xref: /linux/net/ipv4/ping.c (revision e0bf6c5ca2d3281f231c5f0c9bf145e9513644de)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		"Ping" sockets
7  *
8  *		This program is free software; you can redistribute it and/or
9  *		modify it under the terms of the GNU General Public License
10  *		as published by the Free Software Foundation; either version
11  *		2 of the License, or (at your option) any later version.
12  *
13  * Based on ipv4/udp.c code.
14  *
15  * Authors:	Vasiliy Kulikov / Openwall (for Linux 2.6),
16  *		Pavel Kankovsky (for Linux 2.4.32)
17  *
18  * Pavel gave all rights to bugs to Vasiliy,
19  * none of the bugs are Pavel's now.
20  *
21  */
22 
23 #include <linux/uaccess.h>
24 #include <linux/types.h>
25 #include <linux/fcntl.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/in.h>
29 #include <linux/errno.h>
30 #include <linux/timer.h>
31 #include <linux/mm.h>
32 #include <linux/inet.h>
33 #include <linux/netdevice.h>
34 #include <net/snmp.h>
35 #include <net/ip.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <linux/skbuff.h>
39 #include <linux/proc_fs.h>
40 #include <linux/export.h>
41 #include <net/sock.h>
42 #include <net/ping.h>
43 #include <net/udp.h>
44 #include <net/route.h>
45 #include <net/inet_common.h>
46 #include <net/checksum.h>
47 
48 #if IS_ENABLED(CONFIG_IPV6)
49 #include <linux/in6.h>
50 #include <linux/icmpv6.h>
51 #include <net/addrconf.h>
52 #include <net/ipv6.h>
53 #include <net/transp_v6.h>
54 #endif
55 
56 struct ping_table {
57 	struct hlist_nulls_head	hash[PING_HTABLE_SIZE];
58 	rwlock_t		lock;
59 };
60 
61 static struct ping_table ping_table;
62 struct pingv6_ops pingv6_ops;
63 EXPORT_SYMBOL_GPL(pingv6_ops);
64 
65 static u16 ping_port_rover;
66 
67 static inline int ping_hashfn(struct net *net, unsigned int num, unsigned int mask)
68 {
69 	int res = (num + net_hash_mix(net)) & mask;
70 
71 	pr_debug("hash(%d) = %d\n", num, res);
72 	return res;
73 }
74 EXPORT_SYMBOL_GPL(ping_hash);
75 
76 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
77 					     struct net *net, unsigned int num)
78 {
79 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
80 }
81 
82 int ping_get_port(struct sock *sk, unsigned short ident)
83 {
84 	struct hlist_nulls_node *node;
85 	struct hlist_nulls_head *hlist;
86 	struct inet_sock *isk, *isk2;
87 	struct sock *sk2 = NULL;
88 
89 	isk = inet_sk(sk);
90 	write_lock_bh(&ping_table.lock);
91 	if (ident == 0) {
92 		u32 i;
93 		u16 result = ping_port_rover + 1;
94 
95 		for (i = 0; i < (1L << 16); i++, result++) {
96 			if (!result)
97 				result++; /* avoid zero */
98 			hlist = ping_hashslot(&ping_table, sock_net(sk),
99 					    result);
100 			ping_portaddr_for_each_entry(sk2, node, hlist) {
101 				isk2 = inet_sk(sk2);
102 
103 				if (isk2->inet_num == result)
104 					goto next_port;
105 			}
106 
107 			/* found */
108 			ping_port_rover = ident = result;
109 			break;
110 next_port:
111 			;
112 		}
113 		if (i >= (1L << 16))
114 			goto fail;
115 	} else {
116 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
117 		ping_portaddr_for_each_entry(sk2, node, hlist) {
118 			isk2 = inet_sk(sk2);
119 
120 			/* BUG? Why is this reuse and not reuseaddr? ping.c
121 			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
122 			 * that other ping processes can steal its packets.
123 			 */
124 			if ((isk2->inet_num == ident) &&
125 			    (sk2 != sk) &&
126 			    (!sk2->sk_reuse || !sk->sk_reuse))
127 				goto fail;
128 		}
129 	}
130 
131 	pr_debug("found port/ident = %d\n", ident);
132 	isk->inet_num = ident;
133 	if (sk_unhashed(sk)) {
134 		pr_debug("was not hashed\n");
135 		sock_hold(sk);
136 		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
137 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
138 	}
139 	write_unlock_bh(&ping_table.lock);
140 	return 0;
141 
142 fail:
143 	write_unlock_bh(&ping_table.lock);
144 	return 1;
145 }
146 EXPORT_SYMBOL_GPL(ping_get_port);
147 
148 void ping_hash(struct sock *sk)
149 {
150 	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
151 	BUG(); /* "Please do not press this button again." */
152 }
153 
154 void ping_unhash(struct sock *sk)
155 {
156 	struct inet_sock *isk = inet_sk(sk);
157 	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
158 	if (sk_hashed(sk)) {
159 		write_lock_bh(&ping_table.lock);
160 		hlist_nulls_del(&sk->sk_nulls_node);
161 		sock_put(sk);
162 		isk->inet_num = 0;
163 		isk->inet_sport = 0;
164 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
165 		write_unlock_bh(&ping_table.lock);
166 	}
167 }
168 EXPORT_SYMBOL_GPL(ping_unhash);
169 
170 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
171 {
172 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
173 	struct sock *sk = NULL;
174 	struct inet_sock *isk;
175 	struct hlist_nulls_node *hnode;
176 	int dif = skb->dev->ifindex;
177 
178 	if (skb->protocol == htons(ETH_P_IP)) {
179 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
180 			 (int)ident, &ip_hdr(skb)->daddr, dif);
181 #if IS_ENABLED(CONFIG_IPV6)
182 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
183 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
184 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
185 #endif
186 	}
187 
188 	read_lock_bh(&ping_table.lock);
189 
190 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
191 		isk = inet_sk(sk);
192 
193 		pr_debug("iterate\n");
194 		if (isk->inet_num != ident)
195 			continue;
196 
197 		if (skb->protocol == htons(ETH_P_IP) &&
198 		    sk->sk_family == AF_INET) {
199 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
200 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
201 				 sk->sk_bound_dev_if);
202 
203 			if (isk->inet_rcv_saddr &&
204 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
205 				continue;
206 #if IS_ENABLED(CONFIG_IPV6)
207 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
208 			   sk->sk_family == AF_INET6) {
209 
210 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
211 				 (int) isk->inet_num,
212 				 &sk->sk_v6_rcv_saddr,
213 				 sk->sk_bound_dev_if);
214 
215 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
216 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
217 					     &ipv6_hdr(skb)->daddr))
218 				continue;
219 #endif
220 		} else {
221 			continue;
222 		}
223 
224 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
225 			continue;
226 
227 		sock_hold(sk);
228 		goto exit;
229 	}
230 
231 	sk = NULL;
232 exit:
233 	read_unlock_bh(&ping_table.lock);
234 
235 	return sk;
236 }
237 
238 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
239 					  kgid_t *high)
240 {
241 	kgid_t *data = net->ipv4.ping_group_range.range;
242 	unsigned int seq;
243 
244 	do {
245 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
246 
247 		*low = data[0];
248 		*high = data[1];
249 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
250 }
251 
252 
253 int ping_init_sock(struct sock *sk)
254 {
255 	struct net *net = sock_net(sk);
256 	kgid_t group = current_egid();
257 	struct group_info *group_info;
258 	int i, j, count;
259 	kgid_t low, high;
260 	int ret = 0;
261 
262 	if (sk->sk_family == AF_INET6)
263 		sk->sk_ipv6only = 1;
264 
265 	inet_get_ping_group_range_net(net, &low, &high);
266 	if (gid_lte(low, group) && gid_lte(group, high))
267 		return 0;
268 
269 	group_info = get_current_groups();
270 	count = group_info->ngroups;
271 	for (i = 0; i < group_info->nblocks; i++) {
272 		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
273 		for (j = 0; j < cp_count; j++) {
274 			kgid_t gid = group_info->blocks[i][j];
275 			if (gid_lte(low, gid) && gid_lte(gid, high))
276 				goto out_release_group;
277 		}
278 
279 		count -= cp_count;
280 	}
281 
282 	ret = -EACCES;
283 
284 out_release_group:
285 	put_group_info(group_info);
286 	return ret;
287 }
288 EXPORT_SYMBOL_GPL(ping_init_sock);
289 
290 void ping_close(struct sock *sk, long timeout)
291 {
292 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
293 		 inet_sk(sk), inet_sk(sk)->inet_num);
294 	pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
295 
296 	sk_common_release(sk);
297 }
298 EXPORT_SYMBOL_GPL(ping_close);
299 
300 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
301 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
302 				struct sockaddr *uaddr, int addr_len) {
303 	struct net *net = sock_net(sk);
304 	if (sk->sk_family == AF_INET) {
305 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
306 		int chk_addr_ret;
307 
308 		if (addr_len < sizeof(*addr))
309 			return -EINVAL;
310 
311 		if (addr->sin_family != AF_INET &&
312 		    !(addr->sin_family == AF_UNSPEC &&
313 		      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
314 			return -EAFNOSUPPORT;
315 
316 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
317 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
318 
319 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
320 
321 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
322 			chk_addr_ret = RTN_LOCAL;
323 
324 		if ((net->ipv4.sysctl_ip_nonlocal_bind == 0 &&
325 		    isk->freebind == 0 && isk->transparent == 0 &&
326 		     chk_addr_ret != RTN_LOCAL) ||
327 		    chk_addr_ret == RTN_MULTICAST ||
328 		    chk_addr_ret == RTN_BROADCAST)
329 			return -EADDRNOTAVAIL;
330 
331 #if IS_ENABLED(CONFIG_IPV6)
332 	} else if (sk->sk_family == AF_INET6) {
333 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
334 		int addr_type, scoped, has_addr;
335 		struct net_device *dev = NULL;
336 
337 		if (addr_len < sizeof(*addr))
338 			return -EINVAL;
339 
340 		if (addr->sin6_family != AF_INET6)
341 			return -EAFNOSUPPORT;
342 
343 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
344 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
345 
346 		addr_type = ipv6_addr_type(&addr->sin6_addr);
347 		scoped = __ipv6_addr_needs_scope_id(addr_type);
348 		if ((addr_type != IPV6_ADDR_ANY &&
349 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
350 		    (scoped && !addr->sin6_scope_id))
351 			return -EINVAL;
352 
353 		rcu_read_lock();
354 		if (addr->sin6_scope_id) {
355 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
356 			if (!dev) {
357 				rcu_read_unlock();
358 				return -ENODEV;
359 			}
360 		}
361 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
362 						    scoped);
363 		rcu_read_unlock();
364 
365 		if (!(isk->freebind || isk->transparent || has_addr ||
366 		      addr_type == IPV6_ADDR_ANY))
367 			return -EADDRNOTAVAIL;
368 
369 		if (scoped)
370 			sk->sk_bound_dev_if = addr->sin6_scope_id;
371 #endif
372 	} else {
373 		return -EAFNOSUPPORT;
374 	}
375 	return 0;
376 }
377 
378 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
379 {
380 	if (saddr->sa_family == AF_INET) {
381 		struct inet_sock *isk = inet_sk(sk);
382 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
383 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
384 #if IS_ENABLED(CONFIG_IPV6)
385 	} else if (saddr->sa_family == AF_INET6) {
386 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
387 		struct ipv6_pinfo *np = inet6_sk(sk);
388 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
389 #endif
390 	}
391 }
392 
393 static void ping_clear_saddr(struct sock *sk, int dif)
394 {
395 	sk->sk_bound_dev_if = dif;
396 	if (sk->sk_family == AF_INET) {
397 		struct inet_sock *isk = inet_sk(sk);
398 		isk->inet_rcv_saddr = isk->inet_saddr = 0;
399 #if IS_ENABLED(CONFIG_IPV6)
400 	} else if (sk->sk_family == AF_INET6) {
401 		struct ipv6_pinfo *np = inet6_sk(sk);
402 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
403 		memset(&np->saddr, 0, sizeof(np->saddr));
404 #endif
405 	}
406 }
407 /*
408  * We need our own bind because there are no privileged id's == local ports.
409  * Moreover, we don't allow binding to multi- and broadcast addresses.
410  */
411 
412 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
413 {
414 	struct inet_sock *isk = inet_sk(sk);
415 	unsigned short snum;
416 	int err;
417 	int dif = sk->sk_bound_dev_if;
418 
419 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
420 	if (err)
421 		return err;
422 
423 	lock_sock(sk);
424 
425 	err = -EINVAL;
426 	if (isk->inet_num != 0)
427 		goto out;
428 
429 	err = -EADDRINUSE;
430 	ping_set_saddr(sk, uaddr);
431 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
432 	if (ping_get_port(sk, snum) != 0) {
433 		ping_clear_saddr(sk, dif);
434 		goto out;
435 	}
436 
437 	pr_debug("after bind(): num = %d, dif = %d\n",
438 		 (int)isk->inet_num,
439 		 (int)sk->sk_bound_dev_if);
440 
441 	err = 0;
442 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
443 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
444 #if IS_ENABLED(CONFIG_IPV6)
445 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
446 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
447 #endif
448 
449 	if (snum)
450 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
451 	isk->inet_sport = htons(isk->inet_num);
452 	isk->inet_daddr = 0;
453 	isk->inet_dport = 0;
454 
455 #if IS_ENABLED(CONFIG_IPV6)
456 	if (sk->sk_family == AF_INET6)
457 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
458 #endif
459 
460 	sk_dst_reset(sk);
461 out:
462 	release_sock(sk);
463 	pr_debug("ping_v4_bind -> %d\n", err);
464 	return err;
465 }
466 EXPORT_SYMBOL_GPL(ping_bind);
467 
468 /*
469  * Is this a supported type of ICMP message?
470  */
471 
472 static inline int ping_supported(int family, int type, int code)
473 {
474 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
475 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
476 }
477 
478 /*
479  * This routine is called by the ICMP module when it gets some
480  * sort of error condition.
481  */
482 
483 void ping_err(struct sk_buff *skb, int offset, u32 info)
484 {
485 	int family;
486 	struct icmphdr *icmph;
487 	struct inet_sock *inet_sock;
488 	int type;
489 	int code;
490 	struct net *net = dev_net(skb->dev);
491 	struct sock *sk;
492 	int harderr;
493 	int err;
494 
495 	if (skb->protocol == htons(ETH_P_IP)) {
496 		family = AF_INET;
497 		type = icmp_hdr(skb)->type;
498 		code = icmp_hdr(skb)->code;
499 		icmph = (struct icmphdr *)(skb->data + offset);
500 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
501 		family = AF_INET6;
502 		type = icmp6_hdr(skb)->icmp6_type;
503 		code = icmp6_hdr(skb)->icmp6_code;
504 		icmph = (struct icmphdr *) (skb->data + offset);
505 	} else {
506 		BUG();
507 	}
508 
509 	/* We assume the packet has already been checked by icmp_unreach */
510 
511 	if (!ping_supported(family, icmph->type, icmph->code))
512 		return;
513 
514 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
515 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
516 		 ntohs(icmph->un.echo.sequence));
517 
518 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
519 	if (sk == NULL) {
520 		pr_debug("no socket, dropping\n");
521 		return;	/* No socket for error */
522 	}
523 	pr_debug("err on socket %p\n", sk);
524 
525 	err = 0;
526 	harderr = 0;
527 	inet_sock = inet_sk(sk);
528 
529 	if (skb->protocol == htons(ETH_P_IP)) {
530 		switch (type) {
531 		default:
532 		case ICMP_TIME_EXCEEDED:
533 			err = EHOSTUNREACH;
534 			break;
535 		case ICMP_SOURCE_QUENCH:
536 			/* This is not a real error but ping wants to see it.
537 			 * Report it with some fake errno.
538 			 */
539 			err = EREMOTEIO;
540 			break;
541 		case ICMP_PARAMETERPROB:
542 			err = EPROTO;
543 			harderr = 1;
544 			break;
545 		case ICMP_DEST_UNREACH:
546 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
547 				ipv4_sk_update_pmtu(skb, sk, info);
548 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
549 					err = EMSGSIZE;
550 					harderr = 1;
551 					break;
552 				}
553 				goto out;
554 			}
555 			err = EHOSTUNREACH;
556 			if (code <= NR_ICMP_UNREACH) {
557 				harderr = icmp_err_convert[code].fatal;
558 				err = icmp_err_convert[code].errno;
559 			}
560 			break;
561 		case ICMP_REDIRECT:
562 			/* See ICMP_SOURCE_QUENCH */
563 			ipv4_sk_redirect(skb, sk);
564 			err = EREMOTEIO;
565 			break;
566 		}
567 #if IS_ENABLED(CONFIG_IPV6)
568 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
569 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
570 #endif
571 	}
572 
573 	/*
574 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
575 	 *	4.1.3.3.
576 	 */
577 	if ((family == AF_INET && !inet_sock->recverr) ||
578 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
579 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
580 			goto out;
581 	} else {
582 		if (family == AF_INET) {
583 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
584 				      info, (u8 *)icmph);
585 #if IS_ENABLED(CONFIG_IPV6)
586 		} else if (family == AF_INET6) {
587 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
588 						   info, (u8 *)icmph);
589 #endif
590 		}
591 	}
592 	sk->sk_err = err;
593 	sk->sk_error_report(sk);
594 out:
595 	sock_put(sk);
596 }
597 EXPORT_SYMBOL_GPL(ping_err);
598 
599 /*
600  *	Copy and checksum an ICMP Echo packet from user space into a buffer
601  *	starting from the payload.
602  */
603 
604 int ping_getfrag(void *from, char *to,
605 		 int offset, int fraglen, int odd, struct sk_buff *skb)
606 {
607 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
608 
609 	if (offset == 0) {
610 		fraglen -= sizeof(struct icmphdr);
611 		if (fraglen < 0)
612 			BUG();
613 		if (csum_and_copy_from_iter(to + sizeof(struct icmphdr),
614 			    fraglen, &pfh->wcheck,
615 			    &pfh->msg->msg_iter) != fraglen)
616 			return -EFAULT;
617 	} else if (offset < sizeof(struct icmphdr)) {
618 			BUG();
619 	} else {
620 		if (csum_and_copy_from_iter(to, fraglen, &pfh->wcheck,
621 					    &pfh->msg->msg_iter) != fraglen)
622 			return -EFAULT;
623 	}
624 
625 #if IS_ENABLED(CONFIG_IPV6)
626 	/* For IPv6, checksum each skb as we go along, as expected by
627 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
628 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
629 	 */
630 	if (pfh->family == AF_INET6) {
631 		skb->csum = pfh->wcheck;
632 		skb->ip_summed = CHECKSUM_NONE;
633 		pfh->wcheck = 0;
634 	}
635 #endif
636 
637 	return 0;
638 }
639 EXPORT_SYMBOL_GPL(ping_getfrag);
640 
641 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
642 				       struct flowi4 *fl4)
643 {
644 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
645 
646 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
647 		sizeof(struct icmphdr), pfh->wcheck);
648 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
649 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
650 	skb->ip_summed = CHECKSUM_NONE;
651 	return ip_push_pending_frames(sk, fl4);
652 }
653 
654 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
655 			void *user_icmph, size_t icmph_len) {
656 	u8 type, code;
657 
658 	if (len > 0xFFFF)
659 		return -EMSGSIZE;
660 
661 	/*
662 	 *	Check the flags.
663 	 */
664 
665 	/* Mirror BSD error message compatibility */
666 	if (msg->msg_flags & MSG_OOB)
667 		return -EOPNOTSUPP;
668 
669 	/*
670 	 *	Fetch the ICMP header provided by the userland.
671 	 *	iovec is modified! The ICMP header is consumed.
672 	 */
673 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
674 		return -EFAULT;
675 
676 	if (family == AF_INET) {
677 		type = ((struct icmphdr *) user_icmph)->type;
678 		code = ((struct icmphdr *) user_icmph)->code;
679 #if IS_ENABLED(CONFIG_IPV6)
680 	} else if (family == AF_INET6) {
681 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
682 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
683 #endif
684 	} else {
685 		BUG();
686 	}
687 
688 	if (!ping_supported(family, type, code))
689 		return -EINVAL;
690 
691 	return 0;
692 }
693 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
694 
695 static int ping_v4_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
696 			   size_t len)
697 {
698 	struct net *net = sock_net(sk);
699 	struct flowi4 fl4;
700 	struct inet_sock *inet = inet_sk(sk);
701 	struct ipcm_cookie ipc;
702 	struct icmphdr user_icmph;
703 	struct pingfakehdr pfh;
704 	struct rtable *rt = NULL;
705 	struct ip_options_data opt_copy;
706 	int free = 0;
707 	__be32 saddr, daddr, faddr;
708 	u8  tos;
709 	int err;
710 
711 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
712 
713 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
714 				  sizeof(user_icmph));
715 	if (err)
716 		return err;
717 
718 	/*
719 	 *	Get and verify the address.
720 	 */
721 
722 	if (msg->msg_name) {
723 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
724 		if (msg->msg_namelen < sizeof(*usin))
725 			return -EINVAL;
726 		if (usin->sin_family != AF_INET)
727 			return -EAFNOSUPPORT;
728 		daddr = usin->sin_addr.s_addr;
729 		/* no remote port */
730 	} else {
731 		if (sk->sk_state != TCP_ESTABLISHED)
732 			return -EDESTADDRREQ;
733 		daddr = inet->inet_daddr;
734 		/* no remote port */
735 	}
736 
737 	ipc.addr = inet->inet_saddr;
738 	ipc.opt = NULL;
739 	ipc.oif = sk->sk_bound_dev_if;
740 	ipc.tx_flags = 0;
741 	ipc.ttl = 0;
742 	ipc.tos = -1;
743 
744 	sock_tx_timestamp(sk, &ipc.tx_flags);
745 
746 	if (msg->msg_controllen) {
747 		err = ip_cmsg_send(sock_net(sk), msg, &ipc, false);
748 		if (err)
749 			return err;
750 		if (ipc.opt)
751 			free = 1;
752 	}
753 	if (!ipc.opt) {
754 		struct ip_options_rcu *inet_opt;
755 
756 		rcu_read_lock();
757 		inet_opt = rcu_dereference(inet->inet_opt);
758 		if (inet_opt) {
759 			memcpy(&opt_copy, inet_opt,
760 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
761 			ipc.opt = &opt_copy.opt;
762 		}
763 		rcu_read_unlock();
764 	}
765 
766 	saddr = ipc.addr;
767 	ipc.addr = faddr = daddr;
768 
769 	if (ipc.opt && ipc.opt->opt.srr) {
770 		if (!daddr)
771 			return -EINVAL;
772 		faddr = ipc.opt->opt.faddr;
773 	}
774 	tos = get_rttos(&ipc, inet);
775 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
776 	    (msg->msg_flags & MSG_DONTROUTE) ||
777 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
778 		tos |= RTO_ONLINK;
779 	}
780 
781 	if (ipv4_is_multicast(daddr)) {
782 		if (!ipc.oif)
783 			ipc.oif = inet->mc_index;
784 		if (!saddr)
785 			saddr = inet->mc_addr;
786 	} else if (!ipc.oif)
787 		ipc.oif = inet->uc_index;
788 
789 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
790 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
791 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
792 
793 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
794 	rt = ip_route_output_flow(net, &fl4, sk);
795 	if (IS_ERR(rt)) {
796 		err = PTR_ERR(rt);
797 		rt = NULL;
798 		if (err == -ENETUNREACH)
799 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
800 		goto out;
801 	}
802 
803 	err = -EACCES;
804 	if ((rt->rt_flags & RTCF_BROADCAST) &&
805 	    !sock_flag(sk, SOCK_BROADCAST))
806 		goto out;
807 
808 	if (msg->msg_flags & MSG_CONFIRM)
809 		goto do_confirm;
810 back_from_confirm:
811 
812 	if (!ipc.addr)
813 		ipc.addr = fl4.daddr;
814 
815 	lock_sock(sk);
816 
817 	pfh.icmph.type = user_icmph.type; /* already checked */
818 	pfh.icmph.code = user_icmph.code; /* ditto */
819 	pfh.icmph.checksum = 0;
820 	pfh.icmph.un.echo.id = inet->inet_sport;
821 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
822 	pfh.msg = msg;
823 	pfh.wcheck = 0;
824 	pfh.family = AF_INET;
825 
826 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
827 			0, &ipc, &rt, msg->msg_flags);
828 	if (err)
829 		ip_flush_pending_frames(sk);
830 	else
831 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
832 	release_sock(sk);
833 
834 out:
835 	ip_rt_put(rt);
836 	if (free)
837 		kfree(ipc.opt);
838 	if (!err) {
839 		icmp_out_count(sock_net(sk), user_icmph.type);
840 		return len;
841 	}
842 	return err;
843 
844 do_confirm:
845 	dst_confirm(&rt->dst);
846 	if (!(msg->msg_flags & MSG_PROBE) || len)
847 		goto back_from_confirm;
848 	err = 0;
849 	goto out;
850 }
851 
852 int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
853 		 size_t len, int noblock, int flags, int *addr_len)
854 {
855 	struct inet_sock *isk = inet_sk(sk);
856 	int family = sk->sk_family;
857 	struct sk_buff *skb;
858 	int copied, err;
859 
860 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
861 
862 	err = -EOPNOTSUPP;
863 	if (flags & MSG_OOB)
864 		goto out;
865 
866 	if (flags & MSG_ERRQUEUE)
867 		return inet_recv_error(sk, msg, len, addr_len);
868 
869 	skb = skb_recv_datagram(sk, flags, noblock, &err);
870 	if (!skb)
871 		goto out;
872 
873 	copied = skb->len;
874 	if (copied > len) {
875 		msg->msg_flags |= MSG_TRUNC;
876 		copied = len;
877 	}
878 
879 	/* Don't bother checking the checksum */
880 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
881 	if (err)
882 		goto done;
883 
884 	sock_recv_timestamp(msg, sk, skb);
885 
886 	/* Copy the address and add cmsg data. */
887 	if (family == AF_INET) {
888 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
889 
890 		if (sin) {
891 			sin->sin_family = AF_INET;
892 			sin->sin_port = 0 /* skb->h.uh->source */;
893 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
894 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
895 			*addr_len = sizeof(*sin);
896 		}
897 
898 		if (isk->cmsg_flags)
899 			ip_cmsg_recv(msg, skb);
900 
901 #if IS_ENABLED(CONFIG_IPV6)
902 	} else if (family == AF_INET6) {
903 		struct ipv6_pinfo *np = inet6_sk(sk);
904 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
905 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
906 
907 		if (sin6) {
908 			sin6->sin6_family = AF_INET6;
909 			sin6->sin6_port = 0;
910 			sin6->sin6_addr = ip6->saddr;
911 			sin6->sin6_flowinfo = 0;
912 			if (np->sndflow)
913 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
914 			sin6->sin6_scope_id =
915 				ipv6_iface_scope_id(&sin6->sin6_addr,
916 						    inet6_iif(skb));
917 			*addr_len = sizeof(*sin6);
918 		}
919 
920 		if (inet6_sk(sk)->rxopt.all)
921 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
922 		if (skb->protocol == htons(ETH_P_IPV6) &&
923 		    inet6_sk(sk)->rxopt.all)
924 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
925 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
926 			ip_cmsg_recv(msg, skb);
927 #endif
928 	} else {
929 		BUG();
930 	}
931 
932 	err = copied;
933 
934 done:
935 	skb_free_datagram(sk, skb);
936 out:
937 	pr_debug("ping_recvmsg -> %d\n", err);
938 	return err;
939 }
940 EXPORT_SYMBOL_GPL(ping_recvmsg);
941 
942 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
943 {
944 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
945 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
946 	if (sock_queue_rcv_skb(sk, skb) < 0) {
947 		kfree_skb(skb);
948 		pr_debug("ping_queue_rcv_skb -> failed\n");
949 		return -1;
950 	}
951 	return 0;
952 }
953 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
954 
955 
956 /*
957  *	All we need to do is get the socket.
958  */
959 
960 bool ping_rcv(struct sk_buff *skb)
961 {
962 	struct sock *sk;
963 	struct net *net = dev_net(skb->dev);
964 	struct icmphdr *icmph = icmp_hdr(skb);
965 
966 	/* We assume the packet has already been checked by icmp_rcv */
967 
968 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
969 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
970 
971 	/* Push ICMP header back */
972 	skb_push(skb, skb->data - (u8 *)icmph);
973 
974 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
975 	if (sk != NULL) {
976 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
977 
978 		pr_debug("rcv on socket %p\n", sk);
979 		if (skb2)
980 			ping_queue_rcv_skb(sk, skb2);
981 		sock_put(sk);
982 		return true;
983 	}
984 	pr_debug("no socket, dropping\n");
985 
986 	return false;
987 }
988 EXPORT_SYMBOL_GPL(ping_rcv);
989 
990 struct proto ping_prot = {
991 	.name =		"PING",
992 	.owner =	THIS_MODULE,
993 	.init =		ping_init_sock,
994 	.close =	ping_close,
995 	.connect =	ip4_datagram_connect,
996 	.disconnect =	udp_disconnect,
997 	.setsockopt =	ip_setsockopt,
998 	.getsockopt =	ip_getsockopt,
999 	.sendmsg =	ping_v4_sendmsg,
1000 	.recvmsg =	ping_recvmsg,
1001 	.bind =		ping_bind,
1002 	.backlog_rcv =	ping_queue_rcv_skb,
1003 	.release_cb =	ip4_datagram_release_cb,
1004 	.hash =		ping_hash,
1005 	.unhash =	ping_unhash,
1006 	.get_port =	ping_get_port,
1007 	.obj_size =	sizeof(struct inet_sock),
1008 };
1009 EXPORT_SYMBOL(ping_prot);
1010 
1011 #ifdef CONFIG_PROC_FS
1012 
1013 static struct sock *ping_get_first(struct seq_file *seq, int start)
1014 {
1015 	struct sock *sk;
1016 	struct ping_iter_state *state = seq->private;
1017 	struct net *net = seq_file_net(seq);
1018 
1019 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1020 	     ++state->bucket) {
1021 		struct hlist_nulls_node *node;
1022 		struct hlist_nulls_head *hslot;
1023 
1024 		hslot = &ping_table.hash[state->bucket];
1025 
1026 		if (hlist_nulls_empty(hslot))
1027 			continue;
1028 
1029 		sk_nulls_for_each(sk, node, hslot) {
1030 			if (net_eq(sock_net(sk), net) &&
1031 			    sk->sk_family == state->family)
1032 				goto found;
1033 		}
1034 	}
1035 	sk = NULL;
1036 found:
1037 	return sk;
1038 }
1039 
1040 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1041 {
1042 	struct ping_iter_state *state = seq->private;
1043 	struct net *net = seq_file_net(seq);
1044 
1045 	do {
1046 		sk = sk_nulls_next(sk);
1047 	} while (sk && (!net_eq(sock_net(sk), net)));
1048 
1049 	if (!sk)
1050 		return ping_get_first(seq, state->bucket + 1);
1051 	return sk;
1052 }
1053 
1054 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1055 {
1056 	struct sock *sk = ping_get_first(seq, 0);
1057 
1058 	if (sk)
1059 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1060 			--pos;
1061 	return pos ? NULL : sk;
1062 }
1063 
1064 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1065 {
1066 	struct ping_iter_state *state = seq->private;
1067 	state->bucket = 0;
1068 	state->family = family;
1069 
1070 	read_lock_bh(&ping_table.lock);
1071 
1072 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1073 }
1074 EXPORT_SYMBOL_GPL(ping_seq_start);
1075 
1076 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1077 {
1078 	return ping_seq_start(seq, pos, AF_INET);
1079 }
1080 
1081 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1082 {
1083 	struct sock *sk;
1084 
1085 	if (v == SEQ_START_TOKEN)
1086 		sk = ping_get_idx(seq, 0);
1087 	else
1088 		sk = ping_get_next(seq, v);
1089 
1090 	++*pos;
1091 	return sk;
1092 }
1093 EXPORT_SYMBOL_GPL(ping_seq_next);
1094 
1095 void ping_seq_stop(struct seq_file *seq, void *v)
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 %d",
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 		atomic_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 	.show		= ping_v4_seq_show,
1140 	.start		= ping_v4_seq_start,
1141 	.next		= ping_seq_next,
1142 	.stop		= ping_seq_stop,
1143 };
1144 
1145 static int ping_seq_open(struct inode *inode, struct file *file)
1146 {
1147 	struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1148 	return seq_open_net(inode, file, &afinfo->seq_ops,
1149 			   sizeof(struct ping_iter_state));
1150 }
1151 
1152 const struct file_operations ping_seq_fops = {
1153 	.open		= ping_seq_open,
1154 	.read		= seq_read,
1155 	.llseek		= seq_lseek,
1156 	.release	= seq_release_net,
1157 };
1158 EXPORT_SYMBOL_GPL(ping_seq_fops);
1159 
1160 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1161 	.name		= "icmp",
1162 	.family		= AF_INET,
1163 	.seq_fops	= &ping_seq_fops,
1164 	.seq_ops	= {
1165 		.start		= ping_v4_seq_start,
1166 		.show		= ping_v4_seq_show,
1167 		.next		= ping_seq_next,
1168 		.stop		= ping_seq_stop,
1169 	},
1170 };
1171 
1172 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1173 {
1174 	struct proc_dir_entry *p;
1175 	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1176 			     afinfo->seq_fops, afinfo);
1177 	if (!p)
1178 		return -ENOMEM;
1179 	return 0;
1180 }
1181 EXPORT_SYMBOL_GPL(ping_proc_register);
1182 
1183 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1184 {
1185 	remove_proc_entry(afinfo->name, net->proc_net);
1186 }
1187 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1188 
1189 static int __net_init ping_v4_proc_init_net(struct net *net)
1190 {
1191 	return ping_proc_register(net, &ping_v4_seq_afinfo);
1192 }
1193 
1194 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1195 {
1196 	ping_proc_unregister(net, &ping_v4_seq_afinfo);
1197 }
1198 
1199 static struct pernet_operations ping_v4_net_ops = {
1200 	.init = ping_v4_proc_init_net,
1201 	.exit = ping_v4_proc_exit_net,
1202 };
1203 
1204 int __init ping_proc_init(void)
1205 {
1206 	return register_pernet_subsys(&ping_v4_net_ops);
1207 }
1208 
1209 void ping_proc_exit(void)
1210 {
1211 	unregister_pernet_subsys(&ping_v4_net_ops);
1212 }
1213 
1214 #endif
1215 
1216 void __init ping_init(void)
1217 {
1218 	int i;
1219 
1220 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1221 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1222 	rwlock_init(&ping_table.lock);
1223 }
1224