]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - net/ipv4/ping.c
Merge remote-tracking branches 'asoc/topic/davinci', 'asoc/topic/fsl-card' and 'asoc...
[mirror_ubuntu-zesty-kernel.git] / net / ipv4 / ping.c
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 (unlikely(err)) {
750 kfree(ipc.opt);
751 return err;
752 }
753 if (ipc.opt)
754 free = 1;
755 }
756 if (!ipc.opt) {
757 struct ip_options_rcu *inet_opt;
758
759 rcu_read_lock();
760 inet_opt = rcu_dereference(inet->inet_opt);
761 if (inet_opt) {
762 memcpy(&opt_copy, inet_opt,
763 sizeof(*inet_opt) + inet_opt->opt.optlen);
764 ipc.opt = &opt_copy.opt;
765 }
766 rcu_read_unlock();
767 }
768
769 saddr = ipc.addr;
770 ipc.addr = faddr = daddr;
771
772 if (ipc.opt && ipc.opt->opt.srr) {
773 if (!daddr)
774 return -EINVAL;
775 faddr = ipc.opt->opt.faddr;
776 }
777 tos = get_rttos(&ipc, inet);
778 if (sock_flag(sk, SOCK_LOCALROUTE) ||
779 (msg->msg_flags & MSG_DONTROUTE) ||
780 (ipc.opt && ipc.opt->opt.is_strictroute)) {
781 tos |= RTO_ONLINK;
782 }
783
784 if (ipv4_is_multicast(daddr)) {
785 if (!ipc.oif)
786 ipc.oif = inet->mc_index;
787 if (!saddr)
788 saddr = inet->mc_addr;
789 } else if (!ipc.oif)
790 ipc.oif = inet->uc_index;
791
792 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
793 RT_SCOPE_UNIVERSE, sk->sk_protocol,
794 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
795
796 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
797 rt = ip_route_output_flow(net, &fl4, sk);
798 if (IS_ERR(rt)) {
799 err = PTR_ERR(rt);
800 rt = NULL;
801 if (err == -ENETUNREACH)
802 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
803 goto out;
804 }
805
806 err = -EACCES;
807 if ((rt->rt_flags & RTCF_BROADCAST) &&
808 !sock_flag(sk, SOCK_BROADCAST))
809 goto out;
810
811 if (msg->msg_flags & MSG_CONFIRM)
812 goto do_confirm;
813 back_from_confirm:
814
815 if (!ipc.addr)
816 ipc.addr = fl4.daddr;
817
818 lock_sock(sk);
819
820 pfh.icmph.type = user_icmph.type; /* already checked */
821 pfh.icmph.code = user_icmph.code; /* ditto */
822 pfh.icmph.checksum = 0;
823 pfh.icmph.un.echo.id = inet->inet_sport;
824 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
825 pfh.msg = msg;
826 pfh.wcheck = 0;
827 pfh.family = AF_INET;
828
829 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
830 0, &ipc, &rt, msg->msg_flags);
831 if (err)
832 ip_flush_pending_frames(sk);
833 else
834 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
835 release_sock(sk);
836
837 out:
838 ip_rt_put(rt);
839 if (free)
840 kfree(ipc.opt);
841 if (!err) {
842 icmp_out_count(sock_net(sk), user_icmph.type);
843 return len;
844 }
845 return err;
846
847 do_confirm:
848 dst_confirm(&rt->dst);
849 if (!(msg->msg_flags & MSG_PROBE) || len)
850 goto back_from_confirm;
851 err = 0;
852 goto out;
853 }
854
855 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
856 int flags, int *addr_len)
857 {
858 struct inet_sock *isk = inet_sk(sk);
859 int family = sk->sk_family;
860 struct sk_buff *skb;
861 int copied, err;
862
863 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
864
865 err = -EOPNOTSUPP;
866 if (flags & MSG_OOB)
867 goto out;
868
869 if (flags & MSG_ERRQUEUE)
870 return inet_recv_error(sk, msg, len, addr_len);
871
872 skb = skb_recv_datagram(sk, flags, noblock, &err);
873 if (!skb)
874 goto out;
875
876 copied = skb->len;
877 if (copied > len) {
878 msg->msg_flags |= MSG_TRUNC;
879 copied = len;
880 }
881
882 /* Don't bother checking the checksum */
883 err = skb_copy_datagram_msg(skb, 0, msg, copied);
884 if (err)
885 goto done;
886
887 sock_recv_timestamp(msg, sk, skb);
888
889 /* Copy the address and add cmsg data. */
890 if (family == AF_INET) {
891 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
892
893 if (sin) {
894 sin->sin_family = AF_INET;
895 sin->sin_port = 0 /* skb->h.uh->source */;
896 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
897 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
898 *addr_len = sizeof(*sin);
899 }
900
901 if (isk->cmsg_flags)
902 ip_cmsg_recv(msg, skb);
903
904 #if IS_ENABLED(CONFIG_IPV6)
905 } else if (family == AF_INET6) {
906 struct ipv6_pinfo *np = inet6_sk(sk);
907 struct ipv6hdr *ip6 = ipv6_hdr(skb);
908 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
909
910 if (sin6) {
911 sin6->sin6_family = AF_INET6;
912 sin6->sin6_port = 0;
913 sin6->sin6_addr = ip6->saddr;
914 sin6->sin6_flowinfo = 0;
915 if (np->sndflow)
916 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
917 sin6->sin6_scope_id =
918 ipv6_iface_scope_id(&sin6->sin6_addr,
919 inet6_iif(skb));
920 *addr_len = sizeof(*sin6);
921 }
922
923 if (inet6_sk(sk)->rxopt.all)
924 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
925 if (skb->protocol == htons(ETH_P_IPV6) &&
926 inet6_sk(sk)->rxopt.all)
927 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
928 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
929 ip_cmsg_recv(msg, skb);
930 #endif
931 } else {
932 BUG();
933 }
934
935 err = copied;
936
937 done:
938 skb_free_datagram(sk, skb);
939 out:
940 pr_debug("ping_recvmsg -> %d\n", err);
941 return err;
942 }
943 EXPORT_SYMBOL_GPL(ping_recvmsg);
944
945 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
946 {
947 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
948 inet_sk(sk), inet_sk(sk)->inet_num, skb);
949 if (sock_queue_rcv_skb(sk, skb) < 0) {
950 kfree_skb(skb);
951 pr_debug("ping_queue_rcv_skb -> failed\n");
952 return -1;
953 }
954 return 0;
955 }
956 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
957
958
959 /*
960 * All we need to do is get the socket.
961 */
962
963 bool ping_rcv(struct sk_buff *skb)
964 {
965 struct sock *sk;
966 struct net *net = dev_net(skb->dev);
967 struct icmphdr *icmph = icmp_hdr(skb);
968
969 /* We assume the packet has already been checked by icmp_rcv */
970
971 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
972 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
973
974 /* Push ICMP header back */
975 skb_push(skb, skb->data - (u8 *)icmph);
976
977 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
978 if (sk) {
979 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
980
981 pr_debug("rcv on socket %p\n", sk);
982 if (skb2)
983 ping_queue_rcv_skb(sk, skb2);
984 sock_put(sk);
985 return true;
986 }
987 pr_debug("no socket, dropping\n");
988
989 return false;
990 }
991 EXPORT_SYMBOL_GPL(ping_rcv);
992
993 struct proto ping_prot = {
994 .name = "PING",
995 .owner = THIS_MODULE,
996 .init = ping_init_sock,
997 .close = ping_close,
998 .connect = ip4_datagram_connect,
999 .disconnect = udp_disconnect,
1000 .setsockopt = ip_setsockopt,
1001 .getsockopt = ip_getsockopt,
1002 .sendmsg = ping_v4_sendmsg,
1003 .recvmsg = ping_recvmsg,
1004 .bind = ping_bind,
1005 .backlog_rcv = ping_queue_rcv_skb,
1006 .release_cb = ip4_datagram_release_cb,
1007 .hash = ping_hash,
1008 .unhash = ping_unhash,
1009 .get_port = ping_get_port,
1010 .obj_size = sizeof(struct inet_sock),
1011 };
1012 EXPORT_SYMBOL(ping_prot);
1013
1014 #ifdef CONFIG_PROC_FS
1015
1016 static struct sock *ping_get_first(struct seq_file *seq, int start)
1017 {
1018 struct sock *sk;
1019 struct ping_iter_state *state = seq->private;
1020 struct net *net = seq_file_net(seq);
1021
1022 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1023 ++state->bucket) {
1024 struct hlist_nulls_node *node;
1025 struct hlist_nulls_head *hslot;
1026
1027 hslot = &ping_table.hash[state->bucket];
1028
1029 if (hlist_nulls_empty(hslot))
1030 continue;
1031
1032 sk_nulls_for_each(sk, node, hslot) {
1033 if (net_eq(sock_net(sk), net) &&
1034 sk->sk_family == state->family)
1035 goto found;
1036 }
1037 }
1038 sk = NULL;
1039 found:
1040 return sk;
1041 }
1042
1043 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1044 {
1045 struct ping_iter_state *state = seq->private;
1046 struct net *net = seq_file_net(seq);
1047
1048 do {
1049 sk = sk_nulls_next(sk);
1050 } while (sk && (!net_eq(sock_net(sk), net)));
1051
1052 if (!sk)
1053 return ping_get_first(seq, state->bucket + 1);
1054 return sk;
1055 }
1056
1057 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1058 {
1059 struct sock *sk = ping_get_first(seq, 0);
1060
1061 if (sk)
1062 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1063 --pos;
1064 return pos ? NULL : sk;
1065 }
1066
1067 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1068 __acquires(ping_table.lock)
1069 {
1070 struct ping_iter_state *state = seq->private;
1071 state->bucket = 0;
1072 state->family = family;
1073
1074 read_lock_bh(&ping_table.lock);
1075
1076 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1077 }
1078 EXPORT_SYMBOL_GPL(ping_seq_start);
1079
1080 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1081 {
1082 return ping_seq_start(seq, pos, AF_INET);
1083 }
1084
1085 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1086 {
1087 struct sock *sk;
1088
1089 if (v == SEQ_START_TOKEN)
1090 sk = ping_get_idx(seq, 0);
1091 else
1092 sk = ping_get_next(seq, v);
1093
1094 ++*pos;
1095 return sk;
1096 }
1097 EXPORT_SYMBOL_GPL(ping_seq_next);
1098
1099 void ping_seq_stop(struct seq_file *seq, void *v)
1100 __releases(ping_table.lock)
1101 {
1102 read_unlock_bh(&ping_table.lock);
1103 }
1104 EXPORT_SYMBOL_GPL(ping_seq_stop);
1105
1106 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1107 int bucket)
1108 {
1109 struct inet_sock *inet = inet_sk(sp);
1110 __be32 dest = inet->inet_daddr;
1111 __be32 src = inet->inet_rcv_saddr;
1112 __u16 destp = ntohs(inet->inet_dport);
1113 __u16 srcp = ntohs(inet->inet_sport);
1114
1115 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1116 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1117 bucket, src, srcp, dest, destp, sp->sk_state,
1118 sk_wmem_alloc_get(sp),
1119 sk_rmem_alloc_get(sp),
1120 0, 0L, 0,
1121 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1122 0, sock_i_ino(sp),
1123 atomic_read(&sp->sk_refcnt), sp,
1124 atomic_read(&sp->sk_drops));
1125 }
1126
1127 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1128 {
1129 seq_setwidth(seq, 127);
1130 if (v == SEQ_START_TOKEN)
1131 seq_puts(seq, " sl local_address rem_address st tx_queue "
1132 "rx_queue tr tm->when retrnsmt uid timeout "
1133 "inode ref pointer drops");
1134 else {
1135 struct ping_iter_state *state = seq->private;
1136
1137 ping_v4_format_sock(v, seq, state->bucket);
1138 }
1139 seq_pad(seq, '\n');
1140 return 0;
1141 }
1142
1143 static const struct seq_operations ping_v4_seq_ops = {
1144 .show = ping_v4_seq_show,
1145 .start = ping_v4_seq_start,
1146 .next = ping_seq_next,
1147 .stop = ping_seq_stop,
1148 };
1149
1150 static int ping_seq_open(struct inode *inode, struct file *file)
1151 {
1152 struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1153 return seq_open_net(inode, file, &afinfo->seq_ops,
1154 sizeof(struct ping_iter_state));
1155 }
1156
1157 const struct file_operations ping_seq_fops = {
1158 .open = ping_seq_open,
1159 .read = seq_read,
1160 .llseek = seq_lseek,
1161 .release = seq_release_net,
1162 };
1163 EXPORT_SYMBOL_GPL(ping_seq_fops);
1164
1165 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1166 .name = "icmp",
1167 .family = AF_INET,
1168 .seq_fops = &ping_seq_fops,
1169 .seq_ops = {
1170 .start = ping_v4_seq_start,
1171 .show = ping_v4_seq_show,
1172 .next = ping_seq_next,
1173 .stop = ping_seq_stop,
1174 },
1175 };
1176
1177 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1178 {
1179 struct proc_dir_entry *p;
1180 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1181 afinfo->seq_fops, afinfo);
1182 if (!p)
1183 return -ENOMEM;
1184 return 0;
1185 }
1186 EXPORT_SYMBOL_GPL(ping_proc_register);
1187
1188 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1189 {
1190 remove_proc_entry(afinfo->name, net->proc_net);
1191 }
1192 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1193
1194 static int __net_init ping_v4_proc_init_net(struct net *net)
1195 {
1196 return ping_proc_register(net, &ping_v4_seq_afinfo);
1197 }
1198
1199 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1200 {
1201 ping_proc_unregister(net, &ping_v4_seq_afinfo);
1202 }
1203
1204 static struct pernet_operations ping_v4_net_ops = {
1205 .init = ping_v4_proc_init_net,
1206 .exit = ping_v4_proc_exit_net,
1207 };
1208
1209 int __init ping_proc_init(void)
1210 {
1211 return register_pernet_subsys(&ping_v4_net_ops);
1212 }
1213
1214 void ping_proc_exit(void)
1215 {
1216 unregister_pernet_subsys(&ping_v4_net_ops);
1217 }
1218
1219 #endif
1220
1221 void __init ping_init(void)
1222 {
1223 int i;
1224
1225 for (i = 0; i < PING_HTABLE_SIZE; i++)
1226 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1227 rwlock_init(&ping_table.lock);
1228 }