ping.c revision bb0cd2fb539c4e454ed32e32194acc03b75753f3
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 *		"Ping" sockets
7 *
8 *		This program is free software; you can redistribute it and/or
9 *		modify it under the terms of the GNU General Public License
10 *		as published by the Free Software Foundation; either version
11 *		2 of the License, or (at your option) any later version.
12 *
13 * Based on ipv4/udp.c code.
14 *
15 * Authors:	Vasiliy Kulikov / Openwall (for Linux 2.6),
16 *		Pavel Kankovsky (for Linux 2.4.32)
17 *
18 * Pavel gave all rights to bugs to Vasiliy,
19 * none of the bugs are Pavel's now.
20 *
21 */
22
23#include <asm/system.h>
24#include <linux/uaccess.h>
25#include <linux/types.h>
26#include <linux/fcntl.h>
27#include <linux/socket.h>
28#include <linux/sockios.h>
29#include <linux/in.h>
30#include <linux/errno.h>
31#include <linux/timer.h>
32#include <linux/mm.h>
33#include <linux/inet.h>
34#include <linux/netdevice.h>
35#include <net/snmp.h>
36#include <net/ip.h>
37#include <net/ipv6.h>
38#include <net/icmp.h>
39#include <net/protocol.h>
40#include <linux/skbuff.h>
41#include <linux/proc_fs.h>
42#include <net/sock.h>
43#include <net/ping.h>
44#include <net/icmp.h>
45#include <net/udp.h>
46#include <net/route.h>
47#include <net/inet_common.h>
48#include <net/checksum.h>
49
50
51static struct ping_table ping_table;
52
53static u16 ping_port_rover;
54
55static inline int ping_hashfn(struct net *net, unsigned num, unsigned mask)
56{
57	int res = (num + net_hash_mix(net)) & mask;
58	pr_debug("hash(%d) = %d\n", num, res);
59	return res;
60}
61
62static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
63					     struct net *net, unsigned num)
64{
65	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
66}
67
68static int ping_v4_get_port(struct sock *sk, unsigned short ident)
69{
70	struct hlist_nulls_node *node;
71	struct hlist_nulls_head *hlist;
72	struct inet_sock *isk, *isk2;
73	struct sock *sk2 = NULL;
74
75	isk = inet_sk(sk);
76	write_lock_bh(&ping_table.lock);
77	if (ident == 0) {
78		u32 i;
79		u16 result = ping_port_rover + 1;
80
81		for (i = 0; i < (1L << 16); i++, result++) {
82			if (!result)
83				result++; /* avoid zero */
84			hlist = ping_hashslot(&ping_table, sock_net(sk),
85					    result);
86			ping_portaddr_for_each_entry(sk2, node, hlist) {
87				isk2 = inet_sk(sk2);
88
89				if (isk2->inet_num == result)
90					goto next_port;
91			}
92
93			/* found */
94			ping_port_rover = ident = result;
95			break;
96next_port:
97			;
98		}
99		if (i >= (1L << 16))
100			goto fail;
101	} else {
102		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
103		ping_portaddr_for_each_entry(sk2, node, hlist) {
104			isk2 = inet_sk(sk2);
105
106			if ((isk2->inet_num == ident) &&
107			    (sk2 != sk) &&
108			    (!sk2->sk_reuse || !sk->sk_reuse))
109				goto fail;
110		}
111	}
112
113	pr_debug("found port/ident = %d\n", ident);
114	isk->inet_num = ident;
115	if (sk_unhashed(sk)) {
116		pr_debug("was not hashed\n");
117		sock_hold(sk);
118		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
119		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
120	}
121	write_unlock_bh(&ping_table.lock);
122	return 0;
123
124fail:
125	write_unlock_bh(&ping_table.lock);
126	return 1;
127}
128
129static void ping_v4_hash(struct sock *sk)
130{
131	pr_debug("ping_v4_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
132	BUG(); /* "Please do not press this button again." */
133}
134
135static void ping_v4_unhash(struct sock *sk)
136{
137	struct inet_sock *isk = inet_sk(sk);
138	pr_debug("ping_v4_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
139	if (sk_hashed(sk)) {
140		struct hlist_nulls_head *hslot;
141
142		hslot = ping_hashslot(&ping_table, sock_net(sk), isk->inet_num);
143		write_lock_bh(&ping_table.lock);
144		hlist_nulls_del(&sk->sk_nulls_node);
145		sock_put(sk);
146		isk->inet_num = isk->inet_sport = 0;
147		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
148		write_unlock_bh(&ping_table.lock);
149	}
150}
151
152static struct sock *ping_v4_lookup(struct net *net, u32 saddr, u32 daddr,
153				   u16 ident, int dif)
154{
155	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
156	struct sock *sk = NULL;
157	struct inet_sock *isk;
158	struct hlist_nulls_node *hnode;
159
160	pr_debug("try to find: num = %d, daddr = %ld, dif = %d\n",
161			 (int)ident, (unsigned long)daddr, dif);
162	read_lock_bh(&ping_table.lock);
163
164	ping_portaddr_for_each_entry(sk, hnode, hslot) {
165		isk = inet_sk(sk);
166
167		pr_debug("found: %p: num = %d, daddr = %ld, dif = %d\n", sk,
168			 (int)isk->inet_num, (unsigned long)isk->inet_rcv_saddr,
169			 sk->sk_bound_dev_if);
170
171		pr_debug("iterate\n");
172		if (isk->inet_num != ident)
173			continue;
174		if (isk->inet_rcv_saddr && isk->inet_rcv_saddr != daddr)
175			continue;
176		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
177			continue;
178
179		sock_hold(sk);
180		goto exit;
181	}
182
183	sk = NULL;
184exit:
185	read_unlock_bh(&ping_table.lock);
186
187	return sk;
188}
189
190static void inet_get_ping_group_range_net(struct net *net, gid_t *low, gid_t *high)
191{
192	gid_t *data = net->ipv4.sysctl_ping_group_range;
193	unsigned seq;
194	do {
195		seq = read_seqbegin(&sysctl_local_ports.lock);
196
197		*low = data[0];
198		*high = data[1];
199	} while (read_seqretry(&sysctl_local_ports.lock, seq));
200}
201
202
203static int ping_init_sock(struct sock *sk)
204{
205	struct net *net = sock_net(sk);
206	gid_t group = current_egid();
207	gid_t range[2];
208	struct group_info *group_info = get_current_groups();
209	int i, j, count = group_info->ngroups;
210
211	inet_get_ping_group_range_net(net, range, range+1);
212	if (range[0] <= group && group <= range[1])
213		return 0;
214
215	for (i = 0; i < group_info->nblocks; i++) {
216		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
217
218		for (j = 0; j < cp_count; j++) {
219			group = group_info->blocks[i][j];
220			if (range[0] <= group && group <= range[1])
221				return 0;
222		}
223
224		count -= cp_count;
225	}
226
227	return -EACCES;
228}
229
230static void ping_close(struct sock *sk, long timeout)
231{
232	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
233		inet_sk(sk), inet_sk(sk)->inet_num);
234	pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
235
236	sk_common_release(sk);
237}
238
239/*
240 * We need our own bind because there are no privileged id's == local ports.
241 * Moreover, we don't allow binding to multi- and broadcast addresses.
242 */
243
244static int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
245{
246	struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
247	struct inet_sock *isk = inet_sk(sk);
248	unsigned short snum;
249	int chk_addr_ret;
250	int err;
251
252	if (addr_len < sizeof(struct sockaddr_in))
253		return -EINVAL;
254
255	pr_debug("ping_v4_bind(sk=%p,sa_addr=%08x,sa_port=%d)\n",
256		sk, addr->sin_addr.s_addr, ntohs(addr->sin_port));
257
258	chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
259	if (addr->sin_addr.s_addr == INADDR_ANY)
260		chk_addr_ret = RTN_LOCAL;
261
262	if ((sysctl_ip_nonlocal_bind == 0 &&
263	    isk->freebind == 0 && isk->transparent == 0 &&
264	     chk_addr_ret != RTN_LOCAL) ||
265	    chk_addr_ret == RTN_MULTICAST ||
266	    chk_addr_ret == RTN_BROADCAST)
267		return -EADDRNOTAVAIL;
268
269	lock_sock(sk);
270
271	err = -EINVAL;
272	if (isk->inet_num != 0)
273		goto out;
274
275	err = -EADDRINUSE;
276	isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
277	snum = ntohs(addr->sin_port);
278	if (ping_v4_get_port(sk, snum) != 0) {
279		isk->inet_saddr = isk->inet_rcv_saddr = 0;
280		goto out;
281	}
282
283	pr_debug("after bind(): num = %d, daddr = %ld, dif = %d\n",
284		(int)isk->inet_num,
285		(unsigned long) isk->inet_rcv_saddr,
286		(int)sk->sk_bound_dev_if);
287
288	err = 0;
289	if (isk->inet_rcv_saddr)
290		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
291	if (snum)
292		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
293	isk->inet_sport = htons(isk->inet_num);
294	isk->inet_daddr = 0;
295	isk->inet_dport = 0;
296	sk_dst_reset(sk);
297out:
298	release_sock(sk);
299	pr_debug("ping_v4_bind -> %d\n", err);
300	return err;
301}
302
303/*
304 * Is this a supported type of ICMP message?
305 */
306
307static inline int ping_supported(int type, int code)
308{
309	if (type == ICMP_ECHO && code == 0)
310		return 1;
311	return 0;
312}
313
314/*
315 * This routine is called by the ICMP module when it gets some
316 * sort of error condition.
317 */
318
319static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
320
321void ping_err(struct sk_buff *skb, u32 info)
322{
323	struct iphdr *iph = (struct iphdr *)skb->data;
324	struct icmphdr *icmph = (struct icmphdr *)(skb->data+(iph->ihl<<2));
325	struct inet_sock *inet_sock;
326	int type = icmph->type;
327	int code = icmph->code;
328	struct net *net = dev_net(skb->dev);
329	struct sock *sk;
330	int harderr;
331	int err;
332
333	/* We assume the packet has already been checked by icmp_unreach */
334
335	if (!ping_supported(icmph->type, icmph->code))
336		return;
337
338	pr_debug("ping_err(type=%04x,code=%04x,id=%04x,seq=%04x)\n", type,
339		code, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
340
341	sk = ping_v4_lookup(net, iph->daddr, iph->saddr,
342			    ntohs(icmph->un.echo.id), skb->dev->ifindex);
343	if (sk == NULL) {
344		ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
345		pr_debug("no socket, dropping\n");
346		return;	/* No socket for error */
347	}
348	pr_debug("err on socket %p\n", sk);
349
350	err = 0;
351	harderr = 0;
352	inet_sock = inet_sk(sk);
353
354	switch (type) {
355	default:
356	case ICMP_TIME_EXCEEDED:
357		err = EHOSTUNREACH;
358		break;
359	case ICMP_SOURCE_QUENCH:
360		/* This is not a real error but ping wants to see it.
361		 * Report it with some fake errno. */
362		err = EREMOTEIO;
363		break;
364	case ICMP_PARAMETERPROB:
365		err = EPROTO;
366		harderr = 1;
367		break;
368	case ICMP_DEST_UNREACH:
369		if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
370			if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
371				err = EMSGSIZE;
372				harderr = 1;
373				break;
374			}
375			goto out;
376		}
377		err = EHOSTUNREACH;
378		if (code <= NR_ICMP_UNREACH) {
379			harderr = icmp_err_convert[code].fatal;
380			err = icmp_err_convert[code].errno;
381		}
382		break;
383	case ICMP_REDIRECT:
384		/* See ICMP_SOURCE_QUENCH */
385		err = EREMOTEIO;
386		break;
387	}
388
389	/*
390	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
391	 *	4.1.3.3.
392	 */
393	if (!inet_sock->recverr) {
394		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
395			goto out;
396	} else {
397		ip_icmp_error(sk, skb, err, 0 /* no remote port */,
398			 info, (u8 *)icmph);
399	}
400	sk->sk_err = err;
401	sk->sk_error_report(sk);
402out:
403	sock_put(sk);
404}
405
406/*
407 *	Copy and checksum an ICMP Echo packet from user space into a buffer.
408 */
409
410struct pingfakehdr {
411	struct icmphdr icmph;
412	struct iovec *iov;
413	u32 wcheck;
414};
415
416static int ping_getfrag(void *from, char * to,
417			int offset, int fraglen, int odd, struct sk_buff *skb)
418{
419	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
420
421	if (offset == 0) {
422		if (fraglen < sizeof(struct icmphdr))
423			BUG();
424		if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
425			    pfh->iov, 0, fraglen - sizeof(struct icmphdr),
426			    &pfh->wcheck))
427			return -EFAULT;
428
429		return 0;
430	}
431	if (offset < sizeof(struct icmphdr))
432		BUG();
433	if (csum_partial_copy_fromiovecend
434			(to, pfh->iov, offset - sizeof(struct icmphdr),
435			 fraglen, &pfh->wcheck))
436		return -EFAULT;
437	return 0;
438}
439
440static int ping_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh, struct flowi4 *fl4)
441{
442	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
443
444	pfh->wcheck = csum_partial((char *)&pfh->icmph,
445		sizeof(struct icmphdr), pfh->wcheck);
446	pfh->icmph.checksum = csum_fold(pfh->wcheck);
447	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
448	skb->ip_summed = CHECKSUM_NONE;
449	return ip_push_pending_frames(sk, fl4);
450}
451
452static int ping_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
453			size_t len)
454{
455	struct net *net = sock_net(sk);
456	struct flowi4 fl4;
457	struct inet_sock *inet = inet_sk(sk);
458	struct ipcm_cookie ipc;
459	struct icmphdr user_icmph;
460	struct pingfakehdr pfh;
461	struct rtable *rt = NULL;
462	struct ip_options_data opt_copy;
463	int free = 0;
464	u32 saddr, daddr, faddr;
465	u8  tos;
466	int err;
467
468	pr_debug("ping_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
469
470
471	if (len > 0xFFFF)
472		return -EMSGSIZE;
473
474	/*
475	 *	Check the flags.
476	 */
477
478	/* Mirror BSD error message compatibility */
479	if (msg->msg_flags & MSG_OOB)
480		return -EOPNOTSUPP;
481
482	/*
483	 *	Fetch the ICMP header provided by the userland.
484	 *	iovec is modified!
485	 */
486
487	if (memcpy_fromiovec((u8 *)&user_icmph, msg->msg_iov,
488			     sizeof(struct icmphdr)))
489		return -EFAULT;
490	if (!ping_supported(user_icmph.type, user_icmph.code))
491		return -EINVAL;
492
493	/*
494	 *	Get and verify the address.
495	 */
496
497	if (msg->msg_name) {
498		struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
499		if (msg->msg_namelen < sizeof(*usin))
500			return -EINVAL;
501		if (usin->sin_family != AF_INET)
502			return -EINVAL;
503		daddr = usin->sin_addr.s_addr;
504		/* no remote port */
505	} else {
506		if (sk->sk_state != TCP_ESTABLISHED)
507			return -EDESTADDRREQ;
508		daddr = inet->inet_daddr;
509		/* no remote port */
510	}
511
512	ipc.addr = inet->inet_saddr;
513	ipc.opt = NULL;
514	ipc.oif = sk->sk_bound_dev_if;
515	ipc.tx_flags = 0;
516	err = sock_tx_timestamp(sk, &ipc.tx_flags);
517	if (err)
518		return err;
519
520	if (msg->msg_controllen) {
521		err = ip_cmsg_send(sock_net(sk), msg, &ipc);
522		if (err)
523			return err;
524		if (ipc.opt)
525			free = 1;
526	}
527	if (!ipc.opt) {
528		struct ip_options_rcu *inet_opt;
529
530		rcu_read_lock();
531		inet_opt = rcu_dereference(inet->inet_opt);
532		if (inet_opt) {
533			memcpy(&opt_copy, inet_opt,
534			       sizeof(*inet_opt) + inet_opt->opt.optlen);
535			ipc.opt = &opt_copy.opt;
536		}
537		rcu_read_unlock();
538	}
539
540	saddr = ipc.addr;
541	ipc.addr = faddr = daddr;
542
543	if (ipc.opt && ipc.opt->opt.srr) {
544		if (!daddr)
545			return -EINVAL;
546		faddr = ipc.opt->opt.faddr;
547	}
548	tos = RT_TOS(inet->tos);
549	if (sock_flag(sk, SOCK_LOCALROUTE) ||
550	    (msg->msg_flags & MSG_DONTROUTE) ||
551	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
552		tos |= RTO_ONLINK;
553	}
554
555	if (ipv4_is_multicast(daddr)) {
556		if (!ipc.oif)
557			ipc.oif = inet->mc_index;
558		if (!saddr)
559			saddr = inet->mc_addr;
560	}
561
562	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
563			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
564			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
565
566	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
567	rt = ip_route_output_flow(net, &fl4, sk);
568	if (IS_ERR(rt)) {
569		err = PTR_ERR(rt);
570		rt = NULL;
571		if (err == -ENETUNREACH)
572			IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
573		goto out;
574	}
575
576	err = -EACCES;
577	if ((rt->rt_flags & RTCF_BROADCAST) &&
578	    !sock_flag(sk, SOCK_BROADCAST))
579		goto out;
580
581	if (msg->msg_flags & MSG_CONFIRM)
582		goto do_confirm;
583back_from_confirm:
584
585	if (!ipc.addr)
586		ipc.addr = fl4.daddr;
587
588	lock_sock(sk);
589
590	pfh.icmph.type = user_icmph.type; /* already checked */
591	pfh.icmph.code = user_icmph.code; /* ditto */
592	pfh.icmph.checksum = 0;
593	pfh.icmph.un.echo.id = inet->inet_sport;
594	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
595	pfh.iov = msg->msg_iov;
596	pfh.wcheck = 0;
597
598	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
599			0, &ipc, &rt, msg->msg_flags);
600	if (err)
601		ip_flush_pending_frames(sk);
602	else
603		err = ping_push_pending_frames(sk, &pfh, &fl4);
604	release_sock(sk);
605
606out:
607	ip_rt_put(rt);
608	if (free)
609		kfree(ipc.opt);
610	if (!err) {
611		icmp_out_count(sock_net(sk), user_icmph.type);
612		return len;
613	}
614	return err;
615
616do_confirm:
617	dst_confirm(&rt->dst);
618	if (!(msg->msg_flags & MSG_PROBE) || len)
619		goto back_from_confirm;
620	err = 0;
621	goto out;
622}
623
624static int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
625			size_t len, int noblock, int flags, int *addr_len)
626{
627	struct inet_sock *isk = inet_sk(sk);
628	struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
629	struct sk_buff *skb;
630	int copied, err;
631
632	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
633
634	if (flags & MSG_OOB)
635		goto out;
636
637	if (addr_len)
638		*addr_len = sizeof(*sin);
639
640	if (flags & MSG_ERRQUEUE)
641		return ip_recv_error(sk, msg, len);
642
643	skb = skb_recv_datagram(sk, flags, noblock, &err);
644	if (!skb)
645		goto out;
646
647	copied = skb->len;
648	if (copied > len) {
649		msg->msg_flags |= MSG_TRUNC;
650		copied = len;
651	}
652
653	/* Don't bother checking the checksum */
654	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
655	if (err)
656		goto done;
657
658	sock_recv_timestamp(msg, sk, skb);
659
660	/* Copy the address. */
661	if (sin) {
662		sin->sin_family = AF_INET;
663		sin->sin_port = 0 /* skb->h.uh->source */;
664		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
665		memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
666	}
667	if (isk->cmsg_flags)
668		ip_cmsg_recv(msg, skb);
669	err = copied;
670
671done:
672	skb_free_datagram(sk, skb);
673out:
674	pr_debug("ping_recvmsg -> %d\n", err);
675	return err;
676}
677
678static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
679{
680	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
681		inet_sk(sk), inet_sk(sk)->inet_num, skb);
682	if (sock_queue_rcv_skb(sk, skb) < 0) {
683		ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_INERRORS);
684		kfree_skb(skb);
685		pr_debug("ping_queue_rcv_skb -> failed\n");
686		return -1;
687	}
688	return 0;
689}
690
691
692/*
693 *	All we need to do is get the socket.
694 */
695
696void ping_rcv(struct sk_buff *skb)
697{
698	struct sock *sk;
699	struct net *net = dev_net(skb->dev);
700	struct iphdr *iph = ip_hdr(skb);
701	struct icmphdr *icmph = icmp_hdr(skb);
702	u32 saddr = iph->saddr;
703	u32 daddr = iph->daddr;
704
705	/* We assume the packet has already been checked by icmp_rcv */
706
707	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
708		skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
709
710	/* Push ICMP header back */
711	skb_push(skb, skb->data - (u8 *)icmph);
712
713	sk = ping_v4_lookup(net, saddr, daddr, ntohs(icmph->un.echo.id),
714			    skb->dev->ifindex);
715	if (sk != NULL) {
716		pr_debug("rcv on socket %p\n", sk);
717		ping_queue_rcv_skb(sk, skb_get(skb));
718		sock_put(sk);
719		return;
720	}
721	pr_debug("no socket, dropping\n");
722
723	/* We're called from icmp_rcv(). kfree_skb() is done there. */
724}
725
726struct proto ping_prot = {
727	.name =		"PING",
728	.owner =	THIS_MODULE,
729	.init =		ping_init_sock,
730	.close =	ping_close,
731	.connect =	ip4_datagram_connect,
732	.disconnect =	udp_disconnect,
733	.setsockopt =	ip_setsockopt,
734	.getsockopt =	ip_getsockopt,
735	.sendmsg =	ping_sendmsg,
736	.recvmsg =	ping_recvmsg,
737	.bind =		ping_bind,
738	.backlog_rcv =	ping_queue_rcv_skb,
739	.hash =		ping_v4_hash,
740	.unhash =	ping_v4_unhash,
741	.get_port =	ping_v4_get_port,
742	.obj_size =	sizeof(struct inet_sock),
743};
744EXPORT_SYMBOL(ping_prot);
745
746#ifdef CONFIG_PROC_FS
747
748static struct sock *ping_get_first(struct seq_file *seq, int start)
749{
750	struct sock *sk;
751	struct ping_iter_state *state = seq->private;
752	struct net *net = seq_file_net(seq);
753
754	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
755	     ++state->bucket) {
756		struct hlist_nulls_node *node;
757		struct hlist_nulls_head *hslot = &ping_table.hash[state->bucket];
758
759		if (hlist_nulls_empty(hslot))
760			continue;
761
762		sk_nulls_for_each(sk, node, hslot) {
763			if (net_eq(sock_net(sk), net))
764				goto found;
765		}
766	}
767	sk = NULL;
768found:
769	return sk;
770}
771
772static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
773{
774	struct ping_iter_state *state = seq->private;
775	struct net *net = seq_file_net(seq);
776
777	do {
778		sk = sk_nulls_next(sk);
779	} while (sk && (!net_eq(sock_net(sk), net)));
780
781	if (!sk)
782		return ping_get_first(seq, state->bucket + 1);
783	return sk;
784}
785
786static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
787{
788	struct sock *sk = ping_get_first(seq, 0);
789
790	if (sk)
791		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
792			--pos;
793	return pos ? NULL : sk;
794}
795
796static void *ping_seq_start(struct seq_file *seq, loff_t *pos)
797{
798	struct ping_iter_state *state = seq->private;
799	state->bucket = 0;
800
801	read_lock_bh(&ping_table.lock);
802
803	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
804}
805
806static void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
807{
808	struct sock *sk;
809
810	if (v == SEQ_START_TOKEN)
811		sk = ping_get_idx(seq, 0);
812	else
813		sk = ping_get_next(seq, v);
814
815	++*pos;
816	return sk;
817}
818
819static void ping_seq_stop(struct seq_file *seq, void *v)
820{
821	read_unlock_bh(&ping_table.lock);
822}
823
824static void ping_format_sock(struct sock *sp, struct seq_file *f,
825		int bucket, int *len)
826{
827	struct inet_sock *inet = inet_sk(sp);
828	__be32 dest = inet->inet_daddr;
829	__be32 src = inet->inet_rcv_saddr;
830	__u16 destp = ntohs(inet->inet_dport);
831	__u16 srcp = ntohs(inet->inet_sport);
832
833	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
834		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
835		bucket, src, srcp, dest, destp, sp->sk_state,
836		sk_wmem_alloc_get(sp),
837		sk_rmem_alloc_get(sp),
838		0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
839		atomic_read(&sp->sk_refcnt), sp,
840		atomic_read(&sp->sk_drops), len);
841}
842
843static int ping_seq_show(struct seq_file *seq, void *v)
844{
845	if (v == SEQ_START_TOKEN)
846		seq_printf(seq, "%-127s\n",
847			   "  sl  local_address rem_address   st tx_queue "
848			   "rx_queue tr tm->when retrnsmt   uid  timeout "
849			   "inode ref pointer drops");
850	else {
851		struct ping_iter_state *state = seq->private;
852		int len;
853
854		ping_format_sock(v, seq, state->bucket, &len);
855		seq_printf(seq, "%*s\n", 127 - len, "");
856	}
857	return 0;
858}
859
860static const struct seq_operations ping_seq_ops = {
861	.show		= ping_seq_show,
862	.start		= ping_seq_start,
863	.next		= ping_seq_next,
864	.stop		= ping_seq_stop,
865};
866
867static int ping_seq_open(struct inode *inode, struct file *file)
868{
869	return seq_open_net(inode, file, &ping_seq_ops,
870			   sizeof(struct ping_iter_state));
871}
872
873static const struct file_operations ping_seq_fops = {
874	.open		= ping_seq_open,
875	.read		= seq_read,
876	.llseek		= seq_lseek,
877	.release	= seq_release_net,
878};
879
880static int ping_proc_register(struct net *net)
881{
882	struct proc_dir_entry *p;
883	int rc = 0;
884
885	p = proc_net_fops_create(net, "icmp", S_IRUGO, &ping_seq_fops);
886	if (!p)
887		rc = -ENOMEM;
888	return rc;
889}
890
891static void ping_proc_unregister(struct net *net)
892{
893	proc_net_remove(net, "icmp");
894}
895
896
897static int __net_init ping_proc_init_net(struct net *net)
898{
899	return ping_proc_register(net);
900}
901
902static void __net_exit ping_proc_exit_net(struct net *net)
903{
904	ping_proc_unregister(net);
905}
906
907static struct pernet_operations ping_net_ops = {
908	.init = ping_proc_init_net,
909	.exit = ping_proc_exit_net,
910};
911
912int __init ping_proc_init(void)
913{
914	return register_pernet_subsys(&ping_net_ops);
915}
916
917void ping_proc_exit(void)
918{
919	unregister_pernet_subsys(&ping_net_ops);
920}
921
922#endif
923
924void __init ping_init(void)
925{
926	int i;
927
928	for (i = 0; i < PING_HTABLE_SIZE; i++)
929		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
930	rwlock_init(&ping_table.lock);
931}
932