xref: /openbmc/linux/net/ipv4/ping.c (revision dff03381)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		"Ping" sockets
8  *
9  * Based on ipv4/udp.c code.
10  *
11  * Authors:	Vasiliy Kulikov / Openwall (for Linux 2.6),
12  *		Pavel Kankovsky (for Linux 2.4.32)
13  *
14  * Pavel gave all rights to bugs to Vasiliy,
15  * none of the bugs are Pavel's now.
16  */
17 
18 #include <linux/uaccess.h>
19 #include <linux/types.h>
20 #include <linux/fcntl.h>
21 #include <linux/socket.h>
22 #include <linux/sockios.h>
23 #include <linux/in.h>
24 #include <linux/errno.h>
25 #include <linux/timer.h>
26 #include <linux/mm.h>
27 #include <linux/inet.h>
28 #include <linux/netdevice.h>
29 #include <net/snmp.h>
30 #include <net/ip.h>
31 #include <net/icmp.h>
32 #include <net/protocol.h>
33 #include <linux/skbuff.h>
34 #include <linux/proc_fs.h>
35 #include <linux/export.h>
36 #include <net/sock.h>
37 #include <net/ping.h>
38 #include <net/udp.h>
39 #include <net/route.h>
40 #include <net/inet_common.h>
41 #include <net/checksum.h>
42 
43 #if IS_ENABLED(CONFIG_IPV6)
44 #include <linux/in6.h>
45 #include <linux/icmpv6.h>
46 #include <net/addrconf.h>
47 #include <net/ipv6.h>
48 #include <net/transp_v6.h>
49 #endif
50 
51 struct ping_table {
52 	struct hlist_nulls_head	hash[PING_HTABLE_SIZE];
53 	spinlock_t		lock;
54 };
55 
56 static struct ping_table ping_table;
57 struct pingv6_ops pingv6_ops;
58 EXPORT_SYMBOL_GPL(pingv6_ops);
59 
60 static u16 ping_port_rover;
61 
62 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
63 {
64 	u32 res = (num + net_hash_mix(net)) & mask;
65 
66 	pr_debug("hash(%u) = %u\n", num, res);
67 	return res;
68 }
69 EXPORT_SYMBOL_GPL(ping_hash);
70 
71 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
72 					     struct net *net, unsigned int num)
73 {
74 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
75 }
76 
77 int ping_get_port(struct sock *sk, unsigned short ident)
78 {
79 	struct hlist_nulls_node *node;
80 	struct hlist_nulls_head *hlist;
81 	struct inet_sock *isk, *isk2;
82 	struct sock *sk2 = NULL;
83 
84 	isk = inet_sk(sk);
85 	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 			ping_portaddr_for_each_entry(sk2, node, hlist) {
96 				isk2 = inet_sk(sk2);
97 
98 				if (isk2->inet_num == result)
99 					goto next_port;
100 			}
101 
102 			/* found */
103 			ping_port_rover = ident = result;
104 			break;
105 next_port:
106 			;
107 		}
108 		if (i >= (1L << 16))
109 			goto fail;
110 	} else {
111 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
112 		ping_portaddr_for_each_entry(sk2, node, hlist) {
113 			isk2 = inet_sk(sk2);
114 
115 			/* BUG? Why is this reuse and not reuseaddr? ping.c
116 			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
117 			 * that other ping processes can steal its packets.
118 			 */
119 			if ((isk2->inet_num == ident) &&
120 			    (sk2 != sk) &&
121 			    (!sk2->sk_reuse || !sk->sk_reuse))
122 				goto fail;
123 		}
124 	}
125 
126 	pr_debug("found port/ident = %d\n", ident);
127 	isk->inet_num = ident;
128 	if (sk_unhashed(sk)) {
129 		pr_debug("was not hashed\n");
130 		sock_hold(sk);
131 		sock_set_flag(sk, SOCK_RCU_FREE);
132 		hlist_nulls_add_head_rcu(&sk->sk_nulls_node, hlist);
133 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
134 	}
135 	spin_unlock(&ping_table.lock);
136 	return 0;
137 
138 fail:
139 	spin_unlock(&ping_table.lock);
140 	return 1;
141 }
142 EXPORT_SYMBOL_GPL(ping_get_port);
143 
144 int ping_hash(struct sock *sk)
145 {
146 	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
147 	BUG(); /* "Please do not press this button again." */
148 
149 	return 0;
150 }
151 
152 void ping_unhash(struct sock *sk)
153 {
154 	struct inet_sock *isk = inet_sk(sk);
155 
156 	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
157 	spin_lock(&ping_table.lock);
158 	if (sk_hashed(sk)) {
159 		hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
160 		sock_put(sk);
161 		isk->inet_num = 0;
162 		isk->inet_sport = 0;
163 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
164 	}
165 	spin_unlock(&ping_table.lock);
166 }
167 EXPORT_SYMBOL_GPL(ping_unhash);
168 
169 /* Called under rcu_read_lock() */
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, sdif;
177 
178 	if (skb->protocol == htons(ETH_P_IP)) {
179 		dif = inet_iif(skb);
180 		sdif = inet_sdif(skb);
181 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
182 			 (int)ident, &ip_hdr(skb)->daddr, dif);
183 #if IS_ENABLED(CONFIG_IPV6)
184 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
185 		dif = inet6_iif(skb);
186 		sdif = inet6_sdif(skb);
187 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
188 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
189 #endif
190 	} else {
191 		return NULL;
192 	}
193 
194 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
195 		isk = inet_sk(sk);
196 
197 		pr_debug("iterate\n");
198 		if (isk->inet_num != ident)
199 			continue;
200 
201 		if (skb->protocol == htons(ETH_P_IP) &&
202 		    sk->sk_family == AF_INET) {
203 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
204 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
205 				 sk->sk_bound_dev_if);
206 
207 			if (isk->inet_rcv_saddr &&
208 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
209 				continue;
210 #if IS_ENABLED(CONFIG_IPV6)
211 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
212 			   sk->sk_family == AF_INET6) {
213 
214 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
215 				 (int) isk->inet_num,
216 				 &sk->sk_v6_rcv_saddr,
217 				 sk->sk_bound_dev_if);
218 
219 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
220 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
221 					     &ipv6_hdr(skb)->daddr))
222 				continue;
223 #endif
224 		} else {
225 			continue;
226 		}
227 
228 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
229 		    sk->sk_bound_dev_if != sdif)
230 			continue;
231 
232 		goto exit;
233 	}
234 
235 	sk = NULL;
236 exit:
237 
238 	return sk;
239 }
240 
241 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
242 					  kgid_t *high)
243 {
244 	kgid_t *data = net->ipv4.ping_group_range.range;
245 	unsigned int seq;
246 
247 	do {
248 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
249 
250 		*low = data[0];
251 		*high = data[1];
252 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
253 }
254 
255 
256 int ping_init_sock(struct sock *sk)
257 {
258 	struct net *net = sock_net(sk);
259 	kgid_t group = current_egid();
260 	struct group_info *group_info;
261 	int i;
262 	kgid_t low, high;
263 	int ret = 0;
264 
265 	if (sk->sk_family == AF_INET6)
266 		sk->sk_ipv6only = 1;
267 
268 	inet_get_ping_group_range_net(net, &low, &high);
269 	if (gid_lte(low, group) && gid_lte(group, high))
270 		return 0;
271 
272 	group_info = get_current_groups();
273 	for (i = 0; i < group_info->ngroups; i++) {
274 		kgid_t gid = group_info->gid[i];
275 
276 		if (gid_lte(low, gid) && gid_lte(gid, high))
277 			goto out_release_group;
278 	}
279 
280 	ret = -EACCES;
281 
282 out_release_group:
283 	put_group_info(group_info);
284 	return ret;
285 }
286 EXPORT_SYMBOL_GPL(ping_init_sock);
287 
288 void ping_close(struct sock *sk, long timeout)
289 {
290 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
291 		 inet_sk(sk), inet_sk(sk)->inet_num);
292 	pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
293 
294 	sk_common_release(sk);
295 }
296 EXPORT_SYMBOL_GPL(ping_close);
297 
298 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
299 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
300 				struct sockaddr *uaddr, int addr_len)
301 {
302 	struct net *net = sock_net(sk);
303 	if (sk->sk_family == AF_INET) {
304 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
305 		u32 tb_id = RT_TABLE_LOCAL;
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 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
320 			return 0;
321 
322 		tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
323 		chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
324 
325 		if (chk_addr_ret == RTN_MULTICAST ||
326 		    chk_addr_ret == RTN_BROADCAST ||
327 		    (chk_addr_ret != RTN_LOCAL &&
328 		     !inet_can_nonlocal_bind(net, isk)))
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 
362 		if (!dev && sk->sk_bound_dev_if) {
363 			dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
364 			if (!dev) {
365 				rcu_read_unlock();
366 				return -ENODEV;
367 			}
368 		}
369 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
370 						    scoped);
371 		rcu_read_unlock();
372 
373 		if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
374 		      addr_type == IPV6_ADDR_ANY))
375 			return -EADDRNOTAVAIL;
376 
377 		if (scoped)
378 			sk->sk_bound_dev_if = addr->sin6_scope_id;
379 #endif
380 	} else {
381 		return -EAFNOSUPPORT;
382 	}
383 	return 0;
384 }
385 
386 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
387 {
388 	if (saddr->sa_family == AF_INET) {
389 		struct inet_sock *isk = inet_sk(sk);
390 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
391 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
392 #if IS_ENABLED(CONFIG_IPV6)
393 	} else if (saddr->sa_family == AF_INET6) {
394 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
395 		struct ipv6_pinfo *np = inet6_sk(sk);
396 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
397 #endif
398 	}
399 }
400 
401 /*
402  * We need our own bind because there are no privileged id's == local ports.
403  * Moreover, we don't allow binding to multi- and broadcast addresses.
404  */
405 
406 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
407 {
408 	struct inet_sock *isk = inet_sk(sk);
409 	unsigned short snum;
410 	int err;
411 	int dif = sk->sk_bound_dev_if;
412 
413 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
414 	if (err)
415 		return err;
416 
417 	lock_sock(sk);
418 
419 	err = -EINVAL;
420 	if (isk->inet_num != 0)
421 		goto out;
422 
423 	err = -EADDRINUSE;
424 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
425 	if (ping_get_port(sk, snum) != 0) {
426 		/* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
427 		sk->sk_bound_dev_if = dif;
428 		goto out;
429 	}
430 	ping_set_saddr(sk, uaddr);
431 
432 	pr_debug("after bind(): num = %hu, dif = %d\n",
433 		 isk->inet_num,
434 		 sk->sk_bound_dev_if);
435 
436 	err = 0;
437 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
438 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
439 #if IS_ENABLED(CONFIG_IPV6)
440 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
441 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
442 #endif
443 
444 	if (snum)
445 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
446 	isk->inet_sport = htons(isk->inet_num);
447 	isk->inet_daddr = 0;
448 	isk->inet_dport = 0;
449 
450 #if IS_ENABLED(CONFIG_IPV6)
451 	if (sk->sk_family == AF_INET6)
452 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
453 #endif
454 
455 	sk_dst_reset(sk);
456 out:
457 	release_sock(sk);
458 	pr_debug("ping_v4_bind -> %d\n", err);
459 	return err;
460 }
461 EXPORT_SYMBOL_GPL(ping_bind);
462 
463 /*
464  * Is this a supported type of ICMP message?
465  */
466 
467 static inline int ping_supported(int family, int type, int code)
468 {
469 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
470 	       (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
471 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
472 	       (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
473 }
474 
475 /*
476  * This routine is called by the ICMP module when it gets some
477  * sort of error condition.
478  */
479 
480 void ping_err(struct sk_buff *skb, int offset, u32 info)
481 {
482 	int family;
483 	struct icmphdr *icmph;
484 	struct inet_sock *inet_sock;
485 	int type;
486 	int code;
487 	struct net *net = dev_net(skb->dev);
488 	struct sock *sk;
489 	int harderr;
490 	int err;
491 
492 	if (skb->protocol == htons(ETH_P_IP)) {
493 		family = AF_INET;
494 		type = icmp_hdr(skb)->type;
495 		code = icmp_hdr(skb)->code;
496 		icmph = (struct icmphdr *)(skb->data + offset);
497 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
498 		family = AF_INET6;
499 		type = icmp6_hdr(skb)->icmp6_type;
500 		code = icmp6_hdr(skb)->icmp6_code;
501 		icmph = (struct icmphdr *) (skb->data + offset);
502 	} else {
503 		BUG();
504 	}
505 
506 	/* We assume the packet has already been checked by icmp_unreach */
507 
508 	if (!ping_supported(family, icmph->type, icmph->code))
509 		return;
510 
511 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
512 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
513 		 ntohs(icmph->un.echo.sequence));
514 
515 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
516 	if (!sk) {
517 		pr_debug("no socket, dropping\n");
518 		return;	/* No socket for error */
519 	}
520 	pr_debug("err on socket %p\n", sk);
521 
522 	err = 0;
523 	harderr = 0;
524 	inet_sock = inet_sk(sk);
525 
526 	if (skb->protocol == htons(ETH_P_IP)) {
527 		switch (type) {
528 		default:
529 		case ICMP_TIME_EXCEEDED:
530 			err = EHOSTUNREACH;
531 			break;
532 		case ICMP_SOURCE_QUENCH:
533 			/* This is not a real error but ping wants to see it.
534 			 * Report it with some fake errno.
535 			 */
536 			err = EREMOTEIO;
537 			break;
538 		case ICMP_PARAMETERPROB:
539 			err = EPROTO;
540 			harderr = 1;
541 			break;
542 		case ICMP_DEST_UNREACH:
543 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
544 				ipv4_sk_update_pmtu(skb, sk, info);
545 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
546 					err = EMSGSIZE;
547 					harderr = 1;
548 					break;
549 				}
550 				goto out;
551 			}
552 			err = EHOSTUNREACH;
553 			if (code <= NR_ICMP_UNREACH) {
554 				harderr = icmp_err_convert[code].fatal;
555 				err = icmp_err_convert[code].errno;
556 			}
557 			break;
558 		case ICMP_REDIRECT:
559 			/* See ICMP_SOURCE_QUENCH */
560 			ipv4_sk_redirect(skb, sk);
561 			err = EREMOTEIO;
562 			break;
563 		}
564 #if IS_ENABLED(CONFIG_IPV6)
565 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
566 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
567 #endif
568 	}
569 
570 	/*
571 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
572 	 *	4.1.3.3.
573 	 */
574 	if ((family == AF_INET && !inet_sock->recverr) ||
575 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
576 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
577 			goto out;
578 	} else {
579 		if (family == AF_INET) {
580 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
581 				      info, (u8 *)icmph);
582 #if IS_ENABLED(CONFIG_IPV6)
583 		} else if (family == AF_INET6) {
584 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
585 						   info, (u8 *)icmph);
586 #endif
587 		}
588 	}
589 	sk->sk_err = err;
590 	sk_error_report(sk);
591 out:
592 	return;
593 }
594 EXPORT_SYMBOL_GPL(ping_err);
595 
596 /*
597  *	Copy and checksum an ICMP Echo packet from user space into a buffer
598  *	starting from the payload.
599  */
600 
601 int ping_getfrag(void *from, char *to,
602 		 int offset, int fraglen, int odd, struct sk_buff *skb)
603 {
604 	struct pingfakehdr *pfh = from;
605 
606 	if (offset == 0) {
607 		fraglen -= sizeof(struct icmphdr);
608 		if (fraglen < 0)
609 			BUG();
610 		if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
611 			    fraglen, &pfh->wcheck,
612 			    &pfh->msg->msg_iter))
613 			return -EFAULT;
614 	} else if (offset < sizeof(struct icmphdr)) {
615 			BUG();
616 	} else {
617 		if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
618 					    &pfh->msg->msg_iter))
619 			return -EFAULT;
620 	}
621 
622 #if IS_ENABLED(CONFIG_IPV6)
623 	/* For IPv6, checksum each skb as we go along, as expected by
624 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
625 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
626 	 */
627 	if (pfh->family == AF_INET6) {
628 		skb->csum = pfh->wcheck;
629 		skb->ip_summed = CHECKSUM_NONE;
630 		pfh->wcheck = 0;
631 	}
632 #endif
633 
634 	return 0;
635 }
636 EXPORT_SYMBOL_GPL(ping_getfrag);
637 
638 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
639 				       struct flowi4 *fl4)
640 {
641 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
642 
643 	if (!skb)
644 		return 0;
645 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
646 		sizeof(struct icmphdr), pfh->wcheck);
647 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
648 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
649 	skb->ip_summed = CHECKSUM_NONE;
650 	return ip_push_pending_frames(sk, fl4);
651 }
652 
653 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
654 			void *user_icmph, size_t icmph_len)
655 {
656 	u8 type, code;
657 
658 	if (len > 0xFFFF)
659 		return -EMSGSIZE;
660 
661 	/* Must have at least a full ICMP header. */
662 	if (len < icmph_len)
663 		return -EINVAL;
664 
665 	/*
666 	 *	Check the flags.
667 	 */
668 
669 	/* Mirror BSD error message compatibility */
670 	if (msg->msg_flags & MSG_OOB)
671 		return -EOPNOTSUPP;
672 
673 	/*
674 	 *	Fetch the ICMP header provided by the userland.
675 	 *	iovec is modified! The ICMP header is consumed.
676 	 */
677 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
678 		return -EFAULT;
679 
680 	if (family == AF_INET) {
681 		type = ((struct icmphdr *) user_icmph)->type;
682 		code = ((struct icmphdr *) user_icmph)->code;
683 #if IS_ENABLED(CONFIG_IPV6)
684 	} else if (family == AF_INET6) {
685 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
686 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
687 #endif
688 	} else {
689 		BUG();
690 	}
691 
692 	if (!ping_supported(family, type, code))
693 		return -EINVAL;
694 
695 	return 0;
696 }
697 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
698 
699 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
700 {
701 	struct net *net = sock_net(sk);
702 	struct flowi4 fl4;
703 	struct inet_sock *inet = inet_sk(sk);
704 	struct ipcm_cookie ipc;
705 	struct icmphdr user_icmph;
706 	struct pingfakehdr pfh;
707 	struct rtable *rt = NULL;
708 	struct ip_options_data opt_copy;
709 	int free = 0;
710 	__be32 saddr, daddr, faddr;
711 	u8  tos;
712 	int err;
713 
714 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
715 
716 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
717 				  sizeof(user_icmph));
718 	if (err)
719 		return err;
720 
721 	/*
722 	 *	Get and verify the address.
723 	 */
724 
725 	if (msg->msg_name) {
726 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
727 		if (msg->msg_namelen < sizeof(*usin))
728 			return -EINVAL;
729 		if (usin->sin_family != AF_INET)
730 			return -EAFNOSUPPORT;
731 		daddr = usin->sin_addr.s_addr;
732 		/* no remote port */
733 	} else {
734 		if (sk->sk_state != TCP_ESTABLISHED)
735 			return -EDESTADDRREQ;
736 		daddr = inet->inet_daddr;
737 		/* no remote port */
738 	}
739 
740 	ipcm_init_sk(&ipc, inet);
741 
742 	if (msg->msg_controllen) {
743 		err = ip_cmsg_send(sk, msg, &ipc, false);
744 		if (unlikely(err)) {
745 			kfree(ipc.opt);
746 			return err;
747 		}
748 		if (ipc.opt)
749 			free = 1;
750 	}
751 	if (!ipc.opt) {
752 		struct ip_options_rcu *inet_opt;
753 
754 		rcu_read_lock();
755 		inet_opt = rcu_dereference(inet->inet_opt);
756 		if (inet_opt) {
757 			memcpy(&opt_copy, inet_opt,
758 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
759 			ipc.opt = &opt_copy.opt;
760 		}
761 		rcu_read_unlock();
762 	}
763 
764 	saddr = ipc.addr;
765 	ipc.addr = faddr = daddr;
766 
767 	if (ipc.opt && ipc.opt->opt.srr) {
768 		if (!daddr) {
769 			err = -EINVAL;
770 			goto out_free;
771 		}
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 || netif_index_is_l3_master(sock_net(sk), 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, ipc.sockc.mark, tos,
790 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
791 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
792 			   sk->sk_uid);
793 
794 	fl4.fl4_icmp_type = user_icmph.type;
795 	fl4.fl4_icmp_code = user_icmph.code;
796 
797 	security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
798 	rt = ip_route_output_flow(net, &fl4, sk);
799 	if (IS_ERR(rt)) {
800 		err = PTR_ERR(rt);
801 		rt = NULL;
802 		if (err == -ENETUNREACH)
803 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
804 		goto out;
805 	}
806 
807 	err = -EACCES;
808 	if ((rt->rt_flags & RTCF_BROADCAST) &&
809 	    !sock_flag(sk, SOCK_BROADCAST))
810 		goto out;
811 
812 	if (msg->msg_flags & MSG_CONFIRM)
813 		goto do_confirm;
814 back_from_confirm:
815 
816 	if (!ipc.addr)
817 		ipc.addr = fl4.daddr;
818 
819 	lock_sock(sk);
820 
821 	pfh.icmph.type = user_icmph.type; /* already checked */
822 	pfh.icmph.code = user_icmph.code; /* ditto */
823 	pfh.icmph.checksum = 0;
824 	pfh.icmph.un.echo.id = inet->inet_sport;
825 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
826 	pfh.msg = msg;
827 	pfh.wcheck = 0;
828 	pfh.family = AF_INET;
829 
830 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
831 			0, &ipc, &rt, msg->msg_flags);
832 	if (err)
833 		ip_flush_pending_frames(sk);
834 	else
835 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
836 	release_sock(sk);
837 
838 out:
839 	ip_rt_put(rt);
840 out_free:
841 	if (free)
842 		kfree(ipc.opt);
843 	if (!err) {
844 		icmp_out_count(sock_net(sk), user_icmph.type);
845 		return len;
846 	}
847 	return err;
848 
849 do_confirm:
850 	if (msg->msg_flags & MSG_PROBE)
851 		dst_confirm_neigh(&rt->dst, &fl4.daddr);
852 	if (!(msg->msg_flags & MSG_PROBE) || len)
853 		goto back_from_confirm;
854 	err = 0;
855 	goto out;
856 }
857 
858 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int flags,
859 		 int *addr_len)
860 {
861 	struct inet_sock *isk = inet_sk(sk);
862 	int family = sk->sk_family;
863 	struct sk_buff *skb;
864 	int copied, err;
865 
866 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
867 
868 	err = -EOPNOTSUPP;
869 	if (flags & MSG_OOB)
870 		goto out;
871 
872 	if (flags & MSG_ERRQUEUE)
873 		return inet_recv_error(sk, msg, len, addr_len);
874 
875 	skb = skb_recv_datagram(sk, flags, &err);
876 	if (!skb)
877 		goto out;
878 
879 	copied = skb->len;
880 	if (copied > len) {
881 		msg->msg_flags |= MSG_TRUNC;
882 		copied = len;
883 	}
884 
885 	/* Don't bother checking the checksum */
886 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
887 	if (err)
888 		goto done;
889 
890 	sock_recv_timestamp(msg, sk, skb);
891 
892 	/* Copy the address and add cmsg data. */
893 	if (family == AF_INET) {
894 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
895 
896 		if (sin) {
897 			sin->sin_family = AF_INET;
898 			sin->sin_port = 0 /* skb->h.uh->source */;
899 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
900 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
901 			*addr_len = sizeof(*sin);
902 		}
903 
904 		if (isk->cmsg_flags)
905 			ip_cmsg_recv(msg, skb);
906 
907 #if IS_ENABLED(CONFIG_IPV6)
908 	} else if (family == AF_INET6) {
909 		struct ipv6_pinfo *np = inet6_sk(sk);
910 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
911 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
912 
913 		if (sin6) {
914 			sin6->sin6_family = AF_INET6;
915 			sin6->sin6_port = 0;
916 			sin6->sin6_addr = ip6->saddr;
917 			sin6->sin6_flowinfo = 0;
918 			if (np->sndflow)
919 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
920 			sin6->sin6_scope_id =
921 				ipv6_iface_scope_id(&sin6->sin6_addr,
922 						    inet6_iif(skb));
923 			*addr_len = sizeof(*sin6);
924 		}
925 
926 		if (inet6_sk(sk)->rxopt.all)
927 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
928 		if (skb->protocol == htons(ETH_P_IPV6) &&
929 		    inet6_sk(sk)->rxopt.all)
930 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
931 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
932 			ip_cmsg_recv(msg, skb);
933 #endif
934 	} else {
935 		BUG();
936 	}
937 
938 	err = copied;
939 
940 done:
941 	skb_free_datagram(sk, skb);
942 out:
943 	pr_debug("ping_recvmsg -> %d\n", err);
944 	return err;
945 }
946 EXPORT_SYMBOL_GPL(ping_recvmsg);
947 
948 static enum skb_drop_reason __ping_queue_rcv_skb(struct sock *sk,
949 						 struct sk_buff *skb)
950 {
951 	enum skb_drop_reason reason;
952 
953 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
954 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
955 	if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) {
956 		kfree_skb_reason(skb, reason);
957 		pr_debug("ping_queue_rcv_skb -> failed\n");
958 		return reason;
959 	}
960 	return SKB_NOT_DROPPED_YET;
961 }
962 
963 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
964 {
965 	return __ping_queue_rcv_skb(sk, skb) ? -1 : 0;
966 }
967 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
968 
969 
970 /*
971  *	All we need to do is get the socket.
972  */
973 
974 enum skb_drop_reason ping_rcv(struct sk_buff *skb)
975 {
976 	enum skb_drop_reason reason = SKB_DROP_REASON_NO_SOCKET;
977 	struct sock *sk;
978 	struct net *net = dev_net(skb->dev);
979 	struct icmphdr *icmph = icmp_hdr(skb);
980 
981 	/* We assume the packet has already been checked by icmp_rcv */
982 
983 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
984 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
985 
986 	/* Push ICMP header back */
987 	skb_push(skb, skb->data - (u8 *)icmph);
988 
989 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
990 	if (sk) {
991 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
992 
993 		pr_debug("rcv on socket %p\n", sk);
994 		if (skb2)
995 			reason = __ping_queue_rcv_skb(sk, skb2);
996 		else
997 			reason = SKB_DROP_REASON_NOMEM;
998 	}
999 
1000 	if (reason)
1001 		pr_debug("no socket, dropping\n");
1002 
1003 	return reason;
1004 }
1005 EXPORT_SYMBOL_GPL(ping_rcv);
1006 
1007 struct proto ping_prot = {
1008 	.name =		"PING",
1009 	.owner =	THIS_MODULE,
1010 	.init =		ping_init_sock,
1011 	.close =	ping_close,
1012 	.connect =	ip4_datagram_connect,
1013 	.disconnect =	__udp_disconnect,
1014 	.setsockopt =	ip_setsockopt,
1015 	.getsockopt =	ip_getsockopt,
1016 	.sendmsg =	ping_v4_sendmsg,
1017 	.recvmsg =	ping_recvmsg,
1018 	.bind =		ping_bind,
1019 	.backlog_rcv =	ping_queue_rcv_skb,
1020 	.release_cb =	ip4_datagram_release_cb,
1021 	.hash =		ping_hash,
1022 	.unhash =	ping_unhash,
1023 	.get_port =	ping_get_port,
1024 	.put_port =	ping_unhash,
1025 	.obj_size =	sizeof(struct inet_sock),
1026 };
1027 EXPORT_SYMBOL(ping_prot);
1028 
1029 #ifdef CONFIG_PROC_FS
1030 
1031 static struct sock *ping_get_first(struct seq_file *seq, int start)
1032 {
1033 	struct sock *sk;
1034 	struct ping_iter_state *state = seq->private;
1035 	struct net *net = seq_file_net(seq);
1036 
1037 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1038 	     ++state->bucket) {
1039 		struct hlist_nulls_node *node;
1040 		struct hlist_nulls_head *hslot;
1041 
1042 		hslot = &ping_table.hash[state->bucket];
1043 
1044 		if (hlist_nulls_empty(hslot))
1045 			continue;
1046 
1047 		sk_nulls_for_each(sk, node, hslot) {
1048 			if (net_eq(sock_net(sk), net) &&
1049 			    sk->sk_family == state->family)
1050 				goto found;
1051 		}
1052 	}
1053 	sk = NULL;
1054 found:
1055 	return sk;
1056 }
1057 
1058 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1059 {
1060 	struct ping_iter_state *state = seq->private;
1061 	struct net *net = seq_file_net(seq);
1062 
1063 	do {
1064 		sk = sk_nulls_next(sk);
1065 	} while (sk && (!net_eq(sock_net(sk), net)));
1066 
1067 	if (!sk)
1068 		return ping_get_first(seq, state->bucket + 1);
1069 	return sk;
1070 }
1071 
1072 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1073 {
1074 	struct sock *sk = ping_get_first(seq, 0);
1075 
1076 	if (sk)
1077 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1078 			--pos;
1079 	return pos ? NULL : sk;
1080 }
1081 
1082 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1083 	__acquires(RCU)
1084 {
1085 	struct ping_iter_state *state = seq->private;
1086 	state->bucket = 0;
1087 	state->family = family;
1088 
1089 	rcu_read_lock();
1090 
1091 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1092 }
1093 EXPORT_SYMBOL_GPL(ping_seq_start);
1094 
1095 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1096 {
1097 	return ping_seq_start(seq, pos, AF_INET);
1098 }
1099 
1100 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1101 {
1102 	struct sock *sk;
1103 
1104 	if (v == SEQ_START_TOKEN)
1105 		sk = ping_get_idx(seq, 0);
1106 	else
1107 		sk = ping_get_next(seq, v);
1108 
1109 	++*pos;
1110 	return sk;
1111 }
1112 EXPORT_SYMBOL_GPL(ping_seq_next);
1113 
1114 void ping_seq_stop(struct seq_file *seq, void *v)
1115 	__releases(RCU)
1116 {
1117 	rcu_read_unlock();
1118 }
1119 EXPORT_SYMBOL_GPL(ping_seq_stop);
1120 
1121 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1122 		int bucket)
1123 {
1124 	struct inet_sock *inet = inet_sk(sp);
1125 	__be32 dest = inet->inet_daddr;
1126 	__be32 src = inet->inet_rcv_saddr;
1127 	__u16 destp = ntohs(inet->inet_dport);
1128 	__u16 srcp = ntohs(inet->inet_sport);
1129 
1130 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1131 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1132 		bucket, src, srcp, dest, destp, sp->sk_state,
1133 		sk_wmem_alloc_get(sp),
1134 		sk_rmem_alloc_get(sp),
1135 		0, 0L, 0,
1136 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1137 		0, sock_i_ino(sp),
1138 		refcount_read(&sp->sk_refcnt), sp,
1139 		atomic_read(&sp->sk_drops));
1140 }
1141 
1142 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1143 {
1144 	seq_setwidth(seq, 127);
1145 	if (v == SEQ_START_TOKEN)
1146 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1147 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1148 			   "inode ref pointer drops");
1149 	else {
1150 		struct ping_iter_state *state = seq->private;
1151 
1152 		ping_v4_format_sock(v, seq, state->bucket);
1153 	}
1154 	seq_pad(seq, '\n');
1155 	return 0;
1156 }
1157 
1158 static const struct seq_operations ping_v4_seq_ops = {
1159 	.start		= ping_v4_seq_start,
1160 	.show		= ping_v4_seq_show,
1161 	.next		= ping_seq_next,
1162 	.stop		= ping_seq_stop,
1163 };
1164 
1165 static int __net_init ping_v4_proc_init_net(struct net *net)
1166 {
1167 	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1168 			sizeof(struct ping_iter_state)))
1169 		return -ENOMEM;
1170 	return 0;
1171 }
1172 
1173 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1174 {
1175 	remove_proc_entry("icmp", net->proc_net);
1176 }
1177 
1178 static struct pernet_operations ping_v4_net_ops = {
1179 	.init = ping_v4_proc_init_net,
1180 	.exit = ping_v4_proc_exit_net,
1181 };
1182 
1183 int __init ping_proc_init(void)
1184 {
1185 	return register_pernet_subsys(&ping_v4_net_ops);
1186 }
1187 
1188 void ping_proc_exit(void)
1189 {
1190 	unregister_pernet_subsys(&ping_v4_net_ops);
1191 }
1192 
1193 #endif
1194 
1195 void __init ping_init(void)
1196 {
1197 	int i;
1198 
1199 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1200 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1201 	spin_lock_init(&ping_table.lock);
1202 }
1203