net/ipv4/ping.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              "Ping" sockets
   7  *
   8  *              This program is free software; you can redistribute it and/or
   9  *              modify it under the terms of the GNU General Public License
  10  *              as published by the Free Software Foundation; either version
  11  *              2 of the License, or (at your option) any later version.
  12  *
  13  * Based on ipv4/udp.c code.
  14  *
  15  * Authors:     Vasiliy Kulikov / Openwall (for Linux 2.6),
  16  *              Pavel Kankovsky (for Linux 2.4.32)
  17  *
  18  * Pavel gave all rights to bugs to Vasiliy,
  19  * none of the bugs are Pavel's now.
  20  *
  21  */
  22
  23 #include <linux/uaccess.h>
  24 #include <linux/types.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/socket.h>
  27 #include <linux/sockios.h>
  28 #include <linux/in.h>
  29 #include <linux/errno.h>
  30 #include <linux/timer.h>
  31 #include <linux/mm.h>
  32 #include <linux/inet.h>
  33 #include <linux/netdevice.h>
  34 #include <net/snmp.h>
  35 #include <net/ip.h>
  36 #include <net/icmp.h>
  37 #include <net/protocol.h>
  38 #include <linux/skbuff.h>
  39 #include <linux/proc_fs.h>
  40 #include <linux/export.h>
  41 #include <net/sock.h>
  42 #include <net/ping.h>
  43 #include <net/udp.h>
  44 #include <net/route.h>
  45 #include <net/inet_common.h>
  46 #include <net/checksum.h>
  47
  48 #if IS_ENABLED(CONFIG_IPV6)
  49 #include <linux/in6.h>
  50 #include <linux/icmpv6.h>
  51 #include <net/addrconf.h>
  52 #include <net/ipv6.h>
  53 #include <net/transp_v6.h>
  54 #endif
  55
  56 struct ping_table {
  57         struct hlist_nulls_head hash[PING_HTABLE_SIZE];
  58         rwlock_t                lock;
  59 };
  60
  61 static struct ping_table ping_table;
  62 struct pingv6_ops pingv6_ops;
  63 EXPORT_SYMBOL_GPL(pingv6_ops);
  64
  65 static u16 ping_port_rover;
  66
  67 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
  68 {
  69         u32 res = (num + net_hash_mix(net)) & mask;
  70
  71         pr_debug("hash(%u) = %u\n", num, res);
  72         return res;
  73 }
  74 EXPORT_SYMBOL_GPL(ping_hash);
  75
  76 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
  77                                              struct net *net, unsigned int num)
  78 {
  79         return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
  80 }
  81
  82 int ping_get_port(struct sock *sk, unsigned short ident)
  83 {
  84         struct hlist_nulls_node *node;
  85         struct hlist_nulls_head *hlist;
  86         struct inet_sock *isk, *isk2;
  87         struct sock *sk2 = NULL;
  88
  89         isk = inet_sk(sk);
  90         write_lock_bh(&ping_table.lock);
  91         if (ident == 0) {
  92                 u32 i;
  93                 u16 result = ping_port_rover + 1;
  94
  95                 for (i = 0; i < (1L << 16); i++, result++) {
  96                         if (!result)
  97                                 result++; /* avoid zero */
  98                         hlist = ping_hashslot(&ping_table, sock_net(sk),
  99                                             result);
 100                         ping_portaddr_for_each_entry(sk2, node, hlist) {
 101                                 isk2 = inet_sk(sk2);
 102
 103                                 if (isk2->inet_num == result)
 104                                         goto next_port;
 105                         }
 106
 107                         /* found */
 108                         ping_port_rover = ident = result;
 109                         break;
 110 next_port:
 111                         ;
 112                 }
 113                 if (i >= (1L << 16))
 114                         goto fail;
 115         } else {
 116                 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
 117                 ping_portaddr_for_each_entry(sk2, node, hlist) {
 118                         isk2 = inet_sk(sk2);
 119
 120                         /* BUG? Why is this reuse and not reuseaddr? ping.c
 121                          * doesn't turn off SO_REUSEADDR, and it doesn't expect
 122                          * that other ping processes can steal its packets.
 123                          */
 124                         if ((isk2->inet_num == ident) &&
 125                             (sk2 != sk) &&
 126                             (!sk2->sk_reuse || !sk->sk_reuse))
 127                                 goto fail;
 128                 }
 129         }
 130
 131         pr_debug("found port/ident = %d\n", ident);
 132         isk->inet_num = ident;
 133         if (sk_unhashed(sk)) {
 134                 pr_debug("was not hashed\n");
 135                 sock_hold(sk);
 136                 hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
 137                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
 138         }
 139         write_unlock_bh(&ping_table.lock);
 140         return 0;
 141
 142 fail:
 143         write_unlock_bh(&ping_table.lock);
 144         return 1;
 145 }
 146 EXPORT_SYMBOL_GPL(ping_get_port);
 147
 148 int ping_hash(struct sock *sk)
 149 {
 150         pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
 151         BUG(); /* "Please do not press this button again." */
 152
 153         return 0;
 154 }
 155
 156 void ping_unhash(struct sock *sk)
 157 {
 158         struct inet_sock *isk = inet_sk(sk);
 159
 160         pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
 161         write_lock_bh(&ping_table.lock);
 162         if (sk_hashed(sk)) {
 163                 hlist_nulls_del(&sk->sk_nulls_node);
 164                 sk_nulls_node_init(&sk->sk_nulls_node);
 165                 sock_put(sk);
 166                 isk->inet_num = 0;
 167                 isk->inet_sport = 0;
 168                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 169         }
 170         write_unlock_bh(&ping_table.lock);
 171 }
 172 EXPORT_SYMBOL_GPL(ping_unhash);
 173
 174 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
 175 {
 176         struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
 177         struct sock *sk = NULL;
 178         struct inet_sock *isk;
 179         struct hlist_nulls_node *hnode;
 180         int dif = skb->dev->ifindex;
 181
 182         if (skb->protocol == htons(ETH_P_IP)) {
 183                 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
 184                          (int)ident, &ip_hdr(skb)->daddr, dif);
 185 #if IS_ENABLED(CONFIG_IPV6)
 186         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 187                 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
 188                          (int)ident, &ipv6_hdr(skb)->daddr, dif);
 189 #endif
 190         }
 191
 192         read_lock_bh(&ping_table.lock);
 193
 194         ping_portaddr_for_each_entry(sk, hnode, hslot) {
 195                 isk = inet_sk(sk);
 196
 197                 pr_debug("iterate\n");
 198                 if (isk->inet_num != ident)
 199                         continue;
 200
 201                 if (skb->protocol == htons(ETH_P_IP) &&
 202                     sk->sk_family == AF_INET) {
 203                         pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
 204                                  (int) isk->inet_num, &isk->inet_rcv_saddr,
 205                                  sk->sk_bound_dev_if);
 206
 207                         if (isk->inet_rcv_saddr &&
 208                             isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
 209                                 continue;
 210 #if IS_ENABLED(CONFIG_IPV6)
 211                 } else if (skb->protocol == htons(ETH_P_IPV6) &&
 212                            sk->sk_family == AF_INET6) {
 213
 214                         pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
 215                                  (int) isk->inet_num,
 216                                  &sk->sk_v6_rcv_saddr,
 217                                  sk->sk_bound_dev_if);
 218
 219                         if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
 220                             !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
 221                                              &ipv6_hdr(skb)->daddr))
 222                                 continue;
 223 #endif
 224                 } else {
 225                         continue;
 226                 }
 227
 228                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
 229                     sk->sk_bound_dev_if != inet_sdif(skb))
 230                         continue;
 231
 232                 sock_hold(sk);
 233                 goto exit;
 234         }
 235
 236         sk = NULL;
 237 exit:
 238         read_unlock_bh(&ping_table.lock);
 239
 240         return sk;
 241 }
 242
 243 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
 244                                           kgid_t *high)
 245 {
 246         kgid_t *data = net->ipv4.ping_group_range.range;
 247         unsigned int seq;
 248
 249         do {
 250                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 251
 252                 *low = data[0];
 253                 *high = data[1];
 254         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 255 }
 256
 257
 258 int ping_init_sock(struct sock *sk)
 259 {
 260         struct net *net = sock_net(sk);
 261         kgid_t group = current_egid();
 262         struct group_info *group_info;
 263         int i;
 264         kgid_t low, high;
 265         int ret = 0;
 266
 267         if (sk->sk_family == AF_INET6)
 268                 sk->sk_ipv6only = 1;
 269
 270         inet_get_ping_group_range_net(net, &low, &high);
 271         if (gid_lte(low, group) && gid_lte(group, high))
 272                 return 0;
 273
 274         group_info = get_current_groups();
 275         for (i = 0; i < group_info->ngroups; i++) {
 276                 kgid_t gid = group_info->gid[i];
 277
 278                 if (gid_lte(low, gid) && gid_lte(gid, high))
 279                         goto out_release_group;
 280         }
 281
 282         ret = -EACCES;
 283
 284 out_release_group:
 285         put_group_info(group_info);
 286         return ret;
 287 }
 288 EXPORT_SYMBOL_GPL(ping_init_sock);
 289
 290 void ping_close(struct sock *sk, long timeout)
 291 {
 292         pr_debug("ping_close(sk=%p,sk->num=%u)\n",
 293                  inet_sk(sk), inet_sk(sk)->inet_num);
 294         pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
 295
 296         sk_common_release(sk);
 297 }
 298 EXPORT_SYMBOL_GPL(ping_close);
 299
 300 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
 301 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
 302                                 struct sockaddr *uaddr, int addr_len) {
 303         struct net *net = sock_net(sk);
 304         if (sk->sk_family == AF_INET) {
 305                 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 306                 int chk_addr_ret;
 307
 308                 if (addr_len < sizeof(*addr))
 309                         return -EINVAL;
 310
 311                 if (addr->sin_family != AF_INET &&
 312                     !(addr->sin_family == AF_UNSPEC &&
 313                       addr->sin_addr.s_addr == htonl(INADDR_ANY)))
 314                         return -EAFNOSUPPORT;
 315
 316                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
 317                          sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
 318
 319                 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
 320
 321                 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
 322                         chk_addr_ret = RTN_LOCAL;
 323
 324                 if ((net->ipv4.sysctl_ip_nonlocal_bind == 0 &&
 325                     isk->freebind == 0 && isk->transparent == 0 &&
 326                      chk_addr_ret != RTN_LOCAL) ||
 327                     chk_addr_ret == RTN_MULTICAST ||
 328                     chk_addr_ret == RTN_BROADCAST)
 329                         return -EADDRNOTAVAIL;
 330
 331 #if IS_ENABLED(CONFIG_IPV6)
 332         } else if (sk->sk_family == AF_INET6) {
 333                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 334                 int addr_type, scoped, has_addr;
 335                 struct net_device *dev = NULL;
 336
 337                 if (addr_len < sizeof(*addr))
 338                         return -EINVAL;
 339
 340                 if (addr->sin6_family != AF_INET6)
 341                         return -EAFNOSUPPORT;
 342
 343                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 344                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 345
 346                 addr_type = ipv6_addr_type(&addr->sin6_addr);
 347                 scoped = __ipv6_addr_needs_scope_id(addr_type);
 348                 if ((addr_type != IPV6_ADDR_ANY &&
 349                      !(addr_type & IPV6_ADDR_UNICAST)) ||
 350                     (scoped && !addr->sin6_scope_id))
 351                         return -EINVAL;
 352
 353                 rcu_read_lock();
 354                 if (addr->sin6_scope_id) {
 355                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 356                         if (!dev) {
 357                                 rcu_read_unlock();
 358                                 return -ENODEV;
 359                         }
 360                 }
 361                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 362                                                     scoped);
 363                 rcu_read_unlock();
 364
 365                 if (!(net->ipv6.sysctl.ip_nonlocal_bind ||
 366                       isk->freebind || isk->transparent || has_addr ||
 367                       addr_type == IPV6_ADDR_ANY))
 368                         return -EADDRNOTAVAIL;
 369
 370                 if (scoped)
 371                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 372 #endif
 373         } else {
 374                 return -EAFNOSUPPORT;
 375         }
 376         return 0;
 377 }
 378
 379 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
 380 {
 381         if (saddr->sa_family == AF_INET) {
 382                 struct inet_sock *isk = inet_sk(sk);
 383                 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
 384                 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
 385 #if IS_ENABLED(CONFIG_IPV6)
 386         } else if (saddr->sa_family == AF_INET6) {
 387                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
 388                 struct ipv6_pinfo *np = inet6_sk(sk);
 389                 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
 390 #endif
 391         }
 392 }
 393
 394 static void ping_clear_saddr(struct sock *sk, int dif)
 395 {
 396         sk->sk_bound_dev_if = dif;
 397         if (sk->sk_family == AF_INET) {
 398                 struct inet_sock *isk = inet_sk(sk);
 399                 isk->inet_rcv_saddr = isk->inet_saddr = 0;
 400 #if IS_ENABLED(CONFIG_IPV6)
 401         } else if (sk->sk_family == AF_INET6) {
 402                 struct ipv6_pinfo *np = inet6_sk(sk);
 403                 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
 404                 memset(&np->saddr, 0, sizeof(np->saddr));
 405 #endif
 406         }
 407 }
 408 /*
 409  * We need our own bind because there are no privileged id's == local ports.
 410  * Moreover, we don't allow binding to multi- and broadcast addresses.
 411  */
 412
 413 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 414 {
 415         struct inet_sock *isk = inet_sk(sk);
 416         unsigned short snum;
 417         int err;
 418         int dif = sk->sk_bound_dev_if;
 419
 420         err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
 421         if (err)
 422                 return err;
 423
 424         lock_sock(sk);
 425
 426         err = -EINVAL;
 427         if (isk->inet_num != 0)
 428                 goto out;
 429
 430         err = -EADDRINUSE;
 431         ping_set_saddr(sk, uaddr);
 432         snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
 433         if (ping_get_port(sk, snum) != 0) {
 434                 ping_clear_saddr(sk, dif);
 435                 goto out;
 436         }
 437
 438         pr_debug("after bind(): num = %hu, dif = %d\n",
 439                  isk->inet_num,
 440                  sk->sk_bound_dev_if);
 441
 442         err = 0;
 443         if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
 444                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 445 #if IS_ENABLED(CONFIG_IPV6)
 446         if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 447                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 448 #endif
 449
 450         if (snum)
 451                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 452         isk->inet_sport = htons(isk->inet_num);
 453         isk->inet_daddr = 0;
 454         isk->inet_dport = 0;
 455
 456 #if IS_ENABLED(CONFIG_IPV6)
 457         if (sk->sk_family == AF_INET6)
 458                 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
 459 #endif
 460
 461         sk_dst_reset(sk);
 462 out:
 463         release_sock(sk);
 464         pr_debug("ping_v4_bind -> %d\n", err);
 465         return err;
 466 }
 467 EXPORT_SYMBOL_GPL(ping_bind);
 468
 469 /*
 470  * Is this a supported type of ICMP message?
 471  */
 472
 473 static inline int ping_supported(int family, int type, int code)
 474 {
 475         return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
 476                (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
 477 }
 478
 479 /*
 480  * This routine is called by the ICMP module when it gets some
 481  * sort of error condition.
 482  */
 483
 484 void ping_err(struct sk_buff *skb, int offset, u32 info)
 485 {
 486         int family;
 487         struct icmphdr *icmph;
 488         struct inet_sock *inet_sock;
 489         int type;
 490         int code;
 491         struct net *net = dev_net(skb->dev);
 492         struct sock *sk;
 493         int harderr;
 494         int err;
 495
 496         if (skb->protocol == htons(ETH_P_IP)) {
 497                 family = AF_INET;
 498                 type = icmp_hdr(skb)->type;
 499                 code = icmp_hdr(skb)->code;
 500                 icmph = (struct icmphdr *)(skb->data + offset);
 501         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 502                 family = AF_INET6;
 503                 type = icmp6_hdr(skb)->icmp6_type;
 504                 code = icmp6_hdr(skb)->icmp6_code;
 505                 icmph = (struct icmphdr *) (skb->data + offset);
 506         } else {
 507                 BUG();
 508         }
 509
 510         /* We assume the packet has already been checked by icmp_unreach */
 511
 512         if (!ping_supported(family, icmph->type, icmph->code))
 513                 return;
 514
 515         pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
 516                  skb->protocol, type, code, ntohs(icmph->un.echo.id),
 517                  ntohs(icmph->un.echo.sequence));
 518
 519         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 520         if (!sk) {
 521                 pr_debug("no socket, dropping\n");
 522                 return; /* No socket for error */
 523         }
 524         pr_debug("err on socket %p\n", sk);
 525
 526         err = 0;
 527         harderr = 0;
 528         inet_sock = inet_sk(sk);
 529
 530         if (skb->protocol == htons(ETH_P_IP)) {
 531                 switch (type) {
 532                 default:
 533                 case ICMP_TIME_EXCEEDED:
 534                         err = EHOSTUNREACH;
 535                         break;
 536                 case ICMP_SOURCE_QUENCH:
 537                         /* This is not a real error but ping wants to see it.
 538                          * Report it with some fake errno.
 539                          */
 540                         err = EREMOTEIO;
 541                         break;
 542                 case ICMP_PARAMETERPROB:
 543                         err = EPROTO;
 544                         harderr = 1;
 545                         break;
 546                 case ICMP_DEST_UNREACH:
 547                         if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
 548                                 ipv4_sk_update_pmtu(skb, sk, info);
 549                                 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
 550                                         err = EMSGSIZE;
 551                                         harderr = 1;
 552                                         break;
 553                                 }
 554                                 goto out;
 555                         }
 556                         err = EHOSTUNREACH;
 557                         if (code <= NR_ICMP_UNREACH) {
 558                                 harderr = icmp_err_convert[code].fatal;
 559                                 err = icmp_err_convert[code].errno;
 560                         }
 561                         break;
 562                 case ICMP_REDIRECT:
 563                         /* See ICMP_SOURCE_QUENCH */
 564                         ipv4_sk_redirect(skb, sk);
 565                         err = EREMOTEIO;
 566                         break;
 567                 }
 568 #if IS_ENABLED(CONFIG_IPV6)
 569         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 570                 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
 571 #endif
 572         }
 573
 574         /*
 575          *      RFC1122: OK.  Passes ICMP errors back to application, as per
 576          *      4.1.3.3.
 577          */
 578         if ((family == AF_INET && !inet_sock->recverr) ||
 579             (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
 580                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 581                         goto out;
 582         } else {
 583                 if (family == AF_INET) {
 584                         ip_icmp_error(sk, skb, err, 0 /* no remote port */,
 585                                       info, (u8 *)icmph);
 586 #if IS_ENABLED(CONFIG_IPV6)
 587                 } else if (family == AF_INET6) {
 588                         pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
 589                                                    info, (u8 *)icmph);
 590 #endif
 591                 }
 592         }
 593         sk->sk_err = err;
 594         sk->sk_error_report(sk);
 595 out:
 596         sock_put(sk);
 597 }
 598 EXPORT_SYMBOL_GPL(ping_err);
 599
 600 /*
 601  *      Copy and checksum an ICMP Echo packet from user space into a buffer
 602  *      starting from the payload.
 603  */
 604
 605 int ping_getfrag(void *from, char *to,
 606                  int offset, int fraglen, int odd, struct sk_buff *skb)
 607 {
 608         struct pingfakehdr *pfh = (struct pingfakehdr *)from;
 609
 610         if (offset == 0) {
 611                 fraglen -= sizeof(struct icmphdr);
 612                 if (fraglen < 0)
 613                         BUG();
 614                 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
 615                             fraglen, &pfh->wcheck,
 616                             &pfh->msg->msg_iter))
 617                         return -EFAULT;
 618         } else if (offset < sizeof(struct icmphdr)) {
 619                         BUG();
 620         } else {
 621                 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
 622                                             &pfh->msg->msg_iter))
 623                         return -EFAULT;
 624         }
 625
 626 #if IS_ENABLED(CONFIG_IPV6)
 627         /* For IPv6, checksum each skb as we go along, as expected by
 628          * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
 629          * wcheck, it will be finalized in ping_v4_push_pending_frames.
 630          */
 631         if (pfh->family == AF_INET6) {
 632                 skb->csum = pfh->wcheck;
 633                 skb->ip_summed = CHECKSUM_NONE;
 634                 pfh->wcheck = 0;
 635         }
 636 #endif
 637
 638         return 0;
 639 }
 640 EXPORT_SYMBOL_GPL(ping_getfrag);
 641
 642 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
 643                                        struct flowi4 *fl4)
 644 {
 645         struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
 646
 647         if (!skb)
 648                 return 0;
 649         pfh->wcheck = csum_partial((char *)&pfh->icmph,
 650                 sizeof(struct icmphdr), pfh->wcheck);
 651         pfh->icmph.checksum = csum_fold(pfh->wcheck);
 652         memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
 653         skb->ip_summed = CHECKSUM_NONE;
 654         return ip_push_pending_frames(sk, fl4);
 655 }
 656
 657 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
 658                         void *user_icmph, size_t icmph_len) {
 659         u8 type, code;
 660
 661         if (len > 0xFFFF)
 662                 return -EMSGSIZE;
 663
 664         /* Must have at least a full ICMP header. */
 665         if (len < icmph_len)
 666                 return -EINVAL;
 667
 668         /*
 669          *      Check the flags.
 670          */
 671
 672         /* Mirror BSD error message compatibility */
 673         if (msg->msg_flags & MSG_OOB)
 674                 return -EOPNOTSUPP;
 675
 676         /*
 677          *      Fetch the ICMP header provided by the userland.
 678          *      iovec is modified! The ICMP header is consumed.
 679          */
 680         if (memcpy_from_msg(user_icmph, msg, icmph_len))
 681                 return -EFAULT;
 682
 683         if (family == AF_INET) {
 684                 type = ((struct icmphdr *) user_icmph)->type;
 685                 code = ((struct icmphdr *) user_icmph)->code;
 686 #if IS_ENABLED(CONFIG_IPV6)
 687         } else if (family == AF_INET6) {
 688                 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
 689                 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
 690 #endif
 691         } else {
 692                 BUG();
 693         }
 694
 695         if (!ping_supported(family, type, code))
 696                 return -EINVAL;
 697
 698         return 0;
 699 }
 700 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
 701
 702 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 703 {
 704         struct net *net = sock_net(sk);
 705         struct flowi4 fl4;
 706         struct inet_sock *inet = inet_sk(sk);
 707         struct ipcm_cookie ipc;
 708         struct icmphdr user_icmph;
 709         struct pingfakehdr pfh;
 710         struct rtable *rt = NULL;
 711         struct ip_options_data opt_copy;
 712         int free = 0;
 713         __be32 saddr, daddr, faddr;
 714         u8  tos;
 715         int err;
 716
 717         pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
 718
 719         err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
 720                                   sizeof(user_icmph));
 721         if (err)
 722                 return err;
 723
 724         /*
 725          *      Get and verify the address.
 726          */
 727
 728         if (msg->msg_name) {
 729                 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
 730                 if (msg->msg_namelen < sizeof(*usin))
 731                         return -EINVAL;
 732                 if (usin->sin_family != AF_INET)
 733                         return -EAFNOSUPPORT;
 734                 daddr = usin->sin_addr.s_addr;
 735                 /* no remote port */
 736         } else {
 737                 if (sk->sk_state != TCP_ESTABLISHED)
 738                         return -EDESTADDRREQ;
 739                 daddr = inet->inet_daddr;
 740                 /* no remote port */
 741         }
 742
 743         ipc.sockc.tsflags = sk->sk_tsflags;
 744         ipc.addr = inet->inet_saddr;
 745         ipc.opt = NULL;
 746         ipc.oif = sk->sk_bound_dev_if;
 747         ipc.tx_flags = 0;
 748         ipc.ttl = 0;
 749         ipc.tos = -1;
 750
 751         if (msg->msg_controllen) {
 752                 err = ip_cmsg_send(sk, msg, &ipc, false);
 753                 if (unlikely(err)) {
 754                         kfree(ipc.opt);
 755                         return err;
 756                 }
 757                 if (ipc.opt)
 758                         free = 1;
 759         }
 760         if (!ipc.opt) {
 761                 struct ip_options_rcu *inet_opt;
 762
 763                 rcu_read_lock();
 764                 inet_opt = rcu_dereference(inet->inet_opt);
 765                 if (inet_opt) {
 766                         memcpy(&opt_copy, inet_opt,
 767                                sizeof(*inet_opt) + inet_opt->opt.optlen);
 768                         ipc.opt = &opt_copy.opt;
 769                 }
 770                 rcu_read_unlock();
 771         }
 772
 773         sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
 774
 775         saddr = ipc.addr;
 776         ipc.addr = faddr = daddr;
 777
 778         if (ipc.opt && ipc.opt->opt.srr) {
 779                 if (!daddr) {
 780                         err = -EINVAL;
 781                         goto out_free;
 782                 }
 783                 faddr = ipc.opt->opt.faddr;
 784         }
 785         tos = get_rttos(&ipc, inet);
 786         if (sock_flag(sk, SOCK_LOCALROUTE) ||
 787             (msg->msg_flags & MSG_DONTROUTE) ||
 788             (ipc.opt && ipc.opt->opt.is_strictroute)) {
 789                 tos |= RTO_ONLINK;
 790         }
 791
 792         if (ipv4_is_multicast(daddr)) {
 793                 if (!ipc.oif)
 794                         ipc.oif = inet->mc_index;
 795                 if (!saddr)
 796                         saddr = inet->mc_addr;
 797         } else if (!ipc.oif)
 798                 ipc.oif = inet->uc_index;
 799
 800         flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
 801                            RT_SCOPE_UNIVERSE, sk->sk_protocol,
 802                            inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
 803                            sk->sk_uid);
 804
 805         fl4.fl4_icmp_type = user_icmph.type;
 806         fl4.fl4_icmp_code = user_icmph.code;
 807
 808         security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
 809         rt = ip_route_output_flow(net, &fl4, sk);
 810         if (IS_ERR(rt)) {
 811                 err = PTR_ERR(rt);
 812                 rt = NULL;
 813                 if (err == -ENETUNREACH)
 814                         IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
 815                 goto out;
 816         }
 817
 818         err = -EACCES;
 819         if ((rt->rt_flags & RTCF_BROADCAST) &&
 820             !sock_flag(sk, SOCK_BROADCAST))
 821                 goto out;
 822
 823         if (msg->msg_flags & MSG_CONFIRM)
 824                 goto do_confirm;
 825 back_from_confirm:
 826
 827         if (!ipc.addr)
 828                 ipc.addr = fl4.daddr;
 829
 830         lock_sock(sk);
 831
 832         pfh.icmph.type = user_icmph.type; /* already checked */
 833         pfh.icmph.code = user_icmph.code; /* ditto */
 834         pfh.icmph.checksum = 0;
 835         pfh.icmph.un.echo.id = inet->inet_sport;
 836         pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 837         pfh.msg = msg;
 838         pfh.wcheck = 0;
 839         pfh.family = AF_INET;
 840
 841         err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 842                         0, &ipc, &rt, msg->msg_flags);
 843         if (err)
 844                 ip_flush_pending_frames(sk);
 845         else
 846                 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 847         release_sock(sk);
 848
 849 out:
 850         ip_rt_put(rt);
 851 out_free:
 852         if (free)
 853                 kfree(ipc.opt);
 854         if (!err) {
 855                 icmp_out_count(sock_net(sk), user_icmph.type);
 856                 return len;
 857         }
 858         return err;
 859
 860 do_confirm:
 861         if (msg->msg_flags & MSG_PROBE)
 862                 dst_confirm_neigh(&rt->dst, &fl4.daddr);
 863         if (!(msg->msg_flags & MSG_PROBE) || len)
 864                 goto back_from_confirm;
 865         err = 0;
 866         goto out;
 867 }
 868
 869 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
 870                  int flags, int *addr_len)
 871 {
 872         struct inet_sock *isk = inet_sk(sk);
 873         int family = sk->sk_family;
 874         struct sk_buff *skb;
 875         int copied, err;
 876
 877         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 878
 879         err = -EOPNOTSUPP;
 880         if (flags & MSG_OOB)
 881                 goto out;
 882
 883         if (flags & MSG_ERRQUEUE)
 884                 return inet_recv_error(sk, msg, len, addr_len);
 885
 886         skb = skb_recv_datagram(sk, flags, noblock, &err);
 887         if (!skb)
 888                 goto out;
 889
 890         copied = skb->len;
 891         if (copied > len) {
 892                 msg->msg_flags |= MSG_TRUNC;
 893                 copied = len;
 894         }
 895
 896         /* Don't bother checking the checksum */
 897         err = skb_copy_datagram_msg(skb, 0, msg, copied);
 898         if (err)
 899                 goto done;
 900
 901         sock_recv_timestamp(msg, sk, skb);
 902
 903         /* Copy the address and add cmsg data. */
 904         if (family == AF_INET) {
 905                 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 906
 907                 if (sin) {
 908                         sin->sin_family = AF_INET;
 909                         sin->sin_port = 0 /* skb->h.uh->source */;
 910                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 911                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 912                         *addr_len = sizeof(*sin);
 913                 }
 914
 915                 if (isk->cmsg_flags)
 916                         ip_cmsg_recv(msg, skb);
 917
 918 #if IS_ENABLED(CONFIG_IPV6)
 919         } else if (family == AF_INET6) {
 920                 struct ipv6_pinfo *np = inet6_sk(sk);
 921                 struct ipv6hdr *ip6 = ipv6_hdr(skb);
 922                 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 923
 924                 if (sin6) {
 925                         sin6->sin6_family = AF_INET6;
 926                         sin6->sin6_port = 0;
 927                         sin6->sin6_addr = ip6->saddr;
 928                         sin6->sin6_flowinfo = 0;
 929                         if (np->sndflow)
 930                                 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
 931                         sin6->sin6_scope_id =
 932                                 ipv6_iface_scope_id(&sin6->sin6_addr,
 933                                                     inet6_iif(skb));
 934                         *addr_len = sizeof(*sin6);
 935                 }
 936
 937                 if (inet6_sk(sk)->rxopt.all)
 938                         pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
 939                 if (skb->protocol == htons(ETH_P_IPV6) &&
 940                     inet6_sk(sk)->rxopt.all)
 941                         pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
 942                 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
 943                         ip_cmsg_recv(msg, skb);
 944 #endif
 945         } else {
 946                 BUG();
 947         }
 948
 949         err = copied;
 950
 951 done:
 952         skb_free_datagram(sk, skb);
 953 out:
 954         pr_debug("ping_recvmsg -> %d\n", err);
 955         return err;
 956 }
 957 EXPORT_SYMBOL_GPL(ping_recvmsg);
 958
 959 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 960 {
 961         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 962                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
 963         if (sock_queue_rcv_skb(sk, skb) < 0) {
 964                 kfree_skb(skb);
 965                 pr_debug("ping_queue_rcv_skb -> failed\n");
 966                 return -1;
 967         }
 968         return 0;
 969 }
 970 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 971
 972
 973 /*
 974  *      All we need to do is get the socket.
 975  */
 976
 977 bool ping_rcv(struct sk_buff *skb)
 978 {
 979         struct sock *sk;
 980         struct net *net = dev_net(skb->dev);
 981         struct icmphdr *icmph = icmp_hdr(skb);
 982         bool rc = false;
 983
 984         /* We assume the packet has already been checked by icmp_rcv */
 985
 986         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 987                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 988
 989         /* Push ICMP header back */
 990         skb_push(skb, skb->data - (u8 *)icmph);
 991
 992         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 993         if (sk) {
 994                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 995
 996                 pr_debug("rcv on socket %p\n", sk);
 997                 if (skb2 && !ping_queue_rcv_skb(sk, skb2))
 998                         rc = true;
 999                 sock_put(sk);
1000         }
1001
1002         if (!rc)
1003                 pr_debug("no socket, dropping\n");
1004
1005         return rc;
1006 }
1007 EXPORT_SYMBOL_GPL(ping_rcv);
1008
1009 struct proto ping_prot = {
1010         .name =         "PING",
1011         .owner =        THIS_MODULE,
1012         .init =         ping_init_sock,
1013         .close =        ping_close,
1014         .connect =      ip4_datagram_connect,
1015         .disconnect =   __udp_disconnect,
1016         .setsockopt =   ip_setsockopt,
1017         .getsockopt =   ip_getsockopt,
1018         .sendmsg =      ping_v4_sendmsg,
1019         .recvmsg =      ping_recvmsg,
1020         .bind =         ping_bind,
1021         .backlog_rcv =  ping_queue_rcv_skb,
1022         .release_cb =   ip4_datagram_release_cb,
1023         .hash =         ping_hash,
1024         .unhash =       ping_unhash,
1025         .get_port =     ping_get_port,
1026         .obj_size =     sizeof(struct inet_sock),
1027 };
1028 EXPORT_SYMBOL(ping_prot);
1029
1030 #ifdef CONFIG_PROC_FS
1031
1032 static struct sock *ping_get_first(struct seq_file *seq, int start)
1033 {
1034         struct sock *sk;
1035         struct ping_iter_state *state = seq->private;
1036         struct net *net = seq_file_net(seq);
1037
1038         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1039              ++state->bucket) {
1040                 struct hlist_nulls_node *node;
1041                 struct hlist_nulls_head *hslot;
1042
1043                 hslot = &ping_table.hash[state->bucket];
1044
1045                 if (hlist_nulls_empty(hslot))
1046                         continue;
1047
1048                 sk_nulls_for_each(sk, node, hslot) {
1049                         if (net_eq(sock_net(sk), net) &&
1050                             sk->sk_family == state->family)
1051                                 goto found;
1052                 }
1053         }
1054         sk = NULL;
1055 found:
1056         return sk;
1057 }
1058
1059 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1060 {
1061         struct ping_iter_state *state = seq->private;
1062         struct net *net = seq_file_net(seq);
1063
1064         do {
1065                 sk = sk_nulls_next(sk);
1066         } while (sk && (!net_eq(sock_net(sk), net)));
1067
1068         if (!sk)
1069                 return ping_get_first(seq, state->bucket + 1);
1070         return sk;
1071 }
1072
1073 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1074 {
1075         struct sock *sk = ping_get_first(seq, 0);
1076
1077         if (sk)
1078                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1079                         --pos;
1080         return pos ? NULL : sk;
1081 }
1082
1083 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1084         __acquires(ping_table.lock)
1085 {
1086         struct ping_iter_state *state = seq->private;
1087         state->bucket = 0;
1088         state->family = family;
1089
1090         read_lock_bh(&ping_table.lock);
1091
1092         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1093 }
1094 EXPORT_SYMBOL_GPL(ping_seq_start);
1095
1096 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1097 {
1098         return ping_seq_start(seq, pos, AF_INET);
1099 }
1100
1101 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1102 {
1103         struct sock *sk;
1104
1105         if (v == SEQ_START_TOKEN)
1106                 sk = ping_get_idx(seq, 0);
1107         else
1108                 sk = ping_get_next(seq, v);
1109
1110         ++*pos;
1111         return sk;
1112 }
1113 EXPORT_SYMBOL_GPL(ping_seq_next);
1114
1115 void ping_seq_stop(struct seq_file *seq, void *v)
1116         __releases(ping_table.lock)
1117 {
1118         read_unlock_bh(&ping_table.lock);
1119 }
1120 EXPORT_SYMBOL_GPL(ping_seq_stop);
1121
1122 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1123                 int bucket)
1124 {
1125         struct inet_sock *inet = inet_sk(sp);
1126         __be32 dest = inet->inet_daddr;
1127         __be32 src = inet->inet_rcv_saddr;
1128         __u16 destp = ntohs(inet->inet_dport);
1129         __u16 srcp = ntohs(inet->inet_sport);
1130
1131         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1132                 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1133                 bucket, src, srcp, dest, destp, sp->sk_state,
1134                 sk_wmem_alloc_get(sp),
1135                 sk_rmem_alloc_get(sp),
1136                 0, 0L, 0,
1137                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1138                 0, sock_i_ino(sp),
1139                 refcount_read(&sp->sk_refcnt), sp,
1140                 atomic_read(&sp->sk_drops));
1141 }
1142
1143 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1144 {
1145         seq_setwidth(seq, 127);
1146         if (v == SEQ_START_TOKEN)
1147                 seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1148                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1149                            "inode ref pointer drops");
1150         else {
1151                 struct ping_iter_state *state = seq->private;
1152
1153                 ping_v4_format_sock(v, seq, state->bucket);
1154         }
1155         seq_pad(seq, '\n');
1156         return 0;
1157 }
1158
1159 static int ping_seq_open(struct inode *inode, struct file *file)
1160 {
1161         struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1162         return seq_open_net(inode, file, &afinfo->seq_ops,
1163                            sizeof(struct ping_iter_state));
1164 }
1165
1166 const struct file_operations ping_seq_fops = {
1167         .open           = ping_seq_open,
1168         .read           = seq_read,
1169         .llseek         = seq_lseek,
1170         .release        = seq_release_net,
1171 };
1172 EXPORT_SYMBOL_GPL(ping_seq_fops);
1173
1174 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1175         .name           = "icmp",
1176         .family         = AF_INET,
1177         .seq_fops       = &ping_seq_fops,
1178         .seq_ops        = {
1179                 .start          = ping_v4_seq_start,
1180                 .show           = ping_v4_seq_show,
1181                 .next           = ping_seq_next,
1182                 .stop           = ping_seq_stop,
1183         },
1184 };
1185
1186 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1187 {
1188         struct proc_dir_entry *p;
1189         p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1190                              afinfo->seq_fops, afinfo);
1191         if (!p)
1192                 return -ENOMEM;
1193         return 0;
1194 }
1195 EXPORT_SYMBOL_GPL(ping_proc_register);
1196
1197 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1198 {
1199         remove_proc_entry(afinfo->name, net->proc_net);
1200 }
1201 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1202
1203 static int __net_init ping_v4_proc_init_net(struct net *net)
1204 {
1205         return ping_proc_register(net, &ping_v4_seq_afinfo);
1206 }
1207
1208 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1209 {
1210         ping_proc_unregister(net, &ping_v4_seq_afinfo);
1211 }
1212
1213 static struct pernet_operations ping_v4_net_ops = {
1214         .init = ping_v4_proc_init_net,
1215         .exit = ping_v4_proc_exit_net,
1216 };
1217
1218 int __init ping_proc_init(void)
1219 {
1220         return register_pernet_subsys(&ping_v4_net_ops);
1221 }
1222
1223 void ping_proc_exit(void)
1224 {
1225         unregister_pernet_subsys(&ping_v4_net_ops);
1226 }
1227
1228 #endif
1229
1230 void __init ping_init(void)
1231 {
1232         int i;
1233
1234         for (i = 0; i < PING_HTABLE_SIZE; i++)
1235                 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1236         rwlock_init(&ping_table.lock);
1237 }