xref: /openbmc/linux/net/ipv4/ping.c (revision aeefc1a0)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		"Ping" sockets
8  *
9  * Based on ipv4/udp.c code.
10  *
11  * Authors:	Vasiliy Kulikov / Openwall (for Linux 2.6),
12  *		Pavel Kankovsky (for Linux 2.4.32)
13  *
14  * Pavel gave all rights to bugs to Vasiliy,
15  * none of the bugs are Pavel's now.
16  */
17 
18 #include <linux/uaccess.h>
19 #include <linux/types.h>
20 #include <linux/fcntl.h>
21 #include <linux/socket.h>
22 #include <linux/sockios.h>
23 #include <linux/in.h>
24 #include <linux/errno.h>
25 #include <linux/timer.h>
26 #include <linux/mm.h>
27 #include <linux/inet.h>
28 #include <linux/netdevice.h>
29 #include <net/snmp.h>
30 #include <net/ip.h>
31 #include <net/icmp.h>
32 #include <net/protocol.h>
33 #include <linux/skbuff.h>
34 #include <linux/proc_fs.h>
35 #include <linux/export.h>
36 #include <net/sock.h>
37 #include <net/ping.h>
38 #include <net/udp.h>
39 #include <net/route.h>
40 #include <net/inet_common.h>
41 #include <net/checksum.h>
42 
43 #if IS_ENABLED(CONFIG_IPV6)
44 #include <linux/in6.h>
45 #include <linux/icmpv6.h>
46 #include <net/addrconf.h>
47 #include <net/ipv6.h>
48 #include <net/transp_v6.h>
49 #endif
50 
51 struct ping_table {
52 	struct hlist_nulls_head	hash[PING_HTABLE_SIZE];
53 	rwlock_t		lock;
54 };
55 
56 static struct ping_table ping_table;
57 struct pingv6_ops pingv6_ops;
58 EXPORT_SYMBOL_GPL(pingv6_ops);
59 
60 static u16 ping_port_rover;
61 
62 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
63 {
64 	u32 res = (num + net_hash_mix(net)) & mask;
65 
66 	pr_debug("hash(%u) = %u\n", num, res);
67 	return res;
68 }
69 EXPORT_SYMBOL_GPL(ping_hash);
70 
71 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
72 					     struct net *net, unsigned int num)
73 {
74 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
75 }
76 
77 int ping_get_port(struct sock *sk, unsigned short ident)
78 {
79 	struct hlist_nulls_node *node;
80 	struct hlist_nulls_head *hlist;
81 	struct inet_sock *isk, *isk2;
82 	struct sock *sk2 = NULL;
83 
84 	isk = inet_sk(sk);
85 	write_lock_bh(&ping_table.lock);
86 	if (ident == 0) {
87 		u32 i;
88 		u16 result = ping_port_rover + 1;
89 
90 		for (i = 0; i < (1L << 16); i++, result++) {
91 			if (!result)
92 				result++; /* avoid zero */
93 			hlist = ping_hashslot(&ping_table, sock_net(sk),
94 					    result);
95 			ping_portaddr_for_each_entry(sk2, node, hlist) {
96 				isk2 = inet_sk(sk2);
97 
98 				if (isk2->inet_num == result)
99 					goto next_port;
100 			}
101 
102 			/* found */
103 			ping_port_rover = ident = result;
104 			break;
105 next_port:
106 			;
107 		}
108 		if (i >= (1L << 16))
109 			goto fail;
110 	} else {
111 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
112 		ping_portaddr_for_each_entry(sk2, node, hlist) {
113 			isk2 = inet_sk(sk2);
114 
115 			/* BUG? Why is this reuse and not reuseaddr? ping.c
116 			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
117 			 * that other ping processes can steal its packets.
118 			 */
119 			if ((isk2->inet_num == ident) &&
120 			    (sk2 != sk) &&
121 			    (!sk2->sk_reuse || !sk->sk_reuse))
122 				goto fail;
123 		}
124 	}
125 
126 	pr_debug("found port/ident = %d\n", ident);
127 	isk->inet_num = ident;
128 	if (sk_unhashed(sk)) {
129 		pr_debug("was not hashed\n");
130 		sock_hold(sk);
131 		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
132 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
133 	}
134 	write_unlock_bh(&ping_table.lock);
135 	return 0;
136 
137 fail:
138 	write_unlock_bh(&ping_table.lock);
139 	return 1;
140 }
141 EXPORT_SYMBOL_GPL(ping_get_port);
142 
143 int ping_hash(struct sock *sk)
144 {
145 	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
146 	BUG(); /* "Please do not press this button again." */
147 
148 	return 0;
149 }
150 
151 void ping_unhash(struct sock *sk)
152 {
153 	struct inet_sock *isk = inet_sk(sk);
154 
155 	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
156 	write_lock_bh(&ping_table.lock);
157 	if (sk_hashed(sk)) {
158 		hlist_nulls_del(&sk->sk_nulls_node);
159 		sk_nulls_node_init(&sk->sk_nulls_node);
160 		sock_put(sk);
161 		isk->inet_num = 0;
162 		isk->inet_sport = 0;
163 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
164 	}
165 	write_unlock_bh(&ping_table.lock);
166 }
167 EXPORT_SYMBOL_GPL(ping_unhash);
168 
169 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
170 {
171 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
172 	struct sock *sk = NULL;
173 	struct inet_sock *isk;
174 	struct hlist_nulls_node *hnode;
175 	int dif = skb->dev->ifindex;
176 
177 	if (skb->protocol == htons(ETH_P_IP)) {
178 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
179 			 (int)ident, &ip_hdr(skb)->daddr, dif);
180 #if IS_ENABLED(CONFIG_IPV6)
181 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
182 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
183 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
184 #endif
185 	}
186 
187 	read_lock_bh(&ping_table.lock);
188 
189 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
190 		isk = inet_sk(sk);
191 
192 		pr_debug("iterate\n");
193 		if (isk->inet_num != ident)
194 			continue;
195 
196 		if (skb->protocol == htons(ETH_P_IP) &&
197 		    sk->sk_family == AF_INET) {
198 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
199 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
200 				 sk->sk_bound_dev_if);
201 
202 			if (isk->inet_rcv_saddr &&
203 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
204 				continue;
205 #if IS_ENABLED(CONFIG_IPV6)
206 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
207 			   sk->sk_family == AF_INET6) {
208 
209 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
210 				 (int) isk->inet_num,
211 				 &sk->sk_v6_rcv_saddr,
212 				 sk->sk_bound_dev_if);
213 
214 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
215 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
216 					     &ipv6_hdr(skb)->daddr))
217 				continue;
218 #endif
219 		} else {
220 			continue;
221 		}
222 
223 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
224 			continue;
225 
226 		sock_hold(sk);
227 		goto exit;
228 	}
229 
230 	sk = NULL;
231 exit:
232 	read_unlock_bh(&ping_table.lock);
233 
234 	return sk;
235 }
236 
237 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
238 					  kgid_t *high)
239 {
240 	kgid_t *data = net->ipv4.ping_group_range.range;
241 	unsigned int seq;
242 
243 	do {
244 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
245 
246 		*low = data[0];
247 		*high = data[1];
248 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
249 }
250 
251 
252 int ping_init_sock(struct sock *sk)
253 {
254 	struct net *net = sock_net(sk);
255 	kgid_t group = current_egid();
256 	struct group_info *group_info;
257 	int i;
258 	kgid_t low, high;
259 	int ret = 0;
260 
261 	if (sk->sk_family == AF_INET6)
262 		sk->sk_ipv6only = 1;
263 
264 	inet_get_ping_group_range_net(net, &low, &high);
265 	if (gid_lte(low, group) && gid_lte(group, high))
266 		return 0;
267 
268 	group_info = get_current_groups();
269 	for (i = 0; i < group_info->ngroups; i++) {
270 		kgid_t gid = group_info->gid[i];
271 
272 		if (gid_lte(low, gid) && gid_lte(gid, high))
273 			goto out_release_group;
274 	}
275 
276 	ret = -EACCES;
277 
278 out_release_group:
279 	put_group_info(group_info);
280 	return ret;
281 }
282 EXPORT_SYMBOL_GPL(ping_init_sock);
283 
284 void ping_close(struct sock *sk, long timeout)
285 {
286 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
287 		 inet_sk(sk), inet_sk(sk)->inet_num);
288 	pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
289 
290 	sk_common_release(sk);
291 }
292 EXPORT_SYMBOL_GPL(ping_close);
293 
294 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
295 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
296 				struct sockaddr *uaddr, int addr_len)
297 {
298 	struct net *net = sock_net(sk);
299 	if (sk->sk_family == AF_INET) {
300 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
301 		int chk_addr_ret;
302 
303 		if (addr_len < sizeof(*addr))
304 			return -EINVAL;
305 
306 		if (addr->sin_family != AF_INET &&
307 		    !(addr->sin_family == AF_UNSPEC &&
308 		      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
309 			return -EAFNOSUPPORT;
310 
311 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
312 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
313 
314 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
315 			chk_addr_ret = RTN_LOCAL;
316 		else
317 			chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
318 
319 		if ((!inet_can_nonlocal_bind(net, isk) &&
320 		     chk_addr_ret != RTN_LOCAL) ||
321 		    chk_addr_ret == RTN_MULTICAST ||
322 		    chk_addr_ret == RTN_BROADCAST)
323 			return -EADDRNOTAVAIL;
324 
325 #if IS_ENABLED(CONFIG_IPV6)
326 	} else if (sk->sk_family == AF_INET6) {
327 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
328 		int addr_type, scoped, has_addr;
329 		struct net_device *dev = NULL;
330 
331 		if (addr_len < sizeof(*addr))
332 			return -EINVAL;
333 
334 		if (addr->sin6_family != AF_INET6)
335 			return -EAFNOSUPPORT;
336 
337 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
338 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
339 
340 		addr_type = ipv6_addr_type(&addr->sin6_addr);
341 		scoped = __ipv6_addr_needs_scope_id(addr_type);
342 		if ((addr_type != IPV6_ADDR_ANY &&
343 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
344 		    (scoped && !addr->sin6_scope_id))
345 			return -EINVAL;
346 
347 		rcu_read_lock();
348 		if (addr->sin6_scope_id) {
349 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
350 			if (!dev) {
351 				rcu_read_unlock();
352 				return -ENODEV;
353 			}
354 		}
355 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
356 						    scoped);
357 		rcu_read_unlock();
358 
359 		if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
360 		      addr_type == IPV6_ADDR_ANY))
361 			return -EADDRNOTAVAIL;
362 
363 		if (scoped)
364 			sk->sk_bound_dev_if = addr->sin6_scope_id;
365 #endif
366 	} else {
367 		return -EAFNOSUPPORT;
368 	}
369 	return 0;
370 }
371 
372 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
373 {
374 	if (saddr->sa_family == AF_INET) {
375 		struct inet_sock *isk = inet_sk(sk);
376 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
377 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
378 #if IS_ENABLED(CONFIG_IPV6)
379 	} else if (saddr->sa_family == AF_INET6) {
380 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
381 		struct ipv6_pinfo *np = inet6_sk(sk);
382 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
383 #endif
384 	}
385 }
386 
387 /*
388  * We need our own bind because there are no privileged id's == local ports.
389  * Moreover, we don't allow binding to multi- and broadcast addresses.
390  */
391 
392 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
393 {
394 	struct inet_sock *isk = inet_sk(sk);
395 	unsigned short snum;
396 	int err;
397 	int dif = sk->sk_bound_dev_if;
398 
399 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
400 	if (err)
401 		return err;
402 
403 	lock_sock(sk);
404 
405 	err = -EINVAL;
406 	if (isk->inet_num != 0)
407 		goto out;
408 
409 	err = -EADDRINUSE;
410 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
411 	if (ping_get_port(sk, snum) != 0) {
412 		/* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
413 		sk->sk_bound_dev_if = dif;
414 		goto out;
415 	}
416 	ping_set_saddr(sk, uaddr);
417 
418 	pr_debug("after bind(): num = %hu, dif = %d\n",
419 		 isk->inet_num,
420 		 sk->sk_bound_dev_if);
421 
422 	err = 0;
423 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
424 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
425 #if IS_ENABLED(CONFIG_IPV6)
426 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
427 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
428 #endif
429 
430 	if (snum)
431 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
432 	isk->inet_sport = htons(isk->inet_num);
433 	isk->inet_daddr = 0;
434 	isk->inet_dport = 0;
435 
436 #if IS_ENABLED(CONFIG_IPV6)
437 	if (sk->sk_family == AF_INET6)
438 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
439 #endif
440 
441 	sk_dst_reset(sk);
442 out:
443 	release_sock(sk);
444 	pr_debug("ping_v4_bind -> %d\n", err);
445 	return err;
446 }
447 EXPORT_SYMBOL_GPL(ping_bind);
448 
449 /*
450  * Is this a supported type of ICMP message?
451  */
452 
453 static inline int ping_supported(int family, int type, int code)
454 {
455 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
456 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
457 }
458 
459 /*
460  * This routine is called by the ICMP module when it gets some
461  * sort of error condition.
462  */
463 
464 void ping_err(struct sk_buff *skb, int offset, u32 info)
465 {
466 	int family;
467 	struct icmphdr *icmph;
468 	struct inet_sock *inet_sock;
469 	int type;
470 	int code;
471 	struct net *net = dev_net(skb->dev);
472 	struct sock *sk;
473 	int harderr;
474 	int err;
475 
476 	if (skb->protocol == htons(ETH_P_IP)) {
477 		family = AF_INET;
478 		type = icmp_hdr(skb)->type;
479 		code = icmp_hdr(skb)->code;
480 		icmph = (struct icmphdr *)(skb->data + offset);
481 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
482 		family = AF_INET6;
483 		type = icmp6_hdr(skb)->icmp6_type;
484 		code = icmp6_hdr(skb)->icmp6_code;
485 		icmph = (struct icmphdr *) (skb->data + offset);
486 	} else {
487 		BUG();
488 	}
489 
490 	/* We assume the packet has already been checked by icmp_unreach */
491 
492 	if (!ping_supported(family, icmph->type, icmph->code))
493 		return;
494 
495 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
496 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
497 		 ntohs(icmph->un.echo.sequence));
498 
499 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
500 	if (!sk) {
501 		pr_debug("no socket, dropping\n");
502 		return;	/* No socket for error */
503 	}
504 	pr_debug("err on socket %p\n", sk);
505 
506 	err = 0;
507 	harderr = 0;
508 	inet_sock = inet_sk(sk);
509 
510 	if (skb->protocol == htons(ETH_P_IP)) {
511 		switch (type) {
512 		default:
513 		case ICMP_TIME_EXCEEDED:
514 			err = EHOSTUNREACH;
515 			break;
516 		case ICMP_SOURCE_QUENCH:
517 			/* This is not a real error but ping wants to see it.
518 			 * Report it with some fake errno.
519 			 */
520 			err = EREMOTEIO;
521 			break;
522 		case ICMP_PARAMETERPROB:
523 			err = EPROTO;
524 			harderr = 1;
525 			break;
526 		case ICMP_DEST_UNREACH:
527 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
528 				ipv4_sk_update_pmtu(skb, sk, info);
529 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
530 					err = EMSGSIZE;
531 					harderr = 1;
532 					break;
533 				}
534 				goto out;
535 			}
536 			err = EHOSTUNREACH;
537 			if (code <= NR_ICMP_UNREACH) {
538 				harderr = icmp_err_convert[code].fatal;
539 				err = icmp_err_convert[code].errno;
540 			}
541 			break;
542 		case ICMP_REDIRECT:
543 			/* See ICMP_SOURCE_QUENCH */
544 			ipv4_sk_redirect(skb, sk);
545 			err = EREMOTEIO;
546 			break;
547 		}
548 #if IS_ENABLED(CONFIG_IPV6)
549 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
550 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
551 #endif
552 	}
553 
554 	/*
555 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
556 	 *	4.1.3.3.
557 	 */
558 	if ((family == AF_INET && !inet_sock->recverr) ||
559 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
560 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
561 			goto out;
562 	} else {
563 		if (family == AF_INET) {
564 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
565 				      info, (u8 *)icmph);
566 #if IS_ENABLED(CONFIG_IPV6)
567 		} else if (family == AF_INET6) {
568 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
569 						   info, (u8 *)icmph);
570 #endif
571 		}
572 	}
573 	sk->sk_err = err;
574 	sk->sk_error_report(sk);
575 out:
576 	sock_put(sk);
577 }
578 EXPORT_SYMBOL_GPL(ping_err);
579 
580 /*
581  *	Copy and checksum an ICMP Echo packet from user space into a buffer
582  *	starting from the payload.
583  */
584 
585 int ping_getfrag(void *from, char *to,
586 		 int offset, int fraglen, int odd, struct sk_buff *skb)
587 {
588 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
589 
590 	if (offset == 0) {
591 		fraglen -= sizeof(struct icmphdr);
592 		if (fraglen < 0)
593 			BUG();
594 		if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
595 			    fraglen, &pfh->wcheck,
596 			    &pfh->msg->msg_iter))
597 			return -EFAULT;
598 	} else if (offset < sizeof(struct icmphdr)) {
599 			BUG();
600 	} else {
601 		if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
602 					    &pfh->msg->msg_iter))
603 			return -EFAULT;
604 	}
605 
606 #if IS_ENABLED(CONFIG_IPV6)
607 	/* For IPv6, checksum each skb as we go along, as expected by
608 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
609 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
610 	 */
611 	if (pfh->family == AF_INET6) {
612 		skb->csum = pfh->wcheck;
613 		skb->ip_summed = CHECKSUM_NONE;
614 		pfh->wcheck = 0;
615 	}
616 #endif
617 
618 	return 0;
619 }
620 EXPORT_SYMBOL_GPL(ping_getfrag);
621 
622 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
623 				       struct flowi4 *fl4)
624 {
625 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
626 
627 	if (!skb)
628 		return 0;
629 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
630 		sizeof(struct icmphdr), pfh->wcheck);
631 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
632 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
633 	skb->ip_summed = CHECKSUM_NONE;
634 	return ip_push_pending_frames(sk, fl4);
635 }
636 
637 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
638 			void *user_icmph, size_t icmph_len)
639 {
640 	u8 type, code;
641 
642 	if (len > 0xFFFF)
643 		return -EMSGSIZE;
644 
645 	/* Must have at least a full ICMP header. */
646 	if (len < icmph_len)
647 		return -EINVAL;
648 
649 	/*
650 	 *	Check the flags.
651 	 */
652 
653 	/* Mirror BSD error message compatibility */
654 	if (msg->msg_flags & MSG_OOB)
655 		return -EOPNOTSUPP;
656 
657 	/*
658 	 *	Fetch the ICMP header provided by the userland.
659 	 *	iovec is modified! The ICMP header is consumed.
660 	 */
661 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
662 		return -EFAULT;
663 
664 	if (family == AF_INET) {
665 		type = ((struct icmphdr *) user_icmph)->type;
666 		code = ((struct icmphdr *) user_icmph)->code;
667 #if IS_ENABLED(CONFIG_IPV6)
668 	} else if (family == AF_INET6) {
669 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
670 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
671 #endif
672 	} else {
673 		BUG();
674 	}
675 
676 	if (!ping_supported(family, type, code))
677 		return -EINVAL;
678 
679 	return 0;
680 }
681 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
682 
683 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
684 {
685 	struct net *net = sock_net(sk);
686 	struct flowi4 fl4;
687 	struct inet_sock *inet = inet_sk(sk);
688 	struct ipcm_cookie ipc;
689 	struct icmphdr user_icmph;
690 	struct pingfakehdr pfh;
691 	struct rtable *rt = NULL;
692 	struct ip_options_data opt_copy;
693 	int free = 0;
694 	__be32 saddr, daddr, faddr;
695 	u8  tos;
696 	int err;
697 
698 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
699 
700 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
701 				  sizeof(user_icmph));
702 	if (err)
703 		return err;
704 
705 	/*
706 	 *	Get and verify the address.
707 	 */
708 
709 	if (msg->msg_name) {
710 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
711 		if (msg->msg_namelen < sizeof(*usin))
712 			return -EINVAL;
713 		if (usin->sin_family != AF_INET)
714 			return -EAFNOSUPPORT;
715 		daddr = usin->sin_addr.s_addr;
716 		/* no remote port */
717 	} else {
718 		if (sk->sk_state != TCP_ESTABLISHED)
719 			return -EDESTADDRREQ;
720 		daddr = inet->inet_daddr;
721 		/* no remote port */
722 	}
723 
724 	ipcm_init_sk(&ipc, inet);
725 
726 	if (msg->msg_controllen) {
727 		err = ip_cmsg_send(sk, msg, &ipc, false);
728 		if (unlikely(err)) {
729 			kfree(ipc.opt);
730 			return err;
731 		}
732 		if (ipc.opt)
733 			free = 1;
734 	}
735 	if (!ipc.opt) {
736 		struct ip_options_rcu *inet_opt;
737 
738 		rcu_read_lock();
739 		inet_opt = rcu_dereference(inet->inet_opt);
740 		if (inet_opt) {
741 			memcpy(&opt_copy, inet_opt,
742 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
743 			ipc.opt = &opt_copy.opt;
744 		}
745 		rcu_read_unlock();
746 	}
747 
748 	saddr = ipc.addr;
749 	ipc.addr = faddr = daddr;
750 
751 	if (ipc.opt && ipc.opt->opt.srr) {
752 		if (!daddr) {
753 			err = -EINVAL;
754 			goto out_free;
755 		}
756 		faddr = ipc.opt->opt.faddr;
757 	}
758 	tos = get_rttos(&ipc, inet);
759 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
760 	    (msg->msg_flags & MSG_DONTROUTE) ||
761 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
762 		tos |= RTO_ONLINK;
763 	}
764 
765 	if (ipv4_is_multicast(daddr)) {
766 		if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
767 			ipc.oif = inet->mc_index;
768 		if (!saddr)
769 			saddr = inet->mc_addr;
770 	} else if (!ipc.oif)
771 		ipc.oif = inet->uc_index;
772 
773 	flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
774 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
775 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
776 			   sk->sk_uid);
777 
778 	fl4.fl4_icmp_type = user_icmph.type;
779 	fl4.fl4_icmp_code = user_icmph.code;
780 
781 	security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
782 	rt = ip_route_output_flow(net, &fl4, sk);
783 	if (IS_ERR(rt)) {
784 		err = PTR_ERR(rt);
785 		rt = NULL;
786 		if (err == -ENETUNREACH)
787 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
788 		goto out;
789 	}
790 
791 	err = -EACCES;
792 	if ((rt->rt_flags & RTCF_BROADCAST) &&
793 	    !sock_flag(sk, SOCK_BROADCAST))
794 		goto out;
795 
796 	if (msg->msg_flags & MSG_CONFIRM)
797 		goto do_confirm;
798 back_from_confirm:
799 
800 	if (!ipc.addr)
801 		ipc.addr = fl4.daddr;
802 
803 	lock_sock(sk);
804 
805 	pfh.icmph.type = user_icmph.type; /* already checked */
806 	pfh.icmph.code = user_icmph.code; /* ditto */
807 	pfh.icmph.checksum = 0;
808 	pfh.icmph.un.echo.id = inet->inet_sport;
809 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
810 	pfh.msg = msg;
811 	pfh.wcheck = 0;
812 	pfh.family = AF_INET;
813 
814 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
815 			0, &ipc, &rt, msg->msg_flags);
816 	if (err)
817 		ip_flush_pending_frames(sk);
818 	else
819 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
820 	release_sock(sk);
821 
822 out:
823 	ip_rt_put(rt);
824 out_free:
825 	if (free)
826 		kfree(ipc.opt);
827 	if (!err) {
828 		icmp_out_count(sock_net(sk), user_icmph.type);
829 		return len;
830 	}
831 	return err;
832 
833 do_confirm:
834 	if (msg->msg_flags & MSG_PROBE)
835 		dst_confirm_neigh(&rt->dst, &fl4.daddr);
836 	if (!(msg->msg_flags & MSG_PROBE) || len)
837 		goto back_from_confirm;
838 	err = 0;
839 	goto out;
840 }
841 
842 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
843 		 int flags, int *addr_len)
844 {
845 	struct inet_sock *isk = inet_sk(sk);
846 	int family = sk->sk_family;
847 	struct sk_buff *skb;
848 	int copied, err;
849 
850 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
851 
852 	err = -EOPNOTSUPP;
853 	if (flags & MSG_OOB)
854 		goto out;
855 
856 	if (flags & MSG_ERRQUEUE)
857 		return inet_recv_error(sk, msg, len, addr_len);
858 
859 	skb = skb_recv_datagram(sk, flags, noblock, &err);
860 	if (!skb)
861 		goto out;
862 
863 	copied = skb->len;
864 	if (copied > len) {
865 		msg->msg_flags |= MSG_TRUNC;
866 		copied = len;
867 	}
868 
869 	/* Don't bother checking the checksum */
870 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
871 	if (err)
872 		goto done;
873 
874 	sock_recv_timestamp(msg, sk, skb);
875 
876 	/* Copy the address and add cmsg data. */
877 	if (family == AF_INET) {
878 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
879 
880 		if (sin) {
881 			sin->sin_family = AF_INET;
882 			sin->sin_port = 0 /* skb->h.uh->source */;
883 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
884 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
885 			*addr_len = sizeof(*sin);
886 		}
887 
888 		if (isk->cmsg_flags)
889 			ip_cmsg_recv(msg, skb);
890 
891 #if IS_ENABLED(CONFIG_IPV6)
892 	} else if (family == AF_INET6) {
893 		struct ipv6_pinfo *np = inet6_sk(sk);
894 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
895 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
896 
897 		if (sin6) {
898 			sin6->sin6_family = AF_INET6;
899 			sin6->sin6_port = 0;
900 			sin6->sin6_addr = ip6->saddr;
901 			sin6->sin6_flowinfo = 0;
902 			if (np->sndflow)
903 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
904 			sin6->sin6_scope_id =
905 				ipv6_iface_scope_id(&sin6->sin6_addr,
906 						    inet6_iif(skb));
907 			*addr_len = sizeof(*sin6);
908 		}
909 
910 		if (inet6_sk(sk)->rxopt.all)
911 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
912 		if (skb->protocol == htons(ETH_P_IPV6) &&
913 		    inet6_sk(sk)->rxopt.all)
914 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
915 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
916 			ip_cmsg_recv(msg, skb);
917 #endif
918 	} else {
919 		BUG();
920 	}
921 
922 	err = copied;
923 
924 done:
925 	skb_free_datagram(sk, skb);
926 out:
927 	pr_debug("ping_recvmsg -> %d\n", err);
928 	return err;
929 }
930 EXPORT_SYMBOL_GPL(ping_recvmsg);
931 
932 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
933 {
934 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
935 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
936 	if (sock_queue_rcv_skb(sk, skb) < 0) {
937 		kfree_skb(skb);
938 		pr_debug("ping_queue_rcv_skb -> failed\n");
939 		return -1;
940 	}
941 	return 0;
942 }
943 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
944 
945 
946 /*
947  *	All we need to do is get the socket.
948  */
949 
950 bool ping_rcv(struct sk_buff *skb)
951 {
952 	struct sock *sk;
953 	struct net *net = dev_net(skb->dev);
954 	struct icmphdr *icmph = icmp_hdr(skb);
955 
956 	/* We assume the packet has already been checked by icmp_rcv */
957 
958 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
959 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
960 
961 	/* Push ICMP header back */
962 	skb_push(skb, skb->data - (u8 *)icmph);
963 
964 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
965 	if (sk) {
966 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
967 
968 		pr_debug("rcv on socket %p\n", sk);
969 		if (skb2)
970 			ping_queue_rcv_skb(sk, skb2);
971 		sock_put(sk);
972 		return true;
973 	}
974 	pr_debug("no socket, dropping\n");
975 
976 	return false;
977 }
978 EXPORT_SYMBOL_GPL(ping_rcv);
979 
980 struct proto ping_prot = {
981 	.name =		"PING",
982 	.owner =	THIS_MODULE,
983 	.init =		ping_init_sock,
984 	.close =	ping_close,
985 	.connect =	ip4_datagram_connect,
986 	.disconnect =	__udp_disconnect,
987 	.setsockopt =	ip_setsockopt,
988 	.getsockopt =	ip_getsockopt,
989 	.sendmsg =	ping_v4_sendmsg,
990 	.recvmsg =	ping_recvmsg,
991 	.bind =		ping_bind,
992 	.backlog_rcv =	ping_queue_rcv_skb,
993 	.release_cb =	ip4_datagram_release_cb,
994 	.hash =		ping_hash,
995 	.unhash =	ping_unhash,
996 	.get_port =	ping_get_port,
997 	.obj_size =	sizeof(struct inet_sock),
998 };
999 EXPORT_SYMBOL(ping_prot);
1000 
1001 #ifdef CONFIG_PROC_FS
1002 
1003 static struct sock *ping_get_first(struct seq_file *seq, int start)
1004 {
1005 	struct sock *sk;
1006 	struct ping_iter_state *state = seq->private;
1007 	struct net *net = seq_file_net(seq);
1008 
1009 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1010 	     ++state->bucket) {
1011 		struct hlist_nulls_node *node;
1012 		struct hlist_nulls_head *hslot;
1013 
1014 		hslot = &ping_table.hash[state->bucket];
1015 
1016 		if (hlist_nulls_empty(hslot))
1017 			continue;
1018 
1019 		sk_nulls_for_each(sk, node, hslot) {
1020 			if (net_eq(sock_net(sk), net) &&
1021 			    sk->sk_family == state->family)
1022 				goto found;
1023 		}
1024 	}
1025 	sk = NULL;
1026 found:
1027 	return sk;
1028 }
1029 
1030 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1031 {
1032 	struct ping_iter_state *state = seq->private;
1033 	struct net *net = seq_file_net(seq);
1034 
1035 	do {
1036 		sk = sk_nulls_next(sk);
1037 	} while (sk && (!net_eq(sock_net(sk), net)));
1038 
1039 	if (!sk)
1040 		return ping_get_first(seq, state->bucket + 1);
1041 	return sk;
1042 }
1043 
1044 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1045 {
1046 	struct sock *sk = ping_get_first(seq, 0);
1047 
1048 	if (sk)
1049 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1050 			--pos;
1051 	return pos ? NULL : sk;
1052 }
1053 
1054 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1055 	__acquires(ping_table.lock)
1056 {
1057 	struct ping_iter_state *state = seq->private;
1058 	state->bucket = 0;
1059 	state->family = family;
1060 
1061 	read_lock_bh(&ping_table.lock);
1062 
1063 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1064 }
1065 EXPORT_SYMBOL_GPL(ping_seq_start);
1066 
1067 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1068 {
1069 	return ping_seq_start(seq, pos, AF_INET);
1070 }
1071 
1072 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1073 {
1074 	struct sock *sk;
1075 
1076 	if (v == SEQ_START_TOKEN)
1077 		sk = ping_get_idx(seq, 0);
1078 	else
1079 		sk = ping_get_next(seq, v);
1080 
1081 	++*pos;
1082 	return sk;
1083 }
1084 EXPORT_SYMBOL_GPL(ping_seq_next);
1085 
1086 void ping_seq_stop(struct seq_file *seq, void *v)
1087 	__releases(ping_table.lock)
1088 {
1089 	read_unlock_bh(&ping_table.lock);
1090 }
1091 EXPORT_SYMBOL_GPL(ping_seq_stop);
1092 
1093 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1094 		int bucket)
1095 {
1096 	struct inet_sock *inet = inet_sk(sp);
1097 	__be32 dest = inet->inet_daddr;
1098 	__be32 src = inet->inet_rcv_saddr;
1099 	__u16 destp = ntohs(inet->inet_dport);
1100 	__u16 srcp = ntohs(inet->inet_sport);
1101 
1102 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1103 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1104 		bucket, src, srcp, dest, destp, sp->sk_state,
1105 		sk_wmem_alloc_get(sp),
1106 		sk_rmem_alloc_get(sp),
1107 		0, 0L, 0,
1108 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1109 		0, sock_i_ino(sp),
1110 		refcount_read(&sp->sk_refcnt), sp,
1111 		atomic_read(&sp->sk_drops));
1112 }
1113 
1114 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1115 {
1116 	seq_setwidth(seq, 127);
1117 	if (v == SEQ_START_TOKEN)
1118 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1119 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1120 			   "inode ref pointer drops");
1121 	else {
1122 		struct ping_iter_state *state = seq->private;
1123 
1124 		ping_v4_format_sock(v, seq, state->bucket);
1125 	}
1126 	seq_pad(seq, '\n');
1127 	return 0;
1128 }
1129 
1130 static const struct seq_operations ping_v4_seq_ops = {
1131 	.start		= ping_v4_seq_start,
1132 	.show		= ping_v4_seq_show,
1133 	.next		= ping_seq_next,
1134 	.stop		= ping_seq_stop,
1135 };
1136 
1137 static int __net_init ping_v4_proc_init_net(struct net *net)
1138 {
1139 	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1140 			sizeof(struct ping_iter_state)))
1141 		return -ENOMEM;
1142 	return 0;
1143 }
1144 
1145 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1146 {
1147 	remove_proc_entry("icmp", net->proc_net);
1148 }
1149 
1150 static struct pernet_operations ping_v4_net_ops = {
1151 	.init = ping_v4_proc_init_net,
1152 	.exit = ping_v4_proc_exit_net,
1153 };
1154 
1155 int __init ping_proc_init(void)
1156 {
1157 	return register_pernet_subsys(&ping_v4_net_ops);
1158 }
1159 
1160 void ping_proc_exit(void)
1161 {
1162 	unregister_pernet_subsys(&ping_v4_net_ops);
1163 }
1164 
1165 #endif
1166 
1167 void __init ping_init(void)
1168 {
1169 	int i;
1170 
1171 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1172 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1173 	rwlock_init(&ping_table.lock);
1174 }
1175