xref: /linux/net/ipv4/ping.c (revision 0750b8fcf313845b21c71344b4bea8ad7d3cee84)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		"Ping" sockets
8  *
9  * Based on ipv4/udp.c code.
10  *
11  * Authors:	Vasiliy Kulikov / Openwall (for Linux 2.6),
12  *		Pavel Kankovsky (for Linux 2.4.32)
13  *
14  * Pavel gave all rights to bugs to Vasiliy,
15  * none of the bugs are Pavel's now.
16  */
17 
18 #include <linux/uaccess.h>
19 #include <linux/types.h>
20 #include <linux/fcntl.h>
21 #include <linux/socket.h>
22 #include <linux/sockios.h>
23 #include <linux/in.h>
24 #include <linux/errno.h>
25 #include <linux/timer.h>
26 #include <linux/mm.h>
27 #include <linux/inet.h>
28 #include <linux/netdevice.h>
29 #include <net/snmp.h>
30 #include <net/ip.h>
31 #include <net/icmp.h>
32 #include <net/protocol.h>
33 #include <linux/skbuff.h>
34 #include <linux/proc_fs.h>
35 #include <linux/export.h>
36 #include <net/sock.h>
37 #include <net/ping.h>
38 #include <net/udp.h>
39 #include <net/route.h>
40 #include <net/inet_common.h>
41 #include <net/checksum.h>
42 
43 #if IS_ENABLED(CONFIG_IPV6)
44 #include <linux/in6.h>
45 #include <linux/icmpv6.h>
46 #include <net/addrconf.h>
47 #include <net/ipv6.h>
48 #include <net/transp_v6.h>
49 #endif
50 
51 struct ping_table {
52 	struct hlist_nulls_head	hash[PING_HTABLE_SIZE];
53 	rwlock_t		lock;
54 };
55 
56 static struct ping_table ping_table;
57 struct pingv6_ops pingv6_ops;
58 EXPORT_SYMBOL_GPL(pingv6_ops);
59 
60 static u16 ping_port_rover;
61 
62 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
63 {
64 	u32 res = (num + net_hash_mix(net)) & mask;
65 
66 	pr_debug("hash(%u) = %u\n", num, res);
67 	return res;
68 }
69 EXPORT_SYMBOL_GPL(ping_hash);
70 
71 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
72 					     struct net *net, unsigned int num)
73 {
74 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
75 }
76 
77 int ping_get_port(struct sock *sk, unsigned short ident)
78 {
79 	struct hlist_nulls_node *node;
80 	struct hlist_nulls_head *hlist;
81 	struct inet_sock *isk, *isk2;
82 	struct sock *sk2 = NULL;
83 
84 	isk = inet_sk(sk);
85 	write_lock_bh(&ping_table.lock);
86 	if (ident == 0) {
87 		u32 i;
88 		u16 result = ping_port_rover + 1;
89 
90 		for (i = 0; i < (1L << 16); i++, result++) {
91 			if (!result)
92 				result++; /* avoid zero */
93 			hlist = ping_hashslot(&ping_table, sock_net(sk),
94 					    result);
95 			ping_portaddr_for_each_entry(sk2, node, hlist) {
96 				isk2 = inet_sk(sk2);
97 
98 				if (isk2->inet_num == result)
99 					goto next_port;
100 			}
101 
102 			/* found */
103 			ping_port_rover = ident = result;
104 			break;
105 next_port:
106 			;
107 		}
108 		if (i >= (1L << 16))
109 			goto fail;
110 	} else {
111 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
112 		ping_portaddr_for_each_entry(sk2, node, hlist) {
113 			isk2 = inet_sk(sk2);
114 
115 			/* BUG? Why is this reuse and not reuseaddr? ping.c
116 			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
117 			 * that other ping processes can steal its packets.
118 			 */
119 			if ((isk2->inet_num == ident) &&
120 			    (sk2 != sk) &&
121 			    (!sk2->sk_reuse || !sk->sk_reuse))
122 				goto fail;
123 		}
124 	}
125 
126 	pr_debug("found port/ident = %d\n", ident);
127 	isk->inet_num = ident;
128 	if (sk_unhashed(sk)) {
129 		pr_debug("was not hashed\n");
130 		sock_hold(sk);
131 		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
132 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
133 	}
134 	write_unlock_bh(&ping_table.lock);
135 	return 0;
136 
137 fail:
138 	write_unlock_bh(&ping_table.lock);
139 	return 1;
140 }
141 EXPORT_SYMBOL_GPL(ping_get_port);
142 
143 int ping_hash(struct sock *sk)
144 {
145 	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
146 	BUG(); /* "Please do not press this button again." */
147 
148 	return 0;
149 }
150 
151 void ping_unhash(struct sock *sk)
152 {
153 	struct inet_sock *isk = inet_sk(sk);
154 
155 	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
156 	write_lock_bh(&ping_table.lock);
157 	if (sk_hashed(sk)) {
158 		hlist_nulls_del(&sk->sk_nulls_node);
159 		sk_nulls_node_init(&sk->sk_nulls_node);
160 		sock_put(sk);
161 		isk->inet_num = 0;
162 		isk->inet_sport = 0;
163 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
164 	}
165 	write_unlock_bh(&ping_table.lock);
166 }
167 EXPORT_SYMBOL_GPL(ping_unhash);
168 
169 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
170 {
171 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
172 	struct sock *sk = NULL;
173 	struct inet_sock *isk;
174 	struct hlist_nulls_node *hnode;
175 	int dif, sdif;
176 
177 	if (skb->protocol == htons(ETH_P_IP)) {
178 		dif = inet_iif(skb);
179 		sdif = inet_sdif(skb);
180 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
181 			 (int)ident, &ip_hdr(skb)->daddr, dif);
182 #if IS_ENABLED(CONFIG_IPV6)
183 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
184 		dif = inet6_iif(skb);
185 		sdif = inet6_sdif(skb);
186 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
187 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
188 #endif
189 	} else {
190 		pr_err("ping: protocol(%x) is not supported\n", ntohs(skb->protocol));
191 		return NULL;
192 	}
193 
194 	read_lock_bh(&ping_table.lock);
195 
196 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
197 		isk = inet_sk(sk);
198 
199 		pr_debug("iterate\n");
200 		if (isk->inet_num != ident)
201 			continue;
202 
203 		if (skb->protocol == htons(ETH_P_IP) &&
204 		    sk->sk_family == AF_INET) {
205 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
206 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
207 				 sk->sk_bound_dev_if);
208 
209 			if (isk->inet_rcv_saddr &&
210 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
211 				continue;
212 #if IS_ENABLED(CONFIG_IPV6)
213 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
214 			   sk->sk_family == AF_INET6) {
215 
216 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
217 				 (int) isk->inet_num,
218 				 &sk->sk_v6_rcv_saddr,
219 				 sk->sk_bound_dev_if);
220 
221 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
222 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
223 					     &ipv6_hdr(skb)->daddr))
224 				continue;
225 #endif
226 		} else {
227 			continue;
228 		}
229 
230 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
231 		    sk->sk_bound_dev_if != sdif)
232 			continue;
233 
234 		sock_hold(sk);
235 		goto exit;
236 	}
237 
238 	sk = NULL;
239 exit:
240 	read_unlock_bh(&ping_table.lock);
241 
242 	return sk;
243 }
244 
245 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
246 					  kgid_t *high)
247 {
248 	kgid_t *data = net->ipv4.ping_group_range.range;
249 	unsigned int seq;
250 
251 	do {
252 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
253 
254 		*low = data[0];
255 		*high = data[1];
256 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
257 }
258 
259 
260 int ping_init_sock(struct sock *sk)
261 {
262 	struct net *net = sock_net(sk);
263 	kgid_t group = current_egid();
264 	struct group_info *group_info;
265 	int i;
266 	kgid_t low, high;
267 	int ret = 0;
268 
269 	if (sk->sk_family == AF_INET6)
270 		sk->sk_ipv6only = 1;
271 
272 	inet_get_ping_group_range_net(net, &low, &high);
273 	if (gid_lte(low, group) && gid_lte(group, high))
274 		return 0;
275 
276 	group_info = get_current_groups();
277 	for (i = 0; i < group_info->ngroups; i++) {
278 		kgid_t gid = group_info->gid[i];
279 
280 		if (gid_lte(low, gid) && gid_lte(gid, high))
281 			goto out_release_group;
282 	}
283 
284 	ret = -EACCES;
285 
286 out_release_group:
287 	put_group_info(group_info);
288 	return ret;
289 }
290 EXPORT_SYMBOL_GPL(ping_init_sock);
291 
292 void ping_close(struct sock *sk, long timeout)
293 {
294 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
295 		 inet_sk(sk), inet_sk(sk)->inet_num);
296 	pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
297 
298 	sk_common_release(sk);
299 }
300 EXPORT_SYMBOL_GPL(ping_close);
301 
302 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
303 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
304 				struct sockaddr *uaddr, int addr_len)
305 {
306 	struct net *net = sock_net(sk);
307 	if (sk->sk_family == AF_INET) {
308 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
309 		int chk_addr_ret;
310 
311 		if (addr_len < sizeof(*addr))
312 			return -EINVAL;
313 
314 		if (addr->sin_family != AF_INET &&
315 		    !(addr->sin_family == AF_UNSPEC &&
316 		      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
317 			return -EAFNOSUPPORT;
318 
319 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
320 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
321 
322 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
323 
324 		if (!inet_addr_valid_or_nonlocal(net, inet_sk(sk),
325 					         addr->sin_addr.s_addr,
326 	                                         chk_addr_ret))
327 			return -EADDRNOTAVAIL;
328 
329 #if IS_ENABLED(CONFIG_IPV6)
330 	} else if (sk->sk_family == AF_INET6) {
331 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
332 		int addr_type, scoped, has_addr;
333 		struct net_device *dev = NULL;
334 
335 		if (addr_len < sizeof(*addr))
336 			return -EINVAL;
337 
338 		if (addr->sin6_family != AF_INET6)
339 			return -EAFNOSUPPORT;
340 
341 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
342 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
343 
344 		addr_type = ipv6_addr_type(&addr->sin6_addr);
345 		scoped = __ipv6_addr_needs_scope_id(addr_type);
346 		if ((addr_type != IPV6_ADDR_ANY &&
347 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
348 		    (scoped && !addr->sin6_scope_id))
349 			return -EINVAL;
350 
351 		rcu_read_lock();
352 		if (addr->sin6_scope_id) {
353 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
354 			if (!dev) {
355 				rcu_read_unlock();
356 				return -ENODEV;
357 			}
358 		}
359 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
360 						    scoped);
361 		rcu_read_unlock();
362 
363 		if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
364 		      addr_type == IPV6_ADDR_ANY))
365 			return -EADDRNOTAVAIL;
366 
367 		if (scoped)
368 			sk->sk_bound_dev_if = addr->sin6_scope_id;
369 #endif
370 	} else {
371 		return -EAFNOSUPPORT;
372 	}
373 	return 0;
374 }
375 
376 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
377 {
378 	if (saddr->sa_family == AF_INET) {
379 		struct inet_sock *isk = inet_sk(sk);
380 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
381 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
382 #if IS_ENABLED(CONFIG_IPV6)
383 	} else if (saddr->sa_family == AF_INET6) {
384 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
385 		struct ipv6_pinfo *np = inet6_sk(sk);
386 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
387 #endif
388 	}
389 }
390 
391 /*
392  * We need our own bind because there are no privileged id's == local ports.
393  * Moreover, we don't allow binding to multi- and broadcast addresses.
394  */
395 
396 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
397 {
398 	struct inet_sock *isk = inet_sk(sk);
399 	unsigned short snum;
400 	int err;
401 	int dif = sk->sk_bound_dev_if;
402 
403 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
404 	if (err)
405 		return err;
406 
407 	lock_sock(sk);
408 
409 	err = -EINVAL;
410 	if (isk->inet_num != 0)
411 		goto out;
412 
413 	err = -EADDRINUSE;
414 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
415 	if (ping_get_port(sk, snum) != 0) {
416 		/* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
417 		sk->sk_bound_dev_if = dif;
418 		goto out;
419 	}
420 	ping_set_saddr(sk, uaddr);
421 
422 	pr_debug("after bind(): num = %hu, dif = %d\n",
423 		 isk->inet_num,
424 		 sk->sk_bound_dev_if);
425 
426 	err = 0;
427 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
428 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
429 #if IS_ENABLED(CONFIG_IPV6)
430 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
431 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
432 #endif
433 
434 	if (snum)
435 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
436 	isk->inet_sport = htons(isk->inet_num);
437 	isk->inet_daddr = 0;
438 	isk->inet_dport = 0;
439 
440 #if IS_ENABLED(CONFIG_IPV6)
441 	if (sk->sk_family == AF_INET6)
442 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
443 #endif
444 
445 	sk_dst_reset(sk);
446 out:
447 	release_sock(sk);
448 	pr_debug("ping_v4_bind -> %d\n", err);
449 	return err;
450 }
451 EXPORT_SYMBOL_GPL(ping_bind);
452 
453 /*
454  * Is this a supported type of ICMP message?
455  */
456 
457 static inline int ping_supported(int family, int type, int code)
458 {
459 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
460 	       (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
461 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
462 	       (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
463 }
464 
465 /*
466  * This routine is called by the ICMP module when it gets some
467  * sort of error condition.
468  */
469 
470 void ping_err(struct sk_buff *skb, int offset, u32 info)
471 {
472 	int family;
473 	struct icmphdr *icmph;
474 	struct inet_sock *inet_sock;
475 	int type;
476 	int code;
477 	struct net *net = dev_net(skb->dev);
478 	struct sock *sk;
479 	int harderr;
480 	int err;
481 
482 	if (skb->protocol == htons(ETH_P_IP)) {
483 		family = AF_INET;
484 		type = icmp_hdr(skb)->type;
485 		code = icmp_hdr(skb)->code;
486 		icmph = (struct icmphdr *)(skb->data + offset);
487 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
488 		family = AF_INET6;
489 		type = icmp6_hdr(skb)->icmp6_type;
490 		code = icmp6_hdr(skb)->icmp6_code;
491 		icmph = (struct icmphdr *) (skb->data + offset);
492 	} else {
493 		BUG();
494 	}
495 
496 	/* We assume the packet has already been checked by icmp_unreach */
497 
498 	if (!ping_supported(family, icmph->type, icmph->code))
499 		return;
500 
501 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
502 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
503 		 ntohs(icmph->un.echo.sequence));
504 
505 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
506 	if (!sk) {
507 		pr_debug("no socket, dropping\n");
508 		return;	/* No socket for error */
509 	}
510 	pr_debug("err on socket %p\n", sk);
511 
512 	err = 0;
513 	harderr = 0;
514 	inet_sock = inet_sk(sk);
515 
516 	if (skb->protocol == htons(ETH_P_IP)) {
517 		switch (type) {
518 		default:
519 		case ICMP_TIME_EXCEEDED:
520 			err = EHOSTUNREACH;
521 			break;
522 		case ICMP_SOURCE_QUENCH:
523 			/* This is not a real error but ping wants to see it.
524 			 * Report it with some fake errno.
525 			 */
526 			err = EREMOTEIO;
527 			break;
528 		case ICMP_PARAMETERPROB:
529 			err = EPROTO;
530 			harderr = 1;
531 			break;
532 		case ICMP_DEST_UNREACH:
533 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
534 				ipv4_sk_update_pmtu(skb, sk, info);
535 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
536 					err = EMSGSIZE;
537 					harderr = 1;
538 					break;
539 				}
540 				goto out;
541 			}
542 			err = EHOSTUNREACH;
543 			if (code <= NR_ICMP_UNREACH) {
544 				harderr = icmp_err_convert[code].fatal;
545 				err = icmp_err_convert[code].errno;
546 			}
547 			break;
548 		case ICMP_REDIRECT:
549 			/* See ICMP_SOURCE_QUENCH */
550 			ipv4_sk_redirect(skb, sk);
551 			err = EREMOTEIO;
552 			break;
553 		}
554 #if IS_ENABLED(CONFIG_IPV6)
555 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
556 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
557 #endif
558 	}
559 
560 	/*
561 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
562 	 *	4.1.3.3.
563 	 */
564 	if ((family == AF_INET && !inet_sock->recverr) ||
565 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
566 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
567 			goto out;
568 	} else {
569 		if (family == AF_INET) {
570 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
571 				      info, (u8 *)icmph);
572 #if IS_ENABLED(CONFIG_IPV6)
573 		} else if (family == AF_INET6) {
574 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
575 						   info, (u8 *)icmph);
576 #endif
577 		}
578 	}
579 	sk->sk_err = err;
580 	sk_error_report(sk);
581 out:
582 	sock_put(sk);
583 }
584 EXPORT_SYMBOL_GPL(ping_err);
585 
586 /*
587  *	Copy and checksum an ICMP Echo packet from user space into a buffer
588  *	starting from the payload.
589  */
590 
591 int ping_getfrag(void *from, char *to,
592 		 int offset, int fraglen, int odd, struct sk_buff *skb)
593 {
594 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
595 
596 	if (offset == 0) {
597 		fraglen -= sizeof(struct icmphdr);
598 		if (fraglen < 0)
599 			BUG();
600 		if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
601 			    fraglen, &pfh->wcheck,
602 			    &pfh->msg->msg_iter))
603 			return -EFAULT;
604 	} else if (offset < sizeof(struct icmphdr)) {
605 			BUG();
606 	} else {
607 		if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
608 					    &pfh->msg->msg_iter))
609 			return -EFAULT;
610 	}
611 
612 #if IS_ENABLED(CONFIG_IPV6)
613 	/* For IPv6, checksum each skb as we go along, as expected by
614 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
615 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
616 	 */
617 	if (pfh->family == AF_INET6) {
618 		skb->csum = pfh->wcheck;
619 		skb->ip_summed = CHECKSUM_NONE;
620 		pfh->wcheck = 0;
621 	}
622 #endif
623 
624 	return 0;
625 }
626 EXPORT_SYMBOL_GPL(ping_getfrag);
627 
628 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
629 				       struct flowi4 *fl4)
630 {
631 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
632 
633 	if (!skb)
634 		return 0;
635 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
636 		sizeof(struct icmphdr), pfh->wcheck);
637 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
638 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
639 	skb->ip_summed = CHECKSUM_NONE;
640 	return ip_push_pending_frames(sk, fl4);
641 }
642 
643 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
644 			void *user_icmph, size_t icmph_len)
645 {
646 	u8 type, code;
647 
648 	if (len > 0xFFFF)
649 		return -EMSGSIZE;
650 
651 	/* Must have at least a full ICMP header. */
652 	if (len < icmph_len)
653 		return -EINVAL;
654 
655 	/*
656 	 *	Check the flags.
657 	 */
658 
659 	/* Mirror BSD error message compatibility */
660 	if (msg->msg_flags & MSG_OOB)
661 		return -EOPNOTSUPP;
662 
663 	/*
664 	 *	Fetch the ICMP header provided by the userland.
665 	 *	iovec is modified! The ICMP header is consumed.
666 	 */
667 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
668 		return -EFAULT;
669 
670 	if (family == AF_INET) {
671 		type = ((struct icmphdr *) user_icmph)->type;
672 		code = ((struct icmphdr *) user_icmph)->code;
673 #if IS_ENABLED(CONFIG_IPV6)
674 	} else if (family == AF_INET6) {
675 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
676 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
677 #endif
678 	} else {
679 		BUG();
680 	}
681 
682 	if (!ping_supported(family, type, code))
683 		return -EINVAL;
684 
685 	return 0;
686 }
687 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
688 
689 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
690 {
691 	struct net *net = sock_net(sk);
692 	struct flowi4 fl4;
693 	struct inet_sock *inet = inet_sk(sk);
694 	struct ipcm_cookie ipc;
695 	struct icmphdr user_icmph;
696 	struct pingfakehdr pfh;
697 	struct rtable *rt = NULL;
698 	struct ip_options_data opt_copy;
699 	int free = 0;
700 	__be32 saddr, daddr, faddr;
701 	u8  tos;
702 	int err;
703 
704 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
705 
706 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
707 				  sizeof(user_icmph));
708 	if (err)
709 		return err;
710 
711 	/*
712 	 *	Get and verify the address.
713 	 */
714 
715 	if (msg->msg_name) {
716 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
717 		if (msg->msg_namelen < sizeof(*usin))
718 			return -EINVAL;
719 		if (usin->sin_family != AF_INET)
720 			return -EAFNOSUPPORT;
721 		daddr = usin->sin_addr.s_addr;
722 		/* no remote port */
723 	} else {
724 		if (sk->sk_state != TCP_ESTABLISHED)
725 			return -EDESTADDRREQ;
726 		daddr = inet->inet_daddr;
727 		/* no remote port */
728 	}
729 
730 	ipcm_init_sk(&ipc, inet);
731 
732 	if (msg->msg_controllen) {
733 		err = ip_cmsg_send(sk, msg, &ipc, false);
734 		if (unlikely(err)) {
735 			kfree(ipc.opt);
736 			return err;
737 		}
738 		if (ipc.opt)
739 			free = 1;
740 	}
741 	if (!ipc.opt) {
742 		struct ip_options_rcu *inet_opt;
743 
744 		rcu_read_lock();
745 		inet_opt = rcu_dereference(inet->inet_opt);
746 		if (inet_opt) {
747 			memcpy(&opt_copy, inet_opt,
748 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
749 			ipc.opt = &opt_copy.opt;
750 		}
751 		rcu_read_unlock();
752 	}
753 
754 	saddr = ipc.addr;
755 	ipc.addr = faddr = daddr;
756 
757 	if (ipc.opt && ipc.opt->opt.srr) {
758 		if (!daddr) {
759 			err = -EINVAL;
760 			goto out_free;
761 		}
762 		faddr = ipc.opt->opt.faddr;
763 	}
764 	tos = get_rttos(&ipc, inet);
765 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
766 	    (msg->msg_flags & MSG_DONTROUTE) ||
767 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
768 		tos |= RTO_ONLINK;
769 	}
770 
771 	if (ipv4_is_multicast(daddr)) {
772 		if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
773 			ipc.oif = inet->mc_index;
774 		if (!saddr)
775 			saddr = inet->mc_addr;
776 	} else if (!ipc.oif)
777 		ipc.oif = inet->uc_index;
778 
779 	flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
780 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
781 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
782 			   sk->sk_uid);
783 
784 	fl4.fl4_icmp_type = user_icmph.type;
785 	fl4.fl4_icmp_code = user_icmph.code;
786 
787 	security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
788 	rt = ip_route_output_flow(net, &fl4, sk);
789 	if (IS_ERR(rt)) {
790 		err = PTR_ERR(rt);
791 		rt = NULL;
792 		if (err == -ENETUNREACH)
793 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
794 		goto out;
795 	}
796 
797 	err = -EACCES;
798 	if ((rt->rt_flags & RTCF_BROADCAST) &&
799 	    !sock_flag(sk, SOCK_BROADCAST))
800 		goto out;
801 
802 	if (msg->msg_flags & MSG_CONFIRM)
803 		goto do_confirm;
804 back_from_confirm:
805 
806 	if (!ipc.addr)
807 		ipc.addr = fl4.daddr;
808 
809 	lock_sock(sk);
810 
811 	pfh.icmph.type = user_icmph.type; /* already checked */
812 	pfh.icmph.code = user_icmph.code; /* ditto */
813 	pfh.icmph.checksum = 0;
814 	pfh.icmph.un.echo.id = inet->inet_sport;
815 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
816 	pfh.msg = msg;
817 	pfh.wcheck = 0;
818 	pfh.family = AF_INET;
819 
820 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
821 			0, &ipc, &rt, msg->msg_flags);
822 	if (err)
823 		ip_flush_pending_frames(sk);
824 	else
825 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
826 	release_sock(sk);
827 
828 out:
829 	ip_rt_put(rt);
830 out_free:
831 	if (free)
832 		kfree(ipc.opt);
833 	if (!err) {
834 		icmp_out_count(sock_net(sk), user_icmph.type);
835 		return len;
836 	}
837 	return err;
838 
839 do_confirm:
840 	if (msg->msg_flags & MSG_PROBE)
841 		dst_confirm_neigh(&rt->dst, &fl4.daddr);
842 	if (!(msg->msg_flags & MSG_PROBE) || len)
843 		goto back_from_confirm;
844 	err = 0;
845 	goto out;
846 }
847 
848 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
849 		 int flags, int *addr_len)
850 {
851 	struct inet_sock *isk = inet_sk(sk);
852 	int family = sk->sk_family;
853 	struct sk_buff *skb;
854 	int copied, err;
855 
856 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
857 
858 	err = -EOPNOTSUPP;
859 	if (flags & MSG_OOB)
860 		goto out;
861 
862 	if (flags & MSG_ERRQUEUE)
863 		return inet_recv_error(sk, msg, len, addr_len);
864 
865 	skb = skb_recv_datagram(sk, flags, noblock, &err);
866 	if (!skb)
867 		goto out;
868 
869 	copied = skb->len;
870 	if (copied > len) {
871 		msg->msg_flags |= MSG_TRUNC;
872 		copied = len;
873 	}
874 
875 	/* Don't bother checking the checksum */
876 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
877 	if (err)
878 		goto done;
879 
880 	sock_recv_timestamp(msg, sk, skb);
881 
882 	/* Copy the address and add cmsg data. */
883 	if (family == AF_INET) {
884 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
885 
886 		if (sin) {
887 			sin->sin_family = AF_INET;
888 			sin->sin_port = 0 /* skb->h.uh->source */;
889 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
890 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
891 			*addr_len = sizeof(*sin);
892 		}
893 
894 		if (isk->cmsg_flags)
895 			ip_cmsg_recv(msg, skb);
896 
897 #if IS_ENABLED(CONFIG_IPV6)
898 	} else if (family == AF_INET6) {
899 		struct ipv6_pinfo *np = inet6_sk(sk);
900 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
901 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
902 
903 		if (sin6) {
904 			sin6->sin6_family = AF_INET6;
905 			sin6->sin6_port = 0;
906 			sin6->sin6_addr = ip6->saddr;
907 			sin6->sin6_flowinfo = 0;
908 			if (np->sndflow)
909 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
910 			sin6->sin6_scope_id =
911 				ipv6_iface_scope_id(&sin6->sin6_addr,
912 						    inet6_iif(skb));
913 			*addr_len = sizeof(*sin6);
914 		}
915 
916 		if (inet6_sk(sk)->rxopt.all)
917 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
918 		if (skb->protocol == htons(ETH_P_IPV6) &&
919 		    inet6_sk(sk)->rxopt.all)
920 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
921 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
922 			ip_cmsg_recv(msg, skb);
923 #endif
924 	} else {
925 		BUG();
926 	}
927 
928 	err = copied;
929 
930 done:
931 	skb_free_datagram(sk, skb);
932 out:
933 	pr_debug("ping_recvmsg -> %d\n", err);
934 	return err;
935 }
936 EXPORT_SYMBOL_GPL(ping_recvmsg);
937 
938 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
939 {
940 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
941 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
942 	if (sock_queue_rcv_skb(sk, skb) < 0) {
943 		kfree_skb(skb);
944 		pr_debug("ping_queue_rcv_skb -> failed\n");
945 		return -1;
946 	}
947 	return 0;
948 }
949 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
950 
951 
952 /*
953  *	All we need to do is get the socket.
954  */
955 
956 bool ping_rcv(struct sk_buff *skb)
957 {
958 	struct sock *sk;
959 	struct net *net = dev_net(skb->dev);
960 	struct icmphdr *icmph = icmp_hdr(skb);
961 	bool rc = false;
962 
963 	/* We assume the packet has already been checked by icmp_rcv */
964 
965 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
966 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
967 
968 	/* Push ICMP header back */
969 	skb_push(skb, skb->data - (u8 *)icmph);
970 
971 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
972 	if (sk) {
973 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
974 
975 		pr_debug("rcv on socket %p\n", sk);
976 		if (skb2 && !ping_queue_rcv_skb(sk, skb2))
977 			rc = true;
978 		sock_put(sk);
979 	}
980 
981 	if (!rc)
982 		pr_debug("no socket, dropping\n");
983 
984 	return rc;
985 }
986 EXPORT_SYMBOL_GPL(ping_rcv);
987 
988 struct proto ping_prot = {
989 	.name =		"PING",
990 	.owner =	THIS_MODULE,
991 	.init =		ping_init_sock,
992 	.close =	ping_close,
993 	.connect =	ip4_datagram_connect,
994 	.disconnect =	__udp_disconnect,
995 	.setsockopt =	ip_setsockopt,
996 	.getsockopt =	ip_getsockopt,
997 	.sendmsg =	ping_v4_sendmsg,
998 	.recvmsg =	ping_recvmsg,
999 	.bind =		ping_bind,
1000 	.backlog_rcv =	ping_queue_rcv_skb,
1001 	.release_cb =	ip4_datagram_release_cb,
1002 	.hash =		ping_hash,
1003 	.unhash =	ping_unhash,
1004 	.get_port =	ping_get_port,
1005 	.put_port =	ping_unhash,
1006 	.obj_size =	sizeof(struct inet_sock),
1007 };
1008 EXPORT_SYMBOL(ping_prot);
1009 
1010 #ifdef CONFIG_PROC_FS
1011 
1012 static struct sock *ping_get_first(struct seq_file *seq, int start)
1013 {
1014 	struct sock *sk;
1015 	struct ping_iter_state *state = seq->private;
1016 	struct net *net = seq_file_net(seq);
1017 
1018 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1019 	     ++state->bucket) {
1020 		struct hlist_nulls_node *node;
1021 		struct hlist_nulls_head *hslot;
1022 
1023 		hslot = &ping_table.hash[state->bucket];
1024 
1025 		if (hlist_nulls_empty(hslot))
1026 			continue;
1027 
1028 		sk_nulls_for_each(sk, node, hslot) {
1029 			if (net_eq(sock_net(sk), net) &&
1030 			    sk->sk_family == state->family)
1031 				goto found;
1032 		}
1033 	}
1034 	sk = NULL;
1035 found:
1036 	return sk;
1037 }
1038 
1039 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1040 {
1041 	struct ping_iter_state *state = seq->private;
1042 	struct net *net = seq_file_net(seq);
1043 
1044 	do {
1045 		sk = sk_nulls_next(sk);
1046 	} while (sk && (!net_eq(sock_net(sk), net)));
1047 
1048 	if (!sk)
1049 		return ping_get_first(seq, state->bucket + 1);
1050 	return sk;
1051 }
1052 
1053 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1054 {
1055 	struct sock *sk = ping_get_first(seq, 0);
1056 
1057 	if (sk)
1058 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1059 			--pos;
1060 	return pos ? NULL : sk;
1061 }
1062 
1063 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1064 	__acquires(ping_table.lock)
1065 {
1066 	struct ping_iter_state *state = seq->private;
1067 	state->bucket = 0;
1068 	state->family = family;
1069 
1070 	read_lock_bh(&ping_table.lock);
1071 
1072 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1073 }
1074 EXPORT_SYMBOL_GPL(ping_seq_start);
1075 
1076 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1077 {
1078 	return ping_seq_start(seq, pos, AF_INET);
1079 }
1080 
1081 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1082 {
1083 	struct sock *sk;
1084 
1085 	if (v == SEQ_START_TOKEN)
1086 		sk = ping_get_idx(seq, 0);
1087 	else
1088 		sk = ping_get_next(seq, v);
1089 
1090 	++*pos;
1091 	return sk;
1092 }
1093 EXPORT_SYMBOL_GPL(ping_seq_next);
1094 
1095 void ping_seq_stop(struct seq_file *seq, void *v)
1096 	__releases(ping_table.lock)
1097 {
1098 	read_unlock_bh(&ping_table.lock);
1099 }
1100 EXPORT_SYMBOL_GPL(ping_seq_stop);
1101 
1102 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1103 		int bucket)
1104 {
1105 	struct inet_sock *inet = inet_sk(sp);
1106 	__be32 dest = inet->inet_daddr;
1107 	__be32 src = inet->inet_rcv_saddr;
1108 	__u16 destp = ntohs(inet->inet_dport);
1109 	__u16 srcp = ntohs(inet->inet_sport);
1110 
1111 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1112 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1113 		bucket, src, srcp, dest, destp, sp->sk_state,
1114 		sk_wmem_alloc_get(sp),
1115 		sk_rmem_alloc_get(sp),
1116 		0, 0L, 0,
1117 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1118 		0, sock_i_ino(sp),
1119 		refcount_read(&sp->sk_refcnt), sp,
1120 		atomic_read(&sp->sk_drops));
1121 }
1122 
1123 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1124 {
1125 	seq_setwidth(seq, 127);
1126 	if (v == SEQ_START_TOKEN)
1127 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1128 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1129 			   "inode ref pointer drops");
1130 	else {
1131 		struct ping_iter_state *state = seq->private;
1132 
1133 		ping_v4_format_sock(v, seq, state->bucket);
1134 	}
1135 	seq_pad(seq, '\n');
1136 	return 0;
1137 }
1138 
1139 static const struct seq_operations ping_v4_seq_ops = {
1140 	.start		= ping_v4_seq_start,
1141 	.show		= ping_v4_seq_show,
1142 	.next		= ping_seq_next,
1143 	.stop		= ping_seq_stop,
1144 };
1145 
1146 static int __net_init ping_v4_proc_init_net(struct net *net)
1147 {
1148 	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1149 			sizeof(struct ping_iter_state)))
1150 		return -ENOMEM;
1151 	return 0;
1152 }
1153 
1154 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1155 {
1156 	remove_proc_entry("icmp", net->proc_net);
1157 }
1158 
1159 static struct pernet_operations ping_v4_net_ops = {
1160 	.init = ping_v4_proc_init_net,
1161 	.exit = ping_v4_proc_exit_net,
1162 };
1163 
1164 int __init ping_proc_init(void)
1165 {
1166 	return register_pernet_subsys(&ping_v4_net_ops);
1167 }
1168 
1169 void ping_proc_exit(void)
1170 {
1171 	unregister_pernet_subsys(&ping_v4_net_ops);
1172 }
1173 
1174 #endif
1175 
1176 void __init ping_init(void)
1177 {
1178 	int i;
1179 
1180 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1181 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1182 	rwlock_init(&ping_table.lock);
1183 }
1184