net/ipv4/ping.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              "Ping" sockets
   8  *
   9  * Based on ipv4/udp.c code.
  10  *
  11  * Authors:     Vasiliy Kulikov / Openwall (for Linux 2.6),
  12  *              Pavel Kankovsky (for Linux 2.4.32)
  13  *
  14  * Pavel gave all rights to bugs to Vasiliy,
  15  * none of the bugs are Pavel's now.
  16  */
  17
  18 #include <linux/uaccess.h>
  19 #include <linux/types.h>
  20 #include <linux/fcntl.h>
  21 #include <linux/socket.h>
  22 #include <linux/sockios.h>
  23 #include <linux/in.h>
  24 #include <linux/errno.h>
  25 #include <linux/timer.h>
  26 #include <linux/mm.h>
  27 #include <linux/inet.h>
  28 #include <linux/netdevice.h>
  29 #include <net/snmp.h>
  30 #include <net/ip.h>
  31 #include <net/icmp.h>
  32 #include <net/protocol.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/proc_fs.h>
  35 #include <linux/export.h>
  36 #include <net/sock.h>
  37 #include <net/ping.h>
  38 #include <net/udp.h>
  39 #include <net/route.h>
  40 #include <net/inet_common.h>
  41 #include <net/checksum.h>
  42
  43 #if IS_ENABLED(CONFIG_IPV6)
  44 #include <linux/in6.h>
  45 #include <linux/icmpv6.h>
  46 #include <net/addrconf.h>
  47 #include <net/ipv6.h>
  48 #include <net/transp_v6.h>
  49 #endif
  50
  51 struct ping_table {
  52         struct hlist_nulls_head hash[PING_HTABLE_SIZE];
  53         rwlock_t                lock;
  54 };
  55
  56 static struct ping_table ping_table;
  57 struct pingv6_ops pingv6_ops;
  58 EXPORT_SYMBOL_GPL(pingv6_ops);
  59
  60 static u16 ping_port_rover;
  61
  62 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
  63 {
  64         u32 res = (num + net_hash_mix(net)) & mask;
  65
  66         pr_debug("hash(%u) = %u\n", num, res);
  67         return res;
  68 }
  69 EXPORT_SYMBOL_GPL(ping_hash);
  70
  71 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
  72                                              struct net *net, unsigned int num)
  73 {
  74         return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
  75 }
  76
  77 int ping_get_port(struct sock *sk, unsigned short ident)
  78 {
  79         struct hlist_nulls_node *node;
  80         struct hlist_nulls_head *hlist;
  81         struct inet_sock *isk, *isk2;
  82         struct sock *sk2 = NULL;
  83
  84         isk = inet_sk(sk);
  85         write_lock_bh(&ping_table.lock);
  86         if (ident == 0) {
  87                 u32 i;
  88                 u16 result = ping_port_rover + 1;
  89
  90                 for (i = 0; i < (1L << 16); i++, result++) {
  91                         if (!result)
  92                                 result++; /* avoid zero */
  93                         hlist = ping_hashslot(&ping_table, sock_net(sk),
  94                                             result);
  95                         ping_portaddr_for_each_entry(sk2, node, hlist) {
  96                                 isk2 = inet_sk(sk2);
  97
  98                                 if (isk2->inet_num == result)
  99                                         goto next_port;
 100                         }
 101
 102                         /* found */
 103                         ping_port_rover = ident = result;
 104                         break;
 105 next_port:
 106                         ;
 107                 }
 108                 if (i >= (1L << 16))
 109                         goto fail;
 110         } else {
 111                 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
 112                 ping_portaddr_for_each_entry(sk2, node, hlist) {
 113                         isk2 = inet_sk(sk2);
 114
 115                         /* BUG? Why is this reuse and not reuseaddr? ping.c
 116                          * doesn't turn off SO_REUSEADDR, and it doesn't expect
 117                          * that other ping processes can steal its packets.
 118                          */
 119                         if ((isk2->inet_num == ident) &&
 120                             (sk2 != sk) &&
 121                             (!sk2->sk_reuse || !sk->sk_reuse))
 122                                 goto fail;
 123                 }
 124         }
 125
 126         pr_debug("found port/ident = %d\n", ident);
 127         isk->inet_num = ident;
 128         if (sk_unhashed(sk)) {
 129                 pr_debug("was not hashed\n");
 130                 sock_hold(sk);
 131                 hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
 132                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
 133         }
 134         write_unlock_bh(&ping_table.lock);
 135         return 0;
 136
 137 fail:
 138         write_unlock_bh(&ping_table.lock);
 139         return 1;
 140 }
 141 EXPORT_SYMBOL_GPL(ping_get_port);
 142
 143 int ping_hash(struct sock *sk)
 144 {
 145         pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
 146         BUG(); /* "Please do not press this button again." */
 147
 148         return 0;
 149 }
 150
 151 void ping_unhash(struct sock *sk)
 152 {
 153         struct inet_sock *isk = inet_sk(sk);
 154
 155         pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
 156         write_lock_bh(&ping_table.lock);
 157         if (sk_hashed(sk)) {
 158                 hlist_nulls_del(&sk->sk_nulls_node);
 159                 sk_nulls_node_init(&sk->sk_nulls_node);
 160                 sock_put(sk);
 161                 isk->inet_num = 0;
 162                 isk->inet_sport = 0;
 163                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 164         }
 165         write_unlock_bh(&ping_table.lock);
 166 }
 167 EXPORT_SYMBOL_GPL(ping_unhash);
 168
 169 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
 170 {
 171         struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
 172         struct sock *sk = NULL;
 173         struct inet_sock *isk;
 174         struct hlist_nulls_node *hnode;
 175         int dif, sdif;
 176
 177         if (skb->protocol == htons(ETH_P_IP)) {
 178                 dif = inet_iif(skb);
 179                 sdif = inet_sdif(skb);
 180                 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
 181                          (int)ident, &ip_hdr(skb)->daddr, dif);
 182 #if IS_ENABLED(CONFIG_IPV6)
 183         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 184                 dif = inet6_iif(skb);
 185                 sdif = inet6_sdif(skb);
 186                 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
 187                          (int)ident, &ipv6_hdr(skb)->daddr, dif);
 188 #endif
 189         } else {
 190                 return NULL;
 191         }
 192
 193         read_lock_bh(&ping_table.lock);
 194
 195         ping_portaddr_for_each_entry(sk, hnode, hslot) {
 196                 isk = inet_sk(sk);
 197
 198                 pr_debug("iterate\n");
 199                 if (isk->inet_num != ident)
 200                         continue;
 201
 202                 if (skb->protocol == htons(ETH_P_IP) &&
 203                     sk->sk_family == AF_INET) {
 204                         pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
 205                                  (int) isk->inet_num, &isk->inet_rcv_saddr,
 206                                  sk->sk_bound_dev_if);
 207
 208                         if (isk->inet_rcv_saddr &&
 209                             isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
 210                                 continue;
 211 #if IS_ENABLED(CONFIG_IPV6)
 212                 } else if (skb->protocol == htons(ETH_P_IPV6) &&
 213                            sk->sk_family == AF_INET6) {
 214
 215                         pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
 216                                  (int) isk->inet_num,
 217                                  &sk->sk_v6_rcv_saddr,
 218                                  sk->sk_bound_dev_if);
 219
 220                         if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
 221                             !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
 222                                              &ipv6_hdr(skb)->daddr))
 223                                 continue;
 224 #endif
 225                 } else {
 226                         continue;
 227                 }
 228
 229                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
 230                     sk->sk_bound_dev_if != sdif)
 231                         continue;
 232
 233                 sock_hold(sk);
 234                 goto exit;
 235         }
 236
 237         sk = NULL;
 238 exit:
 239         read_unlock_bh(&ping_table.lock);
 240
 241         return sk;
 242 }
 243
 244 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
 245                                           kgid_t *high)
 246 {
 247         kgid_t *data = net->ipv4.ping_group_range.range;
 248         unsigned int seq;
 249
 250         do {
 251                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 252
 253                 *low = data[0];
 254                 *high = data[1];
 255         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 256 }
 257
 258
 259 int ping_init_sock(struct sock *sk)
 260 {
 261         struct net *net = sock_net(sk);
 262         kgid_t group = current_egid();
 263         struct group_info *group_info;
 264         int i;
 265         kgid_t low, high;
 266         int ret = 0;
 267
 268         if (sk->sk_family == AF_INET6)
 269                 sk->sk_ipv6only = 1;
 270
 271         inet_get_ping_group_range_net(net, &low, &high);
 272         if (gid_lte(low, group) && gid_lte(group, high))
 273                 return 0;
 274
 275         group_info = get_current_groups();
 276         for (i = 0; i < group_info->ngroups; i++) {
 277                 kgid_t gid = group_info->gid[i];
 278
 279                 if (gid_lte(low, gid) && gid_lte(gid, high))
 280                         goto out_release_group;
 281         }
 282
 283         ret = -EACCES;
 284
 285 out_release_group:
 286         put_group_info(group_info);
 287         return ret;
 288 }
 289 EXPORT_SYMBOL_GPL(ping_init_sock);
 290
 291 void ping_close(struct sock *sk, long timeout)
 292 {
 293         pr_debug("ping_close(sk=%p,sk->num=%u)\n",
 294                  inet_sk(sk), inet_sk(sk)->inet_num);
 295         pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
 296
 297         sk_common_release(sk);
 298 }
 299 EXPORT_SYMBOL_GPL(ping_close);
 300
 301 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
 302 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
 303                                 struct sockaddr *uaddr, int addr_len)
 304 {
 305         struct net *net = sock_net(sk);
 306         if (sk->sk_family == AF_INET) {
 307                 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 308                 u32 tb_id = RT_TABLE_LOCAL;
 309                 int chk_addr_ret;
 310
 311                 if (addr_len < sizeof(*addr))
 312                         return -EINVAL;
 313
 314                 if (addr->sin_family != AF_INET &&
 315                     !(addr->sin_family == AF_UNSPEC &&
 316                       addr->sin_addr.s_addr == htonl(INADDR_ANY)))
 317                         return -EAFNOSUPPORT;
 318
 319                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
 320                          sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
 321
 322                 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
 323                         chk_addr_ret = RTN_LOCAL;
 324                 else {
 325                         tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
 326                         chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
 327                 }
 328
 329                 if ((!inet_can_nonlocal_bind(net, isk) &&
 330                      chk_addr_ret != RTN_LOCAL) ||
 331                     chk_addr_ret == RTN_MULTICAST ||
 332                     chk_addr_ret == RTN_BROADCAST)
 333                         return -EADDRNOTAVAIL;
 334
 335 #if IS_ENABLED(CONFIG_IPV6)
 336         } else if (sk->sk_family == AF_INET6) {
 337                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 338                 int addr_type, scoped, has_addr;
 339                 struct net_device *dev = NULL;
 340
 341                 if (addr_len < sizeof(*addr))
 342                         return -EINVAL;
 343
 344                 if (addr->sin6_family != AF_INET6)
 345                         return -EAFNOSUPPORT;
 346
 347                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 348                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 349
 350                 addr_type = ipv6_addr_type(&addr->sin6_addr);
 351                 scoped = __ipv6_addr_needs_scope_id(addr_type);
 352                 if ((addr_type != IPV6_ADDR_ANY &&
 353                      !(addr_type & IPV6_ADDR_UNICAST)) ||
 354                     (scoped && !addr->sin6_scope_id))
 355                         return -EINVAL;
 356
 357                 rcu_read_lock();
 358                 if (addr->sin6_scope_id) {
 359                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 360                         if (!dev) {
 361                                 rcu_read_unlock();
 362                                 return -ENODEV;
 363                         }
 364                 }
 365
 366                 if (!dev && sk->sk_bound_dev_if) {
 367                         dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 368                         if (!dev) {
 369                                 rcu_read_unlock();
 370                                 return -ENODEV;
 371                         }
 372                 }
 373                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 374                                                     scoped);
 375                 rcu_read_unlock();
 376
 377                 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
 378                       addr_type == IPV6_ADDR_ANY))
 379                         return -EADDRNOTAVAIL;
 380
 381                 if (scoped)
 382                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 383 #endif
 384         } else {
 385                 return -EAFNOSUPPORT;
 386         }
 387         return 0;
 388 }
 389
 390 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
 391 {
 392         if (saddr->sa_family == AF_INET) {
 393                 struct inet_sock *isk = inet_sk(sk);
 394                 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
 395                 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
 396 #if IS_ENABLED(CONFIG_IPV6)
 397         } else if (saddr->sa_family == AF_INET6) {
 398                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
 399                 struct ipv6_pinfo *np = inet6_sk(sk);
 400                 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
 401 #endif
 402         }
 403 }
 404
 405 /*
 406  * We need our own bind because there are no privileged id's == local ports.
 407  * Moreover, we don't allow binding to multi- and broadcast addresses.
 408  */
 409
 410 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 411 {
 412         struct inet_sock *isk = inet_sk(sk);
 413         unsigned short snum;
 414         int err;
 415         int dif = sk->sk_bound_dev_if;
 416
 417         err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
 418         if (err)
 419                 return err;
 420
 421         lock_sock(sk);
 422
 423         err = -EINVAL;
 424         if (isk->inet_num != 0)
 425                 goto out;
 426
 427         err = -EADDRINUSE;
 428         snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
 429         if (ping_get_port(sk, snum) != 0) {
 430                 /* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
 431                 sk->sk_bound_dev_if = dif;
 432                 goto out;
 433         }
 434         ping_set_saddr(sk, uaddr);
 435
 436         pr_debug("after bind(): num = %hu, dif = %d\n",
 437                  isk->inet_num,
 438                  sk->sk_bound_dev_if);
 439
 440         err = 0;
 441         if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
 442                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 443 #if IS_ENABLED(CONFIG_IPV6)
 444         if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 445                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 446 #endif
 447
 448         if (snum)
 449                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 450         isk->inet_sport = htons(isk->inet_num);
 451         isk->inet_daddr = 0;
 452         isk->inet_dport = 0;
 453
 454 #if IS_ENABLED(CONFIG_IPV6)
 455         if (sk->sk_family == AF_INET6)
 456                 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
 457 #endif
 458
 459         sk_dst_reset(sk);
 460 out:
 461         release_sock(sk);
 462         pr_debug("ping_v4_bind -> %d\n", err);
 463         return err;
 464 }
 465 EXPORT_SYMBOL_GPL(ping_bind);
 466
 467 /*
 468  * Is this a supported type of ICMP message?
 469  */
 470
 471 static inline int ping_supported(int family, int type, int code)
 472 {
 473         return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
 474                (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
 475 }
 476
 477 /*
 478  * This routine is called by the ICMP module when it gets some
 479  * sort of error condition.
 480  */
 481
 482 void ping_err(struct sk_buff *skb, int offset, u32 info)
 483 {
 484         int family;
 485         struct icmphdr *icmph;
 486         struct inet_sock *inet_sock;
 487         int type;
 488         int code;
 489         struct net *net = dev_net(skb->dev);
 490         struct sock *sk;
 491         int harderr;
 492         int err;
 493
 494         if (skb->protocol == htons(ETH_P_IP)) {
 495                 family = AF_INET;
 496                 type = icmp_hdr(skb)->type;
 497                 code = icmp_hdr(skb)->code;
 498                 icmph = (struct icmphdr *)(skb->data + offset);
 499         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 500                 family = AF_INET6;
 501                 type = icmp6_hdr(skb)->icmp6_type;
 502                 code = icmp6_hdr(skb)->icmp6_code;
 503                 icmph = (struct icmphdr *) (skb->data + offset);
 504         } else {
 505                 BUG();
 506         }
 507
 508         /* We assume the packet has already been checked by icmp_unreach */
 509
 510         if (!ping_supported(family, icmph->type, icmph->code))
 511                 return;
 512
 513         pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
 514                  skb->protocol, type, code, ntohs(icmph->un.echo.id),
 515                  ntohs(icmph->un.echo.sequence));
 516
 517         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 518         if (!sk) {
 519                 pr_debug("no socket, dropping\n");
 520                 return; /* No socket for error */
 521         }
 522         pr_debug("err on socket %p\n", sk);
 523
 524         err = 0;
 525         harderr = 0;
 526         inet_sock = inet_sk(sk);
 527
 528         if (skb->protocol == htons(ETH_P_IP)) {
 529                 switch (type) {
 530                 default:
 531                 case ICMP_TIME_EXCEEDED:
 532                         err = EHOSTUNREACH;
 533                         break;
 534                 case ICMP_SOURCE_QUENCH:
 535                         /* This is not a real error but ping wants to see it.
 536                          * Report it with some fake errno.
 537                          */
 538                         err = EREMOTEIO;
 539                         break;
 540                 case ICMP_PARAMETERPROB:
 541                         err = EPROTO;
 542                         harderr = 1;
 543                         break;
 544                 case ICMP_DEST_UNREACH:
 545                         if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
 546                                 ipv4_sk_update_pmtu(skb, sk, info);
 547                                 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
 548                                         err = EMSGSIZE;
 549                                         harderr = 1;
 550                                         break;
 551                                 }
 552                                 goto out;
 553                         }
 554                         err = EHOSTUNREACH;
 555                         if (code <= NR_ICMP_UNREACH) {
 556                                 harderr = icmp_err_convert[code].fatal;
 557                                 err = icmp_err_convert[code].errno;
 558                         }
 559                         break;
 560                 case ICMP_REDIRECT:
 561                         /* See ICMP_SOURCE_QUENCH */
 562                         ipv4_sk_redirect(skb, sk);
 563                         err = EREMOTEIO;
 564                         break;
 565                 }
 566 #if IS_ENABLED(CONFIG_IPV6)
 567         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 568                 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
 569 #endif
 570         }
 571
 572         /*
 573          *      RFC1122: OK.  Passes ICMP errors back to application, as per
 574          *      4.1.3.3.
 575          */
 576         if ((family == AF_INET && !inet_sock->recverr) ||
 577             (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
 578                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 579                         goto out;
 580         } else {
 581                 if (family == AF_INET) {
 582                         ip_icmp_error(sk, skb, err, 0 /* no remote port */,
 583                                       info, (u8 *)icmph);
 584 #if IS_ENABLED(CONFIG_IPV6)
 585                 } else if (family == AF_INET6) {
 586                         pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
 587                                                    info, (u8 *)icmph);
 588 #endif
 589                 }
 590         }
 591         sk->sk_err = err;
 592         sk->sk_error_report(sk);
 593 out:
 594         sock_put(sk);
 595 }
 596 EXPORT_SYMBOL_GPL(ping_err);
 597
 598 /*
 599  *      Copy and checksum an ICMP Echo packet from user space into a buffer
 600  *      starting from the payload.
 601  */
 602
 603 int ping_getfrag(void *from, char *to,
 604                  int offset, int fraglen, int odd, struct sk_buff *skb)
 605 {
 606         struct pingfakehdr *pfh = (struct pingfakehdr *)from;
 607
 608         if (offset == 0) {
 609                 fraglen -= sizeof(struct icmphdr);
 610                 if (fraglen < 0)
 611                         BUG();
 612                 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
 613                             fraglen, &pfh->wcheck,
 614                             &pfh->msg->msg_iter))
 615                         return -EFAULT;
 616         } else if (offset < sizeof(struct icmphdr)) {
 617                         BUG();
 618         } else {
 619                 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
 620                                             &pfh->msg->msg_iter))
 621                         return -EFAULT;
 622         }
 623
 624 #if IS_ENABLED(CONFIG_IPV6)
 625         /* For IPv6, checksum each skb as we go along, as expected by
 626          * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
 627          * wcheck, it will be finalized in ping_v4_push_pending_frames.
 628          */
 629         if (pfh->family == AF_INET6) {
 630                 skb->csum = pfh->wcheck;
 631                 skb->ip_summed = CHECKSUM_NONE;
 632                 pfh->wcheck = 0;
 633         }
 634 #endif
 635
 636         return 0;
 637 }
 638 EXPORT_SYMBOL_GPL(ping_getfrag);
 639
 640 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
 641                                        struct flowi4 *fl4)
 642 {
 643         struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
 644
 645         if (!skb)
 646                 return 0;
 647         pfh->wcheck = csum_partial((char *)&pfh->icmph,
 648                 sizeof(struct icmphdr), pfh->wcheck);
 649         pfh->icmph.checksum = csum_fold(pfh->wcheck);
 650         memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
 651         skb->ip_summed = CHECKSUM_NONE;
 652         return ip_push_pending_frames(sk, fl4);
 653 }
 654
 655 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
 656                         void *user_icmph, size_t icmph_len)
 657 {
 658         u8 type, code;
 659
 660         if (len > 0xFFFF)
 661                 return -EMSGSIZE;
 662
 663         /* Must have at least a full ICMP header. */
 664         if (len < icmph_len)
 665                 return -EINVAL;
 666
 667         /*
 668          *      Check the flags.
 669          */
 670
 671         /* Mirror BSD error message compatibility */
 672         if (msg->msg_flags & MSG_OOB)
 673                 return -EOPNOTSUPP;
 674
 675         /*
 676          *      Fetch the ICMP header provided by the userland.
 677          *      iovec is modified! The ICMP header is consumed.
 678          */
 679         if (memcpy_from_msg(user_icmph, msg, icmph_len))
 680                 return -EFAULT;
 681
 682         if (family == AF_INET) {
 683                 type = ((struct icmphdr *) user_icmph)->type;
 684                 code = ((struct icmphdr *) user_icmph)->code;
 685 #if IS_ENABLED(CONFIG_IPV6)
 686         } else if (family == AF_INET6) {
 687                 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
 688                 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
 689 #endif
 690         } else {
 691                 BUG();
 692         }
 693
 694         if (!ping_supported(family, type, code))
 695                 return -EINVAL;
 696
 697         return 0;
 698 }
 699 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
 700
 701 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 702 {
 703         struct net *net = sock_net(sk);
 704         struct flowi4 fl4;
 705         struct inet_sock *inet = inet_sk(sk);
 706         struct ipcm_cookie ipc;
 707         struct icmphdr user_icmph;
 708         struct pingfakehdr pfh;
 709         struct rtable *rt = NULL;
 710         struct ip_options_data opt_copy;
 711         int free = 0;
 712         __be32 saddr, daddr, faddr;
 713         u8  tos;
 714         int err;
 715
 716         pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
 717
 718         err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
 719                                   sizeof(user_icmph));
 720         if (err)
 721                 return err;
 722
 723         /*
 724          *      Get and verify the address.
 725          */
 726
 727         if (msg->msg_name) {
 728                 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
 729                 if (msg->msg_namelen < sizeof(*usin))
 730                         return -EINVAL;
 731                 if (usin->sin_family != AF_INET)
 732                         return -EAFNOSUPPORT;
 733                 daddr = usin->sin_addr.s_addr;
 734                 /* no remote port */
 735         } else {
 736                 if (sk->sk_state != TCP_ESTABLISHED)
 737                         return -EDESTADDRREQ;
 738                 daddr = inet->inet_daddr;
 739                 /* no remote port */
 740         }
 741
 742         ipcm_init_sk(&ipc, inet);
 743
 744         if (msg->msg_controllen) {
 745                 err = ip_cmsg_send(sk, msg, &ipc, false);
 746                 if (unlikely(err)) {
 747                         kfree(ipc.opt);
 748                         return err;
 749                 }
 750                 if (ipc.opt)
 751                         free = 1;
 752         }
 753         if (!ipc.opt) {
 754                 struct ip_options_rcu *inet_opt;
 755
 756                 rcu_read_lock();
 757                 inet_opt = rcu_dereference(inet->inet_opt);
 758                 if (inet_opt) {
 759                         memcpy(&opt_copy, inet_opt,
 760                                sizeof(*inet_opt) + inet_opt->opt.optlen);
 761                         ipc.opt = &opt_copy.opt;
 762                 }
 763                 rcu_read_unlock();
 764         }
 765
 766         saddr = ipc.addr;
 767         ipc.addr = faddr = daddr;
 768
 769         if (ipc.opt && ipc.opt->opt.srr) {
 770                 if (!daddr) {
 771                         err = -EINVAL;
 772                         goto out_free;
 773                 }
 774                 faddr = ipc.opt->opt.faddr;
 775         }
 776         tos = get_rttos(&ipc, inet);
 777         if (sock_flag(sk, SOCK_LOCALROUTE) ||
 778             (msg->msg_flags & MSG_DONTROUTE) ||
 779             (ipc.opt && ipc.opt->opt.is_strictroute)) {
 780                 tos |= RTO_ONLINK;
 781         }
 782
 783         if (ipv4_is_multicast(daddr)) {
 784                 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
 785                         ipc.oif = inet->mc_index;
 786                 if (!saddr)
 787                         saddr = inet->mc_addr;
 788         } else if (!ipc.oif)
 789                 ipc.oif = inet->uc_index;
 790
 791         flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
 792                            RT_SCOPE_UNIVERSE, sk->sk_protocol,
 793                            inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
 794                            sk->sk_uid);
 795
 796         fl4.fl4_icmp_type = user_icmph.type;
 797         fl4.fl4_icmp_code = user_icmph.code;
 798
 799         security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
 800         rt = ip_route_output_flow(net, &fl4, sk);
 801         if (IS_ERR(rt)) {
 802                 err = PTR_ERR(rt);
 803                 rt = NULL;
 804                 if (err == -ENETUNREACH)
 805                         IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
 806                 goto out;
 807         }
 808
 809         err = -EACCES;
 810         if ((rt->rt_flags & RTCF_BROADCAST) &&
 811             !sock_flag(sk, SOCK_BROADCAST))
 812                 goto out;
 813
 814         if (msg->msg_flags & MSG_CONFIRM)
 815                 goto do_confirm;
 816 back_from_confirm:
 817
 818         if (!ipc.addr)
 819                 ipc.addr = fl4.daddr;
 820
 821         lock_sock(sk);
 822
 823         pfh.icmph.type = user_icmph.type; /* already checked */
 824         pfh.icmph.code = user_icmph.code; /* ditto */
 825         pfh.icmph.checksum = 0;
 826         pfh.icmph.un.echo.id = inet->inet_sport;
 827         pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 828         pfh.msg = msg;
 829         pfh.wcheck = 0;
 830         pfh.family = AF_INET;
 831
 832         err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 833                         0, &ipc, &rt, msg->msg_flags);
 834         if (err)
 835                 ip_flush_pending_frames(sk);
 836         else
 837                 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 838         release_sock(sk);
 839
 840 out:
 841         ip_rt_put(rt);
 842 out_free:
 843         if (free)
 844                 kfree(ipc.opt);
 845         if (!err) {
 846                 icmp_out_count(sock_net(sk), user_icmph.type);
 847                 return len;
 848         }
 849         return err;
 850
 851 do_confirm:
 852         if (msg->msg_flags & MSG_PROBE)
 853                 dst_confirm_neigh(&rt->dst, &fl4.daddr);
 854         if (!(msg->msg_flags & MSG_PROBE) || len)
 855                 goto back_from_confirm;
 856         err = 0;
 857         goto out;
 858 }
 859
 860 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
 861                  int flags, int *addr_len)
 862 {
 863         struct inet_sock *isk = inet_sk(sk);
 864         int family = sk->sk_family;
 865         struct sk_buff *skb;
 866         int copied, err;
 867
 868         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 869
 870         err = -EOPNOTSUPP;
 871         if (flags & MSG_OOB)
 872                 goto out;
 873
 874         if (flags & MSG_ERRQUEUE)
 875                 return inet_recv_error(sk, msg, len, addr_len);
 876
 877         skb = skb_recv_datagram(sk, flags, noblock, &err);
 878         if (!skb)
 879                 goto out;
 880
 881         copied = skb->len;
 882         if (copied > len) {
 883                 msg->msg_flags |= MSG_TRUNC;
 884                 copied = len;
 885         }
 886
 887         /* Don't bother checking the checksum */
 888         err = skb_copy_datagram_msg(skb, 0, msg, copied);
 889         if (err)
 890                 goto done;
 891
 892         sock_recv_timestamp(msg, sk, skb);
 893
 894         /* Copy the address and add cmsg data. */
 895         if (family == AF_INET) {
 896                 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 897
 898                 if (sin) {
 899                         sin->sin_family = AF_INET;
 900                         sin->sin_port = 0 /* skb->h.uh->source */;
 901                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 902                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 903                         *addr_len = sizeof(*sin);
 904                 }
 905
 906                 if (isk->cmsg_flags)
 907                         ip_cmsg_recv(msg, skb);
 908
 909 #if IS_ENABLED(CONFIG_IPV6)
 910         } else if (family == AF_INET6) {
 911                 struct ipv6_pinfo *np = inet6_sk(sk);
 912                 struct ipv6hdr *ip6 = ipv6_hdr(skb);
 913                 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 914
 915                 if (sin6) {
 916                         sin6->sin6_family = AF_INET6;
 917                         sin6->sin6_port = 0;
 918                         sin6->sin6_addr = ip6->saddr;
 919                         sin6->sin6_flowinfo = 0;
 920                         if (np->sndflow)
 921                                 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
 922                         sin6->sin6_scope_id =
 923                                 ipv6_iface_scope_id(&sin6->sin6_addr,
 924                                                     inet6_iif(skb));
 925                         *addr_len = sizeof(*sin6);
 926                 }
 927
 928                 if (inet6_sk(sk)->rxopt.all)
 929                         pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
 930                 if (skb->protocol == htons(ETH_P_IPV6) &&
 931                     inet6_sk(sk)->rxopt.all)
 932                         pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
 933                 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
 934                         ip_cmsg_recv(msg, skb);
 935 #endif
 936         } else {
 937                 BUG();
 938         }
 939
 940         err = copied;
 941
 942 done:
 943         skb_free_datagram(sk, skb);
 944 out:
 945         pr_debug("ping_recvmsg -> %d\n", err);
 946         return err;
 947 }
 948 EXPORT_SYMBOL_GPL(ping_recvmsg);
 949
 950 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 951 {
 952         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 953                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
 954         if (sock_queue_rcv_skb(sk, skb) < 0) {
 955                 kfree_skb(skb);
 956                 pr_debug("ping_queue_rcv_skb -> failed\n");
 957                 return -1;
 958         }
 959         return 0;
 960 }
 961 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 962
 963
 964 /*
 965  *      All we need to do is get the socket.
 966  */
 967
 968 bool ping_rcv(struct sk_buff *skb)
 969 {
 970         struct sock *sk;
 971         struct net *net = dev_net(skb->dev);
 972         struct icmphdr *icmph = icmp_hdr(skb);
 973         bool rc = false;
 974
 975         /* We assume the packet has already been checked by icmp_rcv */
 976
 977         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 978                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 979
 980         /* Push ICMP header back */
 981         skb_push(skb, skb->data - (u8 *)icmph);
 982
 983         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 984         if (sk) {
 985                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 986
 987                 pr_debug("rcv on socket %p\n", sk);
 988                 if (skb2 && !ping_queue_rcv_skb(sk, skb2))
 989                         rc = true;
 990                 sock_put(sk);
 991         }
 992
 993         if (!rc)
 994                 pr_debug("no socket, dropping\n");
 995
 996         return rc;
 997 }
 998 EXPORT_SYMBOL_GPL(ping_rcv);
 999
1000 struct proto ping_prot = {
1001         .name =         "PING",
1002         .owner =        THIS_MODULE,
1003         .init =         ping_init_sock,
1004         .close =        ping_close,
1005         .connect =      ip4_datagram_connect,
1006         .disconnect =   __udp_disconnect,
1007         .setsockopt =   ip_setsockopt,
1008         .getsockopt =   ip_getsockopt,
1009         .sendmsg =      ping_v4_sendmsg,
1010         .recvmsg =      ping_recvmsg,
1011         .bind =         ping_bind,
1012         .backlog_rcv =  ping_queue_rcv_skb,
1013         .release_cb =   ip4_datagram_release_cb,
1014         .hash =         ping_hash,
1015         .unhash =       ping_unhash,
1016         .get_port =     ping_get_port,
1017         .obj_size =     sizeof(struct inet_sock),
1018 };
1019 EXPORT_SYMBOL(ping_prot);
1020
1021 #ifdef CONFIG_PROC_FS
1022
1023 static struct sock *ping_get_first(struct seq_file *seq, int start)
1024 {
1025         struct sock *sk;
1026         struct ping_iter_state *state = seq->private;
1027         struct net *net = seq_file_net(seq);
1028
1029         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1030              ++state->bucket) {
1031                 struct hlist_nulls_node *node;
1032                 struct hlist_nulls_head *hslot;
1033
1034                 hslot = &ping_table.hash[state->bucket];
1035
1036                 if (hlist_nulls_empty(hslot))
1037                         continue;
1038
1039                 sk_nulls_for_each(sk, node, hslot) {
1040                         if (net_eq(sock_net(sk), net) &&
1041                             sk->sk_family == state->family)
1042                                 goto found;
1043                 }
1044         }
1045         sk = NULL;
1046 found:
1047         return sk;
1048 }
1049
1050 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1051 {
1052         struct ping_iter_state *state = seq->private;
1053         struct net *net = seq_file_net(seq);
1054
1055         do {
1056                 sk = sk_nulls_next(sk);
1057         } while (sk && (!net_eq(sock_net(sk), net)));
1058
1059         if (!sk)
1060                 return ping_get_first(seq, state->bucket + 1);
1061         return sk;
1062 }
1063
1064 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1065 {
1066         struct sock *sk = ping_get_first(seq, 0);
1067
1068         if (sk)
1069                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1070                         --pos;
1071         return pos ? NULL : sk;
1072 }
1073
1074 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1075         __acquires(ping_table.lock)
1076 {
1077         struct ping_iter_state *state = seq->private;
1078         state->bucket = 0;
1079         state->family = family;
1080
1081         read_lock_bh(&ping_table.lock);
1082
1083         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1084 }
1085 EXPORT_SYMBOL_GPL(ping_seq_start);
1086
1087 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1088 {
1089         return ping_seq_start(seq, pos, AF_INET);
1090 }
1091
1092 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1093 {
1094         struct sock *sk;
1095
1096         if (v == SEQ_START_TOKEN)
1097                 sk = ping_get_idx(seq, 0);
1098         else
1099                 sk = ping_get_next(seq, v);
1100
1101         ++*pos;
1102         return sk;
1103 }
1104 EXPORT_SYMBOL_GPL(ping_seq_next);
1105
1106 void ping_seq_stop(struct seq_file *seq, void *v)
1107         __releases(ping_table.lock)
1108 {
1109         read_unlock_bh(&ping_table.lock);
1110 }
1111 EXPORT_SYMBOL_GPL(ping_seq_stop);
1112
1113 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1114                 int bucket)
1115 {
1116         struct inet_sock *inet = inet_sk(sp);
1117         __be32 dest = inet->inet_daddr;
1118         __be32 src = inet->inet_rcv_saddr;
1119         __u16 destp = ntohs(inet->inet_dport);
1120         __u16 srcp = ntohs(inet->inet_sport);
1121
1122         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1123                 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1124                 bucket, src, srcp, dest, destp, sp->sk_state,
1125                 sk_wmem_alloc_get(sp),
1126                 sk_rmem_alloc_get(sp),
1127                 0, 0L, 0,
1128                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1129                 0, sock_i_ino(sp),
1130                 refcount_read(&sp->sk_refcnt), sp,
1131                 atomic_read(&sp->sk_drops));
1132 }
1133
1134 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1135 {
1136         seq_setwidth(seq, 127);
1137         if (v == SEQ_START_TOKEN)
1138                 seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1139                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1140                            "inode ref pointer drops");
1141         else {
1142                 struct ping_iter_state *state = seq->private;
1143
1144                 ping_v4_format_sock(v, seq, state->bucket);
1145         }
1146         seq_pad(seq, '\n');
1147         return 0;
1148 }
1149
1150 static const struct seq_operations ping_v4_seq_ops = {
1151         .start          = ping_v4_seq_start,
1152         .show           = ping_v4_seq_show,
1153         .next           = ping_seq_next,
1154         .stop           = ping_seq_stop,
1155 };
1156
1157 static int __net_init ping_v4_proc_init_net(struct net *net)
1158 {
1159         if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1160                         sizeof(struct ping_iter_state)))
1161                 return -ENOMEM;
1162         return 0;
1163 }
1164
1165 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1166 {
1167         remove_proc_entry("icmp", net->proc_net);
1168 }
1169
1170 static struct pernet_operations ping_v4_net_ops = {
1171         .init = ping_v4_proc_init_net,
1172         .exit = ping_v4_proc_exit_net,
1173 };
1174
1175 int __init ping_proc_init(void)
1176 {
1177         return register_pernet_subsys(&ping_v4_net_ops);
1178 }
1179
1180 void ping_proc_exit(void)
1181 {
1182         unregister_pernet_subsys(&ping_v4_net_ops);
1183 }
1184
1185 #endif
1186
1187 void __init ping_init(void)
1188 {
1189         int i;
1190
1191         for (i = 0; i < PING_HTABLE_SIZE; i++)
1192                 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1193         rwlock_init(&ping_table.lock);
1194 }