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 = skb->dev->ifindex;
 176
 177         if (skb->protocol == htons(ETH_P_IP)) {
 178                 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
 179                          (int)ident, &ip_hdr(skb)->daddr, dif);
 180 #if IS_ENABLED(CONFIG_IPV6)
 181         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 182                 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
 183                          (int)ident, &ipv6_hdr(skb)->daddr, dif);
 184 #endif
 185         }
 186
 187         read_lock_bh(&ping_table.lock);
 188
 189         ping_portaddr_for_each_entry(sk, hnode, hslot) {
 190                 isk = inet_sk(sk);
 191
 192                 pr_debug("iterate\n");
 193                 if (isk->inet_num != ident)
 194                         continue;
 195
 196                 if (skb->protocol == htons(ETH_P_IP) &&
 197                     sk->sk_family == AF_INET) {
 198                         pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
 199                                  (int) isk->inet_num, &isk->inet_rcv_saddr,
 200                                  sk->sk_bound_dev_if);
 201
 202                         if (isk->inet_rcv_saddr &&
 203                             isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
 204                                 continue;
 205 #if IS_ENABLED(CONFIG_IPV6)
 206                 } else if (skb->protocol == htons(ETH_P_IPV6) &&
 207                            sk->sk_family == AF_INET6) {
 208
 209                         pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
 210                                  (int) isk->inet_num,
 211                                  &sk->sk_v6_rcv_saddr,
 212                                  sk->sk_bound_dev_if);
 213
 214                         if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
 215                             !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
 216                                              &ipv6_hdr(skb)->daddr))
 217                                 continue;
 218 #endif
 219                 } else {
 220                         continue;
 221                 }
 222
 223                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
 224                     sk->sk_bound_dev_if != inet_sdif(skb))
 225                         continue;
 226
 227                 sock_hold(sk);
 228                 goto exit;
 229         }
 230
 231         sk = NULL;
 232 exit:
 233         read_unlock_bh(&ping_table.lock);
 234
 235         return sk;
 236 }
 237
 238 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
 239                                           kgid_t *high)
 240 {
 241         kgid_t *data = net->ipv4.ping_group_range.range;
 242         unsigned int seq;
 243
 244         do {
 245                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 246
 247                 *low = data[0];
 248                 *high = data[1];
 249         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 250 }
 251
 252
 253 int ping_init_sock(struct sock *sk)
 254 {
 255         struct net *net = sock_net(sk);
 256         kgid_t group = current_egid();
 257         struct group_info *group_info;
 258         int i;
 259         kgid_t low, high;
 260         int ret = 0;
 261
 262         if (sk->sk_family == AF_INET6)
 263                 sk->sk_ipv6only = 1;
 264
 265         inet_get_ping_group_range_net(net, &low, &high);
 266         if (gid_lte(low, group) && gid_lte(group, high))
 267                 return 0;
 268
 269         group_info = get_current_groups();
 270         for (i = 0; i < group_info->ngroups; i++) {
 271                 kgid_t gid = group_info->gid[i];
 272
 273                 if (gid_lte(low, gid) && gid_lte(gid, high))
 274                         goto out_release_group;
 275         }
 276
 277         ret = -EACCES;
 278
 279 out_release_group:
 280         put_group_info(group_info);
 281         return ret;
 282 }
 283 EXPORT_SYMBOL_GPL(ping_init_sock);
 284
 285 void ping_close(struct sock *sk, long timeout)
 286 {
 287         pr_debug("ping_close(sk=%p,sk->num=%u)\n",
 288                  inet_sk(sk), inet_sk(sk)->inet_num);
 289         pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
 290
 291         sk_common_release(sk);
 292 }
 293 EXPORT_SYMBOL_GPL(ping_close);
 294
 295 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
 296 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
 297                                 struct sockaddr *uaddr, int addr_len)
 298 {
 299         struct net *net = sock_net(sk);
 300         if (sk->sk_family == AF_INET) {
 301                 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 302                 int chk_addr_ret;
 303
 304                 if (addr_len < sizeof(*addr))
 305                         return -EINVAL;
 306
 307                 if (addr->sin_family != AF_INET &&
 308                     !(addr->sin_family == AF_UNSPEC &&
 309                       addr->sin_addr.s_addr == htonl(INADDR_ANY)))
 310                         return -EAFNOSUPPORT;
 311
 312                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
 313                          sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
 314
 315                 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
 316
 317                 if (!inet_addr_valid_or_nonlocal(net, inet_sk(sk),
 318                                                  addr->sin_addr.s_addr,
 319                                                  chk_addr_ret))
 320                         return -EADDRNOTAVAIL;
 321
 322 #if IS_ENABLED(CONFIG_IPV6)
 323         } else if (sk->sk_family == AF_INET6) {
 324                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 325                 int addr_type, scoped, has_addr;
 326                 struct net_device *dev = NULL;
 327
 328                 if (addr_len < sizeof(*addr))
 329                         return -EINVAL;
 330
 331                 if (addr->sin6_family != AF_INET6)
 332                         return -EAFNOSUPPORT;
 333
 334                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 335                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 336
 337                 addr_type = ipv6_addr_type(&addr->sin6_addr);
 338                 scoped = __ipv6_addr_needs_scope_id(addr_type);
 339                 if ((addr_type != IPV6_ADDR_ANY &&
 340                      !(addr_type & IPV6_ADDR_UNICAST)) ||
 341                     (scoped && !addr->sin6_scope_id))
 342                         return -EINVAL;
 343
 344                 rcu_read_lock();
 345                 if (addr->sin6_scope_id) {
 346                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 347                         if (!dev) {
 348                                 rcu_read_unlock();
 349                                 return -ENODEV;
 350                         }
 351                 }
 352                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 353                                                     scoped);
 354                 rcu_read_unlock();
 355
 356                 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
 357                       addr_type == IPV6_ADDR_ANY))
 358                         return -EADDRNOTAVAIL;
 359
 360                 if (scoped)
 361                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 362 #endif
 363         } else {
 364                 return -EAFNOSUPPORT;
 365         }
 366         return 0;
 367 }
 368
 369 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
 370 {
 371         if (saddr->sa_family == AF_INET) {
 372                 struct inet_sock *isk = inet_sk(sk);
 373                 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
 374                 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
 375 #if IS_ENABLED(CONFIG_IPV6)
 376         } else if (saddr->sa_family == AF_INET6) {
 377                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
 378                 struct ipv6_pinfo *np = inet6_sk(sk);
 379                 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
 380 #endif
 381         }
 382 }
 383
 384 /*
 385  * We need our own bind because there are no privileged id's == local ports.
 386  * Moreover, we don't allow binding to multi- and broadcast addresses.
 387  */
 388
 389 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 390 {
 391         struct inet_sock *isk = inet_sk(sk);
 392         unsigned short snum;
 393         int err;
 394         int dif = sk->sk_bound_dev_if;
 395
 396         err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
 397         if (err)
 398                 return err;
 399
 400         lock_sock(sk);
 401
 402         err = -EINVAL;
 403         if (isk->inet_num != 0)
 404                 goto out;
 405
 406         err = -EADDRINUSE;
 407         snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
 408         if (ping_get_port(sk, snum) != 0) {
 409                 /* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */
 410                 sk->sk_bound_dev_if = dif;
 411                 goto out;
 412         }
 413         ping_set_saddr(sk, uaddr);
 414
 415         pr_debug("after bind(): num = %hu, dif = %d\n",
 416                  isk->inet_num,
 417                  sk->sk_bound_dev_if);
 418
 419         err = 0;
 420         if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
 421                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 422 #if IS_ENABLED(CONFIG_IPV6)
 423         if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 424                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 425 #endif
 426
 427         if (snum)
 428                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 429         isk->inet_sport = htons(isk->inet_num);
 430         isk->inet_daddr = 0;
 431         isk->inet_dport = 0;
 432
 433 #if IS_ENABLED(CONFIG_IPV6)
 434         if (sk->sk_family == AF_INET6)
 435                 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
 436 #endif
 437
 438         sk_dst_reset(sk);
 439 out:
 440         release_sock(sk);
 441         pr_debug("ping_v4_bind -> %d\n", err);
 442         return err;
 443 }
 444 EXPORT_SYMBOL_GPL(ping_bind);
 445
 446 /*
 447  * Is this a supported type of ICMP message?
 448  */
 449
 450 static inline int ping_supported(int family, int type, int code)
 451 {
 452         return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
 453                (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) ||
 454                (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) ||
 455                (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0);
 456 }
 457
 458 /*
 459  * This routine is called by the ICMP module when it gets some
 460  * sort of error condition.
 461  */
 462
 463 void ping_err(struct sk_buff *skb, int offset, u32 info)
 464 {
 465         int family;
 466         struct icmphdr *icmph;
 467         struct inet_sock *inet_sock;
 468         int type;
 469         int code;
 470         struct net *net = dev_net(skb->dev);
 471         struct sock *sk;
 472         int harderr;
 473         int err;
 474
 475         if (skb->protocol == htons(ETH_P_IP)) {
 476                 family = AF_INET;
 477                 type = icmp_hdr(skb)->type;
 478                 code = icmp_hdr(skb)->code;
 479                 icmph = (struct icmphdr *)(skb->data + offset);
 480         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 481                 family = AF_INET6;
 482                 type = icmp6_hdr(skb)->icmp6_type;
 483                 code = icmp6_hdr(skb)->icmp6_code;
 484                 icmph = (struct icmphdr *) (skb->data + offset);
 485         } else {
 486                 BUG();
 487         }
 488
 489         /* We assume the packet has already been checked by icmp_unreach */
 490
 491         if (!ping_supported(family, icmph->type, icmph->code))
 492                 return;
 493
 494         pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
 495                  skb->protocol, type, code, ntohs(icmph->un.echo.id),
 496                  ntohs(icmph->un.echo.sequence));
 497
 498         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 499         if (!sk) {
 500                 pr_debug("no socket, dropping\n");
 501                 return; /* No socket for error */
 502         }
 503         pr_debug("err on socket %p\n", sk);
 504
 505         err = 0;
 506         harderr = 0;
 507         inet_sock = inet_sk(sk);
 508
 509         if (skb->protocol == htons(ETH_P_IP)) {
 510                 switch (type) {
 511                 default:
 512                 case ICMP_TIME_EXCEEDED:
 513                         err = EHOSTUNREACH;
 514                         break;
 515                 case ICMP_SOURCE_QUENCH:
 516                         /* This is not a real error but ping wants to see it.
 517                          * Report it with some fake errno.
 518                          */
 519                         err = EREMOTEIO;
 520                         break;
 521                 case ICMP_PARAMETERPROB:
 522                         err = EPROTO;
 523                         harderr = 1;
 524                         break;
 525                 case ICMP_DEST_UNREACH:
 526                         if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
 527                                 ipv4_sk_update_pmtu(skb, sk, info);
 528                                 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
 529                                         err = EMSGSIZE;
 530                                         harderr = 1;
 531                                         break;
 532                                 }
 533                                 goto out;
 534                         }
 535                         err = EHOSTUNREACH;
 536                         if (code <= NR_ICMP_UNREACH) {
 537                                 harderr = icmp_err_convert[code].fatal;
 538                                 err = icmp_err_convert[code].errno;
 539                         }
 540                         break;
 541                 case ICMP_REDIRECT:
 542                         /* See ICMP_SOURCE_QUENCH */
 543                         ipv4_sk_redirect(skb, sk);
 544                         err = EREMOTEIO;
 545                         break;
 546                 }
 547 #if IS_ENABLED(CONFIG_IPV6)
 548         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 549                 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
 550 #endif
 551         }
 552
 553         /*
 554          *      RFC1122: OK.  Passes ICMP errors back to application, as per
 555          *      4.1.3.3.
 556          */
 557         if ((family == AF_INET && !inet_sock->recverr) ||
 558             (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
 559                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 560                         goto out;
 561         } else {
 562                 if (family == AF_INET) {
 563                         ip_icmp_error(sk, skb, err, 0 /* no remote port */,
 564                                       info, (u8 *)icmph);
 565 #if IS_ENABLED(CONFIG_IPV6)
 566                 } else if (family == AF_INET6) {
 567                         pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
 568                                                    info, (u8 *)icmph);
 569 #endif
 570                 }
 571         }
 572         sk->sk_err = err;
 573         sk_error_report(sk);
 574 out:
 575         sock_put(sk);
 576 }
 577 EXPORT_SYMBOL_GPL(ping_err);
 578
 579 /*
 580  *      Copy and checksum an ICMP Echo packet from user space into a buffer
 581  *      starting from the payload.
 582  */
 583
 584 int ping_getfrag(void *from, char *to,
 585                  int offset, int fraglen, int odd, struct sk_buff *skb)
 586 {
 587         struct pingfakehdr *pfh = (struct pingfakehdr *)from;
 588
 589         if (offset == 0) {
 590                 fraglen -= sizeof(struct icmphdr);
 591                 if (fraglen < 0)
 592                         BUG();
 593                 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
 594                             fraglen, &pfh->wcheck,
 595                             &pfh->msg->msg_iter))
 596                         return -EFAULT;
 597         } else if (offset < sizeof(struct icmphdr)) {
 598                         BUG();
 599         } else {
 600                 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
 601                                             &pfh->msg->msg_iter))
 602                         return -EFAULT;
 603         }
 604
 605 #if IS_ENABLED(CONFIG_IPV6)
 606         /* For IPv6, checksum each skb as we go along, as expected by
 607          * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
 608          * wcheck, it will be finalized in ping_v4_push_pending_frames.
 609          */
 610         if (pfh->family == AF_INET6) {
 611                 skb->csum = pfh->wcheck;
 612                 skb->ip_summed = CHECKSUM_NONE;
 613                 pfh->wcheck = 0;
 614         }
 615 #endif
 616
 617         return 0;
 618 }
 619 EXPORT_SYMBOL_GPL(ping_getfrag);
 620
 621 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
 622                                        struct flowi4 *fl4)
 623 {
 624         struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
 625
 626         if (!skb)
 627                 return 0;
 628         pfh->wcheck = csum_partial((char *)&pfh->icmph,
 629                 sizeof(struct icmphdr), pfh->wcheck);
 630         pfh->icmph.checksum = csum_fold(pfh->wcheck);
 631         memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
 632         skb->ip_summed = CHECKSUM_NONE;
 633         return ip_push_pending_frames(sk, fl4);
 634 }
 635
 636 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
 637                         void *user_icmph, size_t icmph_len)
 638 {
 639         u8 type, code;
 640
 641         if (len > 0xFFFF)
 642                 return -EMSGSIZE;
 643
 644         /* Must have at least a full ICMP header. */
 645         if (len < icmph_len)
 646                 return -EINVAL;
 647
 648         /*
 649          *      Check the flags.
 650          */
 651
 652         /* Mirror BSD error message compatibility */
 653         if (msg->msg_flags & MSG_OOB)
 654                 return -EOPNOTSUPP;
 655
 656         /*
 657          *      Fetch the ICMP header provided by the userland.
 658          *      iovec is modified! The ICMP header is consumed.
 659          */
 660         if (memcpy_from_msg(user_icmph, msg, icmph_len))
 661                 return -EFAULT;
 662
 663         if (family == AF_INET) {
 664                 type = ((struct icmphdr *) user_icmph)->type;
 665                 code = ((struct icmphdr *) user_icmph)->code;
 666 #if IS_ENABLED(CONFIG_IPV6)
 667         } else if (family == AF_INET6) {
 668                 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
 669                 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
 670 #endif
 671         } else {
 672                 BUG();
 673         }
 674
 675         if (!ping_supported(family, type, code))
 676                 return -EINVAL;
 677
 678         return 0;
 679 }
 680 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
 681
 682 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 683 {
 684         struct net *net = sock_net(sk);
 685         struct flowi4 fl4;
 686         struct inet_sock *inet = inet_sk(sk);
 687         struct ipcm_cookie ipc;
 688         struct icmphdr user_icmph;
 689         struct pingfakehdr pfh;
 690         struct rtable *rt = NULL;
 691         struct ip_options_data opt_copy;
 692         int free = 0;
 693         __be32 saddr, daddr, faddr;
 694         u8  tos;
 695         int err;
 696
 697         pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
 698
 699         err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
 700                                   sizeof(user_icmph));
 701         if (err)
 702                 return err;
 703
 704         /*
 705          *      Get and verify the address.
 706          */
 707
 708         if (msg->msg_name) {
 709                 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
 710                 if (msg->msg_namelen < sizeof(*usin))
 711                         return -EINVAL;
 712                 if (usin->sin_family != AF_INET)
 713                         return -EAFNOSUPPORT;
 714                 daddr = usin->sin_addr.s_addr;
 715                 /* no remote port */
 716         } else {
 717                 if (sk->sk_state != TCP_ESTABLISHED)
 718                         return -EDESTADDRREQ;
 719                 daddr = inet->inet_daddr;
 720                 /* no remote port */
 721         }
 722
 723         ipcm_init_sk(&ipc, inet);
 724
 725         if (msg->msg_controllen) {
 726                 err = ip_cmsg_send(sk, msg, &ipc, false);
 727                 if (unlikely(err)) {
 728                         kfree(ipc.opt);
 729                         return err;
 730                 }
 731                 if (ipc.opt)
 732                         free = 1;
 733         }
 734         if (!ipc.opt) {
 735                 struct ip_options_rcu *inet_opt;
 736
 737                 rcu_read_lock();
 738                 inet_opt = rcu_dereference(inet->inet_opt);
 739                 if (inet_opt) {
 740                         memcpy(&opt_copy, inet_opt,
 741                                sizeof(*inet_opt) + inet_opt->opt.optlen);
 742                         ipc.opt = &opt_copy.opt;
 743                 }
 744                 rcu_read_unlock();
 745         }
 746
 747         saddr = ipc.addr;
 748         ipc.addr = faddr = daddr;
 749
 750         if (ipc.opt && ipc.opt->opt.srr) {
 751                 if (!daddr) {
 752                         err = -EINVAL;
 753                         goto out_free;
 754                 }
 755                 faddr = ipc.opt->opt.faddr;
 756         }
 757         tos = get_rttos(&ipc, inet);
 758         if (sock_flag(sk, SOCK_LOCALROUTE) ||
 759             (msg->msg_flags & MSG_DONTROUTE) ||
 760             (ipc.opt && ipc.opt->opt.is_strictroute)) {
 761                 tos |= RTO_ONLINK;
 762         }
 763
 764         if (ipv4_is_multicast(daddr)) {
 765                 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
 766                         ipc.oif = inet->mc_index;
 767                 if (!saddr)
 768                         saddr = inet->mc_addr;
 769         } else if (!ipc.oif)
 770                 ipc.oif = inet->uc_index;
 771
 772         flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
 773                            RT_SCOPE_UNIVERSE, sk->sk_protocol,
 774                            inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
 775                            sk->sk_uid);
 776
 777         fl4.fl4_icmp_type = user_icmph.type;
 778         fl4.fl4_icmp_code = user_icmph.code;
 779
 780         security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4));
 781         rt = ip_route_output_flow(net, &fl4, sk);
 782         if (IS_ERR(rt)) {
 783                 err = PTR_ERR(rt);
 784                 rt = NULL;
 785                 if (err == -ENETUNREACH)
 786                         IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
 787                 goto out;
 788         }
 789
 790         err = -EACCES;
 791         if ((rt->rt_flags & RTCF_BROADCAST) &&
 792             !sock_flag(sk, SOCK_BROADCAST))
 793                 goto out;
 794
 795         if (msg->msg_flags & MSG_CONFIRM)
 796                 goto do_confirm;
 797 back_from_confirm:
 798
 799         if (!ipc.addr)
 800                 ipc.addr = fl4.daddr;
 801
 802         lock_sock(sk);
 803
 804         pfh.icmph.type = user_icmph.type; /* already checked */
 805         pfh.icmph.code = user_icmph.code; /* ditto */
 806         pfh.icmph.checksum = 0;
 807         pfh.icmph.un.echo.id = inet->inet_sport;
 808         pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 809         pfh.msg = msg;
 810         pfh.wcheck = 0;
 811         pfh.family = AF_INET;
 812
 813         err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 814                         0, &ipc, &rt, msg->msg_flags);
 815         if (err)
 816                 ip_flush_pending_frames(sk);
 817         else
 818                 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 819         release_sock(sk);
 820
 821 out:
 822         ip_rt_put(rt);
 823 out_free:
 824         if (free)
 825                 kfree(ipc.opt);
 826         if (!err) {
 827                 icmp_out_count(sock_net(sk), user_icmph.type);
 828                 return len;
 829         }
 830         return err;
 831
 832 do_confirm:
 833         if (msg->msg_flags & MSG_PROBE)
 834                 dst_confirm_neigh(&rt->dst, &fl4.daddr);
 835         if (!(msg->msg_flags & MSG_PROBE) || len)
 836                 goto back_from_confirm;
 837         err = 0;
 838         goto out;
 839 }
 840
 841 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
 842                  int flags, int *addr_len)
 843 {
 844         struct inet_sock *isk = inet_sk(sk);
 845         int family = sk->sk_family;
 846         struct sk_buff *skb;
 847         int copied, err;
 848
 849         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 850
 851         err = -EOPNOTSUPP;
 852         if (flags & MSG_OOB)
 853                 goto out;
 854
 855         if (flags & MSG_ERRQUEUE)
 856                 return inet_recv_error(sk, msg, len, addr_len);
 857
 858         skb = skb_recv_datagram(sk, flags, noblock, &err);
 859         if (!skb)
 860                 goto out;
 861
 862         copied = skb->len;
 863         if (copied > len) {
 864                 msg->msg_flags |= MSG_TRUNC;
 865                 copied = len;
 866         }
 867
 868         /* Don't bother checking the checksum */
 869         err = skb_copy_datagram_msg(skb, 0, msg, copied);
 870         if (err)
 871                 goto done;
 872
 873         sock_recv_timestamp(msg, sk, skb);
 874
 875         /* Copy the address and add cmsg data. */
 876         if (family == AF_INET) {
 877                 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 878
 879                 if (sin) {
 880                         sin->sin_family = AF_INET;
 881                         sin->sin_port = 0 /* skb->h.uh->source */;
 882                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 883                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 884                         *addr_len = sizeof(*sin);
 885                 }
 886
 887                 if (isk->cmsg_flags)
 888                         ip_cmsg_recv(msg, skb);
 889
 890 #if IS_ENABLED(CONFIG_IPV6)
 891         } else if (family == AF_INET6) {
 892                 struct ipv6_pinfo *np = inet6_sk(sk);
 893                 struct ipv6hdr *ip6 = ipv6_hdr(skb);
 894                 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 895
 896                 if (sin6) {
 897                         sin6->sin6_family = AF_INET6;
 898                         sin6->sin6_port = 0;
 899                         sin6->sin6_addr = ip6->saddr;
 900                         sin6->sin6_flowinfo = 0;
 901                         if (np->sndflow)
 902                                 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
 903                         sin6->sin6_scope_id =
 904                                 ipv6_iface_scope_id(&sin6->sin6_addr,
 905                                                     inet6_iif(skb));
 906                         *addr_len = sizeof(*sin6);
 907                 }
 908
 909                 if (inet6_sk(sk)->rxopt.all)
 910                         pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
 911                 if (skb->protocol == htons(ETH_P_IPV6) &&
 912                     inet6_sk(sk)->rxopt.all)
 913                         pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
 914                 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
 915                         ip_cmsg_recv(msg, skb);
 916 #endif
 917         } else {
 918                 BUG();
 919         }
 920
 921         err = copied;
 922
 923 done:
 924         skb_free_datagram(sk, skb);
 925 out:
 926         pr_debug("ping_recvmsg -> %d\n", err);
 927         return err;
 928 }
 929 EXPORT_SYMBOL_GPL(ping_recvmsg);
 930
 931 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 932 {
 933         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 934                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
 935         if (sock_queue_rcv_skb(sk, skb) < 0) {
 936                 kfree_skb(skb);
 937                 pr_debug("ping_queue_rcv_skb -> failed\n");
 938                 return -1;
 939         }
 940         return 0;
 941 }
 942 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 943
 944
 945 /*
 946  *      All we need to do is get the socket.
 947  */
 948
 949 bool ping_rcv(struct sk_buff *skb)
 950 {
 951         struct sock *sk;
 952         struct net *net = dev_net(skb->dev);
 953         struct icmphdr *icmph = icmp_hdr(skb);
 954         bool rc = false;
 955
 956         /* We assume the packet has already been checked by icmp_rcv */
 957
 958         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 959                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 960
 961         /* Push ICMP header back */
 962         skb_push(skb, skb->data - (u8 *)icmph);
 963
 964         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 965         if (sk) {
 966                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 967
 968                 pr_debug("rcv on socket %p\n", sk);
 969                 if (skb2 && !ping_queue_rcv_skb(sk, skb2))
 970                         rc = true;
 971                 sock_put(sk);
 972         }
 973
 974         if (!rc)
 975                 pr_debug("no socket, dropping\n");
 976
 977         return rc;
 978 }
 979 EXPORT_SYMBOL_GPL(ping_rcv);
 980
 981 struct proto ping_prot = {
 982         .name =         "PING",
 983         .owner =        THIS_MODULE,
 984         .init =         ping_init_sock,
 985         .close =        ping_close,
 986         .connect =      ip4_datagram_connect,
 987         .disconnect =   __udp_disconnect,
 988         .setsockopt =   ip_setsockopt,
 989         .getsockopt =   ip_getsockopt,
 990         .sendmsg =      ping_v4_sendmsg,
 991         .recvmsg =      ping_recvmsg,
 992         .bind =         ping_bind,
 993         .backlog_rcv =  ping_queue_rcv_skb,
 994         .release_cb =   ip4_datagram_release_cb,
 995         .hash =         ping_hash,
 996         .unhash =       ping_unhash,
 997         .get_port =     ping_get_port,
 998         .put_port =     ping_unhash,
 999         .obj_size =     sizeof(struct inet_sock),
1000 };
1001 EXPORT_SYMBOL(ping_prot);
1002
1003 #ifdef CONFIG_PROC_FS
1004
1005 static struct sock *ping_get_first(struct seq_file *seq, int start)
1006 {
1007         struct sock *sk;
1008         struct ping_iter_state *state = seq->private;
1009         struct net *net = seq_file_net(seq);
1010
1011         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1012              ++state->bucket) {
1013                 struct hlist_nulls_node *node;
1014                 struct hlist_nulls_head *hslot;
1015
1016                 hslot = &ping_table.hash[state->bucket];
1017
1018                 if (hlist_nulls_empty(hslot))
1019                         continue;
1020
1021                 sk_nulls_for_each(sk, node, hslot) {
1022                         if (net_eq(sock_net(sk), net) &&
1023                             sk->sk_family == state->family)
1024                                 goto found;
1025                 }
1026         }
1027         sk = NULL;
1028 found:
1029         return sk;
1030 }
1031
1032 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1033 {
1034         struct ping_iter_state *state = seq->private;
1035         struct net *net = seq_file_net(seq);
1036
1037         do {
1038                 sk = sk_nulls_next(sk);
1039         } while (sk && (!net_eq(sock_net(sk), net)));
1040
1041         if (!sk)
1042                 return ping_get_first(seq, state->bucket + 1);
1043         return sk;
1044 }
1045
1046 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1047 {
1048         struct sock *sk = ping_get_first(seq, 0);
1049
1050         if (sk)
1051                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1052                         --pos;
1053         return pos ? NULL : sk;
1054 }
1055
1056 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1057         __acquires(ping_table.lock)
1058 {
1059         struct ping_iter_state *state = seq->private;
1060         state->bucket = 0;
1061         state->family = family;
1062
1063         read_lock_bh(&ping_table.lock);
1064
1065         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1066 }
1067 EXPORT_SYMBOL_GPL(ping_seq_start);
1068
1069 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1070 {
1071         return ping_seq_start(seq, pos, AF_INET);
1072 }
1073
1074 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1075 {
1076         struct sock *sk;
1077
1078         if (v == SEQ_START_TOKEN)
1079                 sk = ping_get_idx(seq, 0);
1080         else
1081                 sk = ping_get_next(seq, v);
1082
1083         ++*pos;
1084         return sk;
1085 }
1086 EXPORT_SYMBOL_GPL(ping_seq_next);
1087
1088 void ping_seq_stop(struct seq_file *seq, void *v)
1089         __releases(ping_table.lock)
1090 {
1091         read_unlock_bh(&ping_table.lock);
1092 }
1093 EXPORT_SYMBOL_GPL(ping_seq_stop);
1094
1095 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1096                 int bucket)
1097 {
1098         struct inet_sock *inet = inet_sk(sp);
1099         __be32 dest = inet->inet_daddr;
1100         __be32 src = inet->inet_rcv_saddr;
1101         __u16 destp = ntohs(inet->inet_dport);
1102         __u16 srcp = ntohs(inet->inet_sport);
1103
1104         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1105                 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1106                 bucket, src, srcp, dest, destp, sp->sk_state,
1107                 sk_wmem_alloc_get(sp),
1108                 sk_rmem_alloc_get(sp),
1109                 0, 0L, 0,
1110                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1111                 0, sock_i_ino(sp),
1112                 refcount_read(&sp->sk_refcnt), sp,
1113                 atomic_read(&sp->sk_drops));
1114 }
1115
1116 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1117 {
1118         seq_setwidth(seq, 127);
1119         if (v == SEQ_START_TOKEN)
1120                 seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1121                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1122                            "inode ref pointer drops");
1123         else {
1124                 struct ping_iter_state *state = seq->private;
1125
1126                 ping_v4_format_sock(v, seq, state->bucket);
1127         }
1128         seq_pad(seq, '\n');
1129         return 0;
1130 }
1131
1132 static const struct seq_operations ping_v4_seq_ops = {
1133         .start          = ping_v4_seq_start,
1134         .show           = ping_v4_seq_show,
1135         .next           = ping_seq_next,
1136         .stop           = ping_seq_stop,
1137 };
1138
1139 static int __net_init ping_v4_proc_init_net(struct net *net)
1140 {
1141         if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1142                         sizeof(struct ping_iter_state)))
1143                 return -ENOMEM;
1144         return 0;
1145 }
1146
1147 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1148 {
1149         remove_proc_entry("icmp", net->proc_net);
1150 }
1151
1152 static struct pernet_operations ping_v4_net_ops = {
1153         .init = ping_v4_proc_init_net,
1154         .exit = ping_v4_proc_exit_net,
1155 };
1156
1157 int __init ping_proc_init(void)
1158 {
1159         return register_pernet_subsys(&ping_v4_net_ops);
1160 }
1161
1162 void ping_proc_exit(void)
1163 {
1164         unregister_pernet_subsys(&ping_v4_net_ops);
1165 }
1166
1167 #endif
1168
1169 void __init ping_init(void)
1170 {
1171         int i;
1172
1173         for (i = 0; i < PING_HTABLE_SIZE; i++)
1174                 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1175         rwlock_init(&ping_table.lock);
1176 }