xref: /openbmc/linux/net/ipv4/raw.c (revision 82003e04)
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/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 static struct raw_hashinfo raw_v4_hashinfo = {
93 	.lock = __RW_LOCK_UNLOCKED(raw_v4_hashinfo.lock),
94 };
95 
96 int raw_hash_sk(struct sock *sk)
97 {
98 	struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
99 	struct hlist_head *head;
100 
101 	head = &h->ht[inet_sk(sk)->inet_num & (RAW_HTABLE_SIZE - 1)];
102 
103 	write_lock_bh(&h->lock);
104 	sk_add_node(sk, head);
105 	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
106 	write_unlock_bh(&h->lock);
107 
108 	return 0;
109 }
110 EXPORT_SYMBOL_GPL(raw_hash_sk);
111 
112 void raw_unhash_sk(struct sock *sk)
113 {
114 	struct raw_hashinfo *h = sk->sk_prot->h.raw_hash;
115 
116 	write_lock_bh(&h->lock);
117 	if (sk_del_node_init(sk))
118 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
119 	write_unlock_bh(&h->lock);
120 }
121 EXPORT_SYMBOL_GPL(raw_unhash_sk);
122 
123 static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk,
124 		unsigned short num, __be32 raddr, __be32 laddr, int dif)
125 {
126 	sk_for_each_from(sk) {
127 		struct inet_sock *inet = inet_sk(sk);
128 
129 		if (net_eq(sock_net(sk), net) && inet->inet_num == num	&&
130 		    !(inet->inet_daddr && inet->inet_daddr != raddr) 	&&
131 		    !(inet->inet_rcv_saddr && inet->inet_rcv_saddr != laddr) &&
132 		    !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
133 			goto found; /* gotcha */
134 	}
135 	sk = NULL;
136 found:
137 	return sk;
138 }
139 
140 /*
141  *	0 - deliver
142  *	1 - block
143  */
144 static int icmp_filter(const struct sock *sk, const struct sk_buff *skb)
145 {
146 	struct icmphdr _hdr;
147 	const struct icmphdr *hdr;
148 
149 	hdr = skb_header_pointer(skb, skb_transport_offset(skb),
150 				 sizeof(_hdr), &_hdr);
151 	if (!hdr)
152 		return 1;
153 
154 	if (hdr->type < 32) {
155 		__u32 data = raw_sk(sk)->filter.data;
156 
157 		return ((1U << hdr->type) & data) != 0;
158 	}
159 
160 	/* Do not block unknown ICMP types */
161 	return 0;
162 }
163 
164 /* IP input processing comes here for RAW socket delivery.
165  * Caller owns SKB, so we must make clones.
166  *
167  * RFC 1122: SHOULD pass TOS value up to the transport layer.
168  * -> It does. And not only TOS, but all IP header.
169  */
170 static int raw_v4_input(struct sk_buff *skb, const struct iphdr *iph, int hash)
171 {
172 	struct sock *sk;
173 	struct hlist_head *head;
174 	int delivered = 0;
175 	struct net *net;
176 
177 	read_lock(&raw_v4_hashinfo.lock);
178 	head = &raw_v4_hashinfo.ht[hash];
179 	if (hlist_empty(head))
180 		goto out;
181 
182 	net = dev_net(skb->dev);
183 	sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol,
184 			     iph->saddr, iph->daddr,
185 			     skb->dev->ifindex);
186 
187 	while (sk) {
188 		delivered = 1;
189 		if ((iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) &&
190 		    ip_mc_sf_allow(sk, iph->daddr, iph->saddr,
191 				   skb->dev->ifindex)) {
192 			struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
193 
194 			/* Not releasing hash table! */
195 			if (clone)
196 				raw_rcv(sk, clone);
197 		}
198 		sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol,
199 				     iph->saddr, iph->daddr,
200 				     skb->dev->ifindex);
201 	}
202 out:
203 	read_unlock(&raw_v4_hashinfo.lock);
204 	return delivered;
205 }
206 
207 int raw_local_deliver(struct sk_buff *skb, int protocol)
208 {
209 	int hash;
210 	struct sock *raw_sk;
211 
212 	hash = protocol & (RAW_HTABLE_SIZE - 1);
213 	raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
214 
215 	/* If there maybe a raw socket we must check - if not we
216 	 * don't care less
217 	 */
218 	if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash))
219 		raw_sk = NULL;
220 
221 	return raw_sk != NULL;
222 
223 }
224 
225 static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info)
226 {
227 	struct inet_sock *inet = inet_sk(sk);
228 	const int type = icmp_hdr(skb)->type;
229 	const int code = icmp_hdr(skb)->code;
230 	int err = 0;
231 	int harderr = 0;
232 
233 	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
234 		ipv4_sk_update_pmtu(skb, sk, info);
235 	else if (type == ICMP_REDIRECT) {
236 		ipv4_sk_redirect(skb, sk);
237 		return;
238 	}
239 
240 	/* Report error on raw socket, if:
241 	   1. User requested ip_recverr.
242 	   2. Socket is connected (otherwise the error indication
243 	      is useless without ip_recverr and error is hard.
244 	 */
245 	if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
246 		return;
247 
248 	switch (type) {
249 	default:
250 	case ICMP_TIME_EXCEEDED:
251 		err = EHOSTUNREACH;
252 		break;
253 	case ICMP_SOURCE_QUENCH:
254 		return;
255 	case ICMP_PARAMETERPROB:
256 		err = EPROTO;
257 		harderr = 1;
258 		break;
259 	case ICMP_DEST_UNREACH:
260 		err = EHOSTUNREACH;
261 		if (code > NR_ICMP_UNREACH)
262 			break;
263 		err = icmp_err_convert[code].errno;
264 		harderr = icmp_err_convert[code].fatal;
265 		if (code == ICMP_FRAG_NEEDED) {
266 			harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
267 			err = EMSGSIZE;
268 		}
269 	}
270 
271 	if (inet->recverr) {
272 		const struct iphdr *iph = (const struct iphdr *)skb->data;
273 		u8 *payload = skb->data + (iph->ihl << 2);
274 
275 		if (inet->hdrincl)
276 			payload = skb->data;
277 		ip_icmp_error(sk, skb, err, 0, info, payload);
278 	}
279 
280 	if (inet->recverr || harderr) {
281 		sk->sk_err = err;
282 		sk->sk_error_report(sk);
283 	}
284 }
285 
286 void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info)
287 {
288 	int hash;
289 	struct sock *raw_sk;
290 	const struct iphdr *iph;
291 	struct net *net;
292 
293 	hash = protocol & (RAW_HTABLE_SIZE - 1);
294 
295 	read_lock(&raw_v4_hashinfo.lock);
296 	raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]);
297 	if (raw_sk) {
298 		iph = (const struct iphdr *)skb->data;
299 		net = dev_net(skb->dev);
300 
301 		while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol,
302 						iph->daddr, iph->saddr,
303 						skb->dev->ifindex)) != NULL) {
304 			raw_err(raw_sk, skb, info);
305 			raw_sk = sk_next(raw_sk);
306 			iph = (const struct iphdr *)skb->data;
307 		}
308 	}
309 	read_unlock(&raw_v4_hashinfo.lock);
310 }
311 
312 static int raw_rcv_skb(struct sock *sk, struct sk_buff *skb)
313 {
314 	/* Charge it to the socket. */
315 
316 	ipv4_pktinfo_prepare(sk, skb);
317 	if (sock_queue_rcv_skb(sk, skb) < 0) {
318 		kfree_skb(skb);
319 		return NET_RX_DROP;
320 	}
321 
322 	return NET_RX_SUCCESS;
323 }
324 
325 int raw_rcv(struct sock *sk, struct sk_buff *skb)
326 {
327 	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
328 		atomic_inc(&sk->sk_drops);
329 		kfree_skb(skb);
330 		return NET_RX_DROP;
331 	}
332 	nf_reset(skb);
333 
334 	skb_push(skb, skb->data - skb_network_header(skb));
335 
336 	raw_rcv_skb(sk, skb);
337 	return 0;
338 }
339 
340 static int raw_send_hdrinc(struct sock *sk, struct flowi4 *fl4,
341 			   struct msghdr *msg, size_t length,
342 			   struct rtable **rtp, unsigned int flags,
343 			   const struct sockcm_cookie *sockc)
344 {
345 	struct inet_sock *inet = inet_sk(sk);
346 	struct net *net = sock_net(sk);
347 	struct iphdr *iph;
348 	struct sk_buff *skb;
349 	unsigned int iphlen;
350 	int err;
351 	struct rtable *rt = *rtp;
352 	int hlen, tlen;
353 
354 	if (length > rt->dst.dev->mtu) {
355 		ip_local_error(sk, EMSGSIZE, fl4->daddr, inet->inet_dport,
356 			       rt->dst.dev->mtu);
357 		return -EMSGSIZE;
358 	}
359 	if (flags&MSG_PROBE)
360 		goto out;
361 
362 	hlen = LL_RESERVED_SPACE(rt->dst.dev);
363 	tlen = rt->dst.dev->needed_tailroom;
364 	skb = sock_alloc_send_skb(sk,
365 				  length + hlen + tlen + 15,
366 				  flags & MSG_DONTWAIT, &err);
367 	if (!skb)
368 		goto error;
369 	skb_reserve(skb, hlen);
370 
371 	skb->priority = sk->sk_priority;
372 	skb->mark = sk->sk_mark;
373 	skb_dst_set(skb, &rt->dst);
374 	*rtp = NULL;
375 
376 	skb_reset_network_header(skb);
377 	iph = ip_hdr(skb);
378 	skb_put(skb, length);
379 
380 	skb->ip_summed = CHECKSUM_NONE;
381 
382 	sock_tx_timestamp(sk, sockc->tsflags, &skb_shinfo(skb)->tx_flags);
383 
384 	skb->transport_header = skb->network_header;
385 	err = -EFAULT;
386 	if (memcpy_from_msg(iph, msg, length))
387 		goto error_free;
388 
389 	iphlen = iph->ihl * 4;
390 
391 	/*
392 	 * We don't want to modify the ip header, but we do need to
393 	 * be sure that it won't cause problems later along the network
394 	 * stack.  Specifically we want to make sure that iph->ihl is a
395 	 * sane value.  If ihl points beyond the length of the buffer passed
396 	 * in, reject the frame as invalid
397 	 */
398 	err = -EINVAL;
399 	if (iphlen > length)
400 		goto error_free;
401 
402 	if (iphlen >= sizeof(*iph)) {
403 		if (!iph->saddr)
404 			iph->saddr = fl4->saddr;
405 		iph->check   = 0;
406 		iph->tot_len = htons(length);
407 		if (!iph->id)
408 			ip_select_ident(net, skb, NULL);
409 
410 		iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
411 		skb->transport_header += iphlen;
412 		if (iph->protocol == IPPROTO_ICMP &&
413 		    length >= iphlen + sizeof(struct icmphdr))
414 			icmp_out_count(net, ((struct icmphdr *)
415 				skb_transport_header(skb))->type);
416 	}
417 
418 	err = NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_OUT,
419 		      net, sk, skb, NULL, rt->dst.dev,
420 		      dst_output);
421 	if (err > 0)
422 		err = net_xmit_errno(err);
423 	if (err)
424 		goto error;
425 out:
426 	return 0;
427 
428 error_free:
429 	kfree_skb(skb);
430 error:
431 	IP_INC_STATS(net, IPSTATS_MIB_OUTDISCARDS);
432 	if (err == -ENOBUFS && !inet->recverr)
433 		err = 0;
434 	return err;
435 }
436 
437 static int raw_probe_proto_opt(struct raw_frag_vec *rfv, struct flowi4 *fl4)
438 {
439 	int err;
440 
441 	if (fl4->flowi4_proto != IPPROTO_ICMP)
442 		return 0;
443 
444 	/* We only need the first two bytes. */
445 	rfv->hlen = 2;
446 
447 	err = memcpy_from_msg(rfv->hdr.c, rfv->msg, rfv->hlen);
448 	if (err)
449 		return err;
450 
451 	fl4->fl4_icmp_type = rfv->hdr.icmph.type;
452 	fl4->fl4_icmp_code = rfv->hdr.icmph.code;
453 
454 	return 0;
455 }
456 
457 static int raw_getfrag(void *from, char *to, int offset, int len, int odd,
458 		       struct sk_buff *skb)
459 {
460 	struct raw_frag_vec *rfv = from;
461 
462 	if (offset < rfv->hlen) {
463 		int copy = min(rfv->hlen - offset, len);
464 
465 		if (skb->ip_summed == CHECKSUM_PARTIAL)
466 			memcpy(to, rfv->hdr.c + offset, copy);
467 		else
468 			skb->csum = csum_block_add(
469 				skb->csum,
470 				csum_partial_copy_nocheck(rfv->hdr.c + offset,
471 							  to, copy, 0),
472 				odd);
473 
474 		odd = 0;
475 		offset += copy;
476 		to += copy;
477 		len -= copy;
478 
479 		if (!len)
480 			return 0;
481 	}
482 
483 	offset -= rfv->hlen;
484 
485 	return ip_generic_getfrag(rfv->msg, to, offset, len, odd, skb);
486 }
487 
488 static int raw_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
489 {
490 	struct inet_sock *inet = inet_sk(sk);
491 	struct net *net = sock_net(sk);
492 	struct ipcm_cookie ipc;
493 	struct rtable *rt = NULL;
494 	struct flowi4 fl4;
495 	int free = 0;
496 	__be32 daddr;
497 	__be32 saddr;
498 	u8  tos;
499 	int err;
500 	struct ip_options_data opt_copy;
501 	struct raw_frag_vec rfv;
502 
503 	err = -EMSGSIZE;
504 	if (len > 0xFFFF)
505 		goto out;
506 
507 	/*
508 	 *	Check the flags.
509 	 */
510 
511 	err = -EOPNOTSUPP;
512 	if (msg->msg_flags & MSG_OOB)	/* Mirror BSD error message */
513 		goto out;               /* compatibility */
514 
515 	/*
516 	 *	Get and verify the address.
517 	 */
518 
519 	if (msg->msg_namelen) {
520 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
521 		err = -EINVAL;
522 		if (msg->msg_namelen < sizeof(*usin))
523 			goto out;
524 		if (usin->sin_family != AF_INET) {
525 			pr_info_once("%s: %s forgot to set AF_INET. Fix it!\n",
526 				     __func__, current->comm);
527 			err = -EAFNOSUPPORT;
528 			if (usin->sin_family)
529 				goto out;
530 		}
531 		daddr = usin->sin_addr.s_addr;
532 		/* ANK: I did not forget to get protocol from port field.
533 		 * I just do not know, who uses this weirdness.
534 		 * IP_HDRINCL is much more convenient.
535 		 */
536 	} else {
537 		err = -EDESTADDRREQ;
538 		if (sk->sk_state != TCP_ESTABLISHED)
539 			goto out;
540 		daddr = inet->inet_daddr;
541 	}
542 
543 	ipc.sockc.tsflags = sk->sk_tsflags;
544 	ipc.addr = inet->inet_saddr;
545 	ipc.opt = NULL;
546 	ipc.tx_flags = 0;
547 	ipc.ttl = 0;
548 	ipc.tos = -1;
549 	ipc.oif = sk->sk_bound_dev_if;
550 
551 	if (msg->msg_controllen) {
552 		err = ip_cmsg_send(sk, msg, &ipc, false);
553 		if (unlikely(err)) {
554 			kfree(ipc.opt);
555 			goto out;
556 		}
557 		if (ipc.opt)
558 			free = 1;
559 	}
560 
561 	saddr = ipc.addr;
562 	ipc.addr = daddr;
563 
564 	if (!ipc.opt) {
565 		struct ip_options_rcu *inet_opt;
566 
567 		rcu_read_lock();
568 		inet_opt = rcu_dereference(inet->inet_opt);
569 		if (inet_opt) {
570 			memcpy(&opt_copy, inet_opt,
571 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
572 			ipc.opt = &opt_copy.opt;
573 		}
574 		rcu_read_unlock();
575 	}
576 
577 	if (ipc.opt) {
578 		err = -EINVAL;
579 		/* Linux does not mangle headers on raw sockets,
580 		 * so that IP options + IP_HDRINCL is non-sense.
581 		 */
582 		if (inet->hdrincl)
583 			goto done;
584 		if (ipc.opt->opt.srr) {
585 			if (!daddr)
586 				goto done;
587 			daddr = ipc.opt->opt.faddr;
588 		}
589 	}
590 	tos = get_rtconn_flags(&ipc, sk);
591 	if (msg->msg_flags & MSG_DONTROUTE)
592 		tos |= RTO_ONLINK;
593 
594 	if (ipv4_is_multicast(daddr)) {
595 		if (!ipc.oif)
596 			ipc.oif = inet->mc_index;
597 		if (!saddr)
598 			saddr = inet->mc_addr;
599 	} else if (!ipc.oif)
600 		ipc.oif = inet->uc_index;
601 
602 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
603 			   RT_SCOPE_UNIVERSE,
604 			   inet->hdrincl ? IPPROTO_RAW : sk->sk_protocol,
605 			   inet_sk_flowi_flags(sk) |
606 			    (inet->hdrincl ? FLOWI_FLAG_KNOWN_NH : 0),
607 			   daddr, saddr, 0, 0);
608 
609 	if (!inet->hdrincl) {
610 		rfv.msg = msg;
611 		rfv.hlen = 0;
612 
613 		err = raw_probe_proto_opt(&rfv, &fl4);
614 		if (err)
615 			goto done;
616 	}
617 
618 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
619 	rt = ip_route_output_flow(net, &fl4, sk);
620 	if (IS_ERR(rt)) {
621 		err = PTR_ERR(rt);
622 		rt = NULL;
623 		goto done;
624 	}
625 
626 	err = -EACCES;
627 	if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
628 		goto done;
629 
630 	if (msg->msg_flags & MSG_CONFIRM)
631 		goto do_confirm;
632 back_from_confirm:
633 
634 	if (inet->hdrincl)
635 		err = raw_send_hdrinc(sk, &fl4, msg, len,
636 				      &rt, msg->msg_flags, &ipc.sockc);
637 
638 	 else {
639 		sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
640 
641 		if (!ipc.addr)
642 			ipc.addr = fl4.daddr;
643 		lock_sock(sk);
644 		err = ip_append_data(sk, &fl4, raw_getfrag,
645 				     &rfv, len, 0,
646 				     &ipc, &rt, msg->msg_flags);
647 		if (err)
648 			ip_flush_pending_frames(sk);
649 		else if (!(msg->msg_flags & MSG_MORE)) {
650 			err = ip_push_pending_frames(sk, &fl4);
651 			if (err == -ENOBUFS && !inet->recverr)
652 				err = 0;
653 		}
654 		release_sock(sk);
655 	}
656 done:
657 	if (free)
658 		kfree(ipc.opt);
659 	ip_rt_put(rt);
660 
661 out:
662 	if (err < 0)
663 		return err;
664 	return len;
665 
666 do_confirm:
667 	dst_confirm(&rt->dst);
668 	if (!(msg->msg_flags & MSG_PROBE) || len)
669 		goto back_from_confirm;
670 	err = 0;
671 	goto done;
672 }
673 
674 static void raw_close(struct sock *sk, long timeout)
675 {
676 	/*
677 	 * Raw sockets may have direct kernel references. Kill them.
678 	 */
679 	ip_ra_control(sk, 0, NULL);
680 
681 	sk_common_release(sk);
682 }
683 
684 static void raw_destroy(struct sock *sk)
685 {
686 	lock_sock(sk);
687 	ip_flush_pending_frames(sk);
688 	release_sock(sk);
689 }
690 
691 /* This gets rid of all the nasties in af_inet. -DaveM */
692 static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
693 {
694 	struct inet_sock *inet = inet_sk(sk);
695 	struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
696 	int ret = -EINVAL;
697 	int chk_addr_ret;
698 
699 	if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
700 		goto out;
701 	chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
702 	ret = -EADDRNOTAVAIL;
703 	if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
704 	    chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
705 		goto out;
706 	inet->inet_rcv_saddr = inet->inet_saddr = addr->sin_addr.s_addr;
707 	if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
708 		inet->inet_saddr = 0;  /* Use device */
709 	sk_dst_reset(sk);
710 	ret = 0;
711 out:	return ret;
712 }
713 
714 /*
715  *	This should be easy, if there is something there
716  *	we return it, otherwise we block.
717  */
718 
719 static int raw_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
720 		       int noblock, int flags, int *addr_len)
721 {
722 	struct inet_sock *inet = inet_sk(sk);
723 	size_t copied = 0;
724 	int err = -EOPNOTSUPP;
725 	DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
726 	struct sk_buff *skb;
727 
728 	if (flags & MSG_OOB)
729 		goto out;
730 
731 	if (flags & MSG_ERRQUEUE) {
732 		err = ip_recv_error(sk, msg, len, addr_len);
733 		goto out;
734 	}
735 
736 	skb = skb_recv_datagram(sk, flags, noblock, &err);
737 	if (!skb)
738 		goto out;
739 
740 	copied = skb->len;
741 	if (len < copied) {
742 		msg->msg_flags |= MSG_TRUNC;
743 		copied = len;
744 	}
745 
746 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
747 	if (err)
748 		goto done;
749 
750 	sock_recv_ts_and_drops(msg, sk, skb);
751 
752 	/* Copy the address. */
753 	if (sin) {
754 		sin->sin_family = AF_INET;
755 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
756 		sin->sin_port = 0;
757 		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
758 		*addr_len = sizeof(*sin);
759 	}
760 	if (inet->cmsg_flags)
761 		ip_cmsg_recv(msg, skb);
762 	if (flags & MSG_TRUNC)
763 		copied = skb->len;
764 done:
765 	skb_free_datagram(sk, skb);
766 out:
767 	if (err)
768 		return err;
769 	return copied;
770 }
771 
772 static int raw_init(struct sock *sk)
773 {
774 	struct raw_sock *rp = raw_sk(sk);
775 
776 	if (inet_sk(sk)->inet_num == IPPROTO_ICMP)
777 		memset(&rp->filter, 0, sizeof(rp->filter));
778 	return 0;
779 }
780 
781 static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
782 {
783 	if (optlen > sizeof(struct icmp_filter))
784 		optlen = sizeof(struct icmp_filter);
785 	if (copy_from_user(&raw_sk(sk)->filter, optval, optlen))
786 		return -EFAULT;
787 	return 0;
788 }
789 
790 static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
791 {
792 	int len, ret = -EFAULT;
793 
794 	if (get_user(len, optlen))
795 		goto out;
796 	ret = -EINVAL;
797 	if (len < 0)
798 		goto out;
799 	if (len > sizeof(struct icmp_filter))
800 		len = sizeof(struct icmp_filter);
801 	ret = -EFAULT;
802 	if (put_user(len, optlen) ||
803 	    copy_to_user(optval, &raw_sk(sk)->filter, len))
804 		goto out;
805 	ret = 0;
806 out:	return ret;
807 }
808 
809 static int do_raw_setsockopt(struct sock *sk, int level, int optname,
810 			  char __user *optval, unsigned int optlen)
811 {
812 	if (optname == ICMP_FILTER) {
813 		if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
814 			return -EOPNOTSUPP;
815 		else
816 			return raw_seticmpfilter(sk, optval, optlen);
817 	}
818 	return -ENOPROTOOPT;
819 }
820 
821 static int raw_setsockopt(struct sock *sk, int level, int optname,
822 			  char __user *optval, unsigned int optlen)
823 {
824 	if (level != SOL_RAW)
825 		return ip_setsockopt(sk, level, optname, optval, optlen);
826 	return do_raw_setsockopt(sk, level, optname, optval, optlen);
827 }
828 
829 #ifdef CONFIG_COMPAT
830 static int compat_raw_setsockopt(struct sock *sk, int level, int optname,
831 				 char __user *optval, unsigned int optlen)
832 {
833 	if (level != SOL_RAW)
834 		return compat_ip_setsockopt(sk, level, optname, optval, optlen);
835 	return do_raw_setsockopt(sk, level, optname, optval, optlen);
836 }
837 #endif
838 
839 static int do_raw_getsockopt(struct sock *sk, int level, int optname,
840 			  char __user *optval, int __user *optlen)
841 {
842 	if (optname == ICMP_FILTER) {
843 		if (inet_sk(sk)->inet_num != IPPROTO_ICMP)
844 			return -EOPNOTSUPP;
845 		else
846 			return raw_geticmpfilter(sk, optval, optlen);
847 	}
848 	return -ENOPROTOOPT;
849 }
850 
851 static int raw_getsockopt(struct sock *sk, int level, int optname,
852 			  char __user *optval, int __user *optlen)
853 {
854 	if (level != SOL_RAW)
855 		return ip_getsockopt(sk, level, optname, optval, optlen);
856 	return do_raw_getsockopt(sk, level, optname, optval, optlen);
857 }
858 
859 #ifdef CONFIG_COMPAT
860 static int compat_raw_getsockopt(struct sock *sk, int level, int optname,
861 				 char __user *optval, int __user *optlen)
862 {
863 	if (level != SOL_RAW)
864 		return compat_ip_getsockopt(sk, level, optname, optval, optlen);
865 	return do_raw_getsockopt(sk, level, optname, optval, optlen);
866 }
867 #endif
868 
869 static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
870 {
871 	switch (cmd) {
872 	case SIOCOUTQ: {
873 		int amount = sk_wmem_alloc_get(sk);
874 
875 		return put_user(amount, (int __user *)arg);
876 	}
877 	case SIOCINQ: {
878 		struct sk_buff *skb;
879 		int amount = 0;
880 
881 		spin_lock_bh(&sk->sk_receive_queue.lock);
882 		skb = skb_peek(&sk->sk_receive_queue);
883 		if (skb)
884 			amount = skb->len;
885 		spin_unlock_bh(&sk->sk_receive_queue.lock);
886 		return put_user(amount, (int __user *)arg);
887 	}
888 
889 	default:
890 #ifdef CONFIG_IP_MROUTE
891 		return ipmr_ioctl(sk, cmd, (void __user *)arg);
892 #else
893 		return -ENOIOCTLCMD;
894 #endif
895 	}
896 }
897 
898 #ifdef CONFIG_COMPAT
899 static int compat_raw_ioctl(struct sock *sk, unsigned int cmd, unsigned long arg)
900 {
901 	switch (cmd) {
902 	case SIOCOUTQ:
903 	case SIOCINQ:
904 		return -ENOIOCTLCMD;
905 	default:
906 #ifdef CONFIG_IP_MROUTE
907 		return ipmr_compat_ioctl(sk, cmd, compat_ptr(arg));
908 #else
909 		return -ENOIOCTLCMD;
910 #endif
911 	}
912 }
913 #endif
914 
915 struct proto raw_prot = {
916 	.name		   = "RAW",
917 	.owner		   = THIS_MODULE,
918 	.close		   = raw_close,
919 	.destroy	   = raw_destroy,
920 	.connect	   = ip4_datagram_connect,
921 	.disconnect	   = __udp_disconnect,
922 	.ioctl		   = raw_ioctl,
923 	.init		   = raw_init,
924 	.setsockopt	   = raw_setsockopt,
925 	.getsockopt	   = raw_getsockopt,
926 	.sendmsg	   = raw_sendmsg,
927 	.recvmsg	   = raw_recvmsg,
928 	.bind		   = raw_bind,
929 	.backlog_rcv	   = raw_rcv_skb,
930 	.release_cb	   = ip4_datagram_release_cb,
931 	.hash		   = raw_hash_sk,
932 	.unhash		   = raw_unhash_sk,
933 	.obj_size	   = sizeof(struct raw_sock),
934 	.h.raw_hash	   = &raw_v4_hashinfo,
935 #ifdef CONFIG_COMPAT
936 	.compat_setsockopt = compat_raw_setsockopt,
937 	.compat_getsockopt = compat_raw_getsockopt,
938 	.compat_ioctl	   = compat_raw_ioctl,
939 #endif
940 };
941 
942 #ifdef CONFIG_PROC_FS
943 static struct sock *raw_get_first(struct seq_file *seq)
944 {
945 	struct sock *sk;
946 	struct raw_iter_state *state = raw_seq_private(seq);
947 
948 	for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE;
949 			++state->bucket) {
950 		sk_for_each(sk, &state->h->ht[state->bucket])
951 			if (sock_net(sk) == seq_file_net(seq))
952 				goto found;
953 	}
954 	sk = NULL;
955 found:
956 	return sk;
957 }
958 
959 static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
960 {
961 	struct raw_iter_state *state = raw_seq_private(seq);
962 
963 	do {
964 		sk = sk_next(sk);
965 try_again:
966 		;
967 	} while (sk && sock_net(sk) != seq_file_net(seq));
968 
969 	if (!sk && ++state->bucket < RAW_HTABLE_SIZE) {
970 		sk = sk_head(&state->h->ht[state->bucket]);
971 		goto try_again;
972 	}
973 	return sk;
974 }
975 
976 static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
977 {
978 	struct sock *sk = raw_get_first(seq);
979 
980 	if (sk)
981 		while (pos && (sk = raw_get_next(seq, sk)) != NULL)
982 			--pos;
983 	return pos ? NULL : sk;
984 }
985 
986 void *raw_seq_start(struct seq_file *seq, loff_t *pos)
987 {
988 	struct raw_iter_state *state = raw_seq_private(seq);
989 
990 	read_lock(&state->h->lock);
991 	return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
992 }
993 EXPORT_SYMBOL_GPL(raw_seq_start);
994 
995 void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
996 {
997 	struct sock *sk;
998 
999 	if (v == SEQ_START_TOKEN)
1000 		sk = raw_get_first(seq);
1001 	else
1002 		sk = raw_get_next(seq, v);
1003 	++*pos;
1004 	return sk;
1005 }
1006 EXPORT_SYMBOL_GPL(raw_seq_next);
1007 
1008 void raw_seq_stop(struct seq_file *seq, void *v)
1009 {
1010 	struct raw_iter_state *state = raw_seq_private(seq);
1011 
1012 	read_unlock(&state->h->lock);
1013 }
1014 EXPORT_SYMBOL_GPL(raw_seq_stop);
1015 
1016 static void raw_sock_seq_show(struct seq_file *seq, struct sock *sp, int i)
1017 {
1018 	struct inet_sock *inet = inet_sk(sp);
1019 	__be32 dest = inet->inet_daddr,
1020 	       src = inet->inet_rcv_saddr;
1021 	__u16 destp = 0,
1022 	      srcp  = inet->inet_num;
1023 
1024 	seq_printf(seq, "%4d: %08X:%04X %08X:%04X"
1025 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d\n",
1026 		i, src, srcp, dest, destp, sp->sk_state,
1027 		sk_wmem_alloc_get(sp),
1028 		sk_rmem_alloc_get(sp),
1029 		0, 0L, 0,
1030 		from_kuid_munged(seq_user_ns(seq), sock_i_uid(sp)),
1031 		0, sock_i_ino(sp),
1032 		atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops));
1033 }
1034 
1035 static int raw_seq_show(struct seq_file *seq, void *v)
1036 {
1037 	if (v == SEQ_START_TOKEN)
1038 		seq_printf(seq, "  sl  local_address rem_address   st tx_queue "
1039 				"rx_queue tr tm->when retrnsmt   uid  timeout "
1040 				"inode ref pointer drops\n");
1041 	else
1042 		raw_sock_seq_show(seq, v, raw_seq_private(seq)->bucket);
1043 	return 0;
1044 }
1045 
1046 static const struct seq_operations raw_seq_ops = {
1047 	.start = raw_seq_start,
1048 	.next  = raw_seq_next,
1049 	.stop  = raw_seq_stop,
1050 	.show  = raw_seq_show,
1051 };
1052 
1053 int raw_seq_open(struct inode *ino, struct file *file,
1054 		 struct raw_hashinfo *h, const struct seq_operations *ops)
1055 {
1056 	int err;
1057 	struct raw_iter_state *i;
1058 
1059 	err = seq_open_net(ino, file, ops, sizeof(struct raw_iter_state));
1060 	if (err < 0)
1061 		return err;
1062 
1063 	i = raw_seq_private((struct seq_file *)file->private_data);
1064 	i->h = h;
1065 	return 0;
1066 }
1067 EXPORT_SYMBOL_GPL(raw_seq_open);
1068 
1069 static int raw_v4_seq_open(struct inode *inode, struct file *file)
1070 {
1071 	return raw_seq_open(inode, file, &raw_v4_hashinfo, &raw_seq_ops);
1072 }
1073 
1074 static const struct file_operations raw_seq_fops = {
1075 	.owner	 = THIS_MODULE,
1076 	.open	 = raw_v4_seq_open,
1077 	.read	 = seq_read,
1078 	.llseek	 = seq_lseek,
1079 	.release = seq_release_net,
1080 };
1081 
1082 static __net_init int raw_init_net(struct net *net)
1083 {
1084 	if (!proc_create("raw", S_IRUGO, net->proc_net, &raw_seq_fops))
1085 		return -ENOMEM;
1086 
1087 	return 0;
1088 }
1089 
1090 static __net_exit void raw_exit_net(struct net *net)
1091 {
1092 	remove_proc_entry("raw", net->proc_net);
1093 }
1094 
1095 static __net_initdata struct pernet_operations raw_net_ops = {
1096 	.init = raw_init_net,
1097 	.exit = raw_exit_net,
1098 };
1099 
1100 int __init raw_proc_init(void)
1101 {
1102 	return register_pernet_subsys(&raw_net_ops);
1103 }
1104 
1105 void __init raw_proc_exit(void)
1106 {
1107 	unregister_pernet_subsys(&raw_net_ops);
1108 }
1109 #endif /* CONFIG_PROC_FS */
1110