");
1327 }
1328 #endif	/* CONFIG_PROC_FS */
1329