net/ipv4/ping.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              "Ping" sockets
   7  *
   8  *              This program is free software; you can redistribute it and/or
   9  *              modify it under the terms of the GNU General Public License
  10  *              as published by the Free Software Foundation; either version
  11  *              2 of the License, or (at your option) any later version.
  12  *
  13  * Based on ipv4/udp.c code.
  14  *
  15  * Authors:     Vasiliy Kulikov / Openwall (for Linux 2.6),
  16  *              Pavel Kankovsky (for Linux 2.4.32)
  17  *
  18  * Pavel gave all rights to bugs to Vasiliy,
  19  * none of the bugs are Pavel's now.
  20  *
  21  */
  22
  23 #include <linux/uaccess.h>
  24 #include <linux/types.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/socket.h>
  27 #include <linux/sockios.h>
  28 #include <linux/in.h>
  29 #include <linux/errno.h>
  30 #include <linux/timer.h>
  31 #include <linux/mm.h>
  32 #include <linux/inet.h>
  33 #include <linux/netdevice.h>
  34 #include <net/snmp.h>
  35 #include <net/ip.h>
  36 #include <net/icmp.h>
  37 #include <net/protocol.h>
  38 #include <linux/skbuff.h>
  39 #include <linux/proc_fs.h>
  40 #include <linux/export.h>
  41 #include <net/sock.h>
  42 #include <net/ping.h>
  43 #include <net/udp.h>
  44 #include <net/route.h>
  45 #include <net/inet_common.h>
  46 #include <net/checksum.h>
  47
  48 #if IS_ENABLED(CONFIG_IPV6)
  49 #include <linux/in6.h>
  50 #include <linux/icmpv6.h>
  51 #include <net/addrconf.h>
  52 #include <net/ipv6.h>
  53 #include <net/transp_v6.h>
  54 #endif
  55
  56
  57 struct ping_table ping_table;
  58 struct pingv6_ops pingv6_ops;
  59 EXPORT_SYMBOL_GPL(pingv6_ops);
  60
  61 static u16 ping_port_rover;
  62
  63 static inline int ping_hashfn(struct net *net, unsigned int num, unsigned int mask)
  64 {
  65         int res = (num + net_hash_mix(net)) & mask;
  66
  67         pr_debug("hash(%d) = %d\n", num, res);
  68         return res;
  69 }
  70 EXPORT_SYMBOL_GPL(ping_hash);
  71
  72 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
  73                                              struct net *net, unsigned int num)
  74 {
  75         return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
  76 }
  77
  78 int ping_get_port(struct sock *sk, unsigned short ident)
  79 {
  80         struct hlist_nulls_node *node;
  81         struct hlist_nulls_head *hlist;
  82         struct inet_sock *isk, *isk2;
  83         struct sock *sk2 = NULL;
  84
  85         isk = inet_sk(sk);
  86         write_lock_bh(&ping_table.lock);
  87         if (ident == 0) {
  88                 u32 i;
  89                 u16 result = ping_port_rover + 1;
  90
  91                 for (i = 0; i < (1L << 16); i++, result++) {
  92                         if (!result)
  93                                 result++; /* avoid zero */
  94                         hlist = ping_hashslot(&ping_table, sock_net(sk),
  95                                             result);
  96                         ping_portaddr_for_each_entry(sk2, node, hlist) {
  97                                 isk2 = inet_sk(sk2);
  98
  99                                 if (isk2->inet_num == result)
 100                                         goto next_port;
 101                         }
 102
 103                         /* found */
 104                         ping_port_rover = ident = result;
 105                         break;
 106 next_port:
 107                         ;
 108                 }
 109                 if (i >= (1L << 16))
 110                         goto fail;
 111         } else {
 112                 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
 113                 ping_portaddr_for_each_entry(sk2, node, hlist) {
 114                         isk2 = inet_sk(sk2);
 115
 116                         /* BUG? Why is this reuse and not reuseaddr? ping.c
 117                          * doesn't turn off SO_REUSEADDR, and it doesn't expect
 118                          * that other ping processes can steal its packets.
 119                          */
 120                         if ((isk2->inet_num == ident) &&
 121                             (sk2 != sk) &&
 122                             (!sk2->sk_reuse || !sk->sk_reuse))
 123                                 goto fail;
 124                 }
 125         }
 126
 127         pr_debug("found port/ident = %d\n", ident);
 128         isk->inet_num = ident;
 129         if (sk_unhashed(sk)) {
 130                 pr_debug("was not hashed\n");
 131                 sock_hold(sk);
 132                 hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
 133                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
 134         }
 135         write_unlock_bh(&ping_table.lock);
 136         return 0;
 137
 138 fail:
 139         write_unlock_bh(&ping_table.lock);
 140         return 1;
 141 }
 142 EXPORT_SYMBOL_GPL(ping_get_port);
 143
 144 void ping_hash(struct sock *sk)
 145 {
 146         pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
 147         BUG(); /* "Please do not press this button again." */
 148 }
 149
 150 void ping_unhash(struct sock *sk)
 151 {
 152         struct inet_sock *isk = inet_sk(sk);
 153         pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
 154         if (sk_hashed(sk)) {
 155                 write_lock_bh(&ping_table.lock);
 156                 hlist_nulls_del(&sk->sk_nulls_node);
 157                 sock_put(sk);
 158                 isk->inet_num = 0;
 159                 isk->inet_sport = 0;
 160                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 161                 write_unlock_bh(&ping_table.lock);
 162         }
 163 }
 164 EXPORT_SYMBOL_GPL(ping_unhash);
 165
 166 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
 167 {
 168         struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
 169         struct sock *sk = NULL;
 170         struct inet_sock *isk;
 171         struct hlist_nulls_node *hnode;
 172         int dif = skb->dev->ifindex;
 173
 174         if (skb->protocol == htons(ETH_P_IP)) {
 175                 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
 176                          (int)ident, &ip_hdr(skb)->daddr, dif);
 177 #if IS_ENABLED(CONFIG_IPV6)
 178         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 179                 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
 180                          (int)ident, &ipv6_hdr(skb)->daddr, dif);
 181 #endif
 182         }
 183
 184         read_lock_bh(&ping_table.lock);
 185
 186         ping_portaddr_for_each_entry(sk, hnode, hslot) {
 187                 isk = inet_sk(sk);
 188
 189                 pr_debug("iterate\n");
 190                 if (isk->inet_num != ident)
 191                         continue;
 192
 193                 if (skb->protocol == htons(ETH_P_IP) &&
 194                     sk->sk_family == AF_INET) {
 195                         pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
 196                                  (int) isk->inet_num, &isk->inet_rcv_saddr,
 197                                  sk->sk_bound_dev_if);
 198
 199                         if (isk->inet_rcv_saddr &&
 200                             isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
 201                                 continue;
 202 #if IS_ENABLED(CONFIG_IPV6)
 203                 } else if (skb->protocol == htons(ETH_P_IPV6) &&
 204                            sk->sk_family == AF_INET6) {
 205                         struct ipv6_pinfo *np = inet6_sk(sk);
 206
 207                         pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
 208                                  (int) isk->inet_num,
 209                                  &inet6_sk(sk)->rcv_saddr,
 210                                  sk->sk_bound_dev_if);
 211
 212                         if (!ipv6_addr_any(&np->rcv_saddr) &&
 213                             !ipv6_addr_equal(&np->rcv_saddr,
 214                                              &ipv6_hdr(skb)->daddr))
 215                                 continue;
 216 #endif
 217                 } else {
 218                         continue;
 219                 }
 220
 221                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
 222                         continue;
 223
 224                 sock_hold(sk);
 225                 goto exit;
 226         }
 227
 228         sk = NULL;
 229 exit:
 230         read_unlock_bh(&ping_table.lock);
 231
 232         return sk;
 233 }
 234
 235 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
 236                                           kgid_t *high)
 237 {
 238         kgid_t *data = net->ipv4.sysctl_ping_group_range;
 239         unsigned int seq;
 240
 241         do {
 242                 seq = read_seqbegin(&sysctl_local_ports.lock);
 243
 244                 *low = data[0];
 245                 *high = data[1];
 246         } while (read_seqretry(&sysctl_local_ports.lock, seq));
 247 }
 248
 249
 250 int ping_init_sock(struct sock *sk)
 251 {
 252         struct net *net = sock_net(sk);
 253         kgid_t group = current_egid();
 254         struct group_info *group_info;
 255         int i, j, count;
 256         kgid_t low, high;
 257         int ret = 0;
 258
 259         if (sk->sk_family == AF_INET6)
 260                 inet6_sk(sk)->ipv6only = 1;
 261
 262         inet_get_ping_group_range_net(net, &low, &high);
 263         if (gid_lte(low, group) && gid_lte(group, high))
 264                 return 0;
 265
 266         group_info = get_current_groups();
 267         count = group_info->ngroups;
 268         for (i = 0; i < group_info->nblocks; i++) {
 269                 int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
 270                 for (j = 0; j < cp_count; j++) {
 271                         kgid_t gid = group_info->blocks[i][j];
 272                         if (gid_lte(low, gid) && gid_lte(gid, high))
 273                                 goto out_release_group;
 274                 }
 275
 276                 count -= cp_count;
 277         }
 278
 279         ret = -EACCES;
 280
 281 out_release_group:
 282         put_group_info(group_info);
 283         return ret;
 284 }
 285 EXPORT_SYMBOL_GPL(ping_init_sock);
 286
 287 void ping_close(struct sock *sk, long timeout)
 288 {
 289         pr_debug("ping_close(sk=%p,sk->num=%u)\n",
 290                  inet_sk(sk), inet_sk(sk)->inet_num);
 291         pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
 292
 293         sk_common_release(sk);
 294 }
 295 EXPORT_SYMBOL_GPL(ping_close);
 296
 297 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
 298 int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
 299                          struct sockaddr *uaddr, int addr_len) {
 300         struct net *net = sock_net(sk);
 301         if (sk->sk_family == AF_INET) {
 302                 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 303                 int chk_addr_ret;
 304
 305                 if (addr_len < sizeof(*addr))
 306                         return -EINVAL;
 307
 308                 if (addr->sin_family != AF_INET &&
 309                     !(addr->sin_family == AF_UNSPEC &&
 310                       addr->sin_addr.s_addr == htonl(INADDR_ANY)))
 311                         return -EAFNOSUPPORT;
 312
 313                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
 314                          sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
 315
 316                 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
 317
 318                 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
 319                         chk_addr_ret = RTN_LOCAL;
 320
 321                 if ((sysctl_ip_nonlocal_bind == 0 &&
 322                     isk->freebind == 0 && isk->transparent == 0 &&
 323                      chk_addr_ret != RTN_LOCAL) ||
 324                     chk_addr_ret == RTN_MULTICAST ||
 325                     chk_addr_ret == RTN_BROADCAST)
 326                         return -EADDRNOTAVAIL;
 327
 328 #if IS_ENABLED(CONFIG_IPV6)
 329         } else if (sk->sk_family == AF_INET6) {
 330                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 331                 int addr_type, scoped, has_addr;
 332                 struct net_device *dev = NULL;
 333
 334                 if (addr_len < sizeof(*addr))
 335                         return -EINVAL;
 336
 337                 if (addr->sin6_family != AF_INET6)
 338                         return -EAFNOSUPPORT;
 339
 340                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 341                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 342
 343                 addr_type = ipv6_addr_type(&addr->sin6_addr);
 344                 scoped = __ipv6_addr_needs_scope_id(addr_type);
 345                 if ((addr_type != IPV6_ADDR_ANY &&
 346                      !(addr_type & IPV6_ADDR_UNICAST)) ||
 347                     (scoped && !addr->sin6_scope_id))
 348                         return -EINVAL;
 349
 350                 rcu_read_lock();
 351                 if (addr->sin6_scope_id) {
 352                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 353                         if (!dev) {
 354                                 rcu_read_unlock();
 355                                 return -ENODEV;
 356                         }
 357                 }
 358                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 359                                                     scoped);
 360                 rcu_read_unlock();
 361
 362                 if (!(isk->freebind || isk->transparent || has_addr ||
 363                       addr_type == IPV6_ADDR_ANY))
 364                         return -EADDRNOTAVAIL;
 365
 366                 if (scoped)
 367                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 368 #endif
 369         } else {
 370                 return -EAFNOSUPPORT;
 371         }
 372         return 0;
 373 }
 374
 375 void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
 376 {
 377         if (saddr->sa_family == AF_INET) {
 378                 struct inet_sock *isk = inet_sk(sk);
 379                 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
 380                 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
 381 #if IS_ENABLED(CONFIG_IPV6)
 382         } else if (saddr->sa_family == AF_INET6) {
 383                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
 384                 struct ipv6_pinfo *np = inet6_sk(sk);
 385                 np->rcv_saddr = np->saddr = addr->sin6_addr;
 386 #endif
 387         }
 388 }
 389
 390 void ping_clear_saddr(struct sock *sk, int dif)
 391 {
 392         sk->sk_bound_dev_if = dif;
 393         if (sk->sk_family == AF_INET) {
 394                 struct inet_sock *isk = inet_sk(sk);
 395                 isk->inet_rcv_saddr = isk->inet_saddr = 0;
 396 #if IS_ENABLED(CONFIG_IPV6)
 397         } else if (sk->sk_family == AF_INET6) {
 398                 struct ipv6_pinfo *np = inet6_sk(sk);
 399                 memset(&np->rcv_saddr, 0, sizeof(np->rcv_saddr));
 400                 memset(&np->saddr, 0, sizeof(np->saddr));
 401 #endif
 402         }
 403 }
 404 /*
 405  * We need our own bind because there are no privileged id's == local ports.
 406  * Moreover, we don't allow binding to multi- and broadcast addresses.
 407  */
 408
 409 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 410 {
 411         struct inet_sock *isk = inet_sk(sk);
 412         unsigned short snum;
 413         int err;
 414         int dif = sk->sk_bound_dev_if;
 415
 416         err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
 417         if (err)
 418                 return err;
 419
 420         lock_sock(sk);
 421
 422         err = -EINVAL;
 423         if (isk->inet_num != 0)
 424                 goto out;
 425
 426         err = -EADDRINUSE;
 427         ping_set_saddr(sk, uaddr);
 428         snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
 429         if (ping_get_port(sk, snum) != 0) {
 430                 ping_clear_saddr(sk, dif);
 431                 goto out;
 432         }
 433
 434         pr_debug("after bind(): num = %d, dif = %d\n",
 435                  (int)isk->inet_num,
 436                  (int)sk->sk_bound_dev_if);
 437
 438         err = 0;
 439         if ((sk->sk_family == AF_INET && isk->inet_rcv_saddr) ||
 440             (sk->sk_family == AF_INET6 &&
 441              !ipv6_addr_any(&inet6_sk(sk)->rcv_saddr)))
 442                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 443
 444         if (snum)
 445                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 446         isk->inet_sport = htons(isk->inet_num);
 447         isk->inet_daddr = 0;
 448         isk->inet_dport = 0;
 449
 450 #if IS_ENABLED(CONFIG_IPV6)
 451         if (sk->sk_family == AF_INET6)
 452                 memset(&inet6_sk(sk)->daddr, 0, sizeof(inet6_sk(sk)->daddr));
 453 #endif
 454
 455         sk_dst_reset(sk);
 456 out:
 457         release_sock(sk);
 458         pr_debug("ping_v4_bind -> %d\n", err);
 459         return err;
 460 }
 461 EXPORT_SYMBOL_GPL(ping_bind);
 462
 463 /*
 464  * Is this a supported type of ICMP message?
 465  */
 466
 467 static inline int ping_supported(int family, int type, int code)
 468 {
 469         return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
 470                (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
 471 }
 472
 473 /*
 474  * This routine is called by the ICMP module when it gets some
 475  * sort of error condition.
 476  */
 477
 478 void ping_err(struct sk_buff *skb, int offset, u32 info)
 479 {
 480         int family;
 481         struct icmphdr *icmph;
 482         struct inet_sock *inet_sock;
 483         int type;
 484         int code;
 485         struct net *net = dev_net(skb->dev);
 486         struct sock *sk;
 487         int harderr;
 488         int err;
 489
 490         if (skb->protocol == htons(ETH_P_IP)) {
 491                 family = AF_INET;
 492                 type = icmp_hdr(skb)->type;
 493                 code = icmp_hdr(skb)->code;
 494                 icmph = (struct icmphdr *)(skb->data + offset);
 495         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 496                 family = AF_INET6;
 497                 type = icmp6_hdr(skb)->icmp6_type;
 498                 code = icmp6_hdr(skb)->icmp6_code;
 499                 icmph = (struct icmphdr *) (skb->data + offset);
 500         } else {
 501                 BUG();
 502         }
 503
 504         /* We assume the packet has already been checked by icmp_unreach */
 505
 506         if (!ping_supported(family, icmph->type, icmph->code))
 507                 return;
 508
 509         pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
 510                  skb->protocol, type, code, ntohs(icmph->un.echo.id),
 511                  ntohs(icmph->un.echo.sequence));
 512
 513         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 514         if (sk == NULL) {
 515                 pr_debug("no socket, dropping\n");
 516                 return; /* No socket for error */
 517         }
 518         pr_debug("err on socket %p\n", sk);
 519
 520         err = 0;
 521         harderr = 0;
 522         inet_sock = inet_sk(sk);
 523
 524         if (skb->protocol == htons(ETH_P_IP)) {
 525                 switch (type) {
 526                 default:
 527                 case ICMP_TIME_EXCEEDED:
 528                         err = EHOSTUNREACH;
 529                         break;
 530                 case ICMP_SOURCE_QUENCH:
 531                         /* This is not a real error but ping wants to see it.
 532                          * Report it with some fake errno.
 533                          */
 534                         err = EREMOTEIO;
 535                         break;
 536                 case ICMP_PARAMETERPROB:
 537                         err = EPROTO;
 538                         harderr = 1;
 539                         break;
 540                 case ICMP_DEST_UNREACH:
 541                         if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
 542                                 ipv4_sk_update_pmtu(skb, sk, info);
 543                                 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
 544                                         err = EMSGSIZE;
 545                                         harderr = 1;
 546                                         break;
 547                                 }
 548                                 goto out;
 549                         }
 550                         err = EHOSTUNREACH;
 551                         if (code <= NR_ICMP_UNREACH) {
 552                                 harderr = icmp_err_convert[code].fatal;
 553                                 err = icmp_err_convert[code].errno;
 554                         }
 555                         break;
 556                 case ICMP_REDIRECT:
 557                         /* See ICMP_SOURCE_QUENCH */
 558                         ipv4_sk_redirect(skb, sk);
 559                         err = EREMOTEIO;
 560                         break;
 561                 }
 562 #if IS_ENABLED(CONFIG_IPV6)
 563         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 564                 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
 565 #endif
 566         }
 567
 568         /*
 569          *      RFC1122: OK.  Passes ICMP errors back to application, as per
 570          *      4.1.3.3.
 571          */
 572         if ((family == AF_INET && !inet_sock->recverr) ||
 573             (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
 574                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 575                         goto out;
 576         } else {
 577                 if (family == AF_INET) {
 578                         ip_icmp_error(sk, skb, err, 0 /* no remote port */,
 579                                       info, (u8 *)icmph);
 580 #if IS_ENABLED(CONFIG_IPV6)
 581                 } else if (family == AF_INET6) {
 582                         pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
 583                                                    info, (u8 *)icmph);
 584 #endif
 585                 }
 586         }
 587         sk->sk_err = err;
 588         sk->sk_error_report(sk);
 589 out:
 590         sock_put(sk);
 591 }
 592 EXPORT_SYMBOL_GPL(ping_err);
 593
 594 /*
 595  *      Copy and checksum an ICMP Echo packet from user space into a buffer
 596  *      starting from the payload.
 597  */
 598
 599 int ping_getfrag(void *from, char *to,
 600                  int offset, int fraglen, int odd, struct sk_buff *skb)
 601 {
 602         struct pingfakehdr *pfh = (struct pingfakehdr *)from;
 603
 604         if (offset == 0) {
 605                 if (fraglen < sizeof(struct icmphdr))
 606                         BUG();
 607                 if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
 608                             pfh->iov, 0, fraglen - sizeof(struct icmphdr),
 609                             &pfh->wcheck))
 610                         return -EFAULT;
 611         } else if (offset < sizeof(struct icmphdr)) {
 612                         BUG();
 613         } else {
 614                 if (csum_partial_copy_fromiovecend
 615                                 (to, pfh->iov, offset - sizeof(struct icmphdr),
 616                                  fraglen, &pfh->wcheck))
 617                         return -EFAULT;
 618         }
 619
 620 #if IS_ENABLED(CONFIG_IPV6)
 621         /* For IPv6, checksum each skb as we go along, as expected by
 622          * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
 623          * wcheck, it will be finalized in ping_v4_push_pending_frames.
 624          */
 625         if (pfh->family == AF_INET6) {
 626                 skb->csum = pfh->wcheck;
 627                 skb->ip_summed = CHECKSUM_NONE;
 628                 pfh->wcheck = 0;
 629         }
 630 #endif
 631
 632         return 0;
 633 }
 634 EXPORT_SYMBOL_GPL(ping_getfrag);
 635
 636 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
 637                                        struct flowi4 *fl4)
 638 {
 639         struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
 640
 641         pfh->wcheck = csum_partial((char *)&pfh->icmph,
 642                 sizeof(struct icmphdr), pfh->wcheck);
 643         pfh->icmph.checksum = csum_fold(pfh->wcheck);
 644         memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
 645         skb->ip_summed = CHECKSUM_NONE;
 646         return ip_push_pending_frames(sk, fl4);
 647 }
 648
 649 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
 650                         void *user_icmph, size_t icmph_len) {
 651         u8 type, code;
 652
 653         if (len > 0xFFFF)
 654                 return -EMSGSIZE;
 655
 656         /*
 657          *      Check the flags.
 658          */
 659
 660         /* Mirror BSD error message compatibility */
 661         if (msg->msg_flags & MSG_OOB)
 662                 return -EOPNOTSUPP;
 663
 664         /*
 665          *      Fetch the ICMP header provided by the userland.
 666          *      iovec is modified! The ICMP header is consumed.
 667          */
 668         if (memcpy_fromiovec(user_icmph, msg->msg_iov, icmph_len))
 669                 return -EFAULT;
 670
 671         if (family == AF_INET) {
 672                 type = ((struct icmphdr *) user_icmph)->type;
 673                 code = ((struct icmphdr *) user_icmph)->code;
 674 #if IS_ENABLED(CONFIG_IPV6)
 675         } else if (family == AF_INET6) {
 676                 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
 677                 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
 678 #endif
 679         } else {
 680                 BUG();
 681         }
 682
 683         if (!ping_supported(family, type, code))
 684                 return -EINVAL;
 685
 686         return 0;
 687 }
 688 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
 689
 690 int ping_v4_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 691                     size_t len)
 692 {
 693         struct net *net = sock_net(sk);
 694         struct flowi4 fl4;
 695         struct inet_sock *inet = inet_sk(sk);
 696         struct ipcm_cookie ipc;
 697         struct icmphdr user_icmph;
 698         struct pingfakehdr pfh;
 699         struct rtable *rt = NULL;
 700         struct ip_options_data opt_copy;
 701         int free = 0;
 702         __be32 saddr, daddr, faddr;
 703         u8  tos;
 704         int err;
 705
 706         pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
 707
 708         err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
 709                                   sizeof(user_icmph));
 710         if (err)
 711                 return err;
 712
 713         /*
 714          *      Get and verify the address.
 715          */
 716
 717         if (msg->msg_name) {
 718                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
 719                 if (msg->msg_namelen < sizeof(*usin))
 720                         return -EINVAL;
 721                 if (usin->sin_family != AF_INET)
 722                         return -EAFNOSUPPORT;
 723                 daddr = usin->sin_addr.s_addr;
 724                 /* no remote port */
 725         } else {
 726                 if (sk->sk_state != TCP_ESTABLISHED)
 727                         return -EDESTADDRREQ;
 728                 daddr = inet->inet_daddr;
 729                 /* no remote port */
 730         }
 731
 732         ipc.addr = inet->inet_saddr;
 733         ipc.opt = NULL;
 734         ipc.oif = sk->sk_bound_dev_if;
 735         ipc.tx_flags = 0;
 736
 737         sock_tx_timestamp(sk, &ipc.tx_flags);
 738
 739         if (msg->msg_controllen) {
 740                 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
 741                 if (err)
 742                         return err;
 743                 if (ipc.opt)
 744                         free = 1;
 745         }
 746         if (!ipc.opt) {
 747                 struct ip_options_rcu *inet_opt;
 748
 749                 rcu_read_lock();
 750                 inet_opt = rcu_dereference(inet->inet_opt);
 751                 if (inet_opt) {
 752                         memcpy(&opt_copy, inet_opt,
 753                                sizeof(*inet_opt) + inet_opt->opt.optlen);
 754                         ipc.opt = &opt_copy.opt;
 755                 }
 756                 rcu_read_unlock();
 757         }
 758
 759         saddr = ipc.addr;
 760         ipc.addr = faddr = daddr;
 761
 762         if (ipc.opt && ipc.opt->opt.srr) {
 763                 if (!daddr)
 764                         return -EINVAL;
 765                 faddr = ipc.opt->opt.faddr;
 766         }
 767         tos = RT_TOS(inet->tos);
 768         if (sock_flag(sk, SOCK_LOCALROUTE) ||
 769             (msg->msg_flags & MSG_DONTROUTE) ||
 770             (ipc.opt && ipc.opt->opt.is_strictroute)) {
 771                 tos |= RTO_ONLINK;
 772         }
 773
 774         if (ipv4_is_multicast(daddr)) {
 775                 if (!ipc.oif)
 776                         ipc.oif = inet->mc_index;
 777                 if (!saddr)
 778                         saddr = inet->mc_addr;
 779         } else if (!ipc.oif)
 780                 ipc.oif = inet->uc_index;
 781
 782         flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
 783                            RT_SCOPE_UNIVERSE, sk->sk_protocol,
 784                            inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
 785                            sock_i_uid(sk));
 786
 787         security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
 788         rt = ip_route_output_flow(net, &fl4, sk);
 789         if (IS_ERR(rt)) {
 790                 err = PTR_ERR(rt);
 791                 rt = NULL;
 792                 if (err == -ENETUNREACH)
 793                         IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
 794                 goto out;
 795         }
 796
 797         err = -EACCES;
 798         if ((rt->rt_flags & RTCF_BROADCAST) &&
 799             !sock_flag(sk, SOCK_BROADCAST))
 800                 goto out;
 801
 802         if (msg->msg_flags & MSG_CONFIRM)
 803                 goto do_confirm;
 804 back_from_confirm:
 805
 806         if (!ipc.addr)
 807                 ipc.addr = fl4.daddr;
 808
 809         lock_sock(sk);
 810
 811         pfh.icmph.type = user_icmph.type; /* already checked */
 812         pfh.icmph.code = user_icmph.code; /* ditto */
 813         pfh.icmph.checksum = 0;
 814         pfh.icmph.un.echo.id = inet->inet_sport;
 815         pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 816         pfh.iov = msg->msg_iov;
 817         pfh.wcheck = 0;
 818         pfh.family = AF_INET;
 819
 820         err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 821                         0, &ipc, &rt, msg->msg_flags);
 822         if (err)
 823                 ip_flush_pending_frames(sk);
 824         else
 825                 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 826         release_sock(sk);
 827
 828 out:
 829         ip_rt_put(rt);
 830         if (free)
 831                 kfree(ipc.opt);
 832         if (!err) {
 833                 icmp_out_count(sock_net(sk), user_icmph.type);
 834                 return len;
 835         }
 836         return err;
 837
 838 do_confirm:
 839         dst_confirm(&rt->dst);
 840         if (!(msg->msg_flags & MSG_PROBE) || len)
 841                 goto back_from_confirm;
 842         err = 0;
 843         goto out;
 844 }
 845
 846 int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 847                  size_t len, int noblock, int flags, int *addr_len)
 848 {
 849         struct inet_sock *isk = inet_sk(sk);
 850         int family = sk->sk_family;
 851         struct sockaddr_in *sin;
 852         struct sockaddr_in6 *sin6;
 853         struct sk_buff *skb;
 854         int copied, err;
 855
 856         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 857
 858         err = -EOPNOTSUPP;
 859         if (flags & MSG_OOB)
 860                 goto out;
 861
 862         if (addr_len) {
 863                 if (family == AF_INET)
 864                         *addr_len = sizeof(*sin);
 865                 else if (family == AF_INET6 && addr_len)
 866                         *addr_len = sizeof(*sin6);
 867         }
 868
 869         if (flags & MSG_ERRQUEUE) {
 870                 if (family == AF_INET) {
 871                         return ip_recv_error(sk, msg, len);
 872 #if IS_ENABLED(CONFIG_IPV6)
 873                 } else if (family == AF_INET6) {
 874                         return pingv6_ops.ipv6_recv_error(sk, msg, len);
 875 #endif
 876                 }
 877         }
 878
 879         skb = skb_recv_datagram(sk, flags, noblock, &err);
 880         if (!skb)
 881                 goto out;
 882
 883         copied = skb->len;
 884         if (copied > len) {
 885                 msg->msg_flags |= MSG_TRUNC;
 886                 copied = len;
 887         }
 888
 889         /* Don't bother checking the checksum */
 890         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 891         if (err)
 892                 goto done;
 893
 894         sock_recv_timestamp(msg, sk, skb);
 895
 896         /* Copy the address and add cmsg data. */
 897         if (family == AF_INET) {
 898                 sin = (struct sockaddr_in *) msg->msg_name;
 899                 if (sin) {
 900                         sin->sin_family = AF_INET;
 901                         sin->sin_port = 0 /* skb->h.uh->source */;
 902                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 903                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 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                 sin6 = (struct sockaddr_in6 *) 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                                                     IP6CB(skb)->iif);
 925                 }
 926
 927                 if (inet6_sk(sk)->rxopt.all)
 928                         pingv6_ops.ip6_datagram_recv_ctl(sk, msg, skb);
 929 #endif
 930         } else {
 931                 BUG();
 932         }
 933
 934         err = copied;
 935
 936 done:
 937         skb_free_datagram(sk, skb);
 938 out:
 939         pr_debug("ping_recvmsg -> %d\n", err);
 940         return err;
 941 }
 942 EXPORT_SYMBOL_GPL(ping_recvmsg);
 943
 944 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 945 {
 946         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 947                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
 948         if (sock_queue_rcv_skb(sk, skb) < 0) {
 949                 kfree_skb(skb);
 950                 pr_debug("ping_queue_rcv_skb -> failed\n");
 951                 return -1;
 952         }
 953         return 0;
 954 }
 955 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 956
 957
 958 /*
 959  *      All we need to do is get the socket.
 960  */
 961
 962 void ping_rcv(struct sk_buff *skb)
 963 {
 964         struct sock *sk;
 965         struct net *net = dev_net(skb->dev);
 966         struct icmphdr *icmph = icmp_hdr(skb);
 967
 968         /* We assume the packet has already been checked by icmp_rcv */
 969
 970         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 971                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 972
 973         /* Push ICMP header back */
 974         skb_push(skb, skb->data - (u8 *)icmph);
 975
 976         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 977         if (sk != NULL) {
 978                 pr_debug("rcv on socket %p\n", sk);
 979                 ping_queue_rcv_skb(sk, skb_get(skb));
 980                 sock_put(sk);
 981                 return;
 982         }
 983         pr_debug("no socket, dropping\n");
 984
 985         /* We're called from icmp_rcv(). kfree_skb() is done there. */
 986 }
 987 EXPORT_SYMBOL_GPL(ping_rcv);
 988
 989 struct proto ping_prot = {
 990         .name =         "PING",
 991         .owner =        THIS_MODULE,
 992         .init =         ping_init_sock,
 993         .close =        ping_close,
 994         .connect =      ip4_datagram_connect,
 995         .disconnect =   udp_disconnect,
 996         .setsockopt =   ip_setsockopt,
 997         .getsockopt =   ip_getsockopt,
 998         .sendmsg =      ping_v4_sendmsg,
 999         .recvmsg =      ping_recvmsg,
1000         .bind =         ping_bind,
1001         .backlog_rcv =  ping_queue_rcv_skb,
1002         .release_cb =   ip4_datagram_release_cb,
1003         .hash =         ping_hash,
1004         .unhash =       ping_unhash,
1005         .get_port =     ping_get_port,
1006         .obj_size =     sizeof(struct inet_sock),
1007 };
1008 EXPORT_SYMBOL(ping_prot);
1009
1010 #ifdef CONFIG_PROC_FS
1011
1012 static struct sock *ping_get_first(struct seq_file *seq, int start)
1013 {
1014         struct sock *sk;
1015         struct ping_iter_state *state = seq->private;
1016         struct net *net = seq_file_net(seq);
1017
1018         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1019              ++state->bucket) {
1020                 struct hlist_nulls_node *node;
1021                 struct hlist_nulls_head *hslot;
1022
1023                 hslot = &ping_table.hash[state->bucket];
1024
1025                 if (hlist_nulls_empty(hslot))
1026                         continue;
1027
1028                 sk_nulls_for_each(sk, node, hslot) {
1029                         if (net_eq(sock_net(sk), net))
1030                                 goto found;
1031                 }
1032         }
1033         sk = NULL;
1034 found:
1035         return sk;
1036 }
1037
1038 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1039 {
1040         struct ping_iter_state *state = seq->private;
1041         struct net *net = seq_file_net(seq);
1042
1043         do {
1044                 sk = sk_nulls_next(sk);
1045         } while (sk && (!net_eq(sock_net(sk), net)));
1046
1047         if (!sk)
1048                 return ping_get_first(seq, state->bucket + 1);
1049         return sk;
1050 }
1051
1052 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1053 {
1054         struct sock *sk = ping_get_first(seq, 0);
1055
1056         if (sk)
1057                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1058                         --pos;
1059         return pos ? NULL : sk;
1060 }
1061
1062 static void *ping_seq_start(struct seq_file *seq, loff_t *pos)
1063 {
1064         struct ping_iter_state *state = seq->private;
1065         state->bucket = 0;
1066
1067         read_lock_bh(&ping_table.lock);
1068
1069         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1070 }
1071
1072 static void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1073 {
1074         struct sock *sk;
1075
1076         if (v == SEQ_START_TOKEN)
1077                 sk = ping_get_idx(seq, 0);
1078         else
1079                 sk = ping_get_next(seq, v);
1080
1081         ++*pos;
1082         return sk;
1083 }
1084
1085 static void ping_seq_stop(struct seq_file *seq, void *v)
1086 {
1087         read_unlock_bh(&ping_table.lock);
1088 }
1089
1090 static void ping_format_sock(struct sock *sp, struct seq_file *f,
1091                 int bucket, int *len)
1092 {
1093         struct inet_sock *inet = inet_sk(sp);
1094         __be32 dest = inet->inet_daddr;
1095         __be32 src = inet->inet_rcv_saddr;
1096         __u16 destp = ntohs(inet->inet_dport);
1097         __u16 srcp = ntohs(inet->inet_sport);
1098
1099         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1100                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
1101                 bucket, src, srcp, dest, destp, sp->sk_state,
1102                 sk_wmem_alloc_get(sp),
1103                 sk_rmem_alloc_get(sp),
1104                 0, 0L, 0,
1105                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1106                 0, sock_i_ino(sp),
1107                 atomic_read(&sp->sk_refcnt), sp,
1108                 atomic_read(&sp->sk_drops), len);
1109 }
1110
1111 static int ping_seq_show(struct seq_file *seq, void *v)
1112 {
1113         if (v == SEQ_START_TOKEN)
1114                 seq_printf(seq, "%-127s\n",
1115                            "  sl  local_address rem_address   st tx_queue "
1116                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1117                            "inode ref pointer drops");
1118         else {
1119                 struct ping_iter_state *state = seq->private;
1120                 int len;
1121
1122                 ping_format_sock(v, seq, state->bucket, &len);
1123                 seq_printf(seq, "%*s\n", 127 - len, "");
1124         }
1125         return 0;
1126 }
1127
1128 static const struct seq_operations ping_seq_ops = {
1129         .show           = ping_seq_show,
1130         .start          = ping_seq_start,
1131         .next           = ping_seq_next,
1132         .stop           = ping_seq_stop,
1133 };
1134
1135 static int ping_seq_open(struct inode *inode, struct file *file)
1136 {
1137         return seq_open_net(inode, file, &ping_seq_ops,
1138                            sizeof(struct ping_iter_state));
1139 }
1140
1141 static const struct file_operations ping_seq_fops = {
1142         .open           = ping_seq_open,
1143         .read           = seq_read,
1144         .llseek         = seq_lseek,
1145         .release        = seq_release_net,
1146 };
1147
1148 static int ping_proc_register(struct net *net)
1149 {
1150         struct proc_dir_entry *p;
1151         int rc = 0;
1152
1153         p = proc_create("icmp", S_IRUGO, net->proc_net, &ping_seq_fops);
1154         if (!p)
1155                 rc = -ENOMEM;
1156         return rc;
1157 }
1158
1159 static void ping_proc_unregister(struct net *net)
1160 {
1161         remove_proc_entry("icmp", net->proc_net);
1162 }
1163
1164
1165 static int __net_init ping_proc_init_net(struct net *net)
1166 {
1167         return ping_proc_register(net);
1168 }
1169
1170 static void __net_exit ping_proc_exit_net(struct net *net)
1171 {
1172         ping_proc_unregister(net);
1173 }
1174
1175 static struct pernet_operations ping_net_ops = {
1176         .init = ping_proc_init_net,
1177         .exit = ping_proc_exit_net,
1178 };
1179
1180 int __init ping_proc_init(void)
1181 {
1182         return register_pernet_subsys(&ping_net_ops);
1183 }
1184
1185 void ping_proc_exit(void)
1186 {
1187         unregister_pernet_subsys(&ping_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 }