Merge remote-tracking branch 'lsk/v3.10/topic/aosp' into linux-linaro-lsk-android
[firefly-linux-kernel-4.4.55.git] / net / ipv4 / ping.c
1 /*
2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
3  *              operating system.  INET is implemented using the  BSD Socket
4  *              interface as the means of communication with the user level.
5  *
6  *              "Ping" sockets
7  *
8  *              This program is free software; you can redistribute it and/or
9  *              modify it under the terms of the GNU General Public License
10  *              as published by the Free Software Foundation; either version
11  *              2 of the License, or (at your option) any later version.
12  *
13  * Based on ipv4/udp.c code.
14  *
15  * Authors:     Vasiliy Kulikov / Openwall (for Linux 2.6),
16  *              Pavel Kankovsky (for Linux 2.4.32)
17  *
18  * Pavel gave all rights to bugs to Vasiliy,
19  * none of the bugs are Pavel's now.
20  *
21  */
22
23 #include <linux/uaccess.h>
24 #include <linux/types.h>
25 #include <linux/fcntl.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/in.h>
29 #include <linux/errno.h>
30 #include <linux/timer.h>
31 #include <linux/mm.h>
32 #include <linux/inet.h>
33 #include <linux/netdevice.h>
34 #include <net/snmp.h>
35 #include <net/ip.h>
36 #include <net/icmp.h>
37 #include <net/protocol.h>
38 #include <linux/skbuff.h>
39 #include <linux/proc_fs.h>
40 #include <linux/export.h>
41 #include <net/sock.h>
42 #include <net/ping.h>
43 #include <net/udp.h>
44 #include <net/route.h>
45 #include <net/inet_common.h>
46 #include <net/checksum.h>
47
48 #if IS_ENABLED(CONFIG_IPV6)
49 #include <linux/in6.h>
50 #include <linux/icmpv6.h>
51 #include <net/addrconf.h>
52 #include <net/ipv6.h>
53 #include <net/transp_v6.h>
54 #endif
55
56
57 struct ping_table ping_table;
58 struct pingv6_ops pingv6_ops;
59 EXPORT_SYMBOL_GPL(pingv6_ops);
60
61 static u16 ping_port_rover;
62
63 static inline int ping_hashfn(struct net *net, unsigned int num, unsigned int mask)
64 {
65         int res = (num + net_hash_mix(net)) & mask;
66
67         pr_debug("hash(%d) = %d\n", num, res);
68         return res;
69 }
70 EXPORT_SYMBOL_GPL(ping_hash);
71
72 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
73                                              struct net *net, unsigned int num)
74 {
75         return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
76 }
77
78 int ping_get_port(struct sock *sk, unsigned short ident)
79 {
80         struct hlist_nulls_node *node;
81         struct hlist_nulls_head *hlist;
82         struct inet_sock *isk, *isk2;
83         struct sock *sk2 = NULL;
84
85         isk = inet_sk(sk);
86         write_lock_bh(&ping_table.lock);
87         if (ident == 0) {
88                 u32 i;
89                 u16 result = ping_port_rover + 1;
90
91                 for (i = 0; i < (1L << 16); i++, result++) {
92                         if (!result)
93                                 result++; /* avoid zero */
94                         hlist = ping_hashslot(&ping_table, sock_net(sk),
95                                             result);
96                         ping_portaddr_for_each_entry(sk2, node, hlist) {
97                                 isk2 = inet_sk(sk2);
98
99                                 if (isk2->inet_num == result)
100                                         goto next_port;
101                         }
102
103                         /* found */
104                         ping_port_rover = ident = result;
105                         break;
106 next_port:
107                         ;
108                 }
109                 if (i >= (1L << 16))
110                         goto fail;
111         } else {
112                 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
113                 ping_portaddr_for_each_entry(sk2, node, hlist) {
114                         isk2 = inet_sk(sk2);
115
116                         /* BUG? Why is this reuse and not reuseaddr? ping.c
117                          * doesn't turn off SO_REUSEADDR, and it doesn't expect
118                          * that other ping processes can steal its packets.
119                          */
120                         if ((isk2->inet_num == ident) &&
121                             (sk2 != sk) &&
122                             (!sk2->sk_reuse || !sk->sk_reuse))
123                                 goto fail;
124                 }
125         }
126
127         pr_debug("found port/ident = %d\n", ident);
128         isk->inet_num = ident;
129         if (sk_unhashed(sk)) {
130                 pr_debug("was not hashed\n");
131                 sock_hold(sk);
132                 hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
133                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
134         }
135         write_unlock_bh(&ping_table.lock);
136         return 0;
137
138 fail:
139         write_unlock_bh(&ping_table.lock);
140         return 1;
141 }
142 EXPORT_SYMBOL_GPL(ping_get_port);
143
144 void ping_hash(struct sock *sk)
145 {
146         pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
147         BUG(); /* "Please do not press this button again." */
148 }
149
150 void ping_unhash(struct sock *sk)
151 {
152         struct inet_sock *isk = inet_sk(sk);
153         pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
154         if (sk_hashed(sk)) {
155                 write_lock_bh(&ping_table.lock);
156                 hlist_nulls_del(&sk->sk_nulls_node);
157                 sock_put(sk);
158                 isk->inet_num = 0;
159                 isk->inet_sport = 0;
160                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
161                 write_unlock_bh(&ping_table.lock);
162         }
163 }
164 EXPORT_SYMBOL_GPL(ping_unhash);
165
166 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
167 {
168         struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
169         struct sock *sk = NULL;
170         struct inet_sock *isk;
171         struct hlist_nulls_node *hnode;
172         int dif = skb->dev->ifindex;
173
174         if (skb->protocol == htons(ETH_P_IP)) {
175                 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
176                          (int)ident, &ip_hdr(skb)->daddr, dif);
177 #if IS_ENABLED(CONFIG_IPV6)
178         } else if (skb->protocol == htons(ETH_P_IPV6)) {
179                 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
180                          (int)ident, &ipv6_hdr(skb)->daddr, dif);
181 #endif
182         }
183
184         read_lock_bh(&ping_table.lock);
185
186         ping_portaddr_for_each_entry(sk, hnode, hslot) {
187                 isk = inet_sk(sk);
188
189                 pr_debug("iterate\n");
190                 if (isk->inet_num != ident)
191                         continue;
192
193                 if (skb->protocol == htons(ETH_P_IP) &&
194                     sk->sk_family == AF_INET) {
195                         pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
196                                  (int) isk->inet_num, &isk->inet_rcv_saddr,
197                                  sk->sk_bound_dev_if);
198
199                         if (isk->inet_rcv_saddr &&
200                             isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
201                                 continue;
202 #if IS_ENABLED(CONFIG_IPV6)
203                 } else if (skb->protocol == htons(ETH_P_IPV6) &&
204                            sk->sk_family == AF_INET6) {
205                         struct ipv6_pinfo *np = inet6_sk(sk);
206
207                         pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
208                                  (int) isk->inet_num,
209                                  &inet6_sk(sk)->rcv_saddr,
210                                  sk->sk_bound_dev_if);
211
212                         if (!ipv6_addr_any(&np->rcv_saddr) &&
213                             !ipv6_addr_equal(&np->rcv_saddr,
214                                              &ipv6_hdr(skb)->daddr))
215                                 continue;
216 #endif
217                 }
218
219                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
220                         continue;
221
222                 sock_hold(sk);
223                 goto exit;
224         }
225
226         sk = NULL;
227 exit:
228         read_unlock_bh(&ping_table.lock);
229
230         return sk;
231 }
232
233 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
234                                           kgid_t *high)
235 {
236         kgid_t *data = net->ipv4.sysctl_ping_group_range;
237         unsigned int seq;
238
239         do {
240                 seq = read_seqbegin(&sysctl_local_ports.lock);
241
242                 *low = data[0];
243                 *high = data[1];
244         } while (read_seqretry(&sysctl_local_ports.lock, seq));
245 }
246
247
248 int ping_init_sock(struct sock *sk)
249 {
250         struct net *net = sock_net(sk);
251         kgid_t group = current_egid();
252         struct group_info *group_info;
253         int i, j, count;
254         kgid_t low, high;
255         int ret = 0;
256
257         inet_get_ping_group_range_net(net, &low, &high);
258         if (gid_lte(low, group) && gid_lte(group, high))
259                 return 0;
260
261         group_info = get_current_groups();
262         count = group_info->ngroups;
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                                 goto out_release_group;
269                 }
270
271                 count -= cp_count;
272         }
273
274         ret = -EACCES;
275
276 out_release_group:
277         put_group_info(group_info);
278         return ret;
279 }
280 EXPORT_SYMBOL_GPL(ping_init_sock);
281
282 void ping_close(struct sock *sk, long timeout)
283 {
284         pr_debug("ping_close(sk=%p,sk->num=%u)\n",
285                  inet_sk(sk), inet_sk(sk)->inet_num);
286         pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
287
288         sk_common_release(sk);
289 }
290 EXPORT_SYMBOL_GPL(ping_close);
291
292 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
293 int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
294                          struct sockaddr *uaddr, int addr_len) {
295         struct net *net = sock_net(sk);
296         if (sk->sk_family == AF_INET) {
297                 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
298                 int chk_addr_ret;
299
300                 if (addr_len < sizeof(*addr))
301                         return -EINVAL;
302
303                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
304                          sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
305
306                 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
307
308                 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
309                         chk_addr_ret = RTN_LOCAL;
310
311                 if ((sysctl_ip_nonlocal_bind == 0 &&
312                     isk->freebind == 0 && isk->transparent == 0 &&
313                      chk_addr_ret != RTN_LOCAL) ||
314                     chk_addr_ret == RTN_MULTICAST ||
315                     chk_addr_ret == RTN_BROADCAST)
316                         return -EADDRNOTAVAIL;
317
318 #if IS_ENABLED(CONFIG_IPV6)
319         } else if (sk->sk_family == AF_INET6) {
320                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
321                 int addr_type, scoped, has_addr;
322                 struct net_device *dev = NULL;
323
324                 if (addr_len < sizeof(*addr))
325                         return -EINVAL;
326
327                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
328                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
329
330                 addr_type = ipv6_addr_type(&addr->sin6_addr);
331                 scoped = __ipv6_addr_needs_scope_id(addr_type);
332                 if ((addr_type != IPV6_ADDR_ANY &&
333                      !(addr_type & IPV6_ADDR_UNICAST)) ||
334                     (scoped && !addr->sin6_scope_id))
335                         return -EINVAL;
336
337                 rcu_read_lock();
338                 if (addr->sin6_scope_id) {
339                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
340                         if (!dev) {
341                                 rcu_read_unlock();
342                                 return -ENODEV;
343                         }
344                 }
345                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
346                                                     scoped);
347                 rcu_read_unlock();
348
349                 if (!(isk->freebind || isk->transparent || has_addr ||
350                       addr_type == IPV6_ADDR_ANY))
351                         return -EADDRNOTAVAIL;
352
353                 if (scoped)
354                         sk->sk_bound_dev_if = addr->sin6_scope_id;
355 #endif
356         } else {
357                 return -EAFNOSUPPORT;
358         }
359         return 0;
360 }
361
362 void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
363 {
364         if (saddr->sa_family == AF_INET) {
365                 struct inet_sock *isk = inet_sk(sk);
366                 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
367                 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
368 #if IS_ENABLED(CONFIG_IPV6)
369         } else if (saddr->sa_family == AF_INET6) {
370                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
371                 struct ipv6_pinfo *np = inet6_sk(sk);
372                 np->rcv_saddr = np->saddr = addr->sin6_addr;
373 #endif
374         }
375 }
376
377 void ping_clear_saddr(struct sock *sk, int dif)
378 {
379         sk->sk_bound_dev_if = dif;
380         if (sk->sk_family == AF_INET) {
381                 struct inet_sock *isk = inet_sk(sk);
382                 isk->inet_rcv_saddr = isk->inet_saddr = 0;
383 #if IS_ENABLED(CONFIG_IPV6)
384         } else if (sk->sk_family == AF_INET6) {
385                 struct ipv6_pinfo *np = inet6_sk(sk);
386                 memset(&np->rcv_saddr, 0, sizeof(np->rcv_saddr));
387                 memset(&np->saddr, 0, sizeof(np->saddr));
388 #endif
389         }
390 }
391 /*
392  * We need our own bind because there are no privileged id's == local ports.
393  * Moreover, we don't allow binding to multi- and broadcast addresses.
394  */
395
396 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
397 {
398         struct inet_sock *isk = inet_sk(sk);
399         unsigned short snum;
400         int err;
401         int dif = sk->sk_bound_dev_if;
402
403         err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
404         if (err)
405                 return err;
406
407         lock_sock(sk);
408
409         err = -EINVAL;
410         if (isk->inet_num != 0)
411                 goto out;
412
413         err = -EADDRINUSE;
414         ping_set_saddr(sk, uaddr);
415         snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
416         if (ping_get_port(sk, snum) != 0) {
417                 ping_clear_saddr(sk, dif);
418                 goto out;
419         }
420
421         pr_debug("after bind(): num = %d, dif = %d\n",
422                  (int)isk->inet_num,
423                  (int)sk->sk_bound_dev_if);
424
425         err = 0;
426         if ((sk->sk_family == AF_INET && isk->inet_rcv_saddr) ||
427             (sk->sk_family == AF_INET6 &&
428              !ipv6_addr_any(&inet6_sk(sk)->rcv_saddr)))
429                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
430
431         if (snum)
432                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
433         isk->inet_sport = htons(isk->inet_num);
434         isk->inet_daddr = 0;
435         isk->inet_dport = 0;
436
437 #if IS_ENABLED(CONFIG_IPV6)
438         if (sk->sk_family == AF_INET6)
439                 memset(&inet6_sk(sk)->daddr, 0, sizeof(inet6_sk(sk)->daddr));
440 #endif
441
442         sk_dst_reset(sk);
443 out:
444         release_sock(sk);
445         pr_debug("ping_v4_bind -> %d\n", err);
446         return err;
447 }
448 EXPORT_SYMBOL_GPL(ping_bind);
449
450 /*
451  * Is this a supported type of ICMP message?
452  */
453
454 static inline int ping_supported(int family, int type, int code)
455 {
456         return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
457                (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
458 }
459
460 /*
461  * This routine is called by the ICMP module when it gets some
462  * sort of error condition.
463  */
464
465 void ping_err(struct sk_buff *skb, int offset, u32 info)
466 {
467         int family;
468         struct icmphdr *icmph;
469         struct inet_sock *inet_sock;
470         int type;
471         int code;
472         struct net *net = dev_net(skb->dev);
473         struct sock *sk;
474         int harderr;
475         int err;
476
477         if (skb->protocol == htons(ETH_P_IP)) {
478                 family = AF_INET;
479                 type = icmp_hdr(skb)->type;
480                 code = icmp_hdr(skb)->code;
481                 icmph = (struct icmphdr *)(skb->data + offset);
482         } else if (skb->protocol == htons(ETH_P_IPV6)) {
483                 family = AF_INET6;
484                 type = icmp6_hdr(skb)->icmp6_type;
485                 code = icmp6_hdr(skb)->icmp6_code;
486                 icmph = (struct icmphdr *) (skb->data + offset);
487         } else {
488                 BUG();
489         }
490
491         /* We assume the packet has already been checked by icmp_unreach */
492
493         if (!ping_supported(family, icmph->type, icmph->code))
494                 return;
495
496         pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
497                  skb->protocol, type, code, ntohs(icmph->un.echo.id),
498                  ntohs(icmph->un.echo.sequence));
499
500         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
501         if (sk == NULL) {
502                 pr_debug("no socket, dropping\n");
503                 return; /* No socket for error */
504         }
505         pr_debug("err on socket %p\n", sk);
506
507         err = 0;
508         harderr = 0;
509         inet_sock = inet_sk(sk);
510
511         if (skb->protocol == htons(ETH_P_IP)) {
512                 switch (type) {
513                 default:
514                 case ICMP_TIME_EXCEEDED:
515                         err = EHOSTUNREACH;
516                         break;
517                 case ICMP_SOURCE_QUENCH:
518                         /* This is not a real error but ping wants to see it.
519                          * Report it with some fake errno.
520                          */
521                         err = EREMOTEIO;
522                         break;
523                 case ICMP_PARAMETERPROB:
524                         err = EPROTO;
525                         harderr = 1;
526                         break;
527                 case ICMP_DEST_UNREACH:
528                         if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
529                                 ipv4_sk_update_pmtu(skb, sk, info);
530                                 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
531                                         err = EMSGSIZE;
532                                         harderr = 1;
533                                         break;
534                                 }
535                                 goto out;
536                         }
537                         err = EHOSTUNREACH;
538                         if (code <= NR_ICMP_UNREACH) {
539                                 harderr = icmp_err_convert[code].fatal;
540                                 err = icmp_err_convert[code].errno;
541                         }
542                         break;
543                 case ICMP_REDIRECT:
544                         /* See ICMP_SOURCE_QUENCH */
545                         ipv4_sk_redirect(skb, sk);
546                         err = EREMOTEIO;
547                         break;
548                 }
549 #if IS_ENABLED(CONFIG_IPV6)
550         } else if (skb->protocol == htons(ETH_P_IPV6)) {
551                 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
552 #endif
553         }
554
555         /*
556          *      RFC1122: OK.  Passes ICMP errors back to application, as per
557          *      4.1.3.3.
558          */
559         if ((family == AF_INET && !inet_sock->recverr) ||
560             (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
561                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
562                         goto out;
563         } else {
564                 if (family == AF_INET) {
565                         ip_icmp_error(sk, skb, err, 0 /* no remote port */,
566                                       info, (u8 *)icmph);
567 #if IS_ENABLED(CONFIG_IPV6)
568                 } else if (family == AF_INET6) {
569                         pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
570                                                    info, (u8 *)icmph);
571 #endif
572                 }
573         }
574         sk->sk_err = err;
575         sk->sk_error_report(sk);
576 out:
577         sock_put(sk);
578 }
579 EXPORT_SYMBOL_GPL(ping_err);
580
581 /*
582  *      Copy and checksum an ICMP Echo packet from user space into a buffer
583  *      starting from the payload.
584  */
585
586 int ping_getfrag(void *from, char *to,
587                  int offset, int fraglen, int odd, struct sk_buff *skb)
588 {
589         struct pingfakehdr *pfh = (struct pingfakehdr *)from;
590
591         if (offset == 0) {
592                 if (fraglen < sizeof(struct icmphdr))
593                         BUG();
594                 if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
595                             pfh->iov, 0, fraglen - sizeof(struct icmphdr),
596                             &pfh->wcheck))
597                         return -EFAULT;
598         } else if (offset < sizeof(struct icmphdr)) {
599                         BUG();
600         } else {
601                 if (csum_partial_copy_fromiovecend
602                                 (to, pfh->iov, offset - sizeof(struct icmphdr),
603                                  fraglen, &pfh->wcheck))
604                         return -EFAULT;
605         }
606
607 #if IS_ENABLED(CONFIG_IPV6)
608         /* For IPv6, checksum each skb as we go along, as expected by
609          * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
610          * wcheck, it will be finalized in ping_v4_push_pending_frames.
611          */
612         if (pfh->family == AF_INET6) {
613                 skb->csum = pfh->wcheck;
614                 skb->ip_summed = CHECKSUM_NONE;
615                 pfh->wcheck = 0;
616         }
617 #endif
618
619         return 0;
620 }
621 EXPORT_SYMBOL_GPL(ping_getfrag);
622
623 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
624                                        struct flowi4 *fl4)
625 {
626         struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
627
628         pfh->wcheck = csum_partial((char *)&pfh->icmph,
629                 sizeof(struct icmphdr), pfh->wcheck);
630         pfh->icmph.checksum = csum_fold(pfh->wcheck);
631         memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
632         skb->ip_summed = CHECKSUM_NONE;
633         return ip_push_pending_frames(sk, fl4);
634 }
635
636 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
637                         void *user_icmph, size_t icmph_len) {
638         u8 type, code;
639
640         if (len > 0xFFFF)
641                 return -EMSGSIZE;
642
643         /*
644          *      Check the flags.
645          */
646
647         /* Mirror BSD error message compatibility */
648         if (msg->msg_flags & MSG_OOB)
649                 return -EOPNOTSUPP;
650
651         /*
652          *      Fetch the ICMP header provided by the userland.
653          *      iovec is modified! The ICMP header is consumed.
654          */
655         if (memcpy_fromiovec(user_icmph, msg->msg_iov, icmph_len))
656                 return -EFAULT;
657
658         if (family == AF_INET) {
659                 type = ((struct icmphdr *) user_icmph)->type;
660                 code = ((struct icmphdr *) user_icmph)->code;
661 #if IS_ENABLED(CONFIG_IPV6)
662         } else if (family == AF_INET6) {
663                 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
664                 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
665 #endif
666         } else {
667                 BUG();
668         }
669
670         if (!ping_supported(family, type, code))
671                 return -EINVAL;
672
673         return 0;
674 }
675 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
676
677 int ping_v4_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
678                     size_t len)
679 {
680         struct net *net = sock_net(sk);
681         struct flowi4 fl4;
682         struct inet_sock *inet = inet_sk(sk);
683         struct ipcm_cookie ipc;
684         struct icmphdr user_icmph;
685         struct pingfakehdr pfh;
686         struct rtable *rt = NULL;
687         struct ip_options_data opt_copy;
688         int free = 0;
689         __be32 saddr, daddr, faddr;
690         u8  tos;
691         int err;
692
693         pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
694
695         err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
696                                   sizeof(user_icmph));
697         if (err)
698                 return err;
699
700         /*
701          *      Get and verify the address.
702          */
703
704         if (msg->msg_name) {
705                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
706                 if (msg->msg_namelen < sizeof(*usin))
707                         return -EINVAL;
708                 if (usin->sin_family != AF_INET)
709                         return -EINVAL;
710                 daddr = usin->sin_addr.s_addr;
711                 /* no remote port */
712         } else {
713                 if (sk->sk_state != TCP_ESTABLISHED)
714                         return -EDESTADDRREQ;
715                 daddr = inet->inet_daddr;
716                 /* no remote port */
717         }
718
719         ipc.addr = inet->inet_saddr;
720         ipc.opt = NULL;
721         ipc.oif = sk->sk_bound_dev_if;
722         ipc.tx_flags = 0;
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 = RT_TOS(inet->tos);
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 sockaddr_in *sin;
838         struct sockaddr_in6 *sin6;
839         struct sk_buff *skb;
840         int copied, err;
841
842         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
843
844         err = -EOPNOTSUPP;
845         if (flags & MSG_OOB)
846                 goto out;
847
848         if (addr_len) {
849                 if (family == AF_INET)
850                         *addr_len = sizeof(*sin);
851                 else if (family == AF_INET6 && addr_len)
852                         *addr_len = sizeof(*sin6);
853         }
854
855         if (flags & MSG_ERRQUEUE) {
856                 if (family == AF_INET) {
857                         return ip_recv_error(sk, msg, len, addr_len);
858 #if IS_ENABLED(CONFIG_IPV6)
859                 } else if (family == AF_INET6) {
860                         return pingv6_ops.ipv6_recv_error(sk, msg, len);
861 #endif
862                 }
863         }
864
865         skb = skb_recv_datagram(sk, flags, noblock, &err);
866         if (!skb)
867                 goto out;
868
869         copied = skb->len;
870         if (copied > len) {
871                 msg->msg_flags |= MSG_TRUNC;
872                 copied = len;
873         }
874
875         /* Don't bother checking the checksum */
876         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
877         if (err)
878                 goto done;
879
880         sock_recv_timestamp(msg, sk, skb);
881
882         /* Copy the address and add cmsg data. */
883         if (family == AF_INET) {
884                 sin = (struct sockaddr_in *) msg->msg_name;
885                 if (sin) {
886                         sin->sin_family = AF_INET;
887                         sin->sin_port = 0 /* skb->h.uh->source */;
888                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
889                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
890                 }
891
892                 if (isk->cmsg_flags)
893                         ip_cmsg_recv(msg, skb);
894
895 #if IS_ENABLED(CONFIG_IPV6)
896         } else if (family == AF_INET6) {
897                 struct ipv6_pinfo *np = inet6_sk(sk);
898                 struct ipv6hdr *ip6 = ipv6_hdr(skb);
899                 sin6 = (struct sockaddr_in6 *) msg->msg_name;
900
901                 if (sin6) {
902                         sin6->sin6_family = AF_INET6;
903                         sin6->sin6_port = 0;
904                         sin6->sin6_addr = ip6->saddr;
905                         sin6->sin6_flowinfo = 0;
906                         if (np->sndflow)
907                                 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
908                         sin6->sin6_scope_id =
909                                 ipv6_iface_scope_id(&sin6->sin6_addr,
910                                                     IP6CB(skb)->iif);
911                 }
912
913                 if (inet6_sk(sk)->rxopt.all)
914                         pingv6_ops.ip6_datagram_recv_ctl(sk, msg, skb);
915 #endif
916         } else {
917                 BUG();
918         }
919
920         err = copied;
921
922 done:
923         skb_free_datagram(sk, skb);
924 out:
925         pr_debug("ping_recvmsg -> %d\n", err);
926         return err;
927 }
928 EXPORT_SYMBOL_GPL(ping_recvmsg);
929
930 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
931 {
932         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
933                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
934         if (sock_queue_rcv_skb(sk, skb) < 0) {
935                 kfree_skb(skb);
936                 pr_debug("ping_queue_rcv_skb -> failed\n");
937                 return -1;
938         }
939         return 0;
940 }
941 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
942
943
944 /*
945  *      All we need to do is get the socket.
946  */
947
948 void ping_rcv(struct sk_buff *skb)
949 {
950         struct sock *sk;
951         struct net *net = dev_net(skb->dev);
952         struct icmphdr *icmph = icmp_hdr(skb);
953
954         /* We assume the packet has already been checked by icmp_rcv */
955
956         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
957                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
958
959         /* Push ICMP header back */
960         skb_push(skb, skb->data - (u8 *)icmph);
961
962         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
963         if (sk != NULL) {
964                 pr_debug("rcv on socket %p\n", sk);
965                 ping_queue_rcv_skb(sk, skb_get(skb));
966                 sock_put(sk);
967                 return;
968         }
969         pr_debug("no socket, dropping\n");
970
971         /* We're called from icmp_rcv(). kfree_skb() is done there. */
972 }
973 EXPORT_SYMBOL_GPL(ping_rcv);
974
975 struct proto ping_prot = {
976         .name =         "PING",
977         .owner =        THIS_MODULE,
978         .init =         ping_init_sock,
979         .close =        ping_close,
980         .connect =      ip4_datagram_connect,
981         .disconnect =   udp_disconnect,
982         .setsockopt =   ip_setsockopt,
983         .getsockopt =   ip_getsockopt,
984         .sendmsg =      ping_v4_sendmsg,
985         .recvmsg =      ping_recvmsg,
986         .bind =         ping_bind,
987         .backlog_rcv =  ping_queue_rcv_skb,
988         .release_cb =   ip4_datagram_release_cb,
989         .hash =         ping_hash,
990         .unhash =       ping_unhash,
991         .get_port =     ping_get_port,
992         .obj_size =     sizeof(struct inet_sock),
993 };
994 EXPORT_SYMBOL(ping_prot);
995
996 #ifdef CONFIG_PROC_FS
997
998 static struct sock *ping_get_first(struct seq_file *seq, int start)
999 {
1000         struct sock *sk;
1001         struct ping_iter_state *state = seq->private;
1002         struct net *net = seq_file_net(seq);
1003
1004         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1005              ++state->bucket) {
1006                 struct hlist_nulls_node *node;
1007                 struct hlist_nulls_head *hslot;
1008
1009                 hslot = &ping_table.hash[state->bucket];
1010
1011                 if (hlist_nulls_empty(hslot))
1012                         continue;
1013
1014                 sk_nulls_for_each(sk, node, hslot) {
1015                         if (net_eq(sock_net(sk), net))
1016                                 goto found;
1017                 }
1018         }
1019         sk = NULL;
1020 found:
1021         return sk;
1022 }
1023
1024 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1025 {
1026         struct ping_iter_state *state = seq->private;
1027         struct net *net = seq_file_net(seq);
1028
1029         do {
1030                 sk = sk_nulls_next(sk);
1031         } while (sk && (!net_eq(sock_net(sk), net)));
1032
1033         if (!sk)
1034                 return ping_get_first(seq, state->bucket + 1);
1035         return sk;
1036 }
1037
1038 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1039 {
1040         struct sock *sk = ping_get_first(seq, 0);
1041
1042         if (sk)
1043                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1044                         --pos;
1045         return pos ? NULL : sk;
1046 }
1047
1048 static void *ping_seq_start(struct seq_file *seq, loff_t *pos)
1049 {
1050         struct ping_iter_state *state = seq->private;
1051         state->bucket = 0;
1052
1053         read_lock_bh(&ping_table.lock);
1054
1055         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1056 }
1057
1058 static void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1059 {
1060         struct sock *sk;
1061
1062         if (v == SEQ_START_TOKEN)
1063                 sk = ping_get_idx(seq, 0);
1064         else
1065                 sk = ping_get_next(seq, v);
1066
1067         ++*pos;
1068         return sk;
1069 }
1070
1071 static void ping_seq_stop(struct seq_file *seq, void *v)
1072 {
1073         read_unlock_bh(&ping_table.lock);
1074 }
1075
1076 static void ping_format_sock(struct sock *sp, struct seq_file *f,
1077                 int bucket, int *len)
1078 {
1079         struct inet_sock *inet = inet_sk(sp);
1080         __be32 dest = inet->inet_daddr;
1081         __be32 src = inet->inet_rcv_saddr;
1082         __u16 destp = ntohs(inet->inet_dport);
1083         __u16 srcp = ntohs(inet->inet_sport);
1084
1085         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1086                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
1087                 bucket, src, srcp, dest, destp, sp->sk_state,
1088                 sk_wmem_alloc_get(sp),
1089                 sk_rmem_alloc_get(sp),
1090                 0, 0L, 0,
1091                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1092                 0, sock_i_ino(sp),
1093                 atomic_read(&sp->sk_refcnt), sp,
1094                 atomic_read(&sp->sk_drops), len);
1095 }
1096
1097 static int ping_seq_show(struct seq_file *seq, void *v)
1098 {
1099         if (v == SEQ_START_TOKEN)
1100                 seq_printf(seq, "%-127s\n",
1101                            "  sl  local_address rem_address   st tx_queue "
1102                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1103                            "inode ref pointer drops");
1104         else {
1105                 struct ping_iter_state *state = seq->private;
1106                 int len;
1107
1108                 ping_format_sock(v, seq, state->bucket, &len);
1109                 seq_printf(seq, "%*s\n", 127 - len, "");
1110         }
1111         return 0;
1112 }
1113
1114 static const struct seq_operations ping_seq_ops = {
1115         .show           = ping_seq_show,
1116         .start          = ping_seq_start,
1117         .next           = ping_seq_next,
1118         .stop           = ping_seq_stop,
1119 };
1120
1121 static int ping_seq_open(struct inode *inode, struct file *file)
1122 {
1123         return seq_open_net(inode, file, &ping_seq_ops,
1124                            sizeof(struct ping_iter_state));
1125 }
1126
1127 static const struct file_operations ping_seq_fops = {
1128         .open           = ping_seq_open,
1129         .read           = seq_read,
1130         .llseek         = seq_lseek,
1131         .release        = seq_release_net,
1132 };
1133
1134 static int ping_proc_register(struct net *net)
1135 {
1136         struct proc_dir_entry *p;
1137         int rc = 0;
1138
1139         p = proc_create("icmp", S_IRUGO, net->proc_net, &ping_seq_fops);
1140         if (!p)
1141                 rc = -ENOMEM;
1142         return rc;
1143 }
1144
1145 static void ping_proc_unregister(struct net *net)
1146 {
1147         remove_proc_entry("icmp", net->proc_net);
1148 }
1149
1150
1151 static int __net_init ping_proc_init_net(struct net *net)
1152 {
1153         return ping_proc_register(net);
1154 }
1155
1156 static void __net_exit ping_proc_exit_net(struct net *net)
1157 {
1158         ping_proc_unregister(net);
1159 }
1160
1161 static struct pernet_operations ping_net_ops = {
1162         .init = ping_proc_init_net,
1163         .exit = ping_proc_exit_net,
1164 };
1165
1166 int __init ping_proc_init(void)
1167 {
1168         return register_pernet_subsys(&ping_net_ops);
1169 }
1170
1171 void ping_proc_exit(void)
1172 {
1173         unregister_pernet_subsys(&ping_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 }