xref: /linux/net/ipv4/ping.c (revision 84f11d5b1f2abc0e22895b7e12e037f0ec03caeb)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		"Ping" sockets
7  *
8  *		This program is free software; you can redistribute it and/or
9  *		modify it under the terms of the GNU General Public License
10  *		as published by the Free Software Foundation; either version
11  *		2 of the License, or (at your option) any later version.
12  *
13  * Based on ipv4/udp.c code.
14  *
15  * Authors:	Vasiliy Kulikov / Openwall (for Linux 2.6),
16  *		Pavel Kankovsky (for Linux 2.4.32)
17  *
18  * Pavel gave all rights to bugs to Vasiliy,
19  * none of the bugs are Pavel's now.
20  *
21  */
22 
23 #include <linux/uaccess.h>
24 #include <linux/types.h>
25 #include <linux/fcntl.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/in.h>
29 #include <linux/errno.h>
30 #include <linux/timer.h>
31 #include <linux/mm.h>
32 #include <linux/inet.h>
33 #include <linux/netdevice.h>
34 #include <net/snmp.h>
35 #include <net/ip.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <linux/skbuff.h>
39 #include <linux/proc_fs.h>
40 #include <linux/export.h>
41 #include <net/sock.h>
42 #include <net/ping.h>
43 #include <net/udp.h>
44 #include <net/route.h>
45 #include <net/inet_common.h>
46 #include <net/checksum.h>
47 
48 #if IS_ENABLED(CONFIG_IPV6)
49 #include <linux/in6.h>
50 #include <linux/icmpv6.h>
51 #include <net/addrconf.h>
52 #include <net/ipv6.h>
53 #include <net/transp_v6.h>
54 #endif
55 
56 struct ping_table {
57 	struct hlist_nulls_head	hash[PING_HTABLE_SIZE];
58 	rwlock_t		lock;
59 };
60 
61 static struct ping_table ping_table;
62 struct pingv6_ops pingv6_ops;
63 EXPORT_SYMBOL_GPL(pingv6_ops);
64 
65 static u16 ping_port_rover;
66 
67 static inline int ping_hashfn(struct net *net, unsigned int num, unsigned int mask)
68 {
69 	int res = (num + net_hash_mix(net)) & mask;
70 
71 	pr_debug("hash(%d) = %d\n", num, res);
72 	return res;
73 }
74 EXPORT_SYMBOL_GPL(ping_hash);
75 
76 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
77 					     struct net *net, unsigned int num)
78 {
79 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
80 }
81 
82 int ping_get_port(struct sock *sk, unsigned short ident)
83 {
84 	struct hlist_nulls_node *node;
85 	struct hlist_nulls_head *hlist;
86 	struct inet_sock *isk, *isk2;
87 	struct sock *sk2 = NULL;
88 
89 	isk = inet_sk(sk);
90 	write_lock_bh(&ping_table.lock);
91 	if (ident == 0) {
92 		u32 i;
93 		u16 result = ping_port_rover + 1;
94 
95 		for (i = 0; i < (1L << 16); i++, result++) {
96 			if (!result)
97 				result++; /* avoid zero */
98 			hlist = ping_hashslot(&ping_table, sock_net(sk),
99 					    result);
100 			ping_portaddr_for_each_entry(sk2, node, hlist) {
101 				isk2 = inet_sk(sk2);
102 
103 				if (isk2->inet_num == result)
104 					goto next_port;
105 			}
106 
107 			/* found */
108 			ping_port_rover = ident = result;
109 			break;
110 next_port:
111 			;
112 		}
113 		if (i >= (1L << 16))
114 			goto fail;
115 	} else {
116 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
117 		ping_portaddr_for_each_entry(sk2, node, hlist) {
118 			isk2 = inet_sk(sk2);
119 
120 			/* BUG? Why is this reuse and not reuseaddr? ping.c
121 			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
122 			 * that other ping processes can steal its packets.
123 			 */
124 			if ((isk2->inet_num == ident) &&
125 			    (sk2 != sk) &&
126 			    (!sk2->sk_reuse || !sk->sk_reuse))
127 				goto fail;
128 		}
129 	}
130 
131 	pr_debug("found port/ident = %d\n", ident);
132 	isk->inet_num = ident;
133 	if (sk_unhashed(sk)) {
134 		pr_debug("was not hashed\n");
135 		sock_hold(sk);
136 		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
137 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
138 	}
139 	write_unlock_bh(&ping_table.lock);
140 	return 0;
141 
142 fail:
143 	write_unlock_bh(&ping_table.lock);
144 	return 1;
145 }
146 EXPORT_SYMBOL_GPL(ping_get_port);
147 
148 void ping_hash(struct sock *sk)
149 {
150 	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
151 	BUG(); /* "Please do not press this button again." */
152 }
153 
154 void ping_unhash(struct sock *sk)
155 {
156 	struct inet_sock *isk = inet_sk(sk);
157 	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
158 	if (sk_hashed(sk)) {
159 		write_lock_bh(&ping_table.lock);
160 		hlist_nulls_del(&sk->sk_nulls_node);
161 		sock_put(sk);
162 		isk->inet_num = 0;
163 		isk->inet_sport = 0;
164 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
165 		write_unlock_bh(&ping_table.lock);
166 	}
167 }
168 EXPORT_SYMBOL_GPL(ping_unhash);
169 
170 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
171 {
172 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
173 	struct sock *sk = NULL;
174 	struct inet_sock *isk;
175 	struct hlist_nulls_node *hnode;
176 	int dif = skb->dev->ifindex;
177 
178 	if (skb->protocol == htons(ETH_P_IP)) {
179 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
180 			 (int)ident, &ip_hdr(skb)->daddr, dif);
181 #if IS_ENABLED(CONFIG_IPV6)
182 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
183 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
184 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
185 #endif
186 	}
187 
188 	read_lock_bh(&ping_table.lock);
189 
190 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
191 		isk = inet_sk(sk);
192 
193 		pr_debug("iterate\n");
194 		if (isk->inet_num != ident)
195 			continue;
196 
197 		if (skb->protocol == htons(ETH_P_IP) &&
198 		    sk->sk_family == AF_INET) {
199 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
200 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
201 				 sk->sk_bound_dev_if);
202 
203 			if (isk->inet_rcv_saddr &&
204 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
205 				continue;
206 #if IS_ENABLED(CONFIG_IPV6)
207 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
208 			   sk->sk_family == AF_INET6) {
209 
210 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
211 				 (int) isk->inet_num,
212 				 &sk->sk_v6_rcv_saddr,
213 				 sk->sk_bound_dev_if);
214 
215 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
216 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
217 					     &ipv6_hdr(skb)->daddr))
218 				continue;
219 #endif
220 		} else {
221 			continue;
222 		}
223 
224 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
225 			continue;
226 
227 		sock_hold(sk);
228 		goto exit;
229 	}
230 
231 	sk = NULL;
232 exit:
233 	read_unlock_bh(&ping_table.lock);
234 
235 	return sk;
236 }
237 
238 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
239 					  kgid_t *high)
240 {
241 	kgid_t *data = net->ipv4.ping_group_range.range;
242 	unsigned int seq;
243 
244 	do {
245 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
246 
247 		*low = data[0];
248 		*high = data[1];
249 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
250 }
251 
252 
253 int ping_init_sock(struct sock *sk)
254 {
255 	struct net *net = sock_net(sk);
256 	kgid_t group = current_egid();
257 	struct group_info *group_info;
258 	int i, j, count;
259 	kgid_t low, high;
260 	int ret = 0;
261 
262 	inet_get_ping_group_range_net(net, &low, &high);
263 	if (gid_lte(low, group) && gid_lte(group, high))
264 		return 0;
265 
266 	group_info = get_current_groups();
267 	count = group_info->ngroups;
268 	for (i = 0; i < group_info->nblocks; i++) {
269 		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
270 		for (j = 0; j < cp_count; j++) {
271 			kgid_t gid = group_info->blocks[i][j];
272 			if (gid_lte(low, gid) && gid_lte(gid, high))
273 				goto out_release_group;
274 		}
275 
276 		count -= cp_count;
277 	}
278 
279 	ret = -EACCES;
280 
281 out_release_group:
282 	put_group_info(group_info);
283 	return ret;
284 }
285 EXPORT_SYMBOL_GPL(ping_init_sock);
286 
287 void ping_close(struct sock *sk, long timeout)
288 {
289 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
290 		 inet_sk(sk), inet_sk(sk)->inet_num);
291 	pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
292 
293 	sk_common_release(sk);
294 }
295 EXPORT_SYMBOL_GPL(ping_close);
296 
297 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
298 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
299 				struct sockaddr *uaddr, int addr_len) {
300 	struct net *net = sock_net(sk);
301 	if (sk->sk_family == AF_INET) {
302 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
303 		int chk_addr_ret;
304 
305 		if (addr_len < sizeof(*addr))
306 			return -EINVAL;
307 
308 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
309 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
310 
311 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
312 
313 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
314 			chk_addr_ret = RTN_LOCAL;
315 
316 		if ((net->ipv4.sysctl_ip_nonlocal_bind == 0 &&
317 		    isk->freebind == 0 && isk->transparent == 0 &&
318 		     chk_addr_ret != RTN_LOCAL) ||
319 		    chk_addr_ret == RTN_MULTICAST ||
320 		    chk_addr_ret == RTN_BROADCAST)
321 			return -EADDRNOTAVAIL;
322 
323 #if IS_ENABLED(CONFIG_IPV6)
324 	} else if (sk->sk_family == AF_INET6) {
325 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
326 		int addr_type, scoped, has_addr;
327 		struct net_device *dev = NULL;
328 
329 		if (addr_len < sizeof(*addr))
330 			return -EINVAL;
331 
332 		if (addr->sin6_family != AF_INET6)
333 			return -EINVAL;
334 
335 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
336 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
337 
338 		addr_type = ipv6_addr_type(&addr->sin6_addr);
339 		scoped = __ipv6_addr_needs_scope_id(addr_type);
340 		if ((addr_type != IPV6_ADDR_ANY &&
341 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
342 		    (scoped && !addr->sin6_scope_id))
343 			return -EINVAL;
344 
345 		rcu_read_lock();
346 		if (addr->sin6_scope_id) {
347 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
348 			if (!dev) {
349 				rcu_read_unlock();
350 				return -ENODEV;
351 			}
352 		}
353 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
354 						    scoped);
355 		rcu_read_unlock();
356 
357 		if (!(isk->freebind || isk->transparent || has_addr ||
358 		      addr_type == IPV6_ADDR_ANY))
359 			return -EADDRNOTAVAIL;
360 
361 		if (scoped)
362 			sk->sk_bound_dev_if = addr->sin6_scope_id;
363 #endif
364 	} else {
365 		return -EAFNOSUPPORT;
366 	}
367 	return 0;
368 }
369 
370 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
371 {
372 	if (saddr->sa_family == AF_INET) {
373 		struct inet_sock *isk = inet_sk(sk);
374 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
375 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
376 #if IS_ENABLED(CONFIG_IPV6)
377 	} else if (saddr->sa_family == AF_INET6) {
378 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
379 		struct ipv6_pinfo *np = inet6_sk(sk);
380 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
381 #endif
382 	}
383 }
384 
385 static void ping_clear_saddr(struct sock *sk, int dif)
386 {
387 	sk->sk_bound_dev_if = dif;
388 	if (sk->sk_family == AF_INET) {
389 		struct inet_sock *isk = inet_sk(sk);
390 		isk->inet_rcv_saddr = isk->inet_saddr = 0;
391 #if IS_ENABLED(CONFIG_IPV6)
392 	} else if (sk->sk_family == AF_INET6) {
393 		struct ipv6_pinfo *np = inet6_sk(sk);
394 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
395 		memset(&np->saddr, 0, sizeof(np->saddr));
396 #endif
397 	}
398 }
399 /*
400  * We need our own bind because there are no privileged id's == local ports.
401  * Moreover, we don't allow binding to multi- and broadcast addresses.
402  */
403 
404 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
405 {
406 	struct inet_sock *isk = inet_sk(sk);
407 	unsigned short snum;
408 	int err;
409 	int dif = sk->sk_bound_dev_if;
410 
411 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
412 	if (err)
413 		return err;
414 
415 	lock_sock(sk);
416 
417 	err = -EINVAL;
418 	if (isk->inet_num != 0)
419 		goto out;
420 
421 	err = -EADDRINUSE;
422 	ping_set_saddr(sk, uaddr);
423 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
424 	if (ping_get_port(sk, snum) != 0) {
425 		ping_clear_saddr(sk, dif);
426 		goto out;
427 	}
428 
429 	pr_debug("after bind(): num = %d, dif = %d\n",
430 		 (int)isk->inet_num,
431 		 (int)sk->sk_bound_dev_if);
432 
433 	err = 0;
434 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
435 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
436 #if IS_ENABLED(CONFIG_IPV6)
437 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
438 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
439 #endif
440 
441 	if (snum)
442 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
443 	isk->inet_sport = htons(isk->inet_num);
444 	isk->inet_daddr = 0;
445 	isk->inet_dport = 0;
446 
447 #if IS_ENABLED(CONFIG_IPV6)
448 	if (sk->sk_family == AF_INET6)
449 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
450 #endif
451 
452 	sk_dst_reset(sk);
453 out:
454 	release_sock(sk);
455 	pr_debug("ping_v4_bind -> %d\n", err);
456 	return err;
457 }
458 EXPORT_SYMBOL_GPL(ping_bind);
459 
460 /*
461  * Is this a supported type of ICMP message?
462  */
463 
464 static inline int ping_supported(int family, int type, int code)
465 {
466 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
467 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
468 }
469 
470 /*
471  * This routine is called by the ICMP module when it gets some
472  * sort of error condition.
473  */
474 
475 void ping_err(struct sk_buff *skb, int offset, u32 info)
476 {
477 	int family;
478 	struct icmphdr *icmph;
479 	struct inet_sock *inet_sock;
480 	int type;
481 	int code;
482 	struct net *net = dev_net(skb->dev);
483 	struct sock *sk;
484 	int harderr;
485 	int err;
486 
487 	if (skb->protocol == htons(ETH_P_IP)) {
488 		family = AF_INET;
489 		type = icmp_hdr(skb)->type;
490 		code = icmp_hdr(skb)->code;
491 		icmph = (struct icmphdr *)(skb->data + offset);
492 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
493 		family = AF_INET6;
494 		type = icmp6_hdr(skb)->icmp6_type;
495 		code = icmp6_hdr(skb)->icmp6_code;
496 		icmph = (struct icmphdr *) (skb->data + offset);
497 	} else {
498 		BUG();
499 	}
500 
501 	/* We assume the packet has already been checked by icmp_unreach */
502 
503 	if (!ping_supported(family, icmph->type, icmph->code))
504 		return;
505 
506 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
507 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
508 		 ntohs(icmph->un.echo.sequence));
509 
510 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
511 	if (sk == NULL) {
512 		pr_debug("no socket, dropping\n");
513 		return;	/* No socket for error */
514 	}
515 	pr_debug("err on socket %p\n", sk);
516 
517 	err = 0;
518 	harderr = 0;
519 	inet_sock = inet_sk(sk);
520 
521 	if (skb->protocol == htons(ETH_P_IP)) {
522 		switch (type) {
523 		default:
524 		case ICMP_TIME_EXCEEDED:
525 			err = EHOSTUNREACH;
526 			break;
527 		case ICMP_SOURCE_QUENCH:
528 			/* This is not a real error but ping wants to see it.
529 			 * Report it with some fake errno.
530 			 */
531 			err = EREMOTEIO;
532 			break;
533 		case ICMP_PARAMETERPROB:
534 			err = EPROTO;
535 			harderr = 1;
536 			break;
537 		case ICMP_DEST_UNREACH:
538 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
539 				ipv4_sk_update_pmtu(skb, sk, info);
540 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
541 					err = EMSGSIZE;
542 					harderr = 1;
543 					break;
544 				}
545 				goto out;
546 			}
547 			err = EHOSTUNREACH;
548 			if (code <= NR_ICMP_UNREACH) {
549 				harderr = icmp_err_convert[code].fatal;
550 				err = icmp_err_convert[code].errno;
551 			}
552 			break;
553 		case ICMP_REDIRECT:
554 			/* See ICMP_SOURCE_QUENCH */
555 			ipv4_sk_redirect(skb, sk);
556 			err = EREMOTEIO;
557 			break;
558 		}
559 #if IS_ENABLED(CONFIG_IPV6)
560 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
561 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
562 #endif
563 	}
564 
565 	/*
566 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
567 	 *	4.1.3.3.
568 	 */
569 	if ((family == AF_INET && !inet_sock->recverr) ||
570 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
571 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
572 			goto out;
573 	} else {
574 		if (family == AF_INET) {
575 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
576 				      info, (u8 *)icmph);
577 #if IS_ENABLED(CONFIG_IPV6)
578 		} else if (family == AF_INET6) {
579 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
580 						   info, (u8 *)icmph);
581 #endif
582 		}
583 	}
584 	sk->sk_err = err;
585 	sk->sk_error_report(sk);
586 out:
587 	sock_put(sk);
588 }
589 EXPORT_SYMBOL_GPL(ping_err);
590 
591 /*
592  *	Copy and checksum an ICMP Echo packet from user space into a buffer
593  *	starting from the payload.
594  */
595 
596 int ping_getfrag(void *from, char *to,
597 		 int offset, int fraglen, int odd, struct sk_buff *skb)
598 {
599 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
600 
601 	if (offset == 0) {
602 		fraglen -= sizeof(struct icmphdr);
603 		if (fraglen < 0)
604 			BUG();
605 		if (csum_and_copy_from_iter(to + sizeof(struct icmphdr),
606 			    fraglen, &pfh->wcheck,
607 			    &pfh->msg->msg_iter) != fraglen)
608 			return -EFAULT;
609 	} else if (offset < sizeof(struct icmphdr)) {
610 			BUG();
611 	} else {
612 		if (csum_and_copy_from_iter(to, fraglen, &pfh->wcheck,
613 					    &pfh->msg->msg_iter) != fraglen)
614 			return -EFAULT;
615 	}
616 
617 #if IS_ENABLED(CONFIG_IPV6)
618 	/* For IPv6, checksum each skb as we go along, as expected by
619 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
620 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
621 	 */
622 	if (pfh->family == AF_INET6) {
623 		skb->csum = pfh->wcheck;
624 		skb->ip_summed = CHECKSUM_NONE;
625 		pfh->wcheck = 0;
626 	}
627 #endif
628 
629 	return 0;
630 }
631 EXPORT_SYMBOL_GPL(ping_getfrag);
632 
633 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
634 				       struct flowi4 *fl4)
635 {
636 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
637 
638 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
639 		sizeof(struct icmphdr), pfh->wcheck);
640 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
641 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
642 	skb->ip_summed = CHECKSUM_NONE;
643 	return ip_push_pending_frames(sk, fl4);
644 }
645 
646 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
647 			void *user_icmph, size_t icmph_len) {
648 	u8 type, code;
649 
650 	if (len > 0xFFFF)
651 		return -EMSGSIZE;
652 
653 	/*
654 	 *	Check the flags.
655 	 */
656 
657 	/* Mirror BSD error message compatibility */
658 	if (msg->msg_flags & MSG_OOB)
659 		return -EOPNOTSUPP;
660 
661 	/*
662 	 *	Fetch the ICMP header provided by the userland.
663 	 *	iovec is modified! The ICMP header is consumed.
664 	 */
665 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
666 		return -EFAULT;
667 
668 	if (family == AF_INET) {
669 		type = ((struct icmphdr *) user_icmph)->type;
670 		code = ((struct icmphdr *) user_icmph)->code;
671 #if IS_ENABLED(CONFIG_IPV6)
672 	} else if (family == AF_INET6) {
673 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
674 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
675 #endif
676 	} else {
677 		BUG();
678 	}
679 
680 	if (!ping_supported(family, type, code))
681 		return -EINVAL;
682 
683 	return 0;
684 }
685 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
686 
687 static int ping_v4_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
688 			   size_t len)
689 {
690 	struct net *net = sock_net(sk);
691 	struct flowi4 fl4;
692 	struct inet_sock *inet = inet_sk(sk);
693 	struct ipcm_cookie ipc;
694 	struct icmphdr user_icmph;
695 	struct pingfakehdr pfh;
696 	struct rtable *rt = NULL;
697 	struct ip_options_data opt_copy;
698 	int free = 0;
699 	__be32 saddr, daddr, faddr;
700 	u8  tos;
701 	int err;
702 
703 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
704 
705 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
706 				  sizeof(user_icmph));
707 	if (err)
708 		return err;
709 
710 	/*
711 	 *	Get and verify the address.
712 	 */
713 
714 	if (msg->msg_name) {
715 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
716 		if (msg->msg_namelen < sizeof(*usin))
717 			return -EINVAL;
718 		if (usin->sin_family != AF_INET)
719 			return -EINVAL;
720 		daddr = usin->sin_addr.s_addr;
721 		/* no remote port */
722 	} else {
723 		if (sk->sk_state != TCP_ESTABLISHED)
724 			return -EDESTADDRREQ;
725 		daddr = inet->inet_daddr;
726 		/* no remote port */
727 	}
728 
729 	ipc.addr = inet->inet_saddr;
730 	ipc.opt = NULL;
731 	ipc.oif = sk->sk_bound_dev_if;
732 	ipc.tx_flags = 0;
733 	ipc.ttl = 0;
734 	ipc.tos = -1;
735 
736 	sock_tx_timestamp(sk, &ipc.tx_flags);
737 
738 	if (msg->msg_controllen) {
739 		err = ip_cmsg_send(sock_net(sk), msg, &ipc, false);
740 		if (err)
741 			return err;
742 		if (ipc.opt)
743 			free = 1;
744 	}
745 	if (!ipc.opt) {
746 		struct ip_options_rcu *inet_opt;
747 
748 		rcu_read_lock();
749 		inet_opt = rcu_dereference(inet->inet_opt);
750 		if (inet_opt) {
751 			memcpy(&opt_copy, inet_opt,
752 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
753 			ipc.opt = &opt_copy.opt;
754 		}
755 		rcu_read_unlock();
756 	}
757 
758 	saddr = ipc.addr;
759 	ipc.addr = faddr = daddr;
760 
761 	if (ipc.opt && ipc.opt->opt.srr) {
762 		if (!daddr)
763 			return -EINVAL;
764 		faddr = ipc.opt->opt.faddr;
765 	}
766 	tos = get_rttos(&ipc, inet);
767 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
768 	    (msg->msg_flags & MSG_DONTROUTE) ||
769 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
770 		tos |= RTO_ONLINK;
771 	}
772 
773 	if (ipv4_is_multicast(daddr)) {
774 		if (!ipc.oif)
775 			ipc.oif = inet->mc_index;
776 		if (!saddr)
777 			saddr = inet->mc_addr;
778 	} else if (!ipc.oif)
779 		ipc.oif = inet->uc_index;
780 
781 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
782 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
783 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
784 
785 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
786 	rt = ip_route_output_flow(net, &fl4, sk);
787 	if (IS_ERR(rt)) {
788 		err = PTR_ERR(rt);
789 		rt = NULL;
790 		if (err == -ENETUNREACH)
791 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
792 		goto out;
793 	}
794 
795 	err = -EACCES;
796 	if ((rt->rt_flags & RTCF_BROADCAST) &&
797 	    !sock_flag(sk, SOCK_BROADCAST))
798 		goto out;
799 
800 	if (msg->msg_flags & MSG_CONFIRM)
801 		goto do_confirm;
802 back_from_confirm:
803 
804 	if (!ipc.addr)
805 		ipc.addr = fl4.daddr;
806 
807 	lock_sock(sk);
808 
809 	pfh.icmph.type = user_icmph.type; /* already checked */
810 	pfh.icmph.code = user_icmph.code; /* ditto */
811 	pfh.icmph.checksum = 0;
812 	pfh.icmph.un.echo.id = inet->inet_sport;
813 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
814 	pfh.msg = msg;
815 	pfh.wcheck = 0;
816 	pfh.family = AF_INET;
817 
818 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
819 			0, &ipc, &rt, msg->msg_flags);
820 	if (err)
821 		ip_flush_pending_frames(sk);
822 	else
823 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
824 	release_sock(sk);
825 
826 out:
827 	ip_rt_put(rt);
828 	if (free)
829 		kfree(ipc.opt);
830 	if (!err) {
831 		icmp_out_count(sock_net(sk), user_icmph.type);
832 		return len;
833 	}
834 	return err;
835 
836 do_confirm:
837 	dst_confirm(&rt->dst);
838 	if (!(msg->msg_flags & MSG_PROBE) || len)
839 		goto back_from_confirm;
840 	err = 0;
841 	goto out;
842 }
843 
844 int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
845 		 size_t len, int noblock, int flags, int *addr_len)
846 {
847 	struct inet_sock *isk = inet_sk(sk);
848 	int family = sk->sk_family;
849 	struct sk_buff *skb;
850 	int copied, err;
851 
852 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
853 
854 	err = -EOPNOTSUPP;
855 	if (flags & MSG_OOB)
856 		goto out;
857 
858 	if (flags & MSG_ERRQUEUE)
859 		return inet_recv_error(sk, msg, len, addr_len);
860 
861 	skb = skb_recv_datagram(sk, flags, noblock, &err);
862 	if (!skb)
863 		goto out;
864 
865 	copied = skb->len;
866 	if (copied > len) {
867 		msg->msg_flags |= MSG_TRUNC;
868 		copied = len;
869 	}
870 
871 	/* Don't bother checking the checksum */
872 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
873 	if (err)
874 		goto done;
875 
876 	sock_recv_timestamp(msg, sk, skb);
877 
878 	/* Copy the address and add cmsg data. */
879 	if (family == AF_INET) {
880 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
881 
882 		if (sin) {
883 			sin->sin_family = AF_INET;
884 			sin->sin_port = 0 /* skb->h.uh->source */;
885 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
886 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
887 			*addr_len = sizeof(*sin);
888 		}
889 
890 		if (isk->cmsg_flags)
891 			ip_cmsg_recv(msg, skb);
892 
893 #if IS_ENABLED(CONFIG_IPV6)
894 	} else if (family == AF_INET6) {
895 		struct ipv6_pinfo *np = inet6_sk(sk);
896 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
897 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
898 
899 		if (sin6) {
900 			sin6->sin6_family = AF_INET6;
901 			sin6->sin6_port = 0;
902 			sin6->sin6_addr = ip6->saddr;
903 			sin6->sin6_flowinfo = 0;
904 			if (np->sndflow)
905 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
906 			sin6->sin6_scope_id =
907 				ipv6_iface_scope_id(&sin6->sin6_addr,
908 						    inet6_iif(skb));
909 			*addr_len = sizeof(*sin6);
910 		}
911 
912 		if (inet6_sk(sk)->rxopt.all)
913 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
914 		if (skb->protocol == htons(ETH_P_IPV6) &&
915 		    inet6_sk(sk)->rxopt.all)
916 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
917 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
918 			ip_cmsg_recv(msg, skb);
919 #endif
920 	} else {
921 		BUG();
922 	}
923 
924 	err = copied;
925 
926 done:
927 	skb_free_datagram(sk, skb);
928 out:
929 	pr_debug("ping_recvmsg -> %d\n", err);
930 	return err;
931 }
932 EXPORT_SYMBOL_GPL(ping_recvmsg);
933 
934 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
935 {
936 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
937 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
938 	if (sock_queue_rcv_skb(sk, skb) < 0) {
939 		kfree_skb(skb);
940 		pr_debug("ping_queue_rcv_skb -> failed\n");
941 		return -1;
942 	}
943 	return 0;
944 }
945 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
946 
947 
948 /*
949  *	All we need to do is get the socket.
950  */
951 
952 bool ping_rcv(struct sk_buff *skb)
953 {
954 	struct sock *sk;
955 	struct net *net = dev_net(skb->dev);
956 	struct icmphdr *icmph = icmp_hdr(skb);
957 
958 	/* We assume the packet has already been checked by icmp_rcv */
959 
960 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
961 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
962 
963 	/* Push ICMP header back */
964 	skb_push(skb, skb->data - (u8 *)icmph);
965 
966 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
967 	if (sk != NULL) {
968 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
969 
970 		pr_debug("rcv on socket %p\n", sk);
971 		if (skb2)
972 			ping_queue_rcv_skb(sk, skb2);
973 		sock_put(sk);
974 		return true;
975 	}
976 	pr_debug("no socket, dropping\n");
977 
978 	return false;
979 }
980 EXPORT_SYMBOL_GPL(ping_rcv);
981 
982 struct proto ping_prot = {
983 	.name =		"PING",
984 	.owner =	THIS_MODULE,
985 	.init =		ping_init_sock,
986 	.close =	ping_close,
987 	.connect =	ip4_datagram_connect,
988 	.disconnect =	udp_disconnect,
989 	.setsockopt =	ip_setsockopt,
990 	.getsockopt =	ip_getsockopt,
991 	.sendmsg =	ping_v4_sendmsg,
992 	.recvmsg =	ping_recvmsg,
993 	.bind =		ping_bind,
994 	.backlog_rcv =	ping_queue_rcv_skb,
995 	.release_cb =	ip4_datagram_release_cb,
996 	.hash =		ping_hash,
997 	.unhash =	ping_unhash,
998 	.get_port =	ping_get_port,
999 	.obj_size =	sizeof(struct inet_sock),
1000 };
1001 EXPORT_SYMBOL(ping_prot);
1002 
1003 #ifdef CONFIG_PROC_FS
1004 
1005 static struct sock *ping_get_first(struct seq_file *seq, int start)
1006 {
1007 	struct sock *sk;
1008 	struct ping_iter_state *state = seq->private;
1009 	struct net *net = seq_file_net(seq);
1010 
1011 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1012 	     ++state->bucket) {
1013 		struct hlist_nulls_node *node;
1014 		struct hlist_nulls_head *hslot;
1015 
1016 		hslot = &ping_table.hash[state->bucket];
1017 
1018 		if (hlist_nulls_empty(hslot))
1019 			continue;
1020 
1021 		sk_nulls_for_each(sk, node, hslot) {
1022 			if (net_eq(sock_net(sk), net) &&
1023 			    sk->sk_family == state->family)
1024 				goto found;
1025 		}
1026 	}
1027 	sk = NULL;
1028 found:
1029 	return sk;
1030 }
1031 
1032 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1033 {
1034 	struct ping_iter_state *state = seq->private;
1035 	struct net *net = seq_file_net(seq);
1036 
1037 	do {
1038 		sk = sk_nulls_next(sk);
1039 	} while (sk && (!net_eq(sock_net(sk), net)));
1040 
1041 	if (!sk)
1042 		return ping_get_first(seq, state->bucket + 1);
1043 	return sk;
1044 }
1045 
1046 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1047 {
1048 	struct sock *sk = ping_get_first(seq, 0);
1049 
1050 	if (sk)
1051 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1052 			--pos;
1053 	return pos ? NULL : sk;
1054 }
1055 
1056 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1057 {
1058 	struct ping_iter_state *state = seq->private;
1059 	state->bucket = 0;
1060 	state->family = family;
1061 
1062 	read_lock_bh(&ping_table.lock);
1063 
1064 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1065 }
1066 EXPORT_SYMBOL_GPL(ping_seq_start);
1067 
1068 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1069 {
1070 	return ping_seq_start(seq, pos, AF_INET);
1071 }
1072 
1073 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1074 {
1075 	struct sock *sk;
1076 
1077 	if (v == SEQ_START_TOKEN)
1078 		sk = ping_get_idx(seq, 0);
1079 	else
1080 		sk = ping_get_next(seq, v);
1081 
1082 	++*pos;
1083 	return sk;
1084 }
1085 EXPORT_SYMBOL_GPL(ping_seq_next);
1086 
1087 void ping_seq_stop(struct seq_file *seq, void *v)
1088 {
1089 	read_unlock_bh(&ping_table.lock);
1090 }
1091 EXPORT_SYMBOL_GPL(ping_seq_stop);
1092 
1093 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1094 		int bucket)
1095 {
1096 	struct inet_sock *inet = inet_sk(sp);
1097 	__be32 dest = inet->inet_daddr;
1098 	__be32 src = inet->inet_rcv_saddr;
1099 	__u16 destp = ntohs(inet->inet_dport);
1100 	__u16 srcp = ntohs(inet->inet_sport);
1101 
1102 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1103 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1104 		bucket, src, srcp, dest, destp, sp->sk_state,
1105 		sk_wmem_alloc_get(sp),
1106 		sk_rmem_alloc_get(sp),
1107 		0, 0L, 0,
1108 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1109 		0, sock_i_ino(sp),
1110 		atomic_read(&sp->sk_refcnt), sp,
1111 		atomic_read(&sp->sk_drops));
1112 }
1113 
1114 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1115 {
1116 	seq_setwidth(seq, 127);
1117 	if (v == SEQ_START_TOKEN)
1118 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1119 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1120 			   "inode ref pointer drops");
1121 	else {
1122 		struct ping_iter_state *state = seq->private;
1123 
1124 		ping_v4_format_sock(v, seq, state->bucket);
1125 	}
1126 	seq_pad(seq, '\n');
1127 	return 0;
1128 }
1129 
1130 static const struct seq_operations ping_v4_seq_ops = {
1131 	.show		= ping_v4_seq_show,
1132 	.start		= ping_v4_seq_start,
1133 	.next		= ping_seq_next,
1134 	.stop		= ping_seq_stop,
1135 };
1136 
1137 static int ping_seq_open(struct inode *inode, struct file *file)
1138 {
1139 	struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1140 	return seq_open_net(inode, file, &afinfo->seq_ops,
1141 			   sizeof(struct ping_iter_state));
1142 }
1143 
1144 const struct file_operations ping_seq_fops = {
1145 	.open		= ping_seq_open,
1146 	.read		= seq_read,
1147 	.llseek		= seq_lseek,
1148 	.release	= seq_release_net,
1149 };
1150 EXPORT_SYMBOL_GPL(ping_seq_fops);
1151 
1152 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1153 	.name		= "icmp",
1154 	.family		= AF_INET,
1155 	.seq_fops	= &ping_seq_fops,
1156 	.seq_ops	= {
1157 		.start		= ping_v4_seq_start,
1158 		.show		= ping_v4_seq_show,
1159 		.next		= ping_seq_next,
1160 		.stop		= ping_seq_stop,
1161 	},
1162 };
1163 
1164 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1165 {
1166 	struct proc_dir_entry *p;
1167 	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1168 			     afinfo->seq_fops, afinfo);
1169 	if (!p)
1170 		return -ENOMEM;
1171 	return 0;
1172 }
1173 EXPORT_SYMBOL_GPL(ping_proc_register);
1174 
1175 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1176 {
1177 	remove_proc_entry(afinfo->name, net->proc_net);
1178 }
1179 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1180 
1181 static int __net_init ping_v4_proc_init_net(struct net *net)
1182 {
1183 	return ping_proc_register(net, &ping_v4_seq_afinfo);
1184 }
1185 
1186 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1187 {
1188 	ping_proc_unregister(net, &ping_v4_seq_afinfo);
1189 }
1190 
1191 static struct pernet_operations ping_v4_net_ops = {
1192 	.init = ping_v4_proc_init_net,
1193 	.exit = ping_v4_proc_exit_net,
1194 };
1195 
1196 int __init ping_proc_init(void)
1197 {
1198 	return register_pernet_subsys(&ping_v4_net_ops);
1199 }
1200 
1201 void ping_proc_exit(void)
1202 {
1203 	unregister_pernet_subsys(&ping_v4_net_ops);
1204 }
1205 
1206 #endif
1207 
1208 void __init ping_init(void)
1209 {
1210 	int i;
1211 
1212 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1213 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1214 	rwlock_init(&ping_table.lock);
1215 }
1216