xref: /linux/net/ipv4/raw.c (revision b43ab901d671e3e3cad425ea5e9a3c74e266dcdd)
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 <asm/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/aio.h>
50 #include <linux/kernel.h>
51 #include <linux/export.h>
52 #include <linux/spinlock.h>
53 #include <linux/sockios.h>
54 #include <linux/socket.h>
55 #include <linux/in.h>
56 #include <linux/mroute.h>
57 #include <linux/netdevice.h>
58 #include <linux/in_route.h>
59 #include <linux/route.h>
60 #include <linux/skbuff.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 
82 static struct raw_hashinfo raw_v4_hashinfo = {
83 	.lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
84 };
85 
86 void raw_hash_sk(struct sock *sk)
87 {
88 	struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
89 	struct hlist_head *head;
90 
91 	head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
92 
93 	write_lock_bh(&h->lock);
94 	sk_add_node(sk, head);
95 	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
96 	write_unlock_bh(&h->lock);
97 }
98 EXPORT_SYMBOL_GPL(raw_hash_sk);
99 
100 void raw_unhash_sk(struct sock *sk)
101 {
102 	struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
103 
104 	write_lock_bh(&h->lock);
105 	if (sk_del_node_init(sk))
106 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
107 	write_unlock_bh(&h->lock);
108 }
109 EXPORT_SYMBOL_GPL(raw_unhash_sk);
110 
111 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
112 		unsigned short num, __be32 raddr, __be32 laddr, int dif)
113 {
114 	struct hlist_node *node;
115 
116 	sk_for_each_from(sk, node) {
117 		struct inet_sock *inet = inet_sk(sk);
118 
119 		if (net_eq(sock_net(sk), net) && inet->inet_num == num	&&
120 		    !(inet->inet_daddr && inet->inet_daddr != raddr) 	&&
121 		    !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
122 		    !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
123 			goto found; /* gotcha */
124 	}
125 	sk = NULL;
126 found:
127 	return sk;
128 }
129 
130 /*
131  *	0 - deliver
132  *	1 - block
133  */
134 static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
135 {
136 	int type;
137 
138 	if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
139 		return 1;
140 
141 	type = icmp_hdr(skb)->type;
142 	if (type < 32) {
143 		__u32 data = raw_sk(sk)->filter.data;
144 
145 		return ((1 << type) & data) != 0;
146 	}
147 
148 	/* Do not block unknown ICMP types */
149 	return 0;
150 }
151 
152 /* IP input processing comes here for RAW socket delivery.
153  * Caller owns SKB, so we must make clones.
154  *
155  * RFC 1122: SHOULD pass TOS value up to the transport layer.
156  * -> It does. And not only TOS, but all IP header.
157  */
158 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
159 {
160 	struct sock *sk;
161 	struct hlist_head *head;
162 	int delivered = 0;
163 	struct net *net;
164 
165 	read_lock(&raw_v4_hashinfo.lock);
166 	head = &raw_v4_hashinfo.ht[hash];
167 	if (hlist_empty(head))
168 		goto out;
169 
170 	net = dev_net(skb->dev);
171 	sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
172 			     iph->saddr, iph->daddr,
173 			     skb->dev->ifindex);
174 
175 	while (sk) {
176 		delivered = 1;
177 		if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
178 			struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
179 
180 			/* Not releasing hash table! */
181 			if (clone)
182 				raw_rcv(sk, clone);
183 		}
184 		sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
185 				     iph->saddr, iph->daddr,
186 				     skb->dev->ifindex);
187 	}
188 out:
189 	read_unlock(&raw_v4_hashinfo.lock);
190 	return delivered;
191 }
192 
193 int raw_local_deliver(struct sk_buff *skb, int protocol)
194 {
195 	int hash;
196 	struct sock *raw_sk;
197 
198 	hash = protocol & (RAW_HTABLE_SIZE - 1);
199 	raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
200 
201 	/* If there maybe a raw socket we must check - if not we
202 	 * don't care less
203 	 */
204 	if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
205 		raw_sk = NULL;
206 
207 	return raw_sk != NULL;
208 
209 }
210 
211 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
212 {
213 	struct inet_sock *inet = inet_sk(sk);
214 	const int type = icmp_hdr(skb)->type;
215 	const int code = icmp_hdr(skb)->code;
216 	int err = 0;
217 	int harderr = 0;
218 
219 	/* Report error on raw socket, if:
220 	   1. User requested ip_recverr.
221 	   2. Socket is connected (otherwise the error indication
222 	      is useless without ip_recverr and error is hard.
223 	 */
224 	if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
225 		return;
226 
227 	switch (type) {
228 	default:
229 	case ICMP_TIME_EXCEEDED:
230 		err = EHOSTUNREACH;
231 		break;
232 	case ICMP_SOURCE_QUENCH:
233 		return;
234 	case ICMP_PARAMETERPROB:
235 		err = EPROTO;
236 		harderr = 1;
237 		break;
238 	case ICMP_DEST_UNREACH:
239 		err = EHOSTUNREACH;
240 		if (code > NR_ICMP_UNREACH)
241 			break;
242 		err = icmp_err_convert[code].errno;
243 		harderr = icmp_err_convert[code].fatal;
244 		if (code == ICMP_FRAG_NEEDED) {
245 			harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
246 			err = EMSGSIZE;
247 		}
248 	}
249 
250 	if (inet->recverr) {
251 		const struct iphdr *iph = (const struct iphdr *)skb->data;
252 		u8 *payload = skb->data + (iph->ihl << 2);
253 
254 		if (inet->hdrincl)
255 			payload = skb->data;
256 		ip_icmp_error(sk, skb, err, 0, info, payload);
257 	}
258 
259 	if (inet->recverr || harderr) {
260 		sk->sk_err = err;
261 		sk->sk_error_report(sk);
262 	}
263 }
264 
265 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
266 {
267 	int hash;
268 	struct sock *raw_sk;
269 	const struct iphdr *iph;
270 	struct net *net;
271 
272 	hash = protocol & (RAW_HTABLE_SIZE - 1);
273 
274 	read_lock(&raw_v4_hashinfo.lock);
275 	raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
276 	if (raw_sk != NULL) {
277 		iph = (const struct iphdr *)skb->data;
278 		net = dev_net(skb->dev);
279 
280 		while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
281 						iph->daddr, iph->saddr,
282 						skb->dev->ifindex)) != NULL) {
283 			raw_err(raw_sk, skb, info);
284 			raw_sk = sk_next(raw_sk);
285 			iph = (const struct iphdr *)skb->data;
286 		}
287 	}
288 	read_unlock(&raw_v4_hashinfo.lock);
289 }
290 
291 static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
292 {
293 	/* Charge it to the socket. */
294 
295 	ipv4_pktinfo_prepare(skb);
296 	if (sock_queue_rcv_skb(sk, skb) < 0) {
297 		kfree_skb(skb);
298 		return NET_RX_DROP;
299 	}
300 
301 	return NET_RX_SUCCESS;
302 }
303 
304 int raw_rcv(struct sock *sk, struct sk_buff *skb)
305 {
306 	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
307 		atomic_inc(&sk->sk_drops);
308 		kfree_skb(skb);
309 		return NET_RX_DROP;
310 	}
311 	nf_reset(skb);
312 
313 	skb_push(skb, skb->data - skb_network_header(skb));
314 
315 	raw_rcv_skb(sk, skb);
316 	return 0;
317 }
318 
319 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
320 			   void *from, size_t length,
321 			   struct rtable **rtp,
322 			   unsigned int flags)
323 {
324 	struct inet_sock *inet = inet_sk(sk);
325 	struct net *net = sock_net(sk);
326 	struct iphdr *iph;
327 	struct sk_buff *skb;
328 	unsigned int iphlen;
329 	int err;
330 	struct rtable *rt = *rtp;
331 	int hlen, tlen;
332 
333 	if (length > rt->dst.dev->mtu) {
334 		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
335 			       rt->dst.dev->mtu);
336 		return -EMSGSIZE;
337 	}
338 	if (flags&MSG_PROBE)
339 		goto out;
340 
341 	hlen = LL_RESERVED_SPACE(rt->dst.dev);
342 	tlen = rt->dst.dev->needed_tailroom;
343 	skb = sock_alloc_send_skb(sk,
344 				  length + hlen + tlen + 15,
345 				  flags & MSG_DONTWAIT, &err);
346 	if (skb == NULL)
347 		goto error;
348 	skb_reserve(skb, hlen);
349 
350 	skb->priority = sk->sk_priority;
351 	skb->mark = sk->sk_mark;
352 	skb_dst_set(skb, &rt->dst);
353 	*rtp = NULL;
354 
355 	skb_reset_network_header(skb);
356 	iph = ip_hdr(skb);
357 	skb_put(skb, length);
358 
359 	skb->ip_summed = CHECKSUM_NONE;
360 
361 	skb->transport_header = skb->network_header;
362 	err = -EFAULT;
363 	if (memcpy_fromiovecend((void *)iph, from, 0, length))
364 		goto error_free;
365 
366 	iphlen = iph->ihl * 4;
367 
368 	/*
369 	 * We don't want to modify the ip header, but we do need to
370 	 * be sure that it won't cause problems later along the network
371 	 * stack.  Specifically we want to make sure that iph->ihl is a
372 	 * sane value.  If ihl points beyond the length of the buffer passed
373 	 * in, reject the frame as invalid
374 	 */
375 	err = -EINVAL;
376 	if (iphlen > length)
377 		goto error_free;
378 
379 	if (iphlen >= sizeof(*iph)) {
380 		if (!iph->saddr)
381 			iph->saddr = fl4->saddr;
382 		iph->check   = 0;
383 		iph->tot_len = htons(length);
384 		if (!iph->id)
385 			ip_select_ident(iph, &rt->dst, NULL);
386 
387 		iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
388 	}
389 	if (iph->protocol == IPPROTO_ICMP)
390 		icmp_out_count(net, ((struct icmphdr *)
391 			skb_transport_header(skb))->type);
392 
393 	err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT, skb, NULL,
394 		      rt->dst.dev, dst_output);
395 	if (err > 0)
396 		err = net_xmit_errno(err);
397 	if (err)
398 		goto error;
399 out:
400 	return 0;
401 
402 error_free:
403 	kfree_skb(skb);
404 error:
405 	IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
406 	if (err == -ENOBUFS && !inet->recverr)
407 		err = 0;
408 	return err;
409 }
410 
411 static int raw_probe_proto_opt(struct flowi4 *fl4, struct msghdr *msg)
412 {
413 	struct iovec *iov;
414 	u8 __user *type = NULL;
415 	u8 __user *code = NULL;
416 	int probed = 0;
417 	unsigned int i;
418 
419 	if (!msg->msg_iov)
420 		return 0;
421 
422 	for (i = 0; i < msg->msg_iovlen; i++) {
423 		iov = &msg->msg_iov[i];
424 		if (!iov)
425 			continue;
426 
427 		switch (fl4->flowi4_proto) {
428 		case IPPROTO_ICMP:
429 			/* check if one-byte field is readable or not. */
430 			if (iov->iov_base && iov->iov_len < 1)
431 				break;
432 
433 			if (!type) {
434 				type = iov->iov_base;
435 				/* check if code field is readable or not. */
436 				if (iov->iov_len > 1)
437 					code = type + 1;
438 			} else if (!code)
439 				code = iov->iov_base;
440 
441 			if (type && code) {
442 				if (get_user(fl4->fl4_icmp_type, type) ||
443 				    get_user(fl4->fl4_icmp_code, code))
444 					return -EFAULT;
445 				probed = 1;
446 			}
447 			break;
448 		default:
449 			probed = 1;
450 			break;
451 		}
452 		if (probed)
453 			break;
454 	}
455 	return 0;
456 }
457 
458 static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
459 		       size_t len)
460 {
461 	struct inet_sock *inet = inet_sk(sk);
462 	struct ipcm_cookie ipc;
463 	struct rtable *rt = NULL;
464 	struct flowi4 fl4;
465 	int free = 0;
466 	__be32 daddr;
467 	__be32 saddr;
468 	u8  tos;
469 	int err;
470 	struct ip_options_data opt_copy;
471 
472 	err = -EMSGSIZE;
473 	if (len > 0xFFFF)
474 		goto out;
475 
476 	/*
477 	 *	Check the flags.
478 	 */
479 
480 	err = -EOPNOTSUPP;
481 	if (msg->msg_flags & MSG_OOB)	/* Mirror BSD error message */
482 		goto out;               /* compatibility */
483 
484 	/*
485 	 *	Get and verify the address.
486 	 */
487 
488 	if (msg->msg_namelen) {
489 		struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
490 		err = -EINVAL;
491 		if (msg->msg_namelen < sizeof(*usin))
492 			goto out;
493 		if (usin->sin_family != AF_INET) {
494 			static int complained;
495 			if (!complained++)
496 				printk(KERN_INFO "%s forgot to set AF_INET in "
497 						 "raw sendmsg. Fix it!\n",
498 						 current->comm);
499 			err = -EAFNOSUPPORT;
500 			if (usin->sin_family)
501 				goto out;
502 		}
503 		daddr = usin->sin_addr.s_addr;
504 		/* ANK: I did not forget to get protocol from port field.
505 		 * I just do not know, who uses this weirdness.
506 		 * IP_HDRINCL is much more convenient.
507 		 */
508 	} else {
509 		err = -EDESTADDRREQ;
510 		if (sk->sk_state != TCP_ESTABLISHED)
511 			goto out;
512 		daddr = inet->inet_daddr;
513 	}
514 
515 	ipc.addr = inet->inet_saddr;
516 	ipc.opt = NULL;
517 	ipc.tx_flags = 0;
518 	ipc.oif = sk->sk_bound_dev_if;
519 
520 	if (msg->msg_controllen) {
521 		err = ip_cmsg_send(sock_net(sk), msg, &ipc);
522 		if (err)
523 			goto out;
524 		if (ipc.opt)
525 			free = 1;
526 	}
527 
528 	saddr = ipc.addr;
529 	ipc.addr = daddr;
530 
531 	if (!ipc.opt) {
532 		struct ip_options_rcu *inet_opt;
533 
534 		rcu_read_lock();
535 		inet_opt = rcu_dereference(inet->inet_opt);
536 		if (inet_opt) {
537 			memcpy(&opt_copy, inet_opt,
538 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
539 			ipc.opt = &opt_copy.opt;
540 		}
541 		rcu_read_unlock();
542 	}
543 
544 	if (ipc.opt) {
545 		err = -EINVAL;
546 		/* Linux does not mangle headers on raw sockets,
547 		 * so that IP options + IP_HDRINCL is non-sense.
548 		 */
549 		if (inet->hdrincl)
550 			goto done;
551 		if (ipc.opt->opt.srr) {
552 			if (!daddr)
553 				goto done;
554 			daddr = ipc.opt->opt.faddr;
555 		}
556 	}
557 	tos = RT_CONN_FLAGS(sk);
558 	if (msg->msg_flags & MSG_DONTROUTE)
559 		tos |= RTO_ONLINK;
560 
561 	if (ipv4_is_multicast(daddr)) {
562 		if (!ipc.oif)
563 			ipc.oif = inet->mc_index;
564 		if (!saddr)
565 			saddr = inet->mc_addr;
566 	}
567 
568 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
569 			   RT_SCOPE_UNIVERSE,
570 			   inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
571 			   inet_sk_flowi_flags(sk) | FLOWI_FLAG_CAN_SLEEP,
572 			   daddr, saddr, 0, 0);
573 
574 	if (!inet->hdrincl) {
575 		err = raw_probe_proto_opt(&fl4, msg);
576 		if (err)
577 			goto done;
578 	}
579 
580 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
581 	rt = ip_route_output_flow(sock_net(sk), &fl4, sk);
582 	if (IS_ERR(rt)) {
583 		err = PTR_ERR(rt);
584 		rt = NULL;
585 		goto done;
586 	}
587 
588 	err = -EACCES;
589 	if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
590 		goto done;
591 
592 	if (msg->msg_flags & MSG_CONFIRM)
593 		goto do_confirm;
594 back_from_confirm:
595 
596 	if (inet->hdrincl)
597 		err = raw_send_hdrinc(sk, &fl4, msg->msg_iov, len,
598 				      &rt, msg->msg_flags);
599 
600 	 else {
601 		if (!ipc.addr)
602 			ipc.addr = fl4.daddr;
603 		lock_sock(sk);
604 		err = ip_append_data(sk, &fl4, ip_generic_getfrag,
605 				     msg->msg_iov, len, 0,
606 				     &ipc, &rt, msg->msg_flags);
607 		if (err)
608 			ip_flush_pending_frames(sk);
609 		else if (!(msg->msg_flags & MSG_MORE)) {
610 			err = ip_push_pending_frames(sk, &fl4);
611 			if (err == -ENOBUFS && !inet->recverr)
612 				err = 0;
613 		}
614 		release_sock(sk);
615 	}
616 done:
617 	if (free)
618 		kfree(ipc.opt);
619 	ip_rt_put(rt);
620 
621 out:
622 	if (err < 0)
623 		return err;
624 	return len;
625 
626 do_confirm:
627 	dst_confirm(&rt->dst);
628 	if (!(msg->msg_flags & MSG_PROBE) || len)
629 		goto back_from_confirm;
630 	err = 0;
631 	goto done;
632 }
633 
634 static void raw_close(struct sock *sk, long timeout)
635 {
636 	/*
637 	 * Raw sockets may have direct kernel references. Kill them.
638 	 */
639 	ip_ra_control(sk, 0, NULL);
640 
641 	sk_common_release(sk);
642 }
643 
644 static void raw_destroy(struct sock *sk)
645 {
646 	lock_sock(sk);
647 	ip_flush_pending_frames(sk);
648 	release_sock(sk);
649 }
650 
651 /* This gets rid of all the nasties in af_inet. -DaveM */
652 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
653 {
654 	struct inet_sock *inet = inet_sk(sk);
655 	struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
656 	int ret = -EINVAL;
657 	int chk_addr_ret;
658 
659 	if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
660 		goto out;
661 	chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
662 	ret = -EADDRNOTAVAIL;
663 	if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
664 	    chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
665 		goto out;
666 	inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
667 	if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
668 		inet->inet_saddr = 0;  /* Use device */
669 	sk_dst_reset(sk);
670 	ret = 0;
671 out:	return ret;
672 }
673 
674 /*
675  *	This should be easy, if there is something there
676  *	we return it, otherwise we block.
677  */
678 
679 static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
680 		       size_t len, int noblock, int flags, int *addr_len)
681 {
682 	struct inet_sock *inet = inet_sk(sk);
683 	size_t copied = 0;
684 	int err = -EOPNOTSUPP;
685 	struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
686 	struct sk_buff *skb;
687 
688 	if (flags & MSG_OOB)
689 		goto out;
690 
691 	if (addr_len)
692 		*addr_len = sizeof(*sin);
693 
694 	if (flags & MSG_ERRQUEUE) {
695 		err = ip_recv_error(sk, msg, len);
696 		goto out;
697 	}
698 
699 	skb = skb_recv_datagram(sk, flags, noblock, &err);
700 	if (!skb)
701 		goto out;
702 
703 	copied = skb->len;
704 	if (len < copied) {
705 		msg->msg_flags |= MSG_TRUNC;
706 		copied = len;
707 	}
708 
709 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
710 	if (err)
711 		goto done;
712 
713 	sock_recv_ts_and_drops(msg, sk, skb);
714 
715 	/* Copy the address. */
716 	if (sin) {
717 		sin->sin_family = AF_INET;
718 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
719 		sin->sin_port = 0;
720 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
721 	}
722 	if (inet->cmsg_flags)
723 		ip_cmsg_recv(msg, skb);
724 	if (flags & MSG_TRUNC)
725 		copied = skb->len;
726 done:
727 	skb_free_datagram(sk, skb);
728 out:
729 	if (err)
730 		return err;
731 	return copied;
732 }
733 
734 static int raw_init(struct sock *sk)
735 {
736 	struct raw_sock *rp = raw_sk(sk);
737 
738 	if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
739 		memset(&rp->filter, 0, sizeof(rp->filter));
740 	return 0;
741 }
742 
743 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
744 {
745 	if (optlen > sizeof(struct icmp_filter))
746 		optlen = sizeof(struct icmp_filter);
747 	if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
748 		return -EFAULT;
749 	return 0;
750 }
751 
752 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
753 {
754 	int len, ret = -EFAULT;
755 
756 	if (get_user(len, optlen))
757 		goto out;
758 	ret = -EINVAL;
759 	if (len < 0)
760 		goto out;
761 	if (len > sizeof(struct icmp_filter))
762 		len = sizeof(struct icmp_filter);
763 	ret = -EFAULT;
764 	if (put_user(len, optlen) ||
765 	    copy_to_user(optval, &raw_sk(sk)->filter, len))
766 		goto out;
767 	ret = 0;
768 out:	return ret;
769 }
770 
771 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
772 			  char __user *optval, unsigned int optlen)
773 {
774 	if (optname == ICMP_FILTER) {
775 		if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
776 			return -EOPNOTSUPP;
777 		else
778 			return raw_seticmpfilter(sk, optval, optlen);
779 	}
780 	return -ENOPROTOOPT;
781 }
782 
783 static int raw_setsockopt(struct sock *sk, int level, int optname,
784 			  char __user *optval, unsigned int optlen)
785 {
786 	if (level != SOL_RAW)
787 		return ip_setsockopt(sk, level, optname, optval, optlen);
788 	return do_raw_setsockopt(sk, level, optname, optval, optlen);
789 }
790 
791 #ifdef CONFIG_COMPAT
792 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
793 				 char __user *optval, unsigned int optlen)
794 {
795 	if (level != SOL_RAW)
796 		return compat_ip_setsockopt(sk, level, optname, optval, optlen);
797 	return do_raw_setsockopt(sk, level, optname, optval, optlen);
798 }
799 #endif
800 
801 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
802 			  char __user *optval, int __user *optlen)
803 {
804 	if (optname == ICMP_FILTER) {
805 		if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
806 			return -EOPNOTSUPP;
807 		else
808 			return raw_geticmpfilter(sk, optval, optlen);
809 	}
810 	return -ENOPROTOOPT;
811 }
812 
813 static int raw_getsockopt(struct sock *sk, int level, int optname,
814 			  char __user *optval, int __user *optlen)
815 {
816 	if (level != SOL_RAW)
817 		return ip_getsockopt(sk, level, optname, optval, optlen);
818 	return do_raw_getsockopt(sk, level, optname, optval, optlen);
819 }
820 
821 #ifdef CONFIG_COMPAT
822 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
823 				 char __user *optval, int __user *optlen)
824 {
825 	if (level != SOL_RAW)
826 		return compat_ip_getsockopt(sk, level, optname, optval, optlen);
827 	return do_raw_getsockopt(sk, level, optname, optval, optlen);
828 }
829 #endif
830 
831 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
832 {
833 	switch (cmd) {
834 	case SIOCOUTQ: {
835 		int amount = sk_wmem_alloc_get(sk);
836 
837 		return put_user(amount, (int __user *)arg);
838 	}
839 	case SIOCINQ: {
840 		struct sk_buff *skb;
841 		int amount = 0;
842 
843 		spin_lock_bh(&sk->sk_receive_queue.lock);
844 		skb = skb_peek(&sk->sk_receive_queue);
845 		if (skb != NULL)
846 			amount = skb->len;
847 		spin_unlock_bh(&sk->sk_receive_queue.lock);
848 		return put_user(amount, (int __user *)arg);
849 	}
850 
851 	default:
852 #ifdef CONFIG_IP_MROUTE
853 		return ipmr_ioctl(sk, cmd, (void __user *)arg);
854 #else
855 		return -ENOIOCTLCMD;
856 #endif
857 	}
858 }
859 
860 #ifdef CONFIG_COMPAT
861 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
862 {
863 	switch (cmd) {
864 	case SIOCOUTQ:
865 	case SIOCINQ:
866 		return -ENOIOCTLCMD;
867 	default:
868 #ifdef CONFIG_IP_MROUTE
869 		return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
870 #else
871 		return -ENOIOCTLCMD;
872 #endif
873 	}
874 }
875 #endif
876 
877 struct proto raw_prot = {
878 	.name		   = "RAW",
879 	.owner		   = THIS_MODULE,
880 	.close		   = raw_close,
881 	.destroy	   = raw_destroy,
882 	.connect	   = ip4_datagram_connect,
883 	.disconnect	   = udp_disconnect,
884 	.ioctl		   = raw_ioctl,
885 	.init		   = raw_init,
886 	.setsockopt	   = raw_setsockopt,
887 	.getsockopt	   = raw_getsockopt,
888 	.sendmsg	   = raw_sendmsg,
889 	.recvmsg	   = raw_recvmsg,
890 	.bind		   = raw_bind,
891 	.backlog_rcv	   = raw_rcv_skb,
892 	.hash		   = raw_hash_sk,
893 	.unhash		   = raw_unhash_sk,
894 	.obj_size	   = sizeof(struct raw_sock),
895 	.h.raw_hash	   = &raw_v4_hashinfo,
896 #ifdef CONFIG_COMPAT
897 	.compat_setsockopt = compat_raw_setsockopt,
898 	.compat_getsockopt = compat_raw_getsockopt,
899 	.compat_ioctl	   = compat_raw_ioctl,
900 #endif
901 };
902 
903 #ifdef CONFIG_PROC_FS
904 static struct sock *raw_get_first(struct seq_file *seq)
905 {
906 	struct sock *sk;
907 	struct raw_iter_state *state = raw_seq_private(seq);
908 
909 	for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
910 			++state->bucket) {
911 		struct hlist_node *node;
912 
913 		sk_for_each(sk, node, &state->h->ht[state->bucket])
914 			if (sock_net(sk) == seq_file_net(seq))
915 				goto found;
916 	}
917 	sk = NULL;
918 found:
919 	return sk;
920 }
921 
922 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
923 {
924 	struct raw_iter_state *state = raw_seq_private(seq);
925 
926 	do {
927 		sk = sk_next(sk);
928 try_again:
929 		;
930 	} while (sk && sock_net(sk) != seq_file_net(seq));
931 
932 	if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
933 		sk = sk_head(&state->h->ht[state->bucket]);
934 		goto try_again;
935 	}
936 	return sk;
937 }
938 
939 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
940 {
941 	struct sock *sk = raw_get_first(seq);
942 
943 	if (sk)
944 		while (pos && (sk = raw_get_next(seq, sk)) != NULL)
945 			--pos;
946 	return pos ? NULL : sk;
947 }
948 
949 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
950 {
951 	struct raw_iter_state *state = raw_seq_private(seq);
952 
953 	read_lock(&state->h->lock);
954 	return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
955 }
956 EXPORT_SYMBOL_GPL(raw_seq_start);
957 
958 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
959 {
960 	struct sock *sk;
961 
962 	if (v == SEQ_START_TOKEN)
963 		sk = raw_get_first(seq);
964 	else
965 		sk = raw_get_next(seq, v);
966 	++*pos;
967 	return sk;
968 }
969 EXPORT_SYMBOL_GPL(raw_seq_next);
970 
971 void raw_seq_stop(struct seq_file *seq, void *v)
972 {
973 	struct raw_iter_state *state = raw_seq_private(seq);
974 
975 	read_unlock(&state->h->lock);
976 }
977 EXPORT_SYMBOL_GPL(raw_seq_stop);
978 
979 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
980 {
981 	struct inet_sock *inet = inet_sk(sp);
982 	__be32 dest = inet->inet_daddr,
983 	       src = inet->inet_rcv_saddr;
984 	__u16 destp = 0,
985 	      srcp  = inet->inet_num;
986 
987 	seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
988 		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
989 		i, src, srcp, dest, destp, sp->sk_state,
990 		sk_wmem_alloc_get(sp),
991 		sk_rmem_alloc_get(sp),
992 		0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
993 		atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
994 }
995 
996 static int raw_seq_show(struct seq_file *seq, void *v)
997 {
998 	if (v == SEQ_START_TOKEN)
999 		seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
1000 				"rx_queue tr tm->when retrnsmt   uid  timeout "
1001 				"inode ref pointer drops\n");
1002 	else
1003 		raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1004 	return 0;
1005 }
1006 
1007 static const struct seq_operations raw_seq_ops = {
1008 	.start = raw_seq_start,
1009 	.next  = raw_seq_next,
1010 	.stop  = raw_seq_stop,
1011 	.show  = raw_seq_show,
1012 };
1013 
1014 int raw_seq_open(struct inode *ino, struct file *file,
1015 		 struct raw_hashinfo *h, const struct seq_operations *ops)
1016 {
1017 	int err;
1018 	struct raw_iter_state *i;
1019 
1020 	err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1021 	if (err < 0)
1022 		return err;
1023 
1024 	i = raw_seq_private((struct seq_file *)file->private_data);
1025 	i->h = h;
1026 	return 0;
1027 }
1028 EXPORT_SYMBOL_GPL(raw_seq_open);
1029 
1030 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1031 {
1032 	return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1033 }
1034 
1035 static const struct file_operations raw_seq_fops = {
1036 	.owner	 = THIS_MODULE,
1037 	.open	 = raw_v4_seq_open,
1038 	.read	 = seq_read,
1039 	.llseek	 = seq_lseek,
1040 	.release = seq_release_net,
1041 };
1042 
1043 static __net_init int raw_init_net(struct net *net)
1044 {
1045 	if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops))
1046 		return -ENOMEM;
1047 
1048 	return 0;
1049 }
1050 
1051 static __net_exit void raw_exit_net(struct net *net)
1052 {
1053 	proc_net_remove(net, "raw");
1054 }
1055 
1056 static __net_initdata struct pernet_operations raw_net_ops = {
1057 	.init = raw_init_net,
1058 	.exit = raw_exit_net,
1059 };
1060 
1061 int __init raw_proc_init(void)
1062 {
1063 	return register_pernet_subsys(&raw_net_ops);
1064 }
1065 
1066 void __init raw_proc_exit(void)
1067 {
1068 	unregister_pernet_subsys(&raw_net_ops);
1069 }
1070 #endif /* CONFIG_PROC_FS */
1071