xref: /linux/net/ipv4/ping.c (revision b86406d42ae3c41ae0ce332ea24350829b88af51)
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 <linux/bpf-cgroup.h>
37 #include <net/sock.h>
38 #include <net/ping.h>
39 #include <net/udp.h>
40 #include <net/route.h>
41 #include <net/inet_common.h>
42 #include <net/checksum.h>
43 
44 #if IS_ENABLED(CONFIG_IPV6)
45 #include <linux/in6.h>
46 #include <linux/icmpv6.h>
47 #include <net/addrconf.h>
48 #include <net/ipv6.h>
49 #include <net/transp_v6.h>
50 #endif
51 
52 struct ping_table {
53 	struct hlist_nulls_head	hash[PING_HTABLE_SIZE];
54 	spinlock_t		lock;
55 };
56 
57 static struct ping_table ping_table;
58 struct pingv6_ops pingv6_ops;
59 EXPORT_SYMBOL_GPL(pingv6_ops);
60 
61 static u16 ping_port_rover;
62 
63 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
64 {
65 	u32 res = (num + net_hash_mix(net)) & mask;
66 
67 	pr_debug("hash(%u) = %u\n", num, res);
68 	return res;
69 }
70 EXPORT_SYMBOL_GPL(ping_hash);
71 
72 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
73 					     struct net *net, unsigned int num)
74 {
75 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
76 }
77 
78 int ping_get_port(struct sock *sk, unsigned short ident)
79 {
80 	struct hlist_nulls_node *node;
81 	struct hlist_nulls_head *hlist;
82 	struct inet_sock *isk, *isk2;
83 	struct sock *sk2 = NULL;
84 
85 	isk = inet_sk(sk);
86 	spin_lock(&ping_table.lock);
87 	if (ident == 0) {
88 		u32 i;
89 		u16 result = ping_port_rover + 1;
90 
91 		for (i = 0; i < (1L << 16); i++, result++) {
92 			if (!result)
93 				result++; /* avoid zero */
94 			hlist = ping_hashslot(&ping_table, sock_net(sk),
95 					    result);
96 			ping_portaddr_for_each_entry(sk2, node, hlist) {
97 				isk2 = inet_sk(sk2);
98 
99 				if (isk2->inet_num == result)
100 					goto next_port;
101 			}
102 
103 			/* found */
104 			ping_port_rover = ident = result;
105 			break;
106 next_port:
107 			;
108 		}
109 		if (i >= (1L << 16))
110 			goto fail;
111 	} else {
112 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
113 		ping_portaddr_for_each_entry(sk2, node, hlist) {
114 			isk2 = inet_sk(sk2);
115 
116 			/* BUG? Why is this reuse and not reuseaddr? ping.c
117 			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
118 			 * that other ping processes can steal its packets.
119 			 */
120 			if ((isk2->inet_num == ident) &&
121 			    (sk2 != sk) &&
122 			    (!sk2->sk_reuse || !sk->sk_reuse))
123 				goto fail;
124 		}
125 	}
126 
127 	pr_debug("found port/ident = %d\n", ident);
128 	isk->inet_num = ident;
129 	if (sk_unhashed(sk)) {
130 		pr_debug("was not hashed\n");
131 		sock_hold(sk);
132 		sock_set_flag(sk, SOCK_RCU_FREE);
133 		hlist_nulls_add_head_rcu(&sk->sk_nulls_node, hlist);
134 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
135 	}
136 	spin_unlock(&ping_table.lock);
137 	return 0;
138 
139 fail:
140 	spin_unlock(&ping_table.lock);
141 	return 1;
142 }
143 EXPORT_SYMBOL_GPL(ping_get_port);
144 
145 int ping_hash(struct sock *sk)
146 {
147 	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
148 	BUG(); /* "Please do not press this button again." */
149 
150 	return 0;
151 }
152 
153 void ping_unhash(struct sock *sk)
154 {
155 	struct inet_sock *isk = inet_sk(sk);
156 
157 	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
158 	spin_lock(&ping_table.lock);
159 	if (sk_hashed(sk)) {
160 		hlist_nulls_del_init_rcu(&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 	}
166 	spin_unlock(&ping_table.lock);
167 }
168 EXPORT_SYMBOL_GPL(ping_unhash);
169 
170 /* Called under rcu_read_lock() */
171 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
172 {
173 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
174 	struct sock *sk = NULL;
175 	struct inet_sock *isk;
176 	struct hlist_nulls_node *hnode;
177 	int dif, sdif;
178 
179 	if (skb->protocol == htons(ETH_P_IP)) {
180 		dif = inet_iif(skb);
181 		sdif = inet_sdif(skb);
182 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
183 			 (int)ident, &ip_hdr(skb)->daddr, dif);
184 #if IS_ENABLED(CONFIG_IPV6)
185 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
186 		dif = inet6_iif(skb);
187 		sdif = inet6_sdif(skb);
188 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
189 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
190 #endif
191 	} else {
192 		return NULL;
193 	}
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 		goto exit;
234 	}
235 
236 	sk = NULL;
237 exit:
238 
239 	return sk;
240 }
241 
242 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
243 					  kgid_t *high)
244 {
245 	kgid_t *data = net->ipv4.ping_group_range.range;
246 	unsigned int seq;
247 
248 	do {
249 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
250 
251 		*low = data[0];
252 		*high = data[1];
253 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
254 }
255 
256 
257 int ping_init_sock(struct sock *sk)
258 {
259 	struct net *net = sock_net(sk);
260 	kgid_t group = current_egid();
261 	struct group_info *group_info;
262 	int i;
263 	kgid_t low, high;
264 	int ret = 0;
265 
266 	if (sk->sk_family == AF_INET6)
267 		sk->sk_ipv6only = 1;
268 
269 	inet_get_ping_group_range_net(net, &low, &high);
270 	if (gid_lte(low, group) && gid_lte(group, high))
271 		return 0;
272 
273 	group_info = get_current_groups();
274 	for (i = 0; i < group_info->ngroups; i++) {
275 		kgid_t gid = group_info->gid[i];
276 
277 		if (gid_lte(low, gid) && gid_lte(gid, high))
278 			goto out_release_group;
279 	}
280 
281 	ret = -EACCES;
282 
283 out_release_group:
284 	put_group_info(group_info);
285 	return ret;
286 }
287 EXPORT_SYMBOL_GPL(ping_init_sock);
288 
289 void ping_close(struct sock *sk, long timeout)
290 {
291 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
292 		 inet_sk(sk), inet_sk(sk)->inet_num);
293 	pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
294 
295 	sk_common_release(sk);
296 }
297 EXPORT_SYMBOL_GPL(ping_close);
298 
299 static int ping_pre_connect(struct sock *sk, struct sockaddr *uaddr,
300 			    int addr_len)
301 {
302 	/* This check is replicated from __ip4_datagram_connect() and
303 	 * intended to prevent BPF program called below from accessing bytes
304 	 * that are out of the bound specified by user in addr_len.
305 	 */
306 	if (addr_len < sizeof(struct sockaddr_in))
307 		return -EINVAL;
308 
309 	return BPF_CGROUP_RUN_PROG_INET4_CONNECT_LOCK(sk, uaddr);
310 }
311 
312 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
313 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
314 				struct sockaddr *uaddr, int addr_len)
315 {
316 	struct net *net = sock_net(sk);
317 	if (sk->sk_family == AF_INET) {
318 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
319 		u32 tb_id = RT_TABLE_LOCAL;
320 		int chk_addr_ret;
321 
322 		if (addr_len < sizeof(*addr))
323 			return -EINVAL;
324 
325 		if (addr->sin_family != AF_INET &&
326 		    !(addr->sin_family == AF_UNSPEC &&
327 		      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
328 			return -EAFNOSUPPORT;
329 
330 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
331 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
332 
333 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
334 			return 0;
335 
336 		tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
337 		chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
338 
339 		if (chk_addr_ret == RTN_MULTICAST ||
340 		    chk_addr_ret == RTN_BROADCAST ||
341 		    (chk_addr_ret != RTN_LOCAL &&
342 		     !inet_can_nonlocal_bind(net, isk)))
343 			return -EADDRNOTAVAIL;
344 
345 #if IS_ENABLED(CONFIG_IPV6)
346 	} else if (sk->sk_family == AF_INET6) {
347 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
348 		int addr_type, scoped, has_addr;
349 		struct net_device *dev = NULL;
350 
351 		if (addr_len < sizeof(*addr))
352 			return -EINVAL;
353 
354 		if (addr->sin6_family != AF_INET6)
355 			return -EAFNOSUPPORT;
356 
357 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
358 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
359 
360 		addr_type = ipv6_addr_type(&addr->sin6_addr);
361 		scoped = __ipv6_addr_needs_scope_id(addr_type);
362 		if ((addr_type != IPV6_ADDR_ANY &&
363 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
364 		    (scoped && !addr->sin6_scope_id))
365 			return -EINVAL;
366 
367 		rcu_read_lock();
368 		if (addr->sin6_scope_id) {
369 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
370 			if (!dev) {
371 				rcu_read_unlock();
372 				return -ENODEV;
373 			}
374 		}
375 
376 		if (!dev && sk->sk_bound_dev_if) {
377 			dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
378 			if (!dev) {
379 				rcu_read_unlock();
380 				return -ENODEV;
381 			}
382 		}
383 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
384 						    scoped);
385 		rcu_read_unlock();
386 
387 		if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
388 		      addr_type == IPV6_ADDR_ANY))
389 			return -EADDRNOTAVAIL;
390 
391 		if (scoped)
392 			sk->sk_bound_dev_if = addr->sin6_scope_id;
393 #endif
394 	} else {
395 		return -EAFNOSUPPORT;
396 	}
397 	return 0;
398 }
399 
400 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
401 {
402 	if (saddr->sa_family == AF_INET) {
403 		struct inet_sock *isk = inet_sk(sk);
404 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
405 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
406 #if IS_ENABLED(CONFIG_IPV6)
407 	} else if (saddr->sa_family == AF_INET6) {
408 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
409 		struct ipv6_pinfo *np = inet6_sk(sk);
410 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
411 #endif
412 	}
413 }
414 
415 /*
416  * We need our own bind because there are no privileged id's == local ports.
417  * Moreover, we don't allow binding to multi- and broadcast addresses.
418  */
419 
420 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
421 {
422 	struct inet_sock *isk = inet_sk(sk);
423 	unsigned short snum;
424 	int err;
425 	int dif = sk->sk_bound_dev_if;
426 
427 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
428 	if (err)
429 		return err;
430 
431 	lock_sock(sk);
432 
433 	err = -EINVAL;
434 	if (isk->inet_num != 0)
435 		goto out;
436 
437 	err = -EADDRINUSE;
438 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
439 	if (ping_get_port(sk, snum) != 0) {
440 		/* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
441 		sk->sk_bound_dev_if = dif;
442 		goto out;
443 	}
444 	ping_set_saddr(sk, uaddr);
445 
446 	pr_debug("after bind(): num = %hu, dif = %d\n",
447 		 isk->inet_num,
448 		 sk->sk_bound_dev_if);
449 
450 	err = 0;
451 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
452 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
453 #if IS_ENABLED(CONFIG_IPV6)
454 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
455 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
456 #endif
457 
458 	if (snum)
459 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
460 	isk->inet_sport = htons(isk->inet_num);
461 	isk->inet_daddr = 0;
462 	isk->inet_dport = 0;
463 
464 #if IS_ENABLED(CONFIG_IPV6)
465 	if (sk->sk_family == AF_INET6)
466 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
467 #endif
468 
469 	sk_dst_reset(sk);
470 out:
471 	release_sock(sk);
472 	pr_debug("ping_v4_bind -> %d\n", err);
473 	return err;
474 }
475 EXPORT_SYMBOL_GPL(ping_bind);
476 
477 /*
478  * Is this a supported type of ICMP message?
479  */
480 
481 static inline int ping_supported(int family, int type, int code)
482 {
483 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
484 	       (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
485 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
486 	       (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
487 }
488 
489 /*
490  * This routine is called by the ICMP module when it gets some
491  * sort of error condition.
492  */
493 
494 void ping_err(struct sk_buff *skb, int offset, u32 info)
495 {
496 	int family;
497 	struct icmphdr *icmph;
498 	struct inet_sock *inet_sock;
499 	int type;
500 	int code;
501 	struct net *net = dev_net(skb->dev);
502 	struct sock *sk;
503 	int harderr;
504 	int err;
505 
506 	if (skb->protocol == htons(ETH_P_IP)) {
507 		family = AF_INET;
508 		type = icmp_hdr(skb)->type;
509 		code = icmp_hdr(skb)->code;
510 		icmph = (struct icmphdr *)(skb->data + offset);
511 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
512 		family = AF_INET6;
513 		type = icmp6_hdr(skb)->icmp6_type;
514 		code = icmp6_hdr(skb)->icmp6_code;
515 		icmph = (struct icmphdr *) (skb->data + offset);
516 	} else {
517 		BUG();
518 	}
519 
520 	/* We assume the packet has already been checked by icmp_unreach */
521 
522 	if (!ping_supported(family, icmph->type, icmph->code))
523 		return;
524 
525 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
526 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
527 		 ntohs(icmph->un.echo.sequence));
528 
529 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
530 	if (!sk) {
531 		pr_debug("no socket, dropping\n");
532 		return;	/* No socket for error */
533 	}
534 	pr_debug("err on socket %p\n", sk);
535 
536 	err = 0;
537 	harderr = 0;
538 	inet_sock = inet_sk(sk);
539 
540 	if (skb->protocol == htons(ETH_P_IP)) {
541 		switch (type) {
542 		default:
543 		case ICMP_TIME_EXCEEDED:
544 			err = EHOSTUNREACH;
545 			break;
546 		case ICMP_SOURCE_QUENCH:
547 			/* This is not a real error but ping wants to see it.
548 			 * Report it with some fake errno.
549 			 */
550 			err = EREMOTEIO;
551 			break;
552 		case ICMP_PARAMETERPROB:
553 			err = EPROTO;
554 			harderr = 1;
555 			break;
556 		case ICMP_DEST_UNREACH:
557 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
558 				ipv4_sk_update_pmtu(skb, sk, info);
559 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
560 					err = EMSGSIZE;
561 					harderr = 1;
562 					break;
563 				}
564 				goto out;
565 			}
566 			err = EHOSTUNREACH;
567 			if (code <= NR_ICMP_UNREACH) {
568 				harderr = icmp_err_convert[code].fatal;
569 				err = icmp_err_convert[code].errno;
570 			}
571 			break;
572 		case ICMP_REDIRECT:
573 			/* See ICMP_SOURCE_QUENCH */
574 			ipv4_sk_redirect(skb, sk);
575 			err = EREMOTEIO;
576 			break;
577 		}
578 #if IS_ENABLED(CONFIG_IPV6)
579 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
580 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
581 #endif
582 	}
583 
584 	/*
585 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
586 	 *	4.1.3.3.
587 	 */
588 	if ((family == AF_INET && !inet_sock->recverr) ||
589 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
590 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
591 			goto out;
592 	} else {
593 		if (family == AF_INET) {
594 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
595 				      info, (u8 *)icmph);
596 #if IS_ENABLED(CONFIG_IPV6)
597 		} else if (family == AF_INET6) {
598 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
599 						   info, (u8 *)icmph);
600 #endif
601 		}
602 	}
603 	sk->sk_err = err;
604 	sk_error_report(sk);
605 out:
606 	return;
607 }
608 EXPORT_SYMBOL_GPL(ping_err);
609 
610 /*
611  *	Copy and checksum an ICMP Echo packet from user space into a buffer
612  *	starting from the payload.
613  */
614 
615 int ping_getfrag(void *from, char *to,
616 		 int offset, int fraglen, int odd, struct sk_buff *skb)
617 {
618 	struct pingfakehdr *pfh = from;
619 
620 	if (offset == 0) {
621 		fraglen -= sizeof(struct icmphdr);
622 		if (fraglen < 0)
623 			BUG();
624 		if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
625 			    fraglen, &pfh->wcheck,
626 			    &pfh->msg->msg_iter))
627 			return -EFAULT;
628 	} else if (offset < sizeof(struct icmphdr)) {
629 			BUG();
630 	} else {
631 		if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
632 					    &pfh->msg->msg_iter))
633 			return -EFAULT;
634 	}
635 
636 #if IS_ENABLED(CONFIG_IPV6)
637 	/* For IPv6, checksum each skb as we go along, as expected by
638 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
639 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
640 	 */
641 	if (pfh->family == AF_INET6) {
642 		skb->csum = pfh->wcheck;
643 		skb->ip_summed = CHECKSUM_NONE;
644 		pfh->wcheck = 0;
645 	}
646 #endif
647 
648 	return 0;
649 }
650 EXPORT_SYMBOL_GPL(ping_getfrag);
651 
652 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
653 				       struct flowi4 *fl4)
654 {
655 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
656 
657 	if (!skb)
658 		return 0;
659 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
660 		sizeof(struct icmphdr), pfh->wcheck);
661 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
662 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
663 	skb->ip_summed = CHECKSUM_NONE;
664 	return ip_push_pending_frames(sk, fl4);
665 }
666 
667 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
668 			void *user_icmph, size_t icmph_len)
669 {
670 	u8 type, code;
671 
672 	if (len > 0xFFFF)
673 		return -EMSGSIZE;
674 
675 	/* Must have at least a full ICMP header. */
676 	if (len < icmph_len)
677 		return -EINVAL;
678 
679 	/*
680 	 *	Check the flags.
681 	 */
682 
683 	/* Mirror BSD error message compatibility */
684 	if (msg->msg_flags & MSG_OOB)
685 		return -EOPNOTSUPP;
686 
687 	/*
688 	 *	Fetch the ICMP header provided by the userland.
689 	 *	iovec is modified! The ICMP header is consumed.
690 	 */
691 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
692 		return -EFAULT;
693 
694 	if (family == AF_INET) {
695 		type = ((struct icmphdr *) user_icmph)->type;
696 		code = ((struct icmphdr *) user_icmph)->code;
697 #if IS_ENABLED(CONFIG_IPV6)
698 	} else if (family == AF_INET6) {
699 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
700 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
701 #endif
702 	} else {
703 		BUG();
704 	}
705 
706 	if (!ping_supported(family, type, code))
707 		return -EINVAL;
708 
709 	return 0;
710 }
711 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
712 
713 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
714 {
715 	struct net *net = sock_net(sk);
716 	struct flowi4 fl4;
717 	struct inet_sock *inet = inet_sk(sk);
718 	struct ipcm_cookie ipc;
719 	struct icmphdr user_icmph;
720 	struct pingfakehdr pfh;
721 	struct rtable *rt = NULL;
722 	struct ip_options_data opt_copy;
723 	int free = 0;
724 	__be32 saddr, daddr, faddr;
725 	u8  tos;
726 	int err;
727 
728 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
729 
730 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
731 				  sizeof(user_icmph));
732 	if (err)
733 		return err;
734 
735 	/*
736 	 *	Get and verify the address.
737 	 */
738 
739 	if (msg->msg_name) {
740 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
741 		if (msg->msg_namelen < sizeof(*usin))
742 			return -EINVAL;
743 		if (usin->sin_family != AF_INET)
744 			return -EAFNOSUPPORT;
745 		daddr = usin->sin_addr.s_addr;
746 		/* no remote port */
747 	} else {
748 		if (sk->sk_state != TCP_ESTABLISHED)
749 			return -EDESTADDRREQ;
750 		daddr = inet->inet_daddr;
751 		/* no remote port */
752 	}
753 
754 	ipcm_init_sk(&ipc, inet);
755 
756 	if (msg->msg_controllen) {
757 		err = ip_cmsg_send(sk, msg, &ipc, false);
758 		if (unlikely(err)) {
759 			kfree(ipc.opt);
760 			return err;
761 		}
762 		if (ipc.opt)
763 			free = 1;
764 	}
765 	if (!ipc.opt) {
766 		struct ip_options_rcu *inet_opt;
767 
768 		rcu_read_lock();
769 		inet_opt = rcu_dereference(inet->inet_opt);
770 		if (inet_opt) {
771 			memcpy(&opt_copy, inet_opt,
772 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
773 			ipc.opt = &opt_copy.opt;
774 		}
775 		rcu_read_unlock();
776 	}
777 
778 	saddr = ipc.addr;
779 	ipc.addr = faddr = daddr;
780 
781 	if (ipc.opt && ipc.opt->opt.srr) {
782 		if (!daddr) {
783 			err = -EINVAL;
784 			goto out_free;
785 		}
786 		faddr = ipc.opt->opt.faddr;
787 	}
788 	tos = get_rttos(&ipc, inet);
789 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
790 	    (msg->msg_flags & MSG_DONTROUTE) ||
791 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
792 		tos |= RTO_ONLINK;
793 	}
794 
795 	if (ipv4_is_multicast(daddr)) {
796 		if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
797 			ipc.oif = inet->mc_index;
798 		if (!saddr)
799 			saddr = inet->mc_addr;
800 	} else if (!ipc.oif)
801 		ipc.oif = inet->uc_index;
802 
803 	flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
804 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
805 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
806 			   sk->sk_uid);
807 
808 	fl4.fl4_icmp_type = user_icmph.type;
809 	fl4.fl4_icmp_code = user_icmph.code;
810 
811 	security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
812 	rt = ip_route_output_flow(net, &fl4, sk);
813 	if (IS_ERR(rt)) {
814 		err = PTR_ERR(rt);
815 		rt = NULL;
816 		if (err == -ENETUNREACH)
817 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
818 		goto out;
819 	}
820 
821 	err = -EACCES;
822 	if ((rt->rt_flags & RTCF_BROADCAST) &&
823 	    !sock_flag(sk, SOCK_BROADCAST))
824 		goto out;
825 
826 	if (msg->msg_flags & MSG_CONFIRM)
827 		goto do_confirm;
828 back_from_confirm:
829 
830 	if (!ipc.addr)
831 		ipc.addr = fl4.daddr;
832 
833 	lock_sock(sk);
834 
835 	pfh.icmph.type = user_icmph.type; /* already checked */
836 	pfh.icmph.code = user_icmph.code; /* ditto */
837 	pfh.icmph.checksum = 0;
838 	pfh.icmph.un.echo.id = inet->inet_sport;
839 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
840 	pfh.msg = msg;
841 	pfh.wcheck = 0;
842 	pfh.family = AF_INET;
843 
844 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
845 			0, &ipc, &rt, msg->msg_flags);
846 	if (err)
847 		ip_flush_pending_frames(sk);
848 	else
849 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
850 	release_sock(sk);
851 
852 out:
853 	ip_rt_put(rt);
854 out_free:
855 	if (free)
856 		kfree(ipc.opt);
857 	if (!err) {
858 		icmp_out_count(sock_net(sk), user_icmph.type);
859 		return len;
860 	}
861 	return err;
862 
863 do_confirm:
864 	if (msg->msg_flags & MSG_PROBE)
865 		dst_confirm_neigh(&rt->dst, &fl4.daddr);
866 	if (!(msg->msg_flags & MSG_PROBE) || len)
867 		goto back_from_confirm;
868 	err = 0;
869 	goto out;
870 }
871 
872 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int flags,
873 		 int *addr_len)
874 {
875 	struct inet_sock *isk = inet_sk(sk);
876 	int family = sk->sk_family;
877 	struct sk_buff *skb;
878 	int copied, err;
879 
880 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
881 
882 	err = -EOPNOTSUPP;
883 	if (flags & MSG_OOB)
884 		goto out;
885 
886 	if (flags & MSG_ERRQUEUE)
887 		return inet_recv_error(sk, msg, len, addr_len);
888 
889 	skb = skb_recv_datagram(sk, flags, &err);
890 	if (!skb)
891 		goto out;
892 
893 	copied = skb->len;
894 	if (copied > len) {
895 		msg->msg_flags |= MSG_TRUNC;
896 		copied = len;
897 	}
898 
899 	/* Don't bother checking the checksum */
900 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
901 	if (err)
902 		goto done;
903 
904 	sock_recv_timestamp(msg, sk, skb);
905 
906 	/* Copy the address and add cmsg data. */
907 	if (family == AF_INET) {
908 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
909 
910 		if (sin) {
911 			sin->sin_family = AF_INET;
912 			sin->sin_port = 0 /* skb->h.uh->source */;
913 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
914 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
915 			*addr_len = sizeof(*sin);
916 		}
917 
918 		if (isk->cmsg_flags)
919 			ip_cmsg_recv(msg, skb);
920 
921 #if IS_ENABLED(CONFIG_IPV6)
922 	} else if (family == AF_INET6) {
923 		struct ipv6_pinfo *np = inet6_sk(sk);
924 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
925 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
926 
927 		if (sin6) {
928 			sin6->sin6_family = AF_INET6;
929 			sin6->sin6_port = 0;
930 			sin6->sin6_addr = ip6->saddr;
931 			sin6->sin6_flowinfo = 0;
932 			if (np->sndflow)
933 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
934 			sin6->sin6_scope_id =
935 				ipv6_iface_scope_id(&sin6->sin6_addr,
936 						    inet6_iif(skb));
937 			*addr_len = sizeof(*sin6);
938 		}
939 
940 		if (inet6_sk(sk)->rxopt.all)
941 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
942 		if (skb->protocol == htons(ETH_P_IPV6) &&
943 		    inet6_sk(sk)->rxopt.all)
944 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
945 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
946 			ip_cmsg_recv(msg, skb);
947 #endif
948 	} else {
949 		BUG();
950 	}
951 
952 	err = copied;
953 
954 done:
955 	skb_free_datagram(sk, skb);
956 out:
957 	pr_debug("ping_recvmsg -> %d\n", err);
958 	return err;
959 }
960 EXPORT_SYMBOL_GPL(ping_recvmsg);
961 
962 static enum skb_drop_reason __ping_queue_rcv_skb(struct sock *sk,
963 						 struct sk_buff *skb)
964 {
965 	enum skb_drop_reason reason;
966 
967 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
968 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
969 	if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) {
970 		kfree_skb_reason(skb, reason);
971 		pr_debug("ping_queue_rcv_skb -> failed\n");
972 		return reason;
973 	}
974 	return SKB_NOT_DROPPED_YET;
975 }
976 
977 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
978 {
979 	return __ping_queue_rcv_skb(sk, skb) ? -1 : 0;
980 }
981 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
982 
983 
984 /*
985  *	All we need to do is get the socket.
986  */
987 
988 enum skb_drop_reason ping_rcv(struct sk_buff *skb)
989 {
990 	enum skb_drop_reason reason = SKB_DROP_REASON_NO_SOCKET;
991 	struct sock *sk;
992 	struct net *net = dev_net(skb->dev);
993 	struct icmphdr *icmph = icmp_hdr(skb);
994 
995 	/* We assume the packet has already been checked by icmp_rcv */
996 
997 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
998 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
999 
1000 	/* Push ICMP header back */
1001 	skb_push(skb, skb->data - (u8 *)icmph);
1002 
1003 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
1004 	if (sk) {
1005 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1006 
1007 		pr_debug("rcv on socket %p\n", sk);
1008 		if (skb2)
1009 			reason = __ping_queue_rcv_skb(sk, skb2);
1010 		else
1011 			reason = SKB_DROP_REASON_NOMEM;
1012 	}
1013 
1014 	if (reason)
1015 		pr_debug("no socket, dropping\n");
1016 
1017 	return reason;
1018 }
1019 EXPORT_SYMBOL_GPL(ping_rcv);
1020 
1021 struct proto ping_prot = {
1022 	.name =		"PING",
1023 	.owner =	THIS_MODULE,
1024 	.init =		ping_init_sock,
1025 	.close =	ping_close,
1026 	.pre_connect =	ping_pre_connect,
1027 	.connect =	ip4_datagram_connect,
1028 	.disconnect =	__udp_disconnect,
1029 	.setsockopt =	ip_setsockopt,
1030 	.getsockopt =	ip_getsockopt,
1031 	.sendmsg =	ping_v4_sendmsg,
1032 	.recvmsg =	ping_recvmsg,
1033 	.bind =		ping_bind,
1034 	.backlog_rcv =	ping_queue_rcv_skb,
1035 	.release_cb =	ip4_datagram_release_cb,
1036 	.hash =		ping_hash,
1037 	.unhash =	ping_unhash,
1038 	.get_port =	ping_get_port,
1039 	.put_port =	ping_unhash,
1040 	.obj_size =	sizeof(struct inet_sock),
1041 };
1042 EXPORT_SYMBOL(ping_prot);
1043 
1044 #ifdef CONFIG_PROC_FS
1045 
1046 static struct sock *ping_get_first(struct seq_file *seq, int start)
1047 {
1048 	struct sock *sk;
1049 	struct ping_iter_state *state = seq->private;
1050 	struct net *net = seq_file_net(seq);
1051 
1052 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1053 	     ++state->bucket) {
1054 		struct hlist_nulls_node *node;
1055 		struct hlist_nulls_head *hslot;
1056 
1057 		hslot = &ping_table.hash[state->bucket];
1058 
1059 		if (hlist_nulls_empty(hslot))
1060 			continue;
1061 
1062 		sk_nulls_for_each(sk, node, hslot) {
1063 			if (net_eq(sock_net(sk), net) &&
1064 			    sk->sk_family == state->family)
1065 				goto found;
1066 		}
1067 	}
1068 	sk = NULL;
1069 found:
1070 	return sk;
1071 }
1072 
1073 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1074 {
1075 	struct ping_iter_state *state = seq->private;
1076 	struct net *net = seq_file_net(seq);
1077 
1078 	do {
1079 		sk = sk_nulls_next(sk);
1080 	} while (sk && (!net_eq(sock_net(sk), net)));
1081 
1082 	if (!sk)
1083 		return ping_get_first(seq, state->bucket + 1);
1084 	return sk;
1085 }
1086 
1087 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1088 {
1089 	struct sock *sk = ping_get_first(seq, 0);
1090 
1091 	if (sk)
1092 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1093 			--pos;
1094 	return pos ? NULL : sk;
1095 }
1096 
1097 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1098 	__acquires(RCU)
1099 {
1100 	struct ping_iter_state *state = seq->private;
1101 	state->bucket = 0;
1102 	state->family = family;
1103 
1104 	rcu_read_lock();
1105 
1106 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1107 }
1108 EXPORT_SYMBOL_GPL(ping_seq_start);
1109 
1110 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1111 {
1112 	return ping_seq_start(seq, pos, AF_INET);
1113 }
1114 
1115 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1116 {
1117 	struct sock *sk;
1118 
1119 	if (v == SEQ_START_TOKEN)
1120 		sk = ping_get_idx(seq, 0);
1121 	else
1122 		sk = ping_get_next(seq, v);
1123 
1124 	++*pos;
1125 	return sk;
1126 }
1127 EXPORT_SYMBOL_GPL(ping_seq_next);
1128 
1129 void ping_seq_stop(struct seq_file *seq, void *v)
1130 	__releases(RCU)
1131 {
1132 	rcu_read_unlock();
1133 }
1134 EXPORT_SYMBOL_GPL(ping_seq_stop);
1135 
1136 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1137 		int bucket)
1138 {
1139 	struct inet_sock *inet = inet_sk(sp);
1140 	__be32 dest = inet->inet_daddr;
1141 	__be32 src = inet->inet_rcv_saddr;
1142 	__u16 destp = ntohs(inet->inet_dport);
1143 	__u16 srcp = ntohs(inet->inet_sport);
1144 
1145 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1146 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1147 		bucket, src, srcp, dest, destp, sp->sk_state,
1148 		sk_wmem_alloc_get(sp),
1149 		sk_rmem_alloc_get(sp),
1150 		0, 0L, 0,
1151 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1152 		0, sock_i_ino(sp),
1153 		refcount_read(&sp->sk_refcnt), sp,
1154 		atomic_read(&sp->sk_drops));
1155 }
1156 
1157 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1158 {
1159 	seq_setwidth(seq, 127);
1160 	if (v == SEQ_START_TOKEN)
1161 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1162 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1163 			   "inode ref pointer drops");
1164 	else {
1165 		struct ping_iter_state *state = seq->private;
1166 
1167 		ping_v4_format_sock(v, seq, state->bucket);
1168 	}
1169 	seq_pad(seq, '\n');
1170 	return 0;
1171 }
1172 
1173 static const struct seq_operations ping_v4_seq_ops = {
1174 	.start		= ping_v4_seq_start,
1175 	.show		= ping_v4_seq_show,
1176 	.next		= ping_seq_next,
1177 	.stop		= ping_seq_stop,
1178 };
1179 
1180 static int __net_init ping_v4_proc_init_net(struct net *net)
1181 {
1182 	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1183 			sizeof(struct ping_iter_state)))
1184 		return -ENOMEM;
1185 	return 0;
1186 }
1187 
1188 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1189 {
1190 	remove_proc_entry("icmp", net->proc_net);
1191 }
1192 
1193 static struct pernet_operations ping_v4_net_ops = {
1194 	.init = ping_v4_proc_init_net,
1195 	.exit = ping_v4_proc_exit_net,
1196 };
1197 
1198 int __init ping_proc_init(void)
1199 {
1200 	return register_pernet_subsys(&ping_v4_net_ops);
1201 }
1202 
1203 void ping_proc_exit(void)
1204 {
1205 	unregister_pernet_subsys(&ping_v4_net_ops);
1206 }
1207 
1208 #endif
1209 
1210 void __init ping_init(void)
1211 {
1212 	int i;
1213 
1214 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1215 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1216 	spin_lock_init(&ping_table.lock);
1217 }
1218