xref: /linux/net/ipv4/raw.c (revision 2fe05e1139a555ae91f00a812cb9520e7d3022ab)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		RAW - implementation of IP "raw" sockets.
7  *
8  * Authors:	Ross Biro
9  *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10  *
11  * Fixes:
12  *		Alan Cox	:	verify_area() fixed up
13  *		Alan Cox	:	ICMP error handling
14  *		Alan Cox	:	EMSGSIZE if you send too big a packet
15  *		Alan Cox	: 	Now uses generic datagrams and shared
16  *					skbuff library. No more peek crashes,
17  *					no more backlogs
18  *		Alan Cox	:	Checks sk->broadcast.
19  *		Alan Cox	:	Uses skb_free_datagram/skb_copy_datagram
20  *		Alan Cox	:	Raw passes ip options too
21  *		Alan Cox	:	Setsocketopt added
22  *		Alan Cox	:	Fixed error return for broadcasts
23  *		Alan Cox	:	Removed wake_up calls
24  *		Alan Cox	:	Use ttl/tos
25  *		Alan Cox	:	Cleaned up old debugging
26  *		Alan Cox	:	Use new kernel side addresses
27  *	Arnt Gulbrandsen	:	Fixed MSG_DONTROUTE in raw sockets.
28  *		Alan Cox	:	BSD style RAW socket demultiplexing.
29  *		Alan Cox	:	Beginnings of mrouted support.
30  *		Alan Cox	:	Added IP_HDRINCL option.
31  *		Alan Cox	:	Skip broadcast check if BSDism set.
32  *		David S. Miller	:	New socket lookup architecture.
33  *
34  *		This program is free software; you can redistribute it and/or
35  *		modify it under the terms of the GNU General Public License
36  *		as published by the Free Software Foundation; either version
37  *		2 of the License, or (at your option) any later version.
38  */
39 
40 #include <linux/types.h>
41 #include <linux/atomic.h>
42 #include <asm/byteorder.h>
43 #include <asm/current.h>
44 #include <linux/uaccess.h>
45 #include <asm/ioctls.h>
46 #include <linux/stddef.h>
47 #include <linux/slab.h>
48 #include <linux/errno.h>
49 #include <linux/kernel.h>
50 #include <linux/export.h>
51 #include <linux/spinlock.h>
52 #include <linux/sockios.h>
53 #include <linux/socket.h>
54 #include <linux/in.h>
55 #include <linux/mroute.h>
56 #include <linux/netdevice.h>
57 #include <linux/in_route.h>
58 #include <linux/route.h>
59 #include <linux/skbuff.h>
60 #include <linux/igmp.h>
61 #include <net/net_namespace.h>
62 #include <net/dst.h>
63 #include <net/sock.h>
64 #include <linux/ip.h>
65 #include <linux/net.h>
66 #include <net/ip.h>
67 #include <net/icmp.h>
68 #include <net/udp.h>
69 #include <net/raw.h>
70 #include <net/snmp.h>
71 #include <net/tcp_states.h>
72 #include <net/inet_common.h>
73 #include <net/checksum.h>
74 #include <net/xfrm.h>
75 #include <linux/rtnetlink.h>
76 #include <linux/proc_fs.h>
77 #include <linux/seq_file.h>
78 #include <linux/netfilter.h>
79 #include <linux/netfilter_ipv4.h>
80 #include <linux/compat.h>
81 #include <linux/uio.h>
82 
83 struct raw_frag_vec {
84 	struct msghdr *msg;
85 	union {
86 		struct icmphdr icmph;
87 		char c[1];
88 	} hdr;
89 	int hlen;
90 };
91 
92 struct raw_hashinfo raw_v4_hashinfo = {
93 	.lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
94 };
95 EXPORT_SYMBOL_GPL(raw_v4_hashinfo);
96 
97 int raw_hash_sk(struct sock *sk)
98 {
99 	struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
100 	struct hlist_head *head;
101 
102 	head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
103 
104 	write_lock_bh(&h->lock);
105 	sk_add_node(sk, head);
106 	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
107 	write_unlock_bh(&h->lock);
108 
109 	return 0;
110 }
111 EXPORT_SYMBOL_GPL(raw_hash_sk);
112 
113 void raw_unhash_sk(struct sock *sk)
114 {
115 	struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
116 
117 	write_lock_bh(&h->lock);
118 	if (sk_del_node_init(sk))
119 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
120 	write_unlock_bh(&h->lock);
121 }
122 EXPORT_SYMBOL_GPL(raw_unhash_sk);
123 
124 struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
125 		unsigned short num, __be32 raddr, __be32 laddr, int dif)
126 {
127 	sk_for_each_from(sk) {
128 		struct inet_sock *inet = inet_sk(sk);
129 
130 		if (net_eq(sock_net(sk), net) && inet->inet_num == num	&&
131 		    !(inet->inet_daddr && inet->inet_daddr != raddr) 	&&
132 		    !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
133 		    !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
134 			goto found; /* gotcha */
135 	}
136 	sk = NULL;
137 found:
138 	return sk;
139 }
140 EXPORT_SYMBOL_GPL(__raw_v4_lookup);
141 
142 /*
143  *	0 - deliver
144  *	1 - block
145  */
146 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
147 {
148 	struct icmphdr _hdr;
149 	const struct icmphdr *hdr;
150 
151 	hdr = skb_header_pointer(skb, skb_transport_offset(skb),
152 				 sizeof(_hdr), &_hdr);
153 	if (!hdr)
154 		return 1;
155 
156 	if (hdr->type < 32) {
157 		__u32 data = raw_sk(sk)->filter.data;
158 
159 		return ((1U << hdr->type) & data) != 0;
160 	}
161 
162 	/* Do not block unknown ICMP types */
163 	return 0;
164 }
165 
166 /* IP input processing comes here for RAW socket delivery.
167  * Caller owns SKB, so we must make clones.
168  *
169  * RFC 1122: SHOULD pass TOS value up to the transport layer.
170  * -> It does. And not only TOS, but all IP header.
171  */
172 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
173 {
174 	struct sock *sk;
175 	struct hlist_head *head;
176 	int delivered = 0;
177 	struct net *net;
178 
179 	read_lock(&raw_v4_hashinfo.lock);
180 	head = &raw_v4_hashinfo.ht[hash];
181 	if (hlist_empty(head))
182 		goto out;
183 
184 	net = dev_net(skb->dev);
185 	sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
186 			     iph->saddr, iph->daddr,
187 			     skb->dev->ifindex);
188 
189 	while (sk) {
190 		delivered = 1;
191 		if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
192 		    ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
193 				   skb->dev->ifindex)) {
194 			struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
195 
196 			/* Not releasing hash table! */
197 			if (clone)
198 				raw_rcv(sk, clone);
199 		}
200 		sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
201 				     iph->saddr, iph->daddr,
202 				     skb->dev->ifindex);
203 	}
204 out:
205 	read_unlock(&raw_v4_hashinfo.lock);
206 	return delivered;
207 }
208 
209 int raw_local_deliver(struct sk_buff *skb, int protocol)
210 {
211 	int hash;
212 	struct sock *raw_sk;
213 
214 	hash = protocol & (RAW_HTABLE_SIZE - 1);
215 	raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
216 
217 	/* If there maybe a raw socket we must check - if not we
218 	 * don't care less
219 	 */
220 	if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
221 		raw_sk = NULL;
222 
223 	return raw_sk != NULL;
224 
225 }
226 
227 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
228 {
229 	struct inet_sock *inet = inet_sk(sk);
230 	const int type = icmp_hdr(skb)->type;
231 	const int code = icmp_hdr(skb)->code;
232 	int err = 0;
233 	int harderr = 0;
234 
235 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
236 		ipv4_sk_update_pmtu(skb, sk, info);
237 	else if (type == ICMP_REDIRECT) {
238 		ipv4_sk_redirect(skb, sk);
239 		return;
240 	}
241 
242 	/* Report error on raw socket, if:
243 	   1. User requested ip_recverr.
244 	   2. Socket is connected (otherwise the error indication
245 	      is useless without ip_recverr and error is hard.
246 	 */
247 	if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
248 		return;
249 
250 	switch (type) {
251 	default:
252 	case ICMP_TIME_EXCEEDED:
253 		err = EHOSTUNREACH;
254 		break;
255 	case ICMP_SOURCE_QUENCH:
256 		return;
257 	case ICMP_PARAMETERPROB:
258 		err = EPROTO;
259 		harderr = 1;
260 		break;
261 	case ICMP_DEST_UNREACH:
262 		err = EHOSTUNREACH;
263 		if (code > NR_ICMP_UNREACH)
264 			break;
265 		err = icmp_err_convert[code].errno;
266 		harderr = icmp_err_convert[code].fatal;
267 		if (code == ICMP_FRAG_NEEDED) {
268 			harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
269 			err = EMSGSIZE;
270 		}
271 	}
272 
273 	if (inet->recverr) {
274 		const struct iphdr *iph = (const struct iphdr *)skb->data;
275 		u8 *payload = skb->data + (iph->ihl << 2);
276 
277 		if (inet->hdrincl)
278 			payload = skb->data;
279 		ip_icmp_error(sk, skb, err, 0, info, payload);
280 	}
281 
282 	if (inet->recverr || harderr) {
283 		sk->sk_err = err;
284 		sk->sk_error_report(sk);
285 	}
286 }
287 
288 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
289 {
290 	int hash;
291 	struct sock *raw_sk;
292 	const struct iphdr *iph;
293 	struct net *net;
294 
295 	hash = protocol & (RAW_HTABLE_SIZE - 1);
296 
297 	read_lock(&raw_v4_hashinfo.lock);
298 	raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
299 	if (raw_sk) {
300 		iph = (const struct iphdr *)skb->data;
301 		net = dev_net(skb->dev);
302 
303 		while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
304 						iph->daddr, iph->saddr,
305 						skb->dev->ifindex)) != NULL) {
306 			raw_err(raw_sk, skb, info);
307 			raw_sk = sk_next(raw_sk);
308 			iph = (const struct iphdr *)skb->data;
309 		}
310 	}
311 	read_unlock(&raw_v4_hashinfo.lock);
312 }
313 
314 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
315 {
316 	/* Charge it to the socket. */
317 
318 	ipv4_pktinfo_prepare(sk, skb);
319 	if (sock_queue_rcv_skb(sk, skb) < 0) {
320 		kfree_skb(skb);
321 		return NET_RX_DROP;
322 	}
323 
324 	return NET_RX_SUCCESS;
325 }
326 
327 int raw_rcv(struct sock *sk, struct sk_buff *skb)
328 {
329 	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
330 		atomic_inc(&sk->sk_drops);
331 		kfree_skb(skb);
332 		return NET_RX_DROP;
333 	}
334 	nf_reset(skb);
335 
336 	skb_push(skb, skb->data - skb_network_header(skb));
337 
338 	raw_rcv_skb(sk, skb);
339 	return 0;
340 }
341 
342 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
343 			   struct msghdr *msg, size_t length,
344 			   struct rtable **rtp, unsigned int flags,
345 			   const struct sockcm_cookie *sockc)
346 {
347 	struct inet_sock *inet = inet_sk(sk);
348 	struct net *net = sock_net(sk);
349 	struct iphdr *iph;
350 	struct sk_buff *skb;
351 	unsigned int iphlen;
352 	int err;
353 	struct rtable *rt = *rtp;
354 	int hlen, tlen;
355 
356 	if (length > rt->dst.dev->mtu) {
357 		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
358 			       rt->dst.dev->mtu);
359 		return -EMSGSIZE;
360 	}
361 	if (length < sizeof(struct iphdr))
362 		return -EINVAL;
363 
364 	if (flags&MSG_PROBE)
365 		goto out;
366 
367 	hlen = LL_RESERVED_SPACE(rt->dst.dev);
368 	tlen = rt->dst.dev->needed_tailroom;
369 	skb = sock_alloc_send_skb(sk,
370 				  length + hlen + tlen + 15,
371 				  flags & MSG_DONTWAIT, &err);
372 	if (!skb)
373 		goto error;
374 	skb_reserve(skb, hlen);
375 
376 	skb->priority = sk->sk_priority;
377 	skb->mark = sk->sk_mark;
378 	skb_dst_set(skb, &rt->dst);
379 	*rtp = NULL;
380 
381 	skb_reset_network_header(skb);
382 	iph = ip_hdr(skb);
383 	skb_put(skb, length);
384 
385 	skb->ip_summed = CHECKSUM_NONE;
386 
387 	sock_tx_timestamp(sk, sockc->tsflags, &skb_shinfo(skb)->tx_flags);
388 
389 	if (flags & MSG_CONFIRM)
390 		skb_set_dst_pending_confirm(skb, 1);
391 
392 	skb->transport_header = skb->network_header;
393 	err = -EFAULT;
394 	if (memcpy_from_msg(iph, msg, length))
395 		goto error_free;
396 
397 	iphlen = iph->ihl * 4;
398 
399 	/*
400 	 * We don't want to modify the ip header, but we do need to
401 	 * be sure that it won't cause problems later along the network
402 	 * stack.  Specifically we want to make sure that iph->ihl is a
403 	 * sane value.  If ihl points beyond the length of the buffer passed
404 	 * in, reject the frame as invalid
405 	 */
406 	err = -EINVAL;
407 	if (iphlen > length)
408 		goto error_free;
409 
410 	if (iphlen >= sizeof(*iph)) {
411 		if (!iph->saddr)
412 			iph->saddr = fl4->saddr;
413 		iph->check   = 0;
414 		iph->tot_len = htons(length);
415 		if (!iph->id)
416 			ip_select_ident(net, skb, NULL);
417 
418 		iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
419 		skb->transport_header += iphlen;
420 		if (iph->protocol == IPPROTO_ICMP &&
421 		    length >= iphlen + sizeof(struct icmphdr))
422 			icmp_out_count(net, ((struct icmphdr *)
423 				skb_transport_header(skb))->type);
424 	}
425 
426 	err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
427 		      net, sk, skb, NULL, rt->dst.dev,
428 		      dst_output);
429 	if (err > 0)
430 		err = net_xmit_errno(err);
431 	if (err)
432 		goto error;
433 out:
434 	return 0;
435 
436 error_free:
437 	kfree_skb(skb);
438 error:
439 	IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
440 	if (err == -ENOBUFS && !inet->recverr)
441 		err = 0;
442 	return err;
443 }
444 
445 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
446 {
447 	int err;
448 
449 	if (fl4->flowi4_proto != IPPROTO_ICMP)
450 		return 0;
451 
452 	/* We only need the first two bytes. */
453 	rfv->hlen = 2;
454 
455 	err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
456 	if (err)
457 		return err;
458 
459 	fl4->fl4_icmp_type = rfv->hdr.icmph.type;
460 	fl4->fl4_icmp_code = rfv->hdr.icmph.code;
461 
462 	return 0;
463 }
464 
465 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
466 		       struct sk_buff *skb)
467 {
468 	struct raw_frag_vec *rfv = from;
469 
470 	if (offset < rfv->hlen) {
471 		int copy = min(rfv->hlen - offset, len);
472 
473 		if (skb->ip_summed == CHECKSUM_PARTIAL)
474 			memcpy(to, rfv->hdr.c + offset, copy);
475 		else
476 			skb->csum = csum_block_add(
477 				skb->csum,
478 				csum_partial_copy_nocheck(rfv->hdr.c + offset,
479 							  to, copy, 0),
480 				odd);
481 
482 		odd = 0;
483 		offset += copy;
484 		to += copy;
485 		len -= copy;
486 
487 		if (!len)
488 			return 0;
489 	}
490 
491 	offset -= rfv->hlen;
492 
493 	return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
494 }
495 
496 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
497 {
498 	struct inet_sock *inet = inet_sk(sk);
499 	struct net *net = sock_net(sk);
500 	struct ipcm_cookie ipc;
501 	struct rtable *rt = NULL;
502 	struct flowi4 fl4;
503 	int free = 0;
504 	__be32 daddr;
505 	__be32 saddr;
506 	u8  tos;
507 	int err;
508 	struct ip_options_data opt_copy;
509 	struct raw_frag_vec rfv;
510 
511 	err = -EMSGSIZE;
512 	if (len > 0xFFFF)
513 		goto out;
514 
515 	/*
516 	 *	Check the flags.
517 	 */
518 
519 	err = -EOPNOTSUPP;
520 	if (msg->msg_flags & MSG_OOB)	/* Mirror BSD error message */
521 		goto out;               /* compatibility */
522 
523 	/*
524 	 *	Get and verify the address.
525 	 */
526 
527 	if (msg->msg_namelen) {
528 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
529 		err = -EINVAL;
530 		if (msg->msg_namelen < sizeof(*usin))
531 			goto out;
532 		if (usin->sin_family != AF_INET) {
533 			pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
534 				     __func__, current->comm);
535 			err = -EAFNOSUPPORT;
536 			if (usin->sin_family)
537 				goto out;
538 		}
539 		daddr = usin->sin_addr.s_addr;
540 		/* ANK: I did not forget to get protocol from port field.
541 		 * I just do not know, who uses this weirdness.
542 		 * IP_HDRINCL is much more convenient.
543 		 */
544 	} else {
545 		err = -EDESTADDRREQ;
546 		if (sk->sk_state != TCP_ESTABLISHED)
547 			goto out;
548 		daddr = inet->inet_daddr;
549 	}
550 
551 	ipc.sockc.tsflags = sk->sk_tsflags;
552 	ipc.addr = inet->inet_saddr;
553 	ipc.opt = NULL;
554 	ipc.tx_flags = 0;
555 	ipc.ttl = 0;
556 	ipc.tos = -1;
557 	ipc.oif = sk->sk_bound_dev_if;
558 
559 	if (msg->msg_controllen) {
560 		err = ip_cmsg_send(sk, msg, &ipc, false);
561 		if (unlikely(err)) {
562 			kfree(ipc.opt);
563 			goto out;
564 		}
565 		if (ipc.opt)
566 			free = 1;
567 	}
568 
569 	saddr = ipc.addr;
570 	ipc.addr = daddr;
571 
572 	if (!ipc.opt) {
573 		struct ip_options_rcu *inet_opt;
574 
575 		rcu_read_lock();
576 		inet_opt = rcu_dereference(inet->inet_opt);
577 		if (inet_opt) {
578 			memcpy(&opt_copy, inet_opt,
579 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
580 			ipc.opt = &opt_copy.opt;
581 		}
582 		rcu_read_unlock();
583 	}
584 
585 	if (ipc.opt) {
586 		err = -EINVAL;
587 		/* Linux does not mangle headers on raw sockets,
588 		 * so that IP options + IP_HDRINCL is non-sense.
589 		 */
590 		if (inet->hdrincl)
591 			goto done;
592 		if (ipc.opt->opt.srr) {
593 			if (!daddr)
594 				goto done;
595 			daddr = ipc.opt->opt.faddr;
596 		}
597 	}
598 	tos = get_rtconn_flags(&ipc, sk);
599 	if (msg->msg_flags & MSG_DONTROUTE)
600 		tos |= RTO_ONLINK;
601 
602 	if (ipv4_is_multicast(daddr)) {
603 		if (!ipc.oif)
604 			ipc.oif = inet->mc_index;
605 		if (!saddr)
606 			saddr = inet->mc_addr;
607 	} else if (!ipc.oif)
608 		ipc.oif = inet->uc_index;
609 
610 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
611 			   RT_SCOPE_UNIVERSE,
612 			   inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
613 			   inet_sk_flowi_flags(sk) |
614 			    (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
615 			   daddr, saddr, 0, 0, sk->sk_uid);
616 
617 	if (!inet->hdrincl) {
618 		rfv.msg = msg;
619 		rfv.hlen = 0;
620 
621 		err = raw_probe_proto_opt(&rfv, &fl4);
622 		if (err)
623 			goto done;
624 	}
625 
626 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
627 	rt = ip_route_output_flow(net, &fl4, sk);
628 	if (IS_ERR(rt)) {
629 		err = PTR_ERR(rt);
630 		rt = NULL;
631 		goto done;
632 	}
633 
634 	err = -EACCES;
635 	if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
636 		goto done;
637 
638 	if (msg->msg_flags & MSG_CONFIRM)
639 		goto do_confirm;
640 back_from_confirm:
641 
642 	if (inet->hdrincl)
643 		err = raw_send_hdrinc(sk, &fl4, msg, len,
644 				      &rt, msg->msg_flags, &ipc.sockc);
645 
646 	 else {
647 		sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
648 
649 		if (!ipc.addr)
650 			ipc.addr = fl4.daddr;
651 		lock_sock(sk);
652 		err = ip_append_data(sk, &fl4, raw_getfrag,
653 				     &rfv, len, 0,
654 				     &ipc, &rt, msg->msg_flags);
655 		if (err)
656 			ip_flush_pending_frames(sk);
657 		else if (!(msg->msg_flags & MSG_MORE)) {
658 			err = ip_push_pending_frames(sk, &fl4);
659 			if (err == -ENOBUFS && !inet->recverr)
660 				err = 0;
661 		}
662 		release_sock(sk);
663 	}
664 done:
665 	if (free)
666 		kfree(ipc.opt);
667 	ip_rt_put(rt);
668 
669 out:
670 	if (err < 0)
671 		return err;
672 	return len;
673 
674 do_confirm:
675 	if (msg->msg_flags & MSG_PROBE)
676 		dst_confirm_neigh(&rt->dst, &fl4.daddr);
677 	if (!(msg->msg_flags & MSG_PROBE) || len)
678 		goto back_from_confirm;
679 	err = 0;
680 	goto done;
681 }
682 
683 static void raw_close(struct sock *sk, long timeout)
684 {
685 	/*
686 	 * Raw sockets may have direct kernel references. Kill them.
687 	 */
688 	rtnl_lock();
689 	ip_ra_control(sk, 0, NULL);
690 	rtnl_unlock();
691 
692 	sk_common_release(sk);
693 }
694 
695 static void raw_destroy(struct sock *sk)
696 {
697 	lock_sock(sk);
698 	ip_flush_pending_frames(sk);
699 	release_sock(sk);
700 }
701 
702 /* This gets rid of all the nasties in af_inet. -DaveM */
703 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
704 {
705 	struct inet_sock *inet = inet_sk(sk);
706 	struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
707 	u32 tb_id = RT_TABLE_LOCAL;
708 	int ret = -EINVAL;
709 	int chk_addr_ret;
710 
711 	if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
712 		goto out;
713 
714 	if (sk->sk_bound_dev_if)
715 		tb_id = l3mdev_fib_table_by_index(sock_net(sk),
716 						 sk->sk_bound_dev_if) ? : tb_id;
717 
718 	chk_addr_ret = inet_addr_type_table(sock_net(sk), addr->sin_addr.s_addr,
719 					    tb_id);
720 
721 	ret = -EADDRNOTAVAIL;
722 	if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
723 	    chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
724 		goto out;
725 	inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
726 	if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
727 		inet->inet_saddr = 0;  /* Use device */
728 	sk_dst_reset(sk);
729 	ret = 0;
730 out:	return ret;
731 }
732 
733 /*
734  *	This should be easy, if there is something there
735  *	we return it, otherwise we block.
736  */
737 
738 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
739 		       int noblock, int flags, int *addr_len)
740 {
741 	struct inet_sock *inet = inet_sk(sk);
742 	size_t copied = 0;
743 	int err = -EOPNOTSUPP;
744 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
745 	struct sk_buff *skb;
746 
747 	if (flags & MSG_OOB)
748 		goto out;
749 
750 	if (flags & MSG_ERRQUEUE) {
751 		err = ip_recv_error(sk, msg, len, addr_len);
752 		goto out;
753 	}
754 
755 	skb = skb_recv_datagram(sk, flags, noblock, &err);
756 	if (!skb)
757 		goto out;
758 
759 	copied = skb->len;
760 	if (len < copied) {
761 		msg->msg_flags |= MSG_TRUNC;
762 		copied = len;
763 	}
764 
765 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
766 	if (err)
767 		goto done;
768 
769 	sock_recv_ts_and_drops(msg, sk, skb);
770 
771 	/* Copy the address. */
772 	if (sin) {
773 		sin->sin_family = AF_INET;
774 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
775 		sin->sin_port = 0;
776 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
777 		*addr_len = sizeof(*sin);
778 	}
779 	if (inet->cmsg_flags)
780 		ip_cmsg_recv(msg, skb);
781 	if (flags & MSG_TRUNC)
782 		copied = skb->len;
783 done:
784 	skb_free_datagram(sk, skb);
785 out:
786 	if (err)
787 		return err;
788 	return copied;
789 }
790 
791 static int raw_init(struct sock *sk)
792 {
793 	struct raw_sock *rp = raw_sk(sk);
794 
795 	if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
796 		memset(&rp->filter, 0, sizeof(rp->filter));
797 	return 0;
798 }
799 
800 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
801 {
802 	if (optlen > sizeof(struct icmp_filter))
803 		optlen = sizeof(struct icmp_filter);
804 	if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
805 		return -EFAULT;
806 	return 0;
807 }
808 
809 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
810 {
811 	int len, ret = -EFAULT;
812 
813 	if (get_user(len, optlen))
814 		goto out;
815 	ret = -EINVAL;
816 	if (len < 0)
817 		goto out;
818 	if (len > sizeof(struct icmp_filter))
819 		len = sizeof(struct icmp_filter);
820 	ret = -EFAULT;
821 	if (put_user(len, optlen) ||
822 	    copy_to_user(optval, &raw_sk(sk)->filter, len))
823 		goto out;
824 	ret = 0;
825 out:	return ret;
826 }
827 
828 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
829 			  char __user *optval, unsigned int optlen)
830 {
831 	if (optname == ICMP_FILTER) {
832 		if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
833 			return -EOPNOTSUPP;
834 		else
835 			return raw_seticmpfilter(sk, optval, optlen);
836 	}
837 	return -ENOPROTOOPT;
838 }
839 
840 static int raw_setsockopt(struct sock *sk, int level, int optname,
841 			  char __user *optval, unsigned int optlen)
842 {
843 	if (level != SOL_RAW)
844 		return ip_setsockopt(sk, level, optname, optval, optlen);
845 	return do_raw_setsockopt(sk, level, optname, optval, optlen);
846 }
847 
848 #ifdef CONFIG_COMPAT
849 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
850 				 char __user *optval, unsigned int optlen)
851 {
852 	if (level != SOL_RAW)
853 		return compat_ip_setsockopt(sk, level, optname, optval, optlen);
854 	return do_raw_setsockopt(sk, level, optname, optval, optlen);
855 }
856 #endif
857 
858 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
859 			  char __user *optval, int __user *optlen)
860 {
861 	if (optname == ICMP_FILTER) {
862 		if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
863 			return -EOPNOTSUPP;
864 		else
865 			return raw_geticmpfilter(sk, optval, optlen);
866 	}
867 	return -ENOPROTOOPT;
868 }
869 
870 static int raw_getsockopt(struct sock *sk, int level, int optname,
871 			  char __user *optval, int __user *optlen)
872 {
873 	if (level != SOL_RAW)
874 		return ip_getsockopt(sk, level, optname, optval, optlen);
875 	return do_raw_getsockopt(sk, level, optname, optval, optlen);
876 }
877 
878 #ifdef CONFIG_COMPAT
879 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
880 				 char __user *optval, int __user *optlen)
881 {
882 	if (level != SOL_RAW)
883 		return compat_ip_getsockopt(sk, level, optname, optval, optlen);
884 	return do_raw_getsockopt(sk, level, optname, optval, optlen);
885 }
886 #endif
887 
888 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
889 {
890 	switch (cmd) {
891 	case SIOCOUTQ: {
892 		int amount = sk_wmem_alloc_get(sk);
893 
894 		return put_user(amount, (int __user *)arg);
895 	}
896 	case SIOCINQ: {
897 		struct sk_buff *skb;
898 		int amount = 0;
899 
900 		spin_lock_bh(&sk->sk_receive_queue.lock);
901 		skb = skb_peek(&sk->sk_receive_queue);
902 		if (skb)
903 			amount = skb->len;
904 		spin_unlock_bh(&sk->sk_receive_queue.lock);
905 		return put_user(amount, (int __user *)arg);
906 	}
907 
908 	default:
909 #ifdef CONFIG_IP_MROUTE
910 		return ipmr_ioctl(sk, cmd, (void __user *)arg);
911 #else
912 		return -ENOIOCTLCMD;
913 #endif
914 	}
915 }
916 
917 #ifdef CONFIG_COMPAT
918 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
919 {
920 	switch (cmd) {
921 	case SIOCOUTQ:
922 	case SIOCINQ:
923 		return -ENOIOCTLCMD;
924 	default:
925 #ifdef CONFIG_IP_MROUTE
926 		return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
927 #else
928 		return -ENOIOCTLCMD;
929 #endif
930 	}
931 }
932 #endif
933 
934 int raw_abort(struct sock *sk, int err)
935 {
936 	lock_sock(sk);
937 
938 	sk->sk_err = err;
939 	sk->sk_error_report(sk);
940 	__udp_disconnect(sk, 0);
941 
942 	release_sock(sk);
943 
944 	return 0;
945 }
946 EXPORT_SYMBOL_GPL(raw_abort);
947 
948 struct proto raw_prot = {
949 	.name		   = "RAW",
950 	.owner		   = THIS_MODULE,
951 	.close		   = raw_close,
952 	.destroy	   = raw_destroy,
953 	.connect	   = ip4_datagram_connect,
954 	.disconnect	   = __udp_disconnect,
955 	.ioctl		   = raw_ioctl,
956 	.init		   = raw_init,
957 	.setsockopt	   = raw_setsockopt,
958 	.getsockopt	   = raw_getsockopt,
959 	.sendmsg	   = raw_sendmsg,
960 	.recvmsg	   = raw_recvmsg,
961 	.bind		   = raw_bind,
962 	.backlog_rcv	   = raw_rcv_skb,
963 	.release_cb	   = ip4_datagram_release_cb,
964 	.hash		   = raw_hash_sk,
965 	.unhash		   = raw_unhash_sk,
966 	.obj_size	   = sizeof(struct raw_sock),
967 	.h.raw_hash	   = &raw_v4_hashinfo,
968 #ifdef CONFIG_COMPAT
969 	.compat_setsockopt = compat_raw_setsockopt,
970 	.compat_getsockopt = compat_raw_getsockopt,
971 	.compat_ioctl	   = compat_raw_ioctl,
972 #endif
973 	.diag_destroy	   = raw_abort,
974 };
975 
976 #ifdef CONFIG_PROC_FS
977 static struct sock *raw_get_first(struct seq_file *seq)
978 {
979 	struct sock *sk;
980 	struct raw_iter_state *state = raw_seq_private(seq);
981 
982 	for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
983 			++state->bucket) {
984 		sk_for_each(sk, &state->h->ht[state->bucket])
985 			if (sock_net(sk) == seq_file_net(seq))
986 				goto found;
987 	}
988 	sk = NULL;
989 found:
990 	return sk;
991 }
992 
993 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
994 {
995 	struct raw_iter_state *state = raw_seq_private(seq);
996 
997 	do {
998 		sk = sk_next(sk);
999 try_again:
1000 		;
1001 	} while (sk && sock_net(sk) != seq_file_net(seq));
1002 
1003 	if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
1004 		sk = sk_head(&state->h->ht[state->bucket]);
1005 		goto try_again;
1006 	}
1007 	return sk;
1008 }
1009 
1010 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
1011 {
1012 	struct sock *sk = raw_get_first(seq);
1013 
1014 	if (sk)
1015 		while (pos && (sk = raw_get_next(seq, sk)) != NULL)
1016 			--pos;
1017 	return pos ? NULL : sk;
1018 }
1019 
1020 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
1021 {
1022 	struct raw_iter_state *state = raw_seq_private(seq);
1023 
1024 	read_lock(&state->h->lock);
1025 	return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
1026 }
1027 EXPORT_SYMBOL_GPL(raw_seq_start);
1028 
1029 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1030 {
1031 	struct sock *sk;
1032 
1033 	if (v == SEQ_START_TOKEN)
1034 		sk = raw_get_first(seq);
1035 	else
1036 		sk = raw_get_next(seq, v);
1037 	++*pos;
1038 	return sk;
1039 }
1040 EXPORT_SYMBOL_GPL(raw_seq_next);
1041 
1042 void raw_seq_stop(struct seq_file *seq, void *v)
1043 {
1044 	struct raw_iter_state *state = raw_seq_private(seq);
1045 
1046 	read_unlock(&state->h->lock);
1047 }
1048 EXPORT_SYMBOL_GPL(raw_seq_stop);
1049 
1050 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1051 {
1052 	struct inet_sock *inet = inet_sk(sp);
1053 	__be32 dest = inet->inet_daddr,
1054 	       src = inet->inet_rcv_saddr;
1055 	__u16 destp = 0,
1056 	      srcp  = inet->inet_num;
1057 
1058 	seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1059 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1060 		i, src, srcp, dest, destp, sp->sk_state,
1061 		sk_wmem_alloc_get(sp),
1062 		sk_rmem_alloc_get(sp),
1063 		0, 0L, 0,
1064 		from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1065 		0, sock_i_ino(sp),
1066 		refcount_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1067 }
1068 
1069 static int raw_seq_show(struct seq_file *seq, void *v)
1070 {
1071 	if (v == SEQ_START_TOKEN)
1072 		seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
1073 				"rx_queue tr tm->when retrnsmt   uid  timeout "
1074 				"inode ref pointer drops\n");
1075 	else
1076 		raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1077 	return 0;
1078 }
1079 
1080 static const struct seq_operations raw_seq_ops = {
1081 	.start = raw_seq_start,
1082 	.next  = raw_seq_next,
1083 	.stop  = raw_seq_stop,
1084 	.show  = raw_seq_show,
1085 };
1086 
1087 int raw_seq_open(struct inode *ino, struct file *file,
1088 		 struct raw_hashinfo *h, const struct seq_operations *ops)
1089 {
1090 	int err;
1091 	struct raw_iter_state *i;
1092 
1093 	err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1094 	if (err < 0)
1095 		return err;
1096 
1097 	i = raw_seq_private((struct seq_file *)file->private_data);
1098 	i->h = h;
1099 	return 0;
1100 }
1101 EXPORT_SYMBOL_GPL(raw_seq_open);
1102 
1103 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1104 {
1105 	return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1106 }
1107 
1108 static const struct file_operations raw_seq_fops = {
1109 	.owner	 = THIS_MODULE,
1110 	.open	 = raw_v4_seq_open,
1111 	.read	 = seq_read,
1112 	.llseek	 = seq_lseek,
1113 	.release = seq_release_net,
1114 };
1115 
1116 static __net_init int raw_init_net(struct net *net)
1117 {
1118 	if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1119 		return -ENOMEM;
1120 
1121 	return 0;
1122 }
1123 
1124 static __net_exit void raw_exit_net(struct net *net)
1125 {
1126 	remove_proc_entry("raw", net->proc_net);
1127 }
1128 
1129 static __net_initdata struct pernet_operations raw_net_ops = {
1130 	.init = raw_init_net,
1131 	.exit = raw_exit_net,
1132 };
1133 
1134 int __init raw_proc_init(void)
1135 {
1136 	return register_pernet_subsys(&raw_net_ops);
1137 }
1138 
1139 void __init raw_proc_exit(void)
1140 {
1141 	unregister_pernet_subsys(&raw_net_ops);
1142 }
1143 #endif /* CONFIG_PROC_FS */
1144