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
56struct ping_table {
57	struct hlist_nulls_head	hash[PING_HTABLE_SIZE];
58	rwlock_t		lock;
59};
60
61static struct ping_table ping_table;
62struct pingv6_ops pingv6_ops;
63EXPORT_SYMBOL_GPL(pingv6_ops);
64
65static u16 ping_port_rover;
66
67static 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}
74EXPORT_SYMBOL_GPL(ping_hash);
75
76static 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
82int 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;
110next_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
142fail:
143	write_unlock_bh(&ping_table.lock);
144	return 1;
145}
146EXPORT_SYMBOL_GPL(ping_get_port);
147
148void 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
154void 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}
169EXPORT_SYMBOL_GPL(ping_unhash);
170
171static 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;
233exit:
234	read_unlock_bh(&ping_table.lock);
235
236	return sk;
237}
238
239static 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
254int 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
285out_release_group:
286	put_group_info(group_info);
287	return ret;
288}
289EXPORT_SYMBOL_GPL(ping_init_sock);
290
291void 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}
299EXPORT_SYMBOL_GPL(ping_close);
300
301/* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
302static 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 (!(isk->freebind || isk->transparent || has_addr ||
367		      addr_type == IPV6_ADDR_ANY))
368			return -EADDRNOTAVAIL;
369
370		if (scoped)
371			sk->sk_bound_dev_if = addr->sin6_scope_id;
372#endif
373	} else {
374		return -EAFNOSUPPORT;
375	}
376	return 0;
377}
378
379static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
380{
381	if (saddr->sa_family == AF_INET) {
382		struct inet_sock *isk = inet_sk(sk);
383		struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
384		isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
385#if IS_ENABLED(CONFIG_IPV6)
386	} else if (saddr->sa_family == AF_INET6) {
387		struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
388		struct ipv6_pinfo *np = inet6_sk(sk);
389		sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
390#endif
391	}
392}
393
394static void ping_clear_saddr(struct sock *sk, int dif)
395{
396	sk->sk_bound_dev_if = dif;
397	if (sk->sk_family == AF_INET) {
398		struct inet_sock *isk = inet_sk(sk);
399		isk->inet_rcv_saddr = isk->inet_saddr = 0;
400#if IS_ENABLED(CONFIG_IPV6)
401	} else if (sk->sk_family == AF_INET6) {
402		struct ipv6_pinfo *np = inet6_sk(sk);
403		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
404		memset(&np->saddr, 0, sizeof(np->saddr));
405#endif
406	}
407}
408/*
409 * We need our own bind because there are no privileged id's == local ports.
410 * Moreover, we don't allow binding to multi- and broadcast addresses.
411 */
412
413int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
414{
415	struct inet_sock *isk = inet_sk(sk);
416	unsigned short snum;
417	int err;
418	int dif = sk->sk_bound_dev_if;
419
420	err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
421	if (err)
422		return err;
423
424	lock_sock(sk);
425
426	err = -EINVAL;
427	if (isk->inet_num != 0)
428		goto out;
429
430	err = -EADDRINUSE;
431	ping_set_saddr(sk, uaddr);
432	snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
433	if (ping_get_port(sk, snum) != 0) {
434		ping_clear_saddr(sk, dif);
435		goto out;
436	}
437
438	pr_debug("after bind(): num = %d, dif = %d\n",
439		 (int)isk->inet_num,
440		 (int)sk->sk_bound_dev_if);
441
442	err = 0;
443	if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
444		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
445#if IS_ENABLED(CONFIG_IPV6)
446	if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
447		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
448#endif
449
450	if (snum)
451		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
452	isk->inet_sport = htons(isk->inet_num);
453	isk->inet_daddr = 0;
454	isk->inet_dport = 0;
455
456#if IS_ENABLED(CONFIG_IPV6)
457	if (sk->sk_family == AF_INET6)
458		memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
459#endif
460
461	sk_dst_reset(sk);
462out:
463	release_sock(sk);
464	pr_debug("ping_v4_bind -> %d\n", err);
465	return err;
466}
467EXPORT_SYMBOL_GPL(ping_bind);
468
469/*
470 * Is this a supported type of ICMP message?
471 */
472
473static inline int ping_supported(int family, int type, int code)
474{
475	return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
476	       (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
477}
478
479/*
480 * This routine is called by the ICMP module when it gets some
481 * sort of error condition.
482 */
483
484void ping_err(struct sk_buff *skb, int offset, u32 info)
485{
486	int family;
487	struct icmphdr *icmph;
488	struct inet_sock *inet_sock;
489	int type;
490	int code;
491	struct net *net = dev_net(skb->dev);
492	struct sock *sk;
493	int harderr;
494	int err;
495
496	if (skb->protocol == htons(ETH_P_IP)) {
497		family = AF_INET;
498		type = icmp_hdr(skb)->type;
499		code = icmp_hdr(skb)->code;
500		icmph = (struct icmphdr *)(skb->data + offset);
501	} else if (skb->protocol == htons(ETH_P_IPV6)) {
502		family = AF_INET6;
503		type = icmp6_hdr(skb)->icmp6_type;
504		code = icmp6_hdr(skb)->icmp6_code;
505		icmph = (struct icmphdr *) (skb->data + offset);
506	} else {
507		BUG();
508	}
509
510	/* We assume the packet has already been checked by icmp_unreach */
511
512	if (!ping_supported(family, icmph->type, icmph->code))
513		return;
514
515	pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
516		 skb->protocol, type, code, ntohs(icmph->un.echo.id),
517		 ntohs(icmph->un.echo.sequence));
518
519	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
520	if (!sk) {
521		pr_debug("no socket, dropping\n");
522		return;	/* No socket for error */
523	}
524	pr_debug("err on socket %p\n", sk);
525
526	err = 0;
527	harderr = 0;
528	inet_sock = inet_sk(sk);
529
530	if (skb->protocol == htons(ETH_P_IP)) {
531		switch (type) {
532		default:
533		case ICMP_TIME_EXCEEDED:
534			err = EHOSTUNREACH;
535			break;
536		case ICMP_SOURCE_QUENCH:
537			/* This is not a real error but ping wants to see it.
538			 * Report it with some fake errno.
539			 */
540			err = EREMOTEIO;
541			break;
542		case ICMP_PARAMETERPROB:
543			err = EPROTO;
544			harderr = 1;
545			break;
546		case ICMP_DEST_UNREACH:
547			if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
548				ipv4_sk_update_pmtu(skb, sk, info);
549				if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
550					err = EMSGSIZE;
551					harderr = 1;
552					break;
553				}
554				goto out;
555			}
556			err = EHOSTUNREACH;
557			if (code <= NR_ICMP_UNREACH) {
558				harderr = icmp_err_convert[code].fatal;
559				err = icmp_err_convert[code].errno;
560			}
561			break;
562		case ICMP_REDIRECT:
563			/* See ICMP_SOURCE_QUENCH */
564			ipv4_sk_redirect(skb, sk);
565			err = EREMOTEIO;
566			break;
567		}
568#if IS_ENABLED(CONFIG_IPV6)
569	} else if (skb->protocol == htons(ETH_P_IPV6)) {
570		harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
571#endif
572	}
573
574	/*
575	 *      RFC1122: OK.  Passes ICMP errors back to application, as per
576	 *	4.1.3.3.
577	 */
578	if ((family == AF_INET && !inet_sock->recverr) ||
579	    (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
580		if (!harderr || sk->sk_state != TCP_ESTABLISHED)
581			goto out;
582	} else {
583		if (family == AF_INET) {
584			ip_icmp_error(sk, skb, err, 0 /* no remote port */,
585				      info, (u8 *)icmph);
586#if IS_ENABLED(CONFIG_IPV6)
587		} else if (family == AF_INET6) {
588			pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
589						   info, (u8 *)icmph);
590#endif
591		}
592	}
593	sk->sk_err = err;
594	sk->sk_error_report(sk);
595out:
596	sock_put(sk);
597}
598EXPORT_SYMBOL_GPL(ping_err);
599
600/*
601 *	Copy and checksum an ICMP Echo packet from user space into a buffer
602 *	starting from the payload.
603 */
604
605int ping_getfrag(void *from, char *to,
606		 int offset, int fraglen, int odd, struct sk_buff *skb)
607{
608	struct pingfakehdr *pfh = (struct pingfakehdr *)from;
609
610	if (offset == 0) {
611		fraglen -= sizeof(struct icmphdr);
612		if (fraglen < 0)
613			BUG();
614		if (csum_and_copy_from_iter(to + sizeof(struct icmphdr),
615			    fraglen, &pfh->wcheck,
616			    &pfh->msg->msg_iter) != fraglen)
617			return -EFAULT;
618	} else if (offset < sizeof(struct icmphdr)) {
619			BUG();
620	} else {
621		if (csum_and_copy_from_iter(to, fraglen, &pfh->wcheck,
622					    &pfh->msg->msg_iter) != fraglen)
623			return -EFAULT;
624	}
625
626#if IS_ENABLED(CONFIG_IPV6)
627	/* For IPv6, checksum each skb as we go along, as expected by
628	 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
629	 * wcheck, it will be finalized in ping_v4_push_pending_frames.
630	 */
631	if (pfh->family == AF_INET6) {
632		skb->csum = pfh->wcheck;
633		skb->ip_summed = CHECKSUM_NONE;
634		pfh->wcheck = 0;
635	}
636#endif
637
638	return 0;
639}
640EXPORT_SYMBOL_GPL(ping_getfrag);
641
642static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
643				       struct flowi4 *fl4)
644{
645	struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
646
647	pfh->wcheck = csum_partial((char *)&pfh->icmph,
648		sizeof(struct icmphdr), pfh->wcheck);
649	pfh->icmph.checksum = csum_fold(pfh->wcheck);
650	memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
651	skb->ip_summed = CHECKSUM_NONE;
652	return ip_push_pending_frames(sk, fl4);
653}
654
655int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
656			void *user_icmph, size_t icmph_len) {
657	u8 type, code;
658
659	if (len > 0xFFFF)
660		return -EMSGSIZE;
661
662	/*
663	 *	Check the flags.
664	 */
665
666	/* Mirror BSD error message compatibility */
667	if (msg->msg_flags & MSG_OOB)
668		return -EOPNOTSUPP;
669
670	/*
671	 *	Fetch the ICMP header provided by the userland.
672	 *	iovec is modified! The ICMP header is consumed.
673	 */
674	if (memcpy_from_msg(user_icmph, msg, icmph_len))
675		return -EFAULT;
676
677	if (family == AF_INET) {
678		type = ((struct icmphdr *) user_icmph)->type;
679		code = ((struct icmphdr *) user_icmph)->code;
680#if IS_ENABLED(CONFIG_IPV6)
681	} else if (family == AF_INET6) {
682		type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
683		code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
684#endif
685	} else {
686		BUG();
687	}
688
689	if (!ping_supported(family, type, code))
690		return -EINVAL;
691
692	return 0;
693}
694EXPORT_SYMBOL_GPL(ping_common_sendmsg);
695
696static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
697{
698	struct net *net = sock_net(sk);
699	struct flowi4 fl4;
700	struct inet_sock *inet = inet_sk(sk);
701	struct ipcm_cookie ipc;
702	struct icmphdr user_icmph;
703	struct pingfakehdr pfh;
704	struct rtable *rt = NULL;
705	struct ip_options_data opt_copy;
706	int free = 0;
707	__be32 saddr, daddr, faddr;
708	u8  tos;
709	int err;
710
711	pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
712
713	err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
714				  sizeof(user_icmph));
715	if (err)
716		return err;
717
718	/*
719	 *	Get and verify the address.
720	 */
721
722	if (msg->msg_name) {
723		DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
724		if (msg->msg_namelen < sizeof(*usin))
725			return -EINVAL;
726		if (usin->sin_family != AF_INET)
727			return -EAFNOSUPPORT;
728		daddr = usin->sin_addr.s_addr;
729		/* no remote port */
730	} else {
731		if (sk->sk_state != TCP_ESTABLISHED)
732			return -EDESTADDRREQ;
733		daddr = inet->inet_daddr;
734		/* no remote port */
735	}
736
737	ipc.addr = inet->inet_saddr;
738	ipc.opt = NULL;
739	ipc.oif = sk->sk_bound_dev_if;
740	ipc.tx_flags = 0;
741	ipc.ttl = 0;
742	ipc.tos = -1;
743
744	sock_tx_timestamp(sk, &ipc.tx_flags);
745
746	if (msg->msg_controllen) {
747		err = ip_cmsg_send(sock_net(sk), msg, &ipc, false);
748		if (unlikely(err)) {
749			kfree(ipc.opt);
750			return err;
751		}
752		if (ipc.opt)
753			free = 1;
754	}
755	if (!ipc.opt) {
756		struct ip_options_rcu *inet_opt;
757
758		rcu_read_lock();
759		inet_opt = rcu_dereference(inet->inet_opt);
760		if (inet_opt) {
761			memcpy(&opt_copy, inet_opt,
762			       sizeof(*inet_opt) + inet_opt->opt.optlen);
763			ipc.opt = &opt_copy.opt;
764		}
765		rcu_read_unlock();
766	}
767
768	saddr = ipc.addr;
769	ipc.addr = faddr = daddr;
770
771	if (ipc.opt && ipc.opt->opt.srr) {
772		if (!daddr)
773			return -EINVAL;
774		faddr = ipc.opt->opt.faddr;
775	}
776	tos = get_rttos(&ipc, inet);
777	if (sock_flag(sk, SOCK_LOCALROUTE) ||
778	    (msg->msg_flags & MSG_DONTROUTE) ||
779	    (ipc.opt && ipc.opt->opt.is_strictroute)) {
780		tos |= RTO_ONLINK;
781	}
782
783	if (ipv4_is_multicast(daddr)) {
784		if (!ipc.oif)
785			ipc.oif = inet->mc_index;
786		if (!saddr)
787			saddr = inet->mc_addr;
788	} else if (!ipc.oif)
789		ipc.oif = inet->uc_index;
790
791	flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
792			   RT_SCOPE_UNIVERSE, sk->sk_protocol,
793			   inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
794
795	security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
796	rt = ip_route_output_flow(net, &fl4, sk);
797	if (IS_ERR(rt)) {
798		err = PTR_ERR(rt);
799		rt = NULL;
800		if (err == -ENETUNREACH)
801			IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
802		goto out;
803	}
804
805	err = -EACCES;
806	if ((rt->rt_flags & RTCF_BROADCAST) &&
807	    !sock_flag(sk, SOCK_BROADCAST))
808		goto out;
809
810	if (msg->msg_flags & MSG_CONFIRM)
811		goto do_confirm;
812back_from_confirm:
813
814	if (!ipc.addr)
815		ipc.addr = fl4.daddr;
816
817	lock_sock(sk);
818
819	pfh.icmph.type = user_icmph.type; /* already checked */
820	pfh.icmph.code = user_icmph.code; /* ditto */
821	pfh.icmph.checksum = 0;
822	pfh.icmph.un.echo.id = inet->inet_sport;
823	pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
824	pfh.msg = msg;
825	pfh.wcheck = 0;
826	pfh.family = AF_INET;
827
828	err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
829			0, &ipc, &rt, msg->msg_flags);
830	if (err)
831		ip_flush_pending_frames(sk);
832	else
833		err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
834	release_sock(sk);
835
836out:
837	ip_rt_put(rt);
838	if (free)
839		kfree(ipc.opt);
840	if (!err) {
841		icmp_out_count(sock_net(sk), user_icmph.type);
842		return len;
843	}
844	return err;
845
846do_confirm:
847	dst_confirm(&rt->dst);
848	if (!(msg->msg_flags & MSG_PROBE) || len)
849		goto back_from_confirm;
850	err = 0;
851	goto out;
852}
853
854int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
855		 int flags, int *addr_len)
856{
857	struct inet_sock *isk = inet_sk(sk);
858	int family = sk->sk_family;
859	struct sk_buff *skb;
860	int copied, err;
861
862	pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
863
864	err = -EOPNOTSUPP;
865	if (flags & MSG_OOB)
866		goto out;
867
868	if (flags & MSG_ERRQUEUE)
869		return inet_recv_error(sk, msg, len, addr_len);
870
871	skb = skb_recv_datagram(sk, flags, noblock, &err);
872	if (!skb)
873		goto out;
874
875	copied = skb->len;
876	if (copied > len) {
877		msg->msg_flags |= MSG_TRUNC;
878		copied = len;
879	}
880
881	/* Don't bother checking the checksum */
882	err = skb_copy_datagram_msg(skb, 0, msg, copied);
883	if (err)
884		goto done;
885
886	sock_recv_timestamp(msg, sk, skb);
887
888	/* Copy the address and add cmsg data. */
889	if (family == AF_INET) {
890		DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
891
892		if (sin) {
893			sin->sin_family = AF_INET;
894			sin->sin_port = 0 /* skb->h.uh->source */;
895			sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
896			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
897			*addr_len = sizeof(*sin);
898		}
899
900		if (isk->cmsg_flags)
901			ip_cmsg_recv(msg, skb);
902
903#if IS_ENABLED(CONFIG_IPV6)
904	} else if (family == AF_INET6) {
905		struct ipv6_pinfo *np = inet6_sk(sk);
906		struct ipv6hdr *ip6 = ipv6_hdr(skb);
907		DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
908
909		if (sin6) {
910			sin6->sin6_family = AF_INET6;
911			sin6->sin6_port = 0;
912			sin6->sin6_addr = ip6->saddr;
913			sin6->sin6_flowinfo = 0;
914			if (np->sndflow)
915				sin6->sin6_flowinfo = ip6_flowinfo(ip6);
916			sin6->sin6_scope_id =
917				ipv6_iface_scope_id(&sin6->sin6_addr,
918						    inet6_iif(skb));
919			*addr_len = sizeof(*sin6);
920		}
921
922		if (inet6_sk(sk)->rxopt.all)
923			pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
924		if (skb->protocol == htons(ETH_P_IPV6) &&
925		    inet6_sk(sk)->rxopt.all)
926			pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
927		else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
928			ip_cmsg_recv(msg, skb);
929#endif
930	} else {
931		BUG();
932	}
933
934	err = copied;
935
936done:
937	skb_free_datagram(sk, skb);
938out:
939	pr_debug("ping_recvmsg -> %d\n", err);
940	return err;
941}
942EXPORT_SYMBOL_GPL(ping_recvmsg);
943
944int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
945{
946	pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
947		 inet_sk(sk), inet_sk(sk)->inet_num, skb);
948	if (sock_queue_rcv_skb(sk, skb) < 0) {
949		kfree_skb(skb);
950		pr_debug("ping_queue_rcv_skb -> failed\n");
951		return -1;
952	}
953	return 0;
954}
955EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
956
957
958/*
959 *	All we need to do is get the socket.
960 */
961
962bool ping_rcv(struct sk_buff *skb)
963{
964	struct sock *sk;
965	struct net *net = dev_net(skb->dev);
966	struct icmphdr *icmph = icmp_hdr(skb);
967
968	/* We assume the packet has already been checked by icmp_rcv */
969
970	pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
971		 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
972
973	/* Push ICMP header back */
974	skb_push(skb, skb->data - (u8 *)icmph);
975
976	sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
977	if (sk) {
978		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
979
980		pr_debug("rcv on socket %p\n", sk);
981		if (skb2)
982			ping_queue_rcv_skb(sk, skb2);
983		sock_put(sk);
984		return true;
985	}
986	pr_debug("no socket, dropping\n");
987
988	return false;
989}
990EXPORT_SYMBOL_GPL(ping_rcv);
991
992struct proto ping_prot = {
993	.name =		"PING",
994	.owner =	THIS_MODULE,
995	.init =		ping_init_sock,
996	.close =	ping_close,
997	.connect =	ip4_datagram_connect,
998	.disconnect =	udp_disconnect,
999	.setsockopt =	ip_setsockopt,
1000	.getsockopt =	ip_getsockopt,
1001	.sendmsg =	ping_v4_sendmsg,
1002	.recvmsg =	ping_recvmsg,
1003	.bind =		ping_bind,
1004	.backlog_rcv =	ping_queue_rcv_skb,
1005	.release_cb =	ip4_datagram_release_cb,
1006	.hash =		ping_hash,
1007	.unhash =	ping_unhash,
1008	.get_port =	ping_get_port,
1009	.obj_size =	sizeof(struct inet_sock),
1010};
1011EXPORT_SYMBOL(ping_prot);
1012
1013#ifdef CONFIG_PROC_FS
1014
1015static struct sock *ping_get_first(struct seq_file *seq, int start)
1016{
1017	struct sock *sk;
1018	struct ping_iter_state *state = seq->private;
1019	struct net *net = seq_file_net(seq);
1020
1021	for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1022	     ++state->bucket) {
1023		struct hlist_nulls_node *node;
1024		struct hlist_nulls_head *hslot;
1025
1026		hslot = &ping_table.hash[state->bucket];
1027
1028		if (hlist_nulls_empty(hslot))
1029			continue;
1030
1031		sk_nulls_for_each(sk, node, hslot) {
1032			if (net_eq(sock_net(sk), net) &&
1033			    sk->sk_family == state->family)
1034				goto found;
1035		}
1036	}
1037	sk = NULL;
1038found:
1039	return sk;
1040}
1041
1042static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1043{
1044	struct ping_iter_state *state = seq->private;
1045	struct net *net = seq_file_net(seq);
1046
1047	do {
1048		sk = sk_nulls_next(sk);
1049	} while (sk && (!net_eq(sock_net(sk), net)));
1050
1051	if (!sk)
1052		return ping_get_first(seq, state->bucket + 1);
1053	return sk;
1054}
1055
1056static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1057{
1058	struct sock *sk = ping_get_first(seq, 0);
1059
1060	if (sk)
1061		while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1062			--pos;
1063	return pos ? NULL : sk;
1064}
1065
1066void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1067{
1068	struct ping_iter_state *state = seq->private;
1069	state->bucket = 0;
1070	state->family = family;
1071
1072	read_lock_bh(&ping_table.lock);
1073
1074	return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1075}
1076EXPORT_SYMBOL_GPL(ping_seq_start);
1077
1078static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1079{
1080	return ping_seq_start(seq, pos, AF_INET);
1081}
1082
1083void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1084{
1085	struct sock *sk;
1086
1087	if (v == SEQ_START_TOKEN)
1088		sk = ping_get_idx(seq, 0);
1089	else
1090		sk = ping_get_next(seq, v);
1091
1092	++*pos;
1093	return sk;
1094}
1095EXPORT_SYMBOL_GPL(ping_seq_next);
1096
1097void ping_seq_stop(struct seq_file *seq, void *v)
1098{
1099	read_unlock_bh(&ping_table.lock);
1100}
1101EXPORT_SYMBOL_GPL(ping_seq_stop);
1102
1103static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1104		int bucket)
1105{
1106	struct inet_sock *inet = inet_sk(sp);
1107	__be32 dest = inet->inet_daddr;
1108	__be32 src = inet->inet_rcv_saddr;
1109	__u16 destp = ntohs(inet->inet_dport);
1110	__u16 srcp = ntohs(inet->inet_sport);
1111
1112	seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1113		" %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1114		bucket, src, srcp, dest, destp, sp->sk_state,
1115		sk_wmem_alloc_get(sp),
1116		sk_rmem_alloc_get(sp),
1117		0, 0L, 0,
1118		from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1119		0, sock_i_ino(sp),
1120		atomic_read(&sp->sk_refcnt), sp,
1121		atomic_read(&sp->sk_drops));
1122}
1123
1124static int ping_v4_seq_show(struct seq_file *seq, void *v)
1125{
1126	seq_setwidth(seq, 127);
1127	if (v == SEQ_START_TOKEN)
1128		seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1129			   "rx_queue tr tm->when retrnsmt   uid  timeout "
1130			   "inode ref pointer drops");
1131	else {
1132		struct ping_iter_state *state = seq->private;
1133
1134		ping_v4_format_sock(v, seq, state->bucket);
1135	}
1136	seq_pad(seq, '\n');
1137	return 0;
1138}
1139
1140static const struct seq_operations ping_v4_seq_ops = {
1141	.show		= ping_v4_seq_show,
1142	.start		= ping_v4_seq_start,
1143	.next		= ping_seq_next,
1144	.stop		= ping_seq_stop,
1145};
1146
1147static int ping_seq_open(struct inode *inode, struct file *file)
1148{
1149	struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1150	return seq_open_net(inode, file, &afinfo->seq_ops,
1151			   sizeof(struct ping_iter_state));
1152}
1153
1154const struct file_operations ping_seq_fops = {
1155	.open		= ping_seq_open,
1156	.read		= seq_read,
1157	.llseek		= seq_lseek,
1158	.release	= seq_release_net,
1159};
1160EXPORT_SYMBOL_GPL(ping_seq_fops);
1161
1162static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1163	.name		= "icmp",
1164	.family		= AF_INET,
1165	.seq_fops	= &ping_seq_fops,
1166	.seq_ops	= {
1167		.start		= ping_v4_seq_start,
1168		.show		= ping_v4_seq_show,
1169		.next		= ping_seq_next,
1170		.stop		= ping_seq_stop,
1171	},
1172};
1173
1174int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1175{
1176	struct proc_dir_entry *p;
1177	p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1178			     afinfo->seq_fops, afinfo);
1179	if (!p)
1180		return -ENOMEM;
1181	return 0;
1182}
1183EXPORT_SYMBOL_GPL(ping_proc_register);
1184
1185void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1186{
1187	remove_proc_entry(afinfo->name, net->proc_net);
1188}
1189EXPORT_SYMBOL_GPL(ping_proc_unregister);
1190
1191static int __net_init ping_v4_proc_init_net(struct net *net)
1192{
1193	return ping_proc_register(net, &ping_v4_seq_afinfo);
1194}
1195
1196static void __net_exit ping_v4_proc_exit_net(struct net *net)
1197{
1198	ping_proc_unregister(net, &ping_v4_seq_afinfo);
1199}
1200
1201static struct pernet_operations ping_v4_net_ops = {
1202	.init = ping_v4_proc_init_net,
1203	.exit = ping_v4_proc_exit_net,
1204};
1205
1206int __init ping_proc_init(void)
1207{
1208	return register_pernet_subsys(&ping_v4_net_ops);
1209}
1210
1211void ping_proc_exit(void)
1212{
1213	unregister_pernet_subsys(&ping_v4_net_ops);
1214}
1215
1216#endif
1217
1218void __init ping_init(void)
1219{
1220	int i;
1221
1222	for (i = 0; i < PING_HTABLE_SIZE; i++)
1223		INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1224	rwlock_init(&ping_table.lock);
1225}
1226