xref: /linux/net/ipv4/ping.c (revision 2c1ba398ac9da3305815f6ae8e95ae2b9fd3b5ff)
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/udp.h>
45 #include <net/route.h>
46 #include <net/inet_common.h>
47 #include <net/checksum.h>
48 
49 
50 static struct ping_table ping_table;
51 
52 static u16 ping_port_rover;
53 
54 static inline int ping_hashfn(struct net *net, unsigned num, unsigned mask)
55 {
56 	int res = (num + net_hash_mix(net)) & mask;
57 	pr_debug("hash(%d) = %d\n", num, res);
58 	return res;
59 }
60 
61 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
62 					     struct net *net, unsigned num)
63 {
64 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
65 }
66 
67 static int ping_v4_get_port(struct sock *sk, unsigned short ident)
68 {
69 	struct hlist_nulls_node *node;
70 	struct hlist_nulls_head *hlist;
71 	struct inet_sock *isk, *isk2;
72 	struct sock *sk2 = NULL;
73 
74 	isk = inet_sk(sk);
75 	write_lock_bh(&ping_table.lock);
76 	if (ident == 0) {
77 		u32 i;
78 		u16 result = ping_port_rover + 1;
79 
80 		for (i = 0; i < (1L << 16); i++, result++) {
81 			if (!result)
82 				result++; /* avoid zero */
83 			hlist = ping_hashslot(&ping_table, sock_net(sk),
84 					    result);
85 			ping_portaddr_for_each_entry(sk2, node, hlist) {
86 				isk2 = inet_sk(sk2);
87 
88 				if (isk2->inet_num == result)
89 					goto next_port;
90 			}
91 
92 			/* found */
93 			ping_port_rover = ident = result;
94 			break;
95 next_port:
96 			;
97 		}
98 		if (i >= (1L << 16))
99 			goto fail;
100 	} else {
101 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
102 		ping_portaddr_for_each_entry(sk2, node, hlist) {
103 			isk2 = inet_sk(sk2);
104 
105 			if ((isk2->inet_num == ident) &&
106 			    (sk2 != sk) &&
107 			    (!sk2->sk_reuse || !sk->sk_reuse))
108 				goto fail;
109 		}
110 	}
111 
112 	pr_debug("found port/ident = %d\n", ident);
113 	isk->inet_num = ident;
114 	if (sk_unhashed(sk)) {
115 		pr_debug("was not hashed\n");
116 		sock_hold(sk);
117 		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
118 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
119 	}
120 	write_unlock_bh(&ping_table.lock);
121 	return 0;
122 
123 fail:
124 	write_unlock_bh(&ping_table.lock);
125 	return 1;
126 }
127 
128 static void ping_v4_hash(struct sock *sk)
129 {
130 	pr_debug("ping_v4_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
131 	BUG(); /* "Please do not press this button again." */
132 }
133 
134 static void ping_v4_unhash(struct sock *sk)
135 {
136 	struct inet_sock *isk = inet_sk(sk);
137 	pr_debug("ping_v4_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
138 	if (sk_hashed(sk)) {
139 		write_lock_bh(&ping_table.lock);
140 		hlist_nulls_del(&sk->sk_nulls_node);
141 		sock_put(sk);
142 		isk->inet_num = isk->inet_sport = 0;
143 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
144 		write_unlock_bh(&ping_table.lock);
145 	}
146 }
147 
148 static struct sock *ping_v4_lookup(struct net *net, u32 saddr, u32 daddr,
149 				   u16 ident, int dif)
150 {
151 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
152 	struct sock *sk = NULL;
153 	struct inet_sock *isk;
154 	struct hlist_nulls_node *hnode;
155 
156 	pr_debug("try to find: num = %d, daddr = %ld, dif = %d\n",
157 			 (int)ident, (unsigned long)daddr, dif);
158 	read_lock_bh(&ping_table.lock);
159 
160 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
161 		isk = inet_sk(sk);
162 
163 		pr_debug("found: %p: num = %d, daddr = %ld, dif = %d\n", sk,
164 			 (int)isk->inet_num, (unsigned long)isk->inet_rcv_saddr,
165 			 sk->sk_bound_dev_if);
166 
167 		pr_debug("iterate\n");
168 		if (isk->inet_num != ident)
169 			continue;
170 		if (isk->inet_rcv_saddr && isk->inet_rcv_saddr != daddr)
171 			continue;
172 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
173 			continue;
174 
175 		sock_hold(sk);
176 		goto exit;
177 	}
178 
179 	sk = NULL;
180 exit:
181 	read_unlock_bh(&ping_table.lock);
182 
183 	return sk;
184 }
185 
186 static void inet_get_ping_group_range_net(struct net *net, gid_t *low,
187 					  gid_t *high)
188 {
189 	gid_t *data = net->ipv4.sysctl_ping_group_range;
190 	unsigned seq;
191 	do {
192 		seq = read_seqbegin(&sysctl_local_ports.lock);
193 
194 		*low = data[0];
195 		*high = data[1];
196 	} while (read_seqretry(&sysctl_local_ports.lock, seq));
197 }
198 
199 
200 static int ping_init_sock(struct sock *sk)
201 {
202 	struct net *net = sock_net(sk);
203 	gid_t group = current_egid();
204 	gid_t range[2];
205 	struct group_info *group_info = get_current_groups();
206 	int i, j, count = group_info->ngroups;
207 
208 	inet_get_ping_group_range_net(net, range, range+1);
209 	if (range[0] <= group && group <= range[1])
210 		return 0;
211 
212 	for (i = 0; i < group_info->nblocks; i++) {
213 		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
214 
215 		for (j = 0; j < cp_count; j++) {
216 			group = group_info->blocks[i][j];
217 			if (range[0] <= group && group <= range[1])
218 				return 0;
219 		}
220 
221 		count -= cp_count;
222 	}
223 
224 	return -EACCES;
225 }
226 
227 static void ping_close(struct sock *sk, long timeout)
228 {
229 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
230 		inet_sk(sk), inet_sk(sk)->inet_num);
231 	pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
232 
233 	sk_common_release(sk);
234 }
235 
236 /*
237  * We need our own bind because there are no privileged id's == local ports.
238  * Moreover, we don't allow binding to multi- and broadcast addresses.
239  */
240 
241 static int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
242 {
243 	struct sockaddr_in *addr = (struct sockaddr_in *)uaddr;
244 	struct inet_sock *isk = inet_sk(sk);
245 	unsigned short snum;
246 	int chk_addr_ret;
247 	int err;
248 
249 	if (addr_len < sizeof(struct sockaddr_in))
250 		return -EINVAL;
251 
252 	pr_debug("ping_v4_bind(sk=%p,sa_addr=%08x,sa_port=%d)\n",
253 		sk, addr->sin_addr.s_addr, ntohs(addr->sin_port));
254 
255 	chk_addr_ret = inet_addr_type(sock_net(sk), addr->sin_addr.s_addr);
256 	if (addr->sin_addr.s_addr == INADDR_ANY)
257 		chk_addr_ret = RTN_LOCAL;
258 
259 	if ((sysctl_ip_nonlocal_bind == 0 &&
260 	    isk->freebind == 0 && isk->transparent == 0 &&
261 	     chk_addr_ret != RTN_LOCAL) ||
262 	    chk_addr_ret == RTN_MULTICAST ||
263 	    chk_addr_ret == RTN_BROADCAST)
264 		return -EADDRNOTAVAIL;
265 
266 	lock_sock(sk);
267 
268 	err = -EINVAL;
269 	if (isk->inet_num != 0)
270 		goto out;
271 
272 	err = -EADDRINUSE;
273 	isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
274 	snum = ntohs(addr->sin_port);
275 	if (ping_v4_get_port(sk, snum) != 0) {
276 		isk->inet_saddr = isk->inet_rcv_saddr = 0;
277 		goto out;
278 	}
279 
280 	pr_debug("after bind(): num = %d, daddr = %ld, dif = %d\n",
281 		(int)isk->inet_num,
282 		(unsigned long) isk->inet_rcv_saddr,
283 		(int)sk->sk_bound_dev_if);
284 
285 	err = 0;
286 	if (isk->inet_rcv_saddr)
287 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
288 	if (snum)
289 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
290 	isk->inet_sport = htons(isk->inet_num);
291 	isk->inet_daddr = 0;
292 	isk->inet_dport = 0;
293 	sk_dst_reset(sk);
294 out:
295 	release_sock(sk);
296 	pr_debug("ping_v4_bind -> %d\n", err);
297 	return err;
298 }
299 
300 /*
301  * Is this a supported type of ICMP message?
302  */
303 
304 static inline int ping_supported(int type, int code)
305 {
306 	if (type == ICMP_ECHO && code == 0)
307 		return 1;
308 	return 0;
309 }
310 
311 /*
312  * This routine is called by the ICMP module when it gets some
313  * sort of error condition.
314  */
315 
316 static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
317 
318 void ping_err(struct sk_buff *skb, u32 info)
319 {
320 	struct iphdr *iph = (struct iphdr *)skb->data;
321 	struct icmphdr *icmph = (struct icmphdr *)(skb->data+(iph->ihl<<2));
322 	struct inet_sock *inet_sock;
323 	int type = icmph->type;
324 	int code = icmph->code;
325 	struct net *net = dev_net(skb->dev);
326 	struct sock *sk;
327 	int harderr;
328 	int err;
329 
330 	/* We assume the packet has already been checked by icmp_unreach */
331 
332 	if (!ping_supported(icmph->type, icmph->code))
333 		return;
334 
335 	pr_debug("ping_err(type=%04x,code=%04x,id=%04x,seq=%04x)\n", type,
336 		code, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
337 
338 	sk = ping_v4_lookup(net, iph->daddr, iph->saddr,
339 			    ntohs(icmph->un.echo.id), skb->dev->ifindex);
340 	if (sk == NULL) {
341 		ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
342 		pr_debug("no socket, dropping\n");
343 		return;	/* No socket for error */
344 	}
345 	pr_debug("err on socket %p\n", sk);
346 
347 	err = 0;
348 	harderr = 0;
349 	inet_sock = inet_sk(sk);
350 
351 	switch (type) {
352 	default:
353 	case ICMP_TIME_EXCEEDED:
354 		err = EHOSTUNREACH;
355 		break;
356 	case ICMP_SOURCE_QUENCH:
357 		/* This is not a real error but ping wants to see it.
358 		 * Report it with some fake errno. */
359 		err = EREMOTEIO;
360 		break;
361 	case ICMP_PARAMETERPROB:
362 		err = EPROTO;
363 		harderr = 1;
364 		break;
365 	case ICMP_DEST_UNREACH:
366 		if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
367 			if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
368 				err = EMSGSIZE;
369 				harderr = 1;
370 				break;
371 			}
372 			goto out;
373 		}
374 		err = EHOSTUNREACH;
375 		if (code <= NR_ICMP_UNREACH) {
376 			harderr = icmp_err_convert[code].fatal;
377 			err = icmp_err_convert[code].errno;
378 		}
379 		break;
380 	case ICMP_REDIRECT:
381 		/* See ICMP_SOURCE_QUENCH */
382 		err = EREMOTEIO;
383 		break;
384 	}
385 
386 	/*
387 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
388 	 *	4.1.3.3.
389 	 */
390 	if (!inet_sock->recverr) {
391 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
392 			goto out;
393 	} else {
394 		ip_icmp_error(sk, skb, err, 0 /* no remote port */,
395 			 info, (u8 *)icmph);
396 	}
397 	sk->sk_err = err;
398 	sk->sk_error_report(sk);
399 out:
400 	sock_put(sk);
401 }
402 
403 /*
404  *	Copy and checksum an ICMP Echo packet from user space into a buffer.
405  */
406 
407 struct pingfakehdr {
408 	struct icmphdr icmph;
409 	struct iovec *iov;
410 	u32 wcheck;
411 };
412 
413 static int ping_getfrag(void *from, char * to,
414 			int offset, int fraglen, int odd, struct sk_buff *skb)
415 {
416 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
417 
418 	if (offset == 0) {
419 		if (fraglen < sizeof(struct icmphdr))
420 			BUG();
421 		if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
422 			    pfh->iov, 0, fraglen - sizeof(struct icmphdr),
423 			    &pfh->wcheck))
424 			return -EFAULT;
425 
426 		return 0;
427 	}
428 	if (offset < sizeof(struct icmphdr))
429 		BUG();
430 	if (csum_partial_copy_fromiovecend
431 			(to, pfh->iov, offset - sizeof(struct icmphdr),
432 			 fraglen, &pfh->wcheck))
433 		return -EFAULT;
434 	return 0;
435 }
436 
437 static int ping_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
438 				    struct flowi4 *fl4)
439 {
440 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
441 
442 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
443 		sizeof(struct icmphdr), pfh->wcheck);
444 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
445 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
446 	skb->ip_summed = CHECKSUM_NONE;
447 	return ip_push_pending_frames(sk, fl4);
448 }
449 
450 static int ping_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
451 			size_t len)
452 {
453 	struct net *net = sock_net(sk);
454 	struct flowi4 fl4;
455 	struct inet_sock *inet = inet_sk(sk);
456 	struct ipcm_cookie ipc;
457 	struct icmphdr user_icmph;
458 	struct pingfakehdr pfh;
459 	struct rtable *rt = NULL;
460 	struct ip_options_data opt_copy;
461 	int free = 0;
462 	u32 saddr, daddr, faddr;
463 	u8  tos;
464 	int err;
465 
466 	pr_debug("ping_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
467 
468 
469 	if (len > 0xFFFF)
470 		return -EMSGSIZE;
471 
472 	/*
473 	 *	Check the flags.
474 	 */
475 
476 	/* Mirror BSD error message compatibility */
477 	if (msg->msg_flags & MSG_OOB)
478 		return -EOPNOTSUPP;
479 
480 	/*
481 	 *	Fetch the ICMP header provided by the userland.
482 	 *	iovec is modified!
483 	 */
484 
485 	if (memcpy_fromiovec((u8 *)&user_icmph, msg->msg_iov,
486 			     sizeof(struct icmphdr)))
487 		return -EFAULT;
488 	if (!ping_supported(user_icmph.type, user_icmph.code))
489 		return -EINVAL;
490 
491 	/*
492 	 *	Get and verify the address.
493 	 */
494 
495 	if (msg->msg_name) {
496 		struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
497 		if (msg->msg_namelen < sizeof(*usin))
498 			return -EINVAL;
499 		if (usin->sin_family != AF_INET)
500 			return -EINVAL;
501 		daddr = usin->sin_addr.s_addr;
502 		/* no remote port */
503 	} else {
504 		if (sk->sk_state != TCP_ESTABLISHED)
505 			return -EDESTADDRREQ;
506 		daddr = inet->inet_daddr;
507 		/* no remote port */
508 	}
509 
510 	ipc.addr = inet->inet_saddr;
511 	ipc.opt = NULL;
512 	ipc.oif = sk->sk_bound_dev_if;
513 	ipc.tx_flags = 0;
514 	err = sock_tx_timestamp(sk, &ipc.tx_flags);
515 	if (err)
516 		return err;
517 
518 	if (msg->msg_controllen) {
519 		err = ip_cmsg_send(sock_net(sk), msg, &ipc);
520 		if (err)
521 			return err;
522 		if (ipc.opt)
523 			free = 1;
524 	}
525 	if (!ipc.opt) {
526 		struct ip_options_rcu *inet_opt;
527 
528 		rcu_read_lock();
529 		inet_opt = rcu_dereference(inet->inet_opt);
530 		if (inet_opt) {
531 			memcpy(&opt_copy, inet_opt,
532 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
533 			ipc.opt = &opt_copy.opt;
534 		}
535 		rcu_read_unlock();
536 	}
537 
538 	saddr = ipc.addr;
539 	ipc.addr = faddr = daddr;
540 
541 	if (ipc.opt && ipc.opt->opt.srr) {
542 		if (!daddr)
543 			return -EINVAL;
544 		faddr = ipc.opt->opt.faddr;
545 	}
546 	tos = RT_TOS(inet->tos);
547 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
548 	    (msg->msg_flags & MSG_DONTROUTE) ||
549 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
550 		tos |= RTO_ONLINK;
551 	}
552 
553 	if (ipv4_is_multicast(daddr)) {
554 		if (!ipc.oif)
555 			ipc.oif = inet->mc_index;
556 		if (!saddr)
557 			saddr = inet->mc_addr;
558 	}
559 
560 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
561 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
562 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
563 
564 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
565 	rt = ip_route_output_flow(net, &fl4, sk);
566 	if (IS_ERR(rt)) {
567 		err = PTR_ERR(rt);
568 		rt = NULL;
569 		if (err == -ENETUNREACH)
570 			IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
571 		goto out;
572 	}
573 
574 	err = -EACCES;
575 	if ((rt->rt_flags & RTCF_BROADCAST) &&
576 	    !sock_flag(sk, SOCK_BROADCAST))
577 		goto out;
578 
579 	if (msg->msg_flags & MSG_CONFIRM)
580 		goto do_confirm;
581 back_from_confirm:
582 
583 	if (!ipc.addr)
584 		ipc.addr = fl4.daddr;
585 
586 	lock_sock(sk);
587 
588 	pfh.icmph.type = user_icmph.type; /* already checked */
589 	pfh.icmph.code = user_icmph.code; /* ditto */
590 	pfh.icmph.checksum = 0;
591 	pfh.icmph.un.echo.id = inet->inet_sport;
592 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
593 	pfh.iov = msg->msg_iov;
594 	pfh.wcheck = 0;
595 
596 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
597 			0, &ipc, &rt, msg->msg_flags);
598 	if (err)
599 		ip_flush_pending_frames(sk);
600 	else
601 		err = ping_push_pending_frames(sk, &pfh, &fl4);
602 	release_sock(sk);
603 
604 out:
605 	ip_rt_put(rt);
606 	if (free)
607 		kfree(ipc.opt);
608 	if (!err) {
609 		icmp_out_count(sock_net(sk), user_icmph.type);
610 		return len;
611 	}
612 	return err;
613 
614 do_confirm:
615 	dst_confirm(&rt->dst);
616 	if (!(msg->msg_flags & MSG_PROBE) || len)
617 		goto back_from_confirm;
618 	err = 0;
619 	goto out;
620 }
621 
622 static int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
623 			size_t len, int noblock, int flags, int *addr_len)
624 {
625 	struct inet_sock *isk = inet_sk(sk);
626 	struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
627 	struct sk_buff *skb;
628 	int copied, err;
629 
630 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
631 
632 	if (flags & MSG_OOB)
633 		goto out;
634 
635 	if (addr_len)
636 		*addr_len = sizeof(*sin);
637 
638 	if (flags & MSG_ERRQUEUE)
639 		return ip_recv_error(sk, msg, len);
640 
641 	skb = skb_recv_datagram(sk, flags, noblock, &err);
642 	if (!skb)
643 		goto out;
644 
645 	copied = skb->len;
646 	if (copied > len) {
647 		msg->msg_flags |= MSG_TRUNC;
648 		copied = len;
649 	}
650 
651 	/* Don't bother checking the checksum */
652 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
653 	if (err)
654 		goto done;
655 
656 	sock_recv_timestamp(msg, sk, skb);
657 
658 	/* Copy the address. */
659 	if (sin) {
660 		sin->sin_family = AF_INET;
661 		sin->sin_port = 0 /* skb->h.uh->source */;
662 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
663 		memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
664 	}
665 	if (isk->cmsg_flags)
666 		ip_cmsg_recv(msg, skb);
667 	err = copied;
668 
669 done:
670 	skb_free_datagram(sk, skb);
671 out:
672 	pr_debug("ping_recvmsg -> %d\n", err);
673 	return err;
674 }
675 
676 static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
677 {
678 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
679 		inet_sk(sk), inet_sk(sk)->inet_num, skb);
680 	if (sock_queue_rcv_skb(sk, skb) < 0) {
681 		ICMP_INC_STATS_BH(sock_net(sk), ICMP_MIB_INERRORS);
682 		kfree_skb(skb);
683 		pr_debug("ping_queue_rcv_skb -> failed\n");
684 		return -1;
685 	}
686 	return 0;
687 }
688 
689 
690 /*
691  *	All we need to do is get the socket.
692  */
693 
694 void ping_rcv(struct sk_buff *skb)
695 {
696 	struct sock *sk;
697 	struct net *net = dev_net(skb->dev);
698 	struct iphdr *iph = ip_hdr(skb);
699 	struct icmphdr *icmph = icmp_hdr(skb);
700 	u32 saddr = iph->saddr;
701 	u32 daddr = iph->daddr;
702 
703 	/* We assume the packet has already been checked by icmp_rcv */
704 
705 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
706 		skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
707 
708 	/* Push ICMP header back */
709 	skb_push(skb, skb->data - (u8 *)icmph);
710 
711 	sk = ping_v4_lookup(net, saddr, daddr, ntohs(icmph->un.echo.id),
712 			    skb->dev->ifindex);
713 	if (sk != NULL) {
714 		pr_debug("rcv on socket %p\n", sk);
715 		ping_queue_rcv_skb(sk, skb_get(skb));
716 		sock_put(sk);
717 		return;
718 	}
719 	pr_debug("no socket, dropping\n");
720 
721 	/* We're called from icmp_rcv(). kfree_skb() is done there. */
722 }
723 
724 struct proto ping_prot = {
725 	.name =		"PING",
726 	.owner =	THIS_MODULE,
727 	.init =		ping_init_sock,
728 	.close =	ping_close,
729 	.connect =	ip4_datagram_connect,
730 	.disconnect =	udp_disconnect,
731 	.setsockopt =	ip_setsockopt,
732 	.getsockopt =	ip_getsockopt,
733 	.sendmsg =	ping_sendmsg,
734 	.recvmsg =	ping_recvmsg,
735 	.bind =		ping_bind,
736 	.backlog_rcv =	ping_queue_rcv_skb,
737 	.hash =		ping_v4_hash,
738 	.unhash =	ping_v4_unhash,
739 	.get_port =	ping_v4_get_port,
740 	.obj_size =	sizeof(struct inet_sock),
741 };
742 EXPORT_SYMBOL(ping_prot);
743 
744 #ifdef CONFIG_PROC_FS
745 
746 static struct sock *ping_get_first(struct seq_file *seq, int start)
747 {
748 	struct sock *sk;
749 	struct ping_iter_state *state = seq->private;
750 	struct net *net = seq_file_net(seq);
751 
752 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
753 	     ++state->bucket) {
754 		struct hlist_nulls_node *node;
755 		struct hlist_nulls_head *hslot;
756 
757 		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;
768 found:
769 	return sk;
770 }
771 
772 static 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 
786 static 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 
796 static 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 
806 static 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 
819 static void ping_seq_stop(struct seq_file *seq, void *v)
820 {
821 	read_unlock_bh(&ping_table.lock);
822 }
823 
824 static 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 
843 static 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 
860 static 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 
867 static 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 
873 static 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 
880 static 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 
891 static void ping_proc_unregister(struct net *net)
892 {
893 	proc_net_remove(net, "icmp");
894 }
895 
896 
897 static int __net_init ping_proc_init_net(struct net *net)
898 {
899 	return ping_proc_register(net);
900 }
901 
902 static void __net_exit ping_proc_exit_net(struct net *net)
903 {
904 	ping_proc_unregister(net);
905 }
906 
907 static struct pernet_operations ping_net_ops = {
908 	.init = ping_proc_init_net,
909 	.exit = ping_proc_exit_net,
910 };
911 
912 int __init ping_proc_init(void)
913 {
914 	return register_pernet_subsys(&ping_net_ops);
915 }
916 
917 void ping_proc_exit(void)
918 {
919 	unregister_pernet_subsys(&ping_net_ops);
920 }
921 
922 #endif
923 
924 void __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