xref: /openbmc/linux/net/ipv4/ping.c (revision 1c0bd035)
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 = skb->dev->ifindex;
176 
177 	if (skb->protocol == htons(ETH_P_IP)) {
178 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
179 			 (int)ident, &ip_hdr(skb)->daddr, dif);
180 #if IS_ENABLED(CONFIG_IPV6)
181 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
182 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
183 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
184 #endif
185 	}
186 
187 	read_lock_bh(&ping_table.lock);
188 
189 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
190 		isk = inet_sk(sk);
191 
192 		pr_debug("iterate\n");
193 		if (isk->inet_num != ident)
194 			continue;
195 
196 		if (skb->protocol == htons(ETH_P_IP) &&
197 		    sk->sk_family == AF_INET) {
198 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
199 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
200 				 sk->sk_bound_dev_if);
201 
202 			if (isk->inet_rcv_saddr &&
203 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
204 				continue;
205 #if IS_ENABLED(CONFIG_IPV6)
206 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
207 			   sk->sk_family == AF_INET6) {
208 
209 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
210 				 (int) isk->inet_num,
211 				 &sk->sk_v6_rcv_saddr,
212 				 sk->sk_bound_dev_if);
213 
214 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
215 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
216 					     &ipv6_hdr(skb)->daddr))
217 				continue;
218 #endif
219 		} else {
220 			continue;
221 		}
222 
223 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
224 			continue;
225 
226 		sock_hold(sk);
227 		goto exit;
228 	}
229 
230 	sk = NULL;
231 exit:
232 	read_unlock_bh(&ping_table.lock);
233 
234 	return sk;
235 }
236 
237 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
238 					  kgid_t *high)
239 {
240 	kgid_t *data = net->ipv4.ping_group_range.range;
241 	unsigned int seq;
242 
243 	do {
244 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
245 
246 		*low = data[0];
247 		*high = data[1];
248 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
249 }
250 
251 
252 int ping_init_sock(struct sock *sk)
253 {
254 	struct net *net = sock_net(sk);
255 	kgid_t group = current_egid();
256 	struct group_info *group_info;
257 	int i;
258 	kgid_t low, high;
259 	int ret = 0;
260 
261 	if (sk->sk_family == AF_INET6)
262 		sk->sk_ipv6only = 1;
263 
264 	inet_get_ping_group_range_net(net, &low, &high);
265 	if (gid_lte(low, group) && gid_lte(group, high))
266 		return 0;
267 
268 	group_info = get_current_groups();
269 	for (i = 0; i < group_info->ngroups; i++) {
270 		kgid_t gid = group_info->gid[i];
271 
272 		if (gid_lte(low, gid) && gid_lte(gid, high))
273 			goto out_release_group;
274 	}
275 
276 	ret = -EACCES;
277 
278 out_release_group:
279 	put_group_info(group_info);
280 	return ret;
281 }
282 EXPORT_SYMBOL_GPL(ping_init_sock);
283 
284 void ping_close(struct sock *sk, long timeout)
285 {
286 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
287 		 inet_sk(sk), inet_sk(sk)->inet_num);
288 	pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
289 
290 	sk_common_release(sk);
291 }
292 EXPORT_SYMBOL_GPL(ping_close);
293 
294 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
295 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
296 				struct sockaddr *uaddr, int addr_len)
297 {
298 	struct net *net = sock_net(sk);
299 	if (sk->sk_family == AF_INET) {
300 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
301 		int chk_addr_ret;
302 
303 		if (addr_len < sizeof(*addr))
304 			return -EINVAL;
305 
306 		if (addr->sin_family != AF_INET &&
307 		    !(addr->sin_family == AF_UNSPEC &&
308 		      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
309 			return -EAFNOSUPPORT;
310 
311 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
312 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
313 
314 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
315 
316 		if (!inet_addr_valid_or_nonlocal(net, inet_sk(sk),
317 					         addr->sin_addr.s_addr,
318 	                                         chk_addr_ret))
319 			return -EADDRNOTAVAIL;
320 
321 #if IS_ENABLED(CONFIG_IPV6)
322 	} else if (sk->sk_family == AF_INET6) {
323 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
324 		int addr_type, scoped, has_addr;
325 		struct net_device *dev = NULL;
326 
327 		if (addr_len < sizeof(*addr))
328 			return -EINVAL;
329 
330 		if (addr->sin6_family != AF_INET6)
331 			return -EAFNOSUPPORT;
332 
333 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
334 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
335 
336 		addr_type = ipv6_addr_type(&addr->sin6_addr);
337 		scoped = __ipv6_addr_needs_scope_id(addr_type);
338 		if ((addr_type != IPV6_ADDR_ANY &&
339 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
340 		    (scoped && !addr->sin6_scope_id))
341 			return -EINVAL;
342 
343 		rcu_read_lock();
344 		if (addr->sin6_scope_id) {
345 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
346 			if (!dev) {
347 				rcu_read_unlock();
348 				return -ENODEV;
349 			}
350 		}
351 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
352 						    scoped);
353 		rcu_read_unlock();
354 
355 		if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
356 		      addr_type == IPV6_ADDR_ANY))
357 			return -EADDRNOTAVAIL;
358 
359 		if (scoped)
360 			sk->sk_bound_dev_if = addr->sin6_scope_id;
361 #endif
362 	} else {
363 		return -EAFNOSUPPORT;
364 	}
365 	return 0;
366 }
367 
368 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
369 {
370 	if (saddr->sa_family == AF_INET) {
371 		struct inet_sock *isk = inet_sk(sk);
372 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
373 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
374 #if IS_ENABLED(CONFIG_IPV6)
375 	} else if (saddr->sa_family == AF_INET6) {
376 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
377 		struct ipv6_pinfo *np = inet6_sk(sk);
378 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
379 #endif
380 	}
381 }
382 
383 /*
384  * We need our own bind because there are no privileged id's == local ports.
385  * Moreover, we don't allow binding to multi- and broadcast addresses.
386  */
387 
388 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
389 {
390 	struct inet_sock *isk = inet_sk(sk);
391 	unsigned short snum;
392 	int err;
393 	int dif = sk->sk_bound_dev_if;
394 
395 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
396 	if (err)
397 		return err;
398 
399 	lock_sock(sk);
400 
401 	err = -EINVAL;
402 	if (isk->inet_num != 0)
403 		goto out;
404 
405 	err = -EADDRINUSE;
406 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
407 	if (ping_get_port(sk, snum) != 0) {
408 		/* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
409 		sk->sk_bound_dev_if = dif;
410 		goto out;
411 	}
412 	ping_set_saddr(sk, uaddr);
413 
414 	pr_debug("after bind(): num = %hu, dif = %d\n",
415 		 isk->inet_num,
416 		 sk->sk_bound_dev_if);
417 
418 	err = 0;
419 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
420 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
421 #if IS_ENABLED(CONFIG_IPV6)
422 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
423 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
424 #endif
425 
426 	if (snum)
427 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
428 	isk->inet_sport = htons(isk->inet_num);
429 	isk->inet_daddr = 0;
430 	isk->inet_dport = 0;
431 
432 #if IS_ENABLED(CONFIG_IPV6)
433 	if (sk->sk_family == AF_INET6)
434 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
435 #endif
436 
437 	sk_dst_reset(sk);
438 out:
439 	release_sock(sk);
440 	pr_debug("ping_v4_bind -> %d\n", err);
441 	return err;
442 }
443 EXPORT_SYMBOL_GPL(ping_bind);
444 
445 /*
446  * Is this a supported type of ICMP message?
447  */
448 
449 static inline int ping_supported(int family, int type, int code)
450 {
451 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
452 	       (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
453 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
454 	       (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
455 }
456 
457 /*
458  * This routine is called by the ICMP module when it gets some
459  * sort of error condition.
460  */
461 
462 void ping_err(struct sk_buff *skb, int offset, u32 info)
463 {
464 	int family;
465 	struct icmphdr *icmph;
466 	struct inet_sock *inet_sock;
467 	int type;
468 	int code;
469 	struct net *net = dev_net(skb->dev);
470 	struct sock *sk;
471 	int harderr;
472 	int err;
473 
474 	if (skb->protocol == htons(ETH_P_IP)) {
475 		family = AF_INET;
476 		type = icmp_hdr(skb)->type;
477 		code = icmp_hdr(skb)->code;
478 		icmph = (struct icmphdr *)(skb->data + offset);
479 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
480 		family = AF_INET6;
481 		type = icmp6_hdr(skb)->icmp6_type;
482 		code = icmp6_hdr(skb)->icmp6_code;
483 		icmph = (struct icmphdr *) (skb->data + offset);
484 	} else {
485 		BUG();
486 	}
487 
488 	/* We assume the packet has already been checked by icmp_unreach */
489 
490 	if (!ping_supported(family, icmph->type, icmph->code))
491 		return;
492 
493 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
494 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
495 		 ntohs(icmph->un.echo.sequence));
496 
497 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
498 	if (!sk) {
499 		pr_debug("no socket, dropping\n");
500 		return;	/* No socket for error */
501 	}
502 	pr_debug("err on socket %p\n", sk);
503 
504 	err = 0;
505 	harderr = 0;
506 	inet_sock = inet_sk(sk);
507 
508 	if (skb->protocol == htons(ETH_P_IP)) {
509 		switch (type) {
510 		default:
511 		case ICMP_TIME_EXCEEDED:
512 			err = EHOSTUNREACH;
513 			break;
514 		case ICMP_SOURCE_QUENCH:
515 			/* This is not a real error but ping wants to see it.
516 			 * Report it with some fake errno.
517 			 */
518 			err = EREMOTEIO;
519 			break;
520 		case ICMP_PARAMETERPROB:
521 			err = EPROTO;
522 			harderr = 1;
523 			break;
524 		case ICMP_DEST_UNREACH:
525 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
526 				ipv4_sk_update_pmtu(skb, sk, info);
527 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
528 					err = EMSGSIZE;
529 					harderr = 1;
530 					break;
531 				}
532 				goto out;
533 			}
534 			err = EHOSTUNREACH;
535 			if (code <= NR_ICMP_UNREACH) {
536 				harderr = icmp_err_convert[code].fatal;
537 				err = icmp_err_convert[code].errno;
538 			}
539 			break;
540 		case ICMP_REDIRECT:
541 			/* See ICMP_SOURCE_QUENCH */
542 			ipv4_sk_redirect(skb, sk);
543 			err = EREMOTEIO;
544 			break;
545 		}
546 #if IS_ENABLED(CONFIG_IPV6)
547 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
548 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
549 #endif
550 	}
551 
552 	/*
553 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
554 	 *	4.1.3.3.
555 	 */
556 	if ((family == AF_INET && !inet_sock->recverr) ||
557 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
558 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
559 			goto out;
560 	} else {
561 		if (family == AF_INET) {
562 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
563 				      info, (u8 *)icmph);
564 #if IS_ENABLED(CONFIG_IPV6)
565 		} else if (family == AF_INET6) {
566 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
567 						   info, (u8 *)icmph);
568 #endif
569 		}
570 	}
571 	sk->sk_err = err;
572 	sk_error_report(sk);
573 out:
574 	sock_put(sk);
575 }
576 EXPORT_SYMBOL_GPL(ping_err);
577 
578 /*
579  *	Copy and checksum an ICMP Echo packet from user space into a buffer
580  *	starting from the payload.
581  */
582 
583 int ping_getfrag(void *from, char *to,
584 		 int offset, int fraglen, int odd, struct sk_buff *skb)
585 {
586 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
587 
588 	if (offset == 0) {
589 		fraglen -= sizeof(struct icmphdr);
590 		if (fraglen < 0)
591 			BUG();
592 		if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
593 			    fraglen, &pfh->wcheck,
594 			    &pfh->msg->msg_iter))
595 			return -EFAULT;
596 	} else if (offset < sizeof(struct icmphdr)) {
597 			BUG();
598 	} else {
599 		if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
600 					    &pfh->msg->msg_iter))
601 			return -EFAULT;
602 	}
603 
604 #if IS_ENABLED(CONFIG_IPV6)
605 	/* For IPv6, checksum each skb as we go along, as expected by
606 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
607 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
608 	 */
609 	if (pfh->family == AF_INET6) {
610 		skb->csum = pfh->wcheck;
611 		skb->ip_summed = CHECKSUM_NONE;
612 		pfh->wcheck = 0;
613 	}
614 #endif
615 
616 	return 0;
617 }
618 EXPORT_SYMBOL_GPL(ping_getfrag);
619 
620 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
621 				       struct flowi4 *fl4)
622 {
623 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
624 
625 	if (!skb)
626 		return 0;
627 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
628 		sizeof(struct icmphdr), pfh->wcheck);
629 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
630 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
631 	skb->ip_summed = CHECKSUM_NONE;
632 	return ip_push_pending_frames(sk, fl4);
633 }
634 
635 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
636 			void *user_icmph, size_t icmph_len)
637 {
638 	u8 type, code;
639 
640 	if (len > 0xFFFF)
641 		return -EMSGSIZE;
642 
643 	/* Must have at least a full ICMP header. */
644 	if (len < icmph_len)
645 		return -EINVAL;
646 
647 	/*
648 	 *	Check the flags.
649 	 */
650 
651 	/* Mirror BSD error message compatibility */
652 	if (msg->msg_flags & MSG_OOB)
653 		return -EOPNOTSUPP;
654 
655 	/*
656 	 *	Fetch the ICMP header provided by the userland.
657 	 *	iovec is modified! The ICMP header is consumed.
658 	 */
659 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
660 		return -EFAULT;
661 
662 	if (family == AF_INET) {
663 		type = ((struct icmphdr *) user_icmph)->type;
664 		code = ((struct icmphdr *) user_icmph)->code;
665 #if IS_ENABLED(CONFIG_IPV6)
666 	} else if (family == AF_INET6) {
667 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
668 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
669 #endif
670 	} else {
671 		BUG();
672 	}
673 
674 	if (!ping_supported(family, type, code))
675 		return -EINVAL;
676 
677 	return 0;
678 }
679 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
680 
681 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
682 {
683 	struct net *net = sock_net(sk);
684 	struct flowi4 fl4;
685 	struct inet_sock *inet = inet_sk(sk);
686 	struct ipcm_cookie ipc;
687 	struct icmphdr user_icmph;
688 	struct pingfakehdr pfh;
689 	struct rtable *rt = NULL;
690 	struct ip_options_data opt_copy;
691 	int free = 0;
692 	__be32 saddr, daddr, faddr;
693 	u8  tos;
694 	int err;
695 
696 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
697 
698 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
699 				  sizeof(user_icmph));
700 	if (err)
701 		return err;
702 
703 	/*
704 	 *	Get and verify the address.
705 	 */
706 
707 	if (msg->msg_name) {
708 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
709 		if (msg->msg_namelen < sizeof(*usin))
710 			return -EINVAL;
711 		if (usin->sin_family != AF_INET)
712 			return -EAFNOSUPPORT;
713 		daddr = usin->sin_addr.s_addr;
714 		/* no remote port */
715 	} else {
716 		if (sk->sk_state != TCP_ESTABLISHED)
717 			return -EDESTADDRREQ;
718 		daddr = inet->inet_daddr;
719 		/* no remote port */
720 	}
721 
722 	ipcm_init_sk(&ipc, inet);
723 
724 	if (msg->msg_controllen) {
725 		err = ip_cmsg_send(sk, msg, &ipc, false);
726 		if (unlikely(err)) {
727 			kfree(ipc.opt);
728 			return err;
729 		}
730 		if (ipc.opt)
731 			free = 1;
732 	}
733 	if (!ipc.opt) {
734 		struct ip_options_rcu *inet_opt;
735 
736 		rcu_read_lock();
737 		inet_opt = rcu_dereference(inet->inet_opt);
738 		if (inet_opt) {
739 			memcpy(&opt_copy, inet_opt,
740 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
741 			ipc.opt = &opt_copy.opt;
742 		}
743 		rcu_read_unlock();
744 	}
745 
746 	saddr = ipc.addr;
747 	ipc.addr = faddr = daddr;
748 
749 	if (ipc.opt && ipc.opt->opt.srr) {
750 		if (!daddr) {
751 			err = -EINVAL;
752 			goto out_free;
753 		}
754 		faddr = ipc.opt->opt.faddr;
755 	}
756 	tos = get_rttos(&ipc, inet);
757 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
758 	    (msg->msg_flags & MSG_DONTROUTE) ||
759 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
760 		tos |= RTO_ONLINK;
761 	}
762 
763 	if (ipv4_is_multicast(daddr)) {
764 		if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
765 			ipc.oif = inet->mc_index;
766 		if (!saddr)
767 			saddr = inet->mc_addr;
768 	} else if (!ipc.oif)
769 		ipc.oif = inet->uc_index;
770 
771 	flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
772 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
773 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
774 			   sk->sk_uid);
775 
776 	fl4.fl4_icmp_type = user_icmph.type;
777 	fl4.fl4_icmp_code = user_icmph.code;
778 
779 	security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
780 	rt = ip_route_output_flow(net, &fl4, sk);
781 	if (IS_ERR(rt)) {
782 		err = PTR_ERR(rt);
783 		rt = NULL;
784 		if (err == -ENETUNREACH)
785 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
786 		goto out;
787 	}
788 
789 	err = -EACCES;
790 	if ((rt->rt_flags & RTCF_BROADCAST) &&
791 	    !sock_flag(sk, SOCK_BROADCAST))
792 		goto out;
793 
794 	if (msg->msg_flags & MSG_CONFIRM)
795 		goto do_confirm;
796 back_from_confirm:
797 
798 	if (!ipc.addr)
799 		ipc.addr = fl4.daddr;
800 
801 	lock_sock(sk);
802 
803 	pfh.icmph.type = user_icmph.type; /* already checked */
804 	pfh.icmph.code = user_icmph.code; /* ditto */
805 	pfh.icmph.checksum = 0;
806 	pfh.icmph.un.echo.id = inet->inet_sport;
807 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
808 	pfh.msg = msg;
809 	pfh.wcheck = 0;
810 	pfh.family = AF_INET;
811 
812 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
813 			0, &ipc, &rt, msg->msg_flags);
814 	if (err)
815 		ip_flush_pending_frames(sk);
816 	else
817 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
818 	release_sock(sk);
819 
820 out:
821 	ip_rt_put(rt);
822 out_free:
823 	if (free)
824 		kfree(ipc.opt);
825 	if (!err) {
826 		icmp_out_count(sock_net(sk), user_icmph.type);
827 		return len;
828 	}
829 	return err;
830 
831 do_confirm:
832 	if (msg->msg_flags & MSG_PROBE)
833 		dst_confirm_neigh(&rt->dst, &fl4.daddr);
834 	if (!(msg->msg_flags & MSG_PROBE) || len)
835 		goto back_from_confirm;
836 	err = 0;
837 	goto out;
838 }
839 
840 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
841 		 int flags, int *addr_len)
842 {
843 	struct inet_sock *isk = inet_sk(sk);
844 	int family = sk->sk_family;
845 	struct sk_buff *skb;
846 	int copied, err;
847 
848 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
849 
850 	err = -EOPNOTSUPP;
851 	if (flags & MSG_OOB)
852 		goto out;
853 
854 	if (flags & MSG_ERRQUEUE)
855 		return inet_recv_error(sk, msg, len, addr_len);
856 
857 	skb = skb_recv_datagram(sk, flags, noblock, &err);
858 	if (!skb)
859 		goto out;
860 
861 	copied = skb->len;
862 	if (copied > len) {
863 		msg->msg_flags |= MSG_TRUNC;
864 		copied = len;
865 	}
866 
867 	/* Don't bother checking the checksum */
868 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
869 	if (err)
870 		goto done;
871 
872 	sock_recv_timestamp(msg, sk, skb);
873 
874 	/* Copy the address and add cmsg data. */
875 	if (family == AF_INET) {
876 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
877 
878 		if (sin) {
879 			sin->sin_family = AF_INET;
880 			sin->sin_port = 0 /* skb->h.uh->source */;
881 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
882 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
883 			*addr_len = sizeof(*sin);
884 		}
885 
886 		if (isk->cmsg_flags)
887 			ip_cmsg_recv(msg, skb);
888 
889 #if IS_ENABLED(CONFIG_IPV6)
890 	} else if (family == AF_INET6) {
891 		struct ipv6_pinfo *np = inet6_sk(sk);
892 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
893 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
894 
895 		if (sin6) {
896 			sin6->sin6_family = AF_INET6;
897 			sin6->sin6_port = 0;
898 			sin6->sin6_addr = ip6->saddr;
899 			sin6->sin6_flowinfo = 0;
900 			if (np->sndflow)
901 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
902 			sin6->sin6_scope_id =
903 				ipv6_iface_scope_id(&sin6->sin6_addr,
904 						    inet6_iif(skb));
905 			*addr_len = sizeof(*sin6);
906 		}
907 
908 		if (inet6_sk(sk)->rxopt.all)
909 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
910 		if (skb->protocol == htons(ETH_P_IPV6) &&
911 		    inet6_sk(sk)->rxopt.all)
912 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
913 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
914 			ip_cmsg_recv(msg, skb);
915 #endif
916 	} else {
917 		BUG();
918 	}
919 
920 	err = copied;
921 
922 done:
923 	skb_free_datagram(sk, skb);
924 out:
925 	pr_debug("ping_recvmsg -> %d\n", err);
926 	return err;
927 }
928 EXPORT_SYMBOL_GPL(ping_recvmsg);
929 
930 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
931 {
932 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
933 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
934 	if (sock_queue_rcv_skb(sk, skb) < 0) {
935 		kfree_skb(skb);
936 		pr_debug("ping_queue_rcv_skb -> failed\n");
937 		return -1;
938 	}
939 	return 0;
940 }
941 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
942 
943 
944 /*
945  *	All we need to do is get the socket.
946  */
947 
948 bool ping_rcv(struct sk_buff *skb)
949 {
950 	struct sock *sk;
951 	struct net *net = dev_net(skb->dev);
952 	struct icmphdr *icmph = icmp_hdr(skb);
953 	bool rc = false;
954 
955 	/* We assume the packet has already been checked by icmp_rcv */
956 
957 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
958 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
959 
960 	/* Push ICMP header back */
961 	skb_push(skb, skb->data - (u8 *)icmph);
962 
963 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
964 	if (sk) {
965 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
966 
967 		pr_debug("rcv on socket %p\n", sk);
968 		if (skb2 && !ping_queue_rcv_skb(sk, skb2))
969 			rc = true;
970 		sock_put(sk);
971 	}
972 
973 	if (!rc)
974 		pr_debug("no socket, dropping\n");
975 
976 	return rc;
977 }
978 EXPORT_SYMBOL_GPL(ping_rcv);
979 
980 struct proto ping_prot = {
981 	.name =		"PING",
982 	.owner =	THIS_MODULE,
983 	.init =		ping_init_sock,
984 	.close =	ping_close,
985 	.connect =	ip4_datagram_connect,
986 	.disconnect =	__udp_disconnect,
987 	.setsockopt =	ip_setsockopt,
988 	.getsockopt =	ip_getsockopt,
989 	.sendmsg =	ping_v4_sendmsg,
990 	.recvmsg =	ping_recvmsg,
991 	.bind =		ping_bind,
992 	.backlog_rcv =	ping_queue_rcv_skb,
993 	.release_cb =	ip4_datagram_release_cb,
994 	.hash =		ping_hash,
995 	.unhash =	ping_unhash,
996 	.get_port =	ping_get_port,
997 	.put_port =	ping_unhash,
998 	.obj_size =	sizeof(struct inet_sock),
999 };
1000 EXPORT_SYMBOL(ping_prot);
1001 
1002 #ifdef CONFIG_PROC_FS
1003 
1004 static struct sock *ping_get_first(struct seq_file *seq, int start)
1005 {
1006 	struct sock *sk;
1007 	struct ping_iter_state *state = seq->private;
1008 	struct net *net = seq_file_net(seq);
1009 
1010 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1011 	     ++state->bucket) {
1012 		struct hlist_nulls_node *node;
1013 		struct hlist_nulls_head *hslot;
1014 
1015 		hslot = &ping_table.hash[state->bucket];
1016 
1017 		if (hlist_nulls_empty(hslot))
1018 			continue;
1019 
1020 		sk_nulls_for_each(sk, node, hslot) {
1021 			if (net_eq(sock_net(sk), net) &&
1022 			    sk->sk_family == state->family)
1023 				goto found;
1024 		}
1025 	}
1026 	sk = NULL;
1027 found:
1028 	return sk;
1029 }
1030 
1031 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1032 {
1033 	struct ping_iter_state *state = seq->private;
1034 	struct net *net = seq_file_net(seq);
1035 
1036 	do {
1037 		sk = sk_nulls_next(sk);
1038 	} while (sk && (!net_eq(sock_net(sk), net)));
1039 
1040 	if (!sk)
1041 		return ping_get_first(seq, state->bucket + 1);
1042 	return sk;
1043 }
1044 
1045 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1046 {
1047 	struct sock *sk = ping_get_first(seq, 0);
1048 
1049 	if (sk)
1050 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1051 			--pos;
1052 	return pos ? NULL : sk;
1053 }
1054 
1055 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1056 	__acquires(ping_table.lock)
1057 {
1058 	struct ping_iter_state *state = seq->private;
1059 	state->bucket = 0;
1060 	state->family = family;
1061 
1062 	read_lock_bh(&ping_table.lock);
1063 
1064 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1065 }
1066 EXPORT_SYMBOL_GPL(ping_seq_start);
1067 
1068 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1069 {
1070 	return ping_seq_start(seq, pos, AF_INET);
1071 }
1072 
1073 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1074 {
1075 	struct sock *sk;
1076 
1077 	if (v == SEQ_START_TOKEN)
1078 		sk = ping_get_idx(seq, 0);
1079 	else
1080 		sk = ping_get_next(seq, v);
1081 
1082 	++*pos;
1083 	return sk;
1084 }
1085 EXPORT_SYMBOL_GPL(ping_seq_next);
1086 
1087 void ping_seq_stop(struct seq_file *seq, void *v)
1088 	__releases(ping_table.lock)
1089 {
1090 	read_unlock_bh(&ping_table.lock);
1091 }
1092 EXPORT_SYMBOL_GPL(ping_seq_stop);
1093 
1094 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1095 		int bucket)
1096 {
1097 	struct inet_sock *inet = inet_sk(sp);
1098 	__be32 dest = inet->inet_daddr;
1099 	__be32 src = inet->inet_rcv_saddr;
1100 	__u16 destp = ntohs(inet->inet_dport);
1101 	__u16 srcp = ntohs(inet->inet_sport);
1102 
1103 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1104 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1105 		bucket, src, srcp, dest, destp, sp->sk_state,
1106 		sk_wmem_alloc_get(sp),
1107 		sk_rmem_alloc_get(sp),
1108 		0, 0L, 0,
1109 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1110 		0, sock_i_ino(sp),
1111 		refcount_read(&sp->sk_refcnt), sp,
1112 		atomic_read(&sp->sk_drops));
1113 }
1114 
1115 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1116 {
1117 	seq_setwidth(seq, 127);
1118 	if (v == SEQ_START_TOKEN)
1119 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1120 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1121 			   "inode ref pointer drops");
1122 	else {
1123 		struct ping_iter_state *state = seq->private;
1124 
1125 		ping_v4_format_sock(v, seq, state->bucket);
1126 	}
1127 	seq_pad(seq, '\n');
1128 	return 0;
1129 }
1130 
1131 static const struct seq_operations ping_v4_seq_ops = {
1132 	.start		= ping_v4_seq_start,
1133 	.show		= ping_v4_seq_show,
1134 	.next		= ping_seq_next,
1135 	.stop		= ping_seq_stop,
1136 };
1137 
1138 static int __net_init ping_v4_proc_init_net(struct net *net)
1139 {
1140 	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1141 			sizeof(struct ping_iter_state)))
1142 		return -ENOMEM;
1143 	return 0;
1144 }
1145 
1146 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1147 {
1148 	remove_proc_entry("icmp", net->proc_net);
1149 }
1150 
1151 static struct pernet_operations ping_v4_net_ops = {
1152 	.init = ping_v4_proc_init_net,
1153 	.exit = ping_v4_proc_exit_net,
1154 };
1155 
1156 int __init ping_proc_init(void)
1157 {
1158 	return register_pernet_subsys(&ping_v4_net_ops);
1159 }
1160 
1161 void ping_proc_exit(void)
1162 {
1163 	unregister_pernet_subsys(&ping_v4_net_ops);
1164 }
1165 
1166 #endif
1167 
1168 void __init ping_init(void)
1169 {
1170 	int i;
1171 
1172 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1173 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1174 	rwlock_init(&ping_table.lock);
1175 }
1176