xref: /openbmc/linux/net/ipv4/ping.c (revision fbb6b31a)
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 		return NULL;
191 	}
192 
193 	read_lock_bh(&ping_table.lock);
194 
195 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
196 		isk = inet_sk(sk);
197 
198 		pr_debug("iterate\n");
199 		if (isk->inet_num != ident)
200 			continue;
201 
202 		if (skb->protocol == htons(ETH_P_IP) &&
203 		    sk->sk_family == AF_INET) {
204 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
205 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
206 				 sk->sk_bound_dev_if);
207 
208 			if (isk->inet_rcv_saddr &&
209 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
210 				continue;
211 #if IS_ENABLED(CONFIG_IPV6)
212 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
213 			   sk->sk_family == AF_INET6) {
214 
215 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
216 				 (int) isk->inet_num,
217 				 &sk->sk_v6_rcv_saddr,
218 				 sk->sk_bound_dev_if);
219 
220 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
221 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
222 					     &ipv6_hdr(skb)->daddr))
223 				continue;
224 #endif
225 		} else {
226 			continue;
227 		}
228 
229 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
230 		    sk->sk_bound_dev_if != sdif)
231 			continue;
232 
233 		sock_hold(sk);
234 		goto exit;
235 	}
236 
237 	sk = NULL;
238 exit:
239 	read_unlock_bh(&ping_table.lock);
240 
241 	return sk;
242 }
243 
244 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
245 					  kgid_t *high)
246 {
247 	kgid_t *data = net->ipv4.ping_group_range.range;
248 	unsigned int seq;
249 
250 	do {
251 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
252 
253 		*low = data[0];
254 		*high = data[1];
255 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
256 }
257 
258 
259 int ping_init_sock(struct sock *sk)
260 {
261 	struct net *net = sock_net(sk);
262 	kgid_t group = current_egid();
263 	struct group_info *group_info;
264 	int i;
265 	kgid_t low, high;
266 	int ret = 0;
267 
268 	if (sk->sk_family == AF_INET6)
269 		sk->sk_ipv6only = 1;
270 
271 	inet_get_ping_group_range_net(net, &low, &high);
272 	if (gid_lte(low, group) && gid_lte(group, high))
273 		return 0;
274 
275 	group_info = get_current_groups();
276 	for (i = 0; i < group_info->ngroups; i++) {
277 		kgid_t gid = group_info->gid[i];
278 
279 		if (gid_lte(low, gid) && gid_lte(gid, high))
280 			goto out_release_group;
281 	}
282 
283 	ret = -EACCES;
284 
285 out_release_group:
286 	put_group_info(group_info);
287 	return ret;
288 }
289 EXPORT_SYMBOL_GPL(ping_init_sock);
290 
291 void ping_close(struct sock *sk, long timeout)
292 {
293 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
294 		 inet_sk(sk), inet_sk(sk)->inet_num);
295 	pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
296 
297 	sk_common_release(sk);
298 }
299 EXPORT_SYMBOL_GPL(ping_close);
300 
301 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
302 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
303 				struct sockaddr *uaddr, int addr_len)
304 {
305 	struct net *net = sock_net(sk);
306 	if (sk->sk_family == AF_INET) {
307 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
308 		u32 tb_id = RT_TABLE_LOCAL;
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 		tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
323 		chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
324 
325 		if (!inet_addr_valid_or_nonlocal(net, inet_sk(sk),
326 					         addr->sin_addr.s_addr,
327 	                                         chk_addr_ret))
328 			return -EADDRNOTAVAIL;
329 
330 #if IS_ENABLED(CONFIG_IPV6)
331 	} else if (sk->sk_family == AF_INET6) {
332 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
333 		int addr_type, scoped, has_addr;
334 		struct net_device *dev = NULL;
335 
336 		if (addr_len < sizeof(*addr))
337 			return -EINVAL;
338 
339 		if (addr->sin6_family != AF_INET6)
340 			return -EAFNOSUPPORT;
341 
342 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
343 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
344 
345 		addr_type = ipv6_addr_type(&addr->sin6_addr);
346 		scoped = __ipv6_addr_needs_scope_id(addr_type);
347 		if ((addr_type != IPV6_ADDR_ANY &&
348 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
349 		    (scoped && !addr->sin6_scope_id))
350 			return -EINVAL;
351 
352 		rcu_read_lock();
353 		if (addr->sin6_scope_id) {
354 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
355 			if (!dev) {
356 				rcu_read_unlock();
357 				return -ENODEV;
358 			}
359 		}
360 
361 		if (!dev && sk->sk_bound_dev_if) {
362 			dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
363 			if (!dev) {
364 				rcu_read_unlock();
365 				return -ENODEV;
366 			}
367 		}
368 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
369 						    scoped);
370 		rcu_read_unlock();
371 
372 		if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
373 		      addr_type == IPV6_ADDR_ANY))
374 			return -EADDRNOTAVAIL;
375 
376 		if (scoped)
377 			sk->sk_bound_dev_if = addr->sin6_scope_id;
378 #endif
379 	} else {
380 		return -EAFNOSUPPORT;
381 	}
382 	return 0;
383 }
384 
385 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
386 {
387 	if (saddr->sa_family == AF_INET) {
388 		struct inet_sock *isk = inet_sk(sk);
389 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
390 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
391 #if IS_ENABLED(CONFIG_IPV6)
392 	} else if (saddr->sa_family == AF_INET6) {
393 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
394 		struct ipv6_pinfo *np = inet6_sk(sk);
395 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
396 #endif
397 	}
398 }
399 
400 /*
401  * We need our own bind because there are no privileged id's == local ports.
402  * Moreover, we don't allow binding to multi- and broadcast addresses.
403  */
404 
405 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
406 {
407 	struct inet_sock *isk = inet_sk(sk);
408 	unsigned short snum;
409 	int err;
410 	int dif = sk->sk_bound_dev_if;
411 
412 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
413 	if (err)
414 		return err;
415 
416 	lock_sock(sk);
417 
418 	err = -EINVAL;
419 	if (isk->inet_num != 0)
420 		goto out;
421 
422 	err = -EADDRINUSE;
423 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
424 	if (ping_get_port(sk, snum) != 0) {
425 		/* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
426 		sk->sk_bound_dev_if = dif;
427 		goto out;
428 	}
429 	ping_set_saddr(sk, uaddr);
430 
431 	pr_debug("after bind(): num = %hu, dif = %d\n",
432 		 isk->inet_num,
433 		 sk->sk_bound_dev_if);
434 
435 	err = 0;
436 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
437 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
438 #if IS_ENABLED(CONFIG_IPV6)
439 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
440 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
441 #endif
442 
443 	if (snum)
444 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
445 	isk->inet_sport = htons(isk->inet_num);
446 	isk->inet_daddr = 0;
447 	isk->inet_dport = 0;
448 
449 #if IS_ENABLED(CONFIG_IPV6)
450 	if (sk->sk_family == AF_INET6)
451 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
452 #endif
453 
454 	sk_dst_reset(sk);
455 out:
456 	release_sock(sk);
457 	pr_debug("ping_v4_bind -> %d\n", err);
458 	return err;
459 }
460 EXPORT_SYMBOL_GPL(ping_bind);
461 
462 /*
463  * Is this a supported type of ICMP message?
464  */
465 
466 static inline int ping_supported(int family, int type, int code)
467 {
468 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
469 	       (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
470 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
471 	       (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
472 }
473 
474 /*
475  * This routine is called by the ICMP module when it gets some
476  * sort of error condition.
477  */
478 
479 void ping_err(struct sk_buff *skb, int offset, u32 info)
480 {
481 	int family;
482 	struct icmphdr *icmph;
483 	struct inet_sock *inet_sock;
484 	int type;
485 	int code;
486 	struct net *net = dev_net(skb->dev);
487 	struct sock *sk;
488 	int harderr;
489 	int err;
490 
491 	if (skb->protocol == htons(ETH_P_IP)) {
492 		family = AF_INET;
493 		type = icmp_hdr(skb)->type;
494 		code = icmp_hdr(skb)->code;
495 		icmph = (struct icmphdr *)(skb->data + offset);
496 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
497 		family = AF_INET6;
498 		type = icmp6_hdr(skb)->icmp6_type;
499 		code = icmp6_hdr(skb)->icmp6_code;
500 		icmph = (struct icmphdr *) (skb->data + offset);
501 	} else {
502 		BUG();
503 	}
504 
505 	/* We assume the packet has already been checked by icmp_unreach */
506 
507 	if (!ping_supported(family, icmph->type, icmph->code))
508 		return;
509 
510 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
511 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
512 		 ntohs(icmph->un.echo.sequence));
513 
514 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
515 	if (!sk) {
516 		pr_debug("no socket, dropping\n");
517 		return;	/* No socket for error */
518 	}
519 	pr_debug("err on socket %p\n", sk);
520 
521 	err = 0;
522 	harderr = 0;
523 	inet_sock = inet_sk(sk);
524 
525 	if (skb->protocol == htons(ETH_P_IP)) {
526 		switch (type) {
527 		default:
528 		case ICMP_TIME_EXCEEDED:
529 			err = EHOSTUNREACH;
530 			break;
531 		case ICMP_SOURCE_QUENCH:
532 			/* This is not a real error but ping wants to see it.
533 			 * Report it with some fake errno.
534 			 */
535 			err = EREMOTEIO;
536 			break;
537 		case ICMP_PARAMETERPROB:
538 			err = EPROTO;
539 			harderr = 1;
540 			break;
541 		case ICMP_DEST_UNREACH:
542 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
543 				ipv4_sk_update_pmtu(skb, sk, info);
544 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
545 					err = EMSGSIZE;
546 					harderr = 1;
547 					break;
548 				}
549 				goto out;
550 			}
551 			err = EHOSTUNREACH;
552 			if (code <= NR_ICMP_UNREACH) {
553 				harderr = icmp_err_convert[code].fatal;
554 				err = icmp_err_convert[code].errno;
555 			}
556 			break;
557 		case ICMP_REDIRECT:
558 			/* See ICMP_SOURCE_QUENCH */
559 			ipv4_sk_redirect(skb, sk);
560 			err = EREMOTEIO;
561 			break;
562 		}
563 #if IS_ENABLED(CONFIG_IPV6)
564 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
565 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
566 #endif
567 	}
568 
569 	/*
570 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
571 	 *	4.1.3.3.
572 	 */
573 	if ((family == AF_INET && !inet_sock->recverr) ||
574 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
575 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
576 			goto out;
577 	} else {
578 		if (family == AF_INET) {
579 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
580 				      info, (u8 *)icmph);
581 #if IS_ENABLED(CONFIG_IPV6)
582 		} else if (family == AF_INET6) {
583 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
584 						   info, (u8 *)icmph);
585 #endif
586 		}
587 	}
588 	sk->sk_err = err;
589 	sk_error_report(sk);
590 out:
591 	sock_put(sk);
592 }
593 EXPORT_SYMBOL_GPL(ping_err);
594 
595 /*
596  *	Copy and checksum an ICMP Echo packet from user space into a buffer
597  *	starting from the payload.
598  */
599 
600 int ping_getfrag(void *from, char *to,
601 		 int offset, int fraglen, int odd, struct sk_buff *skb)
602 {
603 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
604 
605 	if (offset == 0) {
606 		fraglen -= sizeof(struct icmphdr);
607 		if (fraglen < 0)
608 			BUG();
609 		if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
610 			    fraglen, &pfh->wcheck,
611 			    &pfh->msg->msg_iter))
612 			return -EFAULT;
613 	} else if (offset < sizeof(struct icmphdr)) {
614 			BUG();
615 	} else {
616 		if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
617 					    &pfh->msg->msg_iter))
618 			return -EFAULT;
619 	}
620 
621 #if IS_ENABLED(CONFIG_IPV6)
622 	/* For IPv6, checksum each skb as we go along, as expected by
623 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
624 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
625 	 */
626 	if (pfh->family == AF_INET6) {
627 		skb->csum = pfh->wcheck;
628 		skb->ip_summed = CHECKSUM_NONE;
629 		pfh->wcheck = 0;
630 	}
631 #endif
632 
633 	return 0;
634 }
635 EXPORT_SYMBOL_GPL(ping_getfrag);
636 
637 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
638 				       struct flowi4 *fl4)
639 {
640 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
641 
642 	if (!skb)
643 		return 0;
644 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
645 		sizeof(struct icmphdr), pfh->wcheck);
646 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
647 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
648 	skb->ip_summed = CHECKSUM_NONE;
649 	return ip_push_pending_frames(sk, fl4);
650 }
651 
652 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
653 			void *user_icmph, size_t icmph_len)
654 {
655 	u8 type, code;
656 
657 	if (len > 0xFFFF)
658 		return -EMSGSIZE;
659 
660 	/* Must have at least a full ICMP header. */
661 	if (len < icmph_len)
662 		return -EINVAL;
663 
664 	/*
665 	 *	Check the flags.
666 	 */
667 
668 	/* Mirror BSD error message compatibility */
669 	if (msg->msg_flags & MSG_OOB)
670 		return -EOPNOTSUPP;
671 
672 	/*
673 	 *	Fetch the ICMP header provided by the userland.
674 	 *	iovec is modified! The ICMP header is consumed.
675 	 */
676 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
677 		return -EFAULT;
678 
679 	if (family == AF_INET) {
680 		type = ((struct icmphdr *) user_icmph)->type;
681 		code = ((struct icmphdr *) user_icmph)->code;
682 #if IS_ENABLED(CONFIG_IPV6)
683 	} else if (family == AF_INET6) {
684 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
685 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
686 #endif
687 	} else {
688 		BUG();
689 	}
690 
691 	if (!ping_supported(family, type, code))
692 		return -EINVAL;
693 
694 	return 0;
695 }
696 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
697 
698 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
699 {
700 	struct net *net = sock_net(sk);
701 	struct flowi4 fl4;
702 	struct inet_sock *inet = inet_sk(sk);
703 	struct ipcm_cookie ipc;
704 	struct icmphdr user_icmph;
705 	struct pingfakehdr pfh;
706 	struct rtable *rt = NULL;
707 	struct ip_options_data opt_copy;
708 	int free = 0;
709 	__be32 saddr, daddr, faddr;
710 	u8  tos;
711 	int err;
712 
713 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
714 
715 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
716 				  sizeof(user_icmph));
717 	if (err)
718 		return err;
719 
720 	/*
721 	 *	Get and verify the address.
722 	 */
723 
724 	if (msg->msg_name) {
725 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
726 		if (msg->msg_namelen < sizeof(*usin))
727 			return -EINVAL;
728 		if (usin->sin_family != AF_INET)
729 			return -EAFNOSUPPORT;
730 		daddr = usin->sin_addr.s_addr;
731 		/* no remote port */
732 	} else {
733 		if (sk->sk_state != TCP_ESTABLISHED)
734 			return -EDESTADDRREQ;
735 		daddr = inet->inet_daddr;
736 		/* no remote port */
737 	}
738 
739 	ipcm_init_sk(&ipc, inet);
740 
741 	if (msg->msg_controllen) {
742 		err = ip_cmsg_send(sk, msg, &ipc, false);
743 		if (unlikely(err)) {
744 			kfree(ipc.opt);
745 			return err;
746 		}
747 		if (ipc.opt)
748 			free = 1;
749 	}
750 	if (!ipc.opt) {
751 		struct ip_options_rcu *inet_opt;
752 
753 		rcu_read_lock();
754 		inet_opt = rcu_dereference(inet->inet_opt);
755 		if (inet_opt) {
756 			memcpy(&opt_copy, inet_opt,
757 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
758 			ipc.opt = &opt_copy.opt;
759 		}
760 		rcu_read_unlock();
761 	}
762 
763 	saddr = ipc.addr;
764 	ipc.addr = faddr = daddr;
765 
766 	if (ipc.opt && ipc.opt->opt.srr) {
767 		if (!daddr) {
768 			err = -EINVAL;
769 			goto out_free;
770 		}
771 		faddr = ipc.opt->opt.faddr;
772 	}
773 	tos = get_rttos(&ipc, inet);
774 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
775 	    (msg->msg_flags & MSG_DONTROUTE) ||
776 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
777 		tos |= RTO_ONLINK;
778 	}
779 
780 	if (ipv4_is_multicast(daddr)) {
781 		if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
782 			ipc.oif = inet->mc_index;
783 		if (!saddr)
784 			saddr = inet->mc_addr;
785 	} else if (!ipc.oif)
786 		ipc.oif = inet->uc_index;
787 
788 	flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
789 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
790 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
791 			   sk->sk_uid);
792 
793 	fl4.fl4_icmp_type = user_icmph.type;
794 	fl4.fl4_icmp_code = user_icmph.code;
795 
796 	security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
797 	rt = ip_route_output_flow(net, &fl4, sk);
798 	if (IS_ERR(rt)) {
799 		err = PTR_ERR(rt);
800 		rt = NULL;
801 		if (err == -ENETUNREACH)
802 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
803 		goto out;
804 	}
805 
806 	err = -EACCES;
807 	if ((rt->rt_flags & RTCF_BROADCAST) &&
808 	    !sock_flag(sk, SOCK_BROADCAST))
809 		goto out;
810 
811 	if (msg->msg_flags & MSG_CONFIRM)
812 		goto do_confirm;
813 back_from_confirm:
814 
815 	if (!ipc.addr)
816 		ipc.addr = fl4.daddr;
817 
818 	lock_sock(sk);
819 
820 	pfh.icmph.type = user_icmph.type; /* already checked */
821 	pfh.icmph.code = user_icmph.code; /* ditto */
822 	pfh.icmph.checksum = 0;
823 	pfh.icmph.un.echo.id = inet->inet_sport;
824 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
825 	pfh.msg = msg;
826 	pfh.wcheck = 0;
827 	pfh.family = AF_INET;
828 
829 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
830 			0, &ipc, &rt, msg->msg_flags);
831 	if (err)
832 		ip_flush_pending_frames(sk);
833 	else
834 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
835 	release_sock(sk);
836 
837 out:
838 	ip_rt_put(rt);
839 out_free:
840 	if (free)
841 		kfree(ipc.opt);
842 	if (!err) {
843 		icmp_out_count(sock_net(sk), user_icmph.type);
844 		return len;
845 	}
846 	return err;
847 
848 do_confirm:
849 	if (msg->msg_flags & MSG_PROBE)
850 		dst_confirm_neigh(&rt->dst, &fl4.daddr);
851 	if (!(msg->msg_flags & MSG_PROBE) || len)
852 		goto back_from_confirm;
853 	err = 0;
854 	goto out;
855 }
856 
857 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
858 		 int flags, int *addr_len)
859 {
860 	struct inet_sock *isk = inet_sk(sk);
861 	int family = sk->sk_family;
862 	struct sk_buff *skb;
863 	int copied, err;
864 
865 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
866 
867 	err = -EOPNOTSUPP;
868 	if (flags & MSG_OOB)
869 		goto out;
870 
871 	if (flags & MSG_ERRQUEUE)
872 		return inet_recv_error(sk, msg, len, addr_len);
873 
874 	skb = skb_recv_datagram(sk, flags, noblock, &err);
875 	if (!skb)
876 		goto out;
877 
878 	copied = skb->len;
879 	if (copied > len) {
880 		msg->msg_flags |= MSG_TRUNC;
881 		copied = len;
882 	}
883 
884 	/* Don't bother checking the checksum */
885 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
886 	if (err)
887 		goto done;
888 
889 	sock_recv_timestamp(msg, sk, skb);
890 
891 	/* Copy the address and add cmsg data. */
892 	if (family == AF_INET) {
893 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
894 
895 		if (sin) {
896 			sin->sin_family = AF_INET;
897 			sin->sin_port = 0 /* skb->h.uh->source */;
898 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
899 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
900 			*addr_len = sizeof(*sin);
901 		}
902 
903 		if (isk->cmsg_flags)
904 			ip_cmsg_recv(msg, skb);
905 
906 #if IS_ENABLED(CONFIG_IPV6)
907 	} else if (family == AF_INET6) {
908 		struct ipv6_pinfo *np = inet6_sk(sk);
909 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
910 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
911 
912 		if (sin6) {
913 			sin6->sin6_family = AF_INET6;
914 			sin6->sin6_port = 0;
915 			sin6->sin6_addr = ip6->saddr;
916 			sin6->sin6_flowinfo = 0;
917 			if (np->sndflow)
918 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
919 			sin6->sin6_scope_id =
920 				ipv6_iface_scope_id(&sin6->sin6_addr,
921 						    inet6_iif(skb));
922 			*addr_len = sizeof(*sin6);
923 		}
924 
925 		if (inet6_sk(sk)->rxopt.all)
926 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
927 		if (skb->protocol == htons(ETH_P_IPV6) &&
928 		    inet6_sk(sk)->rxopt.all)
929 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
930 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
931 			ip_cmsg_recv(msg, skb);
932 #endif
933 	} else {
934 		BUG();
935 	}
936 
937 	err = copied;
938 
939 done:
940 	skb_free_datagram(sk, skb);
941 out:
942 	pr_debug("ping_recvmsg -> %d\n", err);
943 	return err;
944 }
945 EXPORT_SYMBOL_GPL(ping_recvmsg);
946 
947 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
948 {
949 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
950 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
951 	if (sock_queue_rcv_skb(sk, skb) < 0) {
952 		kfree_skb(skb);
953 		pr_debug("ping_queue_rcv_skb -> failed\n");
954 		return -1;
955 	}
956 	return 0;
957 }
958 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
959 
960 
961 /*
962  *	All we need to do is get the socket.
963  */
964 
965 bool ping_rcv(struct sk_buff *skb)
966 {
967 	struct sock *sk;
968 	struct net *net = dev_net(skb->dev);
969 	struct icmphdr *icmph = icmp_hdr(skb);
970 	bool rc = false;
971 
972 	/* We assume the packet has already been checked by icmp_rcv */
973 
974 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
975 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
976 
977 	/* Push ICMP header back */
978 	skb_push(skb, skb->data - (u8 *)icmph);
979 
980 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
981 	if (sk) {
982 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
983 
984 		pr_debug("rcv on socket %p\n", sk);
985 		if (skb2 && !ping_queue_rcv_skb(sk, skb2))
986 			rc = true;
987 		sock_put(sk);
988 	}
989 
990 	if (!rc)
991 		pr_debug("no socket, dropping\n");
992 
993 	return rc;
994 }
995 EXPORT_SYMBOL_GPL(ping_rcv);
996 
997 struct proto ping_prot = {
998 	.name =		"PING",
999 	.owner =	THIS_MODULE,
1000 	.init =		ping_init_sock,
1001 	.close =	ping_close,
1002 	.connect =	ip4_datagram_connect,
1003 	.disconnect =	__udp_disconnect,
1004 	.setsockopt =	ip_setsockopt,
1005 	.getsockopt =	ip_getsockopt,
1006 	.sendmsg =	ping_v4_sendmsg,
1007 	.recvmsg =	ping_recvmsg,
1008 	.bind =		ping_bind,
1009 	.backlog_rcv =	ping_queue_rcv_skb,
1010 	.release_cb =	ip4_datagram_release_cb,
1011 	.hash =		ping_hash,
1012 	.unhash =	ping_unhash,
1013 	.get_port =	ping_get_port,
1014 	.put_port =	ping_unhash,
1015 	.obj_size =	sizeof(struct inet_sock),
1016 };
1017 EXPORT_SYMBOL(ping_prot);
1018 
1019 #ifdef CONFIG_PROC_FS
1020 
1021 static struct sock *ping_get_first(struct seq_file *seq, int start)
1022 {
1023 	struct sock *sk;
1024 	struct ping_iter_state *state = seq->private;
1025 	struct net *net = seq_file_net(seq);
1026 
1027 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1028 	     ++state->bucket) {
1029 		struct hlist_nulls_node *node;
1030 		struct hlist_nulls_head *hslot;
1031 
1032 		hslot = &ping_table.hash[state->bucket];
1033 
1034 		if (hlist_nulls_empty(hslot))
1035 			continue;
1036 
1037 		sk_nulls_for_each(sk, node, hslot) {
1038 			if (net_eq(sock_net(sk), net) &&
1039 			    sk->sk_family == state->family)
1040 				goto found;
1041 		}
1042 	}
1043 	sk = NULL;
1044 found:
1045 	return sk;
1046 }
1047 
1048 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1049 {
1050 	struct ping_iter_state *state = seq->private;
1051 	struct net *net = seq_file_net(seq);
1052 
1053 	do {
1054 		sk = sk_nulls_next(sk);
1055 	} while (sk && (!net_eq(sock_net(sk), net)));
1056 
1057 	if (!sk)
1058 		return ping_get_first(seq, state->bucket + 1);
1059 	return sk;
1060 }
1061 
1062 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1063 {
1064 	struct sock *sk = ping_get_first(seq, 0);
1065 
1066 	if (sk)
1067 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1068 			--pos;
1069 	return pos ? NULL : sk;
1070 }
1071 
1072 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1073 	__acquires(ping_table.lock)
1074 {
1075 	struct ping_iter_state *state = seq->private;
1076 	state->bucket = 0;
1077 	state->family = family;
1078 
1079 	read_lock_bh(&ping_table.lock);
1080 
1081 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1082 }
1083 EXPORT_SYMBOL_GPL(ping_seq_start);
1084 
1085 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1086 {
1087 	return ping_seq_start(seq, pos, AF_INET);
1088 }
1089 
1090 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1091 {
1092 	struct sock *sk;
1093 
1094 	if (v == SEQ_START_TOKEN)
1095 		sk = ping_get_idx(seq, 0);
1096 	else
1097 		sk = ping_get_next(seq, v);
1098 
1099 	++*pos;
1100 	return sk;
1101 }
1102 EXPORT_SYMBOL_GPL(ping_seq_next);
1103 
1104 void ping_seq_stop(struct seq_file *seq, void *v)
1105 	__releases(ping_table.lock)
1106 {
1107 	read_unlock_bh(&ping_table.lock);
1108 }
1109 EXPORT_SYMBOL_GPL(ping_seq_stop);
1110 
1111 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1112 		int bucket)
1113 {
1114 	struct inet_sock *inet = inet_sk(sp);
1115 	__be32 dest = inet->inet_daddr;
1116 	__be32 src = inet->inet_rcv_saddr;
1117 	__u16 destp = ntohs(inet->inet_dport);
1118 	__u16 srcp = ntohs(inet->inet_sport);
1119 
1120 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1121 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1122 		bucket, src, srcp, dest, destp, sp->sk_state,
1123 		sk_wmem_alloc_get(sp),
1124 		sk_rmem_alloc_get(sp),
1125 		0, 0L, 0,
1126 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1127 		0, sock_i_ino(sp),
1128 		refcount_read(&sp->sk_refcnt), sp,
1129 		atomic_read(&sp->sk_drops));
1130 }
1131 
1132 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1133 {
1134 	seq_setwidth(seq, 127);
1135 	if (v == SEQ_START_TOKEN)
1136 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1137 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1138 			   "inode ref pointer drops");
1139 	else {
1140 		struct ping_iter_state *state = seq->private;
1141 
1142 		ping_v4_format_sock(v, seq, state->bucket);
1143 	}
1144 	seq_pad(seq, '\n');
1145 	return 0;
1146 }
1147 
1148 static const struct seq_operations ping_v4_seq_ops = {
1149 	.start		= ping_v4_seq_start,
1150 	.show		= ping_v4_seq_show,
1151 	.next		= ping_seq_next,
1152 	.stop		= ping_seq_stop,
1153 };
1154 
1155 static int __net_init ping_v4_proc_init_net(struct net *net)
1156 {
1157 	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1158 			sizeof(struct ping_iter_state)))
1159 		return -ENOMEM;
1160 	return 0;
1161 }
1162 
1163 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1164 {
1165 	remove_proc_entry("icmp", net->proc_net);
1166 }
1167 
1168 static struct pernet_operations ping_v4_net_ops = {
1169 	.init = ping_v4_proc_init_net,
1170 	.exit = ping_v4_proc_exit_net,
1171 };
1172 
1173 int __init ping_proc_init(void)
1174 {
1175 	return register_pernet_subsys(&ping_v4_net_ops);
1176 }
1177 
1178 void ping_proc_exit(void)
1179 {
1180 	unregister_pernet_subsys(&ping_v4_net_ops);
1181 }
1182 
1183 #endif
1184 
1185 void __init ping_init(void)
1186 {
1187 	int i;
1188 
1189 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1190 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1191 	rwlock_init(&ping_table.lock);
1192 }
1193