Loading...
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
51struct ping_table {
52 struct hlist_nulls_head hash[PING_HTABLE_SIZE];
53 rwlock_t lock;
54};
55
56static struct ping_table ping_table;
57struct pingv6_ops pingv6_ops;
58EXPORT_SYMBOL_GPL(pingv6_ops);
59
60static u16 ping_port_rover;
61
62static 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}
69EXPORT_SYMBOL_GPL(ping_hash);
70
71static 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
77int 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;
105next_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
137fail:
138 write_unlock_bh(&ping_table.lock);
139 return 1;
140}
141EXPORT_SYMBOL_GPL(ping_get_port);
142
143int 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
151void 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}
167EXPORT_SYMBOL_GPL(ping_unhash);
168
169static 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 continue;
225
226 sock_hold(sk);
227 goto exit;
228 }
229
230 sk = NULL;
231exit:
232 read_unlock_bh(&ping_table.lock);
233
234 return sk;
235}
236
237static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
238 kgid_t *high)
239{
240 kgid_t *data = net->ipv4.ping_group_range.range;
241 unsigned int seq;
242
243 do {
244 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
245
246 *low = data[0];
247 *high = data[1];
248 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
249}
250
251
252int ping_init_sock(struct sock *sk)
253{
254 struct net *net = sock_net(sk);
255 kgid_t group = current_egid();
256 struct group_info *group_info;
257 int i;
258 kgid_t low, high;
259 int ret = 0;
260
261 if (sk->sk_family == AF_INET6)
262 sk->sk_ipv6only = 1;
263
264 inet_get_ping_group_range_net(net, &low, &high);
265 if (gid_lte(low, group) && gid_lte(group, high))
266 return 0;
267
268 group_info = get_current_groups();
269 for (i = 0; i < group_info->ngroups; i++) {
270 kgid_t gid = group_info->gid[i];
271
272 if (gid_lte(low, gid) && gid_lte(gid, high))
273 goto out_release_group;
274 }
275
276 ret = -EACCES;
277
278out_release_group:
279 put_group_info(group_info);
280 return ret;
281}
282EXPORT_SYMBOL_GPL(ping_init_sock);
283
284void ping_close(struct sock *sk, long timeout)
285{
286 pr_debug("ping_close(sk=%p,sk->num=%u)\n",
287 inet_sk(sk), inet_sk(sk)->inet_num);
288 pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
289
290 sk_common_release(sk);
291}
292EXPORT_SYMBOL_GPL(ping_close);
293
294/* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
295static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
296 struct sockaddr *uaddr, int addr_len)
297{
298 struct net *net = sock_net(sk);
299 if (sk->sk_family == AF_INET) {
300 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
301 int chk_addr_ret;
302
303 if (addr_len < sizeof(*addr))
304 return -EINVAL;
305
306 if (addr->sin_family != AF_INET &&
307 !(addr->sin_family == AF_UNSPEC &&
308 addr->sin_addr.s_addr == htonl(INADDR_ANY)))
309 return -EAFNOSUPPORT;
310
311 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
312 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
313
314 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
315 chk_addr_ret = RTN_LOCAL;
316 else
317 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
318
319 if ((!inet_can_nonlocal_bind(net, isk) &&
320 chk_addr_ret != RTN_LOCAL) ||
321 chk_addr_ret == RTN_MULTICAST ||
322 chk_addr_ret == RTN_BROADCAST)
323 return -EADDRNOTAVAIL;
324
325#if IS_ENABLED(CONFIG_IPV6)
326 } else if (sk->sk_family == AF_INET6) {
327 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
328 int addr_type, scoped, has_addr;
329 struct net_device *dev = NULL;
330
331 if (addr_len < sizeof(*addr))
332 return -EINVAL;
333
334 if (addr->sin6_family != AF_INET6)
335 return -EAFNOSUPPORT;
336
337 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
338 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
339
340 addr_type = ipv6_addr_type(&addr->sin6_addr);
341 scoped = __ipv6_addr_needs_scope_id(addr_type);
342 if ((addr_type != IPV6_ADDR_ANY &&
343 !(addr_type & IPV6_ADDR_UNICAST)) ||
344 (scoped && !addr->sin6_scope_id))
345 return -EINVAL;
346
347 rcu_read_lock();
348 if (addr->sin6_scope_id) {
349 dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
350 if (!dev) {
351 rcu_read_unlock();
352 return -ENODEV;
353 }
354 }
355 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
356 scoped);
357 rcu_read_unlock();
358
359 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
360 addr_type == IPV6_ADDR_ANY))
361 return -EADDRNOTAVAIL;
362
363 if (scoped)
364 sk->sk_bound_dev_if = addr->sin6_scope_id;
365#endif
366 } else {
367 return -EAFNOSUPPORT;
368 }
369 return 0;
370}
371
372static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
373{
374 if (saddr->sa_family == AF_INET) {
375 struct inet_sock *isk = inet_sk(sk);
376 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
377 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
378#if IS_ENABLED(CONFIG_IPV6)
379 } else if (saddr->sa_family == AF_INET6) {
380 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
381 struct ipv6_pinfo *np = inet6_sk(sk);
382 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
383#endif
384 }
385}
386
387/*
388 * We need our own bind because there are no privileged id's == local ports.
389 * Moreover, we don't allow binding to multi- and broadcast addresses.
390 */
391
392int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
393{
394 struct inet_sock *isk = inet_sk(sk);
395 unsigned short snum;
396 int err;
397 int dif = sk->sk_bound_dev_if;
398
399 err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
400 if (err)
401 return err;
402
403 lock_sock(sk);
404
405 err = -EINVAL;
406 if (isk->inet_num != 0)
407 goto out;
408
409 err = -EADDRINUSE;
410 snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
411 if (ping_get_port(sk, snum) != 0) {
412 /* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
413 sk->sk_bound_dev_if = dif;
414 goto out;
415 }
416 ping_set_saddr(sk, uaddr);
417
418 pr_debug("after bind(): num = %hu, dif = %d\n",
419 isk->inet_num,
420 sk->sk_bound_dev_if);
421
422 err = 0;
423 if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
424 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
425#if IS_ENABLED(CONFIG_IPV6)
426 if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
427 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
428#endif
429
430 if (snum)
431 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
432 isk->inet_sport = htons(isk->inet_num);
433 isk->inet_daddr = 0;
434 isk->inet_dport = 0;
435
436#if IS_ENABLED(CONFIG_IPV6)
437 if (sk->sk_family == AF_INET6)
438 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
439#endif
440
441 sk_dst_reset(sk);
442out:
443 release_sock(sk);
444 pr_debug("ping_v4_bind -> %d\n", err);
445 return err;
446}
447EXPORT_SYMBOL_GPL(ping_bind);
448
449/*
450 * Is this a supported type of ICMP message?
451 */
452
453static inline int ping_supported(int family, int type, int code)
454{
455 return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
456 (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
457 (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
458 (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
459}
460
461/*
462 * This routine is called by the ICMP module when it gets some
463 * sort of error condition.
464 */
465
466void ping_err(struct sk_buff *skb, int offset, u32 info)
467{
468 int family;
469 struct icmphdr *icmph;
470 struct inet_sock *inet_sock;
471 int type;
472 int code;
473 struct net *net = dev_net(skb->dev);
474 struct sock *sk;
475 int harderr;
476 int err;
477
478 if (skb->protocol == htons(ETH_P_IP)) {
479 family = AF_INET;
480 type = icmp_hdr(skb)->type;
481 code = icmp_hdr(skb)->code;
482 icmph = (struct icmphdr *)(skb->data + offset);
483 } else if (skb->protocol == htons(ETH_P_IPV6)) {
484 family = AF_INET6;
485 type = icmp6_hdr(skb)->icmp6_type;
486 code = icmp6_hdr(skb)->icmp6_code;
487 icmph = (struct icmphdr *) (skb->data + offset);
488 } else {
489 BUG();
490 }
491
492 /* We assume the packet has already been checked by icmp_unreach */
493
494 if (!ping_supported(family, icmph->type, icmph->code))
495 return;
496
497 pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
498 skb->protocol, type, code, ntohs(icmph->un.echo.id),
499 ntohs(icmph->un.echo.sequence));
500
501 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
502 if (!sk) {
503 pr_debug("no socket, dropping\n");
504 return; /* No socket for error */
505 }
506 pr_debug("err on socket %p\n", sk);
507
508 err = 0;
509 harderr = 0;
510 inet_sock = inet_sk(sk);
511
512 if (skb->protocol == htons(ETH_P_IP)) {
513 switch (type) {
514 default:
515 case ICMP_TIME_EXCEEDED:
516 err = EHOSTUNREACH;
517 break;
518 case ICMP_SOURCE_QUENCH:
519 /* This is not a real error but ping wants to see it.
520 * Report it with some fake errno.
521 */
522 err = EREMOTEIO;
523 break;
524 case ICMP_PARAMETERPROB:
525 err = EPROTO;
526 harderr = 1;
527 break;
528 case ICMP_DEST_UNREACH:
529 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
530 ipv4_sk_update_pmtu(skb, sk, info);
531 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
532 err = EMSGSIZE;
533 harderr = 1;
534 break;
535 }
536 goto out;
537 }
538 err = EHOSTUNREACH;
539 if (code <= NR_ICMP_UNREACH) {
540 harderr = icmp_err_convert[code].fatal;
541 err = icmp_err_convert[code].errno;
542 }
543 break;
544 case ICMP_REDIRECT:
545 /* See ICMP_SOURCE_QUENCH */
546 ipv4_sk_redirect(skb, sk);
547 err = EREMOTEIO;
548 break;
549 }
550#if IS_ENABLED(CONFIG_IPV6)
551 } else if (skb->protocol == htons(ETH_P_IPV6)) {
552 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
553#endif
554 }
555
556 /*
557 * RFC1122: OK. Passes ICMP errors back to application, as per
558 * 4.1.3.3.
559 */
560 if ((family == AF_INET && !inet_sock->recverr) ||
561 (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
562 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
563 goto out;
564 } else {
565 if (family == AF_INET) {
566 ip_icmp_error(sk, skb, err, 0 /* no remote port */,
567 info, (u8 *)icmph);
568#if IS_ENABLED(CONFIG_IPV6)
569 } else if (family == AF_INET6) {
570 pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
571 info, (u8 *)icmph);
572#endif
573 }
574 }
575 sk->sk_err = err;
576 sk_error_report(sk);
577out:
578 sock_put(sk);
579}
580EXPORT_SYMBOL_GPL(ping_err);
581
582/*
583 * Copy and checksum an ICMP Echo packet from user space into a buffer
584 * starting from the payload.
585 */
586
587int ping_getfrag(void *from, char *to,
588 int offset, int fraglen, int odd, struct sk_buff *skb)
589{
590 struct pingfakehdr *pfh = (struct pingfakehdr *)from;
591
592 if (offset == 0) {
593 fraglen -= sizeof(struct icmphdr);
594 if (fraglen < 0)
595 BUG();
596 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
597 fraglen, &pfh->wcheck,
598 &pfh->msg->msg_iter))
599 return -EFAULT;
600 } else if (offset < sizeof(struct icmphdr)) {
601 BUG();
602 } else {
603 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
604 &pfh->msg->msg_iter))
605 return -EFAULT;
606 }
607
608#if IS_ENABLED(CONFIG_IPV6)
609 /* For IPv6, checksum each skb as we go along, as expected by
610 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
611 * wcheck, it will be finalized in ping_v4_push_pending_frames.
612 */
613 if (pfh->family == AF_INET6) {
614 skb->csum = pfh->wcheck;
615 skb->ip_summed = CHECKSUM_NONE;
616 pfh->wcheck = 0;
617 }
618#endif
619
620 return 0;
621}
622EXPORT_SYMBOL_GPL(ping_getfrag);
623
624static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
625 struct flowi4 *fl4)
626{
627 struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
628
629 if (!skb)
630 return 0;
631 pfh->wcheck = csum_partial((char *)&pfh->icmph,
632 sizeof(struct icmphdr), pfh->wcheck);
633 pfh->icmph.checksum = csum_fold(pfh->wcheck);
634 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
635 skb->ip_summed = CHECKSUM_NONE;
636 return ip_push_pending_frames(sk, fl4);
637}
638
639int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
640 void *user_icmph, size_t icmph_len)
641{
642 u8 type, code;
643
644 if (len > 0xFFFF)
645 return -EMSGSIZE;
646
647 /* Must have at least a full ICMP header. */
648 if (len < icmph_len)
649 return -EINVAL;
650
651 /*
652 * Check the flags.
653 */
654
655 /* Mirror BSD error message compatibility */
656 if (msg->msg_flags & MSG_OOB)
657 return -EOPNOTSUPP;
658
659 /*
660 * Fetch the ICMP header provided by the userland.
661 * iovec is modified! The ICMP header is consumed.
662 */
663 if (memcpy_from_msg(user_icmph, msg, icmph_len))
664 return -EFAULT;
665
666 if (family == AF_INET) {
667 type = ((struct icmphdr *) user_icmph)->type;
668 code = ((struct icmphdr *) user_icmph)->code;
669#if IS_ENABLED(CONFIG_IPV6)
670 } else if (family == AF_INET6) {
671 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
672 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
673#endif
674 } else {
675 BUG();
676 }
677
678 if (!ping_supported(family, type, code))
679 return -EINVAL;
680
681 return 0;
682}
683EXPORT_SYMBOL_GPL(ping_common_sendmsg);
684
685static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
686{
687 struct net *net = sock_net(sk);
688 struct flowi4 fl4;
689 struct inet_sock *inet = inet_sk(sk);
690 struct ipcm_cookie ipc;
691 struct icmphdr user_icmph;
692 struct pingfakehdr pfh;
693 struct rtable *rt = NULL;
694 struct ip_options_data opt_copy;
695 int free = 0;
696 __be32 saddr, daddr, faddr;
697 u8 tos;
698 int err;
699
700 pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
701
702 err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
703 sizeof(user_icmph));
704 if (err)
705 return err;
706
707 /*
708 * Get and verify the address.
709 */
710
711 if (msg->msg_name) {
712 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
713 if (msg->msg_namelen < sizeof(*usin))
714 return -EINVAL;
715 if (usin->sin_family != AF_INET)
716 return -EAFNOSUPPORT;
717 daddr = usin->sin_addr.s_addr;
718 /* no remote port */
719 } else {
720 if (sk->sk_state != TCP_ESTABLISHED)
721 return -EDESTADDRREQ;
722 daddr = inet->inet_daddr;
723 /* no remote port */
724 }
725
726 ipcm_init_sk(&ipc, inet);
727
728 if (msg->msg_controllen) {
729 err = ip_cmsg_send(sk, msg, &ipc, false);
730 if (unlikely(err)) {
731 kfree(ipc.opt);
732 return err;
733 }
734 if (ipc.opt)
735 free = 1;
736 }
737 if (!ipc.opt) {
738 struct ip_options_rcu *inet_opt;
739
740 rcu_read_lock();
741 inet_opt = rcu_dereference(inet->inet_opt);
742 if (inet_opt) {
743 memcpy(&opt_copy, inet_opt,
744 sizeof(*inet_opt) + inet_opt->opt.optlen);
745 ipc.opt = &opt_copy.opt;
746 }
747 rcu_read_unlock();
748 }
749
750 saddr = ipc.addr;
751 ipc.addr = faddr = daddr;
752
753 if (ipc.opt && ipc.opt->opt.srr) {
754 if (!daddr) {
755 err = -EINVAL;
756 goto out_free;
757 }
758 faddr = ipc.opt->opt.faddr;
759 }
760 tos = get_rttos(&ipc, inet);
761 if (sock_flag(sk, SOCK_LOCALROUTE) ||
762 (msg->msg_flags & MSG_DONTROUTE) ||
763 (ipc.opt && ipc.opt->opt.is_strictroute)) {
764 tos |= RTO_ONLINK;
765 }
766
767 if (ipv4_is_multicast(daddr)) {
768 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
769 ipc.oif = inet->mc_index;
770 if (!saddr)
771 saddr = inet->mc_addr;
772 } else if (!ipc.oif)
773 ipc.oif = inet->uc_index;
774
775 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
776 RT_SCOPE_UNIVERSE, sk->sk_protocol,
777 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
778 sk->sk_uid);
779
780 fl4.fl4_icmp_type = user_icmph.type;
781 fl4.fl4_icmp_code = user_icmph.code;
782
783 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
784 rt = ip_route_output_flow(net, &fl4, sk);
785 if (IS_ERR(rt)) {
786 err = PTR_ERR(rt);
787 rt = NULL;
788 if (err == -ENETUNREACH)
789 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
790 goto out;
791 }
792
793 err = -EACCES;
794 if ((rt->rt_flags & RTCF_BROADCAST) &&
795 !sock_flag(sk, SOCK_BROADCAST))
796 goto out;
797
798 if (msg->msg_flags & MSG_CONFIRM)
799 goto do_confirm;
800back_from_confirm:
801
802 if (!ipc.addr)
803 ipc.addr = fl4.daddr;
804
805 lock_sock(sk);
806
807 pfh.icmph.type = user_icmph.type; /* already checked */
808 pfh.icmph.code = user_icmph.code; /* ditto */
809 pfh.icmph.checksum = 0;
810 pfh.icmph.un.echo.id = inet->inet_sport;
811 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
812 pfh.msg = msg;
813 pfh.wcheck = 0;
814 pfh.family = AF_INET;
815
816 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
817 0, &ipc, &rt, msg->msg_flags);
818 if (err)
819 ip_flush_pending_frames(sk);
820 else
821 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
822 release_sock(sk);
823
824out:
825 ip_rt_put(rt);
826out_free:
827 if (free)
828 kfree(ipc.opt);
829 if (!err) {
830 icmp_out_count(sock_net(sk), user_icmph.type);
831 return len;
832 }
833 return err;
834
835do_confirm:
836 if (msg->msg_flags & MSG_PROBE)
837 dst_confirm_neigh(&rt->dst, &fl4.daddr);
838 if (!(msg->msg_flags & MSG_PROBE) || len)
839 goto back_from_confirm;
840 err = 0;
841 goto out;
842}
843
844int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
845 int flags, int *addr_len)
846{
847 struct inet_sock *isk = inet_sk(sk);
848 int family = sk->sk_family;
849 struct sk_buff *skb;
850 int copied, err;
851
852 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
853
854 err = -EOPNOTSUPP;
855 if (flags & MSG_OOB)
856 goto out;
857
858 if (flags & MSG_ERRQUEUE)
859 return inet_recv_error(sk, msg, len, addr_len);
860
861 skb = skb_recv_datagram(sk, flags, noblock, &err);
862 if (!skb)
863 goto out;
864
865 copied = skb->len;
866 if (copied > len) {
867 msg->msg_flags |= MSG_TRUNC;
868 copied = len;
869 }
870
871 /* Don't bother checking the checksum */
872 err = skb_copy_datagram_msg(skb, 0, msg, copied);
873 if (err)
874 goto done;
875
876 sock_recv_timestamp(msg, sk, skb);
877
878 /* Copy the address and add cmsg data. */
879 if (family == AF_INET) {
880 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
881
882 if (sin) {
883 sin->sin_family = AF_INET;
884 sin->sin_port = 0 /* skb->h.uh->source */;
885 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
886 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
887 *addr_len = sizeof(*sin);
888 }
889
890 if (isk->cmsg_flags)
891 ip_cmsg_recv(msg, skb);
892
893#if IS_ENABLED(CONFIG_IPV6)
894 } else if (family == AF_INET6) {
895 struct ipv6_pinfo *np = inet6_sk(sk);
896 struct ipv6hdr *ip6 = ipv6_hdr(skb);
897 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
898
899 if (sin6) {
900 sin6->sin6_family = AF_INET6;
901 sin6->sin6_port = 0;
902 sin6->sin6_addr = ip6->saddr;
903 sin6->sin6_flowinfo = 0;
904 if (np->sndflow)
905 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
906 sin6->sin6_scope_id =
907 ipv6_iface_scope_id(&sin6->sin6_addr,
908 inet6_iif(skb));
909 *addr_len = sizeof(*sin6);
910 }
911
912 if (inet6_sk(sk)->rxopt.all)
913 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
914 if (skb->protocol == htons(ETH_P_IPV6) &&
915 inet6_sk(sk)->rxopt.all)
916 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
917 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
918 ip_cmsg_recv(msg, skb);
919#endif
920 } else {
921 BUG();
922 }
923
924 err = copied;
925
926done:
927 skb_free_datagram(sk, skb);
928out:
929 pr_debug("ping_recvmsg -> %d\n", err);
930 return err;
931}
932EXPORT_SYMBOL_GPL(ping_recvmsg);
933
934int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
935{
936 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
937 inet_sk(sk), inet_sk(sk)->inet_num, skb);
938 if (sock_queue_rcv_skb(sk, skb) < 0) {
939 kfree_skb(skb);
940 pr_debug("ping_queue_rcv_skb -> failed\n");
941 return -1;
942 }
943 return 0;
944}
945EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
946
947
948/*
949 * All we need to do is get the socket.
950 */
951
952bool ping_rcv(struct sk_buff *skb)
953{
954 struct sock *sk;
955 struct net *net = dev_net(skb->dev);
956 struct icmphdr *icmph = icmp_hdr(skb);
957 bool rc = false;
958
959 /* We assume the packet has already been checked by icmp_rcv */
960
961 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
962 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
963
964 /* Push ICMP header back */
965 skb_push(skb, skb->data - (u8 *)icmph);
966
967 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
968 if (sk) {
969 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
970
971 pr_debug("rcv on socket %p\n", sk);
972 if (skb2 && !ping_queue_rcv_skb(sk, skb2))
973 rc = true;
974 sock_put(sk);
975 }
976
977 if (!rc)
978 pr_debug("no socket, dropping\n");
979
980 return rc;
981}
982EXPORT_SYMBOL_GPL(ping_rcv);
983
984struct proto ping_prot = {
985 .name = "PING",
986 .owner = THIS_MODULE,
987 .init = ping_init_sock,
988 .close = ping_close,
989 .connect = ip4_datagram_connect,
990 .disconnect = __udp_disconnect,
991 .setsockopt = ip_setsockopt,
992 .getsockopt = ip_getsockopt,
993 .sendmsg = ping_v4_sendmsg,
994 .recvmsg = ping_recvmsg,
995 .bind = ping_bind,
996 .backlog_rcv = ping_queue_rcv_skb,
997 .release_cb = ip4_datagram_release_cb,
998 .hash = ping_hash,
999 .unhash = ping_unhash,
1000 .get_port = ping_get_port,
1001 .obj_size = sizeof(struct inet_sock),
1002};
1003EXPORT_SYMBOL(ping_prot);
1004
1005#ifdef CONFIG_PROC_FS
1006
1007static struct sock *ping_get_first(struct seq_file *seq, int start)
1008{
1009 struct sock *sk;
1010 struct ping_iter_state *state = seq->private;
1011 struct net *net = seq_file_net(seq);
1012
1013 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1014 ++state->bucket) {
1015 struct hlist_nulls_node *node;
1016 struct hlist_nulls_head *hslot;
1017
1018 hslot = &ping_table.hash[state->bucket];
1019
1020 if (hlist_nulls_empty(hslot))
1021 continue;
1022
1023 sk_nulls_for_each(sk, node, hslot) {
1024 if (net_eq(sock_net(sk), net) &&
1025 sk->sk_family == state->family)
1026 goto found;
1027 }
1028 }
1029 sk = NULL;
1030found:
1031 return sk;
1032}
1033
1034static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1035{
1036 struct ping_iter_state *state = seq->private;
1037 struct net *net = seq_file_net(seq);
1038
1039 do {
1040 sk = sk_nulls_next(sk);
1041 } while (sk && (!net_eq(sock_net(sk), net)));
1042
1043 if (!sk)
1044 return ping_get_first(seq, state->bucket + 1);
1045 return sk;
1046}
1047
1048static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1049{
1050 struct sock *sk = ping_get_first(seq, 0);
1051
1052 if (sk)
1053 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1054 --pos;
1055 return pos ? NULL : sk;
1056}
1057
1058void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1059 __acquires(ping_table.lock)
1060{
1061 struct ping_iter_state *state = seq->private;
1062 state->bucket = 0;
1063 state->family = family;
1064
1065 read_lock_bh(&ping_table.lock);
1066
1067 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1068}
1069EXPORT_SYMBOL_GPL(ping_seq_start);
1070
1071static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1072{
1073 return ping_seq_start(seq, pos, AF_INET);
1074}
1075
1076void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1077{
1078 struct sock *sk;
1079
1080 if (v == SEQ_START_TOKEN)
1081 sk = ping_get_idx(seq, 0);
1082 else
1083 sk = ping_get_next(seq, v);
1084
1085 ++*pos;
1086 return sk;
1087}
1088EXPORT_SYMBOL_GPL(ping_seq_next);
1089
1090void ping_seq_stop(struct seq_file *seq, void *v)
1091 __releases(ping_table.lock)
1092{
1093 read_unlock_bh(&ping_table.lock);
1094}
1095EXPORT_SYMBOL_GPL(ping_seq_stop);
1096
1097static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1098 int bucket)
1099{
1100 struct inet_sock *inet = inet_sk(sp);
1101 __be32 dest = inet->inet_daddr;
1102 __be32 src = inet->inet_rcv_saddr;
1103 __u16 destp = ntohs(inet->inet_dport);
1104 __u16 srcp = ntohs(inet->inet_sport);
1105
1106 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1107 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1108 bucket, src, srcp, dest, destp, sp->sk_state,
1109 sk_wmem_alloc_get(sp),
1110 sk_rmem_alloc_get(sp),
1111 0, 0L, 0,
1112 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1113 0, sock_i_ino(sp),
1114 refcount_read(&sp->sk_refcnt), sp,
1115 atomic_read(&sp->sk_drops));
1116}
1117
1118static int ping_v4_seq_show(struct seq_file *seq, void *v)
1119{
1120 seq_setwidth(seq, 127);
1121 if (v == SEQ_START_TOKEN)
1122 seq_puts(seq, " sl local_address rem_address st tx_queue "
1123 "rx_queue tr tm->when retrnsmt uid timeout "
1124 "inode ref pointer drops");
1125 else {
1126 struct ping_iter_state *state = seq->private;
1127
1128 ping_v4_format_sock(v, seq, state->bucket);
1129 }
1130 seq_pad(seq, '\n');
1131 return 0;
1132}
1133
1134static const struct seq_operations ping_v4_seq_ops = {
1135 .start = ping_v4_seq_start,
1136 .show = ping_v4_seq_show,
1137 .next = ping_seq_next,
1138 .stop = ping_seq_stop,
1139};
1140
1141static int __net_init ping_v4_proc_init_net(struct net *net)
1142{
1143 if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1144 sizeof(struct ping_iter_state)))
1145 return -ENOMEM;
1146 return 0;
1147}
1148
1149static void __net_exit ping_v4_proc_exit_net(struct net *net)
1150{
1151 remove_proc_entry("icmp", net->proc_net);
1152}
1153
1154static struct pernet_operations ping_v4_net_ops = {
1155 .init = ping_v4_proc_init_net,
1156 .exit = ping_v4_proc_exit_net,
1157};
1158
1159int __init ping_proc_init(void)
1160{
1161 return register_pernet_subsys(&ping_v4_net_ops);
1162}
1163
1164void ping_proc_exit(void)
1165{
1166 unregister_pernet_subsys(&ping_v4_net_ops);
1167}
1168
1169#endif
1170
1171void __init ping_init(void)
1172{
1173 int i;
1174
1175 for (i = 0; i < PING_HTABLE_SIZE; i++)
1176 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1177 rwlock_init(&ping_table.lock);
1178}
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
148int 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 return 0;
154}
155
156void ping_unhash(struct sock *sk)
157{
158 struct inet_sock *isk = inet_sk(sk);
159
160 pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
161 write_lock_bh(&ping_table.lock);
162 if (sk_hashed(sk)) {
163 hlist_nulls_del(&sk->sk_nulls_node);
164 sk_nulls_node_init(&sk->sk_nulls_node);
165 sock_put(sk);
166 isk->inet_num = 0;
167 isk->inet_sport = 0;
168 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
169 }
170 write_unlock_bh(&ping_table.lock);
171}
172EXPORT_SYMBOL_GPL(ping_unhash);
173
174static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
175{
176 struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
177 struct sock *sk = NULL;
178 struct inet_sock *isk;
179 struct hlist_nulls_node *hnode;
180 int dif = skb->dev->ifindex;
181
182 if (skb->protocol == htons(ETH_P_IP)) {
183 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
184 (int)ident, &ip_hdr(skb)->daddr, dif);
185#if IS_ENABLED(CONFIG_IPV6)
186 } else if (skb->protocol == htons(ETH_P_IPV6)) {
187 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
188 (int)ident, &ipv6_hdr(skb)->daddr, dif);
189#endif
190 }
191
192 read_lock_bh(&ping_table.lock);
193
194 ping_portaddr_for_each_entry(sk, hnode, hslot) {
195 isk = inet_sk(sk);
196
197 pr_debug("iterate\n");
198 if (isk->inet_num != ident)
199 continue;
200
201 if (skb->protocol == htons(ETH_P_IP) &&
202 sk->sk_family == AF_INET) {
203 pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
204 (int) isk->inet_num, &isk->inet_rcv_saddr,
205 sk->sk_bound_dev_if);
206
207 if (isk->inet_rcv_saddr &&
208 isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
209 continue;
210#if IS_ENABLED(CONFIG_IPV6)
211 } else if (skb->protocol == htons(ETH_P_IPV6) &&
212 sk->sk_family == AF_INET6) {
213
214 pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
215 (int) isk->inet_num,
216 &sk->sk_v6_rcv_saddr,
217 sk->sk_bound_dev_if);
218
219 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
220 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
221 &ipv6_hdr(skb)->daddr))
222 continue;
223#endif
224 } else {
225 continue;
226 }
227
228 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
229 continue;
230
231 sock_hold(sk);
232 goto exit;
233 }
234
235 sk = NULL;
236exit:
237 read_unlock_bh(&ping_table.lock);
238
239 return sk;
240}
241
242static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
243 kgid_t *high)
244{
245 kgid_t *data = net->ipv4.ping_group_range.range;
246 unsigned int seq;
247
248 do {
249 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
250
251 *low = data[0];
252 *high = data[1];
253 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
254}
255
256
257int ping_init_sock(struct sock *sk)
258{
259 struct net *net = sock_net(sk);
260 kgid_t group = current_egid();
261 struct group_info *group_info;
262 int i;
263 kgid_t low, high;
264 int ret = 0;
265
266 if (sk->sk_family == AF_INET6)
267 sk->sk_ipv6only = 1;
268
269 inet_get_ping_group_range_net(net, &low, &high);
270 if (gid_lte(low, group) && gid_lte(group, high))
271 return 0;
272
273 group_info = get_current_groups();
274 for (i = 0; i < group_info->ngroups; i++) {
275 kgid_t gid = group_info->gid[i];
276
277 if (gid_lte(low, gid) && gid_lte(gid, high))
278 goto out_release_group;
279 }
280
281 ret = -EACCES;
282
283out_release_group:
284 put_group_info(group_info);
285 return ret;
286}
287EXPORT_SYMBOL_GPL(ping_init_sock);
288
289void ping_close(struct sock *sk, long timeout)
290{
291 pr_debug("ping_close(sk=%p,sk->num=%u)\n",
292 inet_sk(sk), inet_sk(sk)->inet_num);
293 pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
294
295 sk_common_release(sk);
296}
297EXPORT_SYMBOL_GPL(ping_close);
298
299/* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
300static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
301 struct sockaddr *uaddr, int addr_len) {
302 struct net *net = sock_net(sk);
303 if (sk->sk_family == AF_INET) {
304 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
305 int chk_addr_ret;
306
307 if (addr_len < sizeof(*addr))
308 return -EINVAL;
309
310 if (addr->sin_family != AF_INET &&
311 !(addr->sin_family == AF_UNSPEC &&
312 addr->sin_addr.s_addr == htonl(INADDR_ANY)))
313 return -EAFNOSUPPORT;
314
315 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
316 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
317
318 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
319
320 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
321 chk_addr_ret = RTN_LOCAL;
322
323 if ((net->ipv4.sysctl_ip_nonlocal_bind == 0 &&
324 isk->freebind == 0 && isk->transparent == 0 &&
325 chk_addr_ret != RTN_LOCAL) ||
326 chk_addr_ret == RTN_MULTICAST ||
327 chk_addr_ret == RTN_BROADCAST)
328 return -EADDRNOTAVAIL;
329
330#if IS_ENABLED(CONFIG_IPV6)
331 } else if (sk->sk_family == AF_INET6) {
332 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
333 int addr_type, scoped, has_addr;
334 struct net_device *dev = NULL;
335
336 if (addr_len < sizeof(*addr))
337 return -EINVAL;
338
339 if (addr->sin6_family != AF_INET6)
340 return -EAFNOSUPPORT;
341
342 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
343 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
344
345 addr_type = ipv6_addr_type(&addr->sin6_addr);
346 scoped = __ipv6_addr_needs_scope_id(addr_type);
347 if ((addr_type != IPV6_ADDR_ANY &&
348 !(addr_type & IPV6_ADDR_UNICAST)) ||
349 (scoped && !addr->sin6_scope_id))
350 return -EINVAL;
351
352 rcu_read_lock();
353 if (addr->sin6_scope_id) {
354 dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
355 if (!dev) {
356 rcu_read_unlock();
357 return -ENODEV;
358 }
359 }
360 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
361 scoped);
362 rcu_read_unlock();
363
364 if (!(net->ipv6.sysctl.ip_nonlocal_bind ||
365 isk->freebind || isk->transparent || has_addr ||
366 addr_type == IPV6_ADDR_ANY))
367 return -EADDRNOTAVAIL;
368
369 if (scoped)
370 sk->sk_bound_dev_if = addr->sin6_scope_id;
371#endif
372 } else {
373 return -EAFNOSUPPORT;
374 }
375 return 0;
376}
377
378static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
379{
380 if (saddr->sa_family == AF_INET) {
381 struct inet_sock *isk = inet_sk(sk);
382 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
383 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
384#if IS_ENABLED(CONFIG_IPV6)
385 } else if (saddr->sa_family == AF_INET6) {
386 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
387 struct ipv6_pinfo *np = inet6_sk(sk);
388 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
389#endif
390 }
391}
392
393static void ping_clear_saddr(struct sock *sk, int dif)
394{
395 sk->sk_bound_dev_if = dif;
396 if (sk->sk_family == AF_INET) {
397 struct inet_sock *isk = inet_sk(sk);
398 isk->inet_rcv_saddr = isk->inet_saddr = 0;
399#if IS_ENABLED(CONFIG_IPV6)
400 } else if (sk->sk_family == AF_INET6) {
401 struct ipv6_pinfo *np = inet6_sk(sk);
402 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
403 memset(&np->saddr, 0, sizeof(np->saddr));
404#endif
405 }
406}
407/*
408 * We need our own bind because there are no privileged id's == local ports.
409 * Moreover, we don't allow binding to multi- and broadcast addresses.
410 */
411
412int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
413{
414 struct inet_sock *isk = inet_sk(sk);
415 unsigned short snum;
416 int err;
417 int dif = sk->sk_bound_dev_if;
418
419 err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
420 if (err)
421 return err;
422
423 lock_sock(sk);
424
425 err = -EINVAL;
426 if (isk->inet_num != 0)
427 goto out;
428
429 err = -EADDRINUSE;
430 ping_set_saddr(sk, uaddr);
431 snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
432 if (ping_get_port(sk, snum) != 0) {
433 ping_clear_saddr(sk, dif);
434 goto out;
435 }
436
437 pr_debug("after bind(): num = %hu, dif = %d\n",
438 isk->inet_num,
439 sk->sk_bound_dev_if);
440
441 err = 0;
442 if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
443 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
444#if IS_ENABLED(CONFIG_IPV6)
445 if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
446 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
447#endif
448
449 if (snum)
450 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
451 isk->inet_sport = htons(isk->inet_num);
452 isk->inet_daddr = 0;
453 isk->inet_dport = 0;
454
455#if IS_ENABLED(CONFIG_IPV6)
456 if (sk->sk_family == AF_INET6)
457 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
458#endif
459
460 sk_dst_reset(sk);
461out:
462 release_sock(sk);
463 pr_debug("ping_v4_bind -> %d\n", err);
464 return err;
465}
466EXPORT_SYMBOL_GPL(ping_bind);
467
468/*
469 * Is this a supported type of ICMP message?
470 */
471
472static inline int ping_supported(int family, int type, int code)
473{
474 return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
475 (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
476}
477
478/*
479 * This routine is called by the ICMP module when it gets some
480 * sort of error condition.
481 */
482
483void ping_err(struct sk_buff *skb, int offset, u32 info)
484{
485 int family;
486 struct icmphdr *icmph;
487 struct inet_sock *inet_sock;
488 int type;
489 int code;
490 struct net *net = dev_net(skb->dev);
491 struct sock *sk;
492 int harderr;
493 int err;
494
495 if (skb->protocol == htons(ETH_P_IP)) {
496 family = AF_INET;
497 type = icmp_hdr(skb)->type;
498 code = icmp_hdr(skb)->code;
499 icmph = (struct icmphdr *)(skb->data + offset);
500 } else if (skb->protocol == htons(ETH_P_IPV6)) {
501 family = AF_INET6;
502 type = icmp6_hdr(skb)->icmp6_type;
503 code = icmp6_hdr(skb)->icmp6_code;
504 icmph = (struct icmphdr *) (skb->data + offset);
505 } else {
506 BUG();
507 }
508
509 /* We assume the packet has already been checked by icmp_unreach */
510
511 if (!ping_supported(family, icmph->type, icmph->code))
512 return;
513
514 pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
515 skb->protocol, type, code, ntohs(icmph->un.echo.id),
516 ntohs(icmph->un.echo.sequence));
517
518 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
519 if (!sk) {
520 pr_debug("no socket, dropping\n");
521 return; /* No socket for error */
522 }
523 pr_debug("err on socket %p\n", sk);
524
525 err = 0;
526 harderr = 0;
527 inet_sock = inet_sk(sk);
528
529 if (skb->protocol == htons(ETH_P_IP)) {
530 switch (type) {
531 default:
532 case ICMP_TIME_EXCEEDED:
533 err = EHOSTUNREACH;
534 break;
535 case ICMP_SOURCE_QUENCH:
536 /* This is not a real error but ping wants to see it.
537 * Report it with some fake errno.
538 */
539 err = EREMOTEIO;
540 break;
541 case ICMP_PARAMETERPROB:
542 err = EPROTO;
543 harderr = 1;
544 break;
545 case ICMP_DEST_UNREACH:
546 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
547 ipv4_sk_update_pmtu(skb, sk, info);
548 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
549 err = EMSGSIZE;
550 harderr = 1;
551 break;
552 }
553 goto out;
554 }
555 err = EHOSTUNREACH;
556 if (code <= NR_ICMP_UNREACH) {
557 harderr = icmp_err_convert[code].fatal;
558 err = icmp_err_convert[code].errno;
559 }
560 break;
561 case ICMP_REDIRECT:
562 /* See ICMP_SOURCE_QUENCH */
563 ipv4_sk_redirect(skb, sk);
564 err = EREMOTEIO;
565 break;
566 }
567#if IS_ENABLED(CONFIG_IPV6)
568 } else if (skb->protocol == htons(ETH_P_IPV6)) {
569 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
570#endif
571 }
572
573 /*
574 * RFC1122: OK. Passes ICMP errors back to application, as per
575 * 4.1.3.3.
576 */
577 if ((family == AF_INET && !inet_sock->recverr) ||
578 (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
579 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
580 goto out;
581 } else {
582 if (family == AF_INET) {
583 ip_icmp_error(sk, skb, err, 0 /* no remote port */,
584 info, (u8 *)icmph);
585#if IS_ENABLED(CONFIG_IPV6)
586 } else if (family == AF_INET6) {
587 pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
588 info, (u8 *)icmph);
589#endif
590 }
591 }
592 sk->sk_err = err;
593 sk->sk_error_report(sk);
594out:
595 sock_put(sk);
596}
597EXPORT_SYMBOL_GPL(ping_err);
598
599/*
600 * Copy and checksum an ICMP Echo packet from user space into a buffer
601 * starting from the payload.
602 */
603
604int ping_getfrag(void *from, char *to,
605 int offset, int fraglen, int odd, struct sk_buff *skb)
606{
607 struct pingfakehdr *pfh = (struct pingfakehdr *)from;
608
609 if (offset == 0) {
610 fraglen -= sizeof(struct icmphdr);
611 if (fraglen < 0)
612 BUG();
613 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
614 fraglen, &pfh->wcheck,
615 &pfh->msg->msg_iter))
616 return -EFAULT;
617 } else if (offset < sizeof(struct icmphdr)) {
618 BUG();
619 } else {
620 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
621 &pfh->msg->msg_iter))
622 return -EFAULT;
623 }
624
625#if IS_ENABLED(CONFIG_IPV6)
626 /* For IPv6, checksum each skb as we go along, as expected by
627 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
628 * wcheck, it will be finalized in ping_v4_push_pending_frames.
629 */
630 if (pfh->family == AF_INET6) {
631 skb->csum = pfh->wcheck;
632 skb->ip_summed = CHECKSUM_NONE;
633 pfh->wcheck = 0;
634 }
635#endif
636
637 return 0;
638}
639EXPORT_SYMBOL_GPL(ping_getfrag);
640
641static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
642 struct flowi4 *fl4)
643{
644 struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
645
646 if (!skb)
647 return 0;
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
656int 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 /* Must have at least a full ICMP header. */
664 if (len < icmph_len)
665 return -EINVAL;
666
667 /*
668 * Check the flags.
669 */
670
671 /* Mirror BSD error message compatibility */
672 if (msg->msg_flags & MSG_OOB)
673 return -EOPNOTSUPP;
674
675 /*
676 * Fetch the ICMP header provided by the userland.
677 * iovec is modified! The ICMP header is consumed.
678 */
679 if (memcpy_from_msg(user_icmph, msg, icmph_len))
680 return -EFAULT;
681
682 if (family == AF_INET) {
683 type = ((struct icmphdr *) user_icmph)->type;
684 code = ((struct icmphdr *) user_icmph)->code;
685#if IS_ENABLED(CONFIG_IPV6)
686 } else if (family == AF_INET6) {
687 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
688 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
689#endif
690 } else {
691 BUG();
692 }
693
694 if (!ping_supported(family, type, code))
695 return -EINVAL;
696
697 return 0;
698}
699EXPORT_SYMBOL_GPL(ping_common_sendmsg);
700
701static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
702{
703 struct net *net = sock_net(sk);
704 struct flowi4 fl4;
705 struct inet_sock *inet = inet_sk(sk);
706 struct ipcm_cookie ipc;
707 struct icmphdr user_icmph;
708 struct pingfakehdr pfh;
709 struct rtable *rt = NULL;
710 struct ip_options_data opt_copy;
711 int free = 0;
712 __be32 saddr, daddr, faddr;
713 u8 tos;
714 int err;
715
716 pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
717
718 err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
719 sizeof(user_icmph));
720 if (err)
721 return err;
722
723 /*
724 * Get and verify the address.
725 */
726
727 if (msg->msg_name) {
728 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
729 if (msg->msg_namelen < sizeof(*usin))
730 return -EINVAL;
731 if (usin->sin_family != AF_INET)
732 return -EAFNOSUPPORT;
733 daddr = usin->sin_addr.s_addr;
734 /* no remote port */
735 } else {
736 if (sk->sk_state != TCP_ESTABLISHED)
737 return -EDESTADDRREQ;
738 daddr = inet->inet_daddr;
739 /* no remote port */
740 }
741
742 ipc.sockc.tsflags = sk->sk_tsflags;
743 ipc.addr = inet->inet_saddr;
744 ipc.opt = NULL;
745 ipc.oif = sk->sk_bound_dev_if;
746 ipc.tx_flags = 0;
747 ipc.ttl = 0;
748 ipc.tos = -1;
749
750 if (msg->msg_controllen) {
751 err = ip_cmsg_send(sk, msg, &ipc, false);
752 if (unlikely(err)) {
753 kfree(ipc.opt);
754 return err;
755 }
756 if (ipc.opt)
757 free = 1;
758 }
759 if (!ipc.opt) {
760 struct ip_options_rcu *inet_opt;
761
762 rcu_read_lock();
763 inet_opt = rcu_dereference(inet->inet_opt);
764 if (inet_opt) {
765 memcpy(&opt_copy, inet_opt,
766 sizeof(*inet_opt) + inet_opt->opt.optlen);
767 ipc.opt = &opt_copy.opt;
768 }
769 rcu_read_unlock();
770 }
771
772 sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
773
774 saddr = ipc.addr;
775 ipc.addr = faddr = daddr;
776
777 if (ipc.opt && ipc.opt->opt.srr) {
778 if (!daddr) {
779 err = -EINVAL;
780 goto out_free;
781 }
782 faddr = ipc.opt->opt.faddr;
783 }
784 tos = get_rttos(&ipc, inet);
785 if (sock_flag(sk, SOCK_LOCALROUTE) ||
786 (msg->msg_flags & MSG_DONTROUTE) ||
787 (ipc.opt && ipc.opt->opt.is_strictroute)) {
788 tos |= RTO_ONLINK;
789 }
790
791 if (ipv4_is_multicast(daddr)) {
792 if (!ipc.oif)
793 ipc.oif = inet->mc_index;
794 if (!saddr)
795 saddr = inet->mc_addr;
796 } else if (!ipc.oif)
797 ipc.oif = inet->uc_index;
798
799 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
800 RT_SCOPE_UNIVERSE, sk->sk_protocol,
801 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
802 sk->sk_uid);
803
804 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
805 rt = ip_route_output_flow(net, &fl4, sk);
806 if (IS_ERR(rt)) {
807 err = PTR_ERR(rt);
808 rt = NULL;
809 if (err == -ENETUNREACH)
810 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
811 goto out;
812 }
813
814 err = -EACCES;
815 if ((rt->rt_flags & RTCF_BROADCAST) &&
816 !sock_flag(sk, SOCK_BROADCAST))
817 goto out;
818
819 if (msg->msg_flags & MSG_CONFIRM)
820 goto do_confirm;
821back_from_confirm:
822
823 if (!ipc.addr)
824 ipc.addr = fl4.daddr;
825
826 lock_sock(sk);
827
828 pfh.icmph.type = user_icmph.type; /* already checked */
829 pfh.icmph.code = user_icmph.code; /* ditto */
830 pfh.icmph.checksum = 0;
831 pfh.icmph.un.echo.id = inet->inet_sport;
832 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
833 pfh.msg = msg;
834 pfh.wcheck = 0;
835 pfh.family = AF_INET;
836
837 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
838 0, &ipc, &rt, msg->msg_flags);
839 if (err)
840 ip_flush_pending_frames(sk);
841 else
842 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
843 release_sock(sk);
844
845out:
846 ip_rt_put(rt);
847out_free:
848 if (free)
849 kfree(ipc.opt);
850 if (!err) {
851 icmp_out_count(sock_net(sk), user_icmph.type);
852 return len;
853 }
854 return err;
855
856do_confirm:
857 if (msg->msg_flags & MSG_PROBE)
858 dst_confirm_neigh(&rt->dst, &fl4.daddr);
859 if (!(msg->msg_flags & MSG_PROBE) || len)
860 goto back_from_confirm;
861 err = 0;
862 goto out;
863}
864
865int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
866 int flags, int *addr_len)
867{
868 struct inet_sock *isk = inet_sk(sk);
869 int family = sk->sk_family;
870 struct sk_buff *skb;
871 int copied, err;
872
873 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
874
875 err = -EOPNOTSUPP;
876 if (flags & MSG_OOB)
877 goto out;
878
879 if (flags & MSG_ERRQUEUE)
880 return inet_recv_error(sk, msg, len, addr_len);
881
882 skb = skb_recv_datagram(sk, flags, noblock, &err);
883 if (!skb)
884 goto out;
885
886 copied = skb->len;
887 if (copied > len) {
888 msg->msg_flags |= MSG_TRUNC;
889 copied = len;
890 }
891
892 /* Don't bother checking the checksum */
893 err = skb_copy_datagram_msg(skb, 0, msg, copied);
894 if (err)
895 goto done;
896
897 sock_recv_timestamp(msg, sk, skb);
898
899 /* Copy the address and add cmsg data. */
900 if (family == AF_INET) {
901 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
902
903 if (sin) {
904 sin->sin_family = AF_INET;
905 sin->sin_port = 0 /* skb->h.uh->source */;
906 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
907 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
908 *addr_len = sizeof(*sin);
909 }
910
911 if (isk->cmsg_flags)
912 ip_cmsg_recv(msg, skb);
913
914#if IS_ENABLED(CONFIG_IPV6)
915 } else if (family == AF_INET6) {
916 struct ipv6_pinfo *np = inet6_sk(sk);
917 struct ipv6hdr *ip6 = ipv6_hdr(skb);
918 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
919
920 if (sin6) {
921 sin6->sin6_family = AF_INET6;
922 sin6->sin6_port = 0;
923 sin6->sin6_addr = ip6->saddr;
924 sin6->sin6_flowinfo = 0;
925 if (np->sndflow)
926 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
927 sin6->sin6_scope_id =
928 ipv6_iface_scope_id(&sin6->sin6_addr,
929 inet6_iif(skb));
930 *addr_len = sizeof(*sin6);
931 }
932
933 if (inet6_sk(sk)->rxopt.all)
934 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
935 if (skb->protocol == htons(ETH_P_IPV6) &&
936 inet6_sk(sk)->rxopt.all)
937 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
938 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
939 ip_cmsg_recv(msg, skb);
940#endif
941 } else {
942 BUG();
943 }
944
945 err = copied;
946
947done:
948 skb_free_datagram(sk, skb);
949out:
950 pr_debug("ping_recvmsg -> %d\n", err);
951 return err;
952}
953EXPORT_SYMBOL_GPL(ping_recvmsg);
954
955int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
956{
957 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
958 inet_sk(sk), inet_sk(sk)->inet_num, skb);
959 if (sock_queue_rcv_skb(sk, skb) < 0) {
960 kfree_skb(skb);
961 pr_debug("ping_queue_rcv_skb -> failed\n");
962 return -1;
963 }
964 return 0;
965}
966EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
967
968
969/*
970 * All we need to do is get the socket.
971 */
972
973bool ping_rcv(struct sk_buff *skb)
974{
975 struct sock *sk;
976 struct net *net = dev_net(skb->dev);
977 struct icmphdr *icmph = icmp_hdr(skb);
978
979 /* We assume the packet has already been checked by icmp_rcv */
980
981 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
982 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
983
984 /* Push ICMP header back */
985 skb_push(skb, skb->data - (u8 *)icmph);
986
987 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
988 if (sk) {
989 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
990
991 pr_debug("rcv on socket %p\n", sk);
992 if (skb2)
993 ping_queue_rcv_skb(sk, skb2);
994 sock_put(sk);
995 return true;
996 }
997 pr_debug("no socket, dropping\n");
998
999 return false;
1000}
1001EXPORT_SYMBOL_GPL(ping_rcv);
1002
1003struct proto ping_prot = {
1004 .name = "PING",
1005 .owner = THIS_MODULE,
1006 .init = ping_init_sock,
1007 .close = ping_close,
1008 .connect = ip4_datagram_connect,
1009 .disconnect = __udp_disconnect,
1010 .setsockopt = ip_setsockopt,
1011 .getsockopt = ip_getsockopt,
1012 .sendmsg = ping_v4_sendmsg,
1013 .recvmsg = ping_recvmsg,
1014 .bind = ping_bind,
1015 .backlog_rcv = ping_queue_rcv_skb,
1016 .release_cb = ip4_datagram_release_cb,
1017 .hash = ping_hash,
1018 .unhash = ping_unhash,
1019 .get_port = ping_get_port,
1020 .obj_size = sizeof(struct inet_sock),
1021};
1022EXPORT_SYMBOL(ping_prot);
1023
1024#ifdef CONFIG_PROC_FS
1025
1026static struct sock *ping_get_first(struct seq_file *seq, int start)
1027{
1028 struct sock *sk;
1029 struct ping_iter_state *state = seq->private;
1030 struct net *net = seq_file_net(seq);
1031
1032 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1033 ++state->bucket) {
1034 struct hlist_nulls_node *node;
1035 struct hlist_nulls_head *hslot;
1036
1037 hslot = &ping_table.hash[state->bucket];
1038
1039 if (hlist_nulls_empty(hslot))
1040 continue;
1041
1042 sk_nulls_for_each(sk, node, hslot) {
1043 if (net_eq(sock_net(sk), net) &&
1044 sk->sk_family == state->family)
1045 goto found;
1046 }
1047 }
1048 sk = NULL;
1049found:
1050 return sk;
1051}
1052
1053static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1054{
1055 struct ping_iter_state *state = seq->private;
1056 struct net *net = seq_file_net(seq);
1057
1058 do {
1059 sk = sk_nulls_next(sk);
1060 } while (sk && (!net_eq(sock_net(sk), net)));
1061
1062 if (!sk)
1063 return ping_get_first(seq, state->bucket + 1);
1064 return sk;
1065}
1066
1067static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1068{
1069 struct sock *sk = ping_get_first(seq, 0);
1070
1071 if (sk)
1072 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1073 --pos;
1074 return pos ? NULL : sk;
1075}
1076
1077void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1078 __acquires(ping_table.lock)
1079{
1080 struct ping_iter_state *state = seq->private;
1081 state->bucket = 0;
1082 state->family = family;
1083
1084 read_lock_bh(&ping_table.lock);
1085
1086 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1087}
1088EXPORT_SYMBOL_GPL(ping_seq_start);
1089
1090static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1091{
1092 return ping_seq_start(seq, pos, AF_INET);
1093}
1094
1095void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1096{
1097 struct sock *sk;
1098
1099 if (v == SEQ_START_TOKEN)
1100 sk = ping_get_idx(seq, 0);
1101 else
1102 sk = ping_get_next(seq, v);
1103
1104 ++*pos;
1105 return sk;
1106}
1107EXPORT_SYMBOL_GPL(ping_seq_next);
1108
1109void ping_seq_stop(struct seq_file *seq, void *v)
1110 __releases(ping_table.lock)
1111{
1112 read_unlock_bh(&ping_table.lock);
1113}
1114EXPORT_SYMBOL_GPL(ping_seq_stop);
1115
1116static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1117 int bucket)
1118{
1119 struct inet_sock *inet = inet_sk(sp);
1120 __be32 dest = inet->inet_daddr;
1121 __be32 src = inet->inet_rcv_saddr;
1122 __u16 destp = ntohs(inet->inet_dport);
1123 __u16 srcp = ntohs(inet->inet_sport);
1124
1125 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1126 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1127 bucket, src, srcp, dest, destp, sp->sk_state,
1128 sk_wmem_alloc_get(sp),
1129 sk_rmem_alloc_get(sp),
1130 0, 0L, 0,
1131 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1132 0, sock_i_ino(sp),
1133 refcount_read(&sp->sk_refcnt), sp,
1134 atomic_read(&sp->sk_drops));
1135}
1136
1137static int ping_v4_seq_show(struct seq_file *seq, void *v)
1138{
1139 seq_setwidth(seq, 127);
1140 if (v == SEQ_START_TOKEN)
1141 seq_puts(seq, " sl local_address rem_address st tx_queue "
1142 "rx_queue tr tm->when retrnsmt uid timeout "
1143 "inode ref pointer drops");
1144 else {
1145 struct ping_iter_state *state = seq->private;
1146
1147 ping_v4_format_sock(v, seq, state->bucket);
1148 }
1149 seq_pad(seq, '\n');
1150 return 0;
1151}
1152
1153static int ping_seq_open(struct inode *inode, struct file *file)
1154{
1155 struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1156 return seq_open_net(inode, file, &afinfo->seq_ops,
1157 sizeof(struct ping_iter_state));
1158}
1159
1160const struct file_operations ping_seq_fops = {
1161 .open = ping_seq_open,
1162 .read = seq_read,
1163 .llseek = seq_lseek,
1164 .release = seq_release_net,
1165};
1166EXPORT_SYMBOL_GPL(ping_seq_fops);
1167
1168static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1169 .name = "icmp",
1170 .family = AF_INET,
1171 .seq_fops = &ping_seq_fops,
1172 .seq_ops = {
1173 .start = ping_v4_seq_start,
1174 .show = ping_v4_seq_show,
1175 .next = ping_seq_next,
1176 .stop = ping_seq_stop,
1177 },
1178};
1179
1180int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1181{
1182 struct proc_dir_entry *p;
1183 p = proc_create_data(afinfo->name, 0444, net->proc_net,
1184 afinfo->seq_fops, afinfo);
1185 if (!p)
1186 return -ENOMEM;
1187 return 0;
1188}
1189EXPORT_SYMBOL_GPL(ping_proc_register);
1190
1191void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1192{
1193 remove_proc_entry(afinfo->name, net->proc_net);
1194}
1195EXPORT_SYMBOL_GPL(ping_proc_unregister);
1196
1197static int __net_init ping_v4_proc_init_net(struct net *net)
1198{
1199 return ping_proc_register(net, &ping_v4_seq_afinfo);
1200}
1201
1202static void __net_exit ping_v4_proc_exit_net(struct net *net)
1203{
1204 ping_proc_unregister(net, &ping_v4_seq_afinfo);
1205}
1206
1207static struct pernet_operations ping_v4_net_ops = {
1208 .init = ping_v4_proc_init_net,
1209 .exit = ping_v4_proc_exit_net,
1210};
1211
1212int __init ping_proc_init(void)
1213{
1214 return register_pernet_subsys(&ping_v4_net_ops);
1215}
1216
1217void ping_proc_exit(void)
1218{
1219 unregister_pernet_subsys(&ping_v4_net_ops);
1220}
1221
1222#endif
1223
1224void __init ping_init(void)
1225{
1226 int i;
1227
1228 for (i = 0; i < PING_HTABLE_SIZE; i++)
1229 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1230 rwlock_init(&ping_table.lock);
1231}