rxrpc: Show some more information through /proc files
[linux/fpc-iii.git] / net / ipv4 / ping.c
blobb54c964ad9256ca2c5ac7eff239366f2869588a9
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.
6 * "Ping" sockets
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.
13 * Based on ipv4/udp.c code.
15 * Authors: Vasiliy Kulikov / Openwall (for Linux 2.6),
16 * Pavel Kankovsky (for Linux 2.4.32)
18 * Pavel gave all rights to bugs to Vasiliy,
19 * none of the bugs are Pavel's now.
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>
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
56 struct ping_table {
57 struct hlist_nulls_head hash[PING_HTABLE_SIZE];
58 rwlock_t lock;
61 static struct ping_table ping_table;
62 struct pingv6_ops pingv6_ops;
63 EXPORT_SYMBOL_GPL(pingv6_ops);
65 static u16 ping_port_rover;
67 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
69 u32 res = (num + net_hash_mix(net)) & mask;
71 pr_debug("hash(%u) = %u\n", num, res);
72 return res;
74 EXPORT_SYMBOL_GPL(ping_hash);
76 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
77 struct net *net, unsigned int num)
79 return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
82 int ping_get_port(struct sock *sk, unsigned short ident)
84 struct hlist_nulls_node *node;
85 struct hlist_nulls_head *hlist;
86 struct inet_sock *isk, *isk2;
87 struct sock *sk2 = NULL;
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;
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);
103 if (isk2->inet_num == result)
104 goto next_port;
107 /* found */
108 ping_port_rover = ident = result;
109 break;
110 next_port:
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);
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.
124 if ((isk2->inet_num == ident) &&
125 (sk2 != sk) &&
126 (!sk2->sk_reuse || !sk->sk_reuse))
127 goto fail;
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);
139 write_unlock_bh(&ping_table.lock);
140 return 0;
142 fail:
143 write_unlock_bh(&ping_table.lock);
144 return 1;
146 EXPORT_SYMBOL_GPL(ping_get_port);
148 int ping_hash(struct sock *sk)
150 pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
151 BUG(); /* "Please do not press this button again." */
153 return 0;
156 void ping_unhash(struct sock *sk)
158 struct inet_sock *isk = inet_sk(sk);
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);
170 write_unlock_bh(&ping_table.lock);
172 EXPORT_SYMBOL_GPL(ping_unhash);
174 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
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;
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
192 read_lock_bh(&ping_table.lock);
194 ping_portaddr_for_each_entry(sk, hnode, hslot) {
195 isk = inet_sk(sk);
197 pr_debug("iterate\n");
198 if (isk->inet_num != ident)
199 continue;
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);
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) {
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);
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;
228 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
229 continue;
231 sock_hold(sk);
232 goto exit;
235 sk = NULL;
236 exit:
237 read_unlock_bh(&ping_table.lock);
239 return sk;
242 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
243 kgid_t *high)
245 kgid_t *data = net->ipv4.ping_group_range.range;
246 unsigned int seq;
248 do {
249 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
251 *low = data[0];
252 *high = data[1];
253 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
257 int ping_init_sock(struct sock *sk)
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;
266 if (sk->sk_family == AF_INET6)
267 sk->sk_ipv6only = 1;
269 inet_get_ping_group_range_net(net, &low, &high);
270 if (gid_lte(low, group) && gid_lte(group, high))
271 return 0;
273 group_info = get_current_groups();
274 for (i = 0; i < group_info->ngroups; i++) {
275 kgid_t gid = group_info->gid[i];
277 if (gid_lte(low, gid) && gid_lte(gid, high))
278 goto out_release_group;
281 ret = -EACCES;
283 out_release_group:
284 put_group_info(group_info);
285 return ret;
287 EXPORT_SYMBOL_GPL(ping_init_sock);
289 void ping_close(struct sock *sk, long timeout)
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));
295 sk_common_release(sk);
297 EXPORT_SYMBOL_GPL(ping_close);
299 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
300 static 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;
307 if (addr_len < sizeof(*addr))
308 return -EINVAL;
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;
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));
318 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
320 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
321 chk_addr_ret = RTN_LOCAL;
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;
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;
336 if (addr_len < sizeof(*addr))
337 return -EINVAL;
339 if (addr->sin6_family != AF_INET6)
340 return -EAFNOSUPPORT;
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));
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;
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;
360 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
361 scoped);
362 rcu_read_unlock();
364 if (!(net->ipv6.sysctl.ip_nonlocal_bind ||
365 isk->freebind || isk->transparent || has_addr ||
366 addr_type == IPV6_ADDR_ANY))
367 return -EADDRNOTAVAIL;
369 if (scoped)
370 sk->sk_bound_dev_if = addr->sin6_scope_id;
371 #endif
372 } else {
373 return -EAFNOSUPPORT;
375 return 0;
378 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
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
393 static void ping_clear_saddr(struct sock *sk, int dif)
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
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.
412 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
414 struct inet_sock *isk = inet_sk(sk);
415 unsigned short snum;
416 int err;
417 int dif = sk->sk_bound_dev_if;
419 err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
420 if (err)
421 return err;
423 lock_sock(sk);
425 err = -EINVAL;
426 if (isk->inet_num != 0)
427 goto out;
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;
437 pr_debug("after bind(): num = %hu, dif = %d\n",
438 isk->inet_num,
439 sk->sk_bound_dev_if);
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
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;
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
460 sk_dst_reset(sk);
461 out:
462 release_sock(sk);
463 pr_debug("ping_v4_bind -> %d\n", err);
464 return err;
466 EXPORT_SYMBOL_GPL(ping_bind);
469 * Is this a supported type of ICMP message?
472 static inline int ping_supported(int family, int type, int code)
474 return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
475 (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
479 * This routine is called by the ICMP module when it gets some
480 * sort of error condition.
483 void ping_err(struct sk_buff *skb, int offset, u32 info)
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;
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();
509 /* We assume the packet has already been checked by icmp_unreach */
511 if (!ping_supported(family, icmph->type, icmph->code))
512 return;
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));
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 */
523 pr_debug("err on socket %p\n", sk);
525 err = 0;
526 harderr = 0;
527 inet_sock = inet_sk(sk);
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.
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;
553 goto out;
555 err = EHOSTUNREACH;
556 if (code <= NR_ICMP_UNREACH) {
557 harderr = icmp_err_convert[code].fatal;
558 err = icmp_err_convert[code].errno;
560 break;
561 case ICMP_REDIRECT:
562 /* See ICMP_SOURCE_QUENCH */
563 ipv4_sk_redirect(skb, sk);
564 err = EREMOTEIO;
565 break;
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
574 * RFC1122: OK. Passes ICMP errors back to application, as per
575 * 4.1.3.3.
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
592 sk->sk_err = err;
593 sk->sk_error_report(sk);
594 out:
595 sock_put(sk);
597 EXPORT_SYMBOL_GPL(ping_err);
600 * Copy and checksum an ICMP Echo packet from user space into a buffer
601 * starting from the payload.
604 int ping_getfrag(void *from, char *to,
605 int offset, int fraglen, int odd, struct sk_buff *skb)
607 struct pingfakehdr *pfh = (struct pingfakehdr *)from;
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;
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.
630 if (pfh->family == AF_INET6) {
631 skb->csum = pfh->wcheck;
632 skb->ip_summed = CHECKSUM_NONE;
633 pfh->wcheck = 0;
635 #endif
637 return 0;
639 EXPORT_SYMBOL_GPL(ping_getfrag);
641 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
642 struct flowi4 *fl4)
644 struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
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);
656 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
657 void *user_icmph, size_t icmph_len) {
658 u8 type, code;
660 if (len > 0xFFFF)
661 return -EMSGSIZE;
663 /* Must have at least a full ICMP header. */
664 if (len < icmph_len)
665 return -EINVAL;
668 * Check the flags.
671 /* Mirror BSD error message compatibility */
672 if (msg->msg_flags & MSG_OOB)
673 return -EOPNOTSUPP;
676 * Fetch the ICMP header provided by the userland.
677 * iovec is modified! The ICMP header is consumed.
679 if (memcpy_from_msg(user_icmph, msg, icmph_len))
680 return -EFAULT;
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();
694 if (!ping_supported(family, type, code))
695 return -EINVAL;
697 return 0;
699 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
701 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
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;
716 pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
718 err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
719 sizeof(user_icmph));
720 if (err)
721 return err;
724 * Get and verify the address.
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 */
742 ipcm_init_sk(&ipc, inet);
744 if (msg->msg_controllen) {
745 err = ip_cmsg_send(sk, msg, &ipc, false);
746 if (unlikely(err)) {
747 kfree(ipc.opt);
748 return err;
750 if (ipc.opt)
751 free = 1;
753 if (!ipc.opt) {
754 struct ip_options_rcu *inet_opt;
756 rcu_read_lock();
757 inet_opt = rcu_dereference(inet->inet_opt);
758 if (inet_opt) {
759 memcpy(&opt_copy, inet_opt,
760 sizeof(*inet_opt) + inet_opt->opt.optlen);
761 ipc.opt = &opt_copy.opt;
763 rcu_read_unlock();
766 saddr = ipc.addr;
767 ipc.addr = faddr = daddr;
769 if (ipc.opt && ipc.opt->opt.srr) {
770 if (!daddr) {
771 err = -EINVAL;
772 goto out_free;
774 faddr = ipc.opt->opt.faddr;
776 tos = get_rttos(&ipc, inet);
777 if (sock_flag(sk, SOCK_LOCALROUTE) ||
778 (msg->msg_flags & MSG_DONTROUTE) ||
779 (ipc.opt && ipc.opt->opt.is_strictroute)) {
780 tos |= RTO_ONLINK;
783 if (ipv4_is_multicast(daddr)) {
784 if (!ipc.oif)
785 ipc.oif = inet->mc_index;
786 if (!saddr)
787 saddr = inet->mc_addr;
788 } else if (!ipc.oif)
789 ipc.oif = inet->uc_index;
791 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
792 RT_SCOPE_UNIVERSE, sk->sk_protocol,
793 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
794 sk->sk_uid);
796 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
797 rt = ip_route_output_flow(net, &fl4, sk);
798 if (IS_ERR(rt)) {
799 err = PTR_ERR(rt);
800 rt = NULL;
801 if (err == -ENETUNREACH)
802 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
803 goto out;
806 err = -EACCES;
807 if ((rt->rt_flags & RTCF_BROADCAST) &&
808 !sock_flag(sk, SOCK_BROADCAST))
809 goto out;
811 if (msg->msg_flags & MSG_CONFIRM)
812 goto do_confirm;
813 back_from_confirm:
815 if (!ipc.addr)
816 ipc.addr = fl4.daddr;
818 lock_sock(sk);
820 pfh.icmph.type = user_icmph.type; /* already checked */
821 pfh.icmph.code = user_icmph.code; /* ditto */
822 pfh.icmph.checksum = 0;
823 pfh.icmph.un.echo.id = inet->inet_sport;
824 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
825 pfh.msg = msg;
826 pfh.wcheck = 0;
827 pfh.family = AF_INET;
829 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
830 0, &ipc, &rt, msg->msg_flags);
831 if (err)
832 ip_flush_pending_frames(sk);
833 else
834 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
835 release_sock(sk);
837 out:
838 ip_rt_put(rt);
839 out_free:
840 if (free)
841 kfree(ipc.opt);
842 if (!err) {
843 icmp_out_count(sock_net(sk), user_icmph.type);
844 return len;
846 return err;
848 do_confirm:
849 if (msg->msg_flags & MSG_PROBE)
850 dst_confirm_neigh(&rt->dst, &fl4.daddr);
851 if (!(msg->msg_flags & MSG_PROBE) || len)
852 goto back_from_confirm;
853 err = 0;
854 goto out;
857 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
858 int flags, int *addr_len)
860 struct inet_sock *isk = inet_sk(sk);
861 int family = sk->sk_family;
862 struct sk_buff *skb;
863 int copied, err;
865 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
867 err = -EOPNOTSUPP;
868 if (flags & MSG_OOB)
869 goto out;
871 if (flags & MSG_ERRQUEUE)
872 return inet_recv_error(sk, msg, len, addr_len);
874 skb = skb_recv_datagram(sk, flags, noblock, &err);
875 if (!skb)
876 goto out;
878 copied = skb->len;
879 if (copied > len) {
880 msg->msg_flags |= MSG_TRUNC;
881 copied = len;
884 /* Don't bother checking the checksum */
885 err = skb_copy_datagram_msg(skb, 0, msg, copied);
886 if (err)
887 goto done;
889 sock_recv_timestamp(msg, sk, skb);
891 /* Copy the address and add cmsg data. */
892 if (family == AF_INET) {
893 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
895 if (sin) {
896 sin->sin_family = AF_INET;
897 sin->sin_port = 0 /* skb->h.uh->source */;
898 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
899 memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
900 *addr_len = sizeof(*sin);
903 if (isk->cmsg_flags)
904 ip_cmsg_recv(msg, skb);
906 #if IS_ENABLED(CONFIG_IPV6)
907 } else if (family == AF_INET6) {
908 struct ipv6_pinfo *np = inet6_sk(sk);
909 struct ipv6hdr *ip6 = ipv6_hdr(skb);
910 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
912 if (sin6) {
913 sin6->sin6_family = AF_INET6;
914 sin6->sin6_port = 0;
915 sin6->sin6_addr = ip6->saddr;
916 sin6->sin6_flowinfo = 0;
917 if (np->sndflow)
918 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
919 sin6->sin6_scope_id =
920 ipv6_iface_scope_id(&sin6->sin6_addr,
921 inet6_iif(skb));
922 *addr_len = sizeof(*sin6);
925 if (inet6_sk(sk)->rxopt.all)
926 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
927 if (skb->protocol == htons(ETH_P_IPV6) &&
928 inet6_sk(sk)->rxopt.all)
929 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
930 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
931 ip_cmsg_recv(msg, skb);
932 #endif
933 } else {
934 BUG();
937 err = copied;
939 done:
940 skb_free_datagram(sk, skb);
941 out:
942 pr_debug("ping_recvmsg -> %d\n", err);
943 return err;
945 EXPORT_SYMBOL_GPL(ping_recvmsg);
947 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
949 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
950 inet_sk(sk), inet_sk(sk)->inet_num, skb);
951 if (sock_queue_rcv_skb(sk, skb) < 0) {
952 kfree_skb(skb);
953 pr_debug("ping_queue_rcv_skb -> failed\n");
954 return -1;
956 return 0;
958 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
962 * All we need to do is get the socket.
965 bool ping_rcv(struct sk_buff *skb)
967 struct sock *sk;
968 struct net *net = dev_net(skb->dev);
969 struct icmphdr *icmph = icmp_hdr(skb);
971 /* We assume the packet has already been checked by icmp_rcv */
973 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
974 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
976 /* Push ICMP header back */
977 skb_push(skb, skb->data - (u8 *)icmph);
979 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
980 if (sk) {
981 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
983 pr_debug("rcv on socket %p\n", sk);
984 if (skb2)
985 ping_queue_rcv_skb(sk, skb2);
986 sock_put(sk);
987 return true;
989 pr_debug("no socket, dropping\n");
991 return false;
993 EXPORT_SYMBOL_GPL(ping_rcv);
995 struct proto ping_prot = {
996 .name = "PING",
997 .owner = THIS_MODULE,
998 .init = ping_init_sock,
999 .close = ping_close,
1000 .connect = ip4_datagram_connect,
1001 .disconnect = __udp_disconnect,
1002 .setsockopt = ip_setsockopt,
1003 .getsockopt = ip_getsockopt,
1004 .sendmsg = ping_v4_sendmsg,
1005 .recvmsg = ping_recvmsg,
1006 .bind = ping_bind,
1007 .backlog_rcv = ping_queue_rcv_skb,
1008 .release_cb = ip4_datagram_release_cb,
1009 .hash = ping_hash,
1010 .unhash = ping_unhash,
1011 .get_port = ping_get_port,
1012 .obj_size = sizeof(struct inet_sock),
1014 EXPORT_SYMBOL(ping_prot);
1016 #ifdef CONFIG_PROC_FS
1018 static struct sock *ping_get_first(struct seq_file *seq, int start)
1020 struct sock *sk;
1021 struct ping_iter_state *state = seq->private;
1022 struct net *net = seq_file_net(seq);
1024 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1025 ++state->bucket) {
1026 struct hlist_nulls_node *node;
1027 struct hlist_nulls_head *hslot;
1029 hslot = &ping_table.hash[state->bucket];
1031 if (hlist_nulls_empty(hslot))
1032 continue;
1034 sk_nulls_for_each(sk, node, hslot) {
1035 if (net_eq(sock_net(sk), net) &&
1036 sk->sk_family == state->family)
1037 goto found;
1040 sk = NULL;
1041 found:
1042 return sk;
1045 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1047 struct ping_iter_state *state = seq->private;
1048 struct net *net = seq_file_net(seq);
1050 do {
1051 sk = sk_nulls_next(sk);
1052 } while (sk && (!net_eq(sock_net(sk), net)));
1054 if (!sk)
1055 return ping_get_first(seq, state->bucket + 1);
1056 return sk;
1059 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1061 struct sock *sk = ping_get_first(seq, 0);
1063 if (sk)
1064 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1065 --pos;
1066 return pos ? NULL : sk;
1069 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1070 __acquires(ping_table.lock)
1072 struct ping_iter_state *state = seq->private;
1073 state->bucket = 0;
1074 state->family = family;
1076 read_lock_bh(&ping_table.lock);
1078 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1080 EXPORT_SYMBOL_GPL(ping_seq_start);
1082 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1084 return ping_seq_start(seq, pos, AF_INET);
1087 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1089 struct sock *sk;
1091 if (v == SEQ_START_TOKEN)
1092 sk = ping_get_idx(seq, 0);
1093 else
1094 sk = ping_get_next(seq, v);
1096 ++*pos;
1097 return sk;
1099 EXPORT_SYMBOL_GPL(ping_seq_next);
1101 void ping_seq_stop(struct seq_file *seq, void *v)
1102 __releases(ping_table.lock)
1104 read_unlock_bh(&ping_table.lock);
1106 EXPORT_SYMBOL_GPL(ping_seq_stop);
1108 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1109 int bucket)
1111 struct inet_sock *inet = inet_sk(sp);
1112 __be32 dest = inet->inet_daddr;
1113 __be32 src = inet->inet_rcv_saddr;
1114 __u16 destp = ntohs(inet->inet_dport);
1115 __u16 srcp = ntohs(inet->inet_sport);
1117 seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1118 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1119 bucket, src, srcp, dest, destp, sp->sk_state,
1120 sk_wmem_alloc_get(sp),
1121 sk_rmem_alloc_get(sp),
1122 0, 0L, 0,
1123 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1124 0, sock_i_ino(sp),
1125 refcount_read(&sp->sk_refcnt), sp,
1126 atomic_read(&sp->sk_drops));
1129 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1131 seq_setwidth(seq, 127);
1132 if (v == SEQ_START_TOKEN)
1133 seq_puts(seq, " sl local_address rem_address st tx_queue "
1134 "rx_queue tr tm->when retrnsmt uid timeout "
1135 "inode ref pointer drops");
1136 else {
1137 struct ping_iter_state *state = seq->private;
1139 ping_v4_format_sock(v, seq, state->bucket);
1141 seq_pad(seq, '\n');
1142 return 0;
1145 static const struct seq_operations ping_v4_seq_ops = {
1146 .start = ping_v4_seq_start,
1147 .show = ping_v4_seq_show,
1148 .next = ping_seq_next,
1149 .stop = ping_seq_stop,
1152 static int __net_init ping_v4_proc_init_net(struct net *net)
1154 if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1155 sizeof(struct ping_iter_state)))
1156 return -ENOMEM;
1157 return 0;
1160 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1162 remove_proc_entry("icmp", net->proc_net);
1165 static struct pernet_operations ping_v4_net_ops = {
1166 .init = ping_v4_proc_init_net,
1167 .exit = ping_v4_proc_exit_net,
1170 int __init ping_proc_init(void)
1172 return register_pernet_subsys(&ping_v4_net_ops);
1175 void ping_proc_exit(void)
1177 unregister_pernet_subsys(&ping_v4_net_ops);
1180 #endif
1182 void __init ping_init(void)
1184 int i;
1186 for (i = 0; i < PING_HTABLE_SIZE; i++)
1187 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1188 rwlock_init(&ping_table.lock);