1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * "Ping" sockets
8 *
9 * Based on ipv4/udp.c code.
10 *
11 * Authors: Vasiliy Kulikov / Openwall (for Linux 2.6),
12 * Pavel Kankovsky (for Linux 2.4.32)
13 *
14 * Pavel gave all rights to bugs to Vasiliy,
15 * none of the bugs are Pavel's now.
16 */
17
18 #include <linux/uaccess.h>
19 #include <linux/types.h>
20 #include <linux/fcntl.h>
21 #include <linux/socket.h>
22 #include <linux/sockios.h>
23 #include <linux/in.h>
24 #include <linux/errno.h>
25 #include <linux/timer.h>
26 #include <linux/mm.h>
27 #include <linux/inet.h>
28 #include <linux/netdevice.h>
29 #include <net/snmp.h>
30 #include <net/ip.h>
31 #include <net/icmp.h>
32 #include <net/protocol.h>
33 #include <linux/skbuff.h>
34 #include <linux/proc_fs.h>
35 #include <linux/export.h>
36 #include <net/sock.h>
37 #include <net/ping.h>
38 #include <net/udp.h>
39 #include <net/route.h>
40 #include <net/inet_common.h>
41 #include <net/checksum.h>
42
43 #if IS_ENABLED(CONFIG_IPV6)
44 #include <linux/in6.h>
45 #include <linux/icmpv6.h>
46 #include <net/addrconf.h>
47 #include <net/ipv6.h>
48 #include <net/transp_v6.h>
49 #endif
50
51 struct ping_table {
52 struct hlist_nulls_head hash[PING_HTABLE_SIZE];
53 rwlock_t lock;
54 };
55
56 static struct ping_table ping_table;
57 struct pingv6_ops pingv6_ops;
58 EXPORT_SYMBOL_GPL(pingv6_ops);
59
60 static u16 ping_port_rover;
61
ping_hashfn(const struct net * net,u32 num,u32 mask)62 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
63 {
64 u32 res = (num + net_hash_mix(net)) & mask;
65
66 pr_debug("hash(%u) = %u\n", num, res);
67 return res;
68 }
69 EXPORT_SYMBOL_GPL(ping_hash);
70
ping_hashslot(struct ping_table * table,struct net * net,unsigned int num)71 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
72 struct net *net, unsigned int num)
73 {
74 return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
75 }
76
ping_get_port(struct sock * sk,unsigned short ident)77 int ping_get_port(struct sock *sk, unsigned short ident)
78 {
79 struct hlist_nulls_node *node;
80 struct hlist_nulls_head *hlist;
81 struct inet_sock *isk, *isk2;
82 struct sock *sk2 = NULL;
83
84 isk = inet_sk(sk);
85 write_lock_bh(&ping_table.lock);
86 if (ident == 0) {
87 u32 i;
88 u16 result = ping_port_rover + 1;
89
90 for (i = 0; i < (1L << 16); i++, result++) {
91 if (!result)
92 result++; /* avoid zero */
93 hlist = ping_hashslot(&ping_table, sock_net(sk),
94 result);
95 ping_portaddr_for_each_entry(sk2, node, hlist) {
96 isk2 = inet_sk(sk2);
97
98 if (isk2->inet_num == result)
99 goto next_port;
100 }
101
102 /* found */
103 ping_port_rover = ident = result;
104 break;
105 next_port:
106 ;
107 }
108 if (i >= (1L << 16))
109 goto fail;
110 } else {
111 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
112 ping_portaddr_for_each_entry(sk2, node, hlist) {
113 isk2 = inet_sk(sk2);
114
115 /* BUG? Why is this reuse and not reuseaddr? ping.c
116 * doesn't turn off SO_REUSEADDR, and it doesn't expect
117 * that other ping processes can steal its packets.
118 */
119 if ((isk2->inet_num == ident) &&
120 (sk2 != sk) &&
121 (!sk2->sk_reuse || !sk->sk_reuse))
122 goto fail;
123 }
124 }
125
126 pr_debug("found port/ident = %d\n", ident);
127 isk->inet_num = ident;
128 if (sk_unhashed(sk)) {
129 pr_debug("was not hashed\n");
130 sock_hold(sk);
131 hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
132 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
133 }
134 write_unlock_bh(&ping_table.lock);
135 return 0;
136
137 fail:
138 write_unlock_bh(&ping_table.lock);
139 return 1;
140 }
141 EXPORT_SYMBOL_GPL(ping_get_port);
142
ping_hash(struct sock * sk)143 int ping_hash(struct sock *sk)
144 {
145 pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
146 BUG(); /* "Please do not press this button again." */
147
148 return 0;
149 }
150
ping_unhash(struct sock * sk)151 void ping_unhash(struct sock *sk)
152 {
153 struct inet_sock *isk = inet_sk(sk);
154
155 pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
156 write_lock_bh(&ping_table.lock);
157 if (sk_hashed(sk)) {
158 hlist_nulls_del(&sk->sk_nulls_node);
159 sk_nulls_node_init(&sk->sk_nulls_node);
160 sock_put(sk);
161 isk->inet_num = 0;
162 isk->inet_sport = 0;
163 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
164 }
165 write_unlock_bh(&ping_table.lock);
166 }
167 EXPORT_SYMBOL_GPL(ping_unhash);
168
ping_lookup(struct net * net,struct sk_buff * skb,u16 ident)169 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
170 {
171 struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
172 struct sock *sk = NULL;
173 struct inet_sock *isk;
174 struct hlist_nulls_node *hnode;
175 int dif = skb->dev->ifindex;
176
177 if (skb->protocol == htons(ETH_P_IP)) {
178 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
179 (int)ident, &ip_hdr(skb)->daddr, dif);
180 #if IS_ENABLED(CONFIG_IPV6)
181 } else if (skb->protocol == htons(ETH_P_IPV6)) {
182 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
183 (int)ident, &ipv6_hdr(skb)->daddr, dif);
184 #endif
185 }
186
187 read_lock_bh(&ping_table.lock);
188
189 ping_portaddr_for_each_entry(sk, hnode, hslot) {
190 isk = inet_sk(sk);
191
192 pr_debug("iterate\n");
193 if (isk->inet_num != ident)
194 continue;
195
196 if (skb->protocol == htons(ETH_P_IP) &&
197 sk->sk_family == AF_INET) {
198 pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
199 (int) isk->inet_num, &isk->inet_rcv_saddr,
200 sk->sk_bound_dev_if);
201
202 if (isk->inet_rcv_saddr &&
203 isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
204 continue;
205 #if IS_ENABLED(CONFIG_IPV6)
206 } else if (skb->protocol == htons(ETH_P_IPV6) &&
207 sk->sk_family == AF_INET6) {
208
209 pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
210 (int) isk->inet_num,
211 &sk->sk_v6_rcv_saddr,
212 sk->sk_bound_dev_if);
213
214 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
215 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
216 &ipv6_hdr(skb)->daddr))
217 continue;
218 #endif
219 } else {
220 continue;
221 }
222
223 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
224 sk->sk_bound_dev_if != inet_sdif(skb))
225 continue;
226
227 sock_hold(sk);
228 goto exit;
229 }
230
231 sk = NULL;
232 exit:
233 read_unlock_bh(&ping_table.lock);
234
235 return sk;
236 }
237
inet_get_ping_group_range_net(struct net * net,kgid_t * low,kgid_t * high)238 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
239 kgid_t *high)
240 {
241 kgid_t *data = net->ipv4.ping_group_range.range;
242 unsigned int seq;
243
244 do {
245 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
246
247 *low = data[0];
248 *high = data[1];
249 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
250 }
251
252
ping_init_sock(struct sock * sk)253 int ping_init_sock(struct sock *sk)
254 {
255 struct net *net = sock_net(sk);
256 kgid_t group = current_egid();
257 struct group_info *group_info;
258 int i;
259 kgid_t low, high;
260 int ret = 0;
261
262 if (sk->sk_family == AF_INET6)
263 sk->sk_ipv6only = 1;
264
265 inet_get_ping_group_range_net(net, &low, &high);
266 if (gid_lte(low, group) && gid_lte(group, high))
267 return 0;
268
269 group_info = get_current_groups();
270 for (i = 0; i < group_info->ngroups; i++) {
271 kgid_t gid = group_info->gid[i];
272
273 if (gid_lte(low, gid) && gid_lte(gid, high))
274 goto out_release_group;
275 }
276
277 ret = -EACCES;
278
279 out_release_group:
280 put_group_info(group_info);
281 return ret;
282 }
283 EXPORT_SYMBOL_GPL(ping_init_sock);
284
ping_close(struct sock * sk,long timeout)285 void ping_close(struct sock *sk, long timeout)
286 {
287 pr_debug("ping_close(sk=%p,sk->num=%u)\n",
288 inet_sk(sk), inet_sk(sk)->inet_num);
289 pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
290
291 sk_common_release(sk);
292 }
293 EXPORT_SYMBOL_GPL(ping_close);
294
295 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
ping_check_bind_addr(struct sock * sk,struct inet_sock * isk,struct sockaddr * uaddr,int addr_len)296 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
297 struct sockaddr *uaddr, int addr_len)
298 {
299 struct net *net = sock_net(sk);
300 if (sk->sk_family == AF_INET) {
301 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
302 int chk_addr_ret;
303
304 if (addr_len < sizeof(*addr))
305 return -EINVAL;
306
307 if (addr->sin_family != AF_INET &&
308 !(addr->sin_family == AF_UNSPEC &&
309 addr->sin_addr.s_addr == htonl(INADDR_ANY)))
310 return -EAFNOSUPPORT;
311
312 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
313 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
314
315 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
316 chk_addr_ret = RTN_LOCAL;
317 else
318 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
319
320 if ((!inet_can_nonlocal_bind(net, isk) &&
321 chk_addr_ret != RTN_LOCAL) ||
322 chk_addr_ret == RTN_MULTICAST ||
323 chk_addr_ret == RTN_BROADCAST)
324 return -EADDRNOTAVAIL;
325
326 #if IS_ENABLED(CONFIG_IPV6)
327 } else if (sk->sk_family == AF_INET6) {
328 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
329 int addr_type, scoped, has_addr;
330 struct net_device *dev = NULL;
331
332 if (addr_len < sizeof(*addr))
333 return -EINVAL;
334
335 if (addr->sin6_family != AF_INET6)
336 return -EAFNOSUPPORT;
337
338 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
339 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
340
341 addr_type = ipv6_addr_type(&addr->sin6_addr);
342 scoped = __ipv6_addr_needs_scope_id(addr_type);
343 if ((addr_type != IPV6_ADDR_ANY &&
344 !(addr_type & IPV6_ADDR_UNICAST)) ||
345 (scoped && !addr->sin6_scope_id))
346 return -EINVAL;
347
348 rcu_read_lock();
349 if (addr->sin6_scope_id) {
350 dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
351 if (!dev) {
352 rcu_read_unlock();
353 return -ENODEV;
354 }
355 }
356 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
357 scoped);
358 rcu_read_unlock();
359
360 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
361 addr_type == IPV6_ADDR_ANY))
362 return -EADDRNOTAVAIL;
363
364 if (scoped)
365 sk->sk_bound_dev_if = addr->sin6_scope_id;
366 #endif
367 } else {
368 return -EAFNOSUPPORT;
369 }
370 return 0;
371 }
372
ping_set_saddr(struct sock * sk,struct sockaddr * saddr)373 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
374 {
375 if (saddr->sa_family == AF_INET) {
376 struct inet_sock *isk = inet_sk(sk);
377 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
378 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
379 #if IS_ENABLED(CONFIG_IPV6)
380 } else if (saddr->sa_family == AF_INET6) {
381 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
382 struct ipv6_pinfo *np = inet6_sk(sk);
383 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
384 #endif
385 }
386 }
387
388 /*
389 * We need our own bind because there are no privileged id's == local ports.
390 * Moreover, we don't allow binding to multi- and broadcast addresses.
391 */
392
ping_bind(struct sock * sk,struct sockaddr * uaddr,int addr_len)393 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
394 {
395 struct inet_sock *isk = inet_sk(sk);
396 unsigned short snum;
397 int err;
398 int dif = sk->sk_bound_dev_if;
399
400 err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
401 if (err)
402 return err;
403
404 lock_sock(sk);
405
406 err = -EINVAL;
407 if (isk->inet_num != 0)
408 goto out;
409
410 err = -EADDRINUSE;
411 snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
412 if (ping_get_port(sk, snum) != 0) {
413 /* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
414 sk->sk_bound_dev_if = dif;
415 goto out;
416 }
417 ping_set_saddr(sk, uaddr);
418
419 pr_debug("after bind(): num = %hu, dif = %d\n",
420 isk->inet_num,
421 sk->sk_bound_dev_if);
422
423 err = 0;
424 if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
425 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
426 #if IS_ENABLED(CONFIG_IPV6)
427 if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
428 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
429 #endif
430
431 if (snum)
432 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
433 isk->inet_sport = htons(isk->inet_num);
434 isk->inet_daddr = 0;
435 isk->inet_dport = 0;
436
437 #if IS_ENABLED(CONFIG_IPV6)
438 if (sk->sk_family == AF_INET6)
439 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
440 #endif
441
442 sk_dst_reset(sk);
443 out:
444 release_sock(sk);
445 pr_debug("ping_v4_bind -> %d\n", err);
446 return err;
447 }
448 EXPORT_SYMBOL_GPL(ping_bind);
449
450 /*
451 * Is this a supported type of ICMP message?
452 */
453
ping_supported(int family,int type,int code)454 static inline int ping_supported(int family, int type, int code)
455 {
456 return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
457 (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
458 }
459
460 /*
461 * This routine is called by the ICMP module when it gets some
462 * sort of error condition.
463 */
464
ping_err(struct sk_buff * skb,int offset,u32 info)465 void ping_err(struct sk_buff *skb, int offset, u32 info)
466 {
467 int family;
468 struct icmphdr *icmph;
469 struct inet_sock *inet_sock;
470 int type;
471 int code;
472 struct net *net = dev_net(skb->dev);
473 struct sock *sk;
474 int harderr;
475 int err;
476
477 if (skb->protocol == htons(ETH_P_IP)) {
478 family = AF_INET;
479 type = icmp_hdr(skb)->type;
480 code = icmp_hdr(skb)->code;
481 icmph = (struct icmphdr *)(skb->data + offset);
482 } else if (skb->protocol == htons(ETH_P_IPV6)) {
483 family = AF_INET6;
484 type = icmp6_hdr(skb)->icmp6_type;
485 code = icmp6_hdr(skb)->icmp6_code;
486 icmph = (struct icmphdr *) (skb->data + offset);
487 } else {
488 BUG();
489 }
490
491 /* We assume the packet has already been checked by icmp_unreach */
492
493 if (!ping_supported(family, icmph->type, icmph->code))
494 return;
495
496 pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
497 skb->protocol, type, code, ntohs(icmph->un.echo.id),
498 ntohs(icmph->un.echo.sequence));
499
500 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
501 if (!sk) {
502 pr_debug("no socket, dropping\n");
503 return; /* No socket for error */
504 }
505 pr_debug("err on socket %p\n", sk);
506
507 err = 0;
508 harderr = 0;
509 inet_sock = inet_sk(sk);
510
511 if (skb->protocol == htons(ETH_P_IP)) {
512 switch (type) {
513 default:
514 case ICMP_TIME_EXCEEDED:
515 err = EHOSTUNREACH;
516 break;
517 case ICMP_SOURCE_QUENCH:
518 /* This is not a real error but ping wants to see it.
519 * Report it with some fake errno.
520 */
521 err = EREMOTEIO;
522 break;
523 case ICMP_PARAMETERPROB:
524 err = EPROTO;
525 harderr = 1;
526 break;
527 case ICMP_DEST_UNREACH:
528 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
529 ipv4_sk_update_pmtu(skb, sk, info);
530 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
531 err = EMSGSIZE;
532 harderr = 1;
533 break;
534 }
535 goto out;
536 }
537 err = EHOSTUNREACH;
538 if (code <= NR_ICMP_UNREACH) {
539 harderr = icmp_err_convert[code].fatal;
540 err = icmp_err_convert[code].errno;
541 }
542 break;
543 case ICMP_REDIRECT:
544 /* See ICMP_SOURCE_QUENCH */
545 ipv4_sk_redirect(skb, sk);
546 err = EREMOTEIO;
547 break;
548 }
549 #if IS_ENABLED(CONFIG_IPV6)
550 } else if (skb->protocol == htons(ETH_P_IPV6)) {
551 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
552 #endif
553 }
554
555 /*
556 * RFC1122: OK. Passes ICMP errors back to application, as per
557 * 4.1.3.3.
558 */
559 if ((family == AF_INET && !inet_sock->recverr) ||
560 (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
561 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
562 goto out;
563 } else {
564 if (family == AF_INET) {
565 ip_icmp_error(sk, skb, err, 0 /* no remote port */,
566 info, (u8 *)icmph);
567 #if IS_ENABLED(CONFIG_IPV6)
568 } else if (family == AF_INET6) {
569 pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
570 info, (u8 *)icmph);
571 #endif
572 }
573 }
574 sk->sk_err = err;
575 sk->sk_error_report(sk);
576 out:
577 sock_put(sk);
578 }
579 EXPORT_SYMBOL_GPL(ping_err);
580
581 /*
582 * Copy and checksum an ICMP Echo packet from user space into a buffer
583 * starting from the payload.
584 */
585
ping_getfrag(void * from,char * to,int offset,int fraglen,int odd,struct sk_buff * skb)586 int ping_getfrag(void *from, char *to,
587 int offset, int fraglen, int odd, struct sk_buff *skb)
588 {
589 struct pingfakehdr *pfh = (struct pingfakehdr *)from;
590
591 if (offset == 0) {
592 fraglen -= sizeof(struct icmphdr);
593 if (fraglen < 0)
594 BUG();
595 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
596 fraglen, &pfh->wcheck,
597 &pfh->msg->msg_iter))
598 return -EFAULT;
599 } else if (offset < sizeof(struct icmphdr)) {
600 BUG();
601 } else {
602 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
603 &pfh->msg->msg_iter))
604 return -EFAULT;
605 }
606
607 #if IS_ENABLED(CONFIG_IPV6)
608 /* For IPv6, checksum each skb as we go along, as expected by
609 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
610 * wcheck, it will be finalized in ping_v4_push_pending_frames.
611 */
612 if (pfh->family == AF_INET6) {
613 skb->csum = pfh->wcheck;
614 skb->ip_summed = CHECKSUM_NONE;
615 pfh->wcheck = 0;
616 }
617 #endif
618
619 return 0;
620 }
621 EXPORT_SYMBOL_GPL(ping_getfrag);
622
ping_v4_push_pending_frames(struct sock * sk,struct pingfakehdr * pfh,struct flowi4 * fl4)623 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
624 struct flowi4 *fl4)
625 {
626 struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
627
628 if (!skb)
629 return 0;
630 pfh->wcheck = csum_partial((char *)&pfh->icmph,
631 sizeof(struct icmphdr), pfh->wcheck);
632 pfh->icmph.checksum = csum_fold(pfh->wcheck);
633 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
634 skb->ip_summed = CHECKSUM_NONE;
635 return ip_push_pending_frames(sk, fl4);
636 }
637
ping_common_sendmsg(int family,struct msghdr * msg,size_t len,void * user_icmph,size_t icmph_len)638 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
639 void *user_icmph, size_t icmph_len)
640 {
641 u8 type, code;
642
643 if (len > 0xFFFF)
644 return -EMSGSIZE;
645
646 /* Must have at least a full ICMP header. */
647 if (len < icmph_len)
648 return -EINVAL;
649
650 /*
651 * Check the flags.
652 */
653
654 /* Mirror BSD error message compatibility */
655 if (msg->msg_flags & MSG_OOB)
656 return -EOPNOTSUPP;
657
658 /*
659 * Fetch the ICMP header provided by the userland.
660 * iovec is modified! The ICMP header is consumed.
661 */
662 if (memcpy_from_msg(user_icmph, msg, icmph_len))
663 return -EFAULT;
664
665 if (family == AF_INET) {
666 type = ((struct icmphdr *) user_icmph)->type;
667 code = ((struct icmphdr *) user_icmph)->code;
668 #if IS_ENABLED(CONFIG_IPV6)
669 } else if (family == AF_INET6) {
670 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
671 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
672 #endif
673 } else {
674 BUG();
675 }
676
677 if (!ping_supported(family, type, code))
678 return -EINVAL;
679
680 return 0;
681 }
682 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
683
ping_v4_sendmsg(struct sock * sk,struct msghdr * msg,size_t len)684 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
685 {
686 struct net *net = sock_net(sk);
687 struct flowi4 fl4;
688 struct inet_sock *inet = inet_sk(sk);
689 struct ipcm_cookie ipc;
690 struct icmphdr user_icmph;
691 struct pingfakehdr pfh;
692 struct rtable *rt = NULL;
693 struct ip_options_data opt_copy;
694 int free = 0;
695 __be32 saddr, daddr, faddr;
696 u8 tos;
697 int err;
698
699 pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
700
701 err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
702 sizeof(user_icmph));
703 if (err)
704 return err;
705
706 /*
707 * Get and verify the address.
708 */
709
710 if (msg->msg_name) {
711 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
712 if (msg->msg_namelen < sizeof(*usin))
713 return -EINVAL;
714 if (usin->sin_family != AF_INET)
715 return -EAFNOSUPPORT;
716 daddr = usin->sin_addr.s_addr;
717 /* no remote port */
718 } else {
719 if (sk->sk_state != TCP_ESTABLISHED)
720 return -EDESTADDRREQ;
721 daddr = inet->inet_daddr;
722 /* no remote port */
723 }
724
725 ipcm_init_sk(&ipc, inet);
726
727 if (msg->msg_controllen) {
728 err = ip_cmsg_send(sk, msg, &ipc, false);
729 if (unlikely(err)) {
730 kfree(ipc.opt);
731 return err;
732 }
733 if (ipc.opt)
734 free = 1;
735 }
736 if (!ipc.opt) {
737 struct ip_options_rcu *inet_opt;
738
739 rcu_read_lock();
740 inet_opt = rcu_dereference(inet->inet_opt);
741 if (inet_opt) {
742 memcpy(&opt_copy, inet_opt,
743 sizeof(*inet_opt) + inet_opt->opt.optlen);
744 ipc.opt = &opt_copy.opt;
745 }
746 rcu_read_unlock();
747 }
748
749 saddr = ipc.addr;
750 ipc.addr = faddr = daddr;
751
752 if (ipc.opt && ipc.opt->opt.srr) {
753 if (!daddr) {
754 err = -EINVAL;
755 goto out_free;
756 }
757 faddr = ipc.opt->opt.faddr;
758 }
759 tos = get_rttos(&ipc, inet);
760 if (sock_flag(sk, SOCK_LOCALROUTE) ||
761 (msg->msg_flags & MSG_DONTROUTE) ||
762 (ipc.opt && ipc.opt->opt.is_strictroute)) {
763 tos |= RTO_ONLINK;
764 }
765
766 if (ipv4_is_multicast(daddr)) {
767 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
768 ipc.oif = inet->mc_index;
769 if (!saddr)
770 saddr = inet->mc_addr;
771 } else if (!ipc.oif)
772 ipc.oif = inet->uc_index;
773
774 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
775 RT_SCOPE_UNIVERSE, sk->sk_protocol,
776 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
777 sk->sk_uid);
778
779 fl4.fl4_icmp_type = user_icmph.type;
780 fl4.fl4_icmp_code = user_icmph.code;
781
782 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
783 rt = ip_route_output_flow(net, &fl4, sk);
784 if (IS_ERR(rt)) {
785 err = PTR_ERR(rt);
786 rt = NULL;
787 if (err == -ENETUNREACH)
788 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
789 goto out;
790 }
791
792 err = -EACCES;
793 if ((rt->rt_flags & RTCF_BROADCAST) &&
794 !sock_flag(sk, SOCK_BROADCAST))
795 goto out;
796
797 if (msg->msg_flags & MSG_CONFIRM)
798 goto do_confirm;
799 back_from_confirm:
800
801 if (!ipc.addr)
802 ipc.addr = fl4.daddr;
803
804 lock_sock(sk);
805
806 pfh.icmph.type = user_icmph.type; /* already checked */
807 pfh.icmph.code = user_icmph.code; /* ditto */
808 pfh.icmph.checksum = 0;
809 pfh.icmph.un.echo.id = inet->inet_sport;
810 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
811 pfh.msg = msg;
812 pfh.wcheck = 0;
813 pfh.family = AF_INET;
814
815 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
816 0, &ipc, &rt, msg->msg_flags);
817 if (err)
818 ip_flush_pending_frames(sk);
819 else
820 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
821 release_sock(sk);
822
823 out:
824 ip_rt_put(rt);
825 out_free:
826 if (free)
827 kfree(ipc.opt);
828 if (!err) {
829 icmp_out_count(sock_net(sk), user_icmph.type);
830 return len;
831 }
832 return err;
833
834 do_confirm:
835 if (msg->msg_flags & MSG_PROBE)
836 dst_confirm_neigh(&rt->dst, &fl4.daddr);
837 if (!(msg->msg_flags & MSG_PROBE) || len)
838 goto back_from_confirm;
839 err = 0;
840 goto out;
841 }
842
ping_recvmsg(struct sock * sk,struct msghdr * msg,size_t len,int noblock,int flags,int * addr_len)843 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
844 int flags, int *addr_len)
845 {
846 struct inet_sock *isk = inet_sk(sk);
847 int family = sk->sk_family;
848 struct sk_buff *skb;
849 int copied, err;
850
851 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
852
853 err = -EOPNOTSUPP;
854 if (flags & MSG_OOB)
855 goto out;
856
857 if (flags & MSG_ERRQUEUE)
858 return inet_recv_error(sk, msg, len, addr_len);
859
860 skb = skb_recv_datagram(sk, flags, noblock, &err);
861 if (!skb)
862 goto out;
863
864 copied = skb->len;
865 if (copied > len) {
866 msg->msg_flags |= MSG_TRUNC;
867 copied = len;
868 }
869
870 /* Don't bother checking the checksum */
871 err = skb_copy_datagram_msg(skb, 0, msg, copied);
872 if (err)
873 goto done;
874
875 sock_recv_timestamp(msg, sk, skb);
876
877 /* Copy the address and add cmsg data. */
878 if (family == AF_INET) {
879 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
880
881 if (sin) {
882 sin->sin_family = AF_INET;
883 sin->sin_port = 0 /* skb->h.uh->source */;
884 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
885 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
886 *addr_len = sizeof(*sin);
887 }
888
889 if (isk->cmsg_flags)
890 ip_cmsg_recv(msg, skb);
891
892 #if IS_ENABLED(CONFIG_IPV6)
893 } else if (family == AF_INET6) {
894 struct ipv6_pinfo *np = inet6_sk(sk);
895 struct ipv6hdr *ip6 = ipv6_hdr(skb);
896 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
897
898 if (sin6) {
899 sin6->sin6_family = AF_INET6;
900 sin6->sin6_port = 0;
901 sin6->sin6_addr = ip6->saddr;
902 sin6->sin6_flowinfo = 0;
903 if (np->sndflow)
904 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
905 sin6->sin6_scope_id =
906 ipv6_iface_scope_id(&sin6->sin6_addr,
907 inet6_iif(skb));
908 *addr_len = sizeof(*sin6);
909 }
910
911 if (inet6_sk(sk)->rxopt.all)
912 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
913 if (skb->protocol == htons(ETH_P_IPV6) &&
914 inet6_sk(sk)->rxopt.all)
915 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
916 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
917 ip_cmsg_recv(msg, skb);
918 #endif
919 } else {
920 BUG();
921 }
922
923 err = copied;
924
925 done:
926 skb_free_datagram(sk, skb);
927 out:
928 pr_debug("ping_recvmsg -> %d\n", err);
929 return err;
930 }
931 EXPORT_SYMBOL_GPL(ping_recvmsg);
932
ping_queue_rcv_skb(struct sock * sk,struct sk_buff * skb)933 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
934 {
935 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
936 inet_sk(sk), inet_sk(sk)->inet_num, skb);
937 if (sock_queue_rcv_skb(sk, skb) < 0) {
938 kfree_skb(skb);
939 pr_debug("ping_queue_rcv_skb -> failed\n");
940 return -1;
941 }
942 return 0;
943 }
944 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
945
946
947 /*
948 * All we need to do is get the socket.
949 */
950
ping_rcv(struct sk_buff * skb)951 bool ping_rcv(struct sk_buff *skb)
952 {
953 struct sock *sk;
954 struct net *net = dev_net(skb->dev);
955 struct icmphdr *icmph = icmp_hdr(skb);
956 bool rc = false;
957
958 /* We assume the packet has already been checked by icmp_rcv */
959
960 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
961 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
962
963 /* Push ICMP header back */
964 skb_push(skb, skb->data - (u8 *)icmph);
965
966 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
967 if (sk) {
968 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
969
970 pr_debug("rcv on socket %p\n", sk);
971 if (skb2 && !ping_queue_rcv_skb(sk, skb2))
972 rc = true;
973 sock_put(sk);
974 }
975
976 if (!rc)
977 pr_debug("no socket, dropping\n");
978
979 return rc;
980 }
981 EXPORT_SYMBOL_GPL(ping_rcv);
982
983 struct proto ping_prot = {
984 .name = "PING",
985 .owner = THIS_MODULE,
986 .init = ping_init_sock,
987 .close = ping_close,
988 .connect = ip4_datagram_connect,
989 .disconnect = __udp_disconnect,
990 .setsockopt = ip_setsockopt,
991 .getsockopt = ip_getsockopt,
992 .sendmsg = ping_v4_sendmsg,
993 .recvmsg = ping_recvmsg,
994 .bind = ping_bind,
995 .backlog_rcv = ping_queue_rcv_skb,
996 .release_cb = ip4_datagram_release_cb,
997 .hash = ping_hash,
998 .unhash = ping_unhash,
999 .get_port = ping_get_port,
1000 .obj_size = sizeof(struct inet_sock),
1001 };
1002 EXPORT_SYMBOL(ping_prot);
1003
1004 #ifdef CONFIG_PROC_FS
1005
ping_get_first(struct seq_file * seq,int start)1006 static struct sock *ping_get_first(struct seq_file *seq, int start)
1007 {
1008 struct sock *sk;
1009 struct ping_iter_state *state = seq->private;
1010 struct net *net = seq_file_net(seq);
1011
1012 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1013 ++state->bucket) {
1014 struct hlist_nulls_node *node;
1015 struct hlist_nulls_head *hslot;
1016
1017 hslot = &ping_table.hash[state->bucket];
1018
1019 if (hlist_nulls_empty(hslot))
1020 continue;
1021
1022 sk_nulls_for_each(sk, node, hslot) {
1023 if (net_eq(sock_net(sk), net) &&
1024 sk->sk_family == state->family)
1025 goto found;
1026 }
1027 }
1028 sk = NULL;
1029 found:
1030 return sk;
1031 }
1032
ping_get_next(struct seq_file * seq,struct sock * sk)1033 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1034 {
1035 struct ping_iter_state *state = seq->private;
1036 struct net *net = seq_file_net(seq);
1037
1038 do {
1039 sk = sk_nulls_next(sk);
1040 } while (sk && (!net_eq(sock_net(sk), net)));
1041
1042 if (!sk)
1043 return ping_get_first(seq, state->bucket + 1);
1044 return sk;
1045 }
1046
ping_get_idx(struct seq_file * seq,loff_t pos)1047 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1048 {
1049 struct sock *sk = ping_get_first(seq, 0);
1050
1051 if (sk)
1052 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1053 --pos;
1054 return pos ? NULL : sk;
1055 }
1056
ping_seq_start(struct seq_file * seq,loff_t * pos,sa_family_t family)1057 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1058 __acquires(ping_table.lock)
1059 {
1060 struct ping_iter_state *state = seq->private;
1061 state->bucket = 0;
1062 state->family = family;
1063
1064 read_lock_bh(&ping_table.lock);
1065
1066 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1067 }
1068 EXPORT_SYMBOL_GPL(ping_seq_start);
1069
ping_v4_seq_start(struct seq_file * seq,loff_t * pos)1070 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1071 {
1072 return ping_seq_start(seq, pos, AF_INET);
1073 }
1074
ping_seq_next(struct seq_file * seq,void * v,loff_t * pos)1075 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1076 {
1077 struct sock *sk;
1078
1079 if (v == SEQ_START_TOKEN)
1080 sk = ping_get_idx(seq, 0);
1081 else
1082 sk = ping_get_next(seq, v);
1083
1084 ++*pos;
1085 return sk;
1086 }
1087 EXPORT_SYMBOL_GPL(ping_seq_next);
1088
ping_seq_stop(struct seq_file * seq,void * v)1089 void ping_seq_stop(struct seq_file *seq, void *v)
1090 __releases(ping_table.lock)
1091 {
1092 read_unlock_bh(&ping_table.lock);
1093 }
1094 EXPORT_SYMBOL_GPL(ping_seq_stop);
1095
ping_v4_format_sock(struct sock * sp,struct seq_file * f,int bucket)1096 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1097 int bucket)
1098 {
1099 struct inet_sock *inet = inet_sk(sp);
1100 __be32 dest = inet->inet_daddr;
1101 __be32 src = inet->inet_rcv_saddr;
1102 __u16 destp = ntohs(inet->inet_dport);
1103 __u16 srcp = ntohs(inet->inet_sport);
1104
1105 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1106 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1107 bucket, src, srcp, dest, destp, sp->sk_state,
1108 sk_wmem_alloc_get(sp),
1109 sk_rmem_alloc_get(sp),
1110 0, 0L, 0,
1111 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1112 0, sock_i_ino(sp),
1113 refcount_read(&sp->sk_refcnt), sp,
1114 atomic_read(&sp->sk_drops));
1115 }
1116
ping_v4_seq_show(struct seq_file * seq,void * v)1117 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1118 {
1119 seq_setwidth(seq, 127);
1120 if (v == SEQ_START_TOKEN)
1121 seq_puts(seq, " sl local_address rem_address st tx_queue "
1122 "rx_queue tr tm->when retrnsmt uid timeout "
1123 "inode ref pointer drops");
1124 else {
1125 struct ping_iter_state *state = seq->private;
1126
1127 ping_v4_format_sock(v, seq, state->bucket);
1128 }
1129 seq_pad(seq, '\n');
1130 return 0;
1131 }
1132
1133 static const struct seq_operations ping_v4_seq_ops = {
1134 .start = ping_v4_seq_start,
1135 .show = ping_v4_seq_show,
1136 .next = ping_seq_next,
1137 .stop = ping_seq_stop,
1138 };
1139
ping_v4_proc_init_net(struct net * net)1140 static int __net_init ping_v4_proc_init_net(struct net *net)
1141 {
1142 if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1143 sizeof(struct ping_iter_state)))
1144 return -ENOMEM;
1145 return 0;
1146 }
1147
ping_v4_proc_exit_net(struct net * net)1148 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1149 {
1150 remove_proc_entry("icmp", net->proc_net);
1151 }
1152
1153 static struct pernet_operations ping_v4_net_ops = {
1154 .init = ping_v4_proc_init_net,
1155 .exit = ping_v4_proc_exit_net,
1156 };
1157
ping_proc_init(void)1158 int __init ping_proc_init(void)
1159 {
1160 return register_pernet_subsys(&ping_v4_net_ops);
1161 }
1162
ping_proc_exit(void)1163 void ping_proc_exit(void)
1164 {
1165 unregister_pernet_subsys(&ping_v4_net_ops);
1166 }
1167
1168 #endif
1169
ping_init(void)1170 void __init ping_init(void)
1171 {
1172 int i;
1173
1174 for (i = 0; i < PING_HTABLE_SIZE; i++)
1175 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1176 rwlock_init(&ping_table.lock);
1177 }
1178