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                 sk_nulls_node_init(&sk->sk_nulls_node);
 158                 sock_put(sk);
 159                 isk->inet_num = 0;
 160                 isk->inet_sport = 0;
 161                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 162                 write_unlock_bh(&ping_table.lock);
 163         }
 164 }
 165 EXPORT_SYMBOL_GPL(ping_unhash);
 166
 167 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
 168 {
 169         struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
 170         struct sock *sk = NULL;
 171         struct inet_sock *isk;
 172         struct hlist_nulls_node *hnode;
 173         int dif = skb->dev->ifindex;
 174
 175         if (skb->protocol == htons(ETH_P_IP)) {
 176                 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
 177                          (int)ident, &ip_hdr(skb)->daddr, dif);
 178 #if IS_ENABLED(CONFIG_IPV6)
 179         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 180                 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
 181                          (int)ident, &ipv6_hdr(skb)->daddr, dif);
 182 #endif
 183         }
 184
 185         read_lock_bh(&ping_table.lock);
 186
 187         ping_portaddr_for_each_entry(sk, hnode, hslot) {
 188                 isk = inet_sk(sk);
 189
 190                 pr_debug("iterate\n");
 191                 if (isk->inet_num != ident)
 192                         continue;
 193
 194                 if (skb->protocol == htons(ETH_P_IP) &&
 195                     sk->sk_family == AF_INET) {
 196                         pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
 197                                  (int) isk->inet_num, &isk->inet_rcv_saddr,
 198                                  sk->sk_bound_dev_if);
 199
 200                         if (isk->inet_rcv_saddr &&
 201                             isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
 202                                 continue;
 203 #if IS_ENABLED(CONFIG_IPV6)
 204                 } else if (skb->protocol == htons(ETH_P_IPV6) &&
 205                            sk->sk_family == AF_INET6) {
 206                         struct ipv6_pinfo *np = inet6_sk(sk);
 207
 208                         pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
 209                                  (int) isk->inet_num,
 210                                  &inet6_sk(sk)->rcv_saddr,
 211                                  sk->sk_bound_dev_if);
 212
 213                         if (!ipv6_addr_any(&np->rcv_saddr) &&
 214                             !ipv6_addr_equal(&np->rcv_saddr,
 215                                              &ipv6_hdr(skb)->daddr))
 216                                 continue;
 217 #endif
 218                 } else {
 219                         continue;
 220                 }
 221
 222                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
 223                         continue;
 224
 225                 sock_hold(sk);
 226                 goto exit;
 227         }
 228
 229         sk = NULL;
 230 exit:
 231         read_unlock_bh(&ping_table.lock);
 232
 233         return sk;
 234 }
 235
 236 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
 237                                           kgid_t *high)
 238 {
 239         kgid_t *data = net->ipv4.sysctl_ping_group_range;
 240         unsigned int seq;
 241
 242         do {
 243                 seq = read_seqbegin(&sysctl_local_ports.lock);
 244
 245                 *low = data[0];
 246                 *high = data[1];
 247         } while (read_seqretry(&sysctl_local_ports.lock, seq));
 248 }
 249
 250
 251 int ping_init_sock(struct sock *sk)
 252 {
 253         struct net *net = sock_net(sk);
 254         kgid_t group = current_egid();
 255         struct group_info *group_info;
 256         int i, j, count;
 257         kgid_t low, high;
 258         int ret = 0;
 259
 260         if (sk->sk_family == AF_INET6)
 261                 inet6_sk(sk)->ipv6only = 1;
 262
 263         inet_get_ping_group_range_net(net, &low, &high);
 264         if (gid_lte(low, group) && gid_lte(group, high))
 265                 return 0;
 266
 267         group_info = get_current_groups();
 268         count = group_info->ngroups;
 269         for (i = 0; i < group_info->nblocks; i++) {
 270                 int cp_count = min_t(int, NGROUPS_PER_BLOCK, count);
 271                 for (j = 0; j < cp_count; j++) {
 272                         kgid_t gid = group_info->blocks[i][j];
 273                         if (gid_lte(low, gid) && gid_lte(gid, high))
 274                                 goto out_release_group;
 275                 }
 276
 277                 count -= cp_count;
 278         }
 279
 280         ret = -EACCES;
 281
 282 out_release_group:
 283         put_group_info(group_info);
 284         return ret;
 285 }
 286 EXPORT_SYMBOL_GPL(ping_init_sock);
 287
 288 void ping_close(struct sock *sk, long timeout)
 289 {
 290         pr_debug("ping_close(sk=%p,sk->num=%u)\n",
 291                  inet_sk(sk), inet_sk(sk)->inet_num);
 292         pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
 293
 294         sk_common_release(sk);
 295 }
 296 EXPORT_SYMBOL_GPL(ping_close);
 297
 298 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
 299 int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
 300                          struct sockaddr *uaddr, int addr_len) {
 301         struct net *net = sock_net(sk);
 302         if (sk->sk_family == AF_INET) {
 303                 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 304                 int chk_addr_ret;
 305
 306                 if (addr_len < sizeof(*addr))
 307                         return -EINVAL;
 308
 309                 if (addr->sin_family != AF_INET &&
 310                     !(addr->sin_family == AF_UNSPEC &&
 311                       addr->sin_addr.s_addr == htonl(INADDR_ANY)))
 312                         return -EAFNOSUPPORT;
 313
 314                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
 315                          sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
 316
 317                 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
 318
 319                 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
 320                         chk_addr_ret = RTN_LOCAL;
 321
 322                 if ((sysctl_ip_nonlocal_bind == 0 &&
 323                     isk->freebind == 0 && isk->transparent == 0 &&
 324                      chk_addr_ret != RTN_LOCAL) ||
 325                     chk_addr_ret == RTN_MULTICAST ||
 326                     chk_addr_ret == RTN_BROADCAST)
 327                         return -EADDRNOTAVAIL;
 328
 329 #if IS_ENABLED(CONFIG_IPV6)
 330         } else if (sk->sk_family == AF_INET6) {
 331                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 332                 int addr_type, scoped, has_addr;
 333                 struct net_device *dev = NULL;
 334
 335                 if (addr_len < sizeof(*addr))
 336                         return -EINVAL;
 337
 338                 if (addr->sin6_family != AF_INET6)
 339                         return -EAFNOSUPPORT;
 340
 341                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 342                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 343
 344                 addr_type = ipv6_addr_type(&addr->sin6_addr);
 345                 scoped = __ipv6_addr_needs_scope_id(addr_type);
 346                 if ((addr_type != IPV6_ADDR_ANY &&
 347                      !(addr_type & IPV6_ADDR_UNICAST)) ||
 348                     (scoped && !addr->sin6_scope_id))
 349                         return -EINVAL;
 350
 351                 rcu_read_lock();
 352                 if (addr->sin6_scope_id) {
 353                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 354                         if (!dev) {
 355                                 rcu_read_unlock();
 356                                 return -ENODEV;
 357                         }
 358                 }
 359                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 360                                                     scoped);
 361                 rcu_read_unlock();
 362
 363                 if (!(isk->freebind || isk->transparent || has_addr ||
 364                       addr_type == IPV6_ADDR_ANY))
 365                         return -EADDRNOTAVAIL;
 366
 367                 if (scoped)
 368                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 369 #endif
 370         } else {
 371                 return -EAFNOSUPPORT;
 372         }
 373         return 0;
 374 }
 375
 376 void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
 377 {
 378         if (saddr->sa_family == AF_INET) {
 379                 struct inet_sock *isk = inet_sk(sk);
 380                 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
 381                 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
 382 #if IS_ENABLED(CONFIG_IPV6)
 383         } else if (saddr->sa_family == AF_INET6) {
 384                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
 385                 struct ipv6_pinfo *np = inet6_sk(sk);
 386                 np->rcv_saddr = np->saddr = addr->sin6_addr;
 387 #endif
 388         }
 389 }
 390
 391 void ping_clear_saddr(struct sock *sk, int dif)
 392 {
 393         sk->sk_bound_dev_if = dif;
 394         if (sk->sk_family == AF_INET) {
 395                 struct inet_sock *isk = inet_sk(sk);
 396                 isk->inet_rcv_saddr = isk->inet_saddr = 0;
 397 #if IS_ENABLED(CONFIG_IPV6)
 398         } else if (sk->sk_family == AF_INET6) {
 399                 struct ipv6_pinfo *np = inet6_sk(sk);
 400                 memset(&np->rcv_saddr, 0, sizeof(np->rcv_saddr));
 401                 memset(&np->saddr, 0, sizeof(np->saddr));
 402 #endif
 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         ping_set_saddr(sk, uaddr);
 429         snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
 430         if (ping_get_port(sk, snum) != 0) {
 431                 ping_clear_saddr(sk, dif);
 432                 goto out;
 433         }
 434
 435         pr_debug("after bind(): num = %d, dif = %d\n",
 436                  (int)isk->inet_num,
 437                  (int)sk->sk_bound_dev_if);
 438
 439         err = 0;
 440         if ((sk->sk_family == AF_INET && isk->inet_rcv_saddr) ||
 441             (sk->sk_family == AF_INET6 &&
 442              !ipv6_addr_any(&inet6_sk(sk)->rcv_saddr)))
 443                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 444
 445         if (snum)
 446                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 447         isk->inet_sport = htons(isk->inet_num);
 448         isk->inet_daddr = 0;
 449         isk->inet_dport = 0;
 450
 451 #if IS_ENABLED(CONFIG_IPV6)
 452         if (sk->sk_family == AF_INET6)
 453                 memset(&inet6_sk(sk)->daddr, 0, sizeof(inet6_sk(sk)->daddr));
 454 #endif
 455
 456         sk_dst_reset(sk);
 457 out:
 458         release_sock(sk);
 459         pr_debug("ping_v4_bind -> %d\n", err);
 460         return err;
 461 }
 462 EXPORT_SYMBOL_GPL(ping_bind);
 463
 464 /*
 465  * Is this a supported type of ICMP message?
 466  */
 467
 468 static inline int ping_supported(int family, int type, int code)
 469 {
 470         return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
 471                (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
 472 }
 473
 474 /*
 475  * This routine is called by the ICMP module when it gets some
 476  * sort of error condition.
 477  */
 478
 479 void ping_err(struct sk_buff *skb, int offset, u32 info)
 480 {
 481         int family;
 482         struct icmphdr *icmph;
 483         struct inet_sock *inet_sock;
 484         int type;
 485         int code;
 486         struct net *net = dev_net(skb->dev);
 487         struct sock *sk;
 488         int harderr;
 489         int err;
 490
 491         if (skb->protocol == htons(ETH_P_IP)) {
 492                 family = AF_INET;
 493                 type = icmp_hdr(skb)->type;
 494                 code = icmp_hdr(skb)->code;
 495                 icmph = (struct icmphdr *)(skb->data + offset);
 496         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 497                 family = AF_INET6;
 498                 type = icmp6_hdr(skb)->icmp6_type;
 499                 code = icmp6_hdr(skb)->icmp6_code;
 500                 icmph = (struct icmphdr *) (skb->data + offset);
 501         } else {
 502                 BUG();
 503         }
 504
 505         /* We assume the packet has already been checked by icmp_unreach */
 506
 507         if (!ping_supported(family, icmph->type, icmph->code))
 508                 return;
 509
 510         pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
 511                  skb->protocol, type, code, ntohs(icmph->un.echo.id),
 512                  ntohs(icmph->un.echo.sequence));
 513
 514         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 515         if (sk == NULL) {
 516                 pr_debug("no socket, dropping\n");
 517                 return; /* No socket for error */
 518         }
 519         pr_debug("err on socket %p\n", sk);
 520
 521         err = 0;
 522         harderr = 0;
 523         inet_sock = inet_sk(sk);
 524
 525         if (skb->protocol == htons(ETH_P_IP)) {
 526                 switch (type) {
 527                 default:
 528                 case ICMP_TIME_EXCEEDED:
 529                         err = EHOSTUNREACH;
 530                         break;
 531                 case ICMP_SOURCE_QUENCH:
 532                         /* This is not a real error but ping wants to see it.
 533                          * Report it with some fake errno.
 534                          */
 535                         err = EREMOTEIO;
 536                         break;
 537                 case ICMP_PARAMETERPROB:
 538                         err = EPROTO;
 539                         harderr = 1;
 540                         break;
 541                 case ICMP_DEST_UNREACH:
 542                         if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
 543                                 ipv4_sk_update_pmtu(skb, sk, info);
 544                                 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
 545                                         err = EMSGSIZE;
 546                                         harderr = 1;
 547                                         break;
 548                                 }
 549                                 goto out;
 550                         }
 551                         err = EHOSTUNREACH;
 552                         if (code <= NR_ICMP_UNREACH) {
 553                                 harderr = icmp_err_convert[code].fatal;
 554                                 err = icmp_err_convert[code].errno;
 555                         }
 556                         break;
 557                 case ICMP_REDIRECT:
 558                         /* See ICMP_SOURCE_QUENCH */
 559                         ipv4_sk_redirect(skb, sk);
 560                         err = EREMOTEIO;
 561                         break;
 562                 }
 563 #if IS_ENABLED(CONFIG_IPV6)
 564         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 565                 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
 566 #endif
 567         }
 568
 569         /*
 570          *      RFC1122: OK.  Passes ICMP errors back to application, as per
 571          *      4.1.3.3.
 572          */
 573         if ((family == AF_INET && !inet_sock->recverr) ||
 574             (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
 575                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 576                         goto out;
 577         } else {
 578                 if (family == AF_INET) {
 579                         ip_icmp_error(sk, skb, err, 0 /* no remote port */,
 580                                       info, (u8 *)icmph);
 581 #if IS_ENABLED(CONFIG_IPV6)
 582                 } else if (family == AF_INET6) {
 583                         pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
 584                                                    info, (u8 *)icmph);
 585 #endif
 586                 }
 587         }
 588         sk->sk_err = err;
 589         sk->sk_error_report(sk);
 590 out:
 591         sock_put(sk);
 592 }
 593 EXPORT_SYMBOL_GPL(ping_err);
 594
 595 /*
 596  *      Copy and checksum an ICMP Echo packet from user space into a buffer
 597  *      starting from the payload.
 598  */
 599
 600 int ping_getfrag(void *from, char *to,
 601                  int offset, int fraglen, int odd, struct sk_buff *skb)
 602 {
 603         struct pingfakehdr *pfh = (struct pingfakehdr *)from;
 604
 605         if (offset == 0) {
 606                 if (fraglen < sizeof(struct icmphdr))
 607                         BUG();
 608                 if (csum_partial_copy_fromiovecend(to + sizeof(struct icmphdr),
 609                             pfh->iov, 0, fraglen - sizeof(struct icmphdr),
 610                             &pfh->wcheck))
 611                         return -EFAULT;
 612         } else if (offset < sizeof(struct icmphdr)) {
 613                         BUG();
 614         } else {
 615                 if (csum_partial_copy_fromiovecend
 616                                 (to, pfh->iov, offset - sizeof(struct icmphdr),
 617                                  fraglen, &pfh->wcheck))
 618                         return -EFAULT;
 619         }
 620
 621 #if IS_ENABLED(CONFIG_IPV6)
 622         /* For IPv6, checksum each skb as we go along, as expected by
 623          * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
 624          * wcheck, it will be finalized in ping_v4_push_pending_frames.
 625          */
 626         if (pfh->family == AF_INET6) {
 627                 skb->csum = pfh->wcheck;
 628                 skb->ip_summed = CHECKSUM_NONE;
 629                 pfh->wcheck = 0;
 630         }
 631 #endif
 632
 633         return 0;
 634 }
 635 EXPORT_SYMBOL_GPL(ping_getfrag);
 636
 637 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
 638                                        struct flowi4 *fl4)
 639 {
 640         struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
 641
 642         pfh->wcheck = csum_partial((char *)&pfh->icmph,
 643                 sizeof(struct icmphdr), pfh->wcheck);
 644         pfh->icmph.checksum = csum_fold(pfh->wcheck);
 645         memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
 646         skb->ip_summed = CHECKSUM_NONE;
 647         return ip_push_pending_frames(sk, fl4);
 648 }
 649
 650 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
 651                         void *user_icmph, size_t icmph_len) {
 652         u8 type, code;
 653
 654         if (len > 0xFFFF)
 655                 return -EMSGSIZE;
 656
 657         /*
 658          *      Check the flags.
 659          */
 660
 661         /* Mirror BSD error message compatibility */
 662         if (msg->msg_flags & MSG_OOB)
 663                 return -EOPNOTSUPP;
 664
 665         /*
 666          *      Fetch the ICMP header provided by the userland.
 667          *      iovec is modified! The ICMP header is consumed.
 668          */
 669         if (memcpy_fromiovec(user_icmph, msg->msg_iov, icmph_len))
 670                 return -EFAULT;
 671
 672         if (family == AF_INET) {
 673                 type = ((struct icmphdr *) user_icmph)->type;
 674                 code = ((struct icmphdr *) user_icmph)->code;
 675 #if IS_ENABLED(CONFIG_IPV6)
 676         } else if (family == AF_INET6) {
 677                 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
 678                 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
 679 #endif
 680         } else {
 681                 BUG();
 682         }
 683
 684         if (!ping_supported(family, type, code))
 685                 return -EINVAL;
 686
 687         return 0;
 688 }
 689 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
 690
 691 int ping_v4_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 692                     size_t len)
 693 {
 694         struct net *net = sock_net(sk);
 695         struct flowi4 fl4;
 696         struct inet_sock *inet = inet_sk(sk);
 697         struct ipcm_cookie ipc;
 698         struct icmphdr user_icmph;
 699         struct pingfakehdr pfh;
 700         struct rtable *rt = NULL;
 701         struct ip_options_data opt_copy;
 702         int free = 0;
 703         __be32 saddr, daddr, faddr;
 704         u8  tos;
 705         int err;
 706
 707         pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
 708
 709         err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
 710                                   sizeof(user_icmph));
 711         if (err)
 712                 return err;
 713
 714         /*
 715          *      Get and verify the address.
 716          */
 717
 718         if (msg->msg_name) {
 719                 struct sockaddr_in *usin = (struct sockaddr_in *)msg->msg_name;
 720                 if (msg->msg_namelen < sizeof(*usin))
 721                         return -EINVAL;
 722                 if (usin->sin_family != AF_INET)
 723                         return -EAFNOSUPPORT;
 724                 daddr = usin->sin_addr.s_addr;
 725                 /* no remote port */
 726         } else {
 727                 if (sk->sk_state != TCP_ESTABLISHED)
 728                         return -EDESTADDRREQ;
 729                 daddr = inet->inet_daddr;
 730                 /* no remote port */
 731         }
 732
 733         ipc.addr = inet->inet_saddr;
 734         ipc.opt = NULL;
 735         ipc.oif = sk->sk_bound_dev_if;
 736         ipc.tx_flags = 0;
 737
 738         sock_tx_timestamp(sk, &ipc.tx_flags);
 739
 740         if (msg->msg_controllen) {
 741                 err = ip_cmsg_send(sock_net(sk), msg, &ipc);
 742                 if (err)
 743                         return err;
 744                 if (ipc.opt)
 745                         free = 1;
 746         }
 747         if (!ipc.opt) {
 748                 struct ip_options_rcu *inet_opt;
 749
 750                 rcu_read_lock();
 751                 inet_opt = rcu_dereference(inet->inet_opt);
 752                 if (inet_opt) {
 753                         memcpy(&opt_copy, inet_opt,
 754                                sizeof(*inet_opt) + inet_opt->opt.optlen);
 755                         ipc.opt = &opt_copy.opt;
 756                 }
 757                 rcu_read_unlock();
 758         }
 759
 760         saddr = ipc.addr;
 761         ipc.addr = faddr = daddr;
 762
 763         if (ipc.opt && ipc.opt->opt.srr) {
 764                 if (!daddr)
 765                         return -EINVAL;
 766                 faddr = ipc.opt->opt.faddr;
 767         }
 768         tos = RT_TOS(inet->tos);
 769         if (sock_flag(sk, SOCK_LOCALROUTE) ||
 770             (msg->msg_flags & MSG_DONTROUTE) ||
 771             (ipc.opt && ipc.opt->opt.is_strictroute)) {
 772                 tos |= RTO_ONLINK;
 773         }
 774
 775         if (ipv4_is_multicast(daddr)) {
 776                 if (!ipc.oif)
 777                         ipc.oif = inet->mc_index;
 778                 if (!saddr)
 779                         saddr = inet->mc_addr;
 780         } else if (!ipc.oif)
 781                 ipc.oif = inet->uc_index;
 782
 783         flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
 784                            RT_SCOPE_UNIVERSE, sk->sk_protocol,
 785                            inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
 786                            sock_i_uid(sk));
 787
 788         security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
 789         rt = ip_route_output_flow(net, &fl4, sk);
 790         if (IS_ERR(rt)) {
 791                 err = PTR_ERR(rt);
 792                 rt = NULL;
 793                 if (err == -ENETUNREACH)
 794                         IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
 795                 goto out;
 796         }
 797
 798         err = -EACCES;
 799         if ((rt->rt_flags & RTCF_BROADCAST) &&
 800             !sock_flag(sk, SOCK_BROADCAST))
 801                 goto out;
 802
 803         if (msg->msg_flags & MSG_CONFIRM)
 804                 goto do_confirm;
 805 back_from_confirm:
 806
 807         if (!ipc.addr)
 808                 ipc.addr = fl4.daddr;
 809
 810         lock_sock(sk);
 811
 812         pfh.icmph.type = user_icmph.type; /* already checked */
 813         pfh.icmph.code = user_icmph.code; /* ditto */
 814         pfh.icmph.checksum = 0;
 815         pfh.icmph.un.echo.id = inet->inet_sport;
 816         pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 817         pfh.iov = msg->msg_iov;
 818         pfh.wcheck = 0;
 819         pfh.family = AF_INET;
 820
 821         err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 822                         0, &ipc, &rt, msg->msg_flags);
 823         if (err)
 824                 ip_flush_pending_frames(sk);
 825         else
 826                 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 827         release_sock(sk);
 828
 829 out:
 830         ip_rt_put(rt);
 831         if (free)
 832                 kfree(ipc.opt);
 833         if (!err) {
 834                 icmp_out_count(sock_net(sk), user_icmph.type);
 835                 return len;
 836         }
 837         return err;
 838
 839 do_confirm:
 840         dst_confirm(&rt->dst);
 841         if (!(msg->msg_flags & MSG_PROBE) || len)
 842                 goto back_from_confirm;
 843         err = 0;
 844         goto out;
 845 }
 846
 847 int ping_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
 848                  size_t len, int noblock, int flags, int *addr_len)
 849 {
 850         struct inet_sock *isk = inet_sk(sk);
 851         int family = sk->sk_family;
 852         struct sockaddr_in *sin;
 853         struct sockaddr_in6 *sin6;
 854         struct sk_buff *skb;
 855         int copied, err;
 856
 857         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 858
 859         err = -EOPNOTSUPP;
 860         if (flags & MSG_OOB)
 861                 goto out;
 862
 863         if (addr_len) {
 864                 if (family == AF_INET)
 865                         *addr_len = sizeof(*sin);
 866                 else if (family == AF_INET6 && addr_len)
 867                         *addr_len = sizeof(*sin6);
 868         }
 869
 870         if (flags & MSG_ERRQUEUE) {
 871                 if (family == AF_INET) {
 872                         return ip_recv_error(sk, msg, len, addr_len);
 873 #if IS_ENABLED(CONFIG_IPV6)
 874                 } else if (family == AF_INET6) {
 875                         return pingv6_ops.ipv6_recv_error(sk, msg, len);
 876 #endif
 877                 }
 878         }
 879
 880         skb = skb_recv_datagram(sk, flags, noblock, &err);
 881         if (!skb)
 882                 goto out;
 883
 884         copied = skb->len;
 885         if (copied > len) {
 886                 msg->msg_flags |= MSG_TRUNC;
 887                 copied = len;
 888         }
 889
 890         /* Don't bother checking the checksum */
 891         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 892         if (err)
 893                 goto done;
 894
 895         sock_recv_timestamp(msg, sk, skb);
 896
 897         /* Copy the address and add cmsg data. */
 898         if (family == AF_INET) {
 899                 sin = (struct sockaddr_in *) msg->msg_name;
 900                 if (sin) {
 901                         sin->sin_family = AF_INET;
 902                         sin->sin_port = 0 /* skb->h.uh->source */;
 903                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 904                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 905                 }
 906
 907                 if (isk->cmsg_flags)
 908                         ip_cmsg_recv(msg, skb);
 909
 910 #if IS_ENABLED(CONFIG_IPV6)
 911         } else if (family == AF_INET6) {
 912                 struct ipv6_pinfo *np = inet6_sk(sk);
 913                 struct ipv6hdr *ip6 = ipv6_hdr(skb);
 914                 sin6 = (struct sockaddr_in6 *) msg->msg_name;
 915
 916                 if (sin6) {
 917                         sin6->sin6_family = AF_INET6;
 918                         sin6->sin6_port = 0;
 919                         sin6->sin6_addr = ip6->saddr;
 920                         sin6->sin6_flowinfo = 0;
 921                         if (np->sndflow)
 922                                 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
 923                         sin6->sin6_scope_id =
 924                                 ipv6_iface_scope_id(&sin6->sin6_addr,
 925                                                     IP6CB(skb)->iif);
 926                 }
 927
 928                 if (inet6_sk(sk)->rxopt.all)
 929                         pingv6_ops.ip6_datagram_recv_ctl(sk, msg, skb);
 930 #endif
 931         } else {
 932                 BUG();
 933         }
 934
 935         err = copied;
 936
 937 done:
 938         skb_free_datagram(sk, skb);
 939 out:
 940         pr_debug("ping_recvmsg -> %d\n", err);
 941         return err;
 942 }
 943 EXPORT_SYMBOL_GPL(ping_recvmsg);
 944
 945 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 946 {
 947         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 948                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
 949         if (sock_queue_rcv_skb(sk, skb) < 0) {
 950                 kfree_skb(skb);
 951                 pr_debug("ping_queue_rcv_skb -> failed\n");
 952                 return -1;
 953         }
 954         return 0;
 955 }
 956 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 957
 958
 959 /*
 960  *      All we need to do is get the socket.
 961  */
 962
 963 void ping_rcv(struct sk_buff *skb)
 964 {
 965         struct sock *sk;
 966         struct net *net = dev_net(skb->dev);
 967         struct icmphdr *icmph = icmp_hdr(skb);
 968
 969         /* We assume the packet has already been checked by icmp_rcv */
 970
 971         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 972                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 973
 974         /* Push ICMP header back */
 975         skb_push(skb, skb->data - (u8 *)icmph);
 976
 977         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 978         if (sk != NULL) {
 979                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 980
 981                 pr_debug("rcv on socket %p\n", sk);
 982                 if (skb2)
 983                         ping_queue_rcv_skb(sk, skb2);
 984                 sock_put(sk);
 985                 return;
 986         }
 987         pr_debug("no socket, dropping\n");
 988
 989         /* We're called from icmp_rcv(). kfree_skb() is done there. */
 990 }
 991 EXPORT_SYMBOL_GPL(ping_rcv);
 992
 993 struct proto ping_prot = {
 994         .name =         "PING",
 995         .owner =        THIS_MODULE,
 996         .init =         ping_init_sock,
 997         .close =        ping_close,
 998         .connect =      ip4_datagram_connect,
 999         .disconnect =   udp_disconnect,
1000         .setsockopt =   ip_setsockopt,
1001         .getsockopt =   ip_getsockopt,
1002         .sendmsg =      ping_v4_sendmsg,
1003         .recvmsg =      ping_recvmsg,
1004         .bind =         ping_bind,
1005         .backlog_rcv =  ping_queue_rcv_skb,
1006         .release_cb =   ip4_datagram_release_cb,
1007         .hash =         ping_hash,
1008         .unhash =       ping_unhash,
1009         .get_port =     ping_get_port,
1010         .obj_size =     sizeof(struct inet_sock),
1011 };
1012 EXPORT_SYMBOL(ping_prot);
1013
1014 #ifdef CONFIG_PROC_FS
1015
1016 static struct sock *ping_get_first(struct seq_file *seq, int start)
1017 {
1018         struct sock *sk;
1019         struct ping_iter_state *state = seq->private;
1020         struct net *net = seq_file_net(seq);
1021
1022         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1023              ++state->bucket) {
1024                 struct hlist_nulls_node *node;
1025                 struct hlist_nulls_head *hslot;
1026
1027                 hslot = &ping_table.hash[state->bucket];
1028
1029                 if (hlist_nulls_empty(hslot))
1030                         continue;
1031
1032                 sk_nulls_for_each(sk, node, hslot) {
1033                         if (net_eq(sock_net(sk), net))
1034                                 goto found;
1035                 }
1036         }
1037         sk = NULL;
1038 found:
1039         return sk;
1040 }
1041
1042 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1043 {
1044         struct ping_iter_state *state = seq->private;
1045         struct net *net = seq_file_net(seq);
1046
1047         do {
1048                 sk = sk_nulls_next(sk);
1049         } while (sk && (!net_eq(sock_net(sk), net)));
1050
1051         if (!sk)
1052                 return ping_get_first(seq, state->bucket + 1);
1053         return sk;
1054 }
1055
1056 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1057 {
1058         struct sock *sk = ping_get_first(seq, 0);
1059
1060         if (sk)
1061                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1062                         --pos;
1063         return pos ? NULL : sk;
1064 }
1065
1066 static void *ping_seq_start(struct seq_file *seq, loff_t *pos)
1067 {
1068         struct ping_iter_state *state = seq->private;
1069         state->bucket = 0;
1070
1071         read_lock_bh(&ping_table.lock);
1072
1073         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1074 }
1075
1076 static void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1077 {
1078         struct sock *sk;
1079
1080         if (v == SEQ_START_TOKEN)
1081                 sk = ping_get_idx(seq, 0);
1082         else
1083                 sk = ping_get_next(seq, v);
1084
1085         ++*pos;
1086         return sk;
1087 }
1088
1089 static void ping_seq_stop(struct seq_file *seq, void *v)
1090 {
1091         read_unlock_bh(&ping_table.lock);
1092 }
1093
1094 static void ping_format_sock(struct sock *sp, struct seq_file *f,
1095                 int bucket, int *len)
1096 {
1097         struct inet_sock *inet = inet_sk(sp);
1098         __be32 dest = inet->inet_daddr;
1099         __be32 src = inet->inet_rcv_saddr;
1100         __u16 destp = ntohs(inet->inet_dport);
1101         __u16 srcp = ntohs(inet->inet_sport);
1102
1103         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1104                 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d%n",
1105                 bucket, src, srcp, dest, destp, sp->sk_state,
1106                 sk_wmem_alloc_get(sp),
1107                 sk_rmem_alloc_get(sp),
1108                 0, 0L, 0,
1109                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1110                 0, sock_i_ino(sp),
1111                 atomic_read(&sp->sk_refcnt), sp,
1112                 atomic_read(&sp->sk_drops), len);
1113 }
1114
1115 static int ping_seq_show(struct seq_file *seq, void *v)
1116 {
1117         if (v == SEQ_START_TOKEN)
1118                 seq_printf(seq, "%-127s\n",
1119                            "  sl  local_address rem_address   st tx_queue "
1120                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1121                            "inode ref pointer drops");
1122         else {
1123                 struct ping_iter_state *state = seq->private;
1124                 int len;
1125
1126                 ping_format_sock(v, seq, state->bucket, &len);
1127                 seq_printf(seq, "%*s\n", 127 - len, "");
1128         }
1129         return 0;
1130 }
1131
1132 static const struct seq_operations ping_seq_ops = {
1133         .show           = ping_seq_show,
1134         .start          = ping_seq_start,
1135         .next           = ping_seq_next,
1136         .stop           = ping_seq_stop,
1137 };
1138
1139 static int ping_seq_open(struct inode *inode, struct file *file)
1140 {
1141         return seq_open_net(inode, file, &ping_seq_ops,
1142                            sizeof(struct ping_iter_state));
1143 }
1144
1145 static const struct file_operations ping_seq_fops = {
1146         .open           = ping_seq_open,
1147         .read           = seq_read,
1148         .llseek         = seq_lseek,
1149         .release        = seq_release_net,
1150 };
1151
1152 static int ping_proc_register(struct net *net)
1153 {
1154         struct proc_dir_entry *p;
1155         int rc = 0;
1156
1157         p = proc_create("icmp", S_IRUGO, net->proc_net, &ping_seq_fops);
1158         if (!p)
1159                 rc = -ENOMEM;
1160         return rc;
1161 }
1162
1163 static void ping_proc_unregister(struct net *net)
1164 {
1165         remove_proc_entry("icmp", net->proc_net);
1166 }
1167
1168
1169 static int __net_init ping_proc_init_net(struct net *net)
1170 {
1171         return ping_proc_register(net);
1172 }
1173
1174 static void __net_exit ping_proc_exit_net(struct net *net)
1175 {
1176         ping_proc_unregister(net);
1177 }
1178
1179 static struct pernet_operations ping_net_ops = {
1180         .init = ping_proc_init_net,
1181         .exit = ping_proc_exit_net,
1182 };
1183
1184 int __init ping_proc_init(void)
1185 {
1186         return register_pernet_subsys(&ping_net_ops);
1187 }
1188
1189 void ping_proc_exit(void)
1190 {
1191         unregister_pernet_subsys(&ping_net_ops);
1192 }
1193
1194 #endif
1195
1196 void __init ping_init(void)
1197 {
1198         int i;
1199
1200         for (i = 0; i < PING_HTABLE_SIZE; i++)
1201                 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1202         rwlock_init(&ping_table.lock);
1203 }