xref: /openbmc/linux/net/ipv6/raw.c (revision d5cb9783536a41df9f9cba5b0a1d78047ed787f7)
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 <asm/uaccess.h>
36 #include <asm/ioctls.h>
37 #include <asm/bug.h>
38 
39 #include <net/ip.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
42 
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/ip6_route.h>
47 #include <net/ip6_checksum.h>
48 #include <net/addrconf.h>
49 #include <net/transp_v6.h>
50 #include <net/udp.h>
51 #include <net/inet_common.h>
52 #include <net/tcp_states.h>
53 
54 #include <net/rawv6.h>
55 #include <net/xfrm.h>
56 
57 #include <linux/proc_fs.h>
58 #include <linux/seq_file.h>
59 
60 struct hlist_head raw_v6_htable[RAWV6_HTABLE_SIZE];
61 DEFINE_RWLOCK(raw_v6_lock);
62 
63 static 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 
74 static 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... */
84 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
85 			     struct in6_addr *loc_addr, struct in6_addr *rmt_addr,
86 			     int dif)
87 {
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 
99 			if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
100 				continue;
101 
102 			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;
113 found:
114 	return sk;
115 }
116 
117 /*
118  *	0 - deliver
119  *	1 - block
120  */
121 static __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  */
145 int ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
146 {
147 	struct in6_addr *saddr;
148 	struct in6_addr *daddr;
149 	struct sock *sk;
150 	int delivered = 0;
151 	__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 
169 	sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr, IP6CB(skb)->iif);
170 
171 	while (sk) {
172 		delivered = 1;
173 		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 		}
180 		sk = __raw_v6_lookup(sk_next(sk), nexthdr, daddr, saddr,
181 				     IP6CB(skb)->iif);
182 	}
183 out:
184 	read_unlock(&raw_v6_lock);
185 	return delivered;
186 }
187 
188 /* This cleans up af_inet6 a bit. -DaveM */
189 static 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;
257 out:
258 	release_sock(sk);
259 	return err;
260 }
261 
262 void 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 
296 static 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  */
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_UNNECESSARY) {
339 		if (skb->ip_summed == CHECKSUM_HW) {
340 			skb_postpull_rcsum(skb, skb->nh.raw,
341 			                   skb->h.raw - skb->nh.raw);
342 			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)) {
346 				LIMIT_NETDEBUG(KERN_DEBUG "raw v6 hw csum failure.\n");
347 				skb->ip_summed = CHECKSUM_NONE;
348 			}
349 		}
350 		if (skb->ip_summed == CHECKSUM_NONE)
351 			skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
352 						     &skb->nh.ipv6h->daddr,
353 						     skb->len, inet->num, 0);
354 	}
355 
356 	if (inet->hdrincl) {
357 		if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
358 		    (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
359 			/* FIXME: increment a raw6 drops counter here */
360 			kfree_skb(skb);
361 			return 0;
362 		}
363 		skb->ip_summed = CHECKSUM_UNNECESSARY;
364 	}
365 
366 	rawv6_rcv_skb(sk, skb);
367 	return 0;
368 }
369 
370 
371 /*
372  *	This should be easy, if there is something there
373  *	we return it, otherwise we block.
374  */
375 
376 static 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 
438 out_free:
439 	skb_free_datagram(sk, skb);
440 out:
441 	return err;
442 
443 csum_copy_err:
444 	/* Clear queue. */
445 	if (flags&MSG_PEEK) {
446 		int clear = 0;
447 		spin_lock_bh(&sk->sk_receive_queue.lock);
448 		if (skb == skb_peek(&sk->sk_receive_queue)) {
449 			__skb_unlink(skb, &sk->sk_receive_queue);
450 			clear = 1;
451 		}
452 		spin_unlock_bh(&sk->sk_receive_queue.lock);
453 		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 
465 static int rawv6_push_pending_frames(struct sock *sk, struct flowi *fl,
466 				     struct raw6_sock *rp)
467 {
468 	struct sk_buff *skb;
469 	int err = 0;
470 	int offset;
471 	int len;
472 	int total_len;
473 	u32 tmp_csum;
474 	u16 csum;
475 
476 	if (!rp->checksum)
477 		goto send;
478 
479 	if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
480 		goto out;
481 
482 	offset = rp->offset;
483 	total_len = inet_sk(sk)->cork.length - (skb->nh.raw - skb->data);
484 	if (offset >= total_len - 1) {
485 		err = -EINVAL;
486 		ip6_flush_pending_frames(sk);
487 		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 {
497 		struct sk_buff *csum_skb = NULL;
498 		tmp_csum = 0;
499 
500 		skb_queue_walk(&sk->sk_write_queue, skb) {
501 			tmp_csum = csum_add(tmp_csum, skb->csum);
502 
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;
513 		}
514 
515 		skb = csum_skb;
516 	}
517 
518 	offset += skb->h.raw - skb->data;
519 	if (skb_copy_bits(skb, offset, &csum, 2))
520 		BUG();
521 
522 	/* in case cksum was not initialized */
523 	if (unlikely(csum))
524 		tmp_csum = csum_sub(tmp_csum, csum);
525 
526 	tmp_csum = csum_ipv6_magic(&fl->fl6_src,
527 				   &fl->fl6_dst,
528 				   total_len, fl->proto, tmp_csum);
529 
530 	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 
537 send:
538 	err = ip6_push_pending_frames(sk);
539 out:
540 	return err;
541 }
542 
543 static int rawv6_send_hdrinc(struct sock *sk, void *from, int length,
544 			struct flowi *fl, struct rt6_info *rt,
545 			unsigned int flags)
546 {
547 	struct ipv6_pinfo *np = inet6_sk(sk);
548 	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)
584 		err = np->recverr ? net_xmit_errno(err) : 0;
585 	if (err)
586 		goto error;
587 out:
588 	return 0;
589 
590 error_fault:
591 	err = -EFAULT;
592 	kfree_skb(skb);
593 error:
594 	IP6_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
595 	return err;
596 }
597 
598 static void rawv6_probe_proto_opt(struct flowi *fl, struct msghdr *msg)
599 {
600 	struct iovec *iov;
601 	u8 __user *type = NULL;
602 	u8 __user *code = NULL;
603 	int probed = 0;
604 	int i;
605 
606 	if (!msg->msg_iov)
607 		return;
608 
609 	for (i = 0; i < msg->msg_iovlen; i++) {
610 		iov = &msg->msg_iov[i];
611 		if (!iov)
612 			continue;
613 
614 		switch (fl->proto) {
615 		case IPPROTO_ICMPV6:
616 			/* check if one-byte field is readable or not. */
617 			if (iov->iov_base && iov->iov_len < 1)
618 				break;
619 
620 			if (!type) {
621 				type = iov->iov_base;
622 				/* check if code field is readable or not. */
623 				if (iov->iov_len > 1)
624 					code = type + 1;
625 			} else if (!code)
626 				code = iov->iov_base;
627 
628 			if (type && code) {
629 				get_user(fl->fl_icmp_type, type);
630 				get_user(fl->fl_icmp_code, code);
631 				probed = 1;
632 			}
633 			break;
634 		default:
635 			probed = 1;
636 			break;
637 		}
638 		if (probed)
639 			break;
640 	}
641 }
642 
643 static int rawv6_sendmsg(struct kiocb *iocb, struct sock *sk,
644 		   struct msghdr *msg, size_t len)
645 {
646 	struct ipv6_txoptions opt_space;
647 	struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
648 	struct in6_addr *daddr, *final_p = NULL, final;
649 	struct inet_sock *inet = inet_sk(sk);
650 	struct ipv6_pinfo *np = inet6_sk(sk);
651 	struct raw6_sock *rp = raw6_sk(sk);
652 	struct ipv6_txoptions *opt = NULL;
653 	struct ip6_flowlabel *flowlabel = NULL;
654 	struct dst_entry *dst = NULL;
655 	struct flowi fl;
656 	int addr_len = msg->msg_namelen;
657 	int hlimit = -1;
658 	int tclass = -1;
659 	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 
744 		err = datagram_send_ctl(msg, &fl, opt, &hlimit, &tclass);
745 		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;
759 	opt = fl6_merge_options(&opt_space, flowlabel, opt);
760 
761 	fl.proto = proto;
762 	rawv6_probe_proto_opt(&fl, msg);
763 
764 	ipv6_addr_copy(&fl.fl6_dst, daddr);
765 	if (ipv6_addr_any(&fl.fl6_src) && !ipv6_addr_any(&np->saddr))
766 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
767 
768 	/* merge ip6_build_xmit from ip6_output */
769 	if (opt && opt->srcrt) {
770 		struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
771 		ipv6_addr_copy(&final, &fl.fl6_dst);
772 		ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
773 		final_p = &final;
774 	}
775 
776 	if (!fl.oif && ipv6_addr_is_multicast(&fl.fl6_dst))
777 		fl.oif = np->mcast_oif;
778 
779 	err = ip6_dst_lookup(sk, &dst, &fl);
780 	if (err)
781 		goto out;
782 	if (final_p)
783 		ipv6_addr_copy(&fl.fl6_dst, final_p);
784 
785 	if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0)
786 		goto out;
787 
788 	if (hlimit < 0) {
789 		if (ipv6_addr_is_multicast(&fl.fl6_dst))
790 			hlimit = np->mcast_hops;
791 		else
792 			hlimit = np->hop_limit;
793 		if (hlimit < 0)
794 			hlimit = dst_metric(dst, RTAX_HOPLIMIT);
795 		if (hlimit < 0)
796 			hlimit = ipv6_get_hoplimit(dst->dev);
797 	}
798 
799 	if (tclass < 0) {
800 		tclass = np->cork.tclass;
801 		if (tclass < 0)
802 			tclass = 0;
803 	}
804 
805 	if (msg->msg_flags&MSG_CONFIRM)
806 		goto do_confirm;
807 
808 back_from_confirm:
809 	if (inet->hdrincl) {
810 		err = rawv6_send_hdrinc(sk, msg->msg_iov, len, &fl, (struct rt6_info*)dst, msg->msg_flags);
811 	} else {
812 		lock_sock(sk);
813 		err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov,
814 			len, 0, hlimit, tclass, opt, &fl, (struct rt6_info*)dst,
815 			msg->msg_flags);
816 
817 		if (err)
818 			ip6_flush_pending_frames(sk);
819 		else if (!(msg->msg_flags & MSG_MORE))
820 			err = rawv6_push_pending_frames(sk, &fl, rp);
821 	}
822 done:
823 	ip6_dst_store(sk, dst,
824 		      ipv6_addr_equal(&fl.fl6_dst, &np->daddr) ?
825 		      &np->daddr : NULL);
826 
827 	release_sock(sk);
828 out:
829 	fl6_sock_release(flowlabel);
830 	return err<0?err:len;
831 do_confirm:
832 	dst_confirm(dst);
833 	if (!(msg->msg_flags & MSG_PROBE) || len)
834 		goto back_from_confirm;
835 	err = 0;
836 	goto done;
837 }
838 
839 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
840 			       char __user *optval, int optlen)
841 {
842 	switch (optname) {
843 	case ICMPV6_FILTER:
844 		if (optlen > sizeof(struct icmp6_filter))
845 			optlen = sizeof(struct icmp6_filter);
846 		if (copy_from_user(&raw6_sk(sk)->filter, optval, optlen))
847 			return -EFAULT;
848 		return 0;
849 	default:
850 		return -ENOPROTOOPT;
851 	};
852 
853 	return 0;
854 }
855 
856 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
857 			       char __user *optval, int __user *optlen)
858 {
859 	int len;
860 
861 	switch (optname) {
862 	case ICMPV6_FILTER:
863 		if (get_user(len, optlen))
864 			return -EFAULT;
865 		if (len < 0)
866 			return -EINVAL;
867 		if (len > sizeof(struct icmp6_filter))
868 			len = sizeof(struct icmp6_filter);
869 		if (put_user(len, optlen))
870 			return -EFAULT;
871 		if (copy_to_user(optval, &raw6_sk(sk)->filter, len))
872 			return -EFAULT;
873 		return 0;
874 	default:
875 		return -ENOPROTOOPT;
876 	};
877 
878 	return 0;
879 }
880 
881 
882 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
883 			    char __user *optval, int optlen)
884 {
885 	struct raw6_sock *rp = raw6_sk(sk);
886 	int val;
887 
888 	switch(level) {
889 		case SOL_RAW:
890 			break;
891 
892 		case SOL_ICMPV6:
893 			if (inet_sk(sk)->num != IPPROTO_ICMPV6)
894 				return -EOPNOTSUPP;
895 			return rawv6_seticmpfilter(sk, level, optname, optval,
896 						   optlen);
897 		case SOL_IPV6:
898 			if (optname == IPV6_CHECKSUM)
899 				break;
900 		default:
901 			return ipv6_setsockopt(sk, level, optname, optval,
902 					       optlen);
903 	};
904 
905   	if (get_user(val, (int __user *)optval))
906 		return -EFAULT;
907 
908 	switch (optname) {
909 		case IPV6_CHECKSUM:
910 			/* You may get strange result with a positive odd offset;
911 			   RFC2292bis agrees with me. */
912 			if (val > 0 && (val&1))
913 				return(-EINVAL);
914 			if (val < 0) {
915 				rp->checksum = 0;
916 			} else {
917 				rp->checksum = 1;
918 				rp->offset = val;
919 			}
920 
921 			return 0;
922 			break;
923 
924 		default:
925 			return(-ENOPROTOOPT);
926 	}
927 }
928 
929 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
930 			    char __user *optval, int __user *optlen)
931 {
932 	struct raw6_sock *rp = raw6_sk(sk);
933 	int val, len;
934 
935 	switch(level) {
936 		case SOL_RAW:
937 			break;
938 
939 		case SOL_ICMPV6:
940 			if (inet_sk(sk)->num != IPPROTO_ICMPV6)
941 				return -EOPNOTSUPP;
942 			return rawv6_geticmpfilter(sk, level, optname, optval,
943 						   optlen);
944 		case SOL_IPV6:
945 			if (optname == IPV6_CHECKSUM)
946 				break;
947 		default:
948 			return ipv6_getsockopt(sk, level, optname, optval,
949 					       optlen);
950 	};
951 
952 	if (get_user(len,optlen))
953 		return -EFAULT;
954 
955 	switch (optname) {
956 	case IPV6_CHECKSUM:
957 		if (rp->checksum == 0)
958 			val = -1;
959 		else
960 			val = rp->offset;
961 		break;
962 
963 	default:
964 		return -ENOPROTOOPT;
965 	}
966 
967 	len = min_t(unsigned int, sizeof(int), len);
968 
969 	if (put_user(len, optlen))
970 		return -EFAULT;
971 	if (copy_to_user(optval,&val,len))
972 		return -EFAULT;
973 	return 0;
974 }
975 
976 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
977 {
978 	switch(cmd) {
979 		case SIOCOUTQ:
980 		{
981 			int amount = atomic_read(&sk->sk_wmem_alloc);
982 			return put_user(amount, (int __user *)arg);
983 		}
984 		case SIOCINQ:
985 		{
986 			struct sk_buff *skb;
987 			int amount = 0;
988 
989 			spin_lock_bh(&sk->sk_receive_queue.lock);
990 			skb = skb_peek(&sk->sk_receive_queue);
991 			if (skb != NULL)
992 				amount = skb->tail - skb->h.raw;
993 			spin_unlock_bh(&sk->sk_receive_queue.lock);
994 			return put_user(amount, (int __user *)arg);
995 		}
996 
997 		default:
998 			return -ENOIOCTLCMD;
999 	}
1000 }
1001 
1002 static void rawv6_close(struct sock *sk, long timeout)
1003 {
1004 	if (inet_sk(sk)->num == IPPROTO_RAW)
1005 		ip6_ra_control(sk, -1, NULL);
1006 
1007 	sk_common_release(sk);
1008 }
1009 
1010 static int rawv6_init_sk(struct sock *sk)
1011 {
1012 	if (inet_sk(sk)->num == IPPROTO_ICMPV6) {
1013 		struct raw6_sock *rp = raw6_sk(sk);
1014 		rp->checksum = 1;
1015 		rp->offset   = 2;
1016 	}
1017 	return(0);
1018 }
1019 
1020 struct proto rawv6_prot = {
1021 	.name =		"RAWv6",
1022 	.owner =	THIS_MODULE,
1023 	.close =	rawv6_close,
1024 	.connect =	ip6_datagram_connect,
1025 	.disconnect =	udp_disconnect,
1026 	.ioctl =	rawv6_ioctl,
1027 	.init =		rawv6_init_sk,
1028 	.destroy =	inet6_destroy_sock,
1029 	.setsockopt =	rawv6_setsockopt,
1030 	.getsockopt =	rawv6_getsockopt,
1031 	.sendmsg =	rawv6_sendmsg,
1032 	.recvmsg =	rawv6_recvmsg,
1033 	.bind =		rawv6_bind,
1034 	.backlog_rcv =	rawv6_rcv_skb,
1035 	.hash =		raw_v6_hash,
1036 	.unhash =	raw_v6_unhash,
1037 	.obj_size =	sizeof(struct raw6_sock),
1038 };
1039 
1040 #ifdef CONFIG_PROC_FS
1041 struct raw6_iter_state {
1042 	int bucket;
1043 };
1044 
1045 #define raw6_seq_private(seq) ((struct raw6_iter_state *)(seq)->private)
1046 
1047 static struct sock *raw6_get_first(struct seq_file *seq)
1048 {
1049 	struct sock *sk;
1050 	struct hlist_node *node;
1051 	struct raw6_iter_state* state = raw6_seq_private(seq);
1052 
1053 	for (state->bucket = 0; state->bucket < RAWV6_HTABLE_SIZE; ++state->bucket)
1054 		sk_for_each(sk, node, &raw_v6_htable[state->bucket])
1055 			if (sk->sk_family == PF_INET6)
1056 				goto out;
1057 	sk = NULL;
1058 out:
1059 	return sk;
1060 }
1061 
1062 static struct sock *raw6_get_next(struct seq_file *seq, struct sock *sk)
1063 {
1064 	struct raw6_iter_state* state = raw6_seq_private(seq);
1065 
1066 	do {
1067 		sk = sk_next(sk);
1068 try_again:
1069 		;
1070 	} while (sk && sk->sk_family != PF_INET6);
1071 
1072 	if (!sk && ++state->bucket < RAWV6_HTABLE_SIZE) {
1073 		sk = sk_head(&raw_v6_htable[state->bucket]);
1074 		goto try_again;
1075 	}
1076 	return sk;
1077 }
1078 
1079 static struct sock *raw6_get_idx(struct seq_file *seq, loff_t pos)
1080 {
1081 	struct sock *sk = raw6_get_first(seq);
1082 	if (sk)
1083 		while (pos && (sk = raw6_get_next(seq, sk)) != NULL)
1084 			--pos;
1085 	return pos ? NULL : sk;
1086 }
1087 
1088 static void *raw6_seq_start(struct seq_file *seq, loff_t *pos)
1089 {
1090 	read_lock(&raw_v6_lock);
1091 	return *pos ? raw6_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1092 }
1093 
1094 static void *raw6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1095 {
1096 	struct sock *sk;
1097 
1098 	if (v == SEQ_START_TOKEN)
1099 		sk = raw6_get_first(seq);
1100 	else
1101 		sk = raw6_get_next(seq, v);
1102 	++*pos;
1103 	return sk;
1104 }
1105 
1106 static void raw6_seq_stop(struct seq_file *seq, void *v)
1107 {
1108 	read_unlock(&raw_v6_lock);
1109 }
1110 
1111 static void raw6_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1112 {
1113 	struct ipv6_pinfo *np = inet6_sk(sp);
1114 	struct in6_addr *dest, *src;
1115 	__u16 destp, srcp;
1116 
1117 	dest  = &np->daddr;
1118 	src   = &np->rcv_saddr;
1119 	destp = 0;
1120 	srcp  = inet_sk(sp)->num;
1121 	seq_printf(seq,
1122 		   "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1123 		   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1124 		   i,
1125 		   src->s6_addr32[0], src->s6_addr32[1],
1126 		   src->s6_addr32[2], src->s6_addr32[3], srcp,
1127 		   dest->s6_addr32[0], dest->s6_addr32[1],
1128 		   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1129 		   sp->sk_state,
1130 		   atomic_read(&sp->sk_wmem_alloc),
1131 		   atomic_read(&sp->sk_rmem_alloc),
1132 		   0, 0L, 0,
1133 		   sock_i_uid(sp), 0,
1134 		   sock_i_ino(sp),
1135 		   atomic_read(&sp->sk_refcnt), sp);
1136 }
1137 
1138 static int raw6_seq_show(struct seq_file *seq, void *v)
1139 {
1140 	if (v == SEQ_START_TOKEN)
1141 		seq_printf(seq,
1142 			   "  sl  "
1143 			   "local_address                         "
1144 			   "remote_address                        "
1145 			   "st tx_queue rx_queue tr tm->when retrnsmt"
1146 			   "   uid  timeout inode\n");
1147 	else
1148 		raw6_sock_seq_show(seq, v, raw6_seq_private(seq)->bucket);
1149 	return 0;
1150 }
1151 
1152 static struct seq_operations raw6_seq_ops = {
1153 	.start =	raw6_seq_start,
1154 	.next =		raw6_seq_next,
1155 	.stop =		raw6_seq_stop,
1156 	.show =		raw6_seq_show,
1157 };
1158 
1159 static int raw6_seq_open(struct inode *inode, struct file *file)
1160 {
1161 	struct seq_file *seq;
1162 	int rc = -ENOMEM;
1163 	struct raw6_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
1164 	if (!s)
1165 		goto out;
1166 	rc = seq_open(file, &raw6_seq_ops);
1167 	if (rc)
1168 		goto out_kfree;
1169 	seq = file->private_data;
1170 	seq->private = s;
1171 	memset(s, 0, sizeof(*s));
1172 out:
1173 	return rc;
1174 out_kfree:
1175 	kfree(s);
1176 	goto out;
1177 }
1178 
1179 static struct file_operations raw6_seq_fops = {
1180 	.owner =	THIS_MODULE,
1181 	.open =		raw6_seq_open,
1182 	.read =		seq_read,
1183 	.llseek =	seq_lseek,
1184 	.release =	seq_release_private,
1185 };
1186 
1187 int __init raw6_proc_init(void)
1188 {
1189 	if (!proc_net_fops_create("raw6", S_IRUGO, &raw6_seq_fops))
1190 		return -ENOMEM;
1191 	return 0;
1192 }
1193 
1194 void raw6_proc_exit(void)
1195 {
1196 	proc_net_remove("raw6");
1197 }
1198 #endif	/* CONFIG_PROC_FS */
1199