xref: /openbmc/linux/net/ipv4/ping.c (revision dfc53baa)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		"Ping" sockets
8  *
9  * Based on ipv4/udp.c code.
10  *
11  * Authors:	Vasiliy Kulikov / Openwall (for Linux 2.6),
12  *		Pavel Kankovsky (for Linux 2.4.32)
13  *
14  * Pavel gave all rights to bugs to Vasiliy,
15  * none of the bugs are Pavel's now.
16  */
17 
18 #include <linux/uaccess.h>
19 #include <linux/types.h>
20 #include <linux/fcntl.h>
21 #include <linux/socket.h>
22 #include <linux/sockios.h>
23 #include <linux/in.h>
24 #include <linux/errno.h>
25 #include <linux/timer.h>
26 #include <linux/mm.h>
27 #include <linux/inet.h>
28 #include <linux/netdevice.h>
29 #include <net/snmp.h>
30 #include <net/ip.h>
31 #include <net/icmp.h>
32 #include <net/protocol.h>
33 #include <linux/skbuff.h>
34 #include <linux/proc_fs.h>
35 #include <linux/export.h>
36 #include <net/sock.h>
37 #include <net/ping.h>
38 #include <net/udp.h>
39 #include <net/route.h>
40 #include <net/inet_common.h>
41 #include <net/checksum.h>
42 
43 #if IS_ENABLED(CONFIG_IPV6)
44 #include <linux/in6.h>
45 #include <linux/icmpv6.h>
46 #include <net/addrconf.h>
47 #include <net/ipv6.h>
48 #include <net/transp_v6.h>
49 #endif
50 
51 struct ping_table {
52 	struct hlist_nulls_head	hash[PING_HTABLE_SIZE];
53 	rwlock_t		lock;
54 };
55 
56 static struct ping_table ping_table;
57 struct pingv6_ops pingv6_ops;
58 EXPORT_SYMBOL_GPL(pingv6_ops);
59 
60 static u16 ping_port_rover;
61 
62 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
63 {
64 	u32 res = (num + net_hash_mix(net)) & mask;
65 
66 	pr_debug("hash(%u) = %u\n", num, res);
67 	return res;
68 }
69 EXPORT_SYMBOL_GPL(ping_hash);
70 
71 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
72 					     struct net *net, unsigned int num)
73 {
74 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
75 }
76 
77 int ping_get_port(struct sock *sk, unsigned short ident)
78 {
79 	struct hlist_nulls_node *node;
80 	struct hlist_nulls_head *hlist;
81 	struct inet_sock *isk, *isk2;
82 	struct sock *sk2 = NULL;
83 
84 	isk = inet_sk(sk);
85 	write_lock_bh(&ping_table.lock);
86 	if (ident == 0) {
87 		u32 i;
88 		u16 result = ping_port_rover + 1;
89 
90 		for (i = 0; i < (1L << 16); i++, result++) {
91 			if (!result)
92 				result++; /* avoid zero */
93 			hlist = ping_hashslot(&ping_table, sock_net(sk),
94 					    result);
95 			ping_portaddr_for_each_entry(sk2, node, hlist) {
96 				isk2 = inet_sk(sk2);
97 
98 				if (isk2->inet_num == result)
99 					goto next_port;
100 			}
101 
102 			/* found */
103 			ping_port_rover = ident = result;
104 			break;
105 next_port:
106 			;
107 		}
108 		if (i >= (1L << 16))
109 			goto fail;
110 	} else {
111 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
112 		ping_portaddr_for_each_entry(sk2, node, hlist) {
113 			isk2 = inet_sk(sk2);
114 
115 			/* BUG? Why is this reuse and not reuseaddr? ping.c
116 			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
117 			 * that other ping processes can steal its packets.
118 			 */
119 			if ((isk2->inet_num == ident) &&
120 			    (sk2 != sk) &&
121 			    (!sk2->sk_reuse || !sk->sk_reuse))
122 				goto fail;
123 		}
124 	}
125 
126 	pr_debug("found port/ident = %d\n", ident);
127 	isk->inet_num = ident;
128 	if (sk_unhashed(sk)) {
129 		pr_debug("was not hashed\n");
130 		sock_hold(sk);
131 		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
132 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
133 	}
134 	write_unlock_bh(&ping_table.lock);
135 	return 0;
136 
137 fail:
138 	write_unlock_bh(&ping_table.lock);
139 	return 1;
140 }
141 EXPORT_SYMBOL_GPL(ping_get_port);
142 
143 int ping_hash(struct sock *sk)
144 {
145 	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
146 	BUG(); /* "Please do not press this button again." */
147 
148 	return 0;
149 }
150 
151 void ping_unhash(struct sock *sk)
152 {
153 	struct inet_sock *isk = inet_sk(sk);
154 
155 	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
156 	write_lock_bh(&ping_table.lock);
157 	if (sk_hashed(sk)) {
158 		hlist_nulls_del(&sk->sk_nulls_node);
159 		sk_nulls_node_init(&sk->sk_nulls_node);
160 		sock_put(sk);
161 		isk->inet_num = 0;
162 		isk->inet_sport = 0;
163 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
164 	}
165 	write_unlock_bh(&ping_table.lock);
166 }
167 EXPORT_SYMBOL_GPL(ping_unhash);
168 
169 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
170 {
171 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
172 	struct sock *sk = NULL;
173 	struct inet_sock *isk;
174 	struct hlist_nulls_node *hnode;
175 	int dif = skb->dev->ifindex;
176 
177 	if (skb->protocol == htons(ETH_P_IP)) {
178 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
179 			 (int)ident, &ip_hdr(skb)->daddr, dif);
180 #if IS_ENABLED(CONFIG_IPV6)
181 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
182 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
183 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
184 #endif
185 	}
186 
187 	read_lock_bh(&ping_table.lock);
188 
189 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
190 		isk = inet_sk(sk);
191 
192 		pr_debug("iterate\n");
193 		if (isk->inet_num != ident)
194 			continue;
195 
196 		if (skb->protocol == htons(ETH_P_IP) &&
197 		    sk->sk_family == AF_INET) {
198 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
199 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
200 				 sk->sk_bound_dev_if);
201 
202 			if (isk->inet_rcv_saddr &&
203 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
204 				continue;
205 #if IS_ENABLED(CONFIG_IPV6)
206 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
207 			   sk->sk_family == AF_INET6) {
208 
209 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
210 				 (int) isk->inet_num,
211 				 &sk->sk_v6_rcv_saddr,
212 				 sk->sk_bound_dev_if);
213 
214 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
215 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
216 					     &ipv6_hdr(skb)->daddr))
217 				continue;
218 #endif
219 		} else {
220 			continue;
221 		}
222 
223 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
224 			continue;
225 
226 		sock_hold(sk);
227 		goto exit;
228 	}
229 
230 	sk = NULL;
231 exit:
232 	read_unlock_bh(&ping_table.lock);
233 
234 	return sk;
235 }
236 
237 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
238 					  kgid_t *high)
239 {
240 	kgid_t *data = net->ipv4.ping_group_range.range;
241 	unsigned int seq;
242 
243 	do {
244 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
245 
246 		*low = data[0];
247 		*high = data[1];
248 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
249 }
250 
251 
252 int ping_init_sock(struct sock *sk)
253 {
254 	struct net *net = sock_net(sk);
255 	kgid_t group = current_egid();
256 	struct group_info *group_info;
257 	int i;
258 	kgid_t low, high;
259 	int ret = 0;
260 
261 	if (sk->sk_family == AF_INET6)
262 		sk->sk_ipv6only = 1;
263 
264 	inet_get_ping_group_range_net(net, &low, &high);
265 	if (gid_lte(low, group) && gid_lte(group, high))
266 		return 0;
267 
268 	group_info = get_current_groups();
269 	for (i = 0; i < group_info->ngroups; i++) {
270 		kgid_t gid = group_info->gid[i];
271 
272 		if (gid_lte(low, gid) && gid_lte(gid, high))
273 			goto out_release_group;
274 	}
275 
276 	ret = -EACCES;
277 
278 out_release_group:
279 	put_group_info(group_info);
280 	return ret;
281 }
282 EXPORT_SYMBOL_GPL(ping_init_sock);
283 
284 void ping_close(struct sock *sk, long timeout)
285 {
286 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
287 		 inet_sk(sk), inet_sk(sk)->inet_num);
288 	pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
289 
290 	sk_common_release(sk);
291 }
292 EXPORT_SYMBOL_GPL(ping_close);
293 
294 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
295 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
296 				struct sockaddr *uaddr, int addr_len) {
297 	struct net *net = sock_net(sk);
298 	if (sk->sk_family == AF_INET) {
299 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
300 		int chk_addr_ret;
301 
302 		if (addr_len < sizeof(*addr))
303 			return -EINVAL;
304 
305 		if (addr->sin_family != AF_INET &&
306 		    !(addr->sin_family == AF_UNSPEC &&
307 		      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
308 			return -EAFNOSUPPORT;
309 
310 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
311 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
312 
313 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
314 
315 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
316 			chk_addr_ret = RTN_LOCAL;
317 
318 		if ((!inet_can_nonlocal_bind(net, isk) &&
319 		     chk_addr_ret != RTN_LOCAL) ||
320 		    chk_addr_ret == RTN_MULTICAST ||
321 		    chk_addr_ret == RTN_BROADCAST)
322 			return -EADDRNOTAVAIL;
323 
324 #if IS_ENABLED(CONFIG_IPV6)
325 	} else if (sk->sk_family == AF_INET6) {
326 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
327 		int addr_type, scoped, has_addr;
328 		struct net_device *dev = NULL;
329 
330 		if (addr_len < sizeof(*addr))
331 			return -EINVAL;
332 
333 		if (addr->sin6_family != AF_INET6)
334 			return -EAFNOSUPPORT;
335 
336 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
337 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
338 
339 		addr_type = ipv6_addr_type(&addr->sin6_addr);
340 		scoped = __ipv6_addr_needs_scope_id(addr_type);
341 		if ((addr_type != IPV6_ADDR_ANY &&
342 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
343 		    (scoped && !addr->sin6_scope_id))
344 			return -EINVAL;
345 
346 		rcu_read_lock();
347 		if (addr->sin6_scope_id) {
348 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
349 			if (!dev) {
350 				rcu_read_unlock();
351 				return -ENODEV;
352 			}
353 		}
354 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
355 						    scoped);
356 		rcu_read_unlock();
357 
358 		if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
359 		      addr_type == IPV6_ADDR_ANY))
360 			return -EADDRNOTAVAIL;
361 
362 		if (scoped)
363 			sk->sk_bound_dev_if = addr->sin6_scope_id;
364 #endif
365 	} else {
366 		return -EAFNOSUPPORT;
367 	}
368 	return 0;
369 }
370 
371 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
372 {
373 	if (saddr->sa_family == AF_INET) {
374 		struct inet_sock *isk = inet_sk(sk);
375 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
376 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
377 #if IS_ENABLED(CONFIG_IPV6)
378 	} else if (saddr->sa_family == AF_INET6) {
379 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
380 		struct ipv6_pinfo *np = inet6_sk(sk);
381 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
382 #endif
383 	}
384 }
385 
386 static void ping_clear_saddr(struct sock *sk, int dif)
387 {
388 	sk->sk_bound_dev_if = dif;
389 	if (sk->sk_family == AF_INET) {
390 		struct inet_sock *isk = inet_sk(sk);
391 		isk->inet_rcv_saddr = isk->inet_saddr = 0;
392 #if IS_ENABLED(CONFIG_IPV6)
393 	} else if (sk->sk_family == AF_INET6) {
394 		struct ipv6_pinfo *np = inet6_sk(sk);
395 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
396 		memset(&np->saddr, 0, sizeof(np->saddr));
397 #endif
398 	}
399 }
400 /*
401  * We need our own bind because there are no privileged id's == local ports.
402  * Moreover, we don't allow binding to multi- and broadcast addresses.
403  */
404 
405 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
406 {
407 	struct inet_sock *isk = inet_sk(sk);
408 	unsigned short snum;
409 	int err;
410 	int dif = sk->sk_bound_dev_if;
411 
412 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
413 	if (err)
414 		return err;
415 
416 	lock_sock(sk);
417 
418 	err = -EINVAL;
419 	if (isk->inet_num != 0)
420 		goto out;
421 
422 	err = -EADDRINUSE;
423 	ping_set_saddr(sk, uaddr);
424 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
425 	if (ping_get_port(sk, snum) != 0) {
426 		ping_clear_saddr(sk, dif);
427 		goto out;
428 	}
429 
430 	pr_debug("after bind(): num = %hu, dif = %d\n",
431 		 isk->inet_num,
432 		 sk->sk_bound_dev_if);
433 
434 	err = 0;
435 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
436 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
437 #if IS_ENABLED(CONFIG_IPV6)
438 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
439 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
440 #endif
441 
442 	if (snum)
443 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
444 	isk->inet_sport = htons(isk->inet_num);
445 	isk->inet_daddr = 0;
446 	isk->inet_dport = 0;
447 
448 #if IS_ENABLED(CONFIG_IPV6)
449 	if (sk->sk_family == AF_INET6)
450 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
451 #endif
452 
453 	sk_dst_reset(sk);
454 out:
455 	release_sock(sk);
456 	pr_debug("ping_v4_bind -> %d\n", err);
457 	return err;
458 }
459 EXPORT_SYMBOL_GPL(ping_bind);
460 
461 /*
462  * Is this a supported type of ICMP message?
463  */
464 
465 static inline int ping_supported(int family, int type, int code)
466 {
467 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
468 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
469 }
470 
471 /*
472  * This routine is called by the ICMP module when it gets some
473  * sort of error condition.
474  */
475 
476 void ping_err(struct sk_buff *skb, int offset, u32 info)
477 {
478 	int family;
479 	struct icmphdr *icmph;
480 	struct inet_sock *inet_sock;
481 	int type;
482 	int code;
483 	struct net *net = dev_net(skb->dev);
484 	struct sock *sk;
485 	int harderr;
486 	int err;
487 
488 	if (skb->protocol == htons(ETH_P_IP)) {
489 		family = AF_INET;
490 		type = icmp_hdr(skb)->type;
491 		code = icmp_hdr(skb)->code;
492 		icmph = (struct icmphdr *)(skb->data + offset);
493 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
494 		family = AF_INET6;
495 		type = icmp6_hdr(skb)->icmp6_type;
496 		code = icmp6_hdr(skb)->icmp6_code;
497 		icmph = (struct icmphdr *) (skb->data + offset);
498 	} else {
499 		BUG();
500 	}
501 
502 	/* We assume the packet has already been checked by icmp_unreach */
503 
504 	if (!ping_supported(family, icmph->type, icmph->code))
505 		return;
506 
507 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
508 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
509 		 ntohs(icmph->un.echo.sequence));
510 
511 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
512 	if (!sk) {
513 		pr_debug("no socket, dropping\n");
514 		return;	/* No socket for error */
515 	}
516 	pr_debug("err on socket %p\n", sk);
517 
518 	err = 0;
519 	harderr = 0;
520 	inet_sock = inet_sk(sk);
521 
522 	if (skb->protocol == htons(ETH_P_IP)) {
523 		switch (type) {
524 		default:
525 		case ICMP_TIME_EXCEEDED:
526 			err = EHOSTUNREACH;
527 			break;
528 		case ICMP_SOURCE_QUENCH:
529 			/* This is not a real error but ping wants to see it.
530 			 * Report it with some fake errno.
531 			 */
532 			err = EREMOTEIO;
533 			break;
534 		case ICMP_PARAMETERPROB:
535 			err = EPROTO;
536 			harderr = 1;
537 			break;
538 		case ICMP_DEST_UNREACH:
539 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
540 				ipv4_sk_update_pmtu(skb, sk, info);
541 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
542 					err = EMSGSIZE;
543 					harderr = 1;
544 					break;
545 				}
546 				goto out;
547 			}
548 			err = EHOSTUNREACH;
549 			if (code <= NR_ICMP_UNREACH) {
550 				harderr = icmp_err_convert[code].fatal;
551 				err = icmp_err_convert[code].errno;
552 			}
553 			break;
554 		case ICMP_REDIRECT:
555 			/* See ICMP_SOURCE_QUENCH */
556 			ipv4_sk_redirect(skb, sk);
557 			err = EREMOTEIO;
558 			break;
559 		}
560 #if IS_ENABLED(CONFIG_IPV6)
561 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
562 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
563 #endif
564 	}
565 
566 	/*
567 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
568 	 *	4.1.3.3.
569 	 */
570 	if ((family == AF_INET && !inet_sock->recverr) ||
571 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
572 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
573 			goto out;
574 	} else {
575 		if (family == AF_INET) {
576 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
577 				      info, (u8 *)icmph);
578 #if IS_ENABLED(CONFIG_IPV6)
579 		} else if (family == AF_INET6) {
580 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
581 						   info, (u8 *)icmph);
582 #endif
583 		}
584 	}
585 	sk->sk_err = err;
586 	sk->sk_error_report(sk);
587 out:
588 	sock_put(sk);
589 }
590 EXPORT_SYMBOL_GPL(ping_err);
591 
592 /*
593  *	Copy and checksum an ICMP Echo packet from user space into a buffer
594  *	starting from the payload.
595  */
596 
597 int ping_getfrag(void *from, char *to,
598 		 int offset, int fraglen, int odd, struct sk_buff *skb)
599 {
600 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
601 
602 	if (offset == 0) {
603 		fraglen -= sizeof(struct icmphdr);
604 		if (fraglen < 0)
605 			BUG();
606 		if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
607 			    fraglen, &pfh->wcheck,
608 			    &pfh->msg->msg_iter))
609 			return -EFAULT;
610 	} else if (offset < sizeof(struct icmphdr)) {
611 			BUG();
612 	} else {
613 		if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
614 					    &pfh->msg->msg_iter))
615 			return -EFAULT;
616 	}
617 
618 #if IS_ENABLED(CONFIG_IPV6)
619 	/* For IPv6, checksum each skb as we go along, as expected by
620 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
621 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
622 	 */
623 	if (pfh->family == AF_INET6) {
624 		skb->csum = pfh->wcheck;
625 		skb->ip_summed = CHECKSUM_NONE;
626 		pfh->wcheck = 0;
627 	}
628 #endif
629 
630 	return 0;
631 }
632 EXPORT_SYMBOL_GPL(ping_getfrag);
633 
634 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
635 				       struct flowi4 *fl4)
636 {
637 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
638 
639 	if (!skb)
640 		return 0;
641 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
642 		sizeof(struct icmphdr), pfh->wcheck);
643 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
644 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
645 	skb->ip_summed = CHECKSUM_NONE;
646 	return ip_push_pending_frames(sk, fl4);
647 }
648 
649 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
650 			void *user_icmph, size_t icmph_len) {
651 	u8 type, code;
652 
653 	if (len > 0xFFFF)
654 		return -EMSGSIZE;
655 
656 	/* Must have at least a full ICMP header. */
657 	if (len < icmph_len)
658 		return -EINVAL;
659 
660 	/*
661 	 *	Check the flags.
662 	 */
663 
664 	/* Mirror BSD error message compatibility */
665 	if (msg->msg_flags & MSG_OOB)
666 		return -EOPNOTSUPP;
667 
668 	/*
669 	 *	Fetch the ICMP header provided by the userland.
670 	 *	iovec is modified! The ICMP header is consumed.
671 	 */
672 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
673 		return -EFAULT;
674 
675 	if (family == AF_INET) {
676 		type = ((struct icmphdr *) user_icmph)->type;
677 		code = ((struct icmphdr *) user_icmph)->code;
678 #if IS_ENABLED(CONFIG_IPV6)
679 	} else if (family == AF_INET6) {
680 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
681 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
682 #endif
683 	} else {
684 		BUG();
685 	}
686 
687 	if (!ping_supported(family, type, code))
688 		return -EINVAL;
689 
690 	return 0;
691 }
692 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
693 
694 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
695 {
696 	struct net *net = sock_net(sk);
697 	struct flowi4 fl4;
698 	struct inet_sock *inet = inet_sk(sk);
699 	struct ipcm_cookie ipc;
700 	struct icmphdr user_icmph;
701 	struct pingfakehdr pfh;
702 	struct rtable *rt = NULL;
703 	struct ip_options_data opt_copy;
704 	int free = 0;
705 	__be32 saddr, daddr, faddr;
706 	u8  tos;
707 	int err;
708 
709 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
710 
711 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
712 				  sizeof(user_icmph));
713 	if (err)
714 		return err;
715 
716 	/*
717 	 *	Get and verify the address.
718 	 */
719 
720 	if (msg->msg_name) {
721 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
722 		if (msg->msg_namelen < sizeof(*usin))
723 			return -EINVAL;
724 		if (usin->sin_family != AF_INET)
725 			return -EAFNOSUPPORT;
726 		daddr = usin->sin_addr.s_addr;
727 		/* no remote port */
728 	} else {
729 		if (sk->sk_state != TCP_ESTABLISHED)
730 			return -EDESTADDRREQ;
731 		daddr = inet->inet_daddr;
732 		/* no remote port */
733 	}
734 
735 	ipcm_init_sk(&ipc, inet);
736 
737 	if (msg->msg_controllen) {
738 		err = ip_cmsg_send(sk, msg, &ipc, false);
739 		if (unlikely(err)) {
740 			kfree(ipc.opt);
741 			return err;
742 		}
743 		if (ipc.opt)
744 			free = 1;
745 	}
746 	if (!ipc.opt) {
747 		struct ip_options_rcu *inet_opt;
748 
749 		rcu_read_lock();
750 		inet_opt = rcu_dereference(inet->inet_opt);
751 		if (inet_opt) {
752 			memcpy(&opt_copy, inet_opt,
753 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
754 			ipc.opt = &opt_copy.opt;
755 		}
756 		rcu_read_unlock();
757 	}
758 
759 	saddr = ipc.addr;
760 	ipc.addr = faddr = daddr;
761 
762 	if (ipc.opt && ipc.opt->opt.srr) {
763 		if (!daddr) {
764 			err = -EINVAL;
765 			goto out_free;
766 		}
767 		faddr = ipc.opt->opt.faddr;
768 	}
769 	tos = get_rttos(&ipc, inet);
770 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
771 	    (msg->msg_flags & MSG_DONTROUTE) ||
772 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
773 		tos |= RTO_ONLINK;
774 	}
775 
776 	if (ipv4_is_multicast(daddr)) {
777 		if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
778 			ipc.oif = inet->mc_index;
779 		if (!saddr)
780 			saddr = inet->mc_addr;
781 	} else if (!ipc.oif)
782 		ipc.oif = inet->uc_index;
783 
784 	flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
785 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
786 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
787 			   sk->sk_uid);
788 
789 	fl4.fl4_icmp_type = user_icmph.type;
790 	fl4.fl4_icmp_code = user_icmph.code;
791 
792 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
793 	rt = ip_route_output_flow(net, &fl4, sk);
794 	if (IS_ERR(rt)) {
795 		err = PTR_ERR(rt);
796 		rt = NULL;
797 		if (err == -ENETUNREACH)
798 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
799 		goto out;
800 	}
801 
802 	err = -EACCES;
803 	if ((rt->rt_flags & RTCF_BROADCAST) &&
804 	    !sock_flag(sk, SOCK_BROADCAST))
805 		goto out;
806 
807 	if (msg->msg_flags & MSG_CONFIRM)
808 		goto do_confirm;
809 back_from_confirm:
810 
811 	if (!ipc.addr)
812 		ipc.addr = fl4.daddr;
813 
814 	lock_sock(sk);
815 
816 	pfh.icmph.type = user_icmph.type; /* already checked */
817 	pfh.icmph.code = user_icmph.code; /* ditto */
818 	pfh.icmph.checksum = 0;
819 	pfh.icmph.un.echo.id = inet->inet_sport;
820 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
821 	pfh.msg = msg;
822 	pfh.wcheck = 0;
823 	pfh.family = AF_INET;
824 
825 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
826 			0, &ipc, &rt, msg->msg_flags);
827 	if (err)
828 		ip_flush_pending_frames(sk);
829 	else
830 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
831 	release_sock(sk);
832 
833 out:
834 	ip_rt_put(rt);
835 out_free:
836 	if (free)
837 		kfree(ipc.opt);
838 	if (!err) {
839 		icmp_out_count(sock_net(sk), user_icmph.type);
840 		return len;
841 	}
842 	return err;
843 
844 do_confirm:
845 	if (msg->msg_flags & MSG_PROBE)
846 		dst_confirm_neigh(&rt->dst, &fl4.daddr);
847 	if (!(msg->msg_flags & MSG_PROBE) || len)
848 		goto back_from_confirm;
849 	err = 0;
850 	goto out;
851 }
852 
853 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
854 		 int flags, int *addr_len)
855 {
856 	struct inet_sock *isk = inet_sk(sk);
857 	int family = sk->sk_family;
858 	struct sk_buff *skb;
859 	int copied, err;
860 
861 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
862 
863 	err = -EOPNOTSUPP;
864 	if (flags & MSG_OOB)
865 		goto out;
866 
867 	if (flags & MSG_ERRQUEUE)
868 		return inet_recv_error(sk, msg, len, addr_len);
869 
870 	skb = skb_recv_datagram(sk, flags, noblock, &err);
871 	if (!skb)
872 		goto out;
873 
874 	copied = skb->len;
875 	if (copied > len) {
876 		msg->msg_flags |= MSG_TRUNC;
877 		copied = len;
878 	}
879 
880 	/* Don't bother checking the checksum */
881 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
882 	if (err)
883 		goto done;
884 
885 	sock_recv_timestamp(msg, sk, skb);
886 
887 	/* Copy the address and add cmsg data. */
888 	if (family == AF_INET) {
889 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
890 
891 		if (sin) {
892 			sin->sin_family = AF_INET;
893 			sin->sin_port = 0 /* skb->h.uh->source */;
894 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
895 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
896 			*addr_len = sizeof(*sin);
897 		}
898 
899 		if (isk->cmsg_flags)
900 			ip_cmsg_recv(msg, skb);
901 
902 #if IS_ENABLED(CONFIG_IPV6)
903 	} else if (family == AF_INET6) {
904 		struct ipv6_pinfo *np = inet6_sk(sk);
905 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
906 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
907 
908 		if (sin6) {
909 			sin6->sin6_family = AF_INET6;
910 			sin6->sin6_port = 0;
911 			sin6->sin6_addr = ip6->saddr;
912 			sin6->sin6_flowinfo = 0;
913 			if (np->sndflow)
914 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
915 			sin6->sin6_scope_id =
916 				ipv6_iface_scope_id(&sin6->sin6_addr,
917 						    inet6_iif(skb));
918 			*addr_len = sizeof(*sin6);
919 		}
920 
921 		if (inet6_sk(sk)->rxopt.all)
922 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
923 		if (skb->protocol == htons(ETH_P_IPV6) &&
924 		    inet6_sk(sk)->rxopt.all)
925 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
926 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
927 			ip_cmsg_recv(msg, skb);
928 #endif
929 	} else {
930 		BUG();
931 	}
932 
933 	err = copied;
934 
935 done:
936 	skb_free_datagram(sk, skb);
937 out:
938 	pr_debug("ping_recvmsg -> %d\n", err);
939 	return err;
940 }
941 EXPORT_SYMBOL_GPL(ping_recvmsg);
942 
943 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
944 {
945 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
946 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
947 	if (sock_queue_rcv_skb(sk, skb) < 0) {
948 		kfree_skb(skb);
949 		pr_debug("ping_queue_rcv_skb -> failed\n");
950 		return -1;
951 	}
952 	return 0;
953 }
954 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
955 
956 
957 /*
958  *	All we need to do is get the socket.
959  */
960 
961 bool ping_rcv(struct sk_buff *skb)
962 {
963 	struct sock *sk;
964 	struct net *net = dev_net(skb->dev);
965 	struct icmphdr *icmph = icmp_hdr(skb);
966 
967 	/* We assume the packet has already been checked by icmp_rcv */
968 
969 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
970 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
971 
972 	/* Push ICMP header back */
973 	skb_push(skb, skb->data - (u8 *)icmph);
974 
975 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
976 	if (sk) {
977 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
978 
979 		pr_debug("rcv on socket %p\n", sk);
980 		if (skb2)
981 			ping_queue_rcv_skb(sk, skb2);
982 		sock_put(sk);
983 		return true;
984 	}
985 	pr_debug("no socket, dropping\n");
986 
987 	return false;
988 }
989 EXPORT_SYMBOL_GPL(ping_rcv);
990 
991 struct proto ping_prot = {
992 	.name =		"PING",
993 	.owner =	THIS_MODULE,
994 	.init =		ping_init_sock,
995 	.close =	ping_close,
996 	.connect =	ip4_datagram_connect,
997 	.disconnect =	__udp_disconnect,
998 	.setsockopt =	ip_setsockopt,
999 	.getsockopt =	ip_getsockopt,
1000 	.sendmsg =	ping_v4_sendmsg,
1001 	.recvmsg =	ping_recvmsg,
1002 	.bind =		ping_bind,
1003 	.backlog_rcv =	ping_queue_rcv_skb,
1004 	.release_cb =	ip4_datagram_release_cb,
1005 	.hash =		ping_hash,
1006 	.unhash =	ping_unhash,
1007 	.get_port =	ping_get_port,
1008 	.obj_size =	sizeof(struct inet_sock),
1009 };
1010 EXPORT_SYMBOL(ping_prot);
1011 
1012 #ifdef CONFIG_PROC_FS
1013 
1014 static struct sock *ping_get_first(struct seq_file *seq, int start)
1015 {
1016 	struct sock *sk;
1017 	struct ping_iter_state *state = seq->private;
1018 	struct net *net = seq_file_net(seq);
1019 
1020 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1021 	     ++state->bucket) {
1022 		struct hlist_nulls_node *node;
1023 		struct hlist_nulls_head *hslot;
1024 
1025 		hslot = &ping_table.hash[state->bucket];
1026 
1027 		if (hlist_nulls_empty(hslot))
1028 			continue;
1029 
1030 		sk_nulls_for_each(sk, node, hslot) {
1031 			if (net_eq(sock_net(sk), net) &&
1032 			    sk->sk_family == state->family)
1033 				goto found;
1034 		}
1035 	}
1036 	sk = NULL;
1037 found:
1038 	return sk;
1039 }
1040 
1041 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1042 {
1043 	struct ping_iter_state *state = seq->private;
1044 	struct net *net = seq_file_net(seq);
1045 
1046 	do {
1047 		sk = sk_nulls_next(sk);
1048 	} while (sk && (!net_eq(sock_net(sk), net)));
1049 
1050 	if (!sk)
1051 		return ping_get_first(seq, state->bucket + 1);
1052 	return sk;
1053 }
1054 
1055 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1056 {
1057 	struct sock *sk = ping_get_first(seq, 0);
1058 
1059 	if (sk)
1060 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1061 			--pos;
1062 	return pos ? NULL : sk;
1063 }
1064 
1065 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1066 	__acquires(ping_table.lock)
1067 {
1068 	struct ping_iter_state *state = seq->private;
1069 	state->bucket = 0;
1070 	state->family = family;
1071 
1072 	read_lock_bh(&ping_table.lock);
1073 
1074 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1075 }
1076 EXPORT_SYMBOL_GPL(ping_seq_start);
1077 
1078 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1079 {
1080 	return ping_seq_start(seq, pos, AF_INET);
1081 }
1082 
1083 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1084 {
1085 	struct sock *sk;
1086 
1087 	if (v == SEQ_START_TOKEN)
1088 		sk = ping_get_idx(seq, 0);
1089 	else
1090 		sk = ping_get_next(seq, v);
1091 
1092 	++*pos;
1093 	return sk;
1094 }
1095 EXPORT_SYMBOL_GPL(ping_seq_next);
1096 
1097 void ping_seq_stop(struct seq_file *seq, void *v)
1098 	__releases(ping_table.lock)
1099 {
1100 	read_unlock_bh(&ping_table.lock);
1101 }
1102 EXPORT_SYMBOL_GPL(ping_seq_stop);
1103 
1104 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1105 		int bucket)
1106 {
1107 	struct inet_sock *inet = inet_sk(sp);
1108 	__be32 dest = inet->inet_daddr;
1109 	__be32 src = inet->inet_rcv_saddr;
1110 	__u16 destp = ntohs(inet->inet_dport);
1111 	__u16 srcp = ntohs(inet->inet_sport);
1112 
1113 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1114 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1115 		bucket, src, srcp, dest, destp, sp->sk_state,
1116 		sk_wmem_alloc_get(sp),
1117 		sk_rmem_alloc_get(sp),
1118 		0, 0L, 0,
1119 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1120 		0, sock_i_ino(sp),
1121 		refcount_read(&sp->sk_refcnt), sp,
1122 		atomic_read(&sp->sk_drops));
1123 }
1124 
1125 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1126 {
1127 	seq_setwidth(seq, 127);
1128 	if (v == SEQ_START_TOKEN)
1129 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1130 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1131 			   "inode ref pointer drops");
1132 	else {
1133 		struct ping_iter_state *state = seq->private;
1134 
1135 		ping_v4_format_sock(v, seq, state->bucket);
1136 	}
1137 	seq_pad(seq, '\n');
1138 	return 0;
1139 }
1140 
1141 static const struct seq_operations ping_v4_seq_ops = {
1142 	.start		= ping_v4_seq_start,
1143 	.show		= ping_v4_seq_show,
1144 	.next		= ping_seq_next,
1145 	.stop		= ping_seq_stop,
1146 };
1147 
1148 static int __net_init ping_v4_proc_init_net(struct net *net)
1149 {
1150 	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1151 			sizeof(struct ping_iter_state)))
1152 		return -ENOMEM;
1153 	return 0;
1154 }
1155 
1156 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1157 {
1158 	remove_proc_entry("icmp", net->proc_net);
1159 }
1160 
1161 static struct pernet_operations ping_v4_net_ops = {
1162 	.init = ping_v4_proc_init_net,
1163 	.exit = ping_v4_proc_exit_net,
1164 };
1165 
1166 int __init ping_proc_init(void)
1167 {
1168 	return register_pernet_subsys(&ping_v4_net_ops);
1169 }
1170 
1171 void ping_proc_exit(void)
1172 {
1173 	unregister_pernet_subsys(&ping_v4_net_ops);
1174 }
1175 
1176 #endif
1177 
1178 void __init ping_init(void)
1179 {
1180 	int i;
1181 
1182 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1183 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1184 	rwlock_init(&ping_table.lock);
1185 }
1186