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 u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
  68 {
  69         u32 res = (num + net_hash_mix(net)) & mask;
  70
  71         pr_debug("hash(%u) = %u\n", num, res);
  72         return res;
  73 }
  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 int ping_hash(struct sock *sk)
 149 {
 150         pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
 151         BUG(); /* "Please do not press this button again." */
 152
 153         return 0;
 154 }
 155
 156 void ping_unhash(struct sock *sk)
 157 {
 158         struct inet_sock *isk = inet_sk(sk);
 159
 160         pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
 161         write_lock_bh(&ping_table.lock);
 162         if (sk_hashed(sk)) {
 163                 hlist_nulls_del(&sk->sk_nulls_node);
 164                 sk_nulls_node_init(&sk->sk_nulls_node);
 165                 sock_put(sk);
 166                 isk->inet_num = 0;
 167                 isk->inet_sport = 0;
 168                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 169         }
 170         write_unlock_bh(&ping_table.lock);
 171 }
 172 EXPORT_SYMBOL_GPL(ping_unhash);
 173
 174 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
 175 {
 176         struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
 177         struct sock *sk = NULL;
 178         struct inet_sock *isk;
 179         struct hlist_nulls_node *hnode;
 180         int dif, sdif;
 181
 182         if (skb->protocol == htons(ETH_P_IP)) {
 183                 dif = inet_iif(skb);
 184                 sdif = inet_sdif(skb);
 185                 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
 186                          (int)ident, &ip_hdr(skb)->daddr, dif);
 187 #if IS_ENABLED(CONFIG_IPV6)
 188         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 189                 dif = inet6_iif(skb);
 190                 sdif = inet6_sdif(skb);
 191                 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
 192                          (int)ident, &ipv6_hdr(skb)->daddr, dif);
 193 #endif
 194         } else {
 195                 return NULL;
 196         }
 197
 198         read_lock_bh(&ping_table.lock);
 199
 200         ping_portaddr_for_each_entry(sk, hnode, hslot) {
 201                 isk = inet_sk(sk);
 202
 203                 pr_debug("iterate\n");
 204                 if (isk->inet_num != ident)
 205                         continue;
 206
 207                 if (skb->protocol == htons(ETH_P_IP) &&
 208                     sk->sk_family == AF_INET) {
 209                         pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
 210                                  (int) isk->inet_num, &isk->inet_rcv_saddr,
 211                                  sk->sk_bound_dev_if);
 212
 213                         if (isk->inet_rcv_saddr &&
 214                             isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
 215                                 continue;
 216 #if IS_ENABLED(CONFIG_IPV6)
 217                 } else if (skb->protocol == htons(ETH_P_IPV6) &&
 218                            sk->sk_family == AF_INET6) {
 219
 220                         pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
 221                                  (int) isk->inet_num,
 222                                  &sk->sk_v6_rcv_saddr,
 223                                  sk->sk_bound_dev_if);
 224
 225                         if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
 226                             !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
 227                                              &ipv6_hdr(skb)->daddr))
 228                                 continue;
 229 #endif
 230                 } else {
 231                         continue;
 232                 }
 233
 234                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
 235                     sk->sk_bound_dev_if != sdif)
 236                         continue;
 237
 238                 sock_hold(sk);
 239                 goto exit;
 240         }
 241
 242         sk = NULL;
 243 exit:
 244         read_unlock_bh(&ping_table.lock);
 245
 246         return sk;
 247 }
 248
 249 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
 250                                           kgid_t *high)
 251 {
 252         kgid_t *data = net->ipv4.ping_group_range.range;
 253         unsigned int seq;
 254
 255         do {
 256                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 257
 258                 *low = data[0];
 259                 *high = data[1];
 260         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 261 }
 262
 263
 264 int ping_init_sock(struct sock *sk)
 265 {
 266         struct net *net = sock_net(sk);
 267         kgid_t group = current_egid();
 268         struct group_info *group_info;
 269         int i;
 270         kgid_t low, high;
 271         int ret = 0;
 272
 273         if (sk->sk_family == AF_INET6)
 274                 sk->sk_ipv6only = 1;
 275
 276         inet_get_ping_group_range_net(net, &low, &high);
 277         if (gid_lte(low, group) && gid_lte(group, high))
 278                 return 0;
 279
 280         group_info = get_current_groups();
 281         for (i = 0; i < group_info->ngroups; i++) {
 282                 kgid_t gid = group_info->gid[i];
 283
 284                 if (gid_lte(low, gid) && gid_lte(gid, high))
 285                         goto out_release_group;
 286         }
 287
 288         ret = -EACCES;
 289
 290 out_release_group:
 291         put_group_info(group_info);
 292         return ret;
 293 }
 294 EXPORT_SYMBOL_GPL(ping_init_sock);
 295
 296 void ping_close(struct sock *sk, long timeout)
 297 {
 298         pr_debug("ping_close(sk=%p,sk->num=%u)\n",
 299                  inet_sk(sk), inet_sk(sk)->inet_num);
 300         pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
 301
 302         sk_common_release(sk);
 303 }
 304 EXPORT_SYMBOL_GPL(ping_close);
 305
 306 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
 307 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
 308                                 struct sockaddr *uaddr, int addr_len) {
 309         struct net *net = sock_net(sk);
 310         if (sk->sk_family == AF_INET) {
 311                 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 312                 int chk_addr_ret;
 313
 314                 if (addr_len < sizeof(*addr))
 315                         return -EINVAL;
 316
 317                 if (addr->sin_family != AF_INET &&
 318                     !(addr->sin_family == AF_UNSPEC &&
 319                       addr->sin_addr.s_addr == htonl(INADDR_ANY)))
 320                         return -EAFNOSUPPORT;
 321
 322                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
 323                          sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
 324
 325                 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
 326
 327                 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
 328                         chk_addr_ret = RTN_LOCAL;
 329
 330                 if ((!inet_can_nonlocal_bind(net, isk) &&
 331                      chk_addr_ret != RTN_LOCAL) ||
 332                     chk_addr_ret == RTN_MULTICAST ||
 333                     chk_addr_ret == RTN_BROADCAST)
 334                         return -EADDRNOTAVAIL;
 335
 336 #if IS_ENABLED(CONFIG_IPV6)
 337         } else if (sk->sk_family == AF_INET6) {
 338                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 339                 int addr_type, scoped, has_addr;
 340                 struct net_device *dev = NULL;
 341
 342                 if (addr_len < sizeof(*addr))
 343                         return -EINVAL;
 344
 345                 if (addr->sin6_family != AF_INET6)
 346                         return -EAFNOSUPPORT;
 347
 348                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 349                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 350
 351                 addr_type = ipv6_addr_type(&addr->sin6_addr);
 352                 scoped = __ipv6_addr_needs_scope_id(addr_type);
 353                 if ((addr_type != IPV6_ADDR_ANY &&
 354                      !(addr_type & IPV6_ADDR_UNICAST)) ||
 355                     (scoped && !addr->sin6_scope_id))
 356                         return -EINVAL;
 357
 358                 rcu_read_lock();
 359                 if (addr->sin6_scope_id) {
 360                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 361                         if (!dev) {
 362                                 rcu_read_unlock();
 363                                 return -ENODEV;
 364                         }
 365                 }
 366                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 367                                                     scoped);
 368                 rcu_read_unlock();
 369
 370                 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
 371                       addr_type == IPV6_ADDR_ANY))
 372                         return -EADDRNOTAVAIL;
 373
 374                 if (scoped)
 375                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 376 #endif
 377         } else {
 378                 return -EAFNOSUPPORT;
 379         }
 380         return 0;
 381 }
 382
 383 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
 384 {
 385         if (saddr->sa_family == AF_INET) {
 386                 struct inet_sock *isk = inet_sk(sk);
 387                 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
 388                 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
 389 #if IS_ENABLED(CONFIG_IPV6)
 390         } else if (saddr->sa_family == AF_INET6) {
 391                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
 392                 struct ipv6_pinfo *np = inet6_sk(sk);
 393                 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
 394 #endif
 395         }
 396 }
 397
 398 static void ping_clear_saddr(struct sock *sk, int dif)
 399 {
 400         sk->sk_bound_dev_if = dif;
 401         if (sk->sk_family == AF_INET) {
 402                 struct inet_sock *isk = inet_sk(sk);
 403                 isk->inet_rcv_saddr = isk->inet_saddr = 0;
 404 #if IS_ENABLED(CONFIG_IPV6)
 405         } else if (sk->sk_family == AF_INET6) {
 406                 struct ipv6_pinfo *np = inet6_sk(sk);
 407                 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
 408                 memset(&np->saddr, 0, sizeof(np->saddr));
 409 #endif
 410         }
 411 }
 412 /*
 413  * We need our own bind because there are no privileged id's == local ports.
 414  * Moreover, we don't allow binding to multi- and broadcast addresses.
 415  */
 416
 417 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 418 {
 419         struct inet_sock *isk = inet_sk(sk);
 420         unsigned short snum;
 421         int err;
 422         int dif = sk->sk_bound_dev_if;
 423
 424         err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
 425         if (err)
 426                 return err;
 427
 428         lock_sock(sk);
 429
 430         err = -EINVAL;
 431         if (isk->inet_num != 0)
 432                 goto out;
 433
 434         err = -EADDRINUSE;
 435         ping_set_saddr(sk, uaddr);
 436         snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
 437         if (ping_get_port(sk, snum) != 0) {
 438                 ping_clear_saddr(sk, dif);
 439                 goto out;
 440         }
 441
 442         pr_debug("after bind(): num = %hu, dif = %d\n",
 443                  isk->inet_num,
 444                  sk->sk_bound_dev_if);
 445
 446         err = 0;
 447         if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
 448                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 449 #if IS_ENABLED(CONFIG_IPV6)
 450         if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 451                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 452 #endif
 453
 454         if (snum)
 455                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 456         isk->inet_sport = htons(isk->inet_num);
 457         isk->inet_daddr = 0;
 458         isk->inet_dport = 0;
 459
 460 #if IS_ENABLED(CONFIG_IPV6)
 461         if (sk->sk_family == AF_INET6)
 462                 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
 463 #endif
 464
 465         sk_dst_reset(sk);
 466 out:
 467         release_sock(sk);
 468         pr_debug("ping_v4_bind -> %d\n", err);
 469         return err;
 470 }
 471 EXPORT_SYMBOL_GPL(ping_bind);
 472
 473 /*
 474  * Is this a supported type of ICMP message?
 475  */
 476
 477 static inline int ping_supported(int family, int type, int code)
 478 {
 479         return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
 480                (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
 481 }
 482
 483 /*
 484  * This routine is called by the ICMP module when it gets some
 485  * sort of error condition.
 486  */
 487
 488 void ping_err(struct sk_buff *skb, int offset, u32 info)
 489 {
 490         int family;
 491         struct icmphdr *icmph;
 492         struct inet_sock *inet_sock;
 493         int type;
 494         int code;
 495         struct net *net = dev_net(skb->dev);
 496         struct sock *sk;
 497         int harderr;
 498         int err;
 499
 500         if (skb->protocol == htons(ETH_P_IP)) {
 501                 family = AF_INET;
 502                 type = icmp_hdr(skb)->type;
 503                 code = icmp_hdr(skb)->code;
 504                 icmph = (struct icmphdr *)(skb->data + offset);
 505         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 506                 family = AF_INET6;
 507                 type = icmp6_hdr(skb)->icmp6_type;
 508                 code = icmp6_hdr(skb)->icmp6_code;
 509                 icmph = (struct icmphdr *) (skb->data + offset);
 510         } else {
 511                 BUG();
 512         }
 513
 514         /* We assume the packet has already been checked by icmp_unreach */
 515
 516         if (!ping_supported(family, icmph->type, icmph->code))
 517                 return;
 518
 519         pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
 520                  skb->protocol, type, code, ntohs(icmph->un.echo.id),
 521                  ntohs(icmph->un.echo.sequence));
 522
 523         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 524         if (!sk) {
 525                 pr_debug("no socket, dropping\n");
 526                 return; /* No socket for error */
 527         }
 528         pr_debug("err on socket %p\n", sk);
 529
 530         err = 0;
 531         harderr = 0;
 532         inet_sock = inet_sk(sk);
 533
 534         if (skb->protocol == htons(ETH_P_IP)) {
 535                 switch (type) {
 536                 default:
 537                 case ICMP_TIME_EXCEEDED:
 538                         err = EHOSTUNREACH;
 539                         break;
 540                 case ICMP_SOURCE_QUENCH:
 541                         /* This is not a real error but ping wants to see it.
 542                          * Report it with some fake errno.
 543                          */
 544                         err = EREMOTEIO;
 545                         break;
 546                 case ICMP_PARAMETERPROB:
 547                         err = EPROTO;
 548                         harderr = 1;
 549                         break;
 550                 case ICMP_DEST_UNREACH:
 551                         if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
 552                                 ipv4_sk_update_pmtu(skb, sk, info);
 553                                 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
 554                                         err = EMSGSIZE;
 555                                         harderr = 1;
 556                                         break;
 557                                 }
 558                                 goto out;
 559                         }
 560                         err = EHOSTUNREACH;
 561                         if (code <= NR_ICMP_UNREACH) {
 562                                 harderr = icmp_err_convert[code].fatal;
 563                                 err = icmp_err_convert[code].errno;
 564                         }
 565                         break;
 566                 case ICMP_REDIRECT:
 567                         /* See ICMP_SOURCE_QUENCH */
 568                         ipv4_sk_redirect(skb, sk);
 569                         err = EREMOTEIO;
 570                         break;
 571                 }
 572 #if IS_ENABLED(CONFIG_IPV6)
 573         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 574                 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
 575 #endif
 576         }
 577
 578         /*
 579          *      RFC1122: OK.  Passes ICMP errors back to application, as per
 580          *      4.1.3.3.
 581          */
 582         if ((family == AF_INET && !inet_sock->recverr) ||
 583             (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
 584                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 585                         goto out;
 586         } else {
 587                 if (family == AF_INET) {
 588                         ip_icmp_error(sk, skb, err, 0 /* no remote port */,
 589                                       info, (u8 *)icmph);
 590 #if IS_ENABLED(CONFIG_IPV6)
 591                 } else if (family == AF_INET6) {
 592                         pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
 593                                                    info, (u8 *)icmph);
 594 #endif
 595                 }
 596         }
 597         sk->sk_err = err;
 598         sk->sk_error_report(sk);
 599 out:
 600         sock_put(sk);
 601 }
 602 EXPORT_SYMBOL_GPL(ping_err);
 603
 604 /*
 605  *      Copy and checksum an ICMP Echo packet from user space into a buffer
 606  *      starting from the payload.
 607  */
 608
 609 int ping_getfrag(void *from, char *to,
 610                  int offset, int fraglen, int odd, struct sk_buff *skb)
 611 {
 612         struct pingfakehdr *pfh = (struct pingfakehdr *)from;
 613
 614         if (offset == 0) {
 615                 fraglen -= sizeof(struct icmphdr);
 616                 if (fraglen < 0)
 617                         BUG();
 618                 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
 619                             fraglen, &pfh->wcheck,
 620                             &pfh->msg->msg_iter))
 621                         return -EFAULT;
 622         } else if (offset < sizeof(struct icmphdr)) {
 623                         BUG();
 624         } else {
 625                 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
 626                                             &pfh->msg->msg_iter))
 627                         return -EFAULT;
 628         }
 629
 630 #if IS_ENABLED(CONFIG_IPV6)
 631         /* For IPv6, checksum each skb as we go along, as expected by
 632          * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
 633          * wcheck, it will be finalized in ping_v4_push_pending_frames.
 634          */
 635         if (pfh->family == AF_INET6) {
 636                 skb->csum = pfh->wcheck;
 637                 skb->ip_summed = CHECKSUM_NONE;
 638                 pfh->wcheck = 0;
 639         }
 640 #endif
 641
 642         return 0;
 643 }
 644 EXPORT_SYMBOL_GPL(ping_getfrag);
 645
 646 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
 647                                        struct flowi4 *fl4)
 648 {
 649         struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
 650
 651         if (!skb)
 652                 return 0;
 653         pfh->wcheck = csum_partial((char *)&pfh->icmph,
 654                 sizeof(struct icmphdr), pfh->wcheck);
 655         pfh->icmph.checksum = csum_fold(pfh->wcheck);
 656         memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
 657         skb->ip_summed = CHECKSUM_NONE;
 658         return ip_push_pending_frames(sk, fl4);
 659 }
 660
 661 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
 662                         void *user_icmph, size_t icmph_len) {
 663         u8 type, code;
 664
 665         if (len > 0xFFFF)
 666                 return -EMSGSIZE;
 667
 668         /* Must have at least a full ICMP header. */
 669         if (len < icmph_len)
 670                 return -EINVAL;
 671
 672         /*
 673          *      Check the flags.
 674          */
 675
 676         /* Mirror BSD error message compatibility */
 677         if (msg->msg_flags & MSG_OOB)
 678                 return -EOPNOTSUPP;
 679
 680         /*
 681          *      Fetch the ICMP header provided by the userland.
 682          *      iovec is modified! The ICMP header is consumed.
 683          */
 684         if (memcpy_from_msg(user_icmph, msg, icmph_len))
 685                 return -EFAULT;
 686
 687         if (family == AF_INET) {
 688                 type = ((struct icmphdr *) user_icmph)->type;
 689                 code = ((struct icmphdr *) user_icmph)->code;
 690 #if IS_ENABLED(CONFIG_IPV6)
 691         } else if (family == AF_INET6) {
 692                 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
 693                 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
 694 #endif
 695         } else {
 696                 BUG();
 697         }
 698
 699         if (!ping_supported(family, type, code))
 700                 return -EINVAL;
 701
 702         return 0;
 703 }
 704 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
 705
 706 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 707 {
 708         struct net *net = sock_net(sk);
 709         struct flowi4 fl4;
 710         struct inet_sock *inet = inet_sk(sk);
 711         struct ipcm_cookie ipc;
 712         struct icmphdr user_icmph;
 713         struct pingfakehdr pfh;
 714         struct rtable *rt = NULL;
 715         struct ip_options_data opt_copy;
 716         int free = 0;
 717         __be32 saddr, daddr, faddr;
 718         u8  tos;
 719         int err;
 720
 721         pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
 722
 723         err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
 724                                   sizeof(user_icmph));
 725         if (err)
 726                 return err;
 727
 728         /*
 729          *      Get and verify the address.
 730          */
 731
 732         if (msg->msg_name) {
 733                 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
 734                 if (msg->msg_namelen < sizeof(*usin))
 735                         return -EINVAL;
 736                 if (usin->sin_family != AF_INET)
 737                         return -EAFNOSUPPORT;
 738                 daddr = usin->sin_addr.s_addr;
 739                 /* no remote port */
 740         } else {
 741                 if (sk->sk_state != TCP_ESTABLISHED)
 742                         return -EDESTADDRREQ;
 743                 daddr = inet->inet_daddr;
 744                 /* no remote port */
 745         }
 746
 747         ipcm_init_sk(&ipc, inet);
 748
 749         if (msg->msg_controllen) {
 750                 err = ip_cmsg_send(sk, msg, &ipc, false);
 751                 if (unlikely(err)) {
 752                         kfree(ipc.opt);
 753                         return err;
 754                 }
 755                 if (ipc.opt)
 756                         free = 1;
 757         }
 758         if (!ipc.opt) {
 759                 struct ip_options_rcu *inet_opt;
 760
 761                 rcu_read_lock();
 762                 inet_opt = rcu_dereference(inet->inet_opt);
 763                 if (inet_opt) {
 764                         memcpy(&opt_copy, inet_opt,
 765                                sizeof(*inet_opt) + inet_opt->opt.optlen);
 766                         ipc.opt = &opt_copy.opt;
 767                 }
 768                 rcu_read_unlock();
 769         }
 770
 771         saddr = ipc.addr;
 772         ipc.addr = faddr = daddr;
 773
 774         if (ipc.opt && ipc.opt->opt.srr) {
 775                 if (!daddr) {
 776                         err = -EINVAL;
 777                         goto out_free;
 778                 }
 779                 faddr = ipc.opt->opt.faddr;
 780         }
 781         tos = get_rttos(&ipc, inet);
 782         if (sock_flag(sk, SOCK_LOCALROUTE) ||
 783             (msg->msg_flags & MSG_DONTROUTE) ||
 784             (ipc.opt && ipc.opt->opt.is_strictroute)) {
 785                 tos |= RTO_ONLINK;
 786         }
 787
 788         if (ipv4_is_multicast(daddr)) {
 789                 if (!ipc.oif)
 790                         ipc.oif = inet->mc_index;
 791                 if (!saddr)
 792                         saddr = inet->mc_addr;
 793         } else if (!ipc.oif)
 794                 ipc.oif = inet->uc_index;
 795
 796         flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
 797                            RT_SCOPE_UNIVERSE, sk->sk_protocol,
 798                            inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
 799                            sk->sk_uid);
 800
 801         fl4.fl4_icmp_type = user_icmph.type;
 802         fl4.fl4_icmp_code = user_icmph.code;
 803
 804         security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
 805         rt = ip_route_output_flow(net, &fl4, sk);
 806         if (IS_ERR(rt)) {
 807                 err = PTR_ERR(rt);
 808                 rt = NULL;
 809                 if (err == -ENETUNREACH)
 810                         IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
 811                 goto out;
 812         }
 813
 814         err = -EACCES;
 815         if ((rt->rt_flags & RTCF_BROADCAST) &&
 816             !sock_flag(sk, SOCK_BROADCAST))
 817                 goto out;
 818
 819         if (msg->msg_flags & MSG_CONFIRM)
 820                 goto do_confirm;
 821 back_from_confirm:
 822
 823         if (!ipc.addr)
 824                 ipc.addr = fl4.daddr;
 825
 826         lock_sock(sk);
 827
 828         pfh.icmph.type = user_icmph.type; /* already checked */
 829         pfh.icmph.code = user_icmph.code; /* ditto */
 830         pfh.icmph.checksum = 0;
 831         pfh.icmph.un.echo.id = inet->inet_sport;
 832         pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 833         pfh.msg = msg;
 834         pfh.wcheck = 0;
 835         pfh.family = AF_INET;
 836
 837         err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 838                         0, &ipc, &rt, msg->msg_flags);
 839         if (err)
 840                 ip_flush_pending_frames(sk);
 841         else
 842                 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 843         release_sock(sk);
 844
 845 out:
 846         ip_rt_put(rt);
 847 out_free:
 848         if (free)
 849                 kfree(ipc.opt);
 850         if (!err) {
 851                 icmp_out_count(sock_net(sk), user_icmph.type);
 852                 return len;
 853         }
 854         return err;
 855
 856 do_confirm:
 857         if (msg->msg_flags & MSG_PROBE)
 858                 dst_confirm_neigh(&rt->dst, &fl4.daddr);
 859         if (!(msg->msg_flags & MSG_PROBE) || len)
 860                 goto back_from_confirm;
 861         err = 0;
 862         goto out;
 863 }
 864
 865 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
 866                  int flags, int *addr_len)
 867 {
 868         struct inet_sock *isk = inet_sk(sk);
 869         int family = sk->sk_family;
 870         struct sk_buff *skb;
 871         int copied, err;
 872
 873         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 874
 875         err = -EOPNOTSUPP;
 876         if (flags & MSG_OOB)
 877                 goto out;
 878
 879         if (flags & MSG_ERRQUEUE)
 880                 return inet_recv_error(sk, msg, len, addr_len);
 881
 882         skb = skb_recv_datagram(sk, flags, noblock, &err);
 883         if (!skb)
 884                 goto out;
 885
 886         copied = skb->len;
 887         if (copied > len) {
 888                 msg->msg_flags |= MSG_TRUNC;
 889                 copied = len;
 890         }
 891
 892         /* Don't bother checking the checksum */
 893         err = skb_copy_datagram_msg(skb, 0, msg, copied);
 894         if (err)
 895                 goto done;
 896
 897         sock_recv_timestamp(msg, sk, skb);
 898
 899         /* Copy the address and add cmsg data. */
 900         if (family == AF_INET) {
 901                 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 902
 903                 if (sin) {
 904                         sin->sin_family = AF_INET;
 905                         sin->sin_port = 0 /* skb->h.uh->source */;
 906                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 907                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 908                         *addr_len = sizeof(*sin);
 909                 }
 910
 911                 if (isk->cmsg_flags)
 912                         ip_cmsg_recv(msg, skb);
 913
 914 #if IS_ENABLED(CONFIG_IPV6)
 915         } else if (family == AF_INET6) {
 916                 struct ipv6_pinfo *np = inet6_sk(sk);
 917                 struct ipv6hdr *ip6 = ipv6_hdr(skb);
 918                 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 919
 920                 if (sin6) {
 921                         sin6->sin6_family = AF_INET6;
 922                         sin6->sin6_port = 0;
 923                         sin6->sin6_addr = ip6->saddr;
 924                         sin6->sin6_flowinfo = 0;
 925                         if (np->sndflow)
 926                                 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
 927                         sin6->sin6_scope_id =
 928                                 ipv6_iface_scope_id(&sin6->sin6_addr,
 929                                                     inet6_iif(skb));
 930                         *addr_len = sizeof(*sin6);
 931                 }
 932
 933                 if (inet6_sk(sk)->rxopt.all)
 934                         pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
 935                 if (skb->protocol == htons(ETH_P_IPV6) &&
 936                     inet6_sk(sk)->rxopt.all)
 937                         pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
 938                 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
 939                         ip_cmsg_recv(msg, skb);
 940 #endif
 941         } else {
 942                 BUG();
 943         }
 944
 945         err = copied;
 946
 947 done:
 948         skb_free_datagram(sk, skb);
 949 out:
 950         pr_debug("ping_recvmsg -> %d\n", err);
 951         return err;
 952 }
 953 EXPORT_SYMBOL_GPL(ping_recvmsg);
 954
 955 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 956 {
 957         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 958                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
 959         if (sock_queue_rcv_skb(sk, skb) < 0) {
 960                 kfree_skb(skb);
 961                 pr_debug("ping_queue_rcv_skb -> failed\n");
 962                 return -1;
 963         }
 964         return 0;
 965 }
 966 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 967
 968
 969 /*
 970  *      All we need to do is get the socket.
 971  */
 972
 973 bool ping_rcv(struct sk_buff *skb)
 974 {
 975         struct sock *sk;
 976         struct net *net = dev_net(skb->dev);
 977         struct icmphdr *icmph = icmp_hdr(skb);
 978         bool rc = false;
 979
 980         /* We assume the packet has already been checked by icmp_rcv */
 981
 982         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 983                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 984
 985         /* Push ICMP header back */
 986         skb_push(skb, skb->data - (u8 *)icmph);
 987
 988         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 989         if (sk) {
 990                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 991
 992                 pr_debug("rcv on socket %p\n", sk);
 993                 if (skb2 && !ping_queue_rcv_skb(sk, skb2))
 994                         rc = true;
 995                 sock_put(sk);
 996         }
 997
 998         if (!rc)
 999                 pr_debug("no socket, dropping\n");
1000
1001         return rc;
1002 }
1003 EXPORT_SYMBOL_GPL(ping_rcv);
1004
1005 struct proto ping_prot = {
1006         .name =         "PING",
1007         .owner =        THIS_MODULE,
1008         .init =         ping_init_sock,
1009         .close =        ping_close,
1010         .connect =      ip4_datagram_connect,
1011         .disconnect =   __udp_disconnect,
1012         .setsockopt =   ip_setsockopt,
1013         .getsockopt =   ip_getsockopt,
1014         .sendmsg =      ping_v4_sendmsg,
1015         .recvmsg =      ping_recvmsg,
1016         .bind =         ping_bind,
1017         .backlog_rcv =  ping_queue_rcv_skb,
1018         .release_cb =   ip4_datagram_release_cb,
1019         .hash =         ping_hash,
1020         .unhash =       ping_unhash,
1021         .get_port =     ping_get_port,
1022         .obj_size =     sizeof(struct inet_sock),
1023 };
1024 EXPORT_SYMBOL(ping_prot);
1025
1026 #ifdef CONFIG_PROC_FS
1027
1028 static struct sock *ping_get_first(struct seq_file *seq, int start)
1029 {
1030         struct sock *sk;
1031         struct ping_iter_state *state = seq->private;
1032         struct net *net = seq_file_net(seq);
1033
1034         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1035              ++state->bucket) {
1036                 struct hlist_nulls_node *node;
1037                 struct hlist_nulls_head *hslot;
1038
1039                 hslot = &ping_table.hash[state->bucket];
1040
1041                 if (hlist_nulls_empty(hslot))
1042                         continue;
1043
1044                 sk_nulls_for_each(sk, node, hslot) {
1045                         if (net_eq(sock_net(sk), net) &&
1046                             sk->sk_family == state->family)
1047                                 goto found;
1048                 }
1049         }
1050         sk = NULL;
1051 found:
1052         return sk;
1053 }
1054
1055 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1056 {
1057         struct ping_iter_state *state = seq->private;
1058         struct net *net = seq_file_net(seq);
1059
1060         do {
1061                 sk = sk_nulls_next(sk);
1062         } while (sk && (!net_eq(sock_net(sk), net)));
1063
1064         if (!sk)
1065                 return ping_get_first(seq, state->bucket + 1);
1066         return sk;
1067 }
1068
1069 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1070 {
1071         struct sock *sk = ping_get_first(seq, 0);
1072
1073         if (sk)
1074                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1075                         --pos;
1076         return pos ? NULL : sk;
1077 }
1078
1079 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1080         __acquires(ping_table.lock)
1081 {
1082         struct ping_iter_state *state = seq->private;
1083         state->bucket = 0;
1084         state->family = family;
1085
1086         read_lock_bh(&ping_table.lock);
1087
1088         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1089 }
1090 EXPORT_SYMBOL_GPL(ping_seq_start);
1091
1092 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1093 {
1094         return ping_seq_start(seq, pos, AF_INET);
1095 }
1096
1097 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1098 {
1099         struct sock *sk;
1100
1101         if (v == SEQ_START_TOKEN)
1102                 sk = ping_get_idx(seq, 0);
1103         else
1104                 sk = ping_get_next(seq, v);
1105
1106         ++*pos;
1107         return sk;
1108 }
1109 EXPORT_SYMBOL_GPL(ping_seq_next);
1110
1111 void ping_seq_stop(struct seq_file *seq, void *v)
1112         __releases(ping_table.lock)
1113 {
1114         read_unlock_bh(&ping_table.lock);
1115 }
1116 EXPORT_SYMBOL_GPL(ping_seq_stop);
1117
1118 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1119                 int bucket)
1120 {
1121         struct inet_sock *inet = inet_sk(sp);
1122         __be32 dest = inet->inet_daddr;
1123         __be32 src = inet->inet_rcv_saddr;
1124         __u16 destp = ntohs(inet->inet_dport);
1125         __u16 srcp = ntohs(inet->inet_sport);
1126
1127         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1128                 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1129                 bucket, src, srcp, dest, destp, sp->sk_state,
1130                 sk_wmem_alloc_get(sp),
1131                 sk_rmem_alloc_get(sp),
1132                 0, 0L, 0,
1133                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1134                 0, sock_i_ino(sp),
1135                 refcount_read(&sp->sk_refcnt), sp,
1136                 atomic_read(&sp->sk_drops));
1137 }
1138
1139 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1140 {
1141         seq_setwidth(seq, 127);
1142         if (v == SEQ_START_TOKEN)
1143                 seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1144                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1145                            "inode ref pointer drops");
1146         else {
1147                 struct ping_iter_state *state = seq->private;
1148
1149                 ping_v4_format_sock(v, seq, state->bucket);
1150         }
1151         seq_pad(seq, '\n');
1152         return 0;
1153 }
1154
1155 static const struct seq_operations ping_v4_seq_ops = {
1156         .start          = ping_v4_seq_start,
1157         .show           = ping_v4_seq_show,
1158         .next           = ping_seq_next,
1159         .stop           = ping_seq_stop,
1160 };
1161
1162 static int __net_init ping_v4_proc_init_net(struct net *net)
1163 {
1164         if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1165                         sizeof(struct ping_iter_state)))
1166                 return -ENOMEM;
1167         return 0;
1168 }
1169
1170 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1171 {
1172         remove_proc_entry("icmp", net->proc_net);
1173 }
1174
1175 static struct pernet_operations ping_v4_net_ops = {
1176         .init = ping_v4_proc_init_net,
1177         .exit = ping_v4_proc_exit_net,
1178 };
1179
1180 int __init ping_proc_init(void)
1181 {
1182         return register_pernet_subsys(&ping_v4_net_ops);
1183 }
1184
1185 void ping_proc_exit(void)
1186 {
1187         unregister_pernet_subsys(&ping_v4_net_ops);
1188 }
1189
1190 #endif
1191
1192 void __init ping_init(void)
1193 {
1194         int i;
1195
1196         for (i = 0; i < PING_HTABLE_SIZE; i++)
1197                 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1198         rwlock_init(&ping_table.lock);
1199 }