]> Pileus Git - ~andy/linux/blob - net/ipv4/ping.c
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec...
[~andy/linux] / net / ipv4 / ping.c
1 /*
2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
3  *              operating system.  INET is implemented using the  BSD Socket
4  *              interface as the means of communication with the user level.
5  *
6  *              "Ping" sockets
7  *
8  *              This program is free software; you can redistribute it and/or
9  *              modify it under the terms of the GNU General Public License
10  *              as published by the Free Software Foundation; either version
11  *              2 of the License, or (at your option) any later version.
12  *
13  * Based on ipv4/udp.c code.
14  *
15  * Authors:     Vasiliy Kulikov / Openwall (for Linux 2.6),
16  *              Pavel Kankovsky (for Linux 2.4.32)
17  *
18  * Pavel gave all rights to bugs to Vasiliy,
19  * none of the bugs are Pavel's now.
20  *
21  */
22
23 #include <linux/uaccess.h>
24 #include <linux/types.h>
25 #include <linux/fcntl.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/in.h>
29 #include <linux/errno.h>
30 #include <linux/timer.h>
31 #include <linux/mm.h>
32 #include <linux/inet.h>
33 #include <linux/netdevice.h>
34 #include <net/snmp.h>
35 #include <net/ip.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <linux/skbuff.h>
39 #include <linux/proc_fs.h>
40 #include <linux/export.h>
41 #include <net/sock.h>
42 #include <net/ping.h>
43 #include <net/udp.h>
44 #include <net/route.h>
45 #include <net/inet_common.h>
46 #include <net/checksum.h>
47
48 #if IS_ENABLED(CONFIG_IPV6)
49 #include <linux/in6.h>
50 #include <linux/icmpv6.h>
51 #include <net/addrconf.h>
52 #include <net/ipv6.h>
53 #include <net/transp_v6.h>
54 #endif
55
56 struct ping_table {
57         struct hlist_nulls_head hash[PING_HTABLE_SIZE];
58         rwlock_t                lock;
59 };
60
61 static struct ping_table ping_table;
62 struct pingv6_ops pingv6_ops;
63 EXPORT_SYMBOL_GPL(pingv6_ops);
64
65 static u16 ping_port_rover;
66
67 static inline int ping_hashfn(struct net *net, unsigned int num, unsigned int mask)
68 {
69         int res = (num + net_hash_mix(net)) & mask;
70
71         pr_debug("hash(%d) = %d\n", num, res);
72         return res;
73 }
74 EXPORT_SYMBOL_GPL(ping_hash);
75
76 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
77                                              struct net *net, unsigned int num)
78 {
79         return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
80 }
81
82 int ping_get_port(struct sock *sk, unsigned short ident)
83 {
84         struct hlist_nulls_node *node;
85         struct hlist_nulls_head *hlist;
86         struct inet_sock *isk, *isk2;
87         struct sock *sk2 = NULL;
88
89         isk = inet_sk(sk);
90         write_lock_bh(&ping_table.lock);
91         if (ident == 0) {
92                 u32 i;
93                 u16 result = ping_port_rover + 1;
94
95                 for (i = 0; i < (1L << 16); i++, result++) {
96                         if (!result)
97                                 result++; /* avoid zero */
98                         hlist = ping_hashslot(&ping_table, sock_net(sk),
99                                             result);
100                         ping_portaddr_for_each_entry(sk2, node, hlist) {
101                                 isk2 = inet_sk(sk2);
102
103                                 if (isk2->inet_num == result)
104                                         goto next_port;
105                         }
106
107                         /* found */
108                         ping_port_rover = ident = result;
109                         break;
110 next_port:
111                         ;
112                 }
113                 if (i >= (1L << 16))
114                         goto fail;
115         } else {
116                 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
117                 ping_portaddr_for_each_entry(sk2, node, hlist) {
118                         isk2 = inet_sk(sk2);
119
120                         /* BUG? Why is this reuse and not reuseaddr? ping.c
121                          * doesn't turn off SO_REUSEADDR, and it doesn't expect
122                          * that other ping processes can steal its packets.
123                          */
124                         if ((isk2->inet_num == ident) &&
125                             (sk2 != sk) &&
126                             (!sk2->sk_reuse || !sk->sk_reuse))
127                                 goto fail;
128                 }
129         }
130
131         pr_debug("found port/ident = %d\n", ident);
132         isk->inet_num = ident;
133         if (sk_unhashed(sk)) {
134                 pr_debug("was not hashed\n");
135                 sock_hold(sk);
136                 hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
137                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
138         }
139         write_unlock_bh(&ping_table.lock);
140         return 0;
141
142 fail:
143         write_unlock_bh(&ping_table.lock);
144         return 1;
145 }
146 EXPORT_SYMBOL_GPL(ping_get_port);
147
148 void ping_hash(struct sock *sk)
149 {
150         pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
151         BUG(); /* "Please do not press this button again." */
152 }
153
154 void ping_unhash(struct sock *sk)
155 {
156         struct inet_sock *isk = inet_sk(sk);
157         pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
158         if (sk_hashed(sk)) {
159                 write_lock_bh(&ping_table.lock);
160                 hlist_nulls_del(&sk->sk_nulls_node);
161                 sock_put(sk);
162                 isk->inet_num = 0;
163                 isk->inet_sport = 0;
164                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
165                 write_unlock_bh(&ping_table.lock);
166         }
167 }
168 EXPORT_SYMBOL_GPL(ping_unhash);
169
170 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
171 {
172         struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
173         struct sock *sk = NULL;
174         struct inet_sock *isk;
175         struct hlist_nulls_node *hnode;
176         int dif = skb->dev->ifindex;
177
178         if (skb->protocol == htons(ETH_P_IP)) {
179                 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
180                          (int)ident, &ip_hdr(skb)->daddr, dif);
181 #if IS_ENABLED(CONFIG_IPV6)
182         } else if (skb->protocol == htons(ETH_P_IPV6)) {
183                 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
184                          (int)ident, &ipv6_hdr(skb)->daddr, dif);
185 #endif
186         }
187
188         read_lock_bh(&ping_table.lock);
189
190         ping_portaddr_for_each_entry(sk, hnode, hslot) {
191                 isk = inet_sk(sk);
192
193                 pr_debug("iterate\n");
194                 if (isk->inet_num != ident)
195                         continue;
196
197                 if (skb->protocol == htons(ETH_P_IP) &&
198                     sk->sk_family == AF_INET) {
199                         pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
200                                  (int) isk->inet_num, &isk->inet_rcv_saddr,
201                                  sk->sk_bound_dev_if);
202
203                         if (isk->inet_rcv_saddr &&
204                             isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
205                                 continue;
206 #if IS_ENABLED(CONFIG_IPV6)
207                 } else if (skb->protocol == htons(ETH_P_IPV6) &&
208                            sk->sk_family == AF_INET6) {
209
210                         pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
211                                  (int) isk->inet_num,
212                                  &sk->sk_v6_rcv_saddr,
213                                  sk->sk_bound_dev_if);
214
215                         if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
216                             !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
217                                              &ipv6_hdr(skb)->daddr))
218                                 continue;
219 #endif
220                 }
221
222                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
223                         continue;
224
225                 sock_hold(sk);
226                 goto exit;
227         }
228
229         sk = NULL;
230 exit:
231         read_unlock_bh(&ping_table.lock);
232
233         return sk;
234 }
235
236 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
237                                           kgid_t *high)
238 {
239         kgid_t *data = net->ipv4.sysctl_ping_group_range;
240         unsigned int seq;
241
242         do {
243                 seq = read_seqbegin(&net->ipv4.sysctl_local_ports.lock);
244
245                 *low = data[0];
246                 *high = data[1];
247         } while (read_seqretry(&net->ipv4.sysctl_local_ports.lock, seq));
248 }
249
250
251 int ping_init_sock(struct sock *sk)
252 {
253         struct net *net = sock_net(sk);
254         kgid_t group = current_egid();
255         struct group_info *group_info = get_current_groups();
256         int i, j, count = group_info->ngroups;
257         kgid_t low, high;
258
259         inet_get_ping_group_range_net(net, &low, &high);
260         if (gid_lte(low, group) && gid_lte(group, high))
261                 return 0;
262
263         for (i = 0; i < group_info->nblocks; i++) {
264                 int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
265                 for (j = 0; j < cp_count; j++) {
266                         kgid_t gid = group_info->blocks[i][j];
267                         if (gid_lte(low, gid) && gid_lte(gid, high))
268                                 return 0;
269                 }
270
271                 count -= cp_count;
272         }
273
274         return -EACCES;
275 }
276 EXPORT_SYMBOL_GPL(ping_init_sock);
277
278 void ping_close(struct sock *sk, long timeout)
279 {
280         pr_debug("ping_close(sk=%p,sk->num=%u)\n",
281                  inet_sk(sk), inet_sk(sk)->inet_num);
282         pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
283
284         sk_common_release(sk);
285 }
286 EXPORT_SYMBOL_GPL(ping_close);
287
288 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
289 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
290                                 struct sockaddr *uaddr, int addr_len) {
291         struct net *net = sock_net(sk);
292         if (sk->sk_family == AF_INET) {
293                 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
294                 int chk_addr_ret;
295
296                 if (addr_len < sizeof(*addr))
297                         return -EINVAL;
298
299                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
300                          sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
301
302                 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
303
304                 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
305                         chk_addr_ret = RTN_LOCAL;
306
307                 if ((sysctl_ip_nonlocal_bind == 0 &&
308                     isk->freebind == 0 && isk->transparent == 0 &&
309                      chk_addr_ret != RTN_LOCAL) ||
310                     chk_addr_ret == RTN_MULTICAST ||
311                     chk_addr_ret == RTN_BROADCAST)
312                         return -EADDRNOTAVAIL;
313
314 #if IS_ENABLED(CONFIG_IPV6)
315         } else if (sk->sk_family == AF_INET6) {
316                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
317                 int addr_type, scoped, has_addr;
318                 struct net_device *dev = NULL;
319
320                 if (addr_len < sizeof(*addr))
321                         return -EINVAL;
322
323                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
324                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
325
326                 addr_type = ipv6_addr_type(&addr->sin6_addr);
327                 scoped = __ipv6_addr_needs_scope_id(addr_type);
328                 if ((addr_type != IPV6_ADDR_ANY &&
329                      !(addr_type & IPV6_ADDR_UNICAST)) ||
330                     (scoped && !addr->sin6_scope_id))
331                         return -EINVAL;
332
333                 rcu_read_lock();
334                 if (addr->sin6_scope_id) {
335                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
336                         if (!dev) {
337                                 rcu_read_unlock();
338                                 return -ENODEV;
339                         }
340                 }
341                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
342                                                     scoped);
343                 rcu_read_unlock();
344
345                 if (!(isk->freebind || isk->transparent || has_addr ||
346                       addr_type == IPV6_ADDR_ANY))
347                         return -EADDRNOTAVAIL;
348
349                 if (scoped)
350                         sk->sk_bound_dev_if = addr->sin6_scope_id;
351 #endif
352         } else {
353                 return -EAFNOSUPPORT;
354         }
355         return 0;
356 }
357
358 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
359 {
360         if (saddr->sa_family == AF_INET) {
361                 struct inet_sock *isk = inet_sk(sk);
362                 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
363                 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
364 #if IS_ENABLED(CONFIG_IPV6)
365         } else if (saddr->sa_family == AF_INET6) {
366                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
367                 struct ipv6_pinfo *np = inet6_sk(sk);
368                 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
369 #endif
370         }
371 }
372
373 static void ping_clear_saddr(struct sock *sk, int dif)
374 {
375         sk->sk_bound_dev_if = dif;
376         if (sk->sk_family == AF_INET) {
377                 struct inet_sock *isk = inet_sk(sk);
378                 isk->inet_rcv_saddr = isk->inet_saddr = 0;
379 #if IS_ENABLED(CONFIG_IPV6)
380         } else if (sk->sk_family == AF_INET6) {
381                 struct ipv6_pinfo *np = inet6_sk(sk);
382                 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
383                 memset(&np->saddr, 0, sizeof(np->saddr));
384 #endif
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
392 int 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         ping_set_saddr(sk, uaddr);
411         snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
412         if (ping_get_port(sk, snum) != 0) {
413                 ping_clear_saddr(sk, dif);
414                 goto out;
415         }
416
417         pr_debug("after bind(): num = %d, dif = %d\n",
418                  (int)isk->inet_num,
419                  (int)sk->sk_bound_dev_if);
420
421         err = 0;
422         if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
423                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
424 #if IS_ENABLED(CONFIG_IPV6)
425         if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
426                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
427 #endif
428
429         if (snum)
430                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
431         isk->inet_sport = htons(isk->inet_num);
432         isk->inet_daddr = 0;
433         isk->inet_dport = 0;
434
435 #if IS_ENABLED(CONFIG_IPV6)
436         if (sk->sk_family == AF_INET6)
437                 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
438 #endif
439
440         sk_dst_reset(sk);
441 out:
442         release_sock(sk);
443         pr_debug("ping_v4_bind -> %d\n", err);
444         return err;
445 }
446 EXPORT_SYMBOL_GPL(ping_bind);
447
448 /*
449  * Is this a supported type of ICMP message?
450  */
451
452 static inline int ping_supported(int family, int type, int code)
453 {
454         return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
455                (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
456 }
457
458 /*
459  * This routine is called by the ICMP module when it gets some
460  * sort of error condition.
461  */
462
463 void ping_err(struct sk_buff *skb, int offset, u32 info)
464 {
465         int family;
466         struct icmphdr *icmph;
467         struct inet_sock *inet_sock;
468         int type;
469         int code;
470         struct net *net = dev_net(skb->dev);
471         struct sock *sk;
472         int harderr;
473         int err;
474
475         if (skb->protocol == htons(ETH_P_IP)) {
476                 family = AF_INET;
477                 type = icmp_hdr(skb)->type;
478                 code = icmp_hdr(skb)->code;
479                 icmph = (struct icmphdr *)(skb->data + offset);
480         } else if (skb->protocol == htons(ETH_P_IPV6)) {
481                 family = AF_INET6;
482                 type = icmp6_hdr(skb)->icmp6_type;
483                 code = icmp6_hdr(skb)->icmp6_code;
484                 icmph = (struct icmphdr *) (skb->data + offset);
485         } else {
486                 BUG();
487         }
488
489         /* We assume the packet has already been checked by icmp_unreach */
490
491         if (!ping_supported(family, icmph->type, icmph->code))
492                 return;
493
494         pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
495                  skb->protocol, type, code, ntohs(icmph->un.echo.id),
496                  ntohs(icmph->un.echo.sequence));
497
498         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
499         if (sk == NULL) {
500                 pr_debug("no socket, dropping\n");
501                 return; /* No socket for error */
502         }
503         pr_debug("err on socket %p\n", sk);
504
505         err = 0;
506         harderr = 0;
507         inet_sock = inet_sk(sk);
508
509         if (skb->protocol == htons(ETH_P_IP)) {
510                 switch (type) {
511                 default:
512                 case ICMP_TIME_EXCEEDED:
513                         err = EHOSTUNREACH;
514                         break;
515                 case ICMP_SOURCE_QUENCH:
516                         /* This is not a real error but ping wants to see it.
517                          * Report it with some fake errno.
518                          */
519                         err = EREMOTEIO;
520                         break;
521                 case ICMP_PARAMETERPROB:
522                         err = EPROTO;
523                         harderr = 1;
524                         break;
525                 case ICMP_DEST_UNREACH:
526                         if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
527                                 ipv4_sk_update_pmtu(skb, sk, info);
528                                 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
529                                         err = EMSGSIZE;
530                                         harderr = 1;
531                                         break;
532                                 }
533                                 goto out;
534                         }
535                         err = EHOSTUNREACH;
536                         if (code <= NR_ICMP_UNREACH) {
537                                 harderr = icmp_err_convert[code].fatal;
538                                 err = icmp_err_convert[code].errno;
539                         }
540                         break;
541                 case ICMP_REDIRECT:
542                         /* See ICMP_SOURCE_QUENCH */
543                         ipv4_sk_redirect(skb, sk);
544                         err = EREMOTEIO;
545                         break;
546                 }
547 #if IS_ENABLED(CONFIG_IPV6)
548         } else if (skb->protocol == htons(ETH_P_IPV6)) {
549                 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
550 #endif
551         }
552
553         /*
554          *      RFC1122: OK.  Passes ICMP errors back to application, as per
555          *      4.1.3.3.
556          */
557         if ((family == AF_INET && !inet_sock->recverr) ||
558             (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
559                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
560                         goto out;
561         } else {
562                 if (family == AF_INET) {
563                         ip_icmp_error(sk, skb, err, 0 /* no remote port */,
564                                       info, (u8 *)icmph);
565 #if IS_ENABLED(CONFIG_IPV6)
566                 } else if (family == AF_INET6) {
567                         pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
568                                                    info, (u8 *)icmph);
569 #endif
570                 }
571         }
572         sk->sk_err = err;
573         sk->sk_error_report(sk);
574 out:
575         sock_put(sk);
576 }
577 EXPORT_SYMBOL_GPL(ping_err);
578
579 /*
580  *      Copy and checksum an ICMP Echo packet from user space into a buffer
581  *      starting from the payload.
582  */
583
584 int ping_getfrag(void *from, char *to,
585                  int offset, int fraglen, int odd, struct sk_buff *skb)
586 {
587         struct pingfakehdr *pfh = (struct pingfakehdr *)from;
588
589         if (offset == 0) {
590                 if (fraglen < sizeof(struct icmphdr))
591                         BUG();
592                 if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
593                             pfh->iov, 0, fraglen - sizeof(struct icmphdr),
594                             &pfh->wcheck))
595                         return -EFAULT;
596         } else if (offset < sizeof(struct icmphdr)) {
597                         BUG();
598         } else {
599                 if (csum_partial_copy_fromiovecend
600                                 (to, pfh->iov, offset - sizeof(struct icmphdr),
601                                  fraglen, &pfh->wcheck))
602                         return -EFAULT;
603         }
604
605 #if IS_ENABLED(CONFIG_IPV6)
606         /* For IPv6, checksum each skb as we go along, as expected by
607          * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
608          * wcheck, it will be finalized in ping_v4_push_pending_frames.
609          */
610         if (pfh->family == AF_INET6) {
611                 skb->csum = pfh->wcheck;
612                 skb->ip_summed = CHECKSUM_NONE;
613                 pfh->wcheck = 0;
614         }
615 #endif
616
617         return 0;
618 }
619 EXPORT_SYMBOL_GPL(ping_getfrag);
620
621 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
622                                        struct flowi4 *fl4)
623 {
624         struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
625
626         pfh->wcheck = csum_partial((char *)&pfh->icmph,
627                 sizeof(struct icmphdr), pfh->wcheck);
628         pfh->icmph.checksum = csum_fold(pfh->wcheck);
629         memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
630         skb->ip_summed = CHECKSUM_NONE;
631         return ip_push_pending_frames(sk, fl4);
632 }
633
634 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
635                         void *user_icmph, size_t icmph_len) {
636         u8 type, code;
637
638         if (len > 0xFFFF)
639                 return -EMSGSIZE;
640
641         /*
642          *      Check the flags.
643          */
644
645         /* Mirror BSD error message compatibility */
646         if (msg->msg_flags & MSG_OOB)
647                 return -EOPNOTSUPP;
648
649         /*
650          *      Fetch the ICMP header provided by the userland.
651          *      iovec is modified! The ICMP header is consumed.
652          */
653         if (memcpy_fromiovec(user_icmph, msg->msg_iov, icmph_len))
654                 return -EFAULT;
655
656         if (family == AF_INET) {
657                 type = ((struct icmphdr *) user_icmph)->type;
658                 code = ((struct icmphdr *) user_icmph)->code;
659 #if IS_ENABLED(CONFIG_IPV6)
660         } else if (family == AF_INET6) {
661                 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
662                 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
663 #endif
664         } else {
665                 BUG();
666         }
667
668         if (!ping_supported(family, type, code))
669                 return -EINVAL;
670
671         return 0;
672 }
673 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
674
675 static int ping_v4_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
676                            size_t len)
677 {
678         struct net *net = sock_net(sk);
679         struct flowi4 fl4;
680         struct inet_sock *inet = inet_sk(sk);
681         struct ipcm_cookie ipc;
682         struct icmphdr user_icmph;
683         struct pingfakehdr pfh;
684         struct rtable *rt = NULL;
685         struct ip_options_data opt_copy;
686         int free = 0;
687         __be32 saddr, daddr, faddr;
688         u8  tos;
689         int err;
690
691         pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
692
693         err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
694                                   sizeof(user_icmph));
695         if (err)
696                 return err;
697
698         /*
699          *      Get and verify the address.
700          */
701
702         if (msg->msg_name) {
703                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
704                 if (msg->msg_namelen < sizeof(*usin))
705                         return -EINVAL;
706                 if (usin->sin_family != AF_INET)
707                         return -EINVAL;
708                 daddr = usin->sin_addr.s_addr;
709                 /* no remote port */
710         } else {
711                 if (sk->sk_state != TCP_ESTABLISHED)
712                         return -EDESTADDRREQ;
713                 daddr = inet->inet_daddr;
714                 /* no remote port */
715         }
716
717         ipc.addr = inet->inet_saddr;
718         ipc.opt = NULL;
719         ipc.oif = sk->sk_bound_dev_if;
720         ipc.tx_flags = 0;
721         ipc.ttl = 0;
722         ipc.tos = -1;
723
724         sock_tx_timestamp(sk, &ipc.tx_flags);
725
726         if (msg->msg_controllen) {
727                 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
728                 if (err)
729                         return err;
730                 if (ipc.opt)
731                         free = 1;
732         }
733         if (!ipc.opt) {
734                 struct ip_options_rcu *inet_opt;
735
736                 rcu_read_lock();
737                 inet_opt = rcu_dereference(inet->inet_opt);
738                 if (inet_opt) {
739                         memcpy(&opt_copy, inet_opt,
740                                sizeof(*inet_opt) + inet_opt->opt.optlen);
741                         ipc.opt = &opt_copy.opt;
742                 }
743                 rcu_read_unlock();
744         }
745
746         saddr = ipc.addr;
747         ipc.addr = faddr = daddr;
748
749         if (ipc.opt && ipc.opt->opt.srr) {
750                 if (!daddr)
751                         return -EINVAL;
752                 faddr = ipc.opt->opt.faddr;
753         }
754         tos = get_rttos(&ipc, inet);
755         if (sock_flag(sk, SOCK_LOCALROUTE) ||
756             (msg->msg_flags & MSG_DONTROUTE) ||
757             (ipc.opt && ipc.opt->opt.is_strictroute)) {
758                 tos |= RTO_ONLINK;
759         }
760
761         if (ipv4_is_multicast(daddr)) {
762                 if (!ipc.oif)
763                         ipc.oif = inet->mc_index;
764                 if (!saddr)
765                         saddr = inet->mc_addr;
766         } else if (!ipc.oif)
767                 ipc.oif = inet->uc_index;
768
769         flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
770                            RT_SCOPE_UNIVERSE, sk->sk_protocol,
771                            inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
772
773         security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
774         rt = ip_route_output_flow(net, &fl4, sk);
775         if (IS_ERR(rt)) {
776                 err = PTR_ERR(rt);
777                 rt = NULL;
778                 if (err == -ENETUNREACH)
779                         IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
780                 goto out;
781         }
782
783         err = -EACCES;
784         if ((rt->rt_flags & RTCF_BROADCAST) &&
785             !sock_flag(sk, SOCK_BROADCAST))
786                 goto out;
787
788         if (msg->msg_flags & MSG_CONFIRM)
789                 goto do_confirm;
790 back_from_confirm:
791
792         if (!ipc.addr)
793                 ipc.addr = fl4.daddr;
794
795         lock_sock(sk);
796
797         pfh.icmph.type = user_icmph.type; /* already checked */
798         pfh.icmph.code = user_icmph.code; /* ditto */
799         pfh.icmph.checksum = 0;
800         pfh.icmph.un.echo.id = inet->inet_sport;
801         pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
802         pfh.iov = msg->msg_iov;
803         pfh.wcheck = 0;
804         pfh.family = AF_INET;
805
806         err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
807                         0, &ipc, &rt, msg->msg_flags);
808         if (err)
809                 ip_flush_pending_frames(sk);
810         else
811                 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
812         release_sock(sk);
813
814 out:
815         ip_rt_put(rt);
816         if (free)
817                 kfree(ipc.opt);
818         if (!err) {
819                 icmp_out_count(sock_net(sk), user_icmph.type);
820                 return len;
821         }
822         return err;
823
824 do_confirm:
825         dst_confirm(&rt->dst);
826         if (!(msg->msg_flags & MSG_PROBE) || len)
827                 goto back_from_confirm;
828         err = 0;
829         goto out;
830 }
831
832 int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
833                  size_t len, int noblock, int flags, int *addr_len)
834 {
835         struct inet_sock *isk = inet_sk(sk);
836         int family = sk->sk_family;
837         struct sk_buff *skb;
838         int copied, err;
839
840         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
841
842         err = -EOPNOTSUPP;
843         if (flags & MSG_OOB)
844                 goto out;
845
846         if (flags & MSG_ERRQUEUE) {
847                 if (family == AF_INET) {
848                         return ip_recv_error(sk, msg, len, addr_len);
849 #if IS_ENABLED(CONFIG_IPV6)
850                 } else if (family == AF_INET6) {
851                         return pingv6_ops.ipv6_recv_error(sk, msg, len,
852                                                           addr_len);
853 #endif
854                 }
855         }
856
857         skb = skb_recv_datagram(sk, flags, noblock, &err);
858         if (!skb)
859                 goto out;
860
861         copied = skb->len;
862         if (copied > len) {
863                 msg->msg_flags |= MSG_TRUNC;
864                 copied = len;
865         }
866
867         /* Don't bother checking the checksum */
868         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
869         if (err)
870                 goto done;
871
872         sock_recv_timestamp(msg, sk, skb);
873
874         /* Copy the address and add cmsg data. */
875         if (family == AF_INET) {
876                 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
877
878                 if (sin) {
879                         sin->sin_family = AF_INET;
880                         sin->sin_port = 0 /* skb->h.uh->source */;
881                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
882                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
883                         *addr_len = sizeof(*sin);
884                 }
885
886                 if (isk->cmsg_flags)
887                         ip_cmsg_recv(msg, skb);
888
889 #if IS_ENABLED(CONFIG_IPV6)
890         } else if (family == AF_INET6) {
891                 struct ipv6_pinfo *np = inet6_sk(sk);
892                 struct ipv6hdr *ip6 = ipv6_hdr(skb);
893                 struct sockaddr_in6 *sin6 =
894                         (struct sockaddr_in6 *)msg->msg_name;
895
896                 if (sin6) {
897                         sin6->sin6_family = AF_INET6;
898                         sin6->sin6_port = 0;
899                         sin6->sin6_addr = ip6->saddr;
900                         sin6->sin6_flowinfo = 0;
901                         if (np->sndflow)
902                                 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
903                         sin6->sin6_scope_id =
904                                 ipv6_iface_scope_id(&sin6->sin6_addr,
905                                                     IP6CB(skb)->iif);
906                         *addr_len = sizeof(*sin6);
907                 }
908
909                 if (inet6_sk(sk)->rxopt.all)
910                         pingv6_ops.ip6_datagram_recv_ctl(sk, msg, skb);
911 #endif
912         } else {
913                 BUG();
914         }
915
916         err = copied;
917
918 done:
919         skb_free_datagram(sk, skb);
920 out:
921         pr_debug("ping_recvmsg -> %d\n", err);
922         return err;
923 }
924 EXPORT_SYMBOL_GPL(ping_recvmsg);
925
926 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
927 {
928         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
929                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
930         if (sock_queue_rcv_skb(sk, skb) < 0) {
931                 kfree_skb(skb);
932                 pr_debug("ping_queue_rcv_skb -> failed\n");
933                 return -1;
934         }
935         return 0;
936 }
937 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
938
939
940 /*
941  *      All we need to do is get the socket.
942  */
943
944 void ping_rcv(struct sk_buff *skb)
945 {
946         struct sock *sk;
947         struct net *net = dev_net(skb->dev);
948         struct icmphdr *icmph = icmp_hdr(skb);
949
950         /* We assume the packet has already been checked by icmp_rcv */
951
952         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
953                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
954
955         /* Push ICMP header back */
956         skb_push(skb, skb->data - (u8 *)icmph);
957
958         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
959         if (sk != NULL) {
960                 pr_debug("rcv on socket %p\n", sk);
961                 ping_queue_rcv_skb(sk, skb_get(skb));
962                 sock_put(sk);
963                 return;
964         }
965         pr_debug("no socket, dropping\n");
966
967         /* We're called from icmp_rcv(). kfree_skb() is done there. */
968 }
969 EXPORT_SYMBOL_GPL(ping_rcv);
970
971 struct proto ping_prot = {
972         .name =         "PING",
973         .owner =        THIS_MODULE,
974         .init =         ping_init_sock,
975         .close =        ping_close,
976         .connect =      ip4_datagram_connect,
977         .disconnect =   udp_disconnect,
978         .setsockopt =   ip_setsockopt,
979         .getsockopt =   ip_getsockopt,
980         .sendmsg =      ping_v4_sendmsg,
981         .recvmsg =      ping_recvmsg,
982         .bind =         ping_bind,
983         .backlog_rcv =  ping_queue_rcv_skb,
984         .release_cb =   ip4_datagram_release_cb,
985         .hash =         ping_hash,
986         .unhash =       ping_unhash,
987         .get_port =     ping_get_port,
988         .obj_size =     sizeof(struct inet_sock),
989 };
990 EXPORT_SYMBOL(ping_prot);
991
992 #ifdef CONFIG_PROC_FS
993
994 static struct sock *ping_get_first(struct seq_file *seq, int start)
995 {
996         struct sock *sk;
997         struct ping_iter_state *state = seq->private;
998         struct net *net = seq_file_net(seq);
999
1000         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1001              ++state->bucket) {
1002                 struct hlist_nulls_node *node;
1003                 struct hlist_nulls_head *hslot;
1004
1005                 hslot = &ping_table.hash[state->bucket];
1006
1007                 if (hlist_nulls_empty(hslot))
1008                         continue;
1009
1010                 sk_nulls_for_each(sk, node, hslot) {
1011                         if (net_eq(sock_net(sk), net) &&
1012                             sk->sk_family == state->family)
1013                                 goto found;
1014                 }
1015         }
1016         sk = NULL;
1017 found:
1018         return sk;
1019 }
1020
1021 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1022 {
1023         struct ping_iter_state *state = seq->private;
1024         struct net *net = seq_file_net(seq);
1025
1026         do {
1027                 sk = sk_nulls_next(sk);
1028         } while (sk && (!net_eq(sock_net(sk), net)));
1029
1030         if (!sk)
1031                 return ping_get_first(seq, state->bucket + 1);
1032         return sk;
1033 }
1034
1035 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1036 {
1037         struct sock *sk = ping_get_first(seq, 0);
1038
1039         if (sk)
1040                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1041                         --pos;
1042         return pos ? NULL : sk;
1043 }
1044
1045 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1046 {
1047         struct ping_iter_state *state = seq->private;
1048         state->bucket = 0;
1049         state->family = family;
1050
1051         read_lock_bh(&ping_table.lock);
1052
1053         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1054 }
1055 EXPORT_SYMBOL_GPL(ping_seq_start);
1056
1057 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1058 {
1059         return ping_seq_start(seq, pos, AF_INET);
1060 }
1061
1062 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1063 {
1064         struct sock *sk;
1065
1066         if (v == SEQ_START_TOKEN)
1067                 sk = ping_get_idx(seq, 0);
1068         else
1069                 sk = ping_get_next(seq, v);
1070
1071         ++*pos;
1072         return sk;
1073 }
1074 EXPORT_SYMBOL_GPL(ping_seq_next);
1075
1076 void ping_seq_stop(struct seq_file *seq, void *v)
1077 {
1078         read_unlock_bh(&ping_table.lock);
1079 }
1080 EXPORT_SYMBOL_GPL(ping_seq_stop);
1081
1082 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1083                 int bucket)
1084 {
1085         struct inet_sock *inet = inet_sk(sp);
1086         __be32 dest = inet->inet_daddr;
1087         __be32 src = inet->inet_rcv_saddr;
1088         __u16 destp = ntohs(inet->inet_dport);
1089         __u16 srcp = ntohs(inet->inet_sport);
1090
1091         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1092                 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1093                 bucket, src, srcp, dest, destp, sp->sk_state,
1094                 sk_wmem_alloc_get(sp),
1095                 sk_rmem_alloc_get(sp),
1096                 0, 0L, 0,
1097                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1098                 0, sock_i_ino(sp),
1099                 atomic_read(&sp->sk_refcnt), sp,
1100                 atomic_read(&sp->sk_drops));
1101 }
1102
1103 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1104 {
1105         seq_setwidth(seq, 127);
1106         if (v == SEQ_START_TOKEN)
1107                 seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1108                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1109                            "inode ref pointer drops");
1110         else {
1111                 struct ping_iter_state *state = seq->private;
1112
1113                 ping_v4_format_sock(v, seq, state->bucket);
1114         }
1115         seq_pad(seq, '\n');
1116         return 0;
1117 }
1118
1119 static const struct seq_operations ping_v4_seq_ops = {
1120         .show           = ping_v4_seq_show,
1121         .start          = ping_v4_seq_start,
1122         .next           = ping_seq_next,
1123         .stop           = ping_seq_stop,
1124 };
1125
1126 static int ping_seq_open(struct inode *inode, struct file *file)
1127 {
1128         struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1129         return seq_open_net(inode, file, &afinfo->seq_ops,
1130                            sizeof(struct ping_iter_state));
1131 }
1132
1133 const struct file_operations ping_seq_fops = {
1134         .open           = ping_seq_open,
1135         .read           = seq_read,
1136         .llseek         = seq_lseek,
1137         .release        = seq_release_net,
1138 };
1139 EXPORT_SYMBOL_GPL(ping_seq_fops);
1140
1141 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1142         .name           = "icmp",
1143         .family         = AF_INET,
1144         .seq_fops       = &ping_seq_fops,
1145         .seq_ops        = {
1146                 .start          = ping_v4_seq_start,
1147                 .show           = ping_v4_seq_show,
1148                 .next           = ping_seq_next,
1149                 .stop           = ping_seq_stop,
1150         },
1151 };
1152
1153 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1154 {
1155         struct proc_dir_entry *p;
1156         p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1157                              afinfo->seq_fops, afinfo);
1158         if (!p)
1159                 return -ENOMEM;
1160         return 0;
1161 }
1162 EXPORT_SYMBOL_GPL(ping_proc_register);
1163
1164 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1165 {
1166         remove_proc_entry(afinfo->name, net->proc_net);
1167 }
1168 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1169
1170 static int __net_init ping_v4_proc_init_net(struct net *net)
1171 {
1172         return ping_proc_register(net, &ping_v4_seq_afinfo);
1173 }
1174
1175 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1176 {
1177         ping_proc_unregister(net, &ping_v4_seq_afinfo);
1178 }
1179
1180 static struct pernet_operations ping_v4_net_ops = {
1181         .init = ping_v4_proc_init_net,
1182         .exit = ping_v4_proc_exit_net,
1183 };
1184
1185 int __init ping_proc_init(void)
1186 {
1187         return register_pernet_subsys(&ping_v4_net_ops);
1188 }
1189
1190 void ping_proc_exit(void)
1191 {
1192         unregister_pernet_subsys(&ping_v4_net_ops);
1193 }
1194
1195 #endif
1196
1197 void __init ping_init(void)
1198 {
1199         int i;
1200
1201         for (i = 0; i < PING_HTABLE_SIZE; i++)
1202                 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1203         rwlock_init(&ping_table.lock);
1204 }