xref: /openbmc/linux/net/ipv4/ping.c (revision a93fbb00)
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 		int chk_addr_ret;
309 
310 		if (addr_len < sizeof(*addr))
311 			return -EINVAL;
312 
313 		if (addr->sin_family != AF_INET &&
314 		    !(addr->sin_family == AF_UNSPEC &&
315 		      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
316 			return -EAFNOSUPPORT;
317 
318 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
319 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
320 
321 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
322 
323 		if (!inet_addr_valid_or_nonlocal(net, inet_sk(sk),
324 					         addr->sin_addr.s_addr,
325 	                                         chk_addr_ret))
326 			return -EADDRNOTAVAIL;
327 
328 #if IS_ENABLED(CONFIG_IPV6)
329 	} else if (sk->sk_family == AF_INET6) {
330 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
331 		int addr_type, scoped, has_addr;
332 		struct net_device *dev = NULL;
333 
334 		if (addr_len < sizeof(*addr))
335 			return -EINVAL;
336 
337 		if (addr->sin6_family != AF_INET6)
338 			return -EAFNOSUPPORT;
339 
340 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
341 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
342 
343 		addr_type = ipv6_addr_type(&addr->sin6_addr);
344 		scoped = __ipv6_addr_needs_scope_id(addr_type);
345 		if ((addr_type != IPV6_ADDR_ANY &&
346 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
347 		    (scoped && !addr->sin6_scope_id))
348 			return -EINVAL;
349 
350 		rcu_read_lock();
351 		if (addr->sin6_scope_id) {
352 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
353 			if (!dev) {
354 				rcu_read_unlock();
355 				return -ENODEV;
356 			}
357 		}
358 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
359 						    scoped);
360 		rcu_read_unlock();
361 
362 		if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
363 		      addr_type == IPV6_ADDR_ANY))
364 			return -EADDRNOTAVAIL;
365 
366 		if (scoped)
367 			sk->sk_bound_dev_if = addr->sin6_scope_id;
368 #endif
369 	} else {
370 		return -EAFNOSUPPORT;
371 	}
372 	return 0;
373 }
374 
375 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
376 {
377 	if (saddr->sa_family == AF_INET) {
378 		struct inet_sock *isk = inet_sk(sk);
379 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
380 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
381 #if IS_ENABLED(CONFIG_IPV6)
382 	} else if (saddr->sa_family == AF_INET6) {
383 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
384 		struct ipv6_pinfo *np = inet6_sk(sk);
385 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
386 #endif
387 	}
388 }
389 
390 /*
391  * We need our own bind because there are no privileged id's == local ports.
392  * Moreover, we don't allow binding to multi- and broadcast addresses.
393  */
394 
395 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
396 {
397 	struct inet_sock *isk = inet_sk(sk);
398 	unsigned short snum;
399 	int err;
400 	int dif = sk->sk_bound_dev_if;
401 
402 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
403 	if (err)
404 		return err;
405 
406 	lock_sock(sk);
407 
408 	err = -EINVAL;
409 	if (isk->inet_num != 0)
410 		goto out;
411 
412 	err = -EADDRINUSE;
413 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
414 	if (ping_get_port(sk, snum) != 0) {
415 		/* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
416 		sk->sk_bound_dev_if = dif;
417 		goto out;
418 	}
419 	ping_set_saddr(sk, uaddr);
420 
421 	pr_debug("after bind(): num = %hu, dif = %d\n",
422 		 isk->inet_num,
423 		 sk->sk_bound_dev_if);
424 
425 	err = 0;
426 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
427 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
428 #if IS_ENABLED(CONFIG_IPV6)
429 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
430 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
431 #endif
432 
433 	if (snum)
434 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
435 	isk->inet_sport = htons(isk->inet_num);
436 	isk->inet_daddr = 0;
437 	isk->inet_dport = 0;
438 
439 #if IS_ENABLED(CONFIG_IPV6)
440 	if (sk->sk_family == AF_INET6)
441 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
442 #endif
443 
444 	sk_dst_reset(sk);
445 out:
446 	release_sock(sk);
447 	pr_debug("ping_v4_bind -> %d\n", err);
448 	return err;
449 }
450 EXPORT_SYMBOL_GPL(ping_bind);
451 
452 /*
453  * Is this a supported type of ICMP message?
454  */
455 
456 static inline int ping_supported(int family, int type, int code)
457 {
458 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
459 	       (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
460 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
461 	       (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
462 }
463 
464 /*
465  * This routine is called by the ICMP module when it gets some
466  * sort of error condition.
467  */
468 
469 void ping_err(struct sk_buff *skb, int offset, u32 info)
470 {
471 	int family;
472 	struct icmphdr *icmph;
473 	struct inet_sock *inet_sock;
474 	int type;
475 	int code;
476 	struct net *net = dev_net(skb->dev);
477 	struct sock *sk;
478 	int harderr;
479 	int err;
480 
481 	if (skb->protocol == htons(ETH_P_IP)) {
482 		family = AF_INET;
483 		type = icmp_hdr(skb)->type;
484 		code = icmp_hdr(skb)->code;
485 		icmph = (struct icmphdr *)(skb->data + offset);
486 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
487 		family = AF_INET6;
488 		type = icmp6_hdr(skb)->icmp6_type;
489 		code = icmp6_hdr(skb)->icmp6_code;
490 		icmph = (struct icmphdr *) (skb->data + offset);
491 	} else {
492 		BUG();
493 	}
494 
495 	/* We assume the packet has already been checked by icmp_unreach */
496 
497 	if (!ping_supported(family, icmph->type, icmph->code))
498 		return;
499 
500 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
501 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
502 		 ntohs(icmph->un.echo.sequence));
503 
504 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
505 	if (!sk) {
506 		pr_debug("no socket, dropping\n");
507 		return;	/* No socket for error */
508 	}
509 	pr_debug("err on socket %p\n", sk);
510 
511 	err = 0;
512 	harderr = 0;
513 	inet_sock = inet_sk(sk);
514 
515 	if (skb->protocol == htons(ETH_P_IP)) {
516 		switch (type) {
517 		default:
518 		case ICMP_TIME_EXCEEDED:
519 			err = EHOSTUNREACH;
520 			break;
521 		case ICMP_SOURCE_QUENCH:
522 			/* This is not a real error but ping wants to see it.
523 			 * Report it with some fake errno.
524 			 */
525 			err = EREMOTEIO;
526 			break;
527 		case ICMP_PARAMETERPROB:
528 			err = EPROTO;
529 			harderr = 1;
530 			break;
531 		case ICMP_DEST_UNREACH:
532 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
533 				ipv4_sk_update_pmtu(skb, sk, info);
534 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
535 					err = EMSGSIZE;
536 					harderr = 1;
537 					break;
538 				}
539 				goto out;
540 			}
541 			err = EHOSTUNREACH;
542 			if (code <= NR_ICMP_UNREACH) {
543 				harderr = icmp_err_convert[code].fatal;
544 				err = icmp_err_convert[code].errno;
545 			}
546 			break;
547 		case ICMP_REDIRECT:
548 			/* See ICMP_SOURCE_QUENCH */
549 			ipv4_sk_redirect(skb, sk);
550 			err = EREMOTEIO;
551 			break;
552 		}
553 #if IS_ENABLED(CONFIG_IPV6)
554 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
555 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
556 #endif
557 	}
558 
559 	/*
560 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
561 	 *	4.1.3.3.
562 	 */
563 	if ((family == AF_INET && !inet_sock->recverr) ||
564 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
565 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
566 			goto out;
567 	} else {
568 		if (family == AF_INET) {
569 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
570 				      info, (u8 *)icmph);
571 #if IS_ENABLED(CONFIG_IPV6)
572 		} else if (family == AF_INET6) {
573 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
574 						   info, (u8 *)icmph);
575 #endif
576 		}
577 	}
578 	sk->sk_err = err;
579 	sk_error_report(sk);
580 out:
581 	sock_put(sk);
582 }
583 EXPORT_SYMBOL_GPL(ping_err);
584 
585 /*
586  *	Copy and checksum an ICMP Echo packet from user space into a buffer
587  *	starting from the payload.
588  */
589 
590 int ping_getfrag(void *from, char *to,
591 		 int offset, int fraglen, int odd, struct sk_buff *skb)
592 {
593 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
594 
595 	if (offset == 0) {
596 		fraglen -= sizeof(struct icmphdr);
597 		if (fraglen < 0)
598 			BUG();
599 		if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
600 			    fraglen, &pfh->wcheck,
601 			    &pfh->msg->msg_iter))
602 			return -EFAULT;
603 	} else if (offset < sizeof(struct icmphdr)) {
604 			BUG();
605 	} else {
606 		if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
607 					    &pfh->msg->msg_iter))
608 			return -EFAULT;
609 	}
610 
611 #if IS_ENABLED(CONFIG_IPV6)
612 	/* For IPv6, checksum each skb as we go along, as expected by
613 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
614 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
615 	 */
616 	if (pfh->family == AF_INET6) {
617 		skb->csum = pfh->wcheck;
618 		skb->ip_summed = CHECKSUM_NONE;
619 		pfh->wcheck = 0;
620 	}
621 #endif
622 
623 	return 0;
624 }
625 EXPORT_SYMBOL_GPL(ping_getfrag);
626 
627 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
628 				       struct flowi4 *fl4)
629 {
630 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
631 
632 	if (!skb)
633 		return 0;
634 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
635 		sizeof(struct icmphdr), pfh->wcheck);
636 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
637 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
638 	skb->ip_summed = CHECKSUM_NONE;
639 	return ip_push_pending_frames(sk, fl4);
640 }
641 
642 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
643 			void *user_icmph, size_t icmph_len)
644 {
645 	u8 type, code;
646 
647 	if (len > 0xFFFF)
648 		return -EMSGSIZE;
649 
650 	/* Must have at least a full ICMP header. */
651 	if (len < icmph_len)
652 		return -EINVAL;
653 
654 	/*
655 	 *	Check the flags.
656 	 */
657 
658 	/* Mirror BSD error message compatibility */
659 	if (msg->msg_flags & MSG_OOB)
660 		return -EOPNOTSUPP;
661 
662 	/*
663 	 *	Fetch the ICMP header provided by the userland.
664 	 *	iovec is modified! The ICMP header is consumed.
665 	 */
666 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
667 		return -EFAULT;
668 
669 	if (family == AF_INET) {
670 		type = ((struct icmphdr *) user_icmph)->type;
671 		code = ((struct icmphdr *) user_icmph)->code;
672 #if IS_ENABLED(CONFIG_IPV6)
673 	} else if (family == AF_INET6) {
674 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
675 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
676 #endif
677 	} else {
678 		BUG();
679 	}
680 
681 	if (!ping_supported(family, type, code))
682 		return -EINVAL;
683 
684 	return 0;
685 }
686 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
687 
688 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
689 {
690 	struct net *net = sock_net(sk);
691 	struct flowi4 fl4;
692 	struct inet_sock *inet = inet_sk(sk);
693 	struct ipcm_cookie ipc;
694 	struct icmphdr user_icmph;
695 	struct pingfakehdr pfh;
696 	struct rtable *rt = NULL;
697 	struct ip_options_data opt_copy;
698 	int free = 0;
699 	__be32 saddr, daddr, faddr;
700 	u8  tos;
701 	int err;
702 
703 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
704 
705 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
706 				  sizeof(user_icmph));
707 	if (err)
708 		return err;
709 
710 	/*
711 	 *	Get and verify the address.
712 	 */
713 
714 	if (msg->msg_name) {
715 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
716 		if (msg->msg_namelen < sizeof(*usin))
717 			return -EINVAL;
718 		if (usin->sin_family != AF_INET)
719 			return -EAFNOSUPPORT;
720 		daddr = usin->sin_addr.s_addr;
721 		/* no remote port */
722 	} else {
723 		if (sk->sk_state != TCP_ESTABLISHED)
724 			return -EDESTADDRREQ;
725 		daddr = inet->inet_daddr;
726 		/* no remote port */
727 	}
728 
729 	ipcm_init_sk(&ipc, inet);
730 
731 	if (msg->msg_controllen) {
732 		err = ip_cmsg_send(sk, msg, &ipc, false);
733 		if (unlikely(err)) {
734 			kfree(ipc.opt);
735 			return err;
736 		}
737 		if (ipc.opt)
738 			free = 1;
739 	}
740 	if (!ipc.opt) {
741 		struct ip_options_rcu *inet_opt;
742 
743 		rcu_read_lock();
744 		inet_opt = rcu_dereference(inet->inet_opt);
745 		if (inet_opt) {
746 			memcpy(&opt_copy, inet_opt,
747 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
748 			ipc.opt = &opt_copy.opt;
749 		}
750 		rcu_read_unlock();
751 	}
752 
753 	saddr = ipc.addr;
754 	ipc.addr = faddr = daddr;
755 
756 	if (ipc.opt && ipc.opt->opt.srr) {
757 		if (!daddr) {
758 			err = -EINVAL;
759 			goto out_free;
760 		}
761 		faddr = ipc.opt->opt.faddr;
762 	}
763 	tos = get_rttos(&ipc, inet);
764 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
765 	    (msg->msg_flags & MSG_DONTROUTE) ||
766 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
767 		tos |= RTO_ONLINK;
768 	}
769 
770 	if (ipv4_is_multicast(daddr)) {
771 		if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
772 			ipc.oif = inet->mc_index;
773 		if (!saddr)
774 			saddr = inet->mc_addr;
775 	} else if (!ipc.oif)
776 		ipc.oif = inet->uc_index;
777 
778 	flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
779 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
780 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
781 			   sk->sk_uid);
782 
783 	fl4.fl4_icmp_type = user_icmph.type;
784 	fl4.fl4_icmp_code = user_icmph.code;
785 
786 	security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
787 	rt = ip_route_output_flow(net, &fl4, sk);
788 	if (IS_ERR(rt)) {
789 		err = PTR_ERR(rt);
790 		rt = NULL;
791 		if (err == -ENETUNREACH)
792 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
793 		goto out;
794 	}
795 
796 	err = -EACCES;
797 	if ((rt->rt_flags & RTCF_BROADCAST) &&
798 	    !sock_flag(sk, SOCK_BROADCAST))
799 		goto out;
800 
801 	if (msg->msg_flags & MSG_CONFIRM)
802 		goto do_confirm;
803 back_from_confirm:
804 
805 	if (!ipc.addr)
806 		ipc.addr = fl4.daddr;
807 
808 	lock_sock(sk);
809 
810 	pfh.icmph.type = user_icmph.type; /* already checked */
811 	pfh.icmph.code = user_icmph.code; /* ditto */
812 	pfh.icmph.checksum = 0;
813 	pfh.icmph.un.echo.id = inet->inet_sport;
814 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
815 	pfh.msg = msg;
816 	pfh.wcheck = 0;
817 	pfh.family = AF_INET;
818 
819 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
820 			0, &ipc, &rt, msg->msg_flags);
821 	if (err)
822 		ip_flush_pending_frames(sk);
823 	else
824 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
825 	release_sock(sk);
826 
827 out:
828 	ip_rt_put(rt);
829 out_free:
830 	if (free)
831 		kfree(ipc.opt);
832 	if (!err) {
833 		icmp_out_count(sock_net(sk), user_icmph.type);
834 		return len;
835 	}
836 	return err;
837 
838 do_confirm:
839 	if (msg->msg_flags & MSG_PROBE)
840 		dst_confirm_neigh(&rt->dst, &fl4.daddr);
841 	if (!(msg->msg_flags & MSG_PROBE) || len)
842 		goto back_from_confirm;
843 	err = 0;
844 	goto out;
845 }
846 
847 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
848 		 int flags, int *addr_len)
849 {
850 	struct inet_sock *isk = inet_sk(sk);
851 	int family = sk->sk_family;
852 	struct sk_buff *skb;
853 	int copied, err;
854 
855 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
856 
857 	err = -EOPNOTSUPP;
858 	if (flags & MSG_OOB)
859 		goto out;
860 
861 	if (flags & MSG_ERRQUEUE)
862 		return inet_recv_error(sk, msg, len, addr_len);
863 
864 	skb = skb_recv_datagram(sk, flags, noblock, &err);
865 	if (!skb)
866 		goto out;
867 
868 	copied = skb->len;
869 	if (copied > len) {
870 		msg->msg_flags |= MSG_TRUNC;
871 		copied = len;
872 	}
873 
874 	/* Don't bother checking the checksum */
875 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
876 	if (err)
877 		goto done;
878 
879 	sock_recv_timestamp(msg, sk, skb);
880 
881 	/* Copy the address and add cmsg data. */
882 	if (family == AF_INET) {
883 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
884 
885 		if (sin) {
886 			sin->sin_family = AF_INET;
887 			sin->sin_port = 0 /* skb->h.uh->source */;
888 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
889 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
890 			*addr_len = sizeof(*sin);
891 		}
892 
893 		if (isk->cmsg_flags)
894 			ip_cmsg_recv(msg, skb);
895 
896 #if IS_ENABLED(CONFIG_IPV6)
897 	} else if (family == AF_INET6) {
898 		struct ipv6_pinfo *np = inet6_sk(sk);
899 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
900 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
901 
902 		if (sin6) {
903 			sin6->sin6_family = AF_INET6;
904 			sin6->sin6_port = 0;
905 			sin6->sin6_addr = ip6->saddr;
906 			sin6->sin6_flowinfo = 0;
907 			if (np->sndflow)
908 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
909 			sin6->sin6_scope_id =
910 				ipv6_iface_scope_id(&sin6->sin6_addr,
911 						    inet6_iif(skb));
912 			*addr_len = sizeof(*sin6);
913 		}
914 
915 		if (inet6_sk(sk)->rxopt.all)
916 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
917 		if (skb->protocol == htons(ETH_P_IPV6) &&
918 		    inet6_sk(sk)->rxopt.all)
919 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
920 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
921 			ip_cmsg_recv(msg, skb);
922 #endif
923 	} else {
924 		BUG();
925 	}
926 
927 	err = copied;
928 
929 done:
930 	skb_free_datagram(sk, skb);
931 out:
932 	pr_debug("ping_recvmsg -> %d\n", err);
933 	return err;
934 }
935 EXPORT_SYMBOL_GPL(ping_recvmsg);
936 
937 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
938 {
939 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
940 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
941 	if (sock_queue_rcv_skb(sk, skb) < 0) {
942 		kfree_skb(skb);
943 		pr_debug("ping_queue_rcv_skb -> failed\n");
944 		return -1;
945 	}
946 	return 0;
947 }
948 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
949 
950 
951 /*
952  *	All we need to do is get the socket.
953  */
954 
955 bool ping_rcv(struct sk_buff *skb)
956 {
957 	struct sock *sk;
958 	struct net *net = dev_net(skb->dev);
959 	struct icmphdr *icmph = icmp_hdr(skb);
960 	bool rc = false;
961 
962 	/* We assume the packet has already been checked by icmp_rcv */
963 
964 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
965 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
966 
967 	/* Push ICMP header back */
968 	skb_push(skb, skb->data - (u8 *)icmph);
969 
970 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
971 	if (sk) {
972 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
973 
974 		pr_debug("rcv on socket %p\n", sk);
975 		if (skb2 && !ping_queue_rcv_skb(sk, skb2))
976 			rc = true;
977 		sock_put(sk);
978 	}
979 
980 	if (!rc)
981 		pr_debug("no socket, dropping\n");
982 
983 	return rc;
984 }
985 EXPORT_SYMBOL_GPL(ping_rcv);
986 
987 struct proto ping_prot = {
988 	.name =		"PING",
989 	.owner =	THIS_MODULE,
990 	.init =		ping_init_sock,
991 	.close =	ping_close,
992 	.connect =	ip4_datagram_connect,
993 	.disconnect =	__udp_disconnect,
994 	.setsockopt =	ip_setsockopt,
995 	.getsockopt =	ip_getsockopt,
996 	.sendmsg =	ping_v4_sendmsg,
997 	.recvmsg =	ping_recvmsg,
998 	.bind =		ping_bind,
999 	.backlog_rcv =	ping_queue_rcv_skb,
1000 	.release_cb =	ip4_datagram_release_cb,
1001 	.hash =		ping_hash,
1002 	.unhash =	ping_unhash,
1003 	.get_port =	ping_get_port,
1004 	.put_port =	ping_unhash,
1005 	.obj_size =	sizeof(struct inet_sock),
1006 };
1007 EXPORT_SYMBOL(ping_prot);
1008 
1009 #ifdef CONFIG_PROC_FS
1010 
1011 static struct sock *ping_get_first(struct seq_file *seq, int start)
1012 {
1013 	struct sock *sk;
1014 	struct ping_iter_state *state = seq->private;
1015 	struct net *net = seq_file_net(seq);
1016 
1017 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1018 	     ++state->bucket) {
1019 		struct hlist_nulls_node *node;
1020 		struct hlist_nulls_head *hslot;
1021 
1022 		hslot = &ping_table.hash[state->bucket];
1023 
1024 		if (hlist_nulls_empty(hslot))
1025 			continue;
1026 
1027 		sk_nulls_for_each(sk, node, hslot) {
1028 			if (net_eq(sock_net(sk), net) &&
1029 			    sk->sk_family == state->family)
1030 				goto found;
1031 		}
1032 	}
1033 	sk = NULL;
1034 found:
1035 	return sk;
1036 }
1037 
1038 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1039 {
1040 	struct ping_iter_state *state = seq->private;
1041 	struct net *net = seq_file_net(seq);
1042 
1043 	do {
1044 		sk = sk_nulls_next(sk);
1045 	} while (sk && (!net_eq(sock_net(sk), net)));
1046 
1047 	if (!sk)
1048 		return ping_get_first(seq, state->bucket + 1);
1049 	return sk;
1050 }
1051 
1052 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1053 {
1054 	struct sock *sk = ping_get_first(seq, 0);
1055 
1056 	if (sk)
1057 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1058 			--pos;
1059 	return pos ? NULL : sk;
1060 }
1061 
1062 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1063 	__acquires(ping_table.lock)
1064 {
1065 	struct ping_iter_state *state = seq->private;
1066 	state->bucket = 0;
1067 	state->family = family;
1068 
1069 	read_lock_bh(&ping_table.lock);
1070 
1071 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1072 }
1073 EXPORT_SYMBOL_GPL(ping_seq_start);
1074 
1075 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1076 {
1077 	return ping_seq_start(seq, pos, AF_INET);
1078 }
1079 
1080 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1081 {
1082 	struct sock *sk;
1083 
1084 	if (v == SEQ_START_TOKEN)
1085 		sk = ping_get_idx(seq, 0);
1086 	else
1087 		sk = ping_get_next(seq, v);
1088 
1089 	++*pos;
1090 	return sk;
1091 }
1092 EXPORT_SYMBOL_GPL(ping_seq_next);
1093 
1094 void ping_seq_stop(struct seq_file *seq, void *v)
1095 	__releases(ping_table.lock)
1096 {
1097 	read_unlock_bh(&ping_table.lock);
1098 }
1099 EXPORT_SYMBOL_GPL(ping_seq_stop);
1100 
1101 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1102 		int bucket)
1103 {
1104 	struct inet_sock *inet = inet_sk(sp);
1105 	__be32 dest = inet->inet_daddr;
1106 	__be32 src = inet->inet_rcv_saddr;
1107 	__u16 destp = ntohs(inet->inet_dport);
1108 	__u16 srcp = ntohs(inet->inet_sport);
1109 
1110 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1111 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1112 		bucket, src, srcp, dest, destp, sp->sk_state,
1113 		sk_wmem_alloc_get(sp),
1114 		sk_rmem_alloc_get(sp),
1115 		0, 0L, 0,
1116 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1117 		0, sock_i_ino(sp),
1118 		refcount_read(&sp->sk_refcnt), sp,
1119 		atomic_read(&sp->sk_drops));
1120 }
1121 
1122 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1123 {
1124 	seq_setwidth(seq, 127);
1125 	if (v == SEQ_START_TOKEN)
1126 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1127 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1128 			   "inode ref pointer drops");
1129 	else {
1130 		struct ping_iter_state *state = seq->private;
1131 
1132 		ping_v4_format_sock(v, seq, state->bucket);
1133 	}
1134 	seq_pad(seq, '\n');
1135 	return 0;
1136 }
1137 
1138 static const struct seq_operations ping_v4_seq_ops = {
1139 	.start		= ping_v4_seq_start,
1140 	.show		= ping_v4_seq_show,
1141 	.next		= ping_seq_next,
1142 	.stop		= ping_seq_stop,
1143 };
1144 
1145 static int __net_init ping_v4_proc_init_net(struct net *net)
1146 {
1147 	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1148 			sizeof(struct ping_iter_state)))
1149 		return -ENOMEM;
1150 	return 0;
1151 }
1152 
1153 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1154 {
1155 	remove_proc_entry("icmp", net->proc_net);
1156 }
1157 
1158 static struct pernet_operations ping_v4_net_ops = {
1159 	.init = ping_v4_proc_init_net,
1160 	.exit = ping_v4_proc_exit_net,
1161 };
1162 
1163 int __init ping_proc_init(void)
1164 {
1165 	return register_pernet_subsys(&ping_v4_net_ops);
1166 }
1167 
1168 void ping_proc_exit(void)
1169 {
1170 	unregister_pernet_subsys(&ping_v4_net_ops);
1171 }
1172 
1173 #endif
1174 
1175 void __init ping_init(void)
1176 {
1177 	int i;
1178 
1179 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1180 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1181 	rwlock_init(&ping_table.lock);
1182 }
1183