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                 u32 tb_id = RT_TABLE_LOCAL;
 313                 int chk_addr_ret;
 314
 315                 if (addr_len < sizeof(*addr))
 316                         return -EINVAL;
 317
 318                 if (addr->sin_family != AF_INET &&
 319                     !(addr->sin_family == AF_UNSPEC &&
 320                       addr->sin_addr.s_addr == htonl(INADDR_ANY)))
 321                         return -EAFNOSUPPORT;
 322
 323                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
 324                          sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
 325
 326                 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
 327                 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
 328
 329                 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
 330                         chk_addr_ret = RTN_LOCAL;
 331
 332                 if ((!inet_can_nonlocal_bind(net, isk) &&
 333                      chk_addr_ret != RTN_LOCAL) ||
 334                     chk_addr_ret == RTN_MULTICAST ||
 335                     chk_addr_ret == RTN_BROADCAST)
 336                         return -EADDRNOTAVAIL;
 337
 338 #if IS_ENABLED(CONFIG_IPV6)
 339         } else if (sk->sk_family == AF_INET6) {
 340                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 341                 int addr_type, scoped, has_addr;
 342                 struct net_device *dev = NULL;
 343
 344                 if (addr_len < sizeof(*addr))
 345                         return -EINVAL;
 346
 347                 if (addr->sin6_family != AF_INET6)
 348                         return -EAFNOSUPPORT;
 349
 350                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 351                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 352
 353                 addr_type = ipv6_addr_type(&addr->sin6_addr);
 354                 scoped = __ipv6_addr_needs_scope_id(addr_type);
 355                 if ((addr_type != IPV6_ADDR_ANY &&
 356                      !(addr_type & IPV6_ADDR_UNICAST)) ||
 357                     (scoped && !addr->sin6_scope_id))
 358                         return -EINVAL;
 359
 360                 rcu_read_lock();
 361                 if (addr->sin6_scope_id) {
 362                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 363                         if (!dev) {
 364                                 rcu_read_unlock();
 365                                 return -ENODEV;
 366                         }
 367                 }
 368
 369                 if (!dev && sk->sk_bound_dev_if) {
 370                         dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 371                         if (!dev) {
 372                                 rcu_read_unlock();
 373                                 return -ENODEV;
 374                         }
 375                 }
 376                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 377                                                     scoped);
 378                 rcu_read_unlock();
 379
 380                 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
 381                       addr_type == IPV6_ADDR_ANY))
 382                         return -EADDRNOTAVAIL;
 383
 384                 if (scoped)
 385                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 386 #endif
 387         } else {
 388                 return -EAFNOSUPPORT;
 389         }
 390         return 0;
 391 }
 392
 393 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
 394 {
 395         if (saddr->sa_family == AF_INET) {
 396                 struct inet_sock *isk = inet_sk(sk);
 397                 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
 398                 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
 399 #if IS_ENABLED(CONFIG_IPV6)
 400         } else if (saddr->sa_family == AF_INET6) {
 401                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
 402                 struct ipv6_pinfo *np = inet6_sk(sk);
 403                 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
 404 #endif
 405         }
 406 }
 407
 408 static void ping_clear_saddr(struct sock *sk, int dif)
 409 {
 410         sk->sk_bound_dev_if = dif;
 411         if (sk->sk_family == AF_INET) {
 412                 struct inet_sock *isk = inet_sk(sk);
 413                 isk->inet_rcv_saddr = isk->inet_saddr = 0;
 414 #if IS_ENABLED(CONFIG_IPV6)
 415         } else if (sk->sk_family == AF_INET6) {
 416                 struct ipv6_pinfo *np = inet6_sk(sk);
 417                 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
 418                 memset(&np->saddr, 0, sizeof(np->saddr));
 419 #endif
 420         }
 421 }
 422 /*
 423  * We need our own bind because there are no privileged id's == local ports.
 424  * Moreover, we don't allow binding to multi- and broadcast addresses.
 425  */
 426
 427 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 428 {
 429         struct inet_sock *isk = inet_sk(sk);
 430         unsigned short snum;
 431         int err;
 432         int dif = sk->sk_bound_dev_if;
 433
 434         err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
 435         if (err)
 436                 return err;
 437
 438         lock_sock(sk);
 439
 440         err = -EINVAL;
 441         if (isk->inet_num != 0)
 442                 goto out;
 443
 444         err = -EADDRINUSE;
 445         ping_set_saddr(sk, uaddr);
 446         snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
 447         if (ping_get_port(sk, snum) != 0) {
 448                 ping_clear_saddr(sk, dif);
 449                 goto out;
 450         }
 451
 452         pr_debug("after bind(): num = %hu, dif = %d\n",
 453                  isk->inet_num,
 454                  sk->sk_bound_dev_if);
 455
 456         err = 0;
 457         if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
 458                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 459 #if IS_ENABLED(CONFIG_IPV6)
 460         if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 461                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 462 #endif
 463
 464         if (snum)
 465                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 466         isk->inet_sport = htons(isk->inet_num);
 467         isk->inet_daddr = 0;
 468         isk->inet_dport = 0;
 469
 470 #if IS_ENABLED(CONFIG_IPV6)
 471         if (sk->sk_family == AF_INET6)
 472                 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
 473 #endif
 474
 475         sk_dst_reset(sk);
 476 out:
 477         release_sock(sk);
 478         pr_debug("ping_v4_bind -> %d\n", err);
 479         return err;
 480 }
 481 EXPORT_SYMBOL_GPL(ping_bind);
 482
 483 /*
 484  * Is this a supported type of ICMP message?
 485  */
 486
 487 static inline int ping_supported(int family, int type, int code)
 488 {
 489         return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
 490                (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
 491 }
 492
 493 /*
 494  * This routine is called by the ICMP module when it gets some
 495  * sort of error condition.
 496  */
 497
 498 void ping_err(struct sk_buff *skb, int offset, u32 info)
 499 {
 500         int family;
 501         struct icmphdr *icmph;
 502         struct inet_sock *inet_sock;
 503         int type;
 504         int code;
 505         struct net *net = dev_net(skb->dev);
 506         struct sock *sk;
 507         int harderr;
 508         int err;
 509
 510         if (skb->protocol == htons(ETH_P_IP)) {
 511                 family = AF_INET;
 512                 type = icmp_hdr(skb)->type;
 513                 code = icmp_hdr(skb)->code;
 514                 icmph = (struct icmphdr *)(skb->data + offset);
 515         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 516                 family = AF_INET6;
 517                 type = icmp6_hdr(skb)->icmp6_type;
 518                 code = icmp6_hdr(skb)->icmp6_code;
 519                 icmph = (struct icmphdr *) (skb->data + offset);
 520         } else {
 521                 BUG();
 522         }
 523
 524         /* We assume the packet has already been checked by icmp_unreach */
 525
 526         if (!ping_supported(family, icmph->type, icmph->code))
 527                 return;
 528
 529         pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
 530                  skb->protocol, type, code, ntohs(icmph->un.echo.id),
 531                  ntohs(icmph->un.echo.sequence));
 532
 533         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 534         if (!sk) {
 535                 pr_debug("no socket, dropping\n");
 536                 return; /* No socket for error */
 537         }
 538         pr_debug("err on socket %p\n", sk);
 539
 540         err = 0;
 541         harderr = 0;
 542         inet_sock = inet_sk(sk);
 543
 544         if (skb->protocol == htons(ETH_P_IP)) {
 545                 switch (type) {
 546                 default:
 547                 case ICMP_TIME_EXCEEDED:
 548                         err = EHOSTUNREACH;
 549                         break;
 550                 case ICMP_SOURCE_QUENCH:
 551                         /* This is not a real error but ping wants to see it.
 552                          * Report it with some fake errno.
 553                          */
 554                         err = EREMOTEIO;
 555                         break;
 556                 case ICMP_PARAMETERPROB:
 557                         err = EPROTO;
 558                         harderr = 1;
 559                         break;
 560                 case ICMP_DEST_UNREACH:
 561                         if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
 562                                 ipv4_sk_update_pmtu(skb, sk, info);
 563                                 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
 564                                         err = EMSGSIZE;
 565                                         harderr = 1;
 566                                         break;
 567                                 }
 568                                 goto out;
 569                         }
 570                         err = EHOSTUNREACH;
 571                         if (code <= NR_ICMP_UNREACH) {
 572                                 harderr = icmp_err_convert[code].fatal;
 573                                 err = icmp_err_convert[code].errno;
 574                         }
 575                         break;
 576                 case ICMP_REDIRECT:
 577                         /* See ICMP_SOURCE_QUENCH */
 578                         ipv4_sk_redirect(skb, sk);
 579                         err = EREMOTEIO;
 580                         break;
 581                 }
 582 #if IS_ENABLED(CONFIG_IPV6)
 583         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 584                 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
 585 #endif
 586         }
 587
 588         /*
 589          *      RFC1122: OK.  Passes ICMP errors back to application, as per
 590          *      4.1.3.3.
 591          */
 592         if ((family == AF_INET && !inet_sock->recverr) ||
 593             (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
 594                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 595                         goto out;
 596         } else {
 597                 if (family == AF_INET) {
 598                         ip_icmp_error(sk, skb, err, 0 /* no remote port */,
 599                                       info, (u8 *)icmph);
 600 #if IS_ENABLED(CONFIG_IPV6)
 601                 } else if (family == AF_INET6) {
 602                         pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
 603                                                    info, (u8 *)icmph);
 604 #endif
 605                 }
 606         }
 607         sk->sk_err = err;
 608         sk->sk_error_report(sk);
 609 out:
 610         sock_put(sk);
 611 }
 612 EXPORT_SYMBOL_GPL(ping_err);
 613
 614 /*
 615  *      Copy and checksum an ICMP Echo packet from user space into a buffer
 616  *      starting from the payload.
 617  */
 618
 619 int ping_getfrag(void *from, char *to,
 620                  int offset, int fraglen, int odd, struct sk_buff *skb)
 621 {
 622         struct pingfakehdr *pfh = (struct pingfakehdr *)from;
 623
 624         if (offset == 0) {
 625                 fraglen -= sizeof(struct icmphdr);
 626                 if (fraglen < 0)
 627                         BUG();
 628                 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
 629                             fraglen, &pfh->wcheck,
 630                             &pfh->msg->msg_iter))
 631                         return -EFAULT;
 632         } else if (offset < sizeof(struct icmphdr)) {
 633                         BUG();
 634         } else {
 635                 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
 636                                             &pfh->msg->msg_iter))
 637                         return -EFAULT;
 638         }
 639
 640 #if IS_ENABLED(CONFIG_IPV6)
 641         /* For IPv6, checksum each skb as we go along, as expected by
 642          * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
 643          * wcheck, it will be finalized in ping_v4_push_pending_frames.
 644          */
 645         if (pfh->family == AF_INET6) {
 646                 skb->csum = pfh->wcheck;
 647                 skb->ip_summed = CHECKSUM_NONE;
 648                 pfh->wcheck = 0;
 649         }
 650 #endif
 651
 652         return 0;
 653 }
 654 EXPORT_SYMBOL_GPL(ping_getfrag);
 655
 656 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
 657                                        struct flowi4 *fl4)
 658 {
 659         struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
 660
 661         if (!skb)
 662                 return 0;
 663         pfh->wcheck = csum_partial((char *)&pfh->icmph,
 664                 sizeof(struct icmphdr), pfh->wcheck);
 665         pfh->icmph.checksum = csum_fold(pfh->wcheck);
 666         memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
 667         skb->ip_summed = CHECKSUM_NONE;
 668         return ip_push_pending_frames(sk, fl4);
 669 }
 670
 671 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
 672                         void *user_icmph, size_t icmph_len) {
 673         u8 type, code;
 674
 675         if (len > 0xFFFF)
 676                 return -EMSGSIZE;
 677
 678         /* Must have at least a full ICMP header. */
 679         if (len < icmph_len)
 680                 return -EINVAL;
 681
 682         /*
 683          *      Check the flags.
 684          */
 685
 686         /* Mirror BSD error message compatibility */
 687         if (msg->msg_flags & MSG_OOB)
 688                 return -EOPNOTSUPP;
 689
 690         /*
 691          *      Fetch the ICMP header provided by the userland.
 692          *      iovec is modified! The ICMP header is consumed.
 693          */
 694         if (memcpy_from_msg(user_icmph, msg, icmph_len))
 695                 return -EFAULT;
 696
 697         if (family == AF_INET) {
 698                 type = ((struct icmphdr *) user_icmph)->type;
 699                 code = ((struct icmphdr *) user_icmph)->code;
 700 #if IS_ENABLED(CONFIG_IPV6)
 701         } else if (family == AF_INET6) {
 702                 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
 703                 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
 704 #endif
 705         } else {
 706                 BUG();
 707         }
 708
 709         if (!ping_supported(family, type, code))
 710                 return -EINVAL;
 711
 712         return 0;
 713 }
 714 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
 715
 716 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 717 {
 718         struct net *net = sock_net(sk);
 719         struct flowi4 fl4;
 720         struct inet_sock *inet = inet_sk(sk);
 721         struct ipcm_cookie ipc;
 722         struct icmphdr user_icmph;
 723         struct pingfakehdr pfh;
 724         struct rtable *rt = NULL;
 725         struct ip_options_data opt_copy;
 726         int free = 0;
 727         __be32 saddr, daddr, faddr;
 728         u8  tos;
 729         int err;
 730
 731         pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
 732
 733         err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
 734                                   sizeof(user_icmph));
 735         if (err)
 736                 return err;
 737
 738         /*
 739          *      Get and verify the address.
 740          */
 741
 742         if (msg->msg_name) {
 743                 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
 744                 if (msg->msg_namelen < sizeof(*usin))
 745                         return -EINVAL;
 746                 if (usin->sin_family != AF_INET)
 747                         return -EAFNOSUPPORT;
 748                 daddr = usin->sin_addr.s_addr;
 749                 /* no remote port */
 750         } else {
 751                 if (sk->sk_state != TCP_ESTABLISHED)
 752                         return -EDESTADDRREQ;
 753                 daddr = inet->inet_daddr;
 754                 /* no remote port */
 755         }
 756
 757         ipcm_init_sk(&ipc, inet);
 758
 759         if (msg->msg_controllen) {
 760                 err = ip_cmsg_send(sk, msg, &ipc, false);
 761                 if (unlikely(err)) {
 762                         kfree(ipc.opt);
 763                         return err;
 764                 }
 765                 if (ipc.opt)
 766                         free = 1;
 767         }
 768         if (!ipc.opt) {
 769                 struct ip_options_rcu *inet_opt;
 770
 771                 rcu_read_lock();
 772                 inet_opt = rcu_dereference(inet->inet_opt);
 773                 if (inet_opt) {
 774                         memcpy(&opt_copy, inet_opt,
 775                                sizeof(*inet_opt) + inet_opt->opt.optlen);
 776                         ipc.opt = &opt_copy.opt;
 777                 }
 778                 rcu_read_unlock();
 779         }
 780
 781         saddr = ipc.addr;
 782         ipc.addr = faddr = daddr;
 783
 784         if (ipc.opt && ipc.opt->opt.srr) {
 785                 if (!daddr) {
 786                         err = -EINVAL;
 787                         goto out_free;
 788                 }
 789                 faddr = ipc.opt->opt.faddr;
 790         }
 791         tos = get_rttos(&ipc, inet);
 792         if (sock_flag(sk, SOCK_LOCALROUTE) ||
 793             (msg->msg_flags & MSG_DONTROUTE) ||
 794             (ipc.opt && ipc.opt->opt.is_strictroute)) {
 795                 tos |= RTO_ONLINK;
 796         }
 797
 798         if (ipv4_is_multicast(daddr)) {
 799                 if (!ipc.oif)
 800                         ipc.oif = inet->mc_index;
 801                 if (!saddr)
 802                         saddr = inet->mc_addr;
 803         } else if (!ipc.oif)
 804                 ipc.oif = inet->uc_index;
 805
 806         flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
 807                            RT_SCOPE_UNIVERSE, sk->sk_protocol,
 808                            inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
 809                            sk->sk_uid);
 810
 811         fl4.fl4_icmp_type = user_icmph.type;
 812         fl4.fl4_icmp_code = user_icmph.code;
 813
 814         security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
 815         rt = ip_route_output_flow(net, &fl4, sk);
 816         if (IS_ERR(rt)) {
 817                 err = PTR_ERR(rt);
 818                 rt = NULL;
 819                 if (err == -ENETUNREACH)
 820                         IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
 821                 goto out;
 822         }
 823
 824         err = -EACCES;
 825         if ((rt->rt_flags & RTCF_BROADCAST) &&
 826             !sock_flag(sk, SOCK_BROADCAST))
 827                 goto out;
 828
 829         if (msg->msg_flags & MSG_CONFIRM)
 830                 goto do_confirm;
 831 back_from_confirm:
 832
 833         if (!ipc.addr)
 834                 ipc.addr = fl4.daddr;
 835
 836         lock_sock(sk);
 837
 838         pfh.icmph.type = user_icmph.type; /* already checked */
 839         pfh.icmph.code = user_icmph.code; /* ditto */
 840         pfh.icmph.checksum = 0;
 841         pfh.icmph.un.echo.id = inet->inet_sport;
 842         pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 843         pfh.msg = msg;
 844         pfh.wcheck = 0;
 845         pfh.family = AF_INET;
 846
 847         err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 848                         0, &ipc, &rt, msg->msg_flags);
 849         if (err)
 850                 ip_flush_pending_frames(sk);
 851         else
 852                 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 853         release_sock(sk);
 854
 855 out:
 856         ip_rt_put(rt);
 857 out_free:
 858         if (free)
 859                 kfree(ipc.opt);
 860         if (!err) {
 861                 icmp_out_count(sock_net(sk), user_icmph.type);
 862                 return len;
 863         }
 864         return err;
 865
 866 do_confirm:
 867         if (msg->msg_flags & MSG_PROBE)
 868                 dst_confirm_neigh(&rt->dst, &fl4.daddr);
 869         if (!(msg->msg_flags & MSG_PROBE) || len)
 870                 goto back_from_confirm;
 871         err = 0;
 872         goto out;
 873 }
 874
 875 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
 876                  int flags, int *addr_len)
 877 {
 878         struct inet_sock *isk = inet_sk(sk);
 879         int family = sk->sk_family;
 880         struct sk_buff *skb;
 881         int copied, err;
 882
 883         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 884
 885         err = -EOPNOTSUPP;
 886         if (flags & MSG_OOB)
 887                 goto out;
 888
 889         if (flags & MSG_ERRQUEUE)
 890                 return inet_recv_error(sk, msg, len, addr_len);
 891
 892         skb = skb_recv_datagram(sk, flags, noblock, &err);
 893         if (!skb)
 894                 goto out;
 895
 896         copied = skb->len;
 897         if (copied > len) {
 898                 msg->msg_flags |= MSG_TRUNC;
 899                 copied = len;
 900         }
 901
 902         /* Don't bother checking the checksum */
 903         err = skb_copy_datagram_msg(skb, 0, msg, copied);
 904         if (err)
 905                 goto done;
 906
 907         sock_recv_timestamp(msg, sk, skb);
 908
 909         /* Copy the address and add cmsg data. */
 910         if (family == AF_INET) {
 911                 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 912
 913                 if (sin) {
 914                         sin->sin_family = AF_INET;
 915                         sin->sin_port = 0 /* skb->h.uh->source */;
 916                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 917                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 918                         *addr_len = sizeof(*sin);
 919                 }
 920
 921                 if (isk->cmsg_flags)
 922                         ip_cmsg_recv(msg, skb);
 923
 924 #if IS_ENABLED(CONFIG_IPV6)
 925         } else if (family == AF_INET6) {
 926                 struct ipv6_pinfo *np = inet6_sk(sk);
 927                 struct ipv6hdr *ip6 = ipv6_hdr(skb);
 928                 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 929
 930                 if (sin6) {
 931                         sin6->sin6_family = AF_INET6;
 932                         sin6->sin6_port = 0;
 933                         sin6->sin6_addr = ip6->saddr;
 934                         sin6->sin6_flowinfo = 0;
 935                         if (np->sndflow)
 936                                 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
 937                         sin6->sin6_scope_id =
 938                                 ipv6_iface_scope_id(&sin6->sin6_addr,
 939                                                     inet6_iif(skb));
 940                         *addr_len = sizeof(*sin6);
 941                 }
 942
 943                 if (inet6_sk(sk)->rxopt.all)
 944                         pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
 945                 if (skb->protocol == htons(ETH_P_IPV6) &&
 946                     inet6_sk(sk)->rxopt.all)
 947                         pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
 948                 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
 949                         ip_cmsg_recv(msg, skb);
 950 #endif
 951         } else {
 952                 BUG();
 953         }
 954
 955         err = copied;
 956
 957 done:
 958         skb_free_datagram(sk, skb);
 959 out:
 960         pr_debug("ping_recvmsg -> %d\n", err);
 961         return err;
 962 }
 963 EXPORT_SYMBOL_GPL(ping_recvmsg);
 964
 965 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 966 {
 967         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 968                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
 969         if (sock_queue_rcv_skb(sk, skb) < 0) {
 970                 kfree_skb(skb);
 971                 pr_debug("ping_queue_rcv_skb -> failed\n");
 972                 return -1;
 973         }
 974         return 0;
 975 }
 976 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 977
 978
 979 /*
 980  *      All we need to do is get the socket.
 981  */
 982
 983 bool ping_rcv(struct sk_buff *skb)
 984 {
 985         struct sock *sk;
 986         struct net *net = dev_net(skb->dev);
 987         struct icmphdr *icmph = icmp_hdr(skb);
 988         bool rc = false;
 989
 990         /* We assume the packet has already been checked by icmp_rcv */
 991
 992         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 993                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 994
 995         /* Push ICMP header back */
 996         skb_push(skb, skb->data - (u8 *)icmph);
 997
 998         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 999         if (sk) {
1000                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1001
1002                 pr_debug("rcv on socket %p\n", sk);
1003                 if (skb2 && !ping_queue_rcv_skb(sk, skb2))
1004                         rc = true;
1005                 sock_put(sk);
1006         }
1007
1008         if (!rc)
1009                 pr_debug("no socket, dropping\n");
1010
1011         return rc;
1012 }
1013 EXPORT_SYMBOL_GPL(ping_rcv);
1014
1015 struct proto ping_prot = {
1016         .name =         "PING",
1017         .owner =        THIS_MODULE,
1018         .init =         ping_init_sock,
1019         .close =        ping_close,
1020         .connect =      ip4_datagram_connect,
1021         .disconnect =   __udp_disconnect,
1022         .setsockopt =   ip_setsockopt,
1023         .getsockopt =   ip_getsockopt,
1024         .sendmsg =      ping_v4_sendmsg,
1025         .recvmsg =      ping_recvmsg,
1026         .bind =         ping_bind,
1027         .backlog_rcv =  ping_queue_rcv_skb,
1028         .release_cb =   ip4_datagram_release_cb,
1029         .hash =         ping_hash,
1030         .unhash =       ping_unhash,
1031         .get_port =     ping_get_port,
1032         .obj_size =     sizeof(struct inet_sock),
1033 };
1034 EXPORT_SYMBOL(ping_prot);
1035
1036 #ifdef CONFIG_PROC_FS
1037
1038 static struct sock *ping_get_first(struct seq_file *seq, int start)
1039 {
1040         struct sock *sk;
1041         struct ping_iter_state *state = seq->private;
1042         struct net *net = seq_file_net(seq);
1043
1044         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1045              ++state->bucket) {
1046                 struct hlist_nulls_node *node;
1047                 struct hlist_nulls_head *hslot;
1048
1049                 hslot = &ping_table.hash[state->bucket];
1050
1051                 if (hlist_nulls_empty(hslot))
1052                         continue;
1053
1054                 sk_nulls_for_each(sk, node, hslot) {
1055                         if (net_eq(sock_net(sk), net) &&
1056                             sk->sk_family == state->family)
1057                                 goto found;
1058                 }
1059         }
1060         sk = NULL;
1061 found:
1062         return sk;
1063 }
1064
1065 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1066 {
1067         struct ping_iter_state *state = seq->private;
1068         struct net *net = seq_file_net(seq);
1069
1070         do {
1071                 sk = sk_nulls_next(sk);
1072         } while (sk && (!net_eq(sock_net(sk), net)));
1073
1074         if (!sk)
1075                 return ping_get_first(seq, state->bucket + 1);
1076         return sk;
1077 }
1078
1079 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1080 {
1081         struct sock *sk = ping_get_first(seq, 0);
1082
1083         if (sk)
1084                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1085                         --pos;
1086         return pos ? NULL : sk;
1087 }
1088
1089 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1090         __acquires(ping_table.lock)
1091 {
1092         struct ping_iter_state *state = seq->private;
1093         state->bucket = 0;
1094         state->family = family;
1095
1096         read_lock_bh(&ping_table.lock);
1097
1098         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1099 }
1100 EXPORT_SYMBOL_GPL(ping_seq_start);
1101
1102 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1103 {
1104         return ping_seq_start(seq, pos, AF_INET);
1105 }
1106
1107 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1108 {
1109         struct sock *sk;
1110
1111         if (v == SEQ_START_TOKEN)
1112                 sk = ping_get_idx(seq, 0);
1113         else
1114                 sk = ping_get_next(seq, v);
1115
1116         ++*pos;
1117         return sk;
1118 }
1119 EXPORT_SYMBOL_GPL(ping_seq_next);
1120
1121 void ping_seq_stop(struct seq_file *seq, void *v)
1122         __releases(ping_table.lock)
1123 {
1124         read_unlock_bh(&ping_table.lock);
1125 }
1126 EXPORT_SYMBOL_GPL(ping_seq_stop);
1127
1128 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1129                 int bucket)
1130 {
1131         struct inet_sock *inet = inet_sk(sp);
1132         __be32 dest = inet->inet_daddr;
1133         __be32 src = inet->inet_rcv_saddr;
1134         __u16 destp = ntohs(inet->inet_dport);
1135         __u16 srcp = ntohs(inet->inet_sport);
1136
1137         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1138                 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1139                 bucket, src, srcp, dest, destp, sp->sk_state,
1140                 sk_wmem_alloc_get(sp),
1141                 sk_rmem_alloc_get(sp),
1142                 0, 0L, 0,
1143                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1144                 0, sock_i_ino(sp),
1145                 refcount_read(&sp->sk_refcnt), sp,
1146                 atomic_read(&sp->sk_drops));
1147 }
1148
1149 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1150 {
1151         seq_setwidth(seq, 127);
1152         if (v == SEQ_START_TOKEN)
1153                 seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1154                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1155                            "inode ref pointer drops");
1156         else {
1157                 struct ping_iter_state *state = seq->private;
1158
1159                 ping_v4_format_sock(v, seq, state->bucket);
1160         }
1161         seq_pad(seq, '\n');
1162         return 0;
1163 }
1164
1165 static const struct seq_operations ping_v4_seq_ops = {
1166         .start          = ping_v4_seq_start,
1167         .show           = ping_v4_seq_show,
1168         .next           = ping_seq_next,
1169         .stop           = ping_seq_stop,
1170 };
1171
1172 static int __net_init ping_v4_proc_init_net(struct net *net)
1173 {
1174         if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1175                         sizeof(struct ping_iter_state)))
1176                 return -ENOMEM;
1177         return 0;
1178 }
1179
1180 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1181 {
1182         remove_proc_entry("icmp", net->proc_net);
1183 }
1184
1185 static struct pernet_operations ping_v4_net_ops = {
1186         .init = ping_v4_proc_init_net,
1187         .exit = ping_v4_proc_exit_net,
1188 };
1189
1190 int __init ping_proc_init(void)
1191 {
1192         return register_pernet_subsys(&ping_v4_net_ops);
1193 }
1194
1195 void ping_proc_exit(void)
1196 {
1197         unregister_pernet_subsys(&ping_v4_net_ops);
1198 }
1199
1200 #endif
1201
1202 void __init ping_init(void)
1203 {
1204         int i;
1205
1206         for (i = 0; i < PING_HTABLE_SIZE; i++)
1207                 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1208         rwlock_init(&ping_table.lock);
1209 }