xref: /openbmc/linux/net/ipv4/ping.c (revision 8e9356c6)
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.sysctl_ping_group_range;
240 	unsigned int seq;
241 
242 	do {
243 		seq = read_seqbegin(&net->ipv4.sysctl_local_ports.lock);
244 
245 		*low = data[0];
246 		*high = data[1];
247 	} while (read_seqretry(&net->ipv4.sysctl_local_ports.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 = get_current_groups();
256 	int i, j, count = group_info->ngroups;
257 	kgid_t low, high;
258 
259 	inet_get_ping_group_range_net(net, &low, &high);
260 	if (gid_lte(low, group) && gid_lte(group, high))
261 		return 0;
262 
263 	for (i = 0; i < group_info->nblocks; i++) {
264 		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
265 		for (j = 0; j < cp_count; j++) {
266 			kgid_t gid = group_info->blocks[i][j];
267 			if (gid_lte(low, gid) && gid_lte(gid, high))
268 				return 0;
269 		}
270 
271 		count -= cp_count;
272 	}
273 
274 	return -EACCES;
275 }
276 EXPORT_SYMBOL_GPL(ping_init_sock);
277 
278 void ping_close(struct sock *sk, long timeout)
279 {
280 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
281 		 inet_sk(sk), inet_sk(sk)->inet_num);
282 	pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
283 
284 	sk_common_release(sk);
285 }
286 EXPORT_SYMBOL_GPL(ping_close);
287 
288 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
289 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
290 				struct sockaddr *uaddr, int addr_len) {
291 	struct net *net = sock_net(sk);
292 	if (sk->sk_family == AF_INET) {
293 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
294 		int chk_addr_ret;
295 
296 		if (addr_len < sizeof(*addr))
297 			return -EINVAL;
298 
299 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
300 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
301 
302 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
303 
304 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
305 			chk_addr_ret = RTN_LOCAL;
306 
307 		if ((sysctl_ip_nonlocal_bind == 0 &&
308 		    isk->freebind == 0 && isk->transparent == 0 &&
309 		     chk_addr_ret != RTN_LOCAL) ||
310 		    chk_addr_ret == RTN_MULTICAST ||
311 		    chk_addr_ret == RTN_BROADCAST)
312 			return -EADDRNOTAVAIL;
313 
314 #if IS_ENABLED(CONFIG_IPV6)
315 	} else if (sk->sk_family == AF_INET6) {
316 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
317 		int addr_type, scoped, has_addr;
318 		struct net_device *dev = NULL;
319 
320 		if (addr_len < sizeof(*addr))
321 			return -EINVAL;
322 
323 		if (addr->sin6_family != AF_INET6)
324 			return -EINVAL;
325 
326 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
327 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
328 
329 		addr_type = ipv6_addr_type(&addr->sin6_addr);
330 		scoped = __ipv6_addr_needs_scope_id(addr_type);
331 		if ((addr_type != IPV6_ADDR_ANY &&
332 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
333 		    (scoped && !addr->sin6_scope_id))
334 			return -EINVAL;
335 
336 		rcu_read_lock();
337 		if (addr->sin6_scope_id) {
338 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
339 			if (!dev) {
340 				rcu_read_unlock();
341 				return -ENODEV;
342 			}
343 		}
344 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
345 						    scoped);
346 		rcu_read_unlock();
347 
348 		if (!(isk->freebind || isk->transparent || has_addr ||
349 		      addr_type == IPV6_ADDR_ANY))
350 			return -EADDRNOTAVAIL;
351 
352 		if (scoped)
353 			sk->sk_bound_dev_if = addr->sin6_scope_id;
354 #endif
355 	} else {
356 		return -EAFNOSUPPORT;
357 	}
358 	return 0;
359 }
360 
361 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
362 {
363 	if (saddr->sa_family == AF_INET) {
364 		struct inet_sock *isk = inet_sk(sk);
365 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
366 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
367 #if IS_ENABLED(CONFIG_IPV6)
368 	} else if (saddr->sa_family == AF_INET6) {
369 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
370 		struct ipv6_pinfo *np = inet6_sk(sk);
371 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
372 #endif
373 	}
374 }
375 
376 static void ping_clear_saddr(struct sock *sk, int dif)
377 {
378 	sk->sk_bound_dev_if = dif;
379 	if (sk->sk_family == AF_INET) {
380 		struct inet_sock *isk = inet_sk(sk);
381 		isk->inet_rcv_saddr = isk->inet_saddr = 0;
382 #if IS_ENABLED(CONFIG_IPV6)
383 	} else if (sk->sk_family == AF_INET6) {
384 		struct ipv6_pinfo *np = inet6_sk(sk);
385 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
386 		memset(&np->saddr, 0, sizeof(np->saddr));
387 #endif
388 	}
389 }
390 /*
391  * We need our own bind because there are no privileged id's == local ports.
392  * Moreover, we don't allow binding to multi- and broadcast addresses.
393  */
394 
395 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
396 {
397 	struct inet_sock *isk = inet_sk(sk);
398 	unsigned short snum;
399 	int err;
400 	int dif = sk->sk_bound_dev_if;
401 
402 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
403 	if (err)
404 		return err;
405 
406 	lock_sock(sk);
407 
408 	err = -EINVAL;
409 	if (isk->inet_num != 0)
410 		goto out;
411 
412 	err = -EADDRINUSE;
413 	ping_set_saddr(sk, uaddr);
414 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
415 	if (ping_get_port(sk, snum) != 0) {
416 		ping_clear_saddr(sk, dif);
417 		goto out;
418 	}
419 
420 	pr_debug("after bind(): num = %d, dif = %d\n",
421 		 (int)isk->inet_num,
422 		 (int)sk->sk_bound_dev_if);
423 
424 	err = 0;
425 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
426 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
427 #if IS_ENABLED(CONFIG_IPV6)
428 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
429 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
430 #endif
431 
432 	if (snum)
433 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
434 	isk->inet_sport = htons(isk->inet_num);
435 	isk->inet_daddr = 0;
436 	isk->inet_dport = 0;
437 
438 #if IS_ENABLED(CONFIG_IPV6)
439 	if (sk->sk_family == AF_INET6)
440 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
441 #endif
442 
443 	sk_dst_reset(sk);
444 out:
445 	release_sock(sk);
446 	pr_debug("ping_v4_bind -> %d\n", err);
447 	return err;
448 }
449 EXPORT_SYMBOL_GPL(ping_bind);
450 
451 /*
452  * Is this a supported type of ICMP message?
453  */
454 
455 static inline int ping_supported(int family, int type, int code)
456 {
457 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
458 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
459 }
460 
461 /*
462  * This routine is called by the ICMP module when it gets some
463  * sort of error condition.
464  */
465 
466 void ping_err(struct sk_buff *skb, int offset, u32 info)
467 {
468 	int family;
469 	struct icmphdr *icmph;
470 	struct inet_sock *inet_sock;
471 	int type;
472 	int code;
473 	struct net *net = dev_net(skb->dev);
474 	struct sock *sk;
475 	int harderr;
476 	int err;
477 
478 	if (skb->protocol == htons(ETH_P_IP)) {
479 		family = AF_INET;
480 		type = icmp_hdr(skb)->type;
481 		code = icmp_hdr(skb)->code;
482 		icmph = (struct icmphdr *)(skb->data + offset);
483 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
484 		family = AF_INET6;
485 		type = icmp6_hdr(skb)->icmp6_type;
486 		code = icmp6_hdr(skb)->icmp6_code;
487 		icmph = (struct icmphdr *) (skb->data + offset);
488 	} else {
489 		BUG();
490 	}
491 
492 	/* We assume the packet has already been checked by icmp_unreach */
493 
494 	if (!ping_supported(family, icmph->type, icmph->code))
495 		return;
496 
497 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
498 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
499 		 ntohs(icmph->un.echo.sequence));
500 
501 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
502 	if (sk == NULL) {
503 		pr_debug("no socket, dropping\n");
504 		return;	/* No socket for error */
505 	}
506 	pr_debug("err on socket %p\n", sk);
507 
508 	err = 0;
509 	harderr = 0;
510 	inet_sock = inet_sk(sk);
511 
512 	if (skb->protocol == htons(ETH_P_IP)) {
513 		switch (type) {
514 		default:
515 		case ICMP_TIME_EXCEEDED:
516 			err = EHOSTUNREACH;
517 			break;
518 		case ICMP_SOURCE_QUENCH:
519 			/* This is not a real error but ping wants to see it.
520 			 * Report it with some fake errno.
521 			 */
522 			err = EREMOTEIO;
523 			break;
524 		case ICMP_PARAMETERPROB:
525 			err = EPROTO;
526 			harderr = 1;
527 			break;
528 		case ICMP_DEST_UNREACH:
529 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
530 				ipv4_sk_update_pmtu(skb, sk, info);
531 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
532 					err = EMSGSIZE;
533 					harderr = 1;
534 					break;
535 				}
536 				goto out;
537 			}
538 			err = EHOSTUNREACH;
539 			if (code <= NR_ICMP_UNREACH) {
540 				harderr = icmp_err_convert[code].fatal;
541 				err = icmp_err_convert[code].errno;
542 			}
543 			break;
544 		case ICMP_REDIRECT:
545 			/* See ICMP_SOURCE_QUENCH */
546 			ipv4_sk_redirect(skb, sk);
547 			err = EREMOTEIO;
548 			break;
549 		}
550 #if IS_ENABLED(CONFIG_IPV6)
551 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
552 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
553 #endif
554 	}
555 
556 	/*
557 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
558 	 *	4.1.3.3.
559 	 */
560 	if ((family == AF_INET && !inet_sock->recverr) ||
561 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
562 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
563 			goto out;
564 	} else {
565 		if (family == AF_INET) {
566 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
567 				      info, (u8 *)icmph);
568 #if IS_ENABLED(CONFIG_IPV6)
569 		} else if (family == AF_INET6) {
570 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
571 						   info, (u8 *)icmph);
572 #endif
573 		}
574 	}
575 	sk->sk_err = err;
576 	sk->sk_error_report(sk);
577 out:
578 	sock_put(sk);
579 }
580 EXPORT_SYMBOL_GPL(ping_err);
581 
582 /*
583  *	Copy and checksum an ICMP Echo packet from user space into a buffer
584  *	starting from the payload.
585  */
586 
587 int ping_getfrag(void *from, char *to,
588 		 int offset, int fraglen, int odd, struct sk_buff *skb)
589 {
590 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
591 
592 	if (offset == 0) {
593 		if (fraglen < sizeof(struct icmphdr))
594 			BUG();
595 		if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
596 			    pfh->iov, 0, fraglen - sizeof(struct icmphdr),
597 			    &pfh->wcheck))
598 			return -EFAULT;
599 	} else if (offset < sizeof(struct icmphdr)) {
600 			BUG();
601 	} else {
602 		if (csum_partial_copy_fromiovecend
603 				(to, pfh->iov, offset - sizeof(struct icmphdr),
604 				 fraglen, &pfh->wcheck))
605 			return -EFAULT;
606 	}
607 
608 #if IS_ENABLED(CONFIG_IPV6)
609 	/* For IPv6, checksum each skb as we go along, as expected by
610 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
611 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
612 	 */
613 	if (pfh->family == AF_INET6) {
614 		skb->csum = pfh->wcheck;
615 		skb->ip_summed = CHECKSUM_NONE;
616 		pfh->wcheck = 0;
617 	}
618 #endif
619 
620 	return 0;
621 }
622 EXPORT_SYMBOL_GPL(ping_getfrag);
623 
624 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
625 				       struct flowi4 *fl4)
626 {
627 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
628 
629 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
630 		sizeof(struct icmphdr), pfh->wcheck);
631 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
632 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
633 	skb->ip_summed = CHECKSUM_NONE;
634 	return ip_push_pending_frames(sk, fl4);
635 }
636 
637 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
638 			void *user_icmph, size_t icmph_len) {
639 	u8 type, code;
640 
641 	if (len > 0xFFFF)
642 		return -EMSGSIZE;
643 
644 	/*
645 	 *	Check the flags.
646 	 */
647 
648 	/* Mirror BSD error message compatibility */
649 	if (msg->msg_flags & MSG_OOB)
650 		return -EOPNOTSUPP;
651 
652 	/*
653 	 *	Fetch the ICMP header provided by the userland.
654 	 *	iovec is modified! The ICMP header is consumed.
655 	 */
656 	if (memcpy_fromiovec(user_icmph, msg->msg_iov, icmph_len))
657 		return -EFAULT;
658 
659 	if (family == AF_INET) {
660 		type = ((struct icmphdr *) user_icmph)->type;
661 		code = ((struct icmphdr *) user_icmph)->code;
662 #if IS_ENABLED(CONFIG_IPV6)
663 	} else if (family == AF_INET6) {
664 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
665 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
666 #endif
667 	} else {
668 		BUG();
669 	}
670 
671 	if (!ping_supported(family, type, code))
672 		return -EINVAL;
673 
674 	return 0;
675 }
676 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
677 
678 static int ping_v4_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
679 			   size_t len)
680 {
681 	struct net *net = sock_net(sk);
682 	struct flowi4 fl4;
683 	struct inet_sock *inet = inet_sk(sk);
684 	struct ipcm_cookie ipc;
685 	struct icmphdr user_icmph;
686 	struct pingfakehdr pfh;
687 	struct rtable *rt = NULL;
688 	struct ip_options_data opt_copy;
689 	int free = 0;
690 	__be32 saddr, daddr, faddr;
691 	u8  tos;
692 	int err;
693 
694 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
695 
696 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
697 				  sizeof(user_icmph));
698 	if (err)
699 		return err;
700 
701 	/*
702 	 *	Get and verify the address.
703 	 */
704 
705 	if (msg->msg_name) {
706 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
707 		if (msg->msg_namelen < sizeof(*usin))
708 			return -EINVAL;
709 		if (usin->sin_family != AF_INET)
710 			return -EINVAL;
711 		daddr = usin->sin_addr.s_addr;
712 		/* no remote port */
713 	} else {
714 		if (sk->sk_state != TCP_ESTABLISHED)
715 			return -EDESTADDRREQ;
716 		daddr = inet->inet_daddr;
717 		/* no remote port */
718 	}
719 
720 	ipc.addr = inet->inet_saddr;
721 	ipc.opt = NULL;
722 	ipc.oif = sk->sk_bound_dev_if;
723 	ipc.tx_flags = 0;
724 	ipc.ttl = 0;
725 	ipc.tos = -1;
726 
727 	sock_tx_timestamp(sk, &ipc.tx_flags);
728 
729 	if (msg->msg_controllen) {
730 		err = ip_cmsg_send(sock_net(sk), msg, &ipc);
731 		if (err)
732 			return err;
733 		if (ipc.opt)
734 			free = 1;
735 	}
736 	if (!ipc.opt) {
737 		struct ip_options_rcu *inet_opt;
738 
739 		rcu_read_lock();
740 		inet_opt = rcu_dereference(inet->inet_opt);
741 		if (inet_opt) {
742 			memcpy(&opt_copy, inet_opt,
743 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
744 			ipc.opt = &opt_copy.opt;
745 		}
746 		rcu_read_unlock();
747 	}
748 
749 	saddr = ipc.addr;
750 	ipc.addr = faddr = daddr;
751 
752 	if (ipc.opt && ipc.opt->opt.srr) {
753 		if (!daddr)
754 			return -EINVAL;
755 		faddr = ipc.opt->opt.faddr;
756 	}
757 	tos = get_rttos(&ipc, inet);
758 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
759 	    (msg->msg_flags & MSG_DONTROUTE) ||
760 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
761 		tos |= RTO_ONLINK;
762 	}
763 
764 	if (ipv4_is_multicast(daddr)) {
765 		if (!ipc.oif)
766 			ipc.oif = inet->mc_index;
767 		if (!saddr)
768 			saddr = inet->mc_addr;
769 	} else if (!ipc.oif)
770 		ipc.oif = inet->uc_index;
771 
772 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
773 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
774 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
775 
776 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
777 	rt = ip_route_output_flow(net, &fl4, sk);
778 	if (IS_ERR(rt)) {
779 		err = PTR_ERR(rt);
780 		rt = NULL;
781 		if (err == -ENETUNREACH)
782 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
783 		goto out;
784 	}
785 
786 	err = -EACCES;
787 	if ((rt->rt_flags & RTCF_BROADCAST) &&
788 	    !sock_flag(sk, SOCK_BROADCAST))
789 		goto out;
790 
791 	if (msg->msg_flags & MSG_CONFIRM)
792 		goto do_confirm;
793 back_from_confirm:
794 
795 	if (!ipc.addr)
796 		ipc.addr = fl4.daddr;
797 
798 	lock_sock(sk);
799 
800 	pfh.icmph.type = user_icmph.type; /* already checked */
801 	pfh.icmph.code = user_icmph.code; /* ditto */
802 	pfh.icmph.checksum = 0;
803 	pfh.icmph.un.echo.id = inet->inet_sport;
804 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
805 	pfh.iov = msg->msg_iov;
806 	pfh.wcheck = 0;
807 	pfh.family = AF_INET;
808 
809 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
810 			0, &ipc, &rt, msg->msg_flags);
811 	if (err)
812 		ip_flush_pending_frames(sk);
813 	else
814 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
815 	release_sock(sk);
816 
817 out:
818 	ip_rt_put(rt);
819 	if (free)
820 		kfree(ipc.opt);
821 	if (!err) {
822 		icmp_out_count(sock_net(sk), user_icmph.type);
823 		return len;
824 	}
825 	return err;
826 
827 do_confirm:
828 	dst_confirm(&rt->dst);
829 	if (!(msg->msg_flags & MSG_PROBE) || len)
830 		goto back_from_confirm;
831 	err = 0;
832 	goto out;
833 }
834 
835 int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
836 		 size_t len, int noblock, int flags, int *addr_len)
837 {
838 	struct inet_sock *isk = inet_sk(sk);
839 	int family = sk->sk_family;
840 	struct sk_buff *skb;
841 	int copied, err;
842 
843 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
844 
845 	err = -EOPNOTSUPP;
846 	if (flags & MSG_OOB)
847 		goto out;
848 
849 	if (flags & MSG_ERRQUEUE) {
850 		if (family == AF_INET) {
851 			return ip_recv_error(sk, msg, len, addr_len);
852 #if IS_ENABLED(CONFIG_IPV6)
853 		} else if (family == AF_INET6) {
854 			return pingv6_ops.ipv6_recv_error(sk, msg, len,
855 							  addr_len);
856 #endif
857 		}
858 	}
859 
860 	skb = skb_recv_datagram(sk, flags, noblock, &err);
861 	if (!skb)
862 		goto out;
863 
864 	copied = skb->len;
865 	if (copied > len) {
866 		msg->msg_flags |= MSG_TRUNC;
867 		copied = len;
868 	}
869 
870 	/* Don't bother checking the checksum */
871 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
872 	if (err)
873 		goto done;
874 
875 	sock_recv_timestamp(msg, sk, skb);
876 
877 	/* Copy the address and add cmsg data. */
878 	if (family == AF_INET) {
879 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
880 
881 		if (sin) {
882 			sin->sin_family = AF_INET;
883 			sin->sin_port = 0 /* skb->h.uh->source */;
884 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
885 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
886 			*addr_len = sizeof(*sin);
887 		}
888 
889 		if (isk->cmsg_flags)
890 			ip_cmsg_recv(msg, skb);
891 
892 #if IS_ENABLED(CONFIG_IPV6)
893 	} else if (family == AF_INET6) {
894 		struct ipv6_pinfo *np = inet6_sk(sk);
895 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
896 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
897 
898 		if (sin6) {
899 			sin6->sin6_family = AF_INET6;
900 			sin6->sin6_port = 0;
901 			sin6->sin6_addr = ip6->saddr;
902 			sin6->sin6_flowinfo = 0;
903 			if (np->sndflow)
904 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
905 			sin6->sin6_scope_id =
906 				ipv6_iface_scope_id(&sin6->sin6_addr,
907 						    IP6CB(skb)->iif);
908 			*addr_len = sizeof(*sin6);
909 		}
910 
911 		if (inet6_sk(sk)->rxopt.all)
912 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
913 		if (skb->protocol == htons(ETH_P_IPV6) &&
914 		    inet6_sk(sk)->rxopt.all)
915 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
916 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
917 			ip_cmsg_recv(msg, skb);
918 #endif
919 	} else {
920 		BUG();
921 	}
922 
923 	err = copied;
924 
925 done:
926 	skb_free_datagram(sk, skb);
927 out:
928 	pr_debug("ping_recvmsg -> %d\n", err);
929 	return err;
930 }
931 EXPORT_SYMBOL_GPL(ping_recvmsg);
932 
933 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
934 {
935 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
936 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
937 	if (sock_queue_rcv_skb(sk, skb) < 0) {
938 		kfree_skb(skb);
939 		pr_debug("ping_queue_rcv_skb -> failed\n");
940 		return -1;
941 	}
942 	return 0;
943 }
944 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
945 
946 
947 /*
948  *	All we need to do is get the socket.
949  */
950 
951 void ping_rcv(struct sk_buff *skb)
952 {
953 	struct sock *sk;
954 	struct net *net = dev_net(skb->dev);
955 	struct icmphdr *icmph = icmp_hdr(skb);
956 
957 	/* We assume the packet has already been checked by icmp_rcv */
958 
959 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
960 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
961 
962 	/* Push ICMP header back */
963 	skb_push(skb, skb->data - (u8 *)icmph);
964 
965 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
966 	if (sk != NULL) {
967 		pr_debug("rcv on socket %p\n", sk);
968 		ping_queue_rcv_skb(sk, skb_get(skb));
969 		sock_put(sk);
970 		return;
971 	}
972 	pr_debug("no socket, dropping\n");
973 
974 	/* We're called from icmp_rcv(). kfree_skb() is done there. */
975 }
976 EXPORT_SYMBOL_GPL(ping_rcv);
977 
978 struct proto ping_prot = {
979 	.name =		"PING",
980 	.owner =	THIS_MODULE,
981 	.init =		ping_init_sock,
982 	.close =	ping_close,
983 	.connect =	ip4_datagram_connect,
984 	.disconnect =	udp_disconnect,
985 	.setsockopt =	ip_setsockopt,
986 	.getsockopt =	ip_getsockopt,
987 	.sendmsg =	ping_v4_sendmsg,
988 	.recvmsg =	ping_recvmsg,
989 	.bind =		ping_bind,
990 	.backlog_rcv =	ping_queue_rcv_skb,
991 	.release_cb =	ip4_datagram_release_cb,
992 	.hash =		ping_hash,
993 	.unhash =	ping_unhash,
994 	.get_port =	ping_get_port,
995 	.obj_size =	sizeof(struct inet_sock),
996 };
997 EXPORT_SYMBOL(ping_prot);
998 
999 #ifdef CONFIG_PROC_FS
1000 
1001 static struct sock *ping_get_first(struct seq_file *seq, int start)
1002 {
1003 	struct sock *sk;
1004 	struct ping_iter_state *state = seq->private;
1005 	struct net *net = seq_file_net(seq);
1006 
1007 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1008 	     ++state->bucket) {
1009 		struct hlist_nulls_node *node;
1010 		struct hlist_nulls_head *hslot;
1011 
1012 		hslot = &ping_table.hash[state->bucket];
1013 
1014 		if (hlist_nulls_empty(hslot))
1015 			continue;
1016 
1017 		sk_nulls_for_each(sk, node, hslot) {
1018 			if (net_eq(sock_net(sk), net) &&
1019 			    sk->sk_family == state->family)
1020 				goto found;
1021 		}
1022 	}
1023 	sk = NULL;
1024 found:
1025 	return sk;
1026 }
1027 
1028 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1029 {
1030 	struct ping_iter_state *state = seq->private;
1031 	struct net *net = seq_file_net(seq);
1032 
1033 	do {
1034 		sk = sk_nulls_next(sk);
1035 	} while (sk && (!net_eq(sock_net(sk), net)));
1036 
1037 	if (!sk)
1038 		return ping_get_first(seq, state->bucket + 1);
1039 	return sk;
1040 }
1041 
1042 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1043 {
1044 	struct sock *sk = ping_get_first(seq, 0);
1045 
1046 	if (sk)
1047 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1048 			--pos;
1049 	return pos ? NULL : sk;
1050 }
1051 
1052 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1053 {
1054 	struct ping_iter_state *state = seq->private;
1055 	state->bucket = 0;
1056 	state->family = family;
1057 
1058 	read_lock_bh(&ping_table.lock);
1059 
1060 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1061 }
1062 EXPORT_SYMBOL_GPL(ping_seq_start);
1063 
1064 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1065 {
1066 	return ping_seq_start(seq, pos, AF_INET);
1067 }
1068 
1069 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1070 {
1071 	struct sock *sk;
1072 
1073 	if (v == SEQ_START_TOKEN)
1074 		sk = ping_get_idx(seq, 0);
1075 	else
1076 		sk = ping_get_next(seq, v);
1077 
1078 	++*pos;
1079 	return sk;
1080 }
1081 EXPORT_SYMBOL_GPL(ping_seq_next);
1082 
1083 void ping_seq_stop(struct seq_file *seq, void *v)
1084 {
1085 	read_unlock_bh(&ping_table.lock);
1086 }
1087 EXPORT_SYMBOL_GPL(ping_seq_stop);
1088 
1089 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1090 		int bucket)
1091 {
1092 	struct inet_sock *inet = inet_sk(sp);
1093 	__be32 dest = inet->inet_daddr;
1094 	__be32 src = inet->inet_rcv_saddr;
1095 	__u16 destp = ntohs(inet->inet_dport);
1096 	__u16 srcp = ntohs(inet->inet_sport);
1097 
1098 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1099 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1100 		bucket, src, srcp, dest, destp, sp->sk_state,
1101 		sk_wmem_alloc_get(sp),
1102 		sk_rmem_alloc_get(sp),
1103 		0, 0L, 0,
1104 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1105 		0, sock_i_ino(sp),
1106 		atomic_read(&sp->sk_refcnt), sp,
1107 		atomic_read(&sp->sk_drops));
1108 }
1109 
1110 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1111 {
1112 	seq_setwidth(seq, 127);
1113 	if (v == SEQ_START_TOKEN)
1114 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1115 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1116 			   "inode ref pointer drops");
1117 	else {
1118 		struct ping_iter_state *state = seq->private;
1119 
1120 		ping_v4_format_sock(v, seq, state->bucket);
1121 	}
1122 	seq_pad(seq, '\n');
1123 	return 0;
1124 }
1125 
1126 static const struct seq_operations ping_v4_seq_ops = {
1127 	.show		= ping_v4_seq_show,
1128 	.start		= ping_v4_seq_start,
1129 	.next		= ping_seq_next,
1130 	.stop		= ping_seq_stop,
1131 };
1132 
1133 static int ping_seq_open(struct inode *inode, struct file *file)
1134 {
1135 	struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1136 	return seq_open_net(inode, file, &afinfo->seq_ops,
1137 			   sizeof(struct ping_iter_state));
1138 }
1139 
1140 const struct file_operations ping_seq_fops = {
1141 	.open		= ping_seq_open,
1142 	.read		= seq_read,
1143 	.llseek		= seq_lseek,
1144 	.release	= seq_release_net,
1145 };
1146 EXPORT_SYMBOL_GPL(ping_seq_fops);
1147 
1148 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1149 	.name		= "icmp",
1150 	.family		= AF_INET,
1151 	.seq_fops	= &ping_seq_fops,
1152 	.seq_ops	= {
1153 		.start		= ping_v4_seq_start,
1154 		.show		= ping_v4_seq_show,
1155 		.next		= ping_seq_next,
1156 		.stop		= ping_seq_stop,
1157 	},
1158 };
1159 
1160 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1161 {
1162 	struct proc_dir_entry *p;
1163 	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1164 			     afinfo->seq_fops, afinfo);
1165 	if (!p)
1166 		return -ENOMEM;
1167 	return 0;
1168 }
1169 EXPORT_SYMBOL_GPL(ping_proc_register);
1170 
1171 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1172 {
1173 	remove_proc_entry(afinfo->name, net->proc_net);
1174 }
1175 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1176 
1177 static int __net_init ping_v4_proc_init_net(struct net *net)
1178 {
1179 	return ping_proc_register(net, &ping_v4_seq_afinfo);
1180 }
1181 
1182 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1183 {
1184 	ping_proc_unregister(net, &ping_v4_seq_afinfo);
1185 }
1186 
1187 static struct pernet_operations ping_v4_net_ops = {
1188 	.init = ping_v4_proc_init_net,
1189 	.exit = ping_v4_proc_exit_net,
1190 };
1191 
1192 int __init ping_proc_init(void)
1193 {
1194 	return register_pernet_subsys(&ping_v4_net_ops);
1195 }
1196 
1197 void ping_proc_exit(void)
1198 {
1199 	unregister_pernet_subsys(&ping_v4_net_ops);
1200 }
1201 
1202 #endif
1203 
1204 void __init ping_init(void)
1205 {
1206 	int i;
1207 
1208 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1209 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1210 	rwlock_init(&ping_table.lock);
1211 }
1212