xref: /linux/net/ipv4/ping.c (revision a508da6cc0093171833efb8376b00473f24221b9)
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 <linux/uaccess.h>
24 #include <linux/types.h>
25 #include <linux/fcntl.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/in.h>
29 #include <linux/errno.h>
30 #include <linux/timer.h>
31 #include <linux/mm.h>
32 #include <linux/inet.h>
33 #include <linux/netdevice.h>
34 #include <net/snmp.h>
35 #include <net/ip.h>
36 #include <net/ipv6.h>
37 #include <net/icmp.h>
38 #include <net/protocol.h>
39 #include <linux/skbuff.h>
40 #include <linux/proc_fs.h>
41 #include <linux/export.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 int num, unsigned int mask)
55 {
56 	int res = (num + net_hash_mix(net)) & mask;
57 
58 	pr_debug("hash(%d) = %d\n", num, res);
59 	return res;
60 }
61 
62 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
63 					     struct net *net, unsigned int num)
64 {
65 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
66 }
67 
68 static 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;
96 next_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 
124 fail:
125 	write_unlock_bh(&ping_table.lock);
126 	return 1;
127 }
128 
129 static 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 
135 static 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 		write_lock_bh(&ping_table.lock);
141 		hlist_nulls_del(&sk->sk_nulls_node);
142 		sock_put(sk);
143 		isk->inet_num = 0;
144 		isk->inet_sport = 0;
145 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
146 		write_unlock_bh(&ping_table.lock);
147 	}
148 }
149 
150 static struct sock *ping_v4_lookup(struct net *net, __be32 saddr, __be32 daddr,
151 				   u16 ident, int dif)
152 {
153 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
154 	struct sock *sk = NULL;
155 	struct inet_sock *isk;
156 	struct hlist_nulls_node *hnode;
157 
158 	pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
159 		 (int)ident, &daddr, dif);
160 	read_lock_bh(&ping_table.lock);
161 
162 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
163 		isk = inet_sk(sk);
164 
165 		pr_debug("found: %p: num = %d, daddr = %pI4, dif = %d\n", sk,
166 			 (int)isk->inet_num, &isk->inet_rcv_saddr,
167 			 sk->sk_bound_dev_if);
168 
169 		pr_debug("iterate\n");
170 		if (isk->inet_num != ident)
171 			continue;
172 		if (isk->inet_rcv_saddr && isk->inet_rcv_saddr != daddr)
173 			continue;
174 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
175 			continue;
176 
177 		sock_hold(sk);
178 		goto exit;
179 	}
180 
181 	sk = NULL;
182 exit:
183 	read_unlock_bh(&ping_table.lock);
184 
185 	return sk;
186 }
187 
188 static void inet_get_ping_group_range_net(struct net *net, gid_t *low,
189 					  gid_t *high)
190 {
191 	gid_t *data = net->ipv4.sysctl_ping_group_range;
192 	unsigned int seq;
193 
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 
203 static 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 
230 static 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 
244 static 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 == htonl(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 = %pI4, dif = %d\n",
284 		 (int)isk->inet_num,
285 		 &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);
297 out:
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 
307 static 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 
319 static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
320 
321 void 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 		pr_debug("no socket, dropping\n");
345 		return;	/* No socket for error */
346 	}
347 	pr_debug("err on socket %p\n", sk);
348 
349 	err = 0;
350 	harderr = 0;
351 	inet_sock = inet_sk(sk);
352 
353 	switch (type) {
354 	default:
355 	case ICMP_TIME_EXCEEDED:
356 		err = EHOSTUNREACH;
357 		break;
358 	case ICMP_SOURCE_QUENCH:
359 		/* This is not a real error but ping wants to see it.
360 		 * Report it with some fake errno. */
361 		err = EREMOTEIO;
362 		break;
363 	case ICMP_PARAMETERPROB:
364 		err = EPROTO;
365 		harderr = 1;
366 		break;
367 	case ICMP_DEST_UNREACH:
368 		if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
369 			if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
370 				err = EMSGSIZE;
371 				harderr = 1;
372 				break;
373 			}
374 			goto out;
375 		}
376 		err = EHOSTUNREACH;
377 		if (code <= NR_ICMP_UNREACH) {
378 			harderr = icmp_err_convert[code].fatal;
379 			err = icmp_err_convert[code].errno;
380 		}
381 		break;
382 	case ICMP_REDIRECT:
383 		/* See ICMP_SOURCE_QUENCH */
384 		err = EREMOTEIO;
385 		break;
386 	}
387 
388 	/*
389 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
390 	 *	4.1.3.3.
391 	 */
392 	if (!inet_sock->recverr) {
393 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
394 			goto out;
395 	} else {
396 		ip_icmp_error(sk, skb, err, 0 /* no remote port */,
397 			 info, (u8 *)icmph);
398 	}
399 	sk->sk_err = err;
400 	sk->sk_error_report(sk);
401 out:
402 	sock_put(sk);
403 }
404 
405 /*
406  *	Copy and checksum an ICMP Echo packet from user space into a buffer.
407  */
408 
409 struct pingfakehdr {
410 	struct icmphdr icmph;
411 	struct iovec *iov;
412 	__wsum wcheck;
413 };
414 
415 static int ping_getfrag(void *from, char *to,
416 			int offset, int fraglen, int odd, struct sk_buff *skb)
417 {
418 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
419 
420 	if (offset == 0) {
421 		if (fraglen < sizeof(struct icmphdr))
422 			BUG();
423 		if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
424 			    pfh->iov, 0, fraglen - sizeof(struct icmphdr),
425 			    &pfh->wcheck))
426 			return -EFAULT;
427 
428 		return 0;
429 	}
430 	if (offset < sizeof(struct icmphdr))
431 		BUG();
432 	if (csum_partial_copy_fromiovecend
433 			(to, pfh->iov, offset - sizeof(struct icmphdr),
434 			 fraglen, &pfh->wcheck))
435 		return -EFAULT;
436 	return 0;
437 }
438 
439 static int ping_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
440 				    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 
452 static 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 	__be32 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 	} else if (!ipc.oif)
561 		ipc.oif = inet->uc_index;
562 
563 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
564 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
565 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
566 
567 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
568 	rt = ip_route_output_flow(net, &fl4, sk);
569 	if (IS_ERR(rt)) {
570 		err = PTR_ERR(rt);
571 		rt = NULL;
572 		if (err == -ENETUNREACH)
573 			IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
574 		goto out;
575 	}
576 
577 	err = -EACCES;
578 	if ((rt->rt_flags & RTCF_BROADCAST) &&
579 	    !sock_flag(sk, SOCK_BROADCAST))
580 		goto out;
581 
582 	if (msg->msg_flags & MSG_CONFIRM)
583 		goto do_confirm;
584 back_from_confirm:
585 
586 	if (!ipc.addr)
587 		ipc.addr = fl4.daddr;
588 
589 	lock_sock(sk);
590 
591 	pfh.icmph.type = user_icmph.type; /* already checked */
592 	pfh.icmph.code = user_icmph.code; /* ditto */
593 	pfh.icmph.checksum = 0;
594 	pfh.icmph.un.echo.id = inet->inet_sport;
595 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
596 	pfh.iov = msg->msg_iov;
597 	pfh.wcheck = 0;
598 
599 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
600 			0, &ipc, &rt, msg->msg_flags);
601 	if (err)
602 		ip_flush_pending_frames(sk);
603 	else
604 		err = ping_push_pending_frames(sk, &pfh, &fl4);
605 	release_sock(sk);
606 
607 out:
608 	ip_rt_put(rt);
609 	if (free)
610 		kfree(ipc.opt);
611 	if (!err) {
612 		icmp_out_count(sock_net(sk), user_icmph.type);
613 		return len;
614 	}
615 	return err;
616 
617 do_confirm:
618 	dst_confirm(&rt->dst);
619 	if (!(msg->msg_flags & MSG_PROBE) || len)
620 		goto back_from_confirm;
621 	err = 0;
622 	goto out;
623 }
624 
625 static int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
626 			size_t len, int noblock, int flags, int *addr_len)
627 {
628 	struct inet_sock *isk = inet_sk(sk);
629 	struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
630 	struct sk_buff *skb;
631 	int copied, err;
632 
633 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
634 
635 	err = -EOPNOTSUPP;
636 	if (flags & MSG_OOB)
637 		goto out;
638 
639 	if (addr_len)
640 		*addr_len = sizeof(*sin);
641 
642 	if (flags & MSG_ERRQUEUE)
643 		return ip_recv_error(sk, msg, len);
644 
645 	skb = skb_recv_datagram(sk, flags, noblock, &err);
646 	if (!skb)
647 		goto out;
648 
649 	copied = skb->len;
650 	if (copied > len) {
651 		msg->msg_flags |= MSG_TRUNC;
652 		copied = len;
653 	}
654 
655 	/* Don't bother checking the checksum */
656 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
657 	if (err)
658 		goto done;
659 
660 	sock_recv_timestamp(msg, sk, skb);
661 
662 	/* Copy the address. */
663 	if (sin) {
664 		sin->sin_family = AF_INET;
665 		sin->sin_port = 0 /* skb->h.uh->source */;
666 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
667 		memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
668 	}
669 	if (isk->cmsg_flags)
670 		ip_cmsg_recv(msg, skb);
671 	err = copied;
672 
673 done:
674 	skb_free_datagram(sk, skb);
675 out:
676 	pr_debug("ping_recvmsg -> %d\n", err);
677 	return err;
678 }
679 
680 static int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
681 {
682 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
683 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
684 	if (sock_queue_rcv_skb(sk, skb) < 0) {
685 		kfree_skb(skb);
686 		pr_debug("ping_queue_rcv_skb -> failed\n");
687 		return -1;
688 	}
689 	return 0;
690 }
691 
692 
693 /*
694  *	All we need to do is get the socket.
695  */
696 
697 void ping_rcv(struct sk_buff *skb)
698 {
699 	struct sock *sk;
700 	struct net *net = dev_net(skb->dev);
701 	struct iphdr *iph = ip_hdr(skb);
702 	struct icmphdr *icmph = icmp_hdr(skb);
703 	__be32 saddr = iph->saddr;
704 	__be32 daddr = iph->daddr;
705 
706 	/* We assume the packet has already been checked by icmp_rcv */
707 
708 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
709 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
710 
711 	/* Push ICMP header back */
712 	skb_push(skb, skb->data - (u8 *)icmph);
713 
714 	sk = ping_v4_lookup(net, saddr, daddr, ntohs(icmph->un.echo.id),
715 			    skb->dev->ifindex);
716 	if (sk != NULL) {
717 		pr_debug("rcv on socket %p\n", sk);
718 		ping_queue_rcv_skb(sk, skb_get(skb));
719 		sock_put(sk);
720 		return;
721 	}
722 	pr_debug("no socket, dropping\n");
723 
724 	/* We're called from icmp_rcv(). kfree_skb() is done there. */
725 }
726 
727 struct proto ping_prot = {
728 	.name =		"PING",
729 	.owner =	THIS_MODULE,
730 	.init =		ping_init_sock,
731 	.close =	ping_close,
732 	.connect =	ip4_datagram_connect,
733 	.disconnect =	udp_disconnect,
734 	.setsockopt =	ip_setsockopt,
735 	.getsockopt =	ip_getsockopt,
736 	.sendmsg =	ping_sendmsg,
737 	.recvmsg =	ping_recvmsg,
738 	.bind =		ping_bind,
739 	.backlog_rcv =	ping_queue_rcv_skb,
740 	.hash =		ping_v4_hash,
741 	.unhash =	ping_v4_unhash,
742 	.get_port =	ping_v4_get_port,
743 	.obj_size =	sizeof(struct inet_sock),
744 };
745 EXPORT_SYMBOL(ping_prot);
746 
747 #ifdef CONFIG_PROC_FS
748 
749 static struct sock *ping_get_first(struct seq_file *seq, int start)
750 {
751 	struct sock *sk;
752 	struct ping_iter_state *state = seq->private;
753 	struct net *net = seq_file_net(seq);
754 
755 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
756 	     ++state->bucket) {
757 		struct hlist_nulls_node *node;
758 		struct hlist_nulls_head *hslot;
759 
760 		hslot = &ping_table.hash[state->bucket];
761 
762 		if (hlist_nulls_empty(hslot))
763 			continue;
764 
765 		sk_nulls_for_each(sk, node, hslot) {
766 			if (net_eq(sock_net(sk), net))
767 				goto found;
768 		}
769 	}
770 	sk = NULL;
771 found:
772 	return sk;
773 }
774 
775 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
776 {
777 	struct ping_iter_state *state = seq->private;
778 	struct net *net = seq_file_net(seq);
779 
780 	do {
781 		sk = sk_nulls_next(sk);
782 	} while (sk && (!net_eq(sock_net(sk), net)));
783 
784 	if (!sk)
785 		return ping_get_first(seq, state->bucket + 1);
786 	return sk;
787 }
788 
789 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
790 {
791 	struct sock *sk = ping_get_first(seq, 0);
792 
793 	if (sk)
794 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
795 			--pos;
796 	return pos ? NULL : sk;
797 }
798 
799 static void *ping_seq_start(struct seq_file *seq, loff_t *pos)
800 {
801 	struct ping_iter_state *state = seq->private;
802 	state->bucket = 0;
803 
804 	read_lock_bh(&ping_table.lock);
805 
806 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
807 }
808 
809 static void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
810 {
811 	struct sock *sk;
812 
813 	if (v == SEQ_START_TOKEN)
814 		sk = ping_get_idx(seq, 0);
815 	else
816 		sk = ping_get_next(seq, v);
817 
818 	++*pos;
819 	return sk;
820 }
821 
822 static void ping_seq_stop(struct seq_file *seq, void *v)
823 {
824 	read_unlock_bh(&ping_table.lock);
825 }
826 
827 static void ping_format_sock(struct sock *sp, struct seq_file *f,
828 		int bucket, int *len)
829 {
830 	struct inet_sock *inet = inet_sk(sp);
831 	__be32 dest = inet->inet_daddr;
832 	__be32 src = inet->inet_rcv_saddr;
833 	__u16 destp = ntohs(inet->inet_dport);
834 	__u16 srcp = ntohs(inet->inet_sport);
835 
836 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
837 		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
838 		bucket, src, srcp, dest, destp, sp->sk_state,
839 		sk_wmem_alloc_get(sp),
840 		sk_rmem_alloc_get(sp),
841 		0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
842 		atomic_read(&sp->sk_refcnt), sp,
843 		atomic_read(&sp->sk_drops), len);
844 }
845 
846 static int ping_seq_show(struct seq_file *seq, void *v)
847 {
848 	if (v == SEQ_START_TOKEN)
849 		seq_printf(seq, "%-127s\n",
850 			   "  sl  local_address rem_address   st tx_queue "
851 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
852 			   "inode ref pointer drops");
853 	else {
854 		struct ping_iter_state *state = seq->private;
855 		int len;
856 
857 		ping_format_sock(v, seq, state->bucket, &len);
858 		seq_printf(seq, "%*s\n", 127 - len, "");
859 	}
860 	return 0;
861 }
862 
863 static const struct seq_operations ping_seq_ops = {
864 	.show		= ping_seq_show,
865 	.start		= ping_seq_start,
866 	.next		= ping_seq_next,
867 	.stop		= ping_seq_stop,
868 };
869 
870 static int ping_seq_open(struct inode *inode, struct file *file)
871 {
872 	return seq_open_net(inode, file, &ping_seq_ops,
873 			   sizeof(struct ping_iter_state));
874 }
875 
876 static const struct file_operations ping_seq_fops = {
877 	.open		= ping_seq_open,
878 	.read		= seq_read,
879 	.llseek		= seq_lseek,
880 	.release	= seq_release_net,
881 };
882 
883 static int ping_proc_register(struct net *net)
884 {
885 	struct proc_dir_entry *p;
886 	int rc = 0;
887 
888 	p = proc_net_fops_create(net, "icmp", S_IRUGO, &ping_seq_fops);
889 	if (!p)
890 		rc = -ENOMEM;
891 	return rc;
892 }
893 
894 static void ping_proc_unregister(struct net *net)
895 {
896 	proc_net_remove(net, "icmp");
897 }
898 
899 
900 static int __net_init ping_proc_init_net(struct net *net)
901 {
902 	return ping_proc_register(net);
903 }
904 
905 static void __net_exit ping_proc_exit_net(struct net *net)
906 {
907 	ping_proc_unregister(net);
908 }
909 
910 static struct pernet_operations ping_net_ops = {
911 	.init = ping_proc_init_net,
912 	.exit = ping_proc_exit_net,
913 };
914 
915 int __init ping_proc_init(void)
916 {
917 	return register_pernet_subsys(&ping_net_ops);
918 }
919 
920 void ping_proc_exit(void)
921 {
922 	unregister_pernet_subsys(&ping_net_ops);
923 }
924 
925 #endif
926 
927 void __init ping_init(void)
928 {
929 	int i;
930 
931 	for (i = 0; i < PING_HTABLE_SIZE; i++)
932 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
933 	rwlock_init(&ping_table.lock);
934 }
935