xref: /openbmc/linux/net/ipv4/ping.c (revision d2168146)
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 		}
221 
222 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
223 			continue;
224 
225 		sock_hold(sk);
226 		goto exit;
227 	}
228 
229 	sk = NULL;
230 exit:
231 	read_unlock_bh(&ping_table.lock);
232 
233 	return sk;
234 }
235 
236 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
237 					  kgid_t *high)
238 {
239 	kgid_t *data = net->ipv4.ping_group_range.range;
240 	unsigned int seq;
241 
242 	do {
243 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
244 
245 		*low = data[0];
246 		*high = data[1];
247 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
248 }
249 
250 
251 int ping_init_sock(struct sock *sk)
252 {
253 	struct net *net = sock_net(sk);
254 	kgid_t group = current_egid();
255 	struct group_info *group_info;
256 	int i, j, count;
257 	kgid_t low, high;
258 	int ret = 0;
259 
260 	inet_get_ping_group_range_net(net, &low, &high);
261 	if (gid_lte(low, group) && gid_lte(group, high))
262 		return 0;
263 
264 	group_info = get_current_groups();
265 	count = group_info->ngroups;
266 	for (i = 0; i < group_info->nblocks; i++) {
267 		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
268 		for (j = 0; j < cp_count; j++) {
269 			kgid_t gid = group_info->blocks[i][j];
270 			if (gid_lte(low, gid) && gid_lte(gid, high))
271 				goto out_release_group;
272 		}
273 
274 		count -= cp_count;
275 	}
276 
277 	ret = -EACCES;
278 
279 out_release_group:
280 	put_group_info(group_info);
281 	return ret;
282 }
283 EXPORT_SYMBOL_GPL(ping_init_sock);
284 
285 void ping_close(struct sock *sk, long timeout)
286 {
287 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
288 		 inet_sk(sk), inet_sk(sk)->inet_num);
289 	pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
290 
291 	sk_common_release(sk);
292 }
293 EXPORT_SYMBOL_GPL(ping_close);
294 
295 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
296 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
297 				struct sockaddr *uaddr, int addr_len) {
298 	struct net *net = sock_net(sk);
299 	if (sk->sk_family == AF_INET) {
300 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
301 		int chk_addr_ret;
302 
303 		if (addr_len < sizeof(*addr))
304 			return -EINVAL;
305 
306 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
307 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
308 
309 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
310 
311 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
312 			chk_addr_ret = RTN_LOCAL;
313 
314 		if ((sysctl_ip_nonlocal_bind == 0 &&
315 		    isk->freebind == 0 && isk->transparent == 0 &&
316 		     chk_addr_ret != RTN_LOCAL) ||
317 		    chk_addr_ret == RTN_MULTICAST ||
318 		    chk_addr_ret == RTN_BROADCAST)
319 			return -EADDRNOTAVAIL;
320 
321 #if IS_ENABLED(CONFIG_IPV6)
322 	} else if (sk->sk_family == AF_INET6) {
323 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
324 		int addr_type, scoped, has_addr;
325 		struct net_device *dev = NULL;
326 
327 		if (addr_len < sizeof(*addr))
328 			return -EINVAL;
329 
330 		if (addr->sin6_family != AF_INET6)
331 			return -EINVAL;
332 
333 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
334 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
335 
336 		addr_type = ipv6_addr_type(&addr->sin6_addr);
337 		scoped = __ipv6_addr_needs_scope_id(addr_type);
338 		if ((addr_type != IPV6_ADDR_ANY &&
339 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
340 		    (scoped && !addr->sin6_scope_id))
341 			return -EINVAL;
342 
343 		rcu_read_lock();
344 		if (addr->sin6_scope_id) {
345 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
346 			if (!dev) {
347 				rcu_read_unlock();
348 				return -ENODEV;
349 			}
350 		}
351 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
352 						    scoped);
353 		rcu_read_unlock();
354 
355 		if (!(isk->freebind || isk->transparent || has_addr ||
356 		      addr_type == IPV6_ADDR_ANY))
357 			return -EADDRNOTAVAIL;
358 
359 		if (scoped)
360 			sk->sk_bound_dev_if = addr->sin6_scope_id;
361 #endif
362 	} else {
363 		return -EAFNOSUPPORT;
364 	}
365 	return 0;
366 }
367 
368 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
369 {
370 	if (saddr->sa_family == AF_INET) {
371 		struct inet_sock *isk = inet_sk(sk);
372 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
373 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
374 #if IS_ENABLED(CONFIG_IPV6)
375 	} else if (saddr->sa_family == AF_INET6) {
376 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
377 		struct ipv6_pinfo *np = inet6_sk(sk);
378 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
379 #endif
380 	}
381 }
382 
383 static void ping_clear_saddr(struct sock *sk, int dif)
384 {
385 	sk->sk_bound_dev_if = dif;
386 	if (sk->sk_family == AF_INET) {
387 		struct inet_sock *isk = inet_sk(sk);
388 		isk->inet_rcv_saddr = isk->inet_saddr = 0;
389 #if IS_ENABLED(CONFIG_IPV6)
390 	} else if (sk->sk_family == AF_INET6) {
391 		struct ipv6_pinfo *np = inet6_sk(sk);
392 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
393 		memset(&np->saddr, 0, sizeof(np->saddr));
394 #endif
395 	}
396 }
397 /*
398  * We need our own bind because there are no privileged id's == local ports.
399  * Moreover, we don't allow binding to multi- and broadcast addresses.
400  */
401 
402 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
403 {
404 	struct inet_sock *isk = inet_sk(sk);
405 	unsigned short snum;
406 	int err;
407 	int dif = sk->sk_bound_dev_if;
408 
409 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
410 	if (err)
411 		return err;
412 
413 	lock_sock(sk);
414 
415 	err = -EINVAL;
416 	if (isk->inet_num != 0)
417 		goto out;
418 
419 	err = -EADDRINUSE;
420 	ping_set_saddr(sk, uaddr);
421 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
422 	if (ping_get_port(sk, snum) != 0) {
423 		ping_clear_saddr(sk, dif);
424 		goto out;
425 	}
426 
427 	pr_debug("after bind(): num = %d, dif = %d\n",
428 		 (int)isk->inet_num,
429 		 (int)sk->sk_bound_dev_if);
430 
431 	err = 0;
432 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
433 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
434 #if IS_ENABLED(CONFIG_IPV6)
435 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
436 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
437 #endif
438 
439 	if (snum)
440 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
441 	isk->inet_sport = htons(isk->inet_num);
442 	isk->inet_daddr = 0;
443 	isk->inet_dport = 0;
444 
445 #if IS_ENABLED(CONFIG_IPV6)
446 	if (sk->sk_family == AF_INET6)
447 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
448 #endif
449 
450 	sk_dst_reset(sk);
451 out:
452 	release_sock(sk);
453 	pr_debug("ping_v4_bind -> %d\n", err);
454 	return err;
455 }
456 EXPORT_SYMBOL_GPL(ping_bind);
457 
458 /*
459  * Is this a supported type of ICMP message?
460  */
461 
462 static inline int ping_supported(int family, int type, int code)
463 {
464 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
465 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
466 }
467 
468 /*
469  * This routine is called by the ICMP module when it gets some
470  * sort of error condition.
471  */
472 
473 void ping_err(struct sk_buff *skb, int offset, u32 info)
474 {
475 	int family;
476 	struct icmphdr *icmph;
477 	struct inet_sock *inet_sock;
478 	int type;
479 	int code;
480 	struct net *net = dev_net(skb->dev);
481 	struct sock *sk;
482 	int harderr;
483 	int err;
484 
485 	if (skb->protocol == htons(ETH_P_IP)) {
486 		family = AF_INET;
487 		type = icmp_hdr(skb)->type;
488 		code = icmp_hdr(skb)->code;
489 		icmph = (struct icmphdr *)(skb->data + offset);
490 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
491 		family = AF_INET6;
492 		type = icmp6_hdr(skb)->icmp6_type;
493 		code = icmp6_hdr(skb)->icmp6_code;
494 		icmph = (struct icmphdr *) (skb->data + offset);
495 	} else {
496 		BUG();
497 	}
498 
499 	/* We assume the packet has already been checked by icmp_unreach */
500 
501 	if (!ping_supported(family, icmph->type, icmph->code))
502 		return;
503 
504 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
505 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
506 		 ntohs(icmph->un.echo.sequence));
507 
508 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
509 	if (sk == NULL) {
510 		pr_debug("no socket, dropping\n");
511 		return;	/* No socket for error */
512 	}
513 	pr_debug("err on socket %p\n", sk);
514 
515 	err = 0;
516 	harderr = 0;
517 	inet_sock = inet_sk(sk);
518 
519 	if (skb->protocol == htons(ETH_P_IP)) {
520 		switch (type) {
521 		default:
522 		case ICMP_TIME_EXCEEDED:
523 			err = EHOSTUNREACH;
524 			break;
525 		case ICMP_SOURCE_QUENCH:
526 			/* This is not a real error but ping wants to see it.
527 			 * Report it with some fake errno.
528 			 */
529 			err = EREMOTEIO;
530 			break;
531 		case ICMP_PARAMETERPROB:
532 			err = EPROTO;
533 			harderr = 1;
534 			break;
535 		case ICMP_DEST_UNREACH:
536 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
537 				ipv4_sk_update_pmtu(skb, sk, info);
538 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
539 					err = EMSGSIZE;
540 					harderr = 1;
541 					break;
542 				}
543 				goto out;
544 			}
545 			err = EHOSTUNREACH;
546 			if (code <= NR_ICMP_UNREACH) {
547 				harderr = icmp_err_convert[code].fatal;
548 				err = icmp_err_convert[code].errno;
549 			}
550 			break;
551 		case ICMP_REDIRECT:
552 			/* See ICMP_SOURCE_QUENCH */
553 			ipv4_sk_redirect(skb, sk);
554 			err = EREMOTEIO;
555 			break;
556 		}
557 #if IS_ENABLED(CONFIG_IPV6)
558 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
559 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
560 #endif
561 	}
562 
563 	/*
564 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
565 	 *	4.1.3.3.
566 	 */
567 	if ((family == AF_INET && !inet_sock->recverr) ||
568 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
569 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
570 			goto out;
571 	} else {
572 		if (family == AF_INET) {
573 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
574 				      info, (u8 *)icmph);
575 #if IS_ENABLED(CONFIG_IPV6)
576 		} else if (family == AF_INET6) {
577 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
578 						   info, (u8 *)icmph);
579 #endif
580 		}
581 	}
582 	sk->sk_err = err;
583 	sk->sk_error_report(sk);
584 out:
585 	sock_put(sk);
586 }
587 EXPORT_SYMBOL_GPL(ping_err);
588 
589 /*
590  *	Copy and checksum an ICMP Echo packet from user space into a buffer
591  *	starting from the payload.
592  */
593 
594 int ping_getfrag(void *from, char *to,
595 		 int offset, int fraglen, int odd, struct sk_buff *skb)
596 {
597 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
598 
599 	if (offset == 0) {
600 		if (fraglen < sizeof(struct icmphdr))
601 			BUG();
602 		if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
603 			    pfh->iov, 0, fraglen - sizeof(struct icmphdr),
604 			    &pfh->wcheck))
605 			return -EFAULT;
606 	} else if (offset < sizeof(struct icmphdr)) {
607 			BUG();
608 	} else {
609 		if (csum_partial_copy_fromiovecend
610 				(to, pfh->iov, offset - sizeof(struct icmphdr),
611 				 fraglen, &pfh->wcheck))
612 			return -EFAULT;
613 	}
614 
615 #if IS_ENABLED(CONFIG_IPV6)
616 	/* For IPv6, checksum each skb as we go along, as expected by
617 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
618 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
619 	 */
620 	if (pfh->family == AF_INET6) {
621 		skb->csum = pfh->wcheck;
622 		skb->ip_summed = CHECKSUM_NONE;
623 		pfh->wcheck = 0;
624 	}
625 #endif
626 
627 	return 0;
628 }
629 EXPORT_SYMBOL_GPL(ping_getfrag);
630 
631 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
632 				       struct flowi4 *fl4)
633 {
634 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
635 
636 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
637 		sizeof(struct icmphdr), pfh->wcheck);
638 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
639 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
640 	skb->ip_summed = CHECKSUM_NONE;
641 	return ip_push_pending_frames(sk, fl4);
642 }
643 
644 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
645 			void *user_icmph, size_t icmph_len) {
646 	u8 type, code;
647 
648 	if (len > 0xFFFF)
649 		return -EMSGSIZE;
650 
651 	/*
652 	 *	Check the flags.
653 	 */
654 
655 	/* Mirror BSD error message compatibility */
656 	if (msg->msg_flags & MSG_OOB)
657 		return -EOPNOTSUPP;
658 
659 	/*
660 	 *	Fetch the ICMP header provided by the userland.
661 	 *	iovec is modified! The ICMP header is consumed.
662 	 */
663 	if (memcpy_fromiovec(user_icmph, msg->msg_iov, icmph_len))
664 		return -EFAULT;
665 
666 	if (family == AF_INET) {
667 		type = ((struct icmphdr *) user_icmph)->type;
668 		code = ((struct icmphdr *) user_icmph)->code;
669 #if IS_ENABLED(CONFIG_IPV6)
670 	} else if (family == AF_INET6) {
671 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
672 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
673 #endif
674 	} else {
675 		BUG();
676 	}
677 
678 	if (!ping_supported(family, type, code))
679 		return -EINVAL;
680 
681 	return 0;
682 }
683 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
684 
685 static int ping_v4_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
686 			   size_t len)
687 {
688 	struct net *net = sock_net(sk);
689 	struct flowi4 fl4;
690 	struct inet_sock *inet = inet_sk(sk);
691 	struct ipcm_cookie ipc;
692 	struct icmphdr user_icmph;
693 	struct pingfakehdr pfh;
694 	struct rtable *rt = NULL;
695 	struct ip_options_data opt_copy;
696 	int free = 0;
697 	__be32 saddr, daddr, faddr;
698 	u8  tos;
699 	int err;
700 
701 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
702 
703 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
704 				  sizeof(user_icmph));
705 	if (err)
706 		return err;
707 
708 	/*
709 	 *	Get and verify the address.
710 	 */
711 
712 	if (msg->msg_name) {
713 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
714 		if (msg->msg_namelen < sizeof(*usin))
715 			return -EINVAL;
716 		if (usin->sin_family != AF_INET)
717 			return -EINVAL;
718 		daddr = usin->sin_addr.s_addr;
719 		/* no remote port */
720 	} else {
721 		if (sk->sk_state != TCP_ESTABLISHED)
722 			return -EDESTADDRREQ;
723 		daddr = inet->inet_daddr;
724 		/* no remote port */
725 	}
726 
727 	ipc.addr = inet->inet_saddr;
728 	ipc.opt = NULL;
729 	ipc.oif = sk->sk_bound_dev_if;
730 	ipc.tx_flags = 0;
731 	ipc.ttl = 0;
732 	ipc.tos = -1;
733 
734 	sock_tx_timestamp(sk, &ipc.tx_flags);
735 
736 	if (msg->msg_controllen) {
737 		err = ip_cmsg_send(sock_net(sk), msg, &ipc, false);
738 		if (err)
739 			return err;
740 		if (ipc.opt)
741 			free = 1;
742 	}
743 	if (!ipc.opt) {
744 		struct ip_options_rcu *inet_opt;
745 
746 		rcu_read_lock();
747 		inet_opt = rcu_dereference(inet->inet_opt);
748 		if (inet_opt) {
749 			memcpy(&opt_copy, inet_opt,
750 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
751 			ipc.opt = &opt_copy.opt;
752 		}
753 		rcu_read_unlock();
754 	}
755 
756 	saddr = ipc.addr;
757 	ipc.addr = faddr = daddr;
758 
759 	if (ipc.opt && ipc.opt->opt.srr) {
760 		if (!daddr)
761 			return -EINVAL;
762 		faddr = ipc.opt->opt.faddr;
763 	}
764 	tos = get_rttos(&ipc, inet);
765 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
766 	    (msg->msg_flags & MSG_DONTROUTE) ||
767 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
768 		tos |= RTO_ONLINK;
769 	}
770 
771 	if (ipv4_is_multicast(daddr)) {
772 		if (!ipc.oif)
773 			ipc.oif = inet->mc_index;
774 		if (!saddr)
775 			saddr = inet->mc_addr;
776 	} else if (!ipc.oif)
777 		ipc.oif = inet->uc_index;
778 
779 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
780 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
781 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
782 
783 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
784 	rt = ip_route_output_flow(net, &fl4, sk);
785 	if (IS_ERR(rt)) {
786 		err = PTR_ERR(rt);
787 		rt = NULL;
788 		if (err == -ENETUNREACH)
789 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
790 		goto out;
791 	}
792 
793 	err = -EACCES;
794 	if ((rt->rt_flags & RTCF_BROADCAST) &&
795 	    !sock_flag(sk, SOCK_BROADCAST))
796 		goto out;
797 
798 	if (msg->msg_flags & MSG_CONFIRM)
799 		goto do_confirm;
800 back_from_confirm:
801 
802 	if (!ipc.addr)
803 		ipc.addr = fl4.daddr;
804 
805 	lock_sock(sk);
806 
807 	pfh.icmph.type = user_icmph.type; /* already checked */
808 	pfh.icmph.code = user_icmph.code; /* ditto */
809 	pfh.icmph.checksum = 0;
810 	pfh.icmph.un.echo.id = inet->inet_sport;
811 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
812 	pfh.iov = msg->msg_iov;
813 	pfh.wcheck = 0;
814 	pfh.family = AF_INET;
815 
816 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
817 			0, &ipc, &rt, msg->msg_flags);
818 	if (err)
819 		ip_flush_pending_frames(sk);
820 	else
821 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
822 	release_sock(sk);
823 
824 out:
825 	ip_rt_put(rt);
826 	if (free)
827 		kfree(ipc.opt);
828 	if (!err) {
829 		icmp_out_count(sock_net(sk), user_icmph.type);
830 		return len;
831 	}
832 	return err;
833 
834 do_confirm:
835 	dst_confirm(&rt->dst);
836 	if (!(msg->msg_flags & MSG_PROBE) || len)
837 		goto back_from_confirm;
838 	err = 0;
839 	goto out;
840 }
841 
842 int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
843 		 size_t len, int noblock, int flags, int *addr_len)
844 {
845 	struct inet_sock *isk = inet_sk(sk);
846 	int family = sk->sk_family;
847 	struct sk_buff *skb;
848 	int copied, err;
849 
850 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
851 
852 	err = -EOPNOTSUPP;
853 	if (flags & MSG_OOB)
854 		goto out;
855 
856 	if (flags & MSG_ERRQUEUE) {
857 		if (family == AF_INET) {
858 			return ip_recv_error(sk, msg, len, addr_len);
859 #if IS_ENABLED(CONFIG_IPV6)
860 		} else if (family == AF_INET6) {
861 			return pingv6_ops.ipv6_recv_error(sk, msg, len,
862 							  addr_len);
863 #endif
864 		}
865 	}
866 
867 	skb = skb_recv_datagram(sk, flags, noblock, &err);
868 	if (!skb)
869 		goto out;
870 
871 	copied = skb->len;
872 	if (copied > len) {
873 		msg->msg_flags |= MSG_TRUNC;
874 		copied = len;
875 	}
876 
877 	/* Don't bother checking the checksum */
878 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
879 	if (err)
880 		goto done;
881 
882 	sock_recv_timestamp(msg, sk, skb);
883 
884 	/* Copy the address and add cmsg data. */
885 	if (family == AF_INET) {
886 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
887 
888 		if (sin) {
889 			sin->sin_family = AF_INET;
890 			sin->sin_port = 0 /* skb->h.uh->source */;
891 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
892 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
893 			*addr_len = sizeof(*sin);
894 		}
895 
896 		if (isk->cmsg_flags)
897 			ip_cmsg_recv(msg, skb);
898 
899 #if IS_ENABLED(CONFIG_IPV6)
900 	} else if (family == AF_INET6) {
901 		struct ipv6_pinfo *np = inet6_sk(sk);
902 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
903 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
904 
905 		if (sin6) {
906 			sin6->sin6_family = AF_INET6;
907 			sin6->sin6_port = 0;
908 			sin6->sin6_addr = ip6->saddr;
909 			sin6->sin6_flowinfo = 0;
910 			if (np->sndflow)
911 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
912 			sin6->sin6_scope_id =
913 				ipv6_iface_scope_id(&sin6->sin6_addr,
914 						    IP6CB(skb)->iif);
915 			*addr_len = sizeof(*sin6);
916 		}
917 
918 		if (inet6_sk(sk)->rxopt.all)
919 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
920 		if (skb->protocol == htons(ETH_P_IPV6) &&
921 		    inet6_sk(sk)->rxopt.all)
922 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
923 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
924 			ip_cmsg_recv(msg, skb);
925 #endif
926 	} else {
927 		BUG();
928 	}
929 
930 	err = copied;
931 
932 done:
933 	skb_free_datagram(sk, skb);
934 out:
935 	pr_debug("ping_recvmsg -> %d\n", err);
936 	return err;
937 }
938 EXPORT_SYMBOL_GPL(ping_recvmsg);
939 
940 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
941 {
942 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
943 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
944 	if (sock_queue_rcv_skb(sk, skb) < 0) {
945 		kfree_skb(skb);
946 		pr_debug("ping_queue_rcv_skb -> failed\n");
947 		return -1;
948 	}
949 	return 0;
950 }
951 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
952 
953 
954 /*
955  *	All we need to do is get the socket.
956  */
957 
958 void ping_rcv(struct sk_buff *skb)
959 {
960 	struct sock *sk;
961 	struct net *net = dev_net(skb->dev);
962 	struct icmphdr *icmph = icmp_hdr(skb);
963 
964 	/* We assume the packet has already been checked by icmp_rcv */
965 
966 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
967 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
968 
969 	/* Push ICMP header back */
970 	skb_push(skb, skb->data - (u8 *)icmph);
971 
972 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
973 	if (sk != NULL) {
974 		pr_debug("rcv on socket %p\n", sk);
975 		ping_queue_rcv_skb(sk, skb_get(skb));
976 		sock_put(sk);
977 		return;
978 	}
979 	pr_debug("no socket, dropping\n");
980 
981 	/* We're called from icmp_rcv(). kfree_skb() is done there. */
982 }
983 EXPORT_SYMBOL_GPL(ping_rcv);
984 
985 struct proto ping_prot = {
986 	.name =		"PING",
987 	.owner =	THIS_MODULE,
988 	.init =		ping_init_sock,
989 	.close =	ping_close,
990 	.connect =	ip4_datagram_connect,
991 	.disconnect =	udp_disconnect,
992 	.setsockopt =	ip_setsockopt,
993 	.getsockopt =	ip_getsockopt,
994 	.sendmsg =	ping_v4_sendmsg,
995 	.recvmsg =	ping_recvmsg,
996 	.bind =		ping_bind,
997 	.backlog_rcv =	ping_queue_rcv_skb,
998 	.release_cb =	ip4_datagram_release_cb,
999 	.hash =		ping_hash,
1000 	.unhash =	ping_unhash,
1001 	.get_port =	ping_get_port,
1002 	.obj_size =	sizeof(struct inet_sock),
1003 };
1004 EXPORT_SYMBOL(ping_prot);
1005 
1006 #ifdef CONFIG_PROC_FS
1007 
1008 static struct sock *ping_get_first(struct seq_file *seq, int start)
1009 {
1010 	struct sock *sk;
1011 	struct ping_iter_state *state = seq->private;
1012 	struct net *net = seq_file_net(seq);
1013 
1014 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1015 	     ++state->bucket) {
1016 		struct hlist_nulls_node *node;
1017 		struct hlist_nulls_head *hslot;
1018 
1019 		hslot = &ping_table.hash[state->bucket];
1020 
1021 		if (hlist_nulls_empty(hslot))
1022 			continue;
1023 
1024 		sk_nulls_for_each(sk, node, hslot) {
1025 			if (net_eq(sock_net(sk), net) &&
1026 			    sk->sk_family == state->family)
1027 				goto found;
1028 		}
1029 	}
1030 	sk = NULL;
1031 found:
1032 	return sk;
1033 }
1034 
1035 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1036 {
1037 	struct ping_iter_state *state = seq->private;
1038 	struct net *net = seq_file_net(seq);
1039 
1040 	do {
1041 		sk = sk_nulls_next(sk);
1042 	} while (sk && (!net_eq(sock_net(sk), net)));
1043 
1044 	if (!sk)
1045 		return ping_get_first(seq, state->bucket + 1);
1046 	return sk;
1047 }
1048 
1049 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1050 {
1051 	struct sock *sk = ping_get_first(seq, 0);
1052 
1053 	if (sk)
1054 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1055 			--pos;
1056 	return pos ? NULL : sk;
1057 }
1058 
1059 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1060 {
1061 	struct ping_iter_state *state = seq->private;
1062 	state->bucket = 0;
1063 	state->family = family;
1064 
1065 	read_lock_bh(&ping_table.lock);
1066 
1067 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1068 }
1069 EXPORT_SYMBOL_GPL(ping_seq_start);
1070 
1071 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1072 {
1073 	return ping_seq_start(seq, pos, AF_INET);
1074 }
1075 
1076 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1077 {
1078 	struct sock *sk;
1079 
1080 	if (v == SEQ_START_TOKEN)
1081 		sk = ping_get_idx(seq, 0);
1082 	else
1083 		sk = ping_get_next(seq, v);
1084 
1085 	++*pos;
1086 	return sk;
1087 }
1088 EXPORT_SYMBOL_GPL(ping_seq_next);
1089 
1090 void ping_seq_stop(struct seq_file *seq, void *v)
1091 {
1092 	read_unlock_bh(&ping_table.lock);
1093 }
1094 EXPORT_SYMBOL_GPL(ping_seq_stop);
1095 
1096 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1097 		int bucket)
1098 {
1099 	struct inet_sock *inet = inet_sk(sp);
1100 	__be32 dest = inet->inet_daddr;
1101 	__be32 src = inet->inet_rcv_saddr;
1102 	__u16 destp = ntohs(inet->inet_dport);
1103 	__u16 srcp = ntohs(inet->inet_sport);
1104 
1105 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1106 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1107 		bucket, src, srcp, dest, destp, sp->sk_state,
1108 		sk_wmem_alloc_get(sp),
1109 		sk_rmem_alloc_get(sp),
1110 		0, 0L, 0,
1111 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1112 		0, sock_i_ino(sp),
1113 		atomic_read(&sp->sk_refcnt), sp,
1114 		atomic_read(&sp->sk_drops));
1115 }
1116 
1117 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1118 {
1119 	seq_setwidth(seq, 127);
1120 	if (v == SEQ_START_TOKEN)
1121 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1122 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1123 			   "inode ref pointer drops");
1124 	else {
1125 		struct ping_iter_state *state = seq->private;
1126 
1127 		ping_v4_format_sock(v, seq, state->bucket);
1128 	}
1129 	seq_pad(seq, '\n');
1130 	return 0;
1131 }
1132 
1133 static const struct seq_operations ping_v4_seq_ops = {
1134 	.show		= ping_v4_seq_show,
1135 	.start		= ping_v4_seq_start,
1136 	.next		= ping_seq_next,
1137 	.stop		= ping_seq_stop,
1138 };
1139 
1140 static int ping_seq_open(struct inode *inode, struct file *file)
1141 {
1142 	struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1143 	return seq_open_net(inode, file, &afinfo->seq_ops,
1144 			   sizeof(struct ping_iter_state));
1145 }
1146 
1147 const struct file_operations ping_seq_fops = {
1148 	.open		= ping_seq_open,
1149 	.read		= seq_read,
1150 	.llseek		= seq_lseek,
1151 	.release	= seq_release_net,
1152 };
1153 EXPORT_SYMBOL_GPL(ping_seq_fops);
1154 
1155 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1156 	.name		= "icmp",
1157 	.family		= AF_INET,
1158 	.seq_fops	= &ping_seq_fops,
1159 	.seq_ops	= {
1160 		.start		= ping_v4_seq_start,
1161 		.show		= ping_v4_seq_show,
1162 		.next		= ping_seq_next,
1163 		.stop		= ping_seq_stop,
1164 	},
1165 };
1166 
1167 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1168 {
1169 	struct proc_dir_entry *p;
1170 	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1171 			     afinfo->seq_fops, afinfo);
1172 	if (!p)
1173 		return -ENOMEM;
1174 	return 0;
1175 }
1176 EXPORT_SYMBOL_GPL(ping_proc_register);
1177 
1178 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1179 {
1180 	remove_proc_entry(afinfo->name, net->proc_net);
1181 }
1182 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1183 
1184 static int __net_init ping_v4_proc_init_net(struct net *net)
1185 {
1186 	return ping_proc_register(net, &ping_v4_seq_afinfo);
1187 }
1188 
1189 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1190 {
1191 	ping_proc_unregister(net, &ping_v4_seq_afinfo);
1192 }
1193 
1194 static struct pernet_operations ping_v4_net_ops = {
1195 	.init = ping_v4_proc_init_net,
1196 	.exit = ping_v4_proc_exit_net,
1197 };
1198 
1199 int __init ping_proc_init(void)
1200 {
1201 	return register_pernet_subsys(&ping_v4_net_ops);
1202 }
1203 
1204 void ping_proc_exit(void)
1205 {
1206 	unregister_pernet_subsys(&ping_v4_net_ops);
1207 }
1208 
1209 #endif
1210 
1211 void __init ping_init(void)
1212 {
1213 	int i;
1214 
1215 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1216 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1217 	rwlock_init(&ping_table.lock);
1218 }
1219