xref: /openbmc/linux/net/ipv4/ping.c (revision b96fc2f3)
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 u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
68 {
69 	u32 res = (num + net_hash_mix(net)) & mask;
70 
71 	pr_debug("hash(%u) = %u\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 		sk_nulls_node_init(&sk->sk_nulls_node);
162 		sock_put(sk);
163 		isk->inet_num = 0;
164 		isk->inet_sport = 0;
165 		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
166 		write_unlock_bh(&ping_table.lock);
167 	}
168 }
169 EXPORT_SYMBOL_GPL(ping_unhash);
170 
171 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
172 {
173 	struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
174 	struct sock *sk = NULL;
175 	struct inet_sock *isk;
176 	struct hlist_nulls_node *hnode;
177 	int dif = skb->dev->ifindex;
178 
179 	if (skb->protocol == htons(ETH_P_IP)) {
180 		pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
181 			 (int)ident, &ip_hdr(skb)->daddr, dif);
182 #if IS_ENABLED(CONFIG_IPV6)
183 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
184 		pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
185 			 (int)ident, &ipv6_hdr(skb)->daddr, dif);
186 #endif
187 	}
188 
189 	read_lock_bh(&ping_table.lock);
190 
191 	ping_portaddr_for_each_entry(sk, hnode, hslot) {
192 		isk = inet_sk(sk);
193 
194 		pr_debug("iterate\n");
195 		if (isk->inet_num != ident)
196 			continue;
197 
198 		if (skb->protocol == htons(ETH_P_IP) &&
199 		    sk->sk_family == AF_INET) {
200 			pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
201 				 (int) isk->inet_num, &isk->inet_rcv_saddr,
202 				 sk->sk_bound_dev_if);
203 
204 			if (isk->inet_rcv_saddr &&
205 			    isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
206 				continue;
207 #if IS_ENABLED(CONFIG_IPV6)
208 		} else if (skb->protocol == htons(ETH_P_IPV6) &&
209 			   sk->sk_family == AF_INET6) {
210 
211 			pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
212 				 (int) isk->inet_num,
213 				 &sk->sk_v6_rcv_saddr,
214 				 sk->sk_bound_dev_if);
215 
216 			if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
217 			    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
218 					     &ipv6_hdr(skb)->daddr))
219 				continue;
220 #endif
221 		} else {
222 			continue;
223 		}
224 
225 		if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
226 			continue;
227 
228 		sock_hold(sk);
229 		goto exit;
230 	}
231 
232 	sk = NULL;
233 exit:
234 	read_unlock_bh(&ping_table.lock);
235 
236 	return sk;
237 }
238 
239 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
240 					  kgid_t *high)
241 {
242 	kgid_t *data = net->ipv4.ping_group_range.range;
243 	unsigned int seq;
244 
245 	do {
246 		seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
247 
248 		*low = data[0];
249 		*high = data[1];
250 	} while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
251 }
252 
253 
254 int ping_init_sock(struct sock *sk)
255 {
256 	struct net *net = sock_net(sk);
257 	kgid_t group = current_egid();
258 	struct group_info *group_info;
259 	int i, j, count;
260 	kgid_t low, high;
261 	int ret = 0;
262 
263 	if (sk->sk_family == AF_INET6)
264 		sk->sk_ipv6only = 1;
265 
266 	inet_get_ping_group_range_net(net, &low, &high);
267 	if (gid_lte(low, group) && gid_lte(group, high))
268 		return 0;
269 
270 	group_info = get_current_groups();
271 	count = group_info->ngroups;
272 	for (i = 0; i < group_info->nblocks; i++) {
273 		int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
274 		for (j = 0; j < cp_count; j++) {
275 			kgid_t gid = group_info->blocks[i][j];
276 			if (gid_lte(low, gid) && gid_lte(gid, high))
277 				goto out_release_group;
278 		}
279 
280 		count -= cp_count;
281 	}
282 
283 	ret = -EACCES;
284 
285 out_release_group:
286 	put_group_info(group_info);
287 	return ret;
288 }
289 EXPORT_SYMBOL_GPL(ping_init_sock);
290 
291 void ping_close(struct sock *sk, long timeout)
292 {
293 	pr_debug("ping_close(sk=%p,sk->num=%u)\n",
294 		 inet_sk(sk), inet_sk(sk)->inet_num);
295 	pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
296 
297 	sk_common_release(sk);
298 }
299 EXPORT_SYMBOL_GPL(ping_close);
300 
301 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
302 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
303 				struct sockaddr *uaddr, int addr_len) {
304 	struct net *net = sock_net(sk);
305 	if (sk->sk_family == AF_INET) {
306 		struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
307 		int chk_addr_ret;
308 
309 		if (addr_len < sizeof(*addr))
310 			return -EINVAL;
311 
312 		if (addr->sin_family != AF_INET &&
313 		    !(addr->sin_family == AF_UNSPEC &&
314 		      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
315 			return -EAFNOSUPPORT;
316 
317 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
318 			 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
319 
320 		chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
321 
322 		if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
323 			chk_addr_ret = RTN_LOCAL;
324 
325 		if ((net->ipv4.sysctl_ip_nonlocal_bind == 0 &&
326 		    isk->freebind == 0 && isk->transparent == 0 &&
327 		     chk_addr_ret != RTN_LOCAL) ||
328 		    chk_addr_ret == RTN_MULTICAST ||
329 		    chk_addr_ret == RTN_BROADCAST)
330 			return -EADDRNOTAVAIL;
331 
332 #if IS_ENABLED(CONFIG_IPV6)
333 	} else if (sk->sk_family == AF_INET6) {
334 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
335 		int addr_type, scoped, has_addr;
336 		struct net_device *dev = NULL;
337 
338 		if (addr_len < sizeof(*addr))
339 			return -EINVAL;
340 
341 		if (addr->sin6_family != AF_INET6)
342 			return -EAFNOSUPPORT;
343 
344 		pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
345 			 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
346 
347 		addr_type = ipv6_addr_type(&addr->sin6_addr);
348 		scoped = __ipv6_addr_needs_scope_id(addr_type);
349 		if ((addr_type != IPV6_ADDR_ANY &&
350 		     !(addr_type & IPV6_ADDR_UNICAST)) ||
351 		    (scoped && !addr->sin6_scope_id))
352 			return -EINVAL;
353 
354 		rcu_read_lock();
355 		if (addr->sin6_scope_id) {
356 			dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
357 			if (!dev) {
358 				rcu_read_unlock();
359 				return -ENODEV;
360 			}
361 		}
362 		has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
363 						    scoped);
364 		rcu_read_unlock();
365 
366 		if (!(net->ipv6.sysctl.ip_nonlocal_bind ||
367 		      isk->freebind || isk->transparent || has_addr ||
368 		      addr_type == IPV6_ADDR_ANY))
369 			return -EADDRNOTAVAIL;
370 
371 		if (scoped)
372 			sk->sk_bound_dev_if = addr->sin6_scope_id;
373 #endif
374 	} else {
375 		return -EAFNOSUPPORT;
376 	}
377 	return 0;
378 }
379 
380 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
381 {
382 	if (saddr->sa_family == AF_INET) {
383 		struct inet_sock *isk = inet_sk(sk);
384 		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
385 		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
386 #if IS_ENABLED(CONFIG_IPV6)
387 	} else if (saddr->sa_family == AF_INET6) {
388 		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
389 		struct ipv6_pinfo *np = inet6_sk(sk);
390 		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
391 #endif
392 	}
393 }
394 
395 static void ping_clear_saddr(struct sock *sk, int dif)
396 {
397 	sk->sk_bound_dev_if = dif;
398 	if (sk->sk_family == AF_INET) {
399 		struct inet_sock *isk = inet_sk(sk);
400 		isk->inet_rcv_saddr = isk->inet_saddr = 0;
401 #if IS_ENABLED(CONFIG_IPV6)
402 	} else if (sk->sk_family == AF_INET6) {
403 		struct ipv6_pinfo *np = inet6_sk(sk);
404 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
405 		memset(&np->saddr, 0, sizeof(np->saddr));
406 #endif
407 	}
408 }
409 /*
410  * We need our own bind because there are no privileged id's == local ports.
411  * Moreover, we don't allow binding to multi- and broadcast addresses.
412  */
413 
414 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
415 {
416 	struct inet_sock *isk = inet_sk(sk);
417 	unsigned short snum;
418 	int err;
419 	int dif = sk->sk_bound_dev_if;
420 
421 	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
422 	if (err)
423 		return err;
424 
425 	lock_sock(sk);
426 
427 	err = -EINVAL;
428 	if (isk->inet_num != 0)
429 		goto out;
430 
431 	err = -EADDRINUSE;
432 	ping_set_saddr(sk, uaddr);
433 	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
434 	if (ping_get_port(sk, snum) != 0) {
435 		ping_clear_saddr(sk, dif);
436 		goto out;
437 	}
438 
439 	pr_debug("after bind(): num = %d, dif = %d\n",
440 		 (int)isk->inet_num,
441 		 (int)sk->sk_bound_dev_if);
442 
443 	err = 0;
444 	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
445 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
446 #if IS_ENABLED(CONFIG_IPV6)
447 	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
448 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
449 #endif
450 
451 	if (snum)
452 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
453 	isk->inet_sport = htons(isk->inet_num);
454 	isk->inet_daddr = 0;
455 	isk->inet_dport = 0;
456 
457 #if IS_ENABLED(CONFIG_IPV6)
458 	if (sk->sk_family == AF_INET6)
459 		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
460 #endif
461 
462 	sk_dst_reset(sk);
463 out:
464 	release_sock(sk);
465 	pr_debug("ping_v4_bind -> %d\n", err);
466 	return err;
467 }
468 EXPORT_SYMBOL_GPL(ping_bind);
469 
470 /*
471  * Is this a supported type of ICMP message?
472  */
473 
474 static inline int ping_supported(int family, int type, int code)
475 {
476 	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
477 	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
478 }
479 
480 /*
481  * This routine is called by the ICMP module when it gets some
482  * sort of error condition.
483  */
484 
485 void ping_err(struct sk_buff *skb, int offset, u32 info)
486 {
487 	int family;
488 	struct icmphdr *icmph;
489 	struct inet_sock *inet_sock;
490 	int type;
491 	int code;
492 	struct net *net = dev_net(skb->dev);
493 	struct sock *sk;
494 	int harderr;
495 	int err;
496 
497 	if (skb->protocol == htons(ETH_P_IP)) {
498 		family = AF_INET;
499 		type = icmp_hdr(skb)->type;
500 		code = icmp_hdr(skb)->code;
501 		icmph = (struct icmphdr *)(skb->data + offset);
502 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
503 		family = AF_INET6;
504 		type = icmp6_hdr(skb)->icmp6_type;
505 		code = icmp6_hdr(skb)->icmp6_code;
506 		icmph = (struct icmphdr *) (skb->data + offset);
507 	} else {
508 		BUG();
509 	}
510 
511 	/* We assume the packet has already been checked by icmp_unreach */
512 
513 	if (!ping_supported(family, icmph->type, icmph->code))
514 		return;
515 
516 	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
517 		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
518 		 ntohs(icmph->un.echo.sequence));
519 
520 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
521 	if (!sk) {
522 		pr_debug("no socket, dropping\n");
523 		return;	/* No socket for error */
524 	}
525 	pr_debug("err on socket %p\n", sk);
526 
527 	err = 0;
528 	harderr = 0;
529 	inet_sock = inet_sk(sk);
530 
531 	if (skb->protocol == htons(ETH_P_IP)) {
532 		switch (type) {
533 		default:
534 		case ICMP_TIME_EXCEEDED:
535 			err = EHOSTUNREACH;
536 			break;
537 		case ICMP_SOURCE_QUENCH:
538 			/* This is not a real error but ping wants to see it.
539 			 * Report it with some fake errno.
540 			 */
541 			err = EREMOTEIO;
542 			break;
543 		case ICMP_PARAMETERPROB:
544 			err = EPROTO;
545 			harderr = 1;
546 			break;
547 		case ICMP_DEST_UNREACH:
548 			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
549 				ipv4_sk_update_pmtu(skb, sk, info);
550 				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
551 					err = EMSGSIZE;
552 					harderr = 1;
553 					break;
554 				}
555 				goto out;
556 			}
557 			err = EHOSTUNREACH;
558 			if (code <= NR_ICMP_UNREACH) {
559 				harderr = icmp_err_convert[code].fatal;
560 				err = icmp_err_convert[code].errno;
561 			}
562 			break;
563 		case ICMP_REDIRECT:
564 			/* See ICMP_SOURCE_QUENCH */
565 			ipv4_sk_redirect(skb, sk);
566 			err = EREMOTEIO;
567 			break;
568 		}
569 #if IS_ENABLED(CONFIG_IPV6)
570 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
571 		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
572 #endif
573 	}
574 
575 	/*
576 	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
577 	 *	4.1.3.3.
578 	 */
579 	if ((family == AF_INET && !inet_sock->recverr) ||
580 	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
581 		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
582 			goto out;
583 	} else {
584 		if (family == AF_INET) {
585 			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
586 				      info, (u8 *)icmph);
587 #if IS_ENABLED(CONFIG_IPV6)
588 		} else if (family == AF_INET6) {
589 			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
590 						   info, (u8 *)icmph);
591 #endif
592 		}
593 	}
594 	sk->sk_err = err;
595 	sk->sk_error_report(sk);
596 out:
597 	sock_put(sk);
598 }
599 EXPORT_SYMBOL_GPL(ping_err);
600 
601 /*
602  *	Copy and checksum an ICMP Echo packet from user space into a buffer
603  *	starting from the payload.
604  */
605 
606 int ping_getfrag(void *from, char *to,
607 		 int offset, int fraglen, int odd, struct sk_buff *skb)
608 {
609 	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
610 
611 	if (offset == 0) {
612 		fraglen -= sizeof(struct icmphdr);
613 		if (fraglen < 0)
614 			BUG();
615 		if (csum_and_copy_from_iter(to + sizeof(struct icmphdr),
616 			    fraglen, &pfh->wcheck,
617 			    &pfh->msg->msg_iter) != fraglen)
618 			return -EFAULT;
619 	} else if (offset < sizeof(struct icmphdr)) {
620 			BUG();
621 	} else {
622 		if (csum_and_copy_from_iter(to, fraglen, &pfh->wcheck,
623 					    &pfh->msg->msg_iter) != fraglen)
624 			return -EFAULT;
625 	}
626 
627 #if IS_ENABLED(CONFIG_IPV6)
628 	/* For IPv6, checksum each skb as we go along, as expected by
629 	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
630 	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
631 	 */
632 	if (pfh->family == AF_INET6) {
633 		skb->csum = pfh->wcheck;
634 		skb->ip_summed = CHECKSUM_NONE;
635 		pfh->wcheck = 0;
636 	}
637 #endif
638 
639 	return 0;
640 }
641 EXPORT_SYMBOL_GPL(ping_getfrag);
642 
643 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
644 				       struct flowi4 *fl4)
645 {
646 	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
647 
648 	pfh->wcheck = csum_partial((char *)&pfh->icmph,
649 		sizeof(struct icmphdr), pfh->wcheck);
650 	pfh->icmph.checksum = csum_fold(pfh->wcheck);
651 	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
652 	skb->ip_summed = CHECKSUM_NONE;
653 	return ip_push_pending_frames(sk, fl4);
654 }
655 
656 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
657 			void *user_icmph, size_t icmph_len) {
658 	u8 type, code;
659 
660 	if (len > 0xFFFF)
661 		return -EMSGSIZE;
662 
663 	/*
664 	 *	Check the flags.
665 	 */
666 
667 	/* Mirror BSD error message compatibility */
668 	if (msg->msg_flags & MSG_OOB)
669 		return -EOPNOTSUPP;
670 
671 	/*
672 	 *	Fetch the ICMP header provided by the userland.
673 	 *	iovec is modified! The ICMP header is consumed.
674 	 */
675 	if (memcpy_from_msg(user_icmph, msg, icmph_len))
676 		return -EFAULT;
677 
678 	if (family == AF_INET) {
679 		type = ((struct icmphdr *) user_icmph)->type;
680 		code = ((struct icmphdr *) user_icmph)->code;
681 #if IS_ENABLED(CONFIG_IPV6)
682 	} else if (family == AF_INET6) {
683 		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
684 		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
685 #endif
686 	} else {
687 		BUG();
688 	}
689 
690 	if (!ping_supported(family, type, code))
691 		return -EINVAL;
692 
693 	return 0;
694 }
695 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
696 
697 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
698 {
699 	struct net *net = sock_net(sk);
700 	struct flowi4 fl4;
701 	struct inet_sock *inet = inet_sk(sk);
702 	struct ipcm_cookie ipc;
703 	struct icmphdr user_icmph;
704 	struct pingfakehdr pfh;
705 	struct rtable *rt = NULL;
706 	struct ip_options_data opt_copy;
707 	int free = 0;
708 	__be32 saddr, daddr, faddr;
709 	u8  tos;
710 	int err;
711 
712 	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
713 
714 	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
715 				  sizeof(user_icmph));
716 	if (err)
717 		return err;
718 
719 	/*
720 	 *	Get and verify the address.
721 	 */
722 
723 	if (msg->msg_name) {
724 		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
725 		if (msg->msg_namelen < sizeof(*usin))
726 			return -EINVAL;
727 		if (usin->sin_family != AF_INET)
728 			return -EAFNOSUPPORT;
729 		daddr = usin->sin_addr.s_addr;
730 		/* no remote port */
731 	} else {
732 		if (sk->sk_state != TCP_ESTABLISHED)
733 			return -EDESTADDRREQ;
734 		daddr = inet->inet_daddr;
735 		/* no remote port */
736 	}
737 
738 	ipc.addr = inet->inet_saddr;
739 	ipc.opt = NULL;
740 	ipc.oif = sk->sk_bound_dev_if;
741 	ipc.tx_flags = 0;
742 	ipc.ttl = 0;
743 	ipc.tos = -1;
744 
745 	sock_tx_timestamp(sk, &ipc.tx_flags);
746 
747 	if (msg->msg_controllen) {
748 		err = ip_cmsg_send(sock_net(sk), msg, &ipc, false);
749 		if (err)
750 			return err;
751 		if (ipc.opt)
752 			free = 1;
753 	}
754 	if (!ipc.opt) {
755 		struct ip_options_rcu *inet_opt;
756 
757 		rcu_read_lock();
758 		inet_opt = rcu_dereference(inet->inet_opt);
759 		if (inet_opt) {
760 			memcpy(&opt_copy, inet_opt,
761 			       sizeof(*inet_opt) + inet_opt->opt.optlen);
762 			ipc.opt = &opt_copy.opt;
763 		}
764 		rcu_read_unlock();
765 	}
766 
767 	saddr = ipc.addr;
768 	ipc.addr = faddr = daddr;
769 
770 	if (ipc.opt && ipc.opt->opt.srr) {
771 		if (!daddr)
772 			return -EINVAL;
773 		faddr = ipc.opt->opt.faddr;
774 	}
775 	tos = get_rttos(&ipc, inet);
776 	if (sock_flag(sk, SOCK_LOCALROUTE) ||
777 	    (msg->msg_flags & MSG_DONTROUTE) ||
778 	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
779 		tos |= RTO_ONLINK;
780 	}
781 
782 	if (ipv4_is_multicast(daddr)) {
783 		if (!ipc.oif)
784 			ipc.oif = inet->mc_index;
785 		if (!saddr)
786 			saddr = inet->mc_addr;
787 	} else if (!ipc.oif)
788 		ipc.oif = inet->uc_index;
789 
790 	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
791 			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
792 			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
793 
794 	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
795 	rt = ip_route_output_flow(net, &fl4, sk);
796 	if (IS_ERR(rt)) {
797 		err = PTR_ERR(rt);
798 		rt = NULL;
799 		if (err == -ENETUNREACH)
800 			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
801 		goto out;
802 	}
803 
804 	err = -EACCES;
805 	if ((rt->rt_flags & RTCF_BROADCAST) &&
806 	    !sock_flag(sk, SOCK_BROADCAST))
807 		goto out;
808 
809 	if (msg->msg_flags & MSG_CONFIRM)
810 		goto do_confirm;
811 back_from_confirm:
812 
813 	if (!ipc.addr)
814 		ipc.addr = fl4.daddr;
815 
816 	lock_sock(sk);
817 
818 	pfh.icmph.type = user_icmph.type; /* already checked */
819 	pfh.icmph.code = user_icmph.code; /* ditto */
820 	pfh.icmph.checksum = 0;
821 	pfh.icmph.un.echo.id = inet->inet_sport;
822 	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
823 	pfh.msg = msg;
824 	pfh.wcheck = 0;
825 	pfh.family = AF_INET;
826 
827 	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
828 			0, &ipc, &rt, msg->msg_flags);
829 	if (err)
830 		ip_flush_pending_frames(sk);
831 	else
832 		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
833 	release_sock(sk);
834 
835 out:
836 	ip_rt_put(rt);
837 	if (free)
838 		kfree(ipc.opt);
839 	if (!err) {
840 		icmp_out_count(sock_net(sk), user_icmph.type);
841 		return len;
842 	}
843 	return err;
844 
845 do_confirm:
846 	dst_confirm(&rt->dst);
847 	if (!(msg->msg_flags & MSG_PROBE) || len)
848 		goto back_from_confirm;
849 	err = 0;
850 	goto out;
851 }
852 
853 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
854 		 int flags, int *addr_len)
855 {
856 	struct inet_sock *isk = inet_sk(sk);
857 	int family = sk->sk_family;
858 	struct sk_buff *skb;
859 	int copied, err;
860 
861 	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
862 
863 	err = -EOPNOTSUPP;
864 	if (flags & MSG_OOB)
865 		goto out;
866 
867 	if (flags & MSG_ERRQUEUE)
868 		return inet_recv_error(sk, msg, len, addr_len);
869 
870 	skb = skb_recv_datagram(sk, flags, noblock, &err);
871 	if (!skb)
872 		goto out;
873 
874 	copied = skb->len;
875 	if (copied > len) {
876 		msg->msg_flags |= MSG_TRUNC;
877 		copied = len;
878 	}
879 
880 	/* Don't bother checking the checksum */
881 	err = skb_copy_datagram_msg(skb, 0, msg, copied);
882 	if (err)
883 		goto done;
884 
885 	sock_recv_timestamp(msg, sk, skb);
886 
887 	/* Copy the address and add cmsg data. */
888 	if (family == AF_INET) {
889 		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
890 
891 		if (sin) {
892 			sin->sin_family = AF_INET;
893 			sin->sin_port = 0 /* skb->h.uh->source */;
894 			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
895 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
896 			*addr_len = sizeof(*sin);
897 		}
898 
899 		if (isk->cmsg_flags)
900 			ip_cmsg_recv(msg, skb);
901 
902 #if IS_ENABLED(CONFIG_IPV6)
903 	} else if (family == AF_INET6) {
904 		struct ipv6_pinfo *np = inet6_sk(sk);
905 		struct ipv6hdr *ip6 = ipv6_hdr(skb);
906 		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
907 
908 		if (sin6) {
909 			sin6->sin6_family = AF_INET6;
910 			sin6->sin6_port = 0;
911 			sin6->sin6_addr = ip6->saddr;
912 			sin6->sin6_flowinfo = 0;
913 			if (np->sndflow)
914 				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
915 			sin6->sin6_scope_id =
916 				ipv6_iface_scope_id(&sin6->sin6_addr,
917 						    inet6_iif(skb));
918 			*addr_len = sizeof(*sin6);
919 		}
920 
921 		if (inet6_sk(sk)->rxopt.all)
922 			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
923 		if (skb->protocol == htons(ETH_P_IPV6) &&
924 		    inet6_sk(sk)->rxopt.all)
925 			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
926 		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
927 			ip_cmsg_recv(msg, skb);
928 #endif
929 	} else {
930 		BUG();
931 	}
932 
933 	err = copied;
934 
935 done:
936 	skb_free_datagram(sk, skb);
937 out:
938 	pr_debug("ping_recvmsg -> %d\n", err);
939 	return err;
940 }
941 EXPORT_SYMBOL_GPL(ping_recvmsg);
942 
943 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
944 {
945 	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
946 		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
947 	if (sock_queue_rcv_skb(sk, skb) < 0) {
948 		kfree_skb(skb);
949 		pr_debug("ping_queue_rcv_skb -> failed\n");
950 		return -1;
951 	}
952 	return 0;
953 }
954 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
955 
956 
957 /*
958  *	All we need to do is get the socket.
959  */
960 
961 bool ping_rcv(struct sk_buff *skb)
962 {
963 	struct sock *sk;
964 	struct net *net = dev_net(skb->dev);
965 	struct icmphdr *icmph = icmp_hdr(skb);
966 
967 	/* We assume the packet has already been checked by icmp_rcv */
968 
969 	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
970 		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
971 
972 	/* Push ICMP header back */
973 	skb_push(skb, skb->data - (u8 *)icmph);
974 
975 	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
976 	if (sk) {
977 		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
978 
979 		pr_debug("rcv on socket %p\n", sk);
980 		if (skb2)
981 			ping_queue_rcv_skb(sk, skb2);
982 		sock_put(sk);
983 		return true;
984 	}
985 	pr_debug("no socket, dropping\n");
986 
987 	return false;
988 }
989 EXPORT_SYMBOL_GPL(ping_rcv);
990 
991 struct proto ping_prot = {
992 	.name =		"PING",
993 	.owner =	THIS_MODULE,
994 	.init =		ping_init_sock,
995 	.close =	ping_close,
996 	.connect =	ip4_datagram_connect,
997 	.disconnect =	udp_disconnect,
998 	.setsockopt =	ip_setsockopt,
999 	.getsockopt =	ip_getsockopt,
1000 	.sendmsg =	ping_v4_sendmsg,
1001 	.recvmsg =	ping_recvmsg,
1002 	.bind =		ping_bind,
1003 	.backlog_rcv =	ping_queue_rcv_skb,
1004 	.release_cb =	ip4_datagram_release_cb,
1005 	.hash =		ping_hash,
1006 	.unhash =	ping_unhash,
1007 	.get_port =	ping_get_port,
1008 	.obj_size =	sizeof(struct inet_sock),
1009 };
1010 EXPORT_SYMBOL(ping_prot);
1011 
1012 #ifdef CONFIG_PROC_FS
1013 
1014 static struct sock *ping_get_first(struct seq_file *seq, int start)
1015 {
1016 	struct sock *sk;
1017 	struct ping_iter_state *state = seq->private;
1018 	struct net *net = seq_file_net(seq);
1019 
1020 	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1021 	     ++state->bucket) {
1022 		struct hlist_nulls_node *node;
1023 		struct hlist_nulls_head *hslot;
1024 
1025 		hslot = &ping_table.hash[state->bucket];
1026 
1027 		if (hlist_nulls_empty(hslot))
1028 			continue;
1029 
1030 		sk_nulls_for_each(sk, node, hslot) {
1031 			if (net_eq(sock_net(sk), net) &&
1032 			    sk->sk_family == state->family)
1033 				goto found;
1034 		}
1035 	}
1036 	sk = NULL;
1037 found:
1038 	return sk;
1039 }
1040 
1041 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1042 {
1043 	struct ping_iter_state *state = seq->private;
1044 	struct net *net = seq_file_net(seq);
1045 
1046 	do {
1047 		sk = sk_nulls_next(sk);
1048 	} while (sk && (!net_eq(sock_net(sk), net)));
1049 
1050 	if (!sk)
1051 		return ping_get_first(seq, state->bucket + 1);
1052 	return sk;
1053 }
1054 
1055 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1056 {
1057 	struct sock *sk = ping_get_first(seq, 0);
1058 
1059 	if (sk)
1060 		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1061 			--pos;
1062 	return pos ? NULL : sk;
1063 }
1064 
1065 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1066 {
1067 	struct ping_iter_state *state = seq->private;
1068 	state->bucket = 0;
1069 	state->family = family;
1070 
1071 	read_lock_bh(&ping_table.lock);
1072 
1073 	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1074 }
1075 EXPORT_SYMBOL_GPL(ping_seq_start);
1076 
1077 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1078 {
1079 	return ping_seq_start(seq, pos, AF_INET);
1080 }
1081 
1082 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1083 {
1084 	struct sock *sk;
1085 
1086 	if (v == SEQ_START_TOKEN)
1087 		sk = ping_get_idx(seq, 0);
1088 	else
1089 		sk = ping_get_next(seq, v);
1090 
1091 	++*pos;
1092 	return sk;
1093 }
1094 EXPORT_SYMBOL_GPL(ping_seq_next);
1095 
1096 void ping_seq_stop(struct seq_file *seq, void *v)
1097 {
1098 	read_unlock_bh(&ping_table.lock);
1099 }
1100 EXPORT_SYMBOL_GPL(ping_seq_stop);
1101 
1102 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1103 		int bucket)
1104 {
1105 	struct inet_sock *inet = inet_sk(sp);
1106 	__be32 dest = inet->inet_daddr;
1107 	__be32 src = inet->inet_rcv_saddr;
1108 	__u16 destp = ntohs(inet->inet_dport);
1109 	__u16 srcp = ntohs(inet->inet_sport);
1110 
1111 	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1112 		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1113 		bucket, src, srcp, dest, destp, sp->sk_state,
1114 		sk_wmem_alloc_get(sp),
1115 		sk_rmem_alloc_get(sp),
1116 		0, 0L, 0,
1117 		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1118 		0, sock_i_ino(sp),
1119 		atomic_read(&sp->sk_refcnt), sp,
1120 		atomic_read(&sp->sk_drops));
1121 }
1122 
1123 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1124 {
1125 	seq_setwidth(seq, 127);
1126 	if (v == SEQ_START_TOKEN)
1127 		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1128 			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1129 			   "inode ref pointer drops");
1130 	else {
1131 		struct ping_iter_state *state = seq->private;
1132 
1133 		ping_v4_format_sock(v, seq, state->bucket);
1134 	}
1135 	seq_pad(seq, '\n');
1136 	return 0;
1137 }
1138 
1139 static const struct seq_operations ping_v4_seq_ops = {
1140 	.show		= ping_v4_seq_show,
1141 	.start		= ping_v4_seq_start,
1142 	.next		= ping_seq_next,
1143 	.stop		= ping_seq_stop,
1144 };
1145 
1146 static int ping_seq_open(struct inode *inode, struct file *file)
1147 {
1148 	struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1149 	return seq_open_net(inode, file, &afinfo->seq_ops,
1150 			   sizeof(struct ping_iter_state));
1151 }
1152 
1153 const struct file_operations ping_seq_fops = {
1154 	.open		= ping_seq_open,
1155 	.read		= seq_read,
1156 	.llseek		= seq_lseek,
1157 	.release	= seq_release_net,
1158 };
1159 EXPORT_SYMBOL_GPL(ping_seq_fops);
1160 
1161 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1162 	.name		= "icmp",
1163 	.family		= AF_INET,
1164 	.seq_fops	= &ping_seq_fops,
1165 	.seq_ops	= {
1166 		.start		= ping_v4_seq_start,
1167 		.show		= ping_v4_seq_show,
1168 		.next		= ping_seq_next,
1169 		.stop		= ping_seq_stop,
1170 	},
1171 };
1172 
1173 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1174 {
1175 	struct proc_dir_entry *p;
1176 	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1177 			     afinfo->seq_fops, afinfo);
1178 	if (!p)
1179 		return -ENOMEM;
1180 	return 0;
1181 }
1182 EXPORT_SYMBOL_GPL(ping_proc_register);
1183 
1184 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1185 {
1186 	remove_proc_entry(afinfo->name, net->proc_net);
1187 }
1188 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1189 
1190 static int __net_init ping_v4_proc_init_net(struct net *net)
1191 {
1192 	return ping_proc_register(net, &ping_v4_seq_afinfo);
1193 }
1194 
1195 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1196 {
1197 	ping_proc_unregister(net, &ping_v4_seq_afinfo);
1198 }
1199 
1200 static struct pernet_operations ping_v4_net_ops = {
1201 	.init = ping_v4_proc_init_net,
1202 	.exit = ping_v4_proc_exit_net,
1203 };
1204 
1205 int __init ping_proc_init(void)
1206 {
1207 	return register_pernet_subsys(&ping_v4_net_ops);
1208 }
1209 
1210 void ping_proc_exit(void)
1211 {
1212 	unregister_pernet_subsys(&ping_v4_net_ops);
1213 }
1214 
1215 #endif
1216 
1217 void __init ping_init(void)
1218 {
1219 	int i;
1220 
1221 	for (i = 0; i < PING_HTABLE_SIZE; i++)
1222 		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1223 	rwlock_init(&ping_table.lock);
1224 }
1225