xref: /openbmc/linux/net/ipv4/ping.c (revision 79f08d9e)
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 
57 struct ping_table ping_table;
58 struct pingv6_ops pingv6_ops;
59 EXPORT_SYMBOL_GPL(pingv6_ops);
60 
61 static u16 ping_port_rover;
62 
63 static inline int ping_hashfn(struct net *net, unsigned int num, unsigned int mask)
64 {
65 	int res = (num + net_hash_mix(net)) & mask;
66 
67 	pr_debug("hash(%d) = %d\n", num, res);
68 	return res;
69 }
70 EXPORT_SYMBOL_GPL(ping_hash);
71 
72 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
73 					     struct net *net, unsigned int num)
74 {
75 	return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
76 }
77 
78 int ping_get_port(struct sock *sk, unsigned short ident)
79 {
80 	struct hlist_nulls_node *node;
81 	struct hlist_nulls_head *hlist;
82 	struct inet_sock *isk, *isk2;
83 	struct sock *sk2 = NULL;
84 
85 	isk = inet_sk(sk);
86 	write_lock_bh(&ping_table.lock);
87 	if (ident == 0) {
88 		u32 i;
89 		u16 result = ping_port_rover + 1;
90 
91 		for (i = 0; i < (1L << 16); i++, result++) {
92 			if (!result)
93 				result++; /* avoid zero */
94 			hlist = ping_hashslot(&ping_table, sock_net(sk),
95 					    result);
96 			ping_portaddr_for_each_entry(sk2, node, hlist) {
97 				isk2 = inet_sk(sk2);
98 
99 				if (isk2->inet_num == result)
100 					goto next_port;
101 			}
102 
103 			/* found */
104 			ping_port_rover = ident = result;
105 			break;
106 next_port:
107 			;
108 		}
109 		if (i >= (1L << 16))
110 			goto fail;
111 	} else {
112 		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
113 		ping_portaddr_for_each_entry(sk2, node, hlist) {
114 			isk2 = inet_sk(sk2);
115 
116 			/* BUG? Why is this reuse and not reuseaddr? ping.c
117 			 * doesn't turn off SO_REUSEADDR, and it doesn't expect
118 			 * that other ping processes can steal its packets.
119 			 */
120 			if ((isk2->inet_num == ident) &&
121 			    (sk2 != sk) &&
122 			    (!sk2->sk_reuse || !sk->sk_reuse))
123 				goto fail;
124 		}
125 	}
126 
127 	pr_debug("found port/ident = %d\n", ident);
128 	isk->inet_num = ident;
129 	if (sk_unhashed(sk)) {
130 		pr_debug("was not hashed\n");
131 		sock_hold(sk);
132 		hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
133 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
134 	}
135 	write_unlock_bh(&ping_table.lock);
136 	return 0;
137 
138 fail:
139 	write_unlock_bh(&ping_table.lock);
140 	return 1;
141 }
142 EXPORT_SYMBOL_GPL(ping_get_port);
143 
144 void ping_hash(struct sock *sk)
145 {
146 	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
147 	BUG(); /* "Please do not press this button again." */
148 }
149 
150 void ping_unhash(struct sock *sk)
151 {
152 	struct inet_sock *isk = inet_sk(sk);
153 	pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
154 	if (sk_hashed(sk)) {
155 		write_lock_bh(&ping_table.lock);
156 		hlist_nulls_del(&sk->sk_nulls_node);
157 		sock_put(sk);
158 		isk->inet_num = 0;
159 		isk->inet_sport = 0;
160 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
161 		write_unlock_bh(&ping_table.lock);
162 	}
163 }
164 EXPORT_SYMBOL_GPL(ping_unhash);
165 
166 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
167 {
168 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
169 	struct sock *sk = NULL;
170 	struct inet_sock *isk;
171 	struct hlist_nulls_node *hnode;
172 	int dif = skb->dev->ifindex;
173 
174 	if (skb->protocol == htons(ETH_P_IP)) {
175 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
176 			 (int)ident, &ip_hdr(skb)->daddr, dif);
177 #if IS_ENABLED(CONFIG_IPV6)
178 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
179 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
180 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
181 #endif
182 	}
183 
184 	read_lock_bh(&ping_table.lock);
185 
186 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
187 		isk = inet_sk(sk);
188 
189 		pr_debug("iterate\n");
190 		if (isk->inet_num != ident)
191 			continue;
192 
193 		if (skb->protocol == htons(ETH_P_IP) &&
194 		    sk->sk_family == AF_INET) {
195 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
196 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
197 				 sk->sk_bound_dev_if);
198 
199 			if (isk->inet_rcv_saddr &&
200 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
201 				continue;
202 #if IS_ENABLED(CONFIG_IPV6)
203 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
204 			   sk->sk_family == AF_INET6) {
205 
206 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
207 				 (int) isk->inet_num,
208 				 &sk->sk_v6_rcv_saddr,
209 				 sk->sk_bound_dev_if);
210 
211 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
212 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
213 					     &ipv6_hdr(skb)->daddr))
214 				continue;
215 #endif
216 		}
217 
218 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
219 			continue;
220 
221 		sock_hold(sk);
222 		goto exit;
223 	}
224 
225 	sk = NULL;
226 exit:
227 	read_unlock_bh(&ping_table.lock);
228 
229 	return sk;
230 }
231 
232 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
233 					  kgid_t *high)
234 {
235 	kgid_t *data = net->ipv4.sysctl_ping_group_range;
236 	unsigned int seq;
237 
238 	do {
239 		seq = read_seqbegin(&net->ipv4.sysctl_local_ports.lock);
240 
241 		*low = data[0];
242 		*high = data[1];
243 	} while (read_seqretry(&net->ipv4.sysctl_local_ports.lock, seq));
244 }
245 
246 
247 int ping_init_sock(struct sock *sk)
248 {
249 	struct net *net = sock_net(sk);
250 	kgid_t group = current_egid();
251 	struct group_info *group_info = get_current_groups();
252 	int i, j, count = group_info->ngroups;
253 	kgid_t low, high;
254 
255 	inet_get_ping_group_range_net(net, &low, &high);
256 	if (gid_lte(low, group) && gid_lte(group, high))
257 		return 0;
258 
259 	for (i = 0; i < group_info->nblocks; i++) {
260 		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
261 		for (j = 0; j < cp_count; j++) {
262 			kgid_t gid = group_info->blocks[i][j];
263 			if (gid_lte(low, gid) && gid_lte(gid, high))
264 				return 0;
265 		}
266 
267 		count -= cp_count;
268 	}
269 
270 	return -EACCES;
271 }
272 EXPORT_SYMBOL_GPL(ping_init_sock);
273 
274 void ping_close(struct sock *sk, long timeout)
275 {
276 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
277 		 inet_sk(sk), inet_sk(sk)->inet_num);
278 	pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
279 
280 	sk_common_release(sk);
281 }
282 EXPORT_SYMBOL_GPL(ping_close);
283 
284 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
285 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
286 				struct sockaddr *uaddr, int addr_len) {
287 	struct net *net = sock_net(sk);
288 	if (sk->sk_family == AF_INET) {
289 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
290 		int chk_addr_ret;
291 
292 		if (addr_len < sizeof(*addr))
293 			return -EINVAL;
294 
295 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
296 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
297 
298 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
299 
300 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
301 			chk_addr_ret = RTN_LOCAL;
302 
303 		if ((sysctl_ip_nonlocal_bind == 0 &&
304 		    isk->freebind == 0 && isk->transparent == 0 &&
305 		     chk_addr_ret != RTN_LOCAL) ||
306 		    chk_addr_ret == RTN_MULTICAST ||
307 		    chk_addr_ret == RTN_BROADCAST)
308 			return -EADDRNOTAVAIL;
309 
310 #if IS_ENABLED(CONFIG_IPV6)
311 	} else if (sk->sk_family == AF_INET6) {
312 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
313 		int addr_type, scoped, has_addr;
314 		struct net_device *dev = NULL;
315 
316 		if (addr_len < sizeof(*addr))
317 			return -EINVAL;
318 
319 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
320 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
321 
322 		addr_type = ipv6_addr_type(&addr->sin6_addr);
323 		scoped = __ipv6_addr_needs_scope_id(addr_type);
324 		if ((addr_type != IPV6_ADDR_ANY &&
325 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
326 		    (scoped && !addr->sin6_scope_id))
327 			return -EINVAL;
328 
329 		rcu_read_lock();
330 		if (addr->sin6_scope_id) {
331 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
332 			if (!dev) {
333 				rcu_read_unlock();
334 				return -ENODEV;
335 			}
336 		}
337 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
338 						    scoped);
339 		rcu_read_unlock();
340 
341 		if (!(isk->freebind || isk->transparent || has_addr ||
342 		      addr_type == IPV6_ADDR_ANY))
343 			return -EADDRNOTAVAIL;
344 
345 		if (scoped)
346 			sk->sk_bound_dev_if = addr->sin6_scope_id;
347 #endif
348 	} else {
349 		return -EAFNOSUPPORT;
350 	}
351 	return 0;
352 }
353 
354 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
355 {
356 	if (saddr->sa_family == AF_INET) {
357 		struct inet_sock *isk = inet_sk(sk);
358 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
359 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
360 #if IS_ENABLED(CONFIG_IPV6)
361 	} else if (saddr->sa_family == AF_INET6) {
362 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
363 		struct ipv6_pinfo *np = inet6_sk(sk);
364 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
365 #endif
366 	}
367 }
368 
369 static void ping_clear_saddr(struct sock *sk, int dif)
370 {
371 	sk->sk_bound_dev_if = dif;
372 	if (sk->sk_family == AF_INET) {
373 		struct inet_sock *isk = inet_sk(sk);
374 		isk->inet_rcv_saddr = isk->inet_saddr = 0;
375 #if IS_ENABLED(CONFIG_IPV6)
376 	} else if (sk->sk_family == AF_INET6) {
377 		struct ipv6_pinfo *np = inet6_sk(sk);
378 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
379 		memset(&np->saddr, 0, sizeof(np->saddr));
380 #endif
381 	}
382 }
383 /*
384  * We need our own bind because there are no privileged id's == local ports.
385  * Moreover, we don't allow binding to multi- and broadcast addresses.
386  */
387 
388 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
389 {
390 	struct inet_sock *isk = inet_sk(sk);
391 	unsigned short snum;
392 	int err;
393 	int dif = sk->sk_bound_dev_if;
394 
395 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
396 	if (err)
397 		return err;
398 
399 	lock_sock(sk);
400 
401 	err = -EINVAL;
402 	if (isk->inet_num != 0)
403 		goto out;
404 
405 	err = -EADDRINUSE;
406 	ping_set_saddr(sk, uaddr);
407 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
408 	if (ping_get_port(sk, snum) != 0) {
409 		ping_clear_saddr(sk, dif);
410 		goto out;
411 	}
412 
413 	pr_debug("after bind(): num = %d, dif = %d\n",
414 		 (int)isk->inet_num,
415 		 (int)sk->sk_bound_dev_if);
416 
417 	err = 0;
418 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
419 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
420 #if IS_ENABLED(CONFIG_IPV6)
421 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
422 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
423 #endif
424 
425 	if (snum)
426 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
427 	isk->inet_sport = htons(isk->inet_num);
428 	isk->inet_daddr = 0;
429 	isk->inet_dport = 0;
430 
431 #if IS_ENABLED(CONFIG_IPV6)
432 	if (sk->sk_family == AF_INET6)
433 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
434 #endif
435 
436 	sk_dst_reset(sk);
437 out:
438 	release_sock(sk);
439 	pr_debug("ping_v4_bind -> %d\n", err);
440 	return err;
441 }
442 EXPORT_SYMBOL_GPL(ping_bind);
443 
444 /*
445  * Is this a supported type of ICMP message?
446  */
447 
448 static inline int ping_supported(int family, int type, int code)
449 {
450 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
451 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
452 }
453 
454 /*
455  * This routine is called by the ICMP module when it gets some
456  * sort of error condition.
457  */
458 
459 void ping_err(struct sk_buff *skb, int offset, u32 info)
460 {
461 	int family;
462 	struct icmphdr *icmph;
463 	struct inet_sock *inet_sock;
464 	int type;
465 	int code;
466 	struct net *net = dev_net(skb->dev);
467 	struct sock *sk;
468 	int harderr;
469 	int err;
470 
471 	if (skb->protocol == htons(ETH_P_IP)) {
472 		family = AF_INET;
473 		type = icmp_hdr(skb)->type;
474 		code = icmp_hdr(skb)->code;
475 		icmph = (struct icmphdr *)(skb->data + offset);
476 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
477 		family = AF_INET6;
478 		type = icmp6_hdr(skb)->icmp6_type;
479 		code = icmp6_hdr(skb)->icmp6_code;
480 		icmph = (struct icmphdr *) (skb->data + offset);
481 	} else {
482 		BUG();
483 	}
484 
485 	/* We assume the packet has already been checked by icmp_unreach */
486 
487 	if (!ping_supported(family, icmph->type, icmph->code))
488 		return;
489 
490 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
491 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
492 		 ntohs(icmph->un.echo.sequence));
493 
494 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
495 	if (sk == NULL) {
496 		pr_debug("no socket, dropping\n");
497 		return;	/* No socket for error */
498 	}
499 	pr_debug("err on socket %p\n", sk);
500 
501 	err = 0;
502 	harderr = 0;
503 	inet_sock = inet_sk(sk);
504 
505 	if (skb->protocol == htons(ETH_P_IP)) {
506 		switch (type) {
507 		default:
508 		case ICMP_TIME_EXCEEDED:
509 			err = EHOSTUNREACH;
510 			break;
511 		case ICMP_SOURCE_QUENCH:
512 			/* This is not a real error but ping wants to see it.
513 			 * Report it with some fake errno.
514 			 */
515 			err = EREMOTEIO;
516 			break;
517 		case ICMP_PARAMETERPROB:
518 			err = EPROTO;
519 			harderr = 1;
520 			break;
521 		case ICMP_DEST_UNREACH:
522 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
523 				ipv4_sk_update_pmtu(skb, sk, info);
524 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
525 					err = EMSGSIZE;
526 					harderr = 1;
527 					break;
528 				}
529 				goto out;
530 			}
531 			err = EHOSTUNREACH;
532 			if (code <= NR_ICMP_UNREACH) {
533 				harderr = icmp_err_convert[code].fatal;
534 				err = icmp_err_convert[code].errno;
535 			}
536 			break;
537 		case ICMP_REDIRECT:
538 			/* See ICMP_SOURCE_QUENCH */
539 			ipv4_sk_redirect(skb, sk);
540 			err = EREMOTEIO;
541 			break;
542 		}
543 #if IS_ENABLED(CONFIG_IPV6)
544 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
545 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
546 #endif
547 	}
548 
549 	/*
550 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
551 	 *	4.1.3.3.
552 	 */
553 	if ((family == AF_INET && !inet_sock->recverr) ||
554 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
555 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
556 			goto out;
557 	} else {
558 		if (family == AF_INET) {
559 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
560 				      info, (u8 *)icmph);
561 #if IS_ENABLED(CONFIG_IPV6)
562 		} else if (family == AF_INET6) {
563 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
564 						   info, (u8 *)icmph);
565 #endif
566 		}
567 	}
568 	sk->sk_err = err;
569 	sk->sk_error_report(sk);
570 out:
571 	sock_put(sk);
572 }
573 EXPORT_SYMBOL_GPL(ping_err);
574 
575 /*
576  *	Copy and checksum an ICMP Echo packet from user space into a buffer
577  *	starting from the payload.
578  */
579 
580 int ping_getfrag(void *from, char *to,
581 		 int offset, int fraglen, int odd, struct sk_buff *skb)
582 {
583 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
584 
585 	if (offset == 0) {
586 		if (fraglen < sizeof(struct icmphdr))
587 			BUG();
588 		if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
589 			    pfh->iov, 0, fraglen - sizeof(struct icmphdr),
590 			    &pfh->wcheck))
591 			return -EFAULT;
592 	} else if (offset < sizeof(struct icmphdr)) {
593 			BUG();
594 	} else {
595 		if (csum_partial_copy_fromiovecend
596 				(to, pfh->iov, offset - sizeof(struct icmphdr),
597 				 fraglen, &pfh->wcheck))
598 			return -EFAULT;
599 	}
600 
601 #if IS_ENABLED(CONFIG_IPV6)
602 	/* For IPv6, checksum each skb as we go along, as expected by
603 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
604 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
605 	 */
606 	if (pfh->family == AF_INET6) {
607 		skb->csum = pfh->wcheck;
608 		skb->ip_summed = CHECKSUM_NONE;
609 		pfh->wcheck = 0;
610 	}
611 #endif
612 
613 	return 0;
614 }
615 EXPORT_SYMBOL_GPL(ping_getfrag);
616 
617 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
618 				       struct flowi4 *fl4)
619 {
620 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
621 
622 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
623 		sizeof(struct icmphdr), pfh->wcheck);
624 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
625 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
626 	skb->ip_summed = CHECKSUM_NONE;
627 	return ip_push_pending_frames(sk, fl4);
628 }
629 
630 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
631 			void *user_icmph, size_t icmph_len) {
632 	u8 type, code;
633 
634 	if (len > 0xFFFF)
635 		return -EMSGSIZE;
636 
637 	/*
638 	 *	Check the flags.
639 	 */
640 
641 	/* Mirror BSD error message compatibility */
642 	if (msg->msg_flags & MSG_OOB)
643 		return -EOPNOTSUPP;
644 
645 	/*
646 	 *	Fetch the ICMP header provided by the userland.
647 	 *	iovec is modified! The ICMP header is consumed.
648 	 */
649 	if (memcpy_fromiovec(user_icmph, msg->msg_iov, icmph_len))
650 		return -EFAULT;
651 
652 	if (family == AF_INET) {
653 		type = ((struct icmphdr *) user_icmph)->type;
654 		code = ((struct icmphdr *) user_icmph)->code;
655 #if IS_ENABLED(CONFIG_IPV6)
656 	} else if (family == AF_INET6) {
657 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
658 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
659 #endif
660 	} else {
661 		BUG();
662 	}
663 
664 	if (!ping_supported(family, type, code))
665 		return -EINVAL;
666 
667 	return 0;
668 }
669 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
670 
671 int ping_v4_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
672 		    size_t len)
673 {
674 	struct net *net = sock_net(sk);
675 	struct flowi4 fl4;
676 	struct inet_sock *inet = inet_sk(sk);
677 	struct ipcm_cookie ipc;
678 	struct icmphdr user_icmph;
679 	struct pingfakehdr pfh;
680 	struct rtable *rt = NULL;
681 	struct ip_options_data opt_copy;
682 	int free = 0;
683 	__be32 saddr, daddr, faddr;
684 	u8  tos;
685 	int err;
686 
687 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
688 
689 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
690 				  sizeof(user_icmph));
691 	if (err)
692 		return err;
693 
694 	/*
695 	 *	Get and verify the address.
696 	 */
697 
698 	if (msg->msg_name) {
699 		struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
700 		if (msg->msg_namelen < sizeof(*usin))
701 			return -EINVAL;
702 		if (usin->sin_family != AF_INET)
703 			return -EINVAL;
704 		daddr = usin->sin_addr.s_addr;
705 		/* no remote port */
706 	} else {
707 		if (sk->sk_state != TCP_ESTABLISHED)
708 			return -EDESTADDRREQ;
709 		daddr = inet->inet_daddr;
710 		/* no remote port */
711 	}
712 
713 	ipc.addr = inet->inet_saddr;
714 	ipc.opt = NULL;
715 	ipc.oif = sk->sk_bound_dev_if;
716 	ipc.tx_flags = 0;
717 	ipc.ttl = 0;
718 	ipc.tos = -1;
719 
720 	sock_tx_timestamp(sk, &ipc.tx_flags);
721 
722 	if (msg->msg_controllen) {
723 		err = ip_cmsg_send(sock_net(sk), msg, &ipc);
724 		if (err)
725 			return err;
726 		if (ipc.opt)
727 			free = 1;
728 	}
729 	if (!ipc.opt) {
730 		struct ip_options_rcu *inet_opt;
731 
732 		rcu_read_lock();
733 		inet_opt = rcu_dereference(inet->inet_opt);
734 		if (inet_opt) {
735 			memcpy(&opt_copy, inet_opt,
736 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
737 			ipc.opt = &opt_copy.opt;
738 		}
739 		rcu_read_unlock();
740 	}
741 
742 	saddr = ipc.addr;
743 	ipc.addr = faddr = daddr;
744 
745 	if (ipc.opt && ipc.opt->opt.srr) {
746 		if (!daddr)
747 			return -EINVAL;
748 		faddr = ipc.opt->opt.faddr;
749 	}
750 	tos = get_rttos(&ipc, inet);
751 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
752 	    (msg->msg_flags & MSG_DONTROUTE) ||
753 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
754 		tos |= RTO_ONLINK;
755 	}
756 
757 	if (ipv4_is_multicast(daddr)) {
758 		if (!ipc.oif)
759 			ipc.oif = inet->mc_index;
760 		if (!saddr)
761 			saddr = inet->mc_addr;
762 	} else if (!ipc.oif)
763 		ipc.oif = inet->uc_index;
764 
765 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
766 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
767 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
768 
769 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
770 	rt = ip_route_output_flow(net, &fl4, sk);
771 	if (IS_ERR(rt)) {
772 		err = PTR_ERR(rt);
773 		rt = NULL;
774 		if (err == -ENETUNREACH)
775 			IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
776 		goto out;
777 	}
778 
779 	err = -EACCES;
780 	if ((rt->rt_flags & RTCF_BROADCAST) &&
781 	    !sock_flag(sk, SOCK_BROADCAST))
782 		goto out;
783 
784 	if (msg->msg_flags & MSG_CONFIRM)
785 		goto do_confirm;
786 back_from_confirm:
787 
788 	if (!ipc.addr)
789 		ipc.addr = fl4.daddr;
790 
791 	lock_sock(sk);
792 
793 	pfh.icmph.type = user_icmph.type; /* already checked */
794 	pfh.icmph.code = user_icmph.code; /* ditto */
795 	pfh.icmph.checksum = 0;
796 	pfh.icmph.un.echo.id = inet->inet_sport;
797 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
798 	pfh.iov = msg->msg_iov;
799 	pfh.wcheck = 0;
800 	pfh.family = AF_INET;
801 
802 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
803 			0, &ipc, &rt, msg->msg_flags);
804 	if (err)
805 		ip_flush_pending_frames(sk);
806 	else
807 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
808 	release_sock(sk);
809 
810 out:
811 	ip_rt_put(rt);
812 	if (free)
813 		kfree(ipc.opt);
814 	if (!err) {
815 		icmp_out_count(sock_net(sk), user_icmph.type);
816 		return len;
817 	}
818 	return err;
819 
820 do_confirm:
821 	dst_confirm(&rt->dst);
822 	if (!(msg->msg_flags & MSG_PROBE) || len)
823 		goto back_from_confirm;
824 	err = 0;
825 	goto out;
826 }
827 
828 int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
829 		 size_t len, int noblock, int flags, int *addr_len)
830 {
831 	struct inet_sock *isk = inet_sk(sk);
832 	int family = sk->sk_family;
833 	struct sk_buff *skb;
834 	int copied, err;
835 
836 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
837 
838 	err = -EOPNOTSUPP;
839 	if (flags & MSG_OOB)
840 		goto out;
841 
842 	if (flags & MSG_ERRQUEUE) {
843 		if (family == AF_INET) {
844 			return ip_recv_error(sk, msg, len);
845 #if IS_ENABLED(CONFIG_IPV6)
846 		} else if (family == AF_INET6) {
847 			return pingv6_ops.ipv6_recv_error(sk, msg, len);
848 #endif
849 		}
850 	}
851 
852 	skb = skb_recv_datagram(sk, flags, noblock, &err);
853 	if (!skb)
854 		goto out;
855 
856 	copied = skb->len;
857 	if (copied > len) {
858 		msg->msg_flags |= MSG_TRUNC;
859 		copied = len;
860 	}
861 
862 	/* Don't bother checking the checksum */
863 	err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
864 	if (err)
865 		goto done;
866 
867 	sock_recv_timestamp(msg, sk, skb);
868 
869 	/* Copy the address and add cmsg data. */
870 	if (family == AF_INET) {
871 		struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
872 
873 		if (sin) {
874 			sin->sin_family = AF_INET;
875 			sin->sin_port = 0 /* skb->h.uh->source */;
876 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
877 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
878 			*addr_len = sizeof(*sin);
879 		}
880 
881 		if (isk->cmsg_flags)
882 			ip_cmsg_recv(msg, skb);
883 
884 #if IS_ENABLED(CONFIG_IPV6)
885 	} else if (family == AF_INET6) {
886 		struct ipv6_pinfo *np = inet6_sk(sk);
887 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
888 		struct sockaddr_in6 *sin6 =
889 			(struct sockaddr_in6 *)msg->msg_name;
890 
891 		if (sin6) {
892 			sin6->sin6_family = AF_INET6;
893 			sin6->sin6_port = 0;
894 			sin6->sin6_addr = ip6->saddr;
895 			sin6->sin6_flowinfo = 0;
896 			if (np->sndflow)
897 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
898 			sin6->sin6_scope_id =
899 				ipv6_iface_scope_id(&sin6->sin6_addr,
900 						    IP6CB(skb)->iif);
901 			*addr_len = sizeof(*sin6);
902 		}
903 
904 		if (inet6_sk(sk)->rxopt.all)
905 			pingv6_ops.ip6_datagram_recv_ctl(sk, msg, skb);
906 #endif
907 	} else {
908 		BUG();
909 	}
910 
911 	err = copied;
912 
913 done:
914 	skb_free_datagram(sk, skb);
915 out:
916 	pr_debug("ping_recvmsg -> %d\n", err);
917 	return err;
918 }
919 EXPORT_SYMBOL_GPL(ping_recvmsg);
920 
921 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
922 {
923 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
924 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
925 	if (sock_queue_rcv_skb(sk, skb) < 0) {
926 		kfree_skb(skb);
927 		pr_debug("ping_queue_rcv_skb -> failed\n");
928 		return -1;
929 	}
930 	return 0;
931 }
932 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
933 
934 
935 /*
936  *	All we need to do is get the socket.
937  */
938 
939 void ping_rcv(struct sk_buff *skb)
940 {
941 	struct sock *sk;
942 	struct net *net = dev_net(skb->dev);
943 	struct icmphdr *icmph = icmp_hdr(skb);
944 
945 	/* We assume the packet has already been checked by icmp_rcv */
946 
947 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
948 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
949 
950 	/* Push ICMP header back */
951 	skb_push(skb, skb->data - (u8 *)icmph);
952 
953 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
954 	if (sk != NULL) {
955 		pr_debug("rcv on socket %p\n", sk);
956 		ping_queue_rcv_skb(sk, skb_get(skb));
957 		sock_put(sk);
958 		return;
959 	}
960 	pr_debug("no socket, dropping\n");
961 
962 	/* We're called from icmp_rcv(). kfree_skb() is done there. */
963 }
964 EXPORT_SYMBOL_GPL(ping_rcv);
965 
966 struct proto ping_prot = {
967 	.name =		"PING",
968 	.owner =	THIS_MODULE,
969 	.init =		ping_init_sock,
970 	.close =	ping_close,
971 	.connect =	ip4_datagram_connect,
972 	.disconnect =	udp_disconnect,
973 	.setsockopt =	ip_setsockopt,
974 	.getsockopt =	ip_getsockopt,
975 	.sendmsg =	ping_v4_sendmsg,
976 	.recvmsg =	ping_recvmsg,
977 	.bind =		ping_bind,
978 	.backlog_rcv =	ping_queue_rcv_skb,
979 	.release_cb =	ip4_datagram_release_cb,
980 	.hash =		ping_hash,
981 	.unhash =	ping_unhash,
982 	.get_port =	ping_get_port,
983 	.obj_size =	sizeof(struct inet_sock),
984 };
985 EXPORT_SYMBOL(ping_prot);
986 
987 #ifdef CONFIG_PROC_FS
988 
989 static struct sock *ping_get_first(struct seq_file *seq, int start)
990 {
991 	struct sock *sk;
992 	struct ping_iter_state *state = seq->private;
993 	struct net *net = seq_file_net(seq);
994 
995 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
996 	     ++state->bucket) {
997 		struct hlist_nulls_node *node;
998 		struct hlist_nulls_head *hslot;
999 
1000 		hslot = &ping_table.hash[state->bucket];
1001 
1002 		if (hlist_nulls_empty(hslot))
1003 			continue;
1004 
1005 		sk_nulls_for_each(sk, node, hslot) {
1006 			if (net_eq(sock_net(sk), net) &&
1007 			    sk->sk_family == state->family)
1008 				goto found;
1009 		}
1010 	}
1011 	sk = NULL;
1012 found:
1013 	return sk;
1014 }
1015 
1016 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1017 {
1018 	struct ping_iter_state *state = seq->private;
1019 	struct net *net = seq_file_net(seq);
1020 
1021 	do {
1022 		sk = sk_nulls_next(sk);
1023 	} while (sk && (!net_eq(sock_net(sk), net)));
1024 
1025 	if (!sk)
1026 		return ping_get_first(seq, state->bucket + 1);
1027 	return sk;
1028 }
1029 
1030 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1031 {
1032 	struct sock *sk = ping_get_first(seq, 0);
1033 
1034 	if (sk)
1035 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1036 			--pos;
1037 	return pos ? NULL : sk;
1038 }
1039 
1040 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1041 {
1042 	struct ping_iter_state *state = seq->private;
1043 	state->bucket = 0;
1044 	state->family = family;
1045 
1046 	read_lock_bh(&ping_table.lock);
1047 
1048 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1049 }
1050 EXPORT_SYMBOL_GPL(ping_seq_start);
1051 
1052 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1053 {
1054 	return ping_seq_start(seq, pos, AF_INET);
1055 }
1056 
1057 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1058 {
1059 	struct sock *sk;
1060 
1061 	if (v == SEQ_START_TOKEN)
1062 		sk = ping_get_idx(seq, 0);
1063 	else
1064 		sk = ping_get_next(seq, v);
1065 
1066 	++*pos;
1067 	return sk;
1068 }
1069 EXPORT_SYMBOL_GPL(ping_seq_next);
1070 
1071 void ping_seq_stop(struct seq_file *seq, void *v)
1072 {
1073 	read_unlock_bh(&ping_table.lock);
1074 }
1075 EXPORT_SYMBOL_GPL(ping_seq_stop);
1076 
1077 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1078 		int bucket)
1079 {
1080 	struct inet_sock *inet = inet_sk(sp);
1081 	__be32 dest = inet->inet_daddr;
1082 	__be32 src = inet->inet_rcv_saddr;
1083 	__u16 destp = ntohs(inet->inet_dport);
1084 	__u16 srcp = ntohs(inet->inet_sport);
1085 
1086 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1087 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1088 		bucket, src, srcp, dest, destp, sp->sk_state,
1089 		sk_wmem_alloc_get(sp),
1090 		sk_rmem_alloc_get(sp),
1091 		0, 0L, 0,
1092 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1093 		0, sock_i_ino(sp),
1094 		atomic_read(&sp->sk_refcnt), sp,
1095 		atomic_read(&sp->sk_drops));
1096 }
1097 
1098 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1099 {
1100 	seq_setwidth(seq, 127);
1101 	if (v == SEQ_START_TOKEN)
1102 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1103 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1104 			   "inode ref pointer drops");
1105 	else {
1106 		struct ping_iter_state *state = seq->private;
1107 
1108 		ping_v4_format_sock(v, seq, state->bucket);
1109 	}
1110 	seq_pad(seq, '\n');
1111 	return 0;
1112 }
1113 
1114 static const struct seq_operations ping_v4_seq_ops = {
1115 	.show		= ping_v4_seq_show,
1116 	.start		= ping_v4_seq_start,
1117 	.next		= ping_seq_next,
1118 	.stop		= ping_seq_stop,
1119 };
1120 
1121 static int ping_seq_open(struct inode *inode, struct file *file)
1122 {
1123 	struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1124 	return seq_open_net(inode, file, &afinfo->seq_ops,
1125 			   sizeof(struct ping_iter_state));
1126 }
1127 
1128 const struct file_operations ping_seq_fops = {
1129 	.open		= ping_seq_open,
1130 	.read		= seq_read,
1131 	.llseek		= seq_lseek,
1132 	.release	= seq_release_net,
1133 };
1134 EXPORT_SYMBOL_GPL(ping_seq_fops);
1135 
1136 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1137 	.name		= "icmp",
1138 	.family		= AF_INET,
1139 	.seq_fops	= &ping_seq_fops,
1140 	.seq_ops	= {
1141 		.start		= ping_v4_seq_start,
1142 		.show		= ping_v4_seq_show,
1143 		.next		= ping_seq_next,
1144 		.stop		= ping_seq_stop,
1145 	},
1146 };
1147 
1148 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1149 {
1150 	struct proc_dir_entry *p;
1151 	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1152 			     afinfo->seq_fops, afinfo);
1153 	if (!p)
1154 		return -ENOMEM;
1155 	return 0;
1156 }
1157 EXPORT_SYMBOL_GPL(ping_proc_register);
1158 
1159 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1160 {
1161 	remove_proc_entry(afinfo->name, net->proc_net);
1162 }
1163 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1164 
1165 static int __net_init ping_v4_proc_init_net(struct net *net)
1166 {
1167 	return ping_proc_register(net, &ping_v4_seq_afinfo);
1168 }
1169 
1170 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1171 {
1172 	ping_proc_unregister(net, &ping_v4_seq_afinfo);
1173 }
1174 
1175 static struct pernet_operations ping_v4_net_ops = {
1176 	.init = ping_v4_proc_init_net,
1177 	.exit = ping_v4_proc_exit_net,
1178 };
1179 
1180 int __init ping_proc_init(void)
1181 {
1182 	return register_pernet_subsys(&ping_v4_net_ops);
1183 }
1184 
1185 void ping_proc_exit(void)
1186 {
1187 	unregister_pernet_subsys(&ping_v4_net_ops);
1188 }
1189 
1190 #endif
1191 
1192 void __init ping_init(void)
1193 {
1194 	int i;
1195 
1196 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1197 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1198 	rwlock_init(&ping_table.lock);
1199 }
1200