xref: /openbmc/linux/net/ipv4/ping.c (revision 5104d265)
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 
57 struct ping_table ping_table;
58 struct pingv6_ops pingv6_ops;
59 EXPORT_SYMBOL_GPL(pingv6_ops);
60 
61 static u16 ping_port_rover;
62 
63 static inline int ping_hashfn(struct net *net, unsigned int num, unsigned int mask)
64 {
65 	int res = (num + net_hash_mix(net)) & mask;
66 
67 	pr_debug("hash(%d) = %d\n", num, res);
68 	return res;
69 }
70 EXPORT_SYMBOL_GPL(ping_hash);
71 
72 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
73 					     struct net *net, unsigned int num)
74 {
75 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
76 }
77 
78 int ping_get_port(struct sock *sk, unsigned short ident)
79 {
80 	struct hlist_nulls_node *node;
81 	struct hlist_nulls_head *hlist;
82 	struct inet_sock *isk, *isk2;
83 	struct sock *sk2 = NULL;
84 
85 	isk = inet_sk(sk);
86 	write_lock_bh(&ping_table.lock);
87 	if (ident == 0) {
88 		u32 i;
89 		u16 result = ping_port_rover + 1;
90 
91 		for (i = 0; i < (1L << 16); i++, result++) {
92 			if (!result)
93 				result++; /* avoid zero */
94 			hlist = ping_hashslot(&ping_table, sock_net(sk),
95 					    result);
96 			ping_portaddr_for_each_entry(sk2, node, hlist) {
97 				isk2 = inet_sk(sk2);
98 
99 				if (isk2->inet_num == result)
100 					goto next_port;
101 			}
102 
103 			/* found */
104 			ping_port_rover = ident = result;
105 			break;
106 next_port:
107 			;
108 		}
109 		if (i >= (1L << 16))
110 			goto fail;
111 	} else {
112 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
113 		ping_portaddr_for_each_entry(sk2, node, hlist) {
114 			isk2 = inet_sk(sk2);
115 
116 			/* BUG? Why is this reuse and not reuseaddr? ping.c
117 			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
118 			 * that other ping processes can steal its packets.
119 			 */
120 			if ((isk2->inet_num == ident) &&
121 			    (sk2 != sk) &&
122 			    (!sk2->sk_reuse || !sk->sk_reuse))
123 				goto fail;
124 		}
125 	}
126 
127 	pr_debug("found port/ident = %d\n", ident);
128 	isk->inet_num = ident;
129 	if (sk_unhashed(sk)) {
130 		pr_debug("was not hashed\n");
131 		sock_hold(sk);
132 		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
133 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
134 	}
135 	write_unlock_bh(&ping_table.lock);
136 	return 0;
137 
138 fail:
139 	write_unlock_bh(&ping_table.lock);
140 	return 1;
141 }
142 EXPORT_SYMBOL_GPL(ping_get_port);
143 
144 void ping_hash(struct sock *sk)
145 {
146 	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
147 	BUG(); /* "Please do not press this button again." */
148 }
149 
150 void ping_unhash(struct sock *sk)
151 {
152 	struct inet_sock *isk = inet_sk(sk);
153 	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
154 	if (sk_hashed(sk)) {
155 		write_lock_bh(&ping_table.lock);
156 		hlist_nulls_del(&sk->sk_nulls_node);
157 		sock_put(sk);
158 		isk->inet_num = 0;
159 		isk->inet_sport = 0;
160 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
161 		write_unlock_bh(&ping_table.lock);
162 	}
163 }
164 EXPORT_SYMBOL_GPL(ping_unhash);
165 
166 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
167 {
168 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
169 	struct sock *sk = NULL;
170 	struct inet_sock *isk;
171 	struct hlist_nulls_node *hnode;
172 	int dif = skb->dev->ifindex;
173 
174 	if (skb->protocol == htons(ETH_P_IP)) {
175 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
176 			 (int)ident, &ip_hdr(skb)->daddr, dif);
177 #if IS_ENABLED(CONFIG_IPV6)
178 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
179 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
180 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
181 #endif
182 	}
183 
184 	read_lock_bh(&ping_table.lock);
185 
186 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
187 		isk = inet_sk(sk);
188 
189 		pr_debug("iterate\n");
190 		if (isk->inet_num != ident)
191 			continue;
192 
193 		if (skb->protocol == htons(ETH_P_IP) &&
194 		    sk->sk_family == AF_INET) {
195 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
196 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
197 				 sk->sk_bound_dev_if);
198 
199 			if (isk->inet_rcv_saddr &&
200 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
201 				continue;
202 #if IS_ENABLED(CONFIG_IPV6)
203 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
204 			   sk->sk_family == AF_INET6) {
205 			struct ipv6_pinfo *np = inet6_sk(sk);
206 
207 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
208 				 (int) isk->inet_num,
209 				 &inet6_sk(sk)->rcv_saddr,
210 				 sk->sk_bound_dev_if);
211 
212 			if (!ipv6_addr_any(&np->rcv_saddr) &&
213 			    !ipv6_addr_equal(&np->rcv_saddr,
214 					     &ipv6_hdr(skb)->daddr))
215 				continue;
216 #endif
217 		}
218 
219 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
220 			continue;
221 
222 		sock_hold(sk);
223 		goto exit;
224 	}
225 
226 	sk = NULL;
227 exit:
228 	read_unlock_bh(&ping_table.lock);
229 
230 	return sk;
231 }
232 
233 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
234 					  kgid_t *high)
235 {
236 	kgid_t *data = net->ipv4.sysctl_ping_group_range;
237 	unsigned int seq;
238 
239 	do {
240 		seq = read_seqbegin(&sysctl_local_ports.lock);
241 
242 		*low = data[0];
243 		*high = data[1];
244 	} while (read_seqretry(&sysctl_local_ports.lock, seq));
245 }
246 
247 
248 int ping_init_sock(struct sock *sk)
249 {
250 	struct net *net = sock_net(sk);
251 	kgid_t group = current_egid();
252 	struct group_info *group_info = get_current_groups();
253 	int i, j, count = group_info->ngroups;
254 	kgid_t low, high;
255 
256 	inet_get_ping_group_range_net(net, &low, &high);
257 	if (gid_lte(low, group) && gid_lte(group, high))
258 		return 0;
259 
260 	for (i = 0; i < group_info->nblocks; i++) {
261 		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
262 		for (j = 0; j < cp_count; j++) {
263 			kgid_t gid = group_info->blocks[i][j];
264 			if (gid_lte(low, gid) && gid_lte(gid, high))
265 				return 0;
266 		}
267 
268 		count -= cp_count;
269 	}
270 
271 	return -EACCES;
272 }
273 EXPORT_SYMBOL_GPL(ping_init_sock);
274 
275 void ping_close(struct sock *sk, long timeout)
276 {
277 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
278 		 inet_sk(sk), inet_sk(sk)->inet_num);
279 	pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
280 
281 	sk_common_release(sk);
282 }
283 EXPORT_SYMBOL_GPL(ping_close);
284 
285 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
286 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
287 				struct sockaddr *uaddr, int addr_len) {
288 	struct net *net = sock_net(sk);
289 	if (sk->sk_family == AF_INET) {
290 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
291 		int chk_addr_ret;
292 
293 		if (addr_len < sizeof(*addr))
294 			return -EINVAL;
295 
296 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
297 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
298 
299 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
300 
301 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
302 			chk_addr_ret = RTN_LOCAL;
303 
304 		if ((sysctl_ip_nonlocal_bind == 0 &&
305 		    isk->freebind == 0 && isk->transparent == 0 &&
306 		     chk_addr_ret != RTN_LOCAL) ||
307 		    chk_addr_ret == RTN_MULTICAST ||
308 		    chk_addr_ret == RTN_BROADCAST)
309 			return -EADDRNOTAVAIL;
310 
311 #if IS_ENABLED(CONFIG_IPV6)
312 	} else if (sk->sk_family == AF_INET6) {
313 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
314 		int addr_type, scoped, has_addr;
315 		struct net_device *dev = NULL;
316 
317 		if (addr_len < sizeof(*addr))
318 			return -EINVAL;
319 
320 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
321 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
322 
323 		addr_type = ipv6_addr_type(&addr->sin6_addr);
324 		scoped = __ipv6_addr_needs_scope_id(addr_type);
325 		if ((addr_type != IPV6_ADDR_ANY &&
326 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
327 		    (scoped && !addr->sin6_scope_id))
328 			return -EINVAL;
329 
330 		rcu_read_lock();
331 		if (addr->sin6_scope_id) {
332 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
333 			if (!dev) {
334 				rcu_read_unlock();
335 				return -ENODEV;
336 			}
337 		}
338 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
339 						    scoped);
340 		rcu_read_unlock();
341 
342 		if (!(isk->freebind || isk->transparent || has_addr ||
343 		      addr_type == IPV6_ADDR_ANY))
344 			return -EADDRNOTAVAIL;
345 
346 		if (scoped)
347 			sk->sk_bound_dev_if = addr->sin6_scope_id;
348 #endif
349 	} else {
350 		return -EAFNOSUPPORT;
351 	}
352 	return 0;
353 }
354 
355 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
356 {
357 	if (saddr->sa_family == AF_INET) {
358 		struct inet_sock *isk = inet_sk(sk);
359 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
360 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
361 #if IS_ENABLED(CONFIG_IPV6)
362 	} else if (saddr->sa_family == AF_INET6) {
363 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
364 		struct ipv6_pinfo *np = inet6_sk(sk);
365 		np->rcv_saddr = np->saddr = addr->sin6_addr;
366 #endif
367 	}
368 }
369 
370 static void ping_clear_saddr(struct sock *sk, int dif)
371 {
372 	sk->sk_bound_dev_if = dif;
373 	if (sk->sk_family == AF_INET) {
374 		struct inet_sock *isk = inet_sk(sk);
375 		isk->inet_rcv_saddr = isk->inet_saddr = 0;
376 #if IS_ENABLED(CONFIG_IPV6)
377 	} else if (sk->sk_family == AF_INET6) {
378 		struct ipv6_pinfo *np = inet6_sk(sk);
379 		memset(&np->rcv_saddr, 0, sizeof(np->rcv_saddr));
380 		memset(&np->saddr, 0, sizeof(np->saddr));
381 #endif
382 	}
383 }
384 /*
385  * We need our own bind because there are no privileged id's == local ports.
386  * Moreover, we don't allow binding to multi- and broadcast addresses.
387  */
388 
389 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
390 {
391 	struct inet_sock *isk = inet_sk(sk);
392 	unsigned short snum;
393 	int err;
394 	int dif = sk->sk_bound_dev_if;
395 
396 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
397 	if (err)
398 		return err;
399 
400 	lock_sock(sk);
401 
402 	err = -EINVAL;
403 	if (isk->inet_num != 0)
404 		goto out;
405 
406 	err = -EADDRINUSE;
407 	ping_set_saddr(sk, uaddr);
408 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
409 	if (ping_get_port(sk, snum) != 0) {
410 		ping_clear_saddr(sk, dif);
411 		goto out;
412 	}
413 
414 	pr_debug("after bind(): num = %d, dif = %d\n",
415 		 (int)isk->inet_num,
416 		 (int)sk->sk_bound_dev_if);
417 
418 	err = 0;
419 	if ((sk->sk_family == AF_INET && isk->inet_rcv_saddr) ||
420 	    (sk->sk_family == AF_INET6 &&
421 	     !ipv6_addr_any(&inet6_sk(sk)->rcv_saddr)))
422 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
423 
424 	if (snum)
425 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
426 	isk->inet_sport = htons(isk->inet_num);
427 	isk->inet_daddr = 0;
428 	isk->inet_dport = 0;
429 
430 #if IS_ENABLED(CONFIG_IPV6)
431 	if (sk->sk_family == AF_INET6)
432 		memset(&inet6_sk(sk)->daddr, 0, sizeof(inet6_sk(sk)->daddr));
433 #endif
434 
435 	sk_dst_reset(sk);
436 out:
437 	release_sock(sk);
438 	pr_debug("ping_v4_bind -> %d\n", err);
439 	return err;
440 }
441 EXPORT_SYMBOL_GPL(ping_bind);
442 
443 /*
444  * Is this a supported type of ICMP message?
445  */
446 
447 static inline int ping_supported(int family, int type, int code)
448 {
449 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
450 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
451 }
452 
453 /*
454  * This routine is called by the ICMP module when it gets some
455  * sort of error condition.
456  */
457 
458 void ping_err(struct sk_buff *skb, int offset, u32 info)
459 {
460 	int family;
461 	struct icmphdr *icmph;
462 	struct inet_sock *inet_sock;
463 	int type;
464 	int code;
465 	struct net *net = dev_net(skb->dev);
466 	struct sock *sk;
467 	int harderr;
468 	int err;
469 
470 	if (skb->protocol == htons(ETH_P_IP)) {
471 		family = AF_INET;
472 		type = icmp_hdr(skb)->type;
473 		code = icmp_hdr(skb)->code;
474 		icmph = (struct icmphdr *)(skb->data + offset);
475 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
476 		family = AF_INET6;
477 		type = icmp6_hdr(skb)->icmp6_type;
478 		code = icmp6_hdr(skb)->icmp6_code;
479 		icmph = (struct icmphdr *) (skb->data + offset);
480 	} else {
481 		BUG();
482 	}
483 
484 	/* We assume the packet has already been checked by icmp_unreach */
485 
486 	if (!ping_supported(family, icmph->type, icmph->code))
487 		return;
488 
489 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
490 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
491 		 ntohs(icmph->un.echo.sequence));
492 
493 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
494 	if (sk == NULL) {
495 		pr_debug("no socket, dropping\n");
496 		return;	/* No socket for error */
497 	}
498 	pr_debug("err on socket %p\n", sk);
499 
500 	err = 0;
501 	harderr = 0;
502 	inet_sock = inet_sk(sk);
503 
504 	if (skb->protocol == htons(ETH_P_IP)) {
505 		switch (type) {
506 		default:
507 		case ICMP_TIME_EXCEEDED:
508 			err = EHOSTUNREACH;
509 			break;
510 		case ICMP_SOURCE_QUENCH:
511 			/* This is not a real error but ping wants to see it.
512 			 * Report it with some fake errno.
513 			 */
514 			err = EREMOTEIO;
515 			break;
516 		case ICMP_PARAMETERPROB:
517 			err = EPROTO;
518 			harderr = 1;
519 			break;
520 		case ICMP_DEST_UNREACH:
521 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
522 				ipv4_sk_update_pmtu(skb, sk, info);
523 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
524 					err = EMSGSIZE;
525 					harderr = 1;
526 					break;
527 				}
528 				goto out;
529 			}
530 			err = EHOSTUNREACH;
531 			if (code <= NR_ICMP_UNREACH) {
532 				harderr = icmp_err_convert[code].fatal;
533 				err = icmp_err_convert[code].errno;
534 			}
535 			break;
536 		case ICMP_REDIRECT:
537 			/* See ICMP_SOURCE_QUENCH */
538 			ipv4_sk_redirect(skb, sk);
539 			err = EREMOTEIO;
540 			break;
541 		}
542 #if IS_ENABLED(CONFIG_IPV6)
543 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
544 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
545 #endif
546 	}
547 
548 	/*
549 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
550 	 *	4.1.3.3.
551 	 */
552 	if ((family == AF_INET && !inet_sock->recverr) ||
553 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
554 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
555 			goto out;
556 	} else {
557 		if (family == AF_INET) {
558 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
559 				      info, (u8 *)icmph);
560 #if IS_ENABLED(CONFIG_IPV6)
561 		} else if (family == AF_INET6) {
562 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
563 						   info, (u8 *)icmph);
564 #endif
565 		}
566 	}
567 	sk->sk_err = err;
568 	sk->sk_error_report(sk);
569 out:
570 	sock_put(sk);
571 }
572 EXPORT_SYMBOL_GPL(ping_err);
573 
574 /*
575  *	Copy and checksum an ICMP Echo packet from user space into a buffer
576  *	starting from the payload.
577  */
578 
579 int ping_getfrag(void *from, char *to,
580 		 int offset, int fraglen, int odd, struct sk_buff *skb)
581 {
582 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
583 
584 	if (offset == 0) {
585 		if (fraglen < sizeof(struct icmphdr))
586 			BUG();
587 		if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
588 			    pfh->iov, 0, fraglen - sizeof(struct icmphdr),
589 			    &pfh->wcheck))
590 			return -EFAULT;
591 	} else if (offset < sizeof(struct icmphdr)) {
592 			BUG();
593 	} else {
594 		if (csum_partial_copy_fromiovecend
595 				(to, pfh->iov, offset - sizeof(struct icmphdr),
596 				 fraglen, &pfh->wcheck))
597 			return -EFAULT;
598 	}
599 
600 #if IS_ENABLED(CONFIG_IPV6)
601 	/* For IPv6, checksum each skb as we go along, as expected by
602 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
603 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
604 	 */
605 	if (pfh->family == AF_INET6) {
606 		skb->csum = pfh->wcheck;
607 		skb->ip_summed = CHECKSUM_NONE;
608 		pfh->wcheck = 0;
609 	}
610 #endif
611 
612 	return 0;
613 }
614 EXPORT_SYMBOL_GPL(ping_getfrag);
615 
616 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
617 				       struct flowi4 *fl4)
618 {
619 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
620 
621 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
622 		sizeof(struct icmphdr), pfh->wcheck);
623 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
624 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
625 	skb->ip_summed = CHECKSUM_NONE;
626 	return ip_push_pending_frames(sk, fl4);
627 }
628 
629 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
630 			void *user_icmph, size_t icmph_len) {
631 	u8 type, code;
632 
633 	if (len > 0xFFFF)
634 		return -EMSGSIZE;
635 
636 	/*
637 	 *	Check the flags.
638 	 */
639 
640 	/* Mirror BSD error message compatibility */
641 	if (msg->msg_flags & MSG_OOB)
642 		return -EOPNOTSUPP;
643 
644 	/*
645 	 *	Fetch the ICMP header provided by the userland.
646 	 *	iovec is modified! The ICMP header is consumed.
647 	 */
648 	if (memcpy_fromiovec(user_icmph, msg->msg_iov, icmph_len))
649 		return -EFAULT;
650 
651 	if (family == AF_INET) {
652 		type = ((struct icmphdr *) user_icmph)->type;
653 		code = ((struct icmphdr *) user_icmph)->code;
654 #if IS_ENABLED(CONFIG_IPV6)
655 	} else if (family == AF_INET6) {
656 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
657 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
658 #endif
659 	} else {
660 		BUG();
661 	}
662 
663 	if (!ping_supported(family, type, code))
664 		return -EINVAL;
665 
666 	return 0;
667 }
668 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
669 
670 int ping_v4_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
671 		    size_t len)
672 {
673 	struct net *net = sock_net(sk);
674 	struct flowi4 fl4;
675 	struct inet_sock *inet = inet_sk(sk);
676 	struct ipcm_cookie ipc;
677 	struct icmphdr user_icmph;
678 	struct pingfakehdr pfh;
679 	struct rtable *rt = NULL;
680 	struct ip_options_data opt_copy;
681 	int free = 0;
682 	__be32 saddr, daddr, faddr;
683 	u8  tos;
684 	int err;
685 
686 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
687 
688 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
689 				  sizeof(user_icmph));
690 	if (err)
691 		return err;
692 
693 	/*
694 	 *	Get and verify the address.
695 	 */
696 
697 	if (msg->msg_name) {
698 		struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
699 		if (msg->msg_namelen < sizeof(*usin))
700 			return -EINVAL;
701 		if (usin->sin_family != AF_INET)
702 			return -EINVAL;
703 		daddr = usin->sin_addr.s_addr;
704 		/* no remote port */
705 	} else {
706 		if (sk->sk_state != TCP_ESTABLISHED)
707 			return -EDESTADDRREQ;
708 		daddr = inet->inet_daddr;
709 		/* no remote port */
710 	}
711 
712 	ipc.addr = inet->inet_saddr;
713 	ipc.opt = NULL;
714 	ipc.oif = sk->sk_bound_dev_if;
715 	ipc.tx_flags = 0;
716 
717 	sock_tx_timestamp(sk, &ipc.tx_flags);
718 
719 	if (msg->msg_controllen) {
720 		err = ip_cmsg_send(sock_net(sk), msg, &ipc);
721 		if (err)
722 			return err;
723 		if (ipc.opt)
724 			free = 1;
725 	}
726 	if (!ipc.opt) {
727 		struct ip_options_rcu *inet_opt;
728 
729 		rcu_read_lock();
730 		inet_opt = rcu_dereference(inet->inet_opt);
731 		if (inet_opt) {
732 			memcpy(&opt_copy, inet_opt,
733 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
734 			ipc.opt = &opt_copy.opt;
735 		}
736 		rcu_read_unlock();
737 	}
738 
739 	saddr = ipc.addr;
740 	ipc.addr = faddr = daddr;
741 
742 	if (ipc.opt && ipc.opt->opt.srr) {
743 		if (!daddr)
744 			return -EINVAL;
745 		faddr = ipc.opt->opt.faddr;
746 	}
747 	tos = RT_TOS(inet->tos);
748 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
749 	    (msg->msg_flags & MSG_DONTROUTE) ||
750 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
751 		tos |= RTO_ONLINK;
752 	}
753 
754 	if (ipv4_is_multicast(daddr)) {
755 		if (!ipc.oif)
756 			ipc.oif = inet->mc_index;
757 		if (!saddr)
758 			saddr = inet->mc_addr;
759 	} else if (!ipc.oif)
760 		ipc.oif = inet->uc_index;
761 
762 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
763 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
764 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
765 
766 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
767 	rt = ip_route_output_flow(net, &fl4, sk);
768 	if (IS_ERR(rt)) {
769 		err = PTR_ERR(rt);
770 		rt = NULL;
771 		if (err == -ENETUNREACH)
772 			IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
773 		goto out;
774 	}
775 
776 	err = -EACCES;
777 	if ((rt->rt_flags & RTCF_BROADCAST) &&
778 	    !sock_flag(sk, SOCK_BROADCAST))
779 		goto out;
780 
781 	if (msg->msg_flags & MSG_CONFIRM)
782 		goto do_confirm;
783 back_from_confirm:
784 
785 	if (!ipc.addr)
786 		ipc.addr = fl4.daddr;
787 
788 	lock_sock(sk);
789 
790 	pfh.icmph.type = user_icmph.type; /* already checked */
791 	pfh.icmph.code = user_icmph.code; /* ditto */
792 	pfh.icmph.checksum = 0;
793 	pfh.icmph.un.echo.id = inet->inet_sport;
794 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
795 	pfh.iov = msg->msg_iov;
796 	pfh.wcheck = 0;
797 	pfh.family = AF_INET;
798 
799 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
800 			0, &ipc, &rt, msg->msg_flags);
801 	if (err)
802 		ip_flush_pending_frames(sk);
803 	else
804 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
805 	release_sock(sk);
806 
807 out:
808 	ip_rt_put(rt);
809 	if (free)
810 		kfree(ipc.opt);
811 	if (!err) {
812 		icmp_out_count(sock_net(sk), user_icmph.type);
813 		return len;
814 	}
815 	return err;
816 
817 do_confirm:
818 	dst_confirm(&rt->dst);
819 	if (!(msg->msg_flags & MSG_PROBE) || len)
820 		goto back_from_confirm;
821 	err = 0;
822 	goto out;
823 }
824 
825 int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
826 		 size_t len, int noblock, int flags, int *addr_len)
827 {
828 	struct inet_sock *isk = inet_sk(sk);
829 	int family = sk->sk_family;
830 	struct sockaddr_in *sin;
831 	struct sockaddr_in6 *sin6;
832 	struct sk_buff *skb;
833 	int copied, err;
834 
835 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
836 
837 	err = -EOPNOTSUPP;
838 	if (flags & MSG_OOB)
839 		goto out;
840 
841 	if (addr_len) {
842 		if (family == AF_INET)
843 			*addr_len = sizeof(*sin);
844 		else if (family == AF_INET6 && addr_len)
845 			*addr_len = sizeof(*sin6);
846 	}
847 
848 	if (flags & MSG_ERRQUEUE) {
849 		if (family == AF_INET) {
850 			return ip_recv_error(sk, msg, len);
851 #if IS_ENABLED(CONFIG_IPV6)
852 		} else if (family == AF_INET6) {
853 			return pingv6_ops.ipv6_recv_error(sk, msg, len);
854 #endif
855 		}
856 	}
857 
858 	skb = skb_recv_datagram(sk, flags, noblock, &err);
859 	if (!skb)
860 		goto out;
861 
862 	copied = skb->len;
863 	if (copied > len) {
864 		msg->msg_flags |= MSG_TRUNC;
865 		copied = len;
866 	}
867 
868 	/* Don't bother checking the checksum */
869 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
870 	if (err)
871 		goto done;
872 
873 	sock_recv_timestamp(msg, sk, skb);
874 
875 	/* Copy the address and add cmsg data. */
876 	if (family == AF_INET) {
877 		sin = (struct sockaddr_in *) msg->msg_name;
878 		sin->sin_family = AF_INET;
879 		sin->sin_port = 0 /* skb->h.uh->source */;
880 		sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
881 		memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
882 
883 		if (isk->cmsg_flags)
884 			ip_cmsg_recv(msg, skb);
885 
886 #if IS_ENABLED(CONFIG_IPV6)
887 	} else if (family == AF_INET6) {
888 		struct ipv6_pinfo *np = inet6_sk(sk);
889 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
890 		sin6 = (struct sockaddr_in6 *) msg->msg_name;
891 		sin6->sin6_family = AF_INET6;
892 		sin6->sin6_port = 0;
893 		sin6->sin6_addr = ip6->saddr;
894 
895 		sin6->sin6_flowinfo = 0;
896 		if (np->sndflow)
897 			sin6->sin6_flowinfo = ip6_flowinfo(ip6);
898 
899 		sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
900 							  IP6CB(skb)->iif);
901 
902 		if (inet6_sk(sk)->rxopt.all)
903 			pingv6_ops.ip6_datagram_recv_ctl(sk, msg, skb);
904 #endif
905 	} else {
906 		BUG();
907 	}
908 
909 	err = copied;
910 
911 done:
912 	skb_free_datagram(sk, skb);
913 out:
914 	pr_debug("ping_recvmsg -> %d\n", err);
915 	return err;
916 }
917 EXPORT_SYMBOL_GPL(ping_recvmsg);
918 
919 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
920 {
921 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
922 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
923 	if (sock_queue_rcv_skb(sk, skb) < 0) {
924 		kfree_skb(skb);
925 		pr_debug("ping_queue_rcv_skb -> failed\n");
926 		return -1;
927 	}
928 	return 0;
929 }
930 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
931 
932 
933 /*
934  *	All we need to do is get the socket.
935  */
936 
937 void ping_rcv(struct sk_buff *skb)
938 {
939 	struct sock *sk;
940 	struct net *net = dev_net(skb->dev);
941 	struct icmphdr *icmph = icmp_hdr(skb);
942 
943 	/* We assume the packet has already been checked by icmp_rcv */
944 
945 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
946 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
947 
948 	/* Push ICMP header back */
949 	skb_push(skb, skb->data - (u8 *)icmph);
950 
951 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
952 	if (sk != NULL) {
953 		pr_debug("rcv on socket %p\n", sk);
954 		ping_queue_rcv_skb(sk, skb_get(skb));
955 		sock_put(sk);
956 		return;
957 	}
958 	pr_debug("no socket, dropping\n");
959 
960 	/* We're called from icmp_rcv(). kfree_skb() is done there. */
961 }
962 EXPORT_SYMBOL_GPL(ping_rcv);
963 
964 struct proto ping_prot = {
965 	.name =		"PING",
966 	.owner =	THIS_MODULE,
967 	.init =		ping_init_sock,
968 	.close =	ping_close,
969 	.connect =	ip4_datagram_connect,
970 	.disconnect =	udp_disconnect,
971 	.setsockopt =	ip_setsockopt,
972 	.getsockopt =	ip_getsockopt,
973 	.sendmsg =	ping_v4_sendmsg,
974 	.recvmsg =	ping_recvmsg,
975 	.bind =		ping_bind,
976 	.backlog_rcv =	ping_queue_rcv_skb,
977 	.release_cb =	ip4_datagram_release_cb,
978 	.hash =		ping_hash,
979 	.unhash =	ping_unhash,
980 	.get_port =	ping_get_port,
981 	.obj_size =	sizeof(struct inet_sock),
982 };
983 EXPORT_SYMBOL(ping_prot);
984 
985 #ifdef CONFIG_PROC_FS
986 
987 static struct sock *ping_get_first(struct seq_file *seq, int start)
988 {
989 	struct sock *sk;
990 	struct ping_iter_state *state = seq->private;
991 	struct net *net = seq_file_net(seq);
992 
993 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
994 	     ++state->bucket) {
995 		struct hlist_nulls_node *node;
996 		struct hlist_nulls_head *hslot;
997 
998 		hslot = &ping_table.hash[state->bucket];
999 
1000 		if (hlist_nulls_empty(hslot))
1001 			continue;
1002 
1003 		sk_nulls_for_each(sk, node, hslot) {
1004 			if (net_eq(sock_net(sk), net) &&
1005 			    sk->sk_family == state->family)
1006 				goto found;
1007 		}
1008 	}
1009 	sk = NULL;
1010 found:
1011 	return sk;
1012 }
1013 
1014 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1015 {
1016 	struct ping_iter_state *state = seq->private;
1017 	struct net *net = seq_file_net(seq);
1018 
1019 	do {
1020 		sk = sk_nulls_next(sk);
1021 	} while (sk && (!net_eq(sock_net(sk), net)));
1022 
1023 	if (!sk)
1024 		return ping_get_first(seq, state->bucket + 1);
1025 	return sk;
1026 }
1027 
1028 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1029 {
1030 	struct sock *sk = ping_get_first(seq, 0);
1031 
1032 	if (sk)
1033 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1034 			--pos;
1035 	return pos ? NULL : sk;
1036 }
1037 
1038 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1039 {
1040 	struct ping_iter_state *state = seq->private;
1041 	state->bucket = 0;
1042 	state->family = family;
1043 
1044 	read_lock_bh(&ping_table.lock);
1045 
1046 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1047 }
1048 EXPORT_SYMBOL_GPL(ping_seq_start);
1049 
1050 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1051 {
1052 	return ping_seq_start(seq, pos, AF_INET);
1053 }
1054 
1055 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1056 {
1057 	struct sock *sk;
1058 
1059 	if (v == SEQ_START_TOKEN)
1060 		sk = ping_get_idx(seq, 0);
1061 	else
1062 		sk = ping_get_next(seq, v);
1063 
1064 	++*pos;
1065 	return sk;
1066 }
1067 EXPORT_SYMBOL_GPL(ping_seq_next);
1068 
1069 void ping_seq_stop(struct seq_file *seq, void *v)
1070 {
1071 	read_unlock_bh(&ping_table.lock);
1072 }
1073 EXPORT_SYMBOL_GPL(ping_seq_stop);
1074 
1075 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1076 		int bucket, int *len)
1077 {
1078 	struct inet_sock *inet = inet_sk(sp);
1079 	__be32 dest = inet->inet_daddr;
1080 	__be32 src = inet->inet_rcv_saddr;
1081 	__u16 destp = ntohs(inet->inet_dport);
1082 	__u16 srcp = ntohs(inet->inet_sport);
1083 
1084 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1085 		" %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
1086 		bucket, src, srcp, dest, destp, sp->sk_state,
1087 		sk_wmem_alloc_get(sp),
1088 		sk_rmem_alloc_get(sp),
1089 		0, 0L, 0,
1090 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1091 		0, sock_i_ino(sp),
1092 		atomic_read(&sp->sk_refcnt), sp,
1093 		atomic_read(&sp->sk_drops), len);
1094 }
1095 
1096 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1097 {
1098 	if (v == SEQ_START_TOKEN)
1099 		seq_printf(seq, "%-127s\n",
1100 			   "  sl  local_address rem_address   st tx_queue "
1101 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1102 			   "inode ref pointer drops");
1103 	else {
1104 		struct ping_iter_state *state = seq->private;
1105 		int len;
1106 
1107 		ping_v4_format_sock(v, seq, state->bucket, &len);
1108 		seq_printf(seq, "%*s\n", 127 - len, "");
1109 	}
1110 	return 0;
1111 }
1112 
1113 static const struct seq_operations ping_v4_seq_ops = {
1114 	.show		= ping_v4_seq_show,
1115 	.start		= ping_v4_seq_start,
1116 	.next		= ping_seq_next,
1117 	.stop		= ping_seq_stop,
1118 };
1119 
1120 static int ping_seq_open(struct inode *inode, struct file *file)
1121 {
1122 	struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1123 	return seq_open_net(inode, file, &afinfo->seq_ops,
1124 			   sizeof(struct ping_iter_state));
1125 }
1126 
1127 const struct file_operations ping_seq_fops = {
1128 	.open		= ping_seq_open,
1129 	.read		= seq_read,
1130 	.llseek		= seq_lseek,
1131 	.release	= seq_release_net,
1132 };
1133 EXPORT_SYMBOL_GPL(ping_seq_fops);
1134 
1135 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1136 	.name		= "icmp",
1137 	.family		= AF_INET,
1138 	.seq_fops	= &ping_seq_fops,
1139 	.seq_ops	= {
1140 		.start		= ping_v4_seq_start,
1141 		.show		= ping_v4_seq_show,
1142 		.next		= ping_seq_next,
1143 		.stop		= ping_seq_stop,
1144 	},
1145 };
1146 
1147 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1148 {
1149 	struct proc_dir_entry *p;
1150 	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1151 			     afinfo->seq_fops, afinfo);
1152 	if (!p)
1153 		return -ENOMEM;
1154 	return 0;
1155 }
1156 EXPORT_SYMBOL_GPL(ping_proc_register);
1157 
1158 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1159 {
1160 	remove_proc_entry(afinfo->name, net->proc_net);
1161 }
1162 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1163 
1164 static int __net_init ping_v4_proc_init_net(struct net *net)
1165 {
1166 	return ping_proc_register(net, &ping_v4_seq_afinfo);
1167 }
1168 
1169 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1170 {
1171 	ping_proc_unregister(net, &ping_v4_seq_afinfo);
1172 }
1173 
1174 static struct pernet_operations ping_v4_net_ops = {
1175 	.init = ping_v4_proc_init_net,
1176 	.exit = ping_v4_proc_exit_net,
1177 };
1178 
1179 int __init ping_proc_init(void)
1180 {
1181 	return register_pernet_subsys(&ping_v4_net_ops);
1182 }
1183 
1184 void ping_proc_exit(void)
1185 {
1186 	unregister_pernet_subsys(&ping_v4_net_ops);
1187 }
1188 
1189 #endif
1190 
1191 void __init ping_init(void)
1192 {
1193 	int i;
1194 
1195 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1196 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1197 	rwlock_init(&ping_table.lock);
1198 }
1199