xref: /linux/net/ipv4/ping.c (revision f85f5ae45ad945270a8884261de8249431e8b5a6)
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_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_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 inet_sock *isk, *isk2;
81 	struct hlist_head *hlist;
82 	struct sock *sk2 = NULL;
83 
84 	isk = inet_sk(sk);
85 	spin_lock(&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 			sk_for_each(sk2, 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 		sk_for_each(sk2, 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 		sk_add_node_rcu(sk, hlist);
131 		sock_set_flag(sk, SOCK_RCU_FREE);
132 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
133 	}
134 	spin_unlock(&ping_table.lock);
135 	return 0;
136 
137 fail:
138 	spin_unlock(&ping_table.lock);
139 	return -EADDRINUSE;
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 	spin_lock(&ping_table.lock);
157 	if (sk_del_node_init_rcu(sk)) {
158 		isk->inet_num = 0;
159 		isk->inet_sport = 0;
160 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
161 	}
162 	spin_unlock(&ping_table.lock);
163 }
164 EXPORT_SYMBOL_GPL(ping_unhash);
165 
166 /* Called under rcu_read_lock() */
167 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
168 {
169 	struct hlist_head *hslot = ping_hashslot(&ping_table, net, ident);
170 	struct sock *sk = NULL;
171 	struct inet_sock *isk;
172 	int dif, sdif;
173 
174 	if (skb->protocol == htons(ETH_P_IP)) {
175 		dif = inet_iif(skb);
176 		sdif = inet_sdif(skb);
177 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
178 			 (int)ident, &ip_hdr(skb)->daddr, dif);
179 #if IS_ENABLED(CONFIG_IPV6)
180 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
181 		dif = inet6_iif(skb);
182 		sdif = inet6_sdif(skb);
183 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
184 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
185 #endif
186 	} else {
187 		return NULL;
188 	}
189 
190 	sk_for_each_rcu(sk, 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 		    sk->sk_bound_dev_if != sdif)
226 			continue;
227 
228 		goto exit;
229 	}
230 
231 	sk = NULL;
232 exit:
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 static int ping_pre_connect(struct sock *sk, struct sockaddr *uaddr,
295 			    int addr_len)
296 {
297 	/* This check is replicated from __ip4_datagram_connect() and
298 	 * intended to prevent BPF program called below from accessing bytes
299 	 * that are out of the bound specified by user in addr_len.
300 	 */
301 	if (addr_len < sizeof(struct sockaddr_in))
302 		return -EINVAL;
303 
304 	return BPF_CGROUP_RUN_PROG_INET4_CONNECT_LOCK(sk, uaddr);
305 }
306 
307 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
308 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
309 				struct sockaddr *uaddr, int addr_len)
310 {
311 	struct net *net = sock_net(sk);
312 	if (sk->sk_family == AF_INET) {
313 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
314 		u32 tb_id = RT_TABLE_LOCAL;
315 		int chk_addr_ret;
316 
317 		if (addr_len < sizeof(*addr))
318 			return -EINVAL;
319 
320 		if (addr->sin_family != AF_INET &&
321 		    !(addr->sin_family == AF_UNSPEC &&
322 		      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
323 			return -EAFNOSUPPORT;
324 
325 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
326 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
327 
328 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
329 			return 0;
330 
331 		tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
332 		chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
333 
334 		if (chk_addr_ret == RTN_MULTICAST ||
335 		    chk_addr_ret == RTN_BROADCAST ||
336 		    (chk_addr_ret != RTN_LOCAL &&
337 		     !inet_can_nonlocal_bind(net, isk)))
338 			return -EADDRNOTAVAIL;
339 
340 #if IS_ENABLED(CONFIG_IPV6)
341 	} else if (sk->sk_family == AF_INET6) {
342 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
343 		int addr_type, scoped, has_addr;
344 		struct net_device *dev = NULL;
345 
346 		if (addr_len < sizeof(*addr))
347 			return -EINVAL;
348 
349 		if (addr->sin6_family != AF_INET6)
350 			return -EAFNOSUPPORT;
351 
352 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
353 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
354 
355 		addr_type = ipv6_addr_type(&addr->sin6_addr);
356 		scoped = __ipv6_addr_needs_scope_id(addr_type);
357 		if ((addr_type != IPV6_ADDR_ANY &&
358 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
359 		    (scoped && !addr->sin6_scope_id))
360 			return -EINVAL;
361 
362 		rcu_read_lock();
363 		if (addr->sin6_scope_id) {
364 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
365 			if (!dev) {
366 				rcu_read_unlock();
367 				return -ENODEV;
368 			}
369 		}
370 
371 		if (!dev && sk->sk_bound_dev_if) {
372 			dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
373 			if (!dev) {
374 				rcu_read_unlock();
375 				return -ENODEV;
376 			}
377 		}
378 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
379 						    scoped);
380 		rcu_read_unlock();
381 
382 		if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
383 		      addr_type == IPV6_ADDR_ANY))
384 			return -EADDRNOTAVAIL;
385 
386 		if (scoped)
387 			sk->sk_bound_dev_if = addr->sin6_scope_id;
388 #endif
389 	} else {
390 		return -EAFNOSUPPORT;
391 	}
392 	return 0;
393 }
394 
395 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
396 {
397 	if (saddr->sa_family == AF_INET) {
398 		struct inet_sock *isk = inet_sk(sk);
399 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
400 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
401 #if IS_ENABLED(CONFIG_IPV6)
402 	} else if (saddr->sa_family == AF_INET6) {
403 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
404 		struct ipv6_pinfo *np = inet6_sk(sk);
405 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
406 #endif
407 	}
408 }
409 
410 /*
411  * We need our own bind because there are no privileged id's == local ports.
412  * Moreover, we don't allow binding to multi- and broadcast addresses.
413  */
414 
415 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
416 {
417 	struct inet_sock *isk = inet_sk(sk);
418 	unsigned short snum;
419 	int err;
420 	int dif = sk->sk_bound_dev_if;
421 
422 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
423 	if (err)
424 		return err;
425 
426 	lock_sock(sk);
427 
428 	err = -EINVAL;
429 	if (isk->inet_num != 0)
430 		goto out;
431 
432 	err = -EADDRINUSE;
433 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
434 	if (ping_get_port(sk, snum) != 0) {
435 		/* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
436 		sk->sk_bound_dev_if = dif;
437 		goto out;
438 	}
439 	ping_set_saddr(sk, uaddr);
440 
441 	pr_debug("after bind(): num = %hu, dif = %d\n",
442 		 isk->inet_num,
443 		 sk->sk_bound_dev_if);
444 
445 	err = 0;
446 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
447 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
448 #if IS_ENABLED(CONFIG_IPV6)
449 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
450 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
451 #endif
452 
453 	if (snum)
454 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
455 	isk->inet_sport = htons(isk->inet_num);
456 	isk->inet_daddr = 0;
457 	isk->inet_dport = 0;
458 
459 #if IS_ENABLED(CONFIG_IPV6)
460 	if (sk->sk_family == AF_INET6)
461 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
462 #endif
463 
464 	sk_dst_reset(sk);
465 out:
466 	release_sock(sk);
467 	pr_debug("ping_v4_bind -> %d\n", err);
468 	return err;
469 }
470 EXPORT_SYMBOL_GPL(ping_bind);
471 
472 /*
473  * Is this a supported type of ICMP message?
474  */
475 
476 static inline int ping_supported(int family, int type, int code)
477 {
478 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
479 	       (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
480 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
481 	       (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
482 }
483 
484 /*
485  * This routine is called by the ICMP module when it gets some
486  * sort of error condition.
487  */
488 
489 void ping_err(struct sk_buff *skb, int offset, u32 info)
490 {
491 	int family;
492 	struct icmphdr *icmph;
493 	struct inet_sock *inet_sock;
494 	int type;
495 	int code;
496 	struct net *net = dev_net(skb->dev);
497 	struct sock *sk;
498 	int harderr;
499 	int err;
500 
501 	if (skb->protocol == htons(ETH_P_IP)) {
502 		family = AF_INET;
503 		type = icmp_hdr(skb)->type;
504 		code = icmp_hdr(skb)->code;
505 		icmph = (struct icmphdr *)(skb->data + offset);
506 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
507 		family = AF_INET6;
508 		type = icmp6_hdr(skb)->icmp6_type;
509 		code = icmp6_hdr(skb)->icmp6_code;
510 		icmph = (struct icmphdr *) (skb->data + offset);
511 	} else {
512 		BUG();
513 	}
514 
515 	/* We assume the packet has already been checked by icmp_unreach */
516 
517 	if (!ping_supported(family, icmph->type, icmph->code))
518 		return;
519 
520 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
521 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
522 		 ntohs(icmph->un.echo.sequence));
523 
524 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
525 	if (!sk) {
526 		pr_debug("no socket, dropping\n");
527 		return;	/* No socket for error */
528 	}
529 	pr_debug("err on socket %p\n", sk);
530 
531 	err = 0;
532 	harderr = 0;
533 	inet_sock = inet_sk(sk);
534 
535 	if (skb->protocol == htons(ETH_P_IP)) {
536 		switch (type) {
537 		default:
538 		case ICMP_TIME_EXCEEDED:
539 			err = EHOSTUNREACH;
540 			break;
541 		case ICMP_SOURCE_QUENCH:
542 			/* This is not a real error but ping wants to see it.
543 			 * Report it with some fake errno.
544 			 */
545 			err = EREMOTEIO;
546 			break;
547 		case ICMP_PARAMETERPROB:
548 			err = EPROTO;
549 			harderr = 1;
550 			break;
551 		case ICMP_DEST_UNREACH:
552 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
553 				ipv4_sk_update_pmtu(skb, sk, info);
554 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
555 					err = EMSGSIZE;
556 					harderr = 1;
557 					break;
558 				}
559 				goto out;
560 			}
561 			err = EHOSTUNREACH;
562 			if (code <= NR_ICMP_UNREACH) {
563 				harderr = icmp_err_convert[code].fatal;
564 				err = icmp_err_convert[code].errno;
565 			}
566 			break;
567 		case ICMP_REDIRECT:
568 			/* See ICMP_SOURCE_QUENCH */
569 			ipv4_sk_redirect(skb, sk);
570 			err = EREMOTEIO;
571 			break;
572 		}
573 #if IS_ENABLED(CONFIG_IPV6)
574 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
575 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
576 #endif
577 	}
578 
579 	/*
580 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
581 	 *	4.1.3.3.
582 	 */
583 	if ((family == AF_INET && !inet_test_bit(RECVERR, sk)) ||
584 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
585 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
586 			goto out;
587 	} else {
588 		if (family == AF_INET) {
589 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
590 				      info, (u8 *)icmph);
591 #if IS_ENABLED(CONFIG_IPV6)
592 		} else if (family == AF_INET6) {
593 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
594 						   info, (u8 *)icmph);
595 #endif
596 		}
597 	}
598 	sk->sk_err = err;
599 	sk_error_report(sk);
600 out:
601 	return;
602 }
603 EXPORT_SYMBOL_GPL(ping_err);
604 
605 /*
606  *	Copy and checksum an ICMP Echo packet from user space into a buffer
607  *	starting from the payload.
608  */
609 
610 int ping_getfrag(void *from, char *to,
611 		 int offset, int fraglen, int odd, struct sk_buff *skb)
612 {
613 	struct pingfakehdr *pfh = from;
614 
615 	if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
616 					  &pfh->msg->msg_iter))
617 		return -EFAULT;
618 
619 #if IS_ENABLED(CONFIG_IPV6)
620 	/* For IPv6, checksum each skb as we go along, as expected by
621 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
622 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
623 	 */
624 	if (pfh->family == AF_INET6) {
625 		skb->csum = csum_block_add(skb->csum, pfh->wcheck, odd);
626 		skb->ip_summed = CHECKSUM_NONE;
627 		pfh->wcheck = 0;
628 	}
629 #endif
630 
631 	return 0;
632 }
633 EXPORT_SYMBOL_GPL(ping_getfrag);
634 
635 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
636 				       struct flowi4 *fl4)
637 {
638 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
639 
640 	if (!skb)
641 		return 0;
642 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
643 		sizeof(struct icmphdr), pfh->wcheck);
644 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
645 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
646 	skb->ip_summed = CHECKSUM_NONE;
647 	return ip_push_pending_frames(sk, fl4);
648 }
649 
650 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
651 			void *user_icmph, size_t icmph_len)
652 {
653 	u8 type, code;
654 
655 	if (len > 0xFFFF)
656 		return -EMSGSIZE;
657 
658 	/* Must have at least a full ICMP header. */
659 	if (len < icmph_len)
660 		return -EINVAL;
661 
662 	/*
663 	 *	Check the flags.
664 	 */
665 
666 	/* Mirror BSD error message compatibility */
667 	if (msg->msg_flags & MSG_OOB)
668 		return -EOPNOTSUPP;
669 
670 	/*
671 	 *	Fetch the ICMP header provided by the userland.
672 	 *	iovec is modified! The ICMP header is consumed.
673 	 */
674 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
675 		return -EFAULT;
676 
677 	if (family == AF_INET) {
678 		type = ((struct icmphdr *) user_icmph)->type;
679 		code = ((struct icmphdr *) user_icmph)->code;
680 #if IS_ENABLED(CONFIG_IPV6)
681 	} else if (family == AF_INET6) {
682 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
683 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
684 #endif
685 	} else {
686 		BUG();
687 	}
688 
689 	if (!ping_supported(family, type, code))
690 		return -EINVAL;
691 
692 	return 0;
693 }
694 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
695 
696 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, 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, scope;
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 	ipcm_init_sk(&ipc, inet);
738 
739 	if (msg->msg_controllen) {
740 		err = ip_cmsg_send(sk, msg, &ipc, false);
741 		if (unlikely(err)) {
742 			kfree(ipc.opt);
743 			return err;
744 		}
745 		if (ipc.opt)
746 			free = 1;
747 	}
748 	if (!ipc.opt) {
749 		struct ip_options_rcu *inet_opt;
750 
751 		rcu_read_lock();
752 		inet_opt = rcu_dereference(inet->inet_opt);
753 		if (inet_opt) {
754 			memcpy(&opt_copy, inet_opt,
755 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
756 			ipc.opt = &opt_copy.opt;
757 		}
758 		rcu_read_unlock();
759 	}
760 
761 	saddr = ipc.addr;
762 	ipc.addr = faddr = daddr;
763 
764 	if (ipc.opt && ipc.opt->opt.srr) {
765 		if (!daddr) {
766 			err = -EINVAL;
767 			goto out_free;
768 		}
769 		faddr = ipc.opt->opt.faddr;
770 	}
771 	tos = get_rttos(&ipc, inet);
772 	scope = ip_sendmsg_scope(inet, &ipc, msg);
773 
774 	if (ipv4_is_multicast(daddr)) {
775 		if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
776 			ipc.oif = inet->mc_index;
777 		if (!saddr)
778 			saddr = inet->mc_addr;
779 	} else if (!ipc.oif)
780 		ipc.oif = inet->uc_index;
781 
782 	flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos, scope,
783 			   sk->sk_protocol, inet_sk_flowi_flags(sk), faddr,
784 			   saddr, 0, 0, sk->sk_uid);
785 
786 	fl4.fl4_icmp_type = user_icmph.type;
787 	fl4.fl4_icmp_code = user_icmph.code;
788 
789 	security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
790 	rt = ip_route_output_flow(net, &fl4, sk);
791 	if (IS_ERR(rt)) {
792 		err = PTR_ERR(rt);
793 		rt = NULL;
794 		if (err == -ENETUNREACH)
795 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
796 		goto out;
797 	}
798 
799 	err = -EACCES;
800 	if ((rt->rt_flags & RTCF_BROADCAST) &&
801 	    !sock_flag(sk, SOCK_BROADCAST))
802 		goto out;
803 
804 	if (msg->msg_flags & MSG_CONFIRM)
805 		goto do_confirm;
806 back_from_confirm:
807 
808 	if (!ipc.addr)
809 		ipc.addr = fl4.daddr;
810 
811 	lock_sock(sk);
812 
813 	pfh.icmph.type = user_icmph.type; /* already checked */
814 	pfh.icmph.code = user_icmph.code; /* ditto */
815 	pfh.icmph.checksum = 0;
816 	pfh.icmph.un.echo.id = inet->inet_sport;
817 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
818 	pfh.msg = msg;
819 	pfh.wcheck = 0;
820 	pfh.family = AF_INET;
821 
822 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
823 			     sizeof(struct icmphdr), &ipc, &rt,
824 			     msg->msg_flags);
825 	if (err)
826 		ip_flush_pending_frames(sk);
827 	else
828 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
829 	release_sock(sk);
830 
831 out:
832 	ip_rt_put(rt);
833 out_free:
834 	if (free)
835 		kfree(ipc.opt);
836 	if (!err) {
837 		icmp_out_count(sock_net(sk), user_icmph.type);
838 		return len;
839 	}
840 	return err;
841 
842 do_confirm:
843 	if (msg->msg_flags & MSG_PROBE)
844 		dst_confirm_neigh(&rt->dst, &fl4.daddr);
845 	if (!(msg->msg_flags & MSG_PROBE) || len)
846 		goto back_from_confirm;
847 	err = 0;
848 	goto out;
849 }
850 
851 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int flags,
852 		 int *addr_len)
853 {
854 	struct inet_sock *isk = inet_sk(sk);
855 	int family = sk->sk_family;
856 	struct sk_buff *skb;
857 	int copied, err;
858 
859 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
860 
861 	err = -EOPNOTSUPP;
862 	if (flags & MSG_OOB)
863 		goto out;
864 
865 	if (flags & MSG_ERRQUEUE)
866 		return inet_recv_error(sk, msg, len, addr_len);
867 
868 	skb = skb_recv_datagram(sk, flags, &err);
869 	if (!skb)
870 		goto out;
871 
872 	copied = skb->len;
873 	if (copied > len) {
874 		msg->msg_flags |= MSG_TRUNC;
875 		copied = len;
876 	}
877 
878 	/* Don't bother checking the checksum */
879 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
880 	if (err)
881 		goto done;
882 
883 	sock_recv_timestamp(msg, sk, skb);
884 
885 	/* Copy the address and add cmsg data. */
886 	if (family == AF_INET) {
887 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
888 
889 		if (sin) {
890 			sin->sin_family = AF_INET;
891 			sin->sin_port = 0 /* skb->h.uh->source */;
892 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
893 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
894 			*addr_len = sizeof(*sin);
895 		}
896 
897 		if (inet_cmsg_flags(isk))
898 			ip_cmsg_recv(msg, skb);
899 
900 #if IS_ENABLED(CONFIG_IPV6)
901 	} else if (family == AF_INET6) {
902 		struct ipv6_pinfo *np = inet6_sk(sk);
903 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
904 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
905 
906 		if (sin6) {
907 			sin6->sin6_family = AF_INET6;
908 			sin6->sin6_port = 0;
909 			sin6->sin6_addr = ip6->saddr;
910 			sin6->sin6_flowinfo = 0;
911 			if (np->sndflow)
912 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
913 			sin6->sin6_scope_id =
914 				ipv6_iface_scope_id(&sin6->sin6_addr,
915 						    inet6_iif(skb));
916 			*addr_len = sizeof(*sin6);
917 		}
918 
919 		if (inet6_sk(sk)->rxopt.all)
920 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
921 		if (skb->protocol == htons(ETH_P_IPV6) &&
922 		    inet6_sk(sk)->rxopt.all)
923 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
924 		else if (skb->protocol == htons(ETH_P_IP) &&
925 			 inet_cmsg_flags(isk))
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 static enum skb_drop_reason __ping_queue_rcv_skb(struct sock *sk,
943 						 struct sk_buff *skb)
944 {
945 	enum skb_drop_reason reason;
946 
947 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
948 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
949 	if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) {
950 		kfree_skb_reason(skb, reason);
951 		pr_debug("ping_queue_rcv_skb -> failed\n");
952 		return reason;
953 	}
954 	return SKB_NOT_DROPPED_YET;
955 }
956 
957 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
958 {
959 	return __ping_queue_rcv_skb(sk, skb) ? -1 : 0;
960 }
961 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
962 
963 
964 /*
965  *	All we need to do is get the socket.
966  */
967 
968 enum skb_drop_reason ping_rcv(struct sk_buff *skb)
969 {
970 	enum skb_drop_reason reason = SKB_DROP_REASON_NO_SOCKET;
971 	struct sock *sk;
972 	struct net *net = dev_net(skb->dev);
973 	struct icmphdr *icmph = icmp_hdr(skb);
974 
975 	/* We assume the packet has already been checked by icmp_rcv */
976 
977 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
978 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
979 
980 	/* Push ICMP header back */
981 	skb_push(skb, skb->data - (u8 *)icmph);
982 
983 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
984 	if (sk) {
985 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
986 
987 		pr_debug("rcv on socket %p\n", sk);
988 		if (skb2)
989 			reason = __ping_queue_rcv_skb(sk, skb2);
990 		else
991 			reason = SKB_DROP_REASON_NOMEM;
992 	}
993 
994 	if (reason)
995 		pr_debug("no socket, dropping\n");
996 
997 	return reason;
998 }
999 EXPORT_SYMBOL_GPL(ping_rcv);
1000 
1001 struct proto ping_prot = {
1002 	.name =		"PING",
1003 	.owner =	THIS_MODULE,
1004 	.init =		ping_init_sock,
1005 	.close =	ping_close,
1006 	.pre_connect =	ping_pre_connect,
1007 	.connect =	ip4_datagram_connect,
1008 	.disconnect =	__udp_disconnect,
1009 	.setsockopt =	ip_setsockopt,
1010 	.getsockopt =	ip_getsockopt,
1011 	.sendmsg =	ping_v4_sendmsg,
1012 	.recvmsg =	ping_recvmsg,
1013 	.bind =		ping_bind,
1014 	.backlog_rcv =	ping_queue_rcv_skb,
1015 	.release_cb =	ip4_datagram_release_cb,
1016 	.hash =		ping_hash,
1017 	.unhash =	ping_unhash,
1018 	.get_port =	ping_get_port,
1019 	.put_port =	ping_unhash,
1020 	.obj_size =	sizeof(struct inet_sock),
1021 };
1022 EXPORT_SYMBOL(ping_prot);
1023 
1024 #ifdef CONFIG_PROC_FS
1025 
1026 static struct sock *ping_get_first(struct seq_file *seq, int start)
1027 {
1028 	struct sock *sk;
1029 	struct ping_iter_state *state = seq->private;
1030 	struct net *net = seq_file_net(seq);
1031 
1032 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1033 	     ++state->bucket) {
1034 		struct hlist_head *hslot;
1035 
1036 		hslot = &ping_table.hash[state->bucket];
1037 
1038 		if (hlist_empty(hslot))
1039 			continue;
1040 
1041 		sk_for_each(sk, hslot) {
1042 			if (net_eq(sock_net(sk), net) &&
1043 			    sk->sk_family == state->family)
1044 				goto found;
1045 		}
1046 	}
1047 	sk = NULL;
1048 found:
1049 	return sk;
1050 }
1051 
1052 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1053 {
1054 	struct ping_iter_state *state = seq->private;
1055 	struct net *net = seq_file_net(seq);
1056 
1057 	do {
1058 		sk = sk_next(sk);
1059 	} while (sk && (!net_eq(sock_net(sk), net)));
1060 
1061 	if (!sk)
1062 		return ping_get_first(seq, state->bucket + 1);
1063 	return sk;
1064 }
1065 
1066 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1067 {
1068 	struct sock *sk = ping_get_first(seq, 0);
1069 
1070 	if (sk)
1071 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1072 			--pos;
1073 	return pos ? NULL : sk;
1074 }
1075 
1076 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1077 	__acquires(ping_table.lock)
1078 {
1079 	struct ping_iter_state *state = seq->private;
1080 	state->bucket = 0;
1081 	state->family = family;
1082 
1083 	spin_lock(&ping_table.lock);
1084 
1085 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1086 }
1087 EXPORT_SYMBOL_GPL(ping_seq_start);
1088 
1089 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1090 {
1091 	return ping_seq_start(seq, pos, AF_INET);
1092 }
1093 
1094 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1095 {
1096 	struct sock *sk;
1097 
1098 	if (v == SEQ_START_TOKEN)
1099 		sk = ping_get_idx(seq, 0);
1100 	else
1101 		sk = ping_get_next(seq, v);
1102 
1103 	++*pos;
1104 	return sk;
1105 }
1106 EXPORT_SYMBOL_GPL(ping_seq_next);
1107 
1108 void ping_seq_stop(struct seq_file *seq, void *v)
1109 	__releases(ping_table.lock)
1110 {
1111 	spin_unlock(&ping_table.lock);
1112 }
1113 EXPORT_SYMBOL_GPL(ping_seq_stop);
1114 
1115 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1116 		int bucket)
1117 {
1118 	struct inet_sock *inet = inet_sk(sp);
1119 	__be32 dest = inet->inet_daddr;
1120 	__be32 src = inet->inet_rcv_saddr;
1121 	__u16 destp = ntohs(inet->inet_dport);
1122 	__u16 srcp = ntohs(inet->inet_sport);
1123 
1124 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1125 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1126 		bucket, src, srcp, dest, destp, sp->sk_state,
1127 		sk_wmem_alloc_get(sp),
1128 		sk_rmem_alloc_get(sp),
1129 		0, 0L, 0,
1130 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1131 		0, sock_i_ino(sp),
1132 		refcount_read(&sp->sk_refcnt), sp,
1133 		atomic_read(&sp->sk_drops));
1134 }
1135 
1136 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1137 {
1138 	seq_setwidth(seq, 127);
1139 	if (v == SEQ_START_TOKEN)
1140 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1141 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1142 			   "inode ref pointer drops");
1143 	else {
1144 		struct ping_iter_state *state = seq->private;
1145 
1146 		ping_v4_format_sock(v, seq, state->bucket);
1147 	}
1148 	seq_pad(seq, '\n');
1149 	return 0;
1150 }
1151 
1152 static const struct seq_operations ping_v4_seq_ops = {
1153 	.start		= ping_v4_seq_start,
1154 	.show		= ping_v4_seq_show,
1155 	.next		= ping_seq_next,
1156 	.stop		= ping_seq_stop,
1157 };
1158 
1159 static int __net_init ping_v4_proc_init_net(struct net *net)
1160 {
1161 	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1162 			sizeof(struct ping_iter_state)))
1163 		return -ENOMEM;
1164 	return 0;
1165 }
1166 
1167 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1168 {
1169 	remove_proc_entry("icmp", net->proc_net);
1170 }
1171 
1172 static struct pernet_operations ping_v4_net_ops = {
1173 	.init = ping_v4_proc_init_net,
1174 	.exit = ping_v4_proc_exit_net,
1175 };
1176 
1177 int __init ping_proc_init(void)
1178 {
1179 	return register_pernet_subsys(&ping_v4_net_ops);
1180 }
1181 
1182 void ping_proc_exit(void)
1183 {
1184 	unregister_pernet_subsys(&ping_v4_net_ops);
1185 }
1186 
1187 #endif
1188 
1189 void __init ping_init(void)
1190 {
1191 	int i;
1192 
1193 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1194 		INIT_HLIST_HEAD(&ping_table.hash[i]);
1195 	spin_lock_init(&ping_table.lock);
1196 }
1197