xref: /openbmc/linux/net/ipv6/raw.c (revision 87c2ce3b)
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/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 <linux/skbuff.h>
36 #include <asm/uaccess.h>
37 #include <asm/ioctls.h>
38 #include <asm/bug.h>
39 
40 #include <net/ip.h>
41 #include <net/sock.h>
42 #include <net/snmp.h>
43 
44 #include <net/ipv6.h>
45 #include <net/ndisc.h>
46 #include <net/protocol.h>
47 #include <net/ip6_route.h>
48 #include <net/ip6_checksum.h>
49 #include <net/addrconf.h>
50 #include <net/transp_v6.h>
51 #include <net/udp.h>
52 #include <net/inet_common.h>
53 #include <net/tcp_states.h>
54 
55 #include <net/rawv6.h>
56 #include <net/xfrm.h>
57 
58 #include <linux/proc_fs.h>
59 #include <linux/seq_file.h>
60 
61 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
62 DEFINE_RWLOCK(raw_v6_lock);
63 
64 static void raw_v6_hash(struct sock *sk)
65 {
66 	struct hlist_head *list = &raw_v6_htable[inet_sk(sk)->num &
67 						 (RAWV6_HTABLE_SIZE - 1)];
68 
69 	write_lock_bh(&raw_v6_lock);
70 	sk_add_node(sk, list);
71 	sock_prot_inc_use(sk->sk_prot);
72  	write_unlock_bh(&raw_v6_lock);
73 }
74 
75 static void raw_v6_unhash(struct sock *sk)
76 {
77  	write_lock_bh(&raw_v6_lock);
78 	if (sk_del_node_init(sk))
79 		sock_prot_dec_use(sk->sk_prot);
80 	write_unlock_bh(&raw_v6_lock);
81 }
82 
83 
84 /* Grumble... icmp and ip_input want to get at this... */
85 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
86 			     struct in6_addr *loc_addr, struct in6_addr *rmt_addr,
87 			     int dif)
88 {
89 	struct hlist_node *node;
90 	int is_multicast = ipv6_addr_is_multicast(loc_addr);
91 
92 	sk_for_each_from(sk, node)
93 		if (inet_sk(sk)->num == num) {
94 			struct ipv6_pinfo *np = inet6_sk(sk);
95 
96 			if (!ipv6_addr_any(&np->daddr) &&
97 			    !ipv6_addr_equal(&np->daddr, rmt_addr))
98 				continue;
99 
100 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
101 				continue;
102 
103 			if (!ipv6_addr_any(&np->rcv_saddr)) {
104 				if (ipv6_addr_equal(&np->rcv_saddr, loc_addr))
105 					goto found;
106 				if (is_multicast &&
107 				    inet6_mc_check(sk, loc_addr, rmt_addr))
108 					goto found;
109 				continue;
110 			}
111 			goto found;
112 		}
113 	sk = NULL;
114 found:
115 	return sk;
116 }
117 
118 /*
119  *	0 - deliver
120  *	1 - block
121  */
122 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
123 {
124 	struct icmp6hdr *icmph;
125 	struct raw6_sock *rp = raw6_sk(sk);
126 
127 	if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
128 		__u32 *data = &rp->filter.data[0];
129 		int bit_nr;
130 
131 		icmph = (struct icmp6hdr *) skb->data;
132 		bit_nr = icmph->icmp6_type;
133 
134 		return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
135 	}
136 	return 0;
137 }
138 
139 /*
140  *	demultiplex raw sockets.
141  *	(should consider queueing the skb in the sock receive_queue
142  *	without calling rawv6.c)
143  *
144  *	Caller owns SKB so we must make clones.
145  */
146 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
147 {
148 	struct in6_addr *saddr;
149 	struct in6_addr *daddr;
150 	struct sock *sk;
151 	int delivered = 0;
152 	__u8 hash;
153 
154 	saddr = &skb->nh.ipv6h->saddr;
155 	daddr = saddr + 1;
156 
157 	hash = nexthdr & (MAX_INET_PROTOS - 1);
158 
159 	read_lock(&raw_v6_lock);
160 	sk = sk_head(&raw_v6_htable[hash]);
161 
162 	/*
163 	 *	The first socket found will be delivered after
164 	 *	delivery to transport protocols.
165 	 */
166 
167 	if (sk == NULL)
168 		goto out;
169 
170 	sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
171 
172 	while (sk) {
173 		delivered = 1;
174 		if (nexthdr != IPPROTO_ICMPV6 || !icmpv6_filter(sk, skb)) {
175 			struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
176 
177 			/* Not releasing hash table! */
178 			if (clone) {
179 				nf_reset(clone);
180 				rawv6_rcv(sk, clone);
181 			}
182 		}
183 		sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
184 				     IP6CB(skb)->iif);
185 	}
186 out:
187 	read_unlock(&raw_v6_lock);
188 	return delivered;
189 }
190 
191 /* This cleans up af_inet6 a bit. -DaveM */
192 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
193 {
194 	struct inet_sock *inet = inet_sk(sk);
195 	struct ipv6_pinfo *np = inet6_sk(sk);
196 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
197 	__u32 v4addr = 0;
198 	int addr_type;
199 	int err;
200 
201 	if (addr_len < SIN6_LEN_RFC2133)
202 		return -EINVAL;
203 	addr_type = ipv6_addr_type(&addr->sin6_addr);
204 
205 	/* Raw sockets are IPv6 only */
206 	if (addr_type == IPV6_ADDR_MAPPED)
207 		return(-EADDRNOTAVAIL);
208 
209 	lock_sock(sk);
210 
211 	err = -EINVAL;
212 	if (sk->sk_state != TCP_CLOSE)
213 		goto out;
214 
215 	/* Check if the address belongs to the host. */
216 	if (addr_type != IPV6_ADDR_ANY) {
217 		struct net_device *dev = NULL;
218 
219 		if (addr_type & IPV6_ADDR_LINKLOCAL) {
220 			if (addr_len >= sizeof(struct sockaddr_in6) &&
221 			    addr->sin6_scope_id) {
222 				/* Override any existing binding, if another
223 				 * one is supplied by user.
224 				 */
225 				sk->sk_bound_dev_if = addr->sin6_scope_id;
226 			}
227 
228 			/* Binding to link-local address requires an interface */
229 			if (!sk->sk_bound_dev_if)
230 				goto out;
231 
232 			dev = dev_get_by_index(sk->sk_bound_dev_if);
233 			if (!dev) {
234 				err = -ENODEV;
235 				goto out;
236 			}
237 		}
238 
239 		/* ipv4 addr of the socket is invalid.  Only the
240 		 * unspecified and mapped address have a v4 equivalent.
241 		 */
242 		v4addr = LOOPBACK4_IPV6;
243 		if (!(addr_type & IPV6_ADDR_MULTICAST))	{
244 			err = -EADDRNOTAVAIL;
245 			if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
246 				if (dev)
247 					dev_put(dev);
248 				goto out;
249 			}
250 		}
251 		if (dev)
252 			dev_put(dev);
253 	}
254 
255 	inet->rcv_saddr = inet->saddr = v4addr;
256 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
257 	if (!(addr_type & IPV6_ADDR_MULTICAST))
258 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
259 	err = 0;
260 out:
261 	release_sock(sk);
262 	return err;
263 }
264 
265 void rawv6_err(struct sock *sk, struct sk_buff *skb,
266 	       struct inet6_skb_parm *opt,
267 	       int type, int code, int offset, u32 info)
268 {
269 	struct inet_sock *inet = inet_sk(sk);
270 	struct ipv6_pinfo *np = inet6_sk(sk);
271 	int err;
272 	int harderr;
273 
274 	/* Report error on raw socket, if:
275 	   1. User requested recverr.
276 	   2. Socket is connected (otherwise the error indication
277 	      is useless without recverr and error is hard.
278 	 */
279 	if (!np->recverr && sk->sk_state != TCP_ESTABLISHED)
280 		return;
281 
282 	harderr = icmpv6_err_convert(type, code, &err);
283 	if (type == ICMPV6_PKT_TOOBIG)
284 		harderr = (np->pmtudisc == IPV6_PMTUDISC_DO);
285 
286 	if (np->recverr) {
287 		u8 *payload = skb->data;
288 		if (!inet->hdrincl)
289 			payload += offset;
290 		ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
291 	}
292 
293 	if (np->recverr || harderr) {
294 		sk->sk_err = err;
295 		sk->sk_error_report(sk);
296 	}
297 }
298 
299 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
300 {
301 	if ((raw6_sk(sk)->checksum || sk->sk_filter) &&
302 	    skb_checksum_complete(skb)) {
303 		/* FIXME: increment a raw6 drops counter here */
304 		kfree_skb(skb);
305 		return 0;
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  */
325 int 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_HW) {
339 		skb_postpull_rcsum(skb, skb->nh.raw,
340 		                   skb->h.raw - skb->nh.raw);
341 		if (!csum_ipv6_magic(&skb->nh.ipv6h->saddr,
342 				     &skb->nh.ipv6h->daddr,
343 				     skb->len, inet->num, skb->csum))
344 			skb->ip_summed = CHECKSUM_UNNECESSARY;
345 	}
346 	if (skb->ip_summed != CHECKSUM_UNNECESSARY)
347 		skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
348 					     &skb->nh.ipv6h->daddr,
349 					     skb->len, inet->num, 0);
350 
351 	if (inet->hdrincl) {
352 		if (skb_checksum_complete(skb)) {
353 			/* FIXME: increment a raw6 drops counter here */
354 			kfree_skb(skb);
355 			return 0;
356 		}
357 	}
358 
359 	rawv6_rcv_skb(sk, skb);
360 	return 0;
361 }
362 
363 
364 /*
365  *	This should be easy, if there is something there
366  *	we return it, otherwise we block.
367  */
368 
369 static int rawv6_recvmsg(struct kiocb *iocb, struct sock *sk,
370 		  struct msghdr *msg, size_t len,
371 		  int noblock, int flags, int *addr_len)
372 {
373 	struct ipv6_pinfo *np = inet6_sk(sk);
374 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
375 	struct sk_buff *skb;
376 	size_t copied;
377 	int err;
378 
379 	if (flags & MSG_OOB)
380 		return -EOPNOTSUPP;
381 
382 	if (addr_len)
383 		*addr_len=sizeof(*sin6);
384 
385 	if (flags & MSG_ERRQUEUE)
386 		return ipv6_recv_error(sk, msg, len);
387 
388 	skb = skb_recv_datagram(sk, flags, noblock, &err);
389 	if (!skb)
390 		goto out;
391 
392 	copied = skb->len;
393   	if (copied > len) {
394   		copied = len;
395   		msg->msg_flags |= MSG_TRUNC;
396   	}
397 
398 	if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
399 		err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
400 	} else if (msg->msg_flags&MSG_TRUNC) {
401 		if (__skb_checksum_complete(skb))
402 			goto csum_copy_err;
403 		err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
404 	} else {
405 		err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
406 		if (err == -EINVAL)
407 			goto csum_copy_err;
408 	}
409 	if (err)
410 		goto out_free;
411 
412 	/* Copy the address. */
413 	if (sin6) {
414 		sin6->sin6_family = AF_INET6;
415 		ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
416 		sin6->sin6_flowinfo = 0;
417 		sin6->sin6_scope_id = 0;
418 		if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
419 			sin6->sin6_scope_id = IP6CB(skb)->iif;
420 	}
421 
422 	sock_recv_timestamp(msg, sk, skb);
423 
424 	if (np->rxopt.all)
425 		datagram_recv_ctl(sk, msg, skb);
426 
427 	err = copied;
428 	if (flags & MSG_TRUNC)
429 		err = skb->len;
430 
431 out_free:
432 	skb_free_datagram(sk, skb);
433 out:
434 	return err;
435 
436 csum_copy_err:
437 	skb_kill_datagram(sk, skb, flags);
438 
439 	/* Error for blocking case is chosen to masquerade
440 	   as some normal condition.
441 	 */
442 	err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
443 	/* FIXME: increment a raw6 drops counter here */
444 	goto out;
445 }
446 
447 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
448 				     struct raw6_sock *rp)
449 {
450 	struct sk_buff *skb;
451 	int err = 0;
452 	int offset;
453 	int len;
454 	int total_len;
455 	u32 tmp_csum;
456 	u16 csum;
457 
458 	if (!rp->checksum)
459 		goto send;
460 
461 	if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
462 		goto out;
463 
464 	offset = rp->offset;
465 	total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data);
466 	if (offset >= total_len - 1) {
467 		err = -EINVAL;
468 		ip6_flush_pending_frames(sk);
469 		goto out;
470 	}
471 
472 	/* should be check HW csum miyazawa */
473 	if (skb_queue_len(&sk->sk_write_queue) == 1) {
474 		/*
475 		 * Only one fragment on the socket.
476 		 */
477 		tmp_csum = skb->csum;
478 	} else {
479 		struct sk_buff *csum_skb = NULL;
480 		tmp_csum = 0;
481 
482 		skb_queue_walk(&sk->sk_write_queue, skb) {
483 			tmp_csum = csum_add(tmp_csum, skb->csum);
484 
485 			if (csum_skb)
486 				continue;
487 
488 			len = skb->len - (skb->h.raw - skb->data);
489 			if (offset >= len) {
490 				offset -= len;
491 				continue;
492 			}
493 
494 			csum_skb = skb;
495 		}
496 
497 		skb = csum_skb;
498 	}
499 
500 	offset += skb->h.raw - skb->data;
501 	if (skb_copy_bits(skb, offset, &csum, 2))
502 		BUG();
503 
504 	/* in case cksum was not initialized */
505 	if (unlikely(csum))
506 		tmp_csum = csum_sub(tmp_csum, csum);
507 
508 	tmp_csum = csum_ipv6_magic(&fl->fl6_src,
509 				   &fl->fl6_dst,
510 				   total_len, fl->proto, tmp_csum);
511 
512 	if (tmp_csum == 0)
513 		tmp_csum = -1;
514 
515 	csum = tmp_csum;
516 	if (skb_store_bits(skb, offset, &csum, 2))
517 		BUG();
518 
519 send:
520 	err = ip6_push_pending_frames(sk);
521 out:
522 	return err;
523 }
524 
525 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
526 			struct flowi *fl, struct rt6_info *rt,
527 			unsigned int flags)
528 {
529 	struct ipv6_pinfo *np = inet6_sk(sk);
530 	struct ipv6hdr *iph;
531 	struct sk_buff *skb;
532 	unsigned int hh_len;
533 	int err;
534 
535 	if (length > rt->u.dst.dev->mtu) {
536 		ipv6_local_error(sk, EMSGSIZE, fl, rt->u.dst.dev->mtu);
537 		return -EMSGSIZE;
538 	}
539 	if (flags&MSG_PROBE)
540 		goto out;
541 
542 	hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
543 
544 	skb = sock_alloc_send_skb(sk, length+hh_len+15,
545 				  flags&MSG_DONTWAIT, &err);
546 	if (skb == NULL)
547 		goto error;
548 	skb_reserve(skb, hh_len);
549 
550 	skb->priority = sk->sk_priority;
551 	skb->dst = dst_clone(&rt->u.dst);
552 
553 	skb->nh.ipv6h = iph = (struct ipv6hdr *)skb_put(skb, length);
554 
555 	skb->ip_summed = CHECKSUM_NONE;
556 
557 	skb->h.raw = skb->nh.raw;
558 	err = memcpy_fromiovecend((void *)iph, from, 0, length);
559 	if (err)
560 		goto error_fault;
561 
562 	IP6_INC_STATS(IPSTATS_MIB_OUTREQUESTS);
563 	err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
564 		      dst_output);
565 	if (err > 0)
566 		err = np->recverr ? net_xmit_errno(err) : 0;
567 	if (err)
568 		goto error;
569 out:
570 	return 0;
571 
572 error_fault:
573 	err = -EFAULT;
574 	kfree_skb(skb);
575 error:
576 	IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
577 	return err;
578 }
579 
580 static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
581 {
582 	struct iovec *iov;
583 	u8 __user *type = NULL;
584 	u8 __user *code = NULL;
585 	int probed = 0;
586 	int i;
587 
588 	if (!msg->msg_iov)
589 		return;
590 
591 	for (i = 0; i < msg->msg_iovlen; i++) {
592 		iov = &msg->msg_iov[i];
593 		if (!iov)
594 			continue;
595 
596 		switch (fl->proto) {
597 		case IPPROTO_ICMPV6:
598 			/* check if one-byte field is readable or not. */
599 			if (iov->iov_base && iov->iov_len < 1)
600 				break;
601 
602 			if (!type) {
603 				type = iov->iov_base;
604 				/* check if code field is readable or not. */
605 				if (iov->iov_len > 1)
606 					code = type + 1;
607 			} else if (!code)
608 				code = iov->iov_base;
609 
610 			if (type && code) {
611 				get_user(fl->fl_icmp_type, type);
612 				get_user(fl->fl_icmp_code, code);
613 				probed = 1;
614 			}
615 			break;
616 		default:
617 			probed = 1;
618 			break;
619 		}
620 		if (probed)
621 			break;
622 	}
623 }
624 
625 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
626 		   struct msghdr *msg, size_t len)
627 {
628 	struct ipv6_txoptions opt_space;
629 	struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
630 	struct in6_addr *daddr, *final_p = NULL, final;
631 	struct inet_sock *inet = inet_sk(sk);
632 	struct ipv6_pinfo *np = inet6_sk(sk);
633 	struct raw6_sock *rp = raw6_sk(sk);
634 	struct ipv6_txoptions *opt = NULL;
635 	struct ip6_flowlabel *flowlabel = NULL;
636 	struct dst_entry *dst = NULL;
637 	struct flowi fl;
638 	int addr_len = msg->msg_namelen;
639 	int hlimit = -1;
640 	int tclass = -1;
641 	u16 proto;
642 	int err;
643 
644 	/* Rough check on arithmetic overflow,
645 	   better check is made in ip6_build_xmit
646 	 */
647 	if (len < 0)
648 		return -EMSGSIZE;
649 
650 	/* Mirror BSD error message compatibility */
651 	if (msg->msg_flags & MSG_OOB)
652 		return -EOPNOTSUPP;
653 
654 	/*
655 	 *	Get and verify the address.
656 	 */
657 	memset(&fl, 0, sizeof(fl));
658 
659 	if (sin6) {
660 		if (addr_len < SIN6_LEN_RFC2133)
661 			return -EINVAL;
662 
663 		if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
664 			return(-EAFNOSUPPORT);
665 
666 		/* port is the proto value [0..255] carried in nexthdr */
667 		proto = ntohs(sin6->sin6_port);
668 
669 		if (!proto)
670 			proto = inet->num;
671 		else if (proto != inet->num)
672 			return(-EINVAL);
673 
674 		if (proto > 255)
675 			return(-EINVAL);
676 
677 		daddr = &sin6->sin6_addr;
678 		if (np->sndflow) {
679 			fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
680 			if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
681 				flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
682 				if (flowlabel == NULL)
683 					return -EINVAL;
684 				daddr = &flowlabel->dst;
685 			}
686 		}
687 
688 		/*
689 		 * Otherwise it will be difficult to maintain
690 		 * sk->sk_dst_cache.
691 		 */
692 		if (sk->sk_state == TCP_ESTABLISHED &&
693 		    ipv6_addr_equal(daddr, &np->daddr))
694 			daddr = &np->daddr;
695 
696 		if (addr_len >= sizeof(struct sockaddr_in6) &&
697 		    sin6->sin6_scope_id &&
698 		    ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
699 			fl.oif = sin6->sin6_scope_id;
700 	} else {
701 		if (sk->sk_state != TCP_ESTABLISHED)
702 			return -EDESTADDRREQ;
703 
704 		proto = inet->num;
705 		daddr = &np->daddr;
706 		fl.fl6_flowlabel = np->flow_label;
707 	}
708 
709 	if (ipv6_addr_any(daddr)) {
710 		/*
711 		 * unspecified destination address
712 		 * treated as error... is this correct ?
713 		 */
714 		fl6_sock_release(flowlabel);
715 		return(-EINVAL);
716 	}
717 
718 	if (fl.oif == 0)
719 		fl.oif = sk->sk_bound_dev_if;
720 
721 	if (msg->msg_controllen) {
722 		opt = &opt_space;
723 		memset(opt, 0, sizeof(struct ipv6_txoptions));
724 		opt->tot_len = sizeof(struct ipv6_txoptions);
725 
726 		err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
727 		if (err < 0) {
728 			fl6_sock_release(flowlabel);
729 			return err;
730 		}
731 		if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
732 			flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
733 			if (flowlabel == NULL)
734 				return -EINVAL;
735 		}
736 		if (!(opt->opt_nflen|opt->opt_flen))
737 			opt = NULL;
738 	}
739 	if (opt == NULL)
740 		opt = np->opt;
741 	if (flowlabel)
742 		opt = fl6_merge_options(&opt_space, flowlabel, opt);
743 	opt = ipv6_fixup_options(&opt_space, opt);
744 
745 	fl.proto = proto;
746 	rawv6_probe_proto_opt(&fl, msg);
747 
748 	ipv6_addr_copy(&fl.fl6_dst, daddr);
749 	if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
750 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
751 
752 	/* merge ip6_build_xmit from ip6_output */
753 	if (opt && opt->srcrt) {
754 		struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
755 		ipv6_addr_copy(&final, &fl.fl6_dst);
756 		ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
757 		final_p = &final;
758 	}
759 
760 	if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
761 		fl.oif = np->mcast_oif;
762 
763 	err = ip6_dst_lookup(sk, &dst, &fl);
764 	if (err)
765 		goto out;
766 	if (final_p)
767 		ipv6_addr_copy(&fl.fl6_dst, final_p);
768 
769 	if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
770 		goto out;
771 
772 	if (hlimit < 0) {
773 		if (ipv6_addr_is_multicast(&fl.fl6_dst))
774 			hlimit = np->mcast_hops;
775 		else
776 			hlimit = np->hop_limit;
777 		if (hlimit < 0)
778 			hlimit = dst_metric(dst, RTAX_HOPLIMIT);
779 		if (hlimit < 0)
780 			hlimit = ipv6_get_hoplimit(dst->dev);
781 	}
782 
783 	if (tclass < 0) {
784 		tclass = np->cork.tclass;
785 		if (tclass < 0)
786 			tclass = 0;
787 	}
788 
789 	if (msg->msg_flags&MSG_CONFIRM)
790 		goto do_confirm;
791 
792 back_from_confirm:
793 	if (inet->hdrincl) {
794 		err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
795 	} else {
796 		lock_sock(sk);
797 		err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
798 			len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
799 			msg->msg_flags);
800 
801 		if (err)
802 			ip6_flush_pending_frames(sk);
803 		else if (!(msg->msg_flags & MSG_MORE))
804 			err = rawv6_push_pending_frames(sk, &fl, rp);
805 	}
806 done:
807 	ip6_dst_store(sk, dst,
808 		      ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
809 		      &np->daddr : NULL);
810 
811 	release_sock(sk);
812 out:
813 	fl6_sock_release(flowlabel);
814 	return err<0?err:len;
815 do_confirm:
816 	dst_confirm(dst);
817 	if (!(msg->msg_flags & MSG_PROBE) || len)
818 		goto back_from_confirm;
819 	err = 0;
820 	goto done;
821 }
822 
823 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
824 			       char __user *optval, int optlen)
825 {
826 	switch (optname) {
827 	case ICMPV6_FILTER:
828 		if (optlen > sizeof(struct icmp6_filter))
829 			optlen = sizeof(struct icmp6_filter);
830 		if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
831 			return -EFAULT;
832 		return 0;
833 	default:
834 		return -ENOPROTOOPT;
835 	};
836 
837 	return 0;
838 }
839 
840 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
841 			       char __user *optval, int __user *optlen)
842 {
843 	int len;
844 
845 	switch (optname) {
846 	case ICMPV6_FILTER:
847 		if (get_user(len, optlen))
848 			return -EFAULT;
849 		if (len < 0)
850 			return -EINVAL;
851 		if (len > sizeof(struct icmp6_filter))
852 			len = sizeof(struct icmp6_filter);
853 		if (put_user(len, optlen))
854 			return -EFAULT;
855 		if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
856 			return -EFAULT;
857 		return 0;
858 	default:
859 		return -ENOPROTOOPT;
860 	};
861 
862 	return 0;
863 }
864 
865 
866 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
867 			    char __user *optval, int optlen)
868 {
869 	struct raw6_sock *rp = raw6_sk(sk);
870 	int val;
871 
872 	switch(level) {
873 		case SOL_RAW:
874 			break;
875 
876 		case SOL_ICMPV6:
877 			if (inet_sk(sk)->num != IPPROTO_ICMPV6)
878 				return -EOPNOTSUPP;
879 			return rawv6_seticmpfilter(sk, level, optname, optval,
880 						   optlen);
881 		case SOL_IPV6:
882 			if (optname == IPV6_CHECKSUM)
883 				break;
884 		default:
885 			return ipv6_setsockopt(sk, level, optname, optval,
886 					       optlen);
887 	};
888 
889   	if (get_user(val, (int __user *)optval))
890 		return -EFAULT;
891 
892 	switch (optname) {
893 		case IPV6_CHECKSUM:
894 			/* You may get strange result with a positive odd offset;
895 			   RFC2292bis agrees with me. */
896 			if (val > 0 && (val&1))
897 				return(-EINVAL);
898 			if (val < 0) {
899 				rp->checksum = 0;
900 			} else {
901 				rp->checksum = 1;
902 				rp->offset = val;
903 			}
904 
905 			return 0;
906 			break;
907 
908 		default:
909 			return(-ENOPROTOOPT);
910 	}
911 }
912 
913 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
914 			    char __user *optval, int __user *optlen)
915 {
916 	struct raw6_sock *rp = raw6_sk(sk);
917 	int val, len;
918 
919 	switch(level) {
920 		case SOL_RAW:
921 			break;
922 
923 		case SOL_ICMPV6:
924 			if (inet_sk(sk)->num != IPPROTO_ICMPV6)
925 				return -EOPNOTSUPP;
926 			return rawv6_geticmpfilter(sk, level, optname, optval,
927 						   optlen);
928 		case SOL_IPV6:
929 			if (optname == IPV6_CHECKSUM)
930 				break;
931 		default:
932 			return ipv6_getsockopt(sk, level, optname, optval,
933 					       optlen);
934 	};
935 
936 	if (get_user(len,optlen))
937 		return -EFAULT;
938 
939 	switch (optname) {
940 	case IPV6_CHECKSUM:
941 		if (rp->checksum == 0)
942 			val = -1;
943 		else
944 			val = rp->offset;
945 		break;
946 
947 	default:
948 		return -ENOPROTOOPT;
949 	}
950 
951 	len = min_t(unsigned int, sizeof(int), len);
952 
953 	if (put_user(len, optlen))
954 		return -EFAULT;
955 	if (copy_to_user(optval,&val,len))
956 		return -EFAULT;
957 	return 0;
958 }
959 
960 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
961 {
962 	switch(cmd) {
963 		case SIOCOUTQ:
964 		{
965 			int amount = atomic_read(&sk->sk_wmem_alloc);
966 			return put_user(amount, (int __user *)arg);
967 		}
968 		case SIOCINQ:
969 		{
970 			struct sk_buff *skb;
971 			int amount = 0;
972 
973 			spin_lock_bh(&sk->sk_receive_queue.lock);
974 			skb = skb_peek(&sk->sk_receive_queue);
975 			if (skb != NULL)
976 				amount = skb->tail - skb->h.raw;
977 			spin_unlock_bh(&sk->sk_receive_queue.lock);
978 			return put_user(amount, (int __user *)arg);
979 		}
980 
981 		default:
982 			return -ENOIOCTLCMD;
983 	}
984 }
985 
986 static void rawv6_close(struct sock *sk, long timeout)
987 {
988 	if (inet_sk(sk)->num == IPPROTO_RAW)
989 		ip6_ra_control(sk, -1, NULL);
990 
991 	sk_common_release(sk);
992 }
993 
994 static int rawv6_init_sk(struct sock *sk)
995 {
996 	if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
997 		struct raw6_sock *rp = raw6_sk(sk);
998 		rp->checksum = 1;
999 		rp->offset   = 2;
1000 	}
1001 	return(0);
1002 }
1003 
1004 struct proto rawv6_prot = {
1005 	.name =		"RAWv6",
1006 	.owner =	THIS_MODULE,
1007 	.close =	rawv6_close,
1008 	.connect =	ip6_datagram_connect,
1009 	.disconnect =	udp_disconnect,
1010 	.ioctl =	rawv6_ioctl,
1011 	.init =		rawv6_init_sk,
1012 	.destroy =	inet6_destroy_sock,
1013 	.setsockopt =	rawv6_setsockopt,
1014 	.getsockopt =	rawv6_getsockopt,
1015 	.sendmsg =	rawv6_sendmsg,
1016 	.recvmsg =	rawv6_recvmsg,
1017 	.bind =		rawv6_bind,
1018 	.backlog_rcv =	rawv6_rcv_skb,
1019 	.hash =		raw_v6_hash,
1020 	.unhash =	raw_v6_unhash,
1021 	.obj_size =	sizeof(struct raw6_sock),
1022 };
1023 
1024 #ifdef CONFIG_PROC_FS
1025 struct raw6_iter_state {
1026 	int bucket;
1027 };
1028 
1029 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1030 
1031 static struct sock *raw6_get_first(struct seq_file *seq)
1032 {
1033 	struct sock *sk;
1034 	struct hlist_node *node;
1035 	struct raw6_iter_state* state = raw6_seq_private(seq);
1036 
1037 	for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1038 		sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1039 			if (sk->sk_family == PF_INET6)
1040 				goto out;
1041 	sk = NULL;
1042 out:
1043 	return sk;
1044 }
1045 
1046 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1047 {
1048 	struct raw6_iter_state* state = raw6_seq_private(seq);
1049 
1050 	do {
1051 		sk = sk_next(sk);
1052 try_again:
1053 		;
1054 	} while (sk && sk->sk_family != PF_INET6);
1055 
1056 	if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1057 		sk = sk_head(&raw_v6_htable[state->bucket]);
1058 		goto try_again;
1059 	}
1060 	return sk;
1061 }
1062 
1063 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1064 {
1065 	struct sock *sk = raw6_get_first(seq);
1066 	if (sk)
1067 		while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1068 			--pos;
1069 	return pos ? NULL : sk;
1070 }
1071 
1072 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1073 {
1074 	read_lock(&raw_v6_lock);
1075 	return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1076 }
1077 
1078 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1079 {
1080 	struct sock *sk;
1081 
1082 	if (v == SEQ_START_TOKEN)
1083 		sk = raw6_get_first(seq);
1084 	else
1085 		sk = raw6_get_next(seq, v);
1086 	++*pos;
1087 	return sk;
1088 }
1089 
1090 static void raw6_seq_stop(struct seq_file *seq, void *v)
1091 {
1092 	read_unlock(&raw_v6_lock);
1093 }
1094 
1095 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1096 {
1097 	struct ipv6_pinfo *np = inet6_sk(sp);
1098 	struct in6_addr *dest, *src;
1099 	__u16 destp, srcp;
1100 
1101 	dest  = &np->daddr;
1102 	src   = &np->rcv_saddr;
1103 	destp = 0;
1104 	srcp  = inet_sk(sp)->num;
1105 	seq_printf(seq,
1106 		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1107 		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1108 		   i,
1109 		   src->s6_addr32[0], src->s6_addr32[1],
1110 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1111 		   dest->s6_addr32[0], dest->s6_addr32[1],
1112 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1113 		   sp->sk_state,
1114 		   atomic_read(&sp->sk_wmem_alloc),
1115 		   atomic_read(&sp->sk_rmem_alloc),
1116 		   0, 0L, 0,
1117 		   sock_i_uid(sp), 0,
1118 		   sock_i_ino(sp),
1119 		   atomic_read(&sp->sk_refcnt), sp);
1120 }
1121 
1122 static int raw6_seq_show(struct seq_file *seq, void *v)
1123 {
1124 	if (v == SEQ_START_TOKEN)
1125 		seq_printf(seq,
1126 			   "  sl  "
1127 			   "local_address                         "
1128 			   "remote_address                        "
1129 			   "st tx_queue rx_queue tr tm->when retrnsmt"
1130 			   "   uid  timeout inode\n");
1131 	else
1132 		raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1133 	return 0;
1134 }
1135 
1136 static struct seq_operations raw6_seq_ops = {
1137 	.start =	raw6_seq_start,
1138 	.next =		raw6_seq_next,
1139 	.stop =		raw6_seq_stop,
1140 	.show =		raw6_seq_show,
1141 };
1142 
1143 static int raw6_seq_open(struct inode *inode, struct file *file)
1144 {
1145 	struct seq_file *seq;
1146 	int rc = -ENOMEM;
1147 	struct raw6_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1148 	if (!s)
1149 		goto out;
1150 	rc = seq_open(file, &raw6_seq_ops);
1151 	if (rc)
1152 		goto out_kfree;
1153 	seq = file->private_data;
1154 	seq->private = s;
1155 	memset(s, 0, sizeof(*s));
1156 out:
1157 	return rc;
1158 out_kfree:
1159 	kfree(s);
1160 	goto out;
1161 }
1162 
1163 static struct file_operations raw6_seq_fops = {
1164 	.owner =	THIS_MODULE,
1165 	.open =		raw6_seq_open,
1166 	.read =		seq_read,
1167 	.llseek =	seq_lseek,
1168 	.release =	seq_release_private,
1169 };
1170 
1171 int __init raw6_proc_init(void)
1172 {
1173 	if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1174 		return -ENOMEM;
1175 	return 0;
1176 }
1177 
1178 void raw6_proc_exit(void)
1179 {
1180 	proc_net_remove("raw6");
1181 }
1182 #endif	/* CONFIG_PROC_FS */
1183