xref: /linux/net/ipv4/ping.c (revision 95db3b255fde4e830e5f8cc011eb404023f669d4)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		"Ping" sockets
7  *
8  *		This program is free software; you can redistribute it and/or
9  *		modify it under the terms of the GNU General Public License
10  *		as published by the Free Software Foundation; either version
11  *		2 of the License, or (at your option) any later version.
12  *
13  * Based on ipv4/udp.c code.
14  *
15  * Authors:	Vasiliy Kulikov / Openwall (for Linux 2.6),
16  *		Pavel Kankovsky (for Linux 2.4.32)
17  *
18  * Pavel gave all rights to bugs to Vasiliy,
19  * none of the bugs are Pavel's now.
20  *
21  */
22 
23 #include <linux/uaccess.h>
24 #include <linux/types.h>
25 #include <linux/fcntl.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/in.h>
29 #include <linux/errno.h>
30 #include <linux/timer.h>
31 #include <linux/mm.h>
32 #include <linux/inet.h>
33 #include <linux/netdevice.h>
34 #include <net/snmp.h>
35 #include <net/ip.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <linux/skbuff.h>
39 #include <linux/proc_fs.h>
40 #include <linux/export.h>
41 #include <net/sock.h>
42 #include <net/ping.h>
43 #include <net/udp.h>
44 #include <net/route.h>
45 #include <net/inet_common.h>
46 #include <net/checksum.h>
47 
48 #if IS_ENABLED(CONFIG_IPV6)
49 #include <linux/in6.h>
50 #include <linux/icmpv6.h>
51 #include <net/addrconf.h>
52 #include <net/ipv6.h>
53 #include <net/transp_v6.h>
54 #endif
55 
56 struct ping_table {
57 	struct hlist_nulls_head	hash[PING_HTABLE_SIZE];
58 	rwlock_t		lock;
59 };
60 
61 static struct ping_table ping_table;
62 struct pingv6_ops pingv6_ops;
63 EXPORT_SYMBOL_GPL(pingv6_ops);
64 
65 static u16 ping_port_rover;
66 
67 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
68 {
69 	u32 res = (num + net_hash_mix(net)) & mask;
70 
71 	pr_debug("hash(%u) = %u\n", num, res);
72 	return res;
73 }
74 EXPORT_SYMBOL_GPL(ping_hash);
75 
76 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
77 					     struct net *net, unsigned int num)
78 {
79 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
80 }
81 
82 int ping_get_port(struct sock *sk, unsigned short ident)
83 {
84 	struct hlist_nulls_node *node;
85 	struct hlist_nulls_head *hlist;
86 	struct inet_sock *isk, *isk2;
87 	struct sock *sk2 = NULL;
88 
89 	isk = inet_sk(sk);
90 	write_lock_bh(&ping_table.lock);
91 	if (ident == 0) {
92 		u32 i;
93 		u16 result = ping_port_rover + 1;
94 
95 		for (i = 0; i < (1L << 16); i++, result++) {
96 			if (!result)
97 				result++; /* avoid zero */
98 			hlist = ping_hashslot(&ping_table, sock_net(sk),
99 					    result);
100 			ping_portaddr_for_each_entry(sk2, node, hlist) {
101 				isk2 = inet_sk(sk2);
102 
103 				if (isk2->inet_num == result)
104 					goto next_port;
105 			}
106 
107 			/* found */
108 			ping_port_rover = ident = result;
109 			break;
110 next_port:
111 			;
112 		}
113 		if (i >= (1L << 16))
114 			goto fail;
115 	} else {
116 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
117 		ping_portaddr_for_each_entry(sk2, node, hlist) {
118 			isk2 = inet_sk(sk2);
119 
120 			/* BUG? Why is this reuse and not reuseaddr? ping.c
121 			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
122 			 * that other ping processes can steal its packets.
123 			 */
124 			if ((isk2->inet_num == ident) &&
125 			    (sk2 != sk) &&
126 			    (!sk2->sk_reuse || !sk->sk_reuse))
127 				goto fail;
128 		}
129 	}
130 
131 	pr_debug("found port/ident = %d\n", ident);
132 	isk->inet_num = ident;
133 	if (sk_unhashed(sk)) {
134 		pr_debug("was not hashed\n");
135 		sock_hold(sk);
136 		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
137 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
138 	}
139 	write_unlock_bh(&ping_table.lock);
140 	return 0;
141 
142 fail:
143 	write_unlock_bh(&ping_table.lock);
144 	return 1;
145 }
146 EXPORT_SYMBOL_GPL(ping_get_port);
147 
148 int ping_hash(struct sock *sk)
149 {
150 	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
151 	BUG(); /* "Please do not press this button again." */
152 
153 	return 0;
154 }
155 
156 void ping_unhash(struct sock *sk)
157 {
158 	struct inet_sock *isk = inet_sk(sk);
159 	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
160 	if (sk_hashed(sk)) {
161 		write_lock_bh(&ping_table.lock);
162 		hlist_nulls_del(&sk->sk_nulls_node);
163 		sk_nulls_node_init(&sk->sk_nulls_node);
164 		sock_put(sk);
165 		isk->inet_num = 0;
166 		isk->inet_sport = 0;
167 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
168 		write_unlock_bh(&ping_table.lock);
169 	}
170 }
171 EXPORT_SYMBOL_GPL(ping_unhash);
172 
173 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
174 {
175 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
176 	struct sock *sk = NULL;
177 	struct inet_sock *isk;
178 	struct hlist_nulls_node *hnode;
179 	int dif = skb->dev->ifindex;
180 
181 	if (skb->protocol == htons(ETH_P_IP)) {
182 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
183 			 (int)ident, &ip_hdr(skb)->daddr, dif);
184 #if IS_ENABLED(CONFIG_IPV6)
185 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
186 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
187 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
188 #endif
189 	}
190 
191 	read_lock_bh(&ping_table.lock);
192 
193 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
194 		isk = inet_sk(sk);
195 
196 		pr_debug("iterate\n");
197 		if (isk->inet_num != ident)
198 			continue;
199 
200 		if (skb->protocol == htons(ETH_P_IP) &&
201 		    sk->sk_family == AF_INET) {
202 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
203 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
204 				 sk->sk_bound_dev_if);
205 
206 			if (isk->inet_rcv_saddr &&
207 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
208 				continue;
209 #if IS_ENABLED(CONFIG_IPV6)
210 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
211 			   sk->sk_family == AF_INET6) {
212 
213 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
214 				 (int) isk->inet_num,
215 				 &sk->sk_v6_rcv_saddr,
216 				 sk->sk_bound_dev_if);
217 
218 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
219 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
220 					     &ipv6_hdr(skb)->daddr))
221 				continue;
222 #endif
223 		} else {
224 			continue;
225 		}
226 
227 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
228 			continue;
229 
230 		sock_hold(sk);
231 		goto exit;
232 	}
233 
234 	sk = NULL;
235 exit:
236 	read_unlock_bh(&ping_table.lock);
237 
238 	return sk;
239 }
240 
241 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
242 					  kgid_t *high)
243 {
244 	kgid_t *data = net->ipv4.ping_group_range.range;
245 	unsigned int seq;
246 
247 	do {
248 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
249 
250 		*low = data[0];
251 		*high = data[1];
252 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
253 }
254 
255 
256 int ping_init_sock(struct sock *sk)
257 {
258 	struct net *net = sock_net(sk);
259 	kgid_t group = current_egid();
260 	struct group_info *group_info;
261 	int i, j, count;
262 	kgid_t low, high;
263 	int ret = 0;
264 
265 	if (sk->sk_family == AF_INET6)
266 		sk->sk_ipv6only = 1;
267 
268 	inet_get_ping_group_range_net(net, &low, &high);
269 	if (gid_lte(low, group) && gid_lte(group, high))
270 		return 0;
271 
272 	group_info = get_current_groups();
273 	count = group_info->ngroups;
274 	for (i = 0; i < group_info->nblocks; i++) {
275 		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
276 		for (j = 0; j < cp_count; j++) {
277 			kgid_t gid = group_info->blocks[i][j];
278 			if (gid_lte(low, gid) && gid_lte(gid, high))
279 				goto out_release_group;
280 		}
281 
282 		count -= cp_count;
283 	}
284 
285 	ret = -EACCES;
286 
287 out_release_group:
288 	put_group_info(group_info);
289 	return ret;
290 }
291 EXPORT_SYMBOL_GPL(ping_init_sock);
292 
293 void ping_close(struct sock *sk, long timeout)
294 {
295 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
296 		 inet_sk(sk), inet_sk(sk)->inet_num);
297 	pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
298 
299 	sk_common_release(sk);
300 }
301 EXPORT_SYMBOL_GPL(ping_close);
302 
303 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
304 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
305 				struct sockaddr *uaddr, int addr_len) {
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 (addr->sin_addr.s_addr == htonl(INADDR_ANY))
325 			chk_addr_ret = RTN_LOCAL;
326 
327 		if ((net->ipv4.sysctl_ip_nonlocal_bind == 0 &&
328 		    isk->freebind == 0 && isk->transparent == 0 &&
329 		     chk_addr_ret != RTN_LOCAL) ||
330 		    chk_addr_ret == RTN_MULTICAST ||
331 		    chk_addr_ret == RTN_BROADCAST)
332 			return -EADDRNOTAVAIL;
333 
334 #if IS_ENABLED(CONFIG_IPV6)
335 	} else if (sk->sk_family == AF_INET6) {
336 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
337 		int addr_type, scoped, has_addr;
338 		struct net_device *dev = NULL;
339 
340 		if (addr_len < sizeof(*addr))
341 			return -EINVAL;
342 
343 		if (addr->sin6_family != AF_INET6)
344 			return -EAFNOSUPPORT;
345 
346 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
347 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
348 
349 		addr_type = ipv6_addr_type(&addr->sin6_addr);
350 		scoped = __ipv6_addr_needs_scope_id(addr_type);
351 		if ((addr_type != IPV6_ADDR_ANY &&
352 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
353 		    (scoped && !addr->sin6_scope_id))
354 			return -EINVAL;
355 
356 		rcu_read_lock();
357 		if (addr->sin6_scope_id) {
358 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
359 			if (!dev) {
360 				rcu_read_unlock();
361 				return -ENODEV;
362 			}
363 		}
364 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
365 						    scoped);
366 		rcu_read_unlock();
367 
368 		if (!(net->ipv6.sysctl.ip_nonlocal_bind ||
369 		      isk->freebind || isk->transparent || has_addr ||
370 		      addr_type == IPV6_ADDR_ANY))
371 			return -EADDRNOTAVAIL;
372 
373 		if (scoped)
374 			sk->sk_bound_dev_if = addr->sin6_scope_id;
375 #endif
376 	} else {
377 		return -EAFNOSUPPORT;
378 	}
379 	return 0;
380 }
381 
382 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
383 {
384 	if (saddr->sa_family == AF_INET) {
385 		struct inet_sock *isk = inet_sk(sk);
386 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
387 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
388 #if IS_ENABLED(CONFIG_IPV6)
389 	} else if (saddr->sa_family == AF_INET6) {
390 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
391 		struct ipv6_pinfo *np = inet6_sk(sk);
392 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
393 #endif
394 	}
395 }
396 
397 static void ping_clear_saddr(struct sock *sk, int dif)
398 {
399 	sk->sk_bound_dev_if = dif;
400 	if (sk->sk_family == AF_INET) {
401 		struct inet_sock *isk = inet_sk(sk);
402 		isk->inet_rcv_saddr = isk->inet_saddr = 0;
403 #if IS_ENABLED(CONFIG_IPV6)
404 	} else if (sk->sk_family == AF_INET6) {
405 		struct ipv6_pinfo *np = inet6_sk(sk);
406 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
407 		memset(&np->saddr, 0, sizeof(np->saddr));
408 #endif
409 	}
410 }
411 /*
412  * We need our own bind because there are no privileged id's == local ports.
413  * Moreover, we don't allow binding to multi- and broadcast addresses.
414  */
415 
416 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
417 {
418 	struct inet_sock *isk = inet_sk(sk);
419 	unsigned short snum;
420 	int err;
421 	int dif = sk->sk_bound_dev_if;
422 
423 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
424 	if (err)
425 		return err;
426 
427 	lock_sock(sk);
428 
429 	err = -EINVAL;
430 	if (isk->inet_num != 0)
431 		goto out;
432 
433 	err = -EADDRINUSE;
434 	ping_set_saddr(sk, uaddr);
435 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
436 	if (ping_get_port(sk, snum) != 0) {
437 		ping_clear_saddr(sk, dif);
438 		goto out;
439 	}
440 
441 	pr_debug("after bind(): num = %d, dif = %d\n",
442 		 (int)isk->inet_num,
443 		 (int)sk->sk_bound_dev_if);
444 
445 	err = 0;
446 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
447 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
448 #if IS_ENABLED(CONFIG_IPV6)
449 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
450 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
451 #endif
452 
453 	if (snum)
454 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
455 	isk->inet_sport = htons(isk->inet_num);
456 	isk->inet_daddr = 0;
457 	isk->inet_dport = 0;
458 
459 #if IS_ENABLED(CONFIG_IPV6)
460 	if (sk->sk_family == AF_INET6)
461 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
462 #endif
463 
464 	sk_dst_reset(sk);
465 out:
466 	release_sock(sk);
467 	pr_debug("ping_v4_bind -> %d\n", err);
468 	return err;
469 }
470 EXPORT_SYMBOL_GPL(ping_bind);
471 
472 /*
473  * Is this a supported type of ICMP message?
474  */
475 
476 static inline int ping_supported(int family, int type, int code)
477 {
478 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
479 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
480 }
481 
482 /*
483  * This routine is called by the ICMP module when it gets some
484  * sort of error condition.
485  */
486 
487 void ping_err(struct sk_buff *skb, int offset, u32 info)
488 {
489 	int family;
490 	struct icmphdr *icmph;
491 	struct inet_sock *inet_sock;
492 	int type;
493 	int code;
494 	struct net *net = dev_net(skb->dev);
495 	struct sock *sk;
496 	int harderr;
497 	int err;
498 
499 	if (skb->protocol == htons(ETH_P_IP)) {
500 		family = AF_INET;
501 		type = icmp_hdr(skb)->type;
502 		code = icmp_hdr(skb)->code;
503 		icmph = (struct icmphdr *)(skb->data + offset);
504 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
505 		family = AF_INET6;
506 		type = icmp6_hdr(skb)->icmp6_type;
507 		code = icmp6_hdr(skb)->icmp6_code;
508 		icmph = (struct icmphdr *) (skb->data + offset);
509 	} else {
510 		BUG();
511 	}
512 
513 	/* We assume the packet has already been checked by icmp_unreach */
514 
515 	if (!ping_supported(family, icmph->type, icmph->code))
516 		return;
517 
518 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
519 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
520 		 ntohs(icmph->un.echo.sequence));
521 
522 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
523 	if (!sk) {
524 		pr_debug("no socket, dropping\n");
525 		return;	/* No socket for error */
526 	}
527 	pr_debug("err on socket %p\n", sk);
528 
529 	err = 0;
530 	harderr = 0;
531 	inet_sock = inet_sk(sk);
532 
533 	if (skb->protocol == htons(ETH_P_IP)) {
534 		switch (type) {
535 		default:
536 		case ICMP_TIME_EXCEEDED:
537 			err = EHOSTUNREACH;
538 			break;
539 		case ICMP_SOURCE_QUENCH:
540 			/* This is not a real error but ping wants to see it.
541 			 * Report it with some fake errno.
542 			 */
543 			err = EREMOTEIO;
544 			break;
545 		case ICMP_PARAMETERPROB:
546 			err = EPROTO;
547 			harderr = 1;
548 			break;
549 		case ICMP_DEST_UNREACH:
550 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
551 				ipv4_sk_update_pmtu(skb, sk, info);
552 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
553 					err = EMSGSIZE;
554 					harderr = 1;
555 					break;
556 				}
557 				goto out;
558 			}
559 			err = EHOSTUNREACH;
560 			if (code <= NR_ICMP_UNREACH) {
561 				harderr = icmp_err_convert[code].fatal;
562 				err = icmp_err_convert[code].errno;
563 			}
564 			break;
565 		case ICMP_REDIRECT:
566 			/* See ICMP_SOURCE_QUENCH */
567 			ipv4_sk_redirect(skb, sk);
568 			err = EREMOTEIO;
569 			break;
570 		}
571 #if IS_ENABLED(CONFIG_IPV6)
572 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
573 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
574 #endif
575 	}
576 
577 	/*
578 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
579 	 *	4.1.3.3.
580 	 */
581 	if ((family == AF_INET && !inet_sock->recverr) ||
582 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
583 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
584 			goto out;
585 	} else {
586 		if (family == AF_INET) {
587 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
588 				      info, (u8 *)icmph);
589 #if IS_ENABLED(CONFIG_IPV6)
590 		} else if (family == AF_INET6) {
591 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
592 						   info, (u8 *)icmph);
593 #endif
594 		}
595 	}
596 	sk->sk_err = err;
597 	sk->sk_error_report(sk);
598 out:
599 	sock_put(sk);
600 }
601 EXPORT_SYMBOL_GPL(ping_err);
602 
603 /*
604  *	Copy and checksum an ICMP Echo packet from user space into a buffer
605  *	starting from the payload.
606  */
607 
608 int ping_getfrag(void *from, char *to,
609 		 int offset, int fraglen, int odd, struct sk_buff *skb)
610 {
611 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
612 
613 	if (offset == 0) {
614 		fraglen -= sizeof(struct icmphdr);
615 		if (fraglen < 0)
616 			BUG();
617 		if (csum_and_copy_from_iter(to + sizeof(struct icmphdr),
618 			    fraglen, &pfh->wcheck,
619 			    &pfh->msg->msg_iter) != fraglen)
620 			return -EFAULT;
621 	} else if (offset < sizeof(struct icmphdr)) {
622 			BUG();
623 	} else {
624 		if (csum_and_copy_from_iter(to, fraglen, &pfh->wcheck,
625 					    &pfh->msg->msg_iter) != fraglen)
626 			return -EFAULT;
627 	}
628 
629 #if IS_ENABLED(CONFIG_IPV6)
630 	/* For IPv6, checksum each skb as we go along, as expected by
631 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
632 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
633 	 */
634 	if (pfh->family == AF_INET6) {
635 		skb->csum = pfh->wcheck;
636 		skb->ip_summed = CHECKSUM_NONE;
637 		pfh->wcheck = 0;
638 	}
639 #endif
640 
641 	return 0;
642 }
643 EXPORT_SYMBOL_GPL(ping_getfrag);
644 
645 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
646 				       struct flowi4 *fl4)
647 {
648 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
649 
650 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
651 		sizeof(struct icmphdr), pfh->wcheck);
652 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
653 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
654 	skb->ip_summed = CHECKSUM_NONE;
655 	return ip_push_pending_frames(sk, fl4);
656 }
657 
658 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
659 			void *user_icmph, size_t icmph_len) {
660 	u8 type, code;
661 
662 	if (len > 0xFFFF)
663 		return -EMSGSIZE;
664 
665 	/*
666 	 *	Check the flags.
667 	 */
668 
669 	/* Mirror BSD error message compatibility */
670 	if (msg->msg_flags & MSG_OOB)
671 		return -EOPNOTSUPP;
672 
673 	/*
674 	 *	Fetch the ICMP header provided by the userland.
675 	 *	iovec is modified! The ICMP header is consumed.
676 	 */
677 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
678 		return -EFAULT;
679 
680 	if (family == AF_INET) {
681 		type = ((struct icmphdr *) user_icmph)->type;
682 		code = ((struct icmphdr *) user_icmph)->code;
683 #if IS_ENABLED(CONFIG_IPV6)
684 	} else if (family == AF_INET6) {
685 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
686 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
687 #endif
688 	} else {
689 		BUG();
690 	}
691 
692 	if (!ping_supported(family, type, code))
693 		return -EINVAL;
694 
695 	return 0;
696 }
697 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
698 
699 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
700 {
701 	struct net *net = sock_net(sk);
702 	struct flowi4 fl4;
703 	struct inet_sock *inet = inet_sk(sk);
704 	struct ipcm_cookie ipc;
705 	struct icmphdr user_icmph;
706 	struct pingfakehdr pfh;
707 	struct rtable *rt = NULL;
708 	struct ip_options_data opt_copy;
709 	int free = 0;
710 	__be32 saddr, daddr, faddr;
711 	u8  tos;
712 	int err;
713 
714 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
715 
716 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
717 				  sizeof(user_icmph));
718 	if (err)
719 		return err;
720 
721 	/*
722 	 *	Get and verify the address.
723 	 */
724 
725 	if (msg->msg_name) {
726 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
727 		if (msg->msg_namelen < sizeof(*usin))
728 			return -EINVAL;
729 		if (usin->sin_family != AF_INET)
730 			return -EAFNOSUPPORT;
731 		daddr = usin->sin_addr.s_addr;
732 		/* no remote port */
733 	} else {
734 		if (sk->sk_state != TCP_ESTABLISHED)
735 			return -EDESTADDRREQ;
736 		daddr = inet->inet_daddr;
737 		/* no remote port */
738 	}
739 
740 	ipc.sockc.tsflags = sk->sk_tsflags;
741 	ipc.addr = inet->inet_saddr;
742 	ipc.opt = NULL;
743 	ipc.oif = sk->sk_bound_dev_if;
744 	ipc.tx_flags = 0;
745 	ipc.ttl = 0;
746 	ipc.tos = -1;
747 
748 	if (msg->msg_controllen) {
749 		err = ip_cmsg_send(sk, msg, &ipc, false);
750 		if (unlikely(err)) {
751 			kfree(ipc.opt);
752 			return err;
753 		}
754 		if (ipc.opt)
755 			free = 1;
756 	}
757 	if (!ipc.opt) {
758 		struct ip_options_rcu *inet_opt;
759 
760 		rcu_read_lock();
761 		inet_opt = rcu_dereference(inet->inet_opt);
762 		if (inet_opt) {
763 			memcpy(&opt_copy, inet_opt,
764 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
765 			ipc.opt = &opt_copy.opt;
766 		}
767 		rcu_read_unlock();
768 	}
769 
770 	sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
771 
772 	saddr = ipc.addr;
773 	ipc.addr = faddr = daddr;
774 
775 	if (ipc.opt && ipc.opt->opt.srr) {
776 		if (!daddr)
777 			return -EINVAL;
778 		faddr = ipc.opt->opt.faddr;
779 	}
780 	tos = get_rttos(&ipc, inet);
781 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
782 	    (msg->msg_flags & MSG_DONTROUTE) ||
783 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
784 		tos |= RTO_ONLINK;
785 	}
786 
787 	if (ipv4_is_multicast(daddr)) {
788 		if (!ipc.oif)
789 			ipc.oif = inet->mc_index;
790 		if (!saddr)
791 			saddr = inet->mc_addr;
792 	} else if (!ipc.oif)
793 		ipc.oif = inet->uc_index;
794 
795 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
796 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
797 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
798 
799 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
800 	rt = ip_route_output_flow(net, &fl4, sk);
801 	if (IS_ERR(rt)) {
802 		err = PTR_ERR(rt);
803 		rt = NULL;
804 		if (err == -ENETUNREACH)
805 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
806 		goto out;
807 	}
808 
809 	err = -EACCES;
810 	if ((rt->rt_flags & RTCF_BROADCAST) &&
811 	    !sock_flag(sk, SOCK_BROADCAST))
812 		goto out;
813 
814 	if (msg->msg_flags & MSG_CONFIRM)
815 		goto do_confirm;
816 back_from_confirm:
817 
818 	if (!ipc.addr)
819 		ipc.addr = fl4.daddr;
820 
821 	lock_sock(sk);
822 
823 	pfh.icmph.type = user_icmph.type; /* already checked */
824 	pfh.icmph.code = user_icmph.code; /* ditto */
825 	pfh.icmph.checksum = 0;
826 	pfh.icmph.un.echo.id = inet->inet_sport;
827 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
828 	pfh.msg = msg;
829 	pfh.wcheck = 0;
830 	pfh.family = AF_INET;
831 
832 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
833 			0, &ipc, &rt, msg->msg_flags);
834 	if (err)
835 		ip_flush_pending_frames(sk);
836 	else
837 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
838 	release_sock(sk);
839 
840 out:
841 	ip_rt_put(rt);
842 	if (free)
843 		kfree(ipc.opt);
844 	if (!err) {
845 		icmp_out_count(sock_net(sk), user_icmph.type);
846 		return len;
847 	}
848 	return err;
849 
850 do_confirm:
851 	dst_confirm(&rt->dst);
852 	if (!(msg->msg_flags & MSG_PROBE) || len)
853 		goto back_from_confirm;
854 	err = 0;
855 	goto out;
856 }
857 
858 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
859 		 int flags, int *addr_len)
860 {
861 	struct inet_sock *isk = inet_sk(sk);
862 	int family = sk->sk_family;
863 	struct sk_buff *skb;
864 	int copied, err;
865 
866 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
867 
868 	err = -EOPNOTSUPP;
869 	if (flags & MSG_OOB)
870 		goto out;
871 
872 	if (flags & MSG_ERRQUEUE)
873 		return inet_recv_error(sk, msg, len, addr_len);
874 
875 	skb = skb_recv_datagram(sk, flags, noblock, &err);
876 	if (!skb)
877 		goto out;
878 
879 	copied = skb->len;
880 	if (copied > len) {
881 		msg->msg_flags |= MSG_TRUNC;
882 		copied = len;
883 	}
884 
885 	/* Don't bother checking the checksum */
886 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
887 	if (err)
888 		goto done;
889 
890 	sock_recv_timestamp(msg, sk, skb);
891 
892 	/* Copy the address and add cmsg data. */
893 	if (family == AF_INET) {
894 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
895 
896 		if (sin) {
897 			sin->sin_family = AF_INET;
898 			sin->sin_port = 0 /* skb->h.uh->source */;
899 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
900 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
901 			*addr_len = sizeof(*sin);
902 		}
903 
904 		if (isk->cmsg_flags)
905 			ip_cmsg_recv(msg, skb);
906 
907 #if IS_ENABLED(CONFIG_IPV6)
908 	} else if (family == AF_INET6) {
909 		struct ipv6_pinfo *np = inet6_sk(sk);
910 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
911 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
912 
913 		if (sin6) {
914 			sin6->sin6_family = AF_INET6;
915 			sin6->sin6_port = 0;
916 			sin6->sin6_addr = ip6->saddr;
917 			sin6->sin6_flowinfo = 0;
918 			if (np->sndflow)
919 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
920 			sin6->sin6_scope_id =
921 				ipv6_iface_scope_id(&sin6->sin6_addr,
922 						    inet6_iif(skb));
923 			*addr_len = sizeof(*sin6);
924 		}
925 
926 		if (inet6_sk(sk)->rxopt.all)
927 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
928 		if (skb->protocol == htons(ETH_P_IPV6) &&
929 		    inet6_sk(sk)->rxopt.all)
930 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
931 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
932 			ip_cmsg_recv(msg, skb);
933 #endif
934 	} else {
935 		BUG();
936 	}
937 
938 	err = copied;
939 
940 done:
941 	skb_free_datagram(sk, skb);
942 out:
943 	pr_debug("ping_recvmsg -> %d\n", err);
944 	return err;
945 }
946 EXPORT_SYMBOL_GPL(ping_recvmsg);
947 
948 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
949 {
950 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
951 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
952 	if (sock_queue_rcv_skb(sk, skb) < 0) {
953 		kfree_skb(skb);
954 		pr_debug("ping_queue_rcv_skb -> failed\n");
955 		return -1;
956 	}
957 	return 0;
958 }
959 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
960 
961 
962 /*
963  *	All we need to do is get the socket.
964  */
965 
966 bool ping_rcv(struct sk_buff *skb)
967 {
968 	struct sock *sk;
969 	struct net *net = dev_net(skb->dev);
970 	struct icmphdr *icmph = icmp_hdr(skb);
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)
986 			ping_queue_rcv_skb(sk, skb2);
987 		sock_put(sk);
988 		return true;
989 	}
990 	pr_debug("no socket, dropping\n");
991 
992 	return false;
993 }
994 EXPORT_SYMBOL_GPL(ping_rcv);
995 
996 struct proto ping_prot = {
997 	.name =		"PING",
998 	.owner =	THIS_MODULE,
999 	.init =		ping_init_sock,
1000 	.close =	ping_close,
1001 	.connect =	ip4_datagram_connect,
1002 	.disconnect =	udp_disconnect,
1003 	.setsockopt =	ip_setsockopt,
1004 	.getsockopt =	ip_getsockopt,
1005 	.sendmsg =	ping_v4_sendmsg,
1006 	.recvmsg =	ping_recvmsg,
1007 	.bind =		ping_bind,
1008 	.backlog_rcv =	ping_queue_rcv_skb,
1009 	.release_cb =	ip4_datagram_release_cb,
1010 	.hash =		ping_hash,
1011 	.unhash =	ping_unhash,
1012 	.get_port =	ping_get_port,
1013 	.obj_size =	sizeof(struct inet_sock),
1014 };
1015 EXPORT_SYMBOL(ping_prot);
1016 
1017 #ifdef CONFIG_PROC_FS
1018 
1019 static struct sock *ping_get_first(struct seq_file *seq, int start)
1020 {
1021 	struct sock *sk;
1022 	struct ping_iter_state *state = seq->private;
1023 	struct net *net = seq_file_net(seq);
1024 
1025 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1026 	     ++state->bucket) {
1027 		struct hlist_nulls_node *node;
1028 		struct hlist_nulls_head *hslot;
1029 
1030 		hslot = &ping_table.hash[state->bucket];
1031 
1032 		if (hlist_nulls_empty(hslot))
1033 			continue;
1034 
1035 		sk_nulls_for_each(sk, node, hslot) {
1036 			if (net_eq(sock_net(sk), net) &&
1037 			    sk->sk_family == state->family)
1038 				goto found;
1039 		}
1040 	}
1041 	sk = NULL;
1042 found:
1043 	return sk;
1044 }
1045 
1046 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1047 {
1048 	struct ping_iter_state *state = seq->private;
1049 	struct net *net = seq_file_net(seq);
1050 
1051 	do {
1052 		sk = sk_nulls_next(sk);
1053 	} while (sk && (!net_eq(sock_net(sk), net)));
1054 
1055 	if (!sk)
1056 		return ping_get_first(seq, state->bucket + 1);
1057 	return sk;
1058 }
1059 
1060 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1061 {
1062 	struct sock *sk = ping_get_first(seq, 0);
1063 
1064 	if (sk)
1065 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1066 			--pos;
1067 	return pos ? NULL : sk;
1068 }
1069 
1070 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1071 	__acquires(ping_table.lock)
1072 {
1073 	struct ping_iter_state *state = seq->private;
1074 	state->bucket = 0;
1075 	state->family = family;
1076 
1077 	read_lock_bh(&ping_table.lock);
1078 
1079 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1080 }
1081 EXPORT_SYMBOL_GPL(ping_seq_start);
1082 
1083 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1084 {
1085 	return ping_seq_start(seq, pos, AF_INET);
1086 }
1087 
1088 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1089 {
1090 	struct sock *sk;
1091 
1092 	if (v == SEQ_START_TOKEN)
1093 		sk = ping_get_idx(seq, 0);
1094 	else
1095 		sk = ping_get_next(seq, v);
1096 
1097 	++*pos;
1098 	return sk;
1099 }
1100 EXPORT_SYMBOL_GPL(ping_seq_next);
1101 
1102 void ping_seq_stop(struct seq_file *seq, void *v)
1103 	__releases(ping_table.lock)
1104 {
1105 	read_unlock_bh(&ping_table.lock);
1106 }
1107 EXPORT_SYMBOL_GPL(ping_seq_stop);
1108 
1109 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1110 		int bucket)
1111 {
1112 	struct inet_sock *inet = inet_sk(sp);
1113 	__be32 dest = inet->inet_daddr;
1114 	__be32 src = inet->inet_rcv_saddr;
1115 	__u16 destp = ntohs(inet->inet_dport);
1116 	__u16 srcp = ntohs(inet->inet_sport);
1117 
1118 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1119 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1120 		bucket, src, srcp, dest, destp, sp->sk_state,
1121 		sk_wmem_alloc_get(sp),
1122 		sk_rmem_alloc_get(sp),
1123 		0, 0L, 0,
1124 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1125 		0, sock_i_ino(sp),
1126 		atomic_read(&sp->sk_refcnt), sp,
1127 		atomic_read(&sp->sk_drops));
1128 }
1129 
1130 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1131 {
1132 	seq_setwidth(seq, 127);
1133 	if (v == SEQ_START_TOKEN)
1134 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1135 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1136 			   "inode ref pointer drops");
1137 	else {
1138 		struct ping_iter_state *state = seq->private;
1139 
1140 		ping_v4_format_sock(v, seq, state->bucket);
1141 	}
1142 	seq_pad(seq, '\n');
1143 	return 0;
1144 }
1145 
1146 static int ping_seq_open(struct inode *inode, struct file *file)
1147 {
1148 	struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1149 	return seq_open_net(inode, file, &afinfo->seq_ops,
1150 			   sizeof(struct ping_iter_state));
1151 }
1152 
1153 const struct file_operations ping_seq_fops = {
1154 	.open		= ping_seq_open,
1155 	.read		= seq_read,
1156 	.llseek		= seq_lseek,
1157 	.release	= seq_release_net,
1158 };
1159 EXPORT_SYMBOL_GPL(ping_seq_fops);
1160 
1161 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1162 	.name		= "icmp",
1163 	.family		= AF_INET,
1164 	.seq_fops	= &ping_seq_fops,
1165 	.seq_ops	= {
1166 		.start		= ping_v4_seq_start,
1167 		.show		= ping_v4_seq_show,
1168 		.next		= ping_seq_next,
1169 		.stop		= ping_seq_stop,
1170 	},
1171 };
1172 
1173 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1174 {
1175 	struct proc_dir_entry *p;
1176 	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1177 			     afinfo->seq_fops, afinfo);
1178 	if (!p)
1179 		return -ENOMEM;
1180 	return 0;
1181 }
1182 EXPORT_SYMBOL_GPL(ping_proc_register);
1183 
1184 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1185 {
1186 	remove_proc_entry(afinfo->name, net->proc_net);
1187 }
1188 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1189 
1190 static int __net_init ping_v4_proc_init_net(struct net *net)
1191 {
1192 	return ping_proc_register(net, &ping_v4_seq_afinfo);
1193 }
1194 
1195 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1196 {
1197 	ping_proc_unregister(net, &ping_v4_seq_afinfo);
1198 }
1199 
1200 static struct pernet_operations ping_v4_net_ops = {
1201 	.init = ping_v4_proc_init_net,
1202 	.exit = ping_v4_proc_exit_net,
1203 };
1204 
1205 int __init ping_proc_init(void)
1206 {
1207 	return register_pernet_subsys(&ping_v4_net_ops);
1208 }
1209 
1210 void ping_proc_exit(void)
1211 {
1212 	unregister_pernet_subsys(&ping_v4_net_ops);
1213 }
1214 
1215 #endif
1216 
1217 void __init ping_init(void)
1218 {
1219 	int i;
1220 
1221 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1222 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1223 	rwlock_init(&ping_table.lock);
1224 }
1225