]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - net/ipv4/ping.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
UBUNTU: Ubuntu-5.15.0-39.42
[mirror_ubuntu-jammy-kernel.git] / net / ipv4 / ping.c
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * "Ping" sockets
8 *
9 * Based on ipv4/udp.c code.
10 *
11 * Authors: Vasiliy Kulikov / Openwall (for Linux 2.6),
12 * Pavel Kankovsky (for Linux 2.4.32)
13 *
14 * Pavel gave all rights to bugs to Vasiliy,
15 * none of the bugs are Pavel's now.
16 */
17
18 #include <linux/uaccess.h>
19 #include <linux/types.h>
20 #include <linux/fcntl.h>
21 #include <linux/socket.h>
22 #include <linux/sockios.h>
23 #include <linux/in.h>
24 #include <linux/errno.h>
25 #include <linux/timer.h>
26 #include <linux/mm.h>
27 #include <linux/inet.h>
28 #include <linux/netdevice.h>
29 #include <net/snmp.h>
30 #include <net/ip.h>
31 #include <net/icmp.h>
32 #include <net/protocol.h>
33 #include <linux/skbuff.h>
34 #include <linux/proc_fs.h>
35 #include <linux/export.h>
36 #include <net/sock.h>
37 #include <net/ping.h>
38 #include <net/udp.h>
39 #include <net/route.h>
40 #include <net/inet_common.h>
41 #include <net/checksum.h>
42
43 #if IS_ENABLED(CONFIG_IPV6)
44 #include <linux/in6.h>
45 #include <linux/icmpv6.h>
46 #include <net/addrconf.h>
47 #include <net/ipv6.h>
48 #include <net/transp_v6.h>
49 #endif
50
51 struct ping_table {
52 struct hlist_nulls_head hash[PING_HTABLE_SIZE];
53 rwlock_t lock;
54 };
55
56 static struct ping_table ping_table;
57 struct pingv6_ops pingv6_ops;
58 EXPORT_SYMBOL_GPL(pingv6_ops);
59
60 static u16 ping_port_rover;
61
62 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
63 {
64 u32 res = (num + net_hash_mix(net)) & mask;
65
66 pr_debug("hash(%u) = %u\n", num, res);
67 return res;
68 }
69 EXPORT_SYMBOL_GPL(ping_hash);
70
71 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
72 struct net *net, unsigned int num)
73 {
74 return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
75 }
76
77 int ping_get_port(struct sock *sk, unsigned short ident)
78 {
79 struct hlist_nulls_node *node;
80 struct hlist_nulls_head *hlist;
81 struct inet_sock *isk, *isk2;
82 struct sock *sk2 = NULL;
83
84 isk = inet_sk(sk);
85 write_lock_bh(&ping_table.lock);
86 if (ident == 0) {
87 u32 i;
88 u16 result = ping_port_rover + 1;
89
90 for (i = 0; i < (1L << 16); i++, result++) {
91 if (!result)
92 result++; /* avoid zero */
93 hlist = ping_hashslot(&ping_table, sock_net(sk),
94 result);
95 ping_portaddr_for_each_entry(sk2, node, hlist) {
96 isk2 = inet_sk(sk2);
97
98 if (isk2->inet_num == result)
99 goto next_port;
100 }
101
102 /* found */
103 ping_port_rover = ident = result;
104 break;
105 next_port:
106 ;
107 }
108 if (i >= (1L << 16))
109 goto fail;
110 } else {
111 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
112 ping_portaddr_for_each_entry(sk2, node, hlist) {
113 isk2 = inet_sk(sk2);
114
115 /* BUG? Why is this reuse and not reuseaddr? ping.c
116 * doesn't turn off SO_REUSEADDR, and it doesn't expect
117 * that other ping processes can steal its packets.
118 */
119 if ((isk2->inet_num == ident) &&
120 (sk2 != sk) &&
121 (!sk2->sk_reuse || !sk->sk_reuse))
122 goto fail;
123 }
124 }
125
126 pr_debug("found port/ident = %d\n", ident);
127 isk->inet_num = ident;
128 if (sk_unhashed(sk)) {
129 pr_debug("was not hashed\n");
130 sock_hold(sk);
131 hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
132 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
133 }
134 write_unlock_bh(&ping_table.lock);
135 return 0;
136
137 fail:
138 write_unlock_bh(&ping_table.lock);
139 return 1;
140 }
141 EXPORT_SYMBOL_GPL(ping_get_port);
142
143 int ping_hash(struct sock *sk)
144 {
145 pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
146 BUG(); /* "Please do not press this button again." */
147
148 return 0;
149 }
150
151 void ping_unhash(struct sock *sk)
152 {
153 struct inet_sock *isk = inet_sk(sk);
154
155 pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
156 write_lock_bh(&ping_table.lock);
157 if (sk_hashed(sk)) {
158 hlist_nulls_del(&sk->sk_nulls_node);
159 sk_nulls_node_init(&sk->sk_nulls_node);
160 sock_put(sk);
161 isk->inet_num = 0;
162 isk->inet_sport = 0;
163 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
164 }
165 write_unlock_bh(&ping_table.lock);
166 }
167 EXPORT_SYMBOL_GPL(ping_unhash);
168
169 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
170 {
171 struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
172 struct sock *sk = NULL;
173 struct inet_sock *isk;
174 struct hlist_nulls_node *hnode;
175 int dif, sdif;
176
177 if (skb->protocol == htons(ETH_P_IP)) {
178 dif = inet_iif(skb);
179 sdif = inet_sdif(skb);
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 dif = inet6_iif(skb);
185 sdif = inet6_sdif(skb);
186 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
187 (int)ident, &ipv6_hdr(skb)->daddr, dif);
188 #endif
189 } else {
190 return NULL;
191 }
192
193 read_lock_bh(&ping_table.lock);
194
195 ping_portaddr_for_each_entry(sk, hnode, hslot) {
196 isk = inet_sk(sk);
197
198 pr_debug("iterate\n");
199 if (isk->inet_num != ident)
200 continue;
201
202 if (skb->protocol == htons(ETH_P_IP) &&
203 sk->sk_family == AF_INET) {
204 pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
205 (int) isk->inet_num, &isk->inet_rcv_saddr,
206 sk->sk_bound_dev_if);
207
208 if (isk->inet_rcv_saddr &&
209 isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
210 continue;
211 #if IS_ENABLED(CONFIG_IPV6)
212 } else if (skb->protocol == htons(ETH_P_IPV6) &&
213 sk->sk_family == AF_INET6) {
214
215 pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
216 (int) isk->inet_num,
217 &sk->sk_v6_rcv_saddr,
218 sk->sk_bound_dev_if);
219
220 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
221 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
222 &ipv6_hdr(skb)->daddr))
223 continue;
224 #endif
225 } else {
226 continue;
227 }
228
229 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
230 sk->sk_bound_dev_if != sdif)
231 continue;
232
233 sock_hold(sk);
234 goto exit;
235 }
236
237 sk = NULL;
238 exit:
239 read_unlock_bh(&ping_table.lock);
240
241 return sk;
242 }
243
244 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
245 kgid_t *high)
246 {
247 kgid_t *data = net->ipv4.ping_group_range.range;
248 unsigned int seq;
249
250 do {
251 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
252
253 *low = data[0];
254 *high = data[1];
255 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
256 }
257
258
259 int ping_init_sock(struct sock *sk)
260 {
261 struct net *net = sock_net(sk);
262 kgid_t group = current_egid();
263 struct group_info *group_info;
264 int i;
265 kgid_t low, high;
266 int ret = 0;
267
268 if (sk->sk_family == AF_INET6)
269 sk->sk_ipv6only = 1;
270
271 inet_get_ping_group_range_net(net, &low, &high);
272 if (gid_lte(low, group) && gid_lte(group, high))
273 return 0;
274
275 group_info = get_current_groups();
276 for (i = 0; i < group_info->ngroups; i++) {
277 kgid_t gid = group_info->gid[i];
278
279 if (gid_lte(low, gid) && gid_lte(gid, high))
280 goto out_release_group;
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", refcount_read(&sk->sk_refcnt));
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 {
305 struct net *net = sock_net(sk);
306 if (sk->sk_family == AF_INET) {
307 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
308 int chk_addr_ret;
309
310 if (addr_len < sizeof(*addr))
311 return -EINVAL;
312
313 if (addr->sin_family != AF_INET &&
314 !(addr->sin_family == AF_UNSPEC &&
315 addr->sin_addr.s_addr == htonl(INADDR_ANY)))
316 return -EAFNOSUPPORT;
317
318 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
319 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
320
321 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
322 chk_addr_ret = RTN_LOCAL;
323 else
324 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
325
326 if ((!inet_can_nonlocal_bind(net, isk) &&
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 (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
367 addr_type == IPV6_ADDR_ANY))
368 return -EADDRNOTAVAIL;
369
370 if (scoped)
371 sk->sk_bound_dev_if = addr->sin6_scope_id;
372 #endif
373 } else {
374 return -EAFNOSUPPORT;
375 }
376 return 0;
377 }
378
379 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
380 {
381 if (saddr->sa_family == AF_INET) {
382 struct inet_sock *isk = inet_sk(sk);
383 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
384 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
385 #if IS_ENABLED(CONFIG_IPV6)
386 } else if (saddr->sa_family == AF_INET6) {
387 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
388 struct ipv6_pinfo *np = inet6_sk(sk);
389 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
390 #endif
391 }
392 }
393
394 /*
395 * We need our own bind because there are no privileged id's == local ports.
396 * Moreover, we don't allow binding to multi- and broadcast addresses.
397 */
398
399 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
400 {
401 struct inet_sock *isk = inet_sk(sk);
402 unsigned short snum;
403 int err;
404 int dif = sk->sk_bound_dev_if;
405
406 err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
407 if (err)
408 return err;
409
410 lock_sock(sk);
411
412 err = -EINVAL;
413 if (isk->inet_num != 0)
414 goto out;
415
416 err = -EADDRINUSE;
417 snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
418 if (ping_get_port(sk, snum) != 0) {
419 /* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
420 sk->sk_bound_dev_if = dif;
421 goto out;
422 }
423 ping_set_saddr(sk, uaddr);
424
425 pr_debug("after bind(): num = %hu, dif = %d\n",
426 isk->inet_num,
427 sk->sk_bound_dev_if);
428
429 err = 0;
430 if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
431 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
432 #if IS_ENABLED(CONFIG_IPV6)
433 if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
434 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
435 #endif
436
437 if (snum)
438 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
439 isk->inet_sport = htons(isk->inet_num);
440 isk->inet_daddr = 0;
441 isk->inet_dport = 0;
442
443 #if IS_ENABLED(CONFIG_IPV6)
444 if (sk->sk_family == AF_INET6)
445 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
446 #endif
447
448 sk_dst_reset(sk);
449 out:
450 release_sock(sk);
451 pr_debug("ping_v4_bind -> %d\n", err);
452 return err;
453 }
454 EXPORT_SYMBOL_GPL(ping_bind);
455
456 /*
457 * Is this a supported type of ICMP message?
458 */
459
460 static inline int ping_supported(int family, int type, int code)
461 {
462 return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
463 (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
464 (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
465 (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
466 }
467
468 /*
469 * This routine is called by the ICMP module when it gets some
470 * sort of error condition.
471 */
472
473 void ping_err(struct sk_buff *skb, int offset, u32 info)
474 {
475 int family;
476 struct icmphdr *icmph;
477 struct inet_sock *inet_sock;
478 int type;
479 int code;
480 struct net *net = dev_net(skb->dev);
481 struct sock *sk;
482 int harderr;
483 int err;
484
485 if (skb->protocol == htons(ETH_P_IP)) {
486 family = AF_INET;
487 type = icmp_hdr(skb)->type;
488 code = icmp_hdr(skb)->code;
489 icmph = (struct icmphdr *)(skb->data + offset);
490 } else if (skb->protocol == htons(ETH_P_IPV6)) {
491 family = AF_INET6;
492 type = icmp6_hdr(skb)->icmp6_type;
493 code = icmp6_hdr(skb)->icmp6_code;
494 icmph = (struct icmphdr *) (skb->data + offset);
495 } else {
496 BUG();
497 }
498
499 /* We assume the packet has already been checked by icmp_unreach */
500
501 if (!ping_supported(family, icmph->type, icmph->code))
502 return;
503
504 pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
505 skb->protocol, type, code, ntohs(icmph->un.echo.id),
506 ntohs(icmph->un.echo.sequence));
507
508 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
509 if (!sk) {
510 pr_debug("no socket, dropping\n");
511 return; /* No socket for error */
512 }
513 pr_debug("err on socket %p\n", sk);
514
515 err = 0;
516 harderr = 0;
517 inet_sock = inet_sk(sk);
518
519 if (skb->protocol == htons(ETH_P_IP)) {
520 switch (type) {
521 default:
522 case ICMP_TIME_EXCEEDED:
523 err = EHOSTUNREACH;
524 break;
525 case ICMP_SOURCE_QUENCH:
526 /* This is not a real error but ping wants to see it.
527 * Report it with some fake errno.
528 */
529 err = EREMOTEIO;
530 break;
531 case ICMP_PARAMETERPROB:
532 err = EPROTO;
533 harderr = 1;
534 break;
535 case ICMP_DEST_UNREACH:
536 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
537 ipv4_sk_update_pmtu(skb, sk, info);
538 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
539 err = EMSGSIZE;
540 harderr = 1;
541 break;
542 }
543 goto out;
544 }
545 err = EHOSTUNREACH;
546 if (code <= NR_ICMP_UNREACH) {
547 harderr = icmp_err_convert[code].fatal;
548 err = icmp_err_convert[code].errno;
549 }
550 break;
551 case ICMP_REDIRECT:
552 /* See ICMP_SOURCE_QUENCH */
553 ipv4_sk_redirect(skb, sk);
554 err = EREMOTEIO;
555 break;
556 }
557 #if IS_ENABLED(CONFIG_IPV6)
558 } else if (skb->protocol == htons(ETH_P_IPV6)) {
559 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
560 #endif
561 }
562
563 /*
564 * RFC1122: OK. Passes ICMP errors back to application, as per
565 * 4.1.3.3.
566 */
567 if ((family == AF_INET && !inet_sock->recverr) ||
568 (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
569 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
570 goto out;
571 } else {
572 if (family == AF_INET) {
573 ip_icmp_error(sk, skb, err, 0 /* no remote port */,
574 info, (u8 *)icmph);
575 #if IS_ENABLED(CONFIG_IPV6)
576 } else if (family == AF_INET6) {
577 pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
578 info, (u8 *)icmph);
579 #endif
580 }
581 }
582 sk->sk_err = err;
583 sk_error_report(sk);
584 out:
585 sock_put(sk);
586 }
587 EXPORT_SYMBOL_GPL(ping_err);
588
589 /*
590 * Copy and checksum an ICMP Echo packet from user space into a buffer
591 * starting from the payload.
592 */
593
594 int ping_getfrag(void *from, char *to,
595 int offset, int fraglen, int odd, struct sk_buff *skb)
596 {
597 struct pingfakehdr *pfh = (struct pingfakehdr *)from;
598
599 if (offset == 0) {
600 fraglen -= sizeof(struct icmphdr);
601 if (fraglen < 0)
602 BUG();
603 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
604 fraglen, &pfh->wcheck,
605 &pfh->msg->msg_iter))
606 return -EFAULT;
607 } else if (offset < sizeof(struct icmphdr)) {
608 BUG();
609 } else {
610 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
611 &pfh->msg->msg_iter))
612 return -EFAULT;
613 }
614
615 #if IS_ENABLED(CONFIG_IPV6)
616 /* For IPv6, checksum each skb as we go along, as expected by
617 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
618 * wcheck, it will be finalized in ping_v4_push_pending_frames.
619 */
620 if (pfh->family == AF_INET6) {
621 skb->csum = pfh->wcheck;
622 skb->ip_summed = CHECKSUM_NONE;
623 pfh->wcheck = 0;
624 }
625 #endif
626
627 return 0;
628 }
629 EXPORT_SYMBOL_GPL(ping_getfrag);
630
631 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
632 struct flowi4 *fl4)
633 {
634 struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
635
636 if (!skb)
637 return 0;
638 pfh->wcheck = csum_partial((char *)&pfh->icmph,
639 sizeof(struct icmphdr), pfh->wcheck);
640 pfh->icmph.checksum = csum_fold(pfh->wcheck);
641 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
642 skb->ip_summed = CHECKSUM_NONE;
643 return ip_push_pending_frames(sk, fl4);
644 }
645
646 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
647 void *user_icmph, size_t icmph_len)
648 {
649 u8 type, code;
650
651 if (len > 0xFFFF)
652 return -EMSGSIZE;
653
654 /* Must have at least a full ICMP header. */
655 if (len < icmph_len)
656 return -EINVAL;
657
658 /*
659 * Check the flags.
660 */
661
662 /* Mirror BSD error message compatibility */
663 if (msg->msg_flags & MSG_OOB)
664 return -EOPNOTSUPP;
665
666 /*
667 * Fetch the ICMP header provided by the userland.
668 * iovec is modified! The ICMP header is consumed.
669 */
670 if (memcpy_from_msg(user_icmph, msg, icmph_len))
671 return -EFAULT;
672
673 if (family == AF_INET) {
674 type = ((struct icmphdr *) user_icmph)->type;
675 code = ((struct icmphdr *) user_icmph)->code;
676 #if IS_ENABLED(CONFIG_IPV6)
677 } else if (family == AF_INET6) {
678 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
679 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
680 #endif
681 } else {
682 BUG();
683 }
684
685 if (!ping_supported(family, type, code))
686 return -EINVAL;
687
688 return 0;
689 }
690 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
691
692 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
693 {
694 struct net *net = sock_net(sk);
695 struct flowi4 fl4;
696 struct inet_sock *inet = inet_sk(sk);
697 struct ipcm_cookie ipc;
698 struct icmphdr user_icmph;
699 struct pingfakehdr pfh;
700 struct rtable *rt = NULL;
701 struct ip_options_data opt_copy;
702 int free = 0;
703 __be32 saddr, daddr, faddr;
704 u8 tos;
705 int err;
706
707 pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
708
709 err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
710 sizeof(user_icmph));
711 if (err)
712 return err;
713
714 /*
715 * Get and verify the address.
716 */
717
718 if (msg->msg_name) {
719 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
720 if (msg->msg_namelen < sizeof(*usin))
721 return -EINVAL;
722 if (usin->sin_family != AF_INET)
723 return -EAFNOSUPPORT;
724 daddr = usin->sin_addr.s_addr;
725 /* no remote port */
726 } else {
727 if (sk->sk_state != TCP_ESTABLISHED)
728 return -EDESTADDRREQ;
729 daddr = inet->inet_daddr;
730 /* no remote port */
731 }
732
733 ipcm_init_sk(&ipc, inet);
734
735 if (msg->msg_controllen) {
736 err = ip_cmsg_send(sk, msg, &ipc, false);
737 if (unlikely(err)) {
738 kfree(ipc.opt);
739 return err;
740 }
741 if (ipc.opt)
742 free = 1;
743 }
744 if (!ipc.opt) {
745 struct ip_options_rcu *inet_opt;
746
747 rcu_read_lock();
748 inet_opt = rcu_dereference(inet->inet_opt);
749 if (inet_opt) {
750 memcpy(&opt_copy, inet_opt,
751 sizeof(*inet_opt) + inet_opt->opt.optlen);
752 ipc.opt = &opt_copy.opt;
753 }
754 rcu_read_unlock();
755 }
756
757 saddr = ipc.addr;
758 ipc.addr = faddr = daddr;
759
760 if (ipc.opt && ipc.opt->opt.srr) {
761 if (!daddr) {
762 err = -EINVAL;
763 goto out_free;
764 }
765 faddr = ipc.opt->opt.faddr;
766 }
767 tos = get_rttos(&ipc, inet);
768 if (sock_flag(sk, SOCK_LOCALROUTE) ||
769 (msg->msg_flags & MSG_DONTROUTE) ||
770 (ipc.opt && ipc.opt->opt.is_strictroute)) {
771 tos |= RTO_ONLINK;
772 }
773
774 if (ipv4_is_multicast(daddr)) {
775 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
776 ipc.oif = inet->mc_index;
777 if (!saddr)
778 saddr = inet->mc_addr;
779 } else if (!ipc.oif)
780 ipc.oif = inet->uc_index;
781
782 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
783 RT_SCOPE_UNIVERSE, sk->sk_protocol,
784 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
785 sk->sk_uid);
786
787 fl4.fl4_icmp_type = user_icmph.type;
788 fl4.fl4_icmp_code = user_icmph.code;
789
790 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
791 rt = ip_route_output_flow(net, &fl4, sk);
792 if (IS_ERR(rt)) {
793 err = PTR_ERR(rt);
794 rt = NULL;
795 if (err == -ENETUNREACH)
796 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
797 goto out;
798 }
799
800 err = -EACCES;
801 if ((rt->rt_flags & RTCF_BROADCAST) &&
802 !sock_flag(sk, SOCK_BROADCAST))
803 goto out;
804
805 if (msg->msg_flags & MSG_CONFIRM)
806 goto do_confirm;
807 back_from_confirm:
808
809 if (!ipc.addr)
810 ipc.addr = fl4.daddr;
811
812 lock_sock(sk);
813
814 pfh.icmph.type = user_icmph.type; /* already checked */
815 pfh.icmph.code = user_icmph.code; /* ditto */
816 pfh.icmph.checksum = 0;
817 pfh.icmph.un.echo.id = inet->inet_sport;
818 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
819 pfh.msg = msg;
820 pfh.wcheck = 0;
821 pfh.family = AF_INET;
822
823 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
824 0, &ipc, &rt, msg->msg_flags);
825 if (err)
826 ip_flush_pending_frames(sk);
827 else
828 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
829 release_sock(sk);
830
831 out:
832 ip_rt_put(rt);
833 out_free:
834 if (free)
835 kfree(ipc.opt);
836 if (!err) {
837 icmp_out_count(sock_net(sk), user_icmph.type);
838 return len;
839 }
840 return err;
841
842 do_confirm:
843 if (msg->msg_flags & MSG_PROBE)
844 dst_confirm_neigh(&rt->dst, &fl4.daddr);
845 if (!(msg->msg_flags & MSG_PROBE) || len)
846 goto back_from_confirm;
847 err = 0;
848 goto out;
849 }
850
851 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
852 int flags, int *addr_len)
853 {
854 struct inet_sock *isk = inet_sk(sk);
855 int family = sk->sk_family;
856 struct sk_buff *skb;
857 int copied, err;
858
859 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
860
861 err = -EOPNOTSUPP;
862 if (flags & MSG_OOB)
863 goto out;
864
865 if (flags & MSG_ERRQUEUE)
866 return inet_recv_error(sk, msg, len, addr_len);
867
868 skb = skb_recv_datagram(sk, flags, noblock, &err);
869 if (!skb)
870 goto out;
871
872 copied = skb->len;
873 if (copied > len) {
874 msg->msg_flags |= MSG_TRUNC;
875 copied = len;
876 }
877
878 /* Don't bother checking the checksum */
879 err = skb_copy_datagram_msg(skb, 0, msg, copied);
880 if (err)
881 goto done;
882
883 sock_recv_timestamp(msg, sk, skb);
884
885 /* Copy the address and add cmsg data. */
886 if (family == AF_INET) {
887 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
888
889 if (sin) {
890 sin->sin_family = AF_INET;
891 sin->sin_port = 0 /* skb->h.uh->source */;
892 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
893 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
894 *addr_len = sizeof(*sin);
895 }
896
897 if (isk->cmsg_flags)
898 ip_cmsg_recv(msg, skb);
899
900 #if IS_ENABLED(CONFIG_IPV6)
901 } else if (family == AF_INET6) {
902 struct ipv6_pinfo *np = inet6_sk(sk);
903 struct ipv6hdr *ip6 = ipv6_hdr(skb);
904 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
905
906 if (sin6) {
907 sin6->sin6_family = AF_INET6;
908 sin6->sin6_port = 0;
909 sin6->sin6_addr = ip6->saddr;
910 sin6->sin6_flowinfo = 0;
911 if (np->sndflow)
912 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
913 sin6->sin6_scope_id =
914 ipv6_iface_scope_id(&sin6->sin6_addr,
915 inet6_iif(skb));
916 *addr_len = sizeof(*sin6);
917 }
918
919 if (inet6_sk(sk)->rxopt.all)
920 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
921 if (skb->protocol == htons(ETH_P_IPV6) &&
922 inet6_sk(sk)->rxopt.all)
923 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
924 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
925 ip_cmsg_recv(msg, skb);
926 #endif
927 } else {
928 BUG();
929 }
930
931 err = copied;
932
933 done:
934 skb_free_datagram(sk, skb);
935 out:
936 pr_debug("ping_recvmsg -> %d\n", err);
937 return err;
938 }
939 EXPORT_SYMBOL_GPL(ping_recvmsg);
940
941 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
942 {
943 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
944 inet_sk(sk), inet_sk(sk)->inet_num, skb);
945 if (sock_queue_rcv_skb(sk, skb) < 0) {
946 kfree_skb(skb);
947 pr_debug("ping_queue_rcv_skb -> failed\n");
948 return -1;
949 }
950 return 0;
951 }
952 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
953
954
955 /*
956 * All we need to do is get the socket.
957 */
958
959 bool ping_rcv(struct sk_buff *skb)
960 {
961 struct sock *sk;
962 struct net *net = dev_net(skb->dev);
963 struct icmphdr *icmph = icmp_hdr(skb);
964 bool rc = false;
965
966 /* We assume the packet has already been checked by icmp_rcv */
967
968 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
969 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
970
971 /* Push ICMP header back */
972 skb_push(skb, skb->data - (u8 *)icmph);
973
974 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
975 if (sk) {
976 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
977
978 pr_debug("rcv on socket %p\n", sk);
979 if (skb2 && !ping_queue_rcv_skb(sk, skb2))
980 rc = true;
981 sock_put(sk);
982 }
983
984 if (!rc)
985 pr_debug("no socket, dropping\n");
986
987 return rc;
988 }
989 EXPORT_SYMBOL_GPL(ping_rcv);
990
991 struct proto ping_prot = {
992 .name = "PING",
993 .owner = THIS_MODULE,
994 .init = ping_init_sock,
995 .close = ping_close,
996 .connect = ip4_datagram_connect,
997 .disconnect = __udp_disconnect,
998 .setsockopt = ip_setsockopt,
999 .getsockopt = ip_getsockopt,
1000 .sendmsg = ping_v4_sendmsg,
1001 .recvmsg = ping_recvmsg,
1002 .bind = ping_bind,
1003 .backlog_rcv = ping_queue_rcv_skb,
1004 .release_cb = ip4_datagram_release_cb,
1005 .hash = ping_hash,
1006 .unhash = ping_unhash,
1007 .get_port = ping_get_port,
1008 .obj_size = sizeof(struct inet_sock),
1009 };
1010 EXPORT_SYMBOL(ping_prot);
1011
1012 #ifdef CONFIG_PROC_FS
1013
1014 static struct sock *ping_get_first(struct seq_file *seq, int start)
1015 {
1016 struct sock *sk;
1017 struct ping_iter_state *state = seq->private;
1018 struct net *net = seq_file_net(seq);
1019
1020 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1021 ++state->bucket) {
1022 struct hlist_nulls_node *node;
1023 struct hlist_nulls_head *hslot;
1024
1025 hslot = &ping_table.hash[state->bucket];
1026
1027 if (hlist_nulls_empty(hslot))
1028 continue;
1029
1030 sk_nulls_for_each(sk, node, hslot) {
1031 if (net_eq(sock_net(sk), net) &&
1032 sk->sk_family == state->family)
1033 goto found;
1034 }
1035 }
1036 sk = NULL;
1037 found:
1038 return sk;
1039 }
1040
1041 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1042 {
1043 struct ping_iter_state *state = seq->private;
1044 struct net *net = seq_file_net(seq);
1045
1046 do {
1047 sk = sk_nulls_next(sk);
1048 } while (sk && (!net_eq(sock_net(sk), net)));
1049
1050 if (!sk)
1051 return ping_get_first(seq, state->bucket + 1);
1052 return sk;
1053 }
1054
1055 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1056 {
1057 struct sock *sk = ping_get_first(seq, 0);
1058
1059 if (sk)
1060 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1061 --pos;
1062 return pos ? NULL : sk;
1063 }
1064
1065 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1066 __acquires(ping_table.lock)
1067 {
1068 struct ping_iter_state *state = seq->private;
1069 state->bucket = 0;
1070 state->family = family;
1071
1072 read_lock_bh(&ping_table.lock);
1073
1074 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1075 }
1076 EXPORT_SYMBOL_GPL(ping_seq_start);
1077
1078 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1079 {
1080 return ping_seq_start(seq, pos, AF_INET);
1081 }
1082
1083 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1084 {
1085 struct sock *sk;
1086
1087 if (v == SEQ_START_TOKEN)
1088 sk = ping_get_idx(seq, 0);
1089 else
1090 sk = ping_get_next(seq, v);
1091
1092 ++*pos;
1093 return sk;
1094 }
1095 EXPORT_SYMBOL_GPL(ping_seq_next);
1096
1097 void ping_seq_stop(struct seq_file *seq, void *v)
1098 __releases(ping_table.lock)
1099 {
1100 read_unlock_bh(&ping_table.lock);
1101 }
1102 EXPORT_SYMBOL_GPL(ping_seq_stop);
1103
1104 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1105 int bucket)
1106 {
1107 struct inet_sock *inet = inet_sk(sp);
1108 __be32 dest = inet->inet_daddr;
1109 __be32 src = inet->inet_rcv_saddr;
1110 __u16 destp = ntohs(inet->inet_dport);
1111 __u16 srcp = ntohs(inet->inet_sport);
1112
1113 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1114 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1115 bucket, src, srcp, dest, destp, sp->sk_state,
1116 sk_wmem_alloc_get(sp),
1117 sk_rmem_alloc_get(sp),
1118 0, 0L, 0,
1119 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1120 0, sock_i_ino(sp),
1121 refcount_read(&sp->sk_refcnt), sp,
1122 atomic_read(&sp->sk_drops));
1123 }
1124
1125 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1126 {
1127 seq_setwidth(seq, 127);
1128 if (v == SEQ_START_TOKEN)
1129 seq_puts(seq, " sl local_address rem_address st tx_queue "
1130 "rx_queue tr tm->when retrnsmt uid timeout "
1131 "inode ref pointer drops");
1132 else {
1133 struct ping_iter_state *state = seq->private;
1134
1135 ping_v4_format_sock(v, seq, state->bucket);
1136 }
1137 seq_pad(seq, '\n');
1138 return 0;
1139 }
1140
1141 static const struct seq_operations ping_v4_seq_ops = {
1142 .start = ping_v4_seq_start,
1143 .show = ping_v4_seq_show,
1144 .next = ping_seq_next,
1145 .stop = ping_seq_stop,
1146 };
1147
1148 static int __net_init ping_v4_proc_init_net(struct net *net)
1149 {
1150 if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1151 sizeof(struct ping_iter_state)))
1152 return -ENOMEM;
1153 return 0;
1154 }
1155
1156 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1157 {
1158 remove_proc_entry("icmp", net->proc_net);
1159 }
1160
1161 static struct pernet_operations ping_v4_net_ops = {
1162 .init = ping_v4_proc_init_net,
1163 .exit = ping_v4_proc_exit_net,
1164 };
1165
1166 int __init ping_proc_init(void)
1167 {
1168 return register_pernet_subsys(&ping_v4_net_ops);
1169 }
1170
1171 void ping_proc_exit(void)
1172 {
1173 unregister_pernet_subsys(&ping_v4_net_ops);
1174 }
1175
1176 #endif
1177
1178 void __init ping_init(void)
1179 {
1180 int i;
1181
1182 for (i = 0; i < PING_HTABLE_SIZE; i++)
1183 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1184 rwlock_init(&ping_table.lock);
1185 }
Ubuntu Jammy Linux kernel mirror