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