net/ipv4/ping.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              "Ping" sockets
   8  *
   9  * Based on ipv4/udp.c code.
  10  *
  11  * Authors:     Vasiliy Kulikov / Openwall (for Linux 2.6),
  12  *              Pavel Kankovsky (for Linux 2.4.32)
  13  *
  14  * Pavel gave all rights to bugs to Vasiliy,
  15  * none of the bugs are Pavel's now.
  16  */
  17
  18 #include <linux/uaccess.h>
  19 #include <linux/types.h>
  20 #include <linux/fcntl.h>
  21 #include <linux/socket.h>
  22 #include <linux/sockios.h>
  23 #include <linux/in.h>
  24 #include <linux/errno.h>
  25 #include <linux/timer.h>
  26 #include <linux/mm.h>
  27 #include <linux/inet.h>
  28 #include <linux/netdevice.h>
  29 #include <net/snmp.h>
  30 #include <net/ip.h>
  31 #include <net/icmp.h>
  32 #include <net/protocol.h>
  33 #include <linux/skbuff.h>
  34 #include <linux/proc_fs.h>
  35 #include <linux/export.h>
  36 #include <net/sock.h>
  37 #include <net/ping.h>
  38 #include <net/udp.h>
  39 #include <net/route.h>
  40 #include <net/inet_common.h>
  41 #include <net/checksum.h>
  42
  43 #if IS_ENABLED(CONFIG_IPV6)
  44 #include <linux/in6.h>
  45 #include <linux/icmpv6.h>
  46 #include <net/addrconf.h>
  47 #include <net/ipv6.h>
  48 #include <net/transp_v6.h>
  49 #endif
  50
  51 struct ping_table {
  52         struct hlist_nulls_head hash[PING_HTABLE_SIZE];
  53         rwlock_t                lock;
  54 };
  55
  56 static struct ping_table ping_table;
  57 struct pingv6_ops pingv6_ops;
  58 EXPORT_SYMBOL_GPL(pingv6_ops);
  59
  60 static u16 ping_port_rover;
  61
  62 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
  63 {
  64         u32 res = (num + net_hash_mix(net)) & mask;
  65
  66         pr_debug("hash(%u) = %u\n", num, res);
  67         return res;
  68 }
  69 EXPORT_SYMBOL_GPL(ping_hash);
  70
  71 static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
  72                                              struct net *net, unsigned int num)
  73 {
  74         return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
  75 }
  76
  77 int ping_get_port(struct sock *sk, unsigned short ident)
  78 {
  79         struct hlist_nulls_node *node;
  80         struct hlist_nulls_head *hlist;
  81         struct inet_sock *isk, *isk2;
  82         struct sock *sk2 = NULL;
  83
  84         isk = inet_sk(sk);
  85         write_lock_bh(&ping_table.lock);
  86         if (ident == 0) {
  87                 u32 i;
  88                 u16 result = ping_port_rover + 1;
  89
  90                 for (i = 0; i < (1L << 16); i++, result++) {
  91                         if (!result)
  92                                 result++; /* avoid zero */
  93                         hlist = ping_hashslot(&ping_table, sock_net(sk),
  94                                             result);
  95                         ping_portaddr_for_each_entry(sk2, node, hlist) {
  96                                 isk2 = inet_sk(sk2);
  97
  98                                 if (isk2->inet_num == result)
  99                                         goto next_port;
 100                         }
 101
 102                         /* found */
 103                         ping_port_rover = ident = result;
 104                         break;
 105 next_port:
 106                         ;
 107                 }
 108                 if (i >= (1L << 16))
 109                         goto fail;
 110         } else {
 111                 hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
 112                 ping_portaddr_for_each_entry(sk2, node, hlist) {
 113                         isk2 = inet_sk(sk2);
 114
 115                         /* BUG? Why is this reuse and not reuseaddr? ping.c
 116                          * doesn't turn off SO_REUSEADDR, and it doesn't expect
 117                          * that other ping processes can steal its packets.
 118                          */
 119                         if ((isk2->inet_num == ident) &&
 120                             (sk2 != sk) &&
 121                             (!sk2->sk_reuse || !sk->sk_reuse))
 122                                 goto fail;
 123                 }
 124         }
 125
 126         pr_debug("found port/ident = %d\n", ident);
 127         isk->inet_num = ident;
 128         if (sk_unhashed(sk)) {
 129                 pr_debug("was not hashed\n");
 130                 sock_hold(sk);
 131                 hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
 132                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
 133         }
 134         write_unlock_bh(&ping_table.lock);
 135         return 0;
 136
 137 fail:
 138         write_unlock_bh(&ping_table.lock);
 139         return 1;
 140 }
 141 EXPORT_SYMBOL_GPL(ping_get_port);
 142
 143 int ping_hash(struct sock *sk)
 144 {
 145         pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
 146         BUG(); /* "Please do not press this button again." */
 147
 148         return 0;
 149 }
 150
 151 void ping_unhash(struct sock *sk)
 152 {
 153         struct inet_sock *isk = inet_sk(sk);
 154
 155         pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
 156         write_lock_bh(&ping_table.lock);
 157         if (sk_hashed(sk)) {
 158                 hlist_nulls_del(&sk->sk_nulls_node);
 159                 sk_nulls_node_init(&sk->sk_nulls_node);
 160                 sock_put(sk);
 161                 isk->inet_num = 0;
 162                 isk->inet_sport = 0;
 163                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 164         }
 165         write_unlock_bh(&ping_table.lock);
 166 }
 167 EXPORT_SYMBOL_GPL(ping_unhash);
 168
 169 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
 170 {
 171         struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
 172         struct sock *sk = NULL;
 173         struct inet_sock *isk;
 174         struct hlist_nulls_node *hnode;
 175         int dif, sdif;
 176
 177         if (skb->protocol == htons(ETH_P_IP)) {
 178                 dif = inet_iif(skb);
 179                 sdif = inet_sdif(skb);
 180                 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
 181                          (int)ident, &ip_hdr(skb)->daddr, dif);
 182 #if IS_ENABLED(CONFIG_IPV6)
 183         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 184                 dif = inet6_iif(skb);
 185                 sdif = inet6_sdif(skb);
 186                 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
 187                          (int)ident, &ipv6_hdr(skb)->daddr, dif);
 188 #endif
 189         } else {
 190                 return NULL;
 191         }
 192
 193         read_lock_bh(&ping_table.lock);
 194
 195         ping_portaddr_for_each_entry(sk, hnode, hslot) {
 196                 isk = inet_sk(sk);
 197
 198                 pr_debug("iterate\n");
 199                 if (isk->inet_num != ident)
 200                         continue;
 201
 202                 if (skb->protocol == htons(ETH_P_IP) &&
 203                     sk->sk_family == AF_INET) {
 204                         pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
 205                                  (int) isk->inet_num, &isk->inet_rcv_saddr,
 206                                  sk->sk_bound_dev_if);
 207
 208                         if (isk->inet_rcv_saddr &&
 209                             isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
 210                                 continue;
 211 #if IS_ENABLED(CONFIG_IPV6)
 212                 } else if (skb->protocol == htons(ETH_P_IPV6) &&
 213                            sk->sk_family == AF_INET6) {
 214
 215                         pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
 216                                  (int) isk->inet_num,
 217                                  &sk->sk_v6_rcv_saddr,
 218                                  sk->sk_bound_dev_if);
 219
 220                         if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
 221                             !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
 222                                              &ipv6_hdr(skb)->daddr))
 223                                 continue;
 224 #endif
 225                 } else {
 226                         continue;
 227                 }
 228
 229                 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif &&
 230                     sk->sk_bound_dev_if != sdif)
 231                         continue;
 232
 233                 sock_hold(sk);
 234                 goto exit;
 235         }
 236
 237         sk = NULL;
 238 exit:
 239         read_unlock_bh(&ping_table.lock);
 240
 241         return sk;
 242 }
 243
 244 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
 245                                           kgid_t *high)
 246 {
 247         kgid_t *data = net->ipv4.ping_group_range.range;
 248         unsigned int seq;
 249
 250         do {
 251                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 252
 253                 *low = data[0];
 254                 *high = data[1];
 255         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 256 }
 257
 258
 259 int ping_init_sock(struct sock *sk)
 260 {
 261         struct net *net = sock_net(sk);
 262         kgid_t group = current_egid();
 263         struct group_info *group_info;
 264         int i;
 265         kgid_t low, high;
 266         int ret = 0;
 267
 268         if (sk->sk_family == AF_INET6)
 269                 sk->sk_ipv6only = 1;
 270
 271         inet_get_ping_group_range_net(net, &low, &high);
 272         if (gid_lte(low, group) && gid_lte(group, high))
 273                 return 0;
 274
 275         group_info = get_current_groups();
 276         for (i = 0; i < group_info->ngroups; i++) {
 277                 kgid_t gid = group_info->gid[i];
 278
 279                 if (gid_lte(low, gid) && gid_lte(gid, high))
 280                         goto out_release_group;
 281         }
 282
 283         ret = -EACCES;
 284
 285 out_release_group:
 286         put_group_info(group_info);
 287         return ret;
 288 }
 289 EXPORT_SYMBOL_GPL(ping_init_sock);
 290
 291 void ping_close(struct sock *sk, long timeout)
 292 {
 293         pr_debug("ping_close(sk=%p,sk->num=%u)\n",
 294                  inet_sk(sk), inet_sk(sk)->inet_num);
 295         pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
 296
 297         sk_common_release(sk);
 298 }
 299 EXPORT_SYMBOL_GPL(ping_close);
 300
 301 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
 302 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
 303                                 struct sockaddr *uaddr, int addr_len) {
 304         struct net *net = sock_net(sk);
 305         if (sk->sk_family == AF_INET) {
 306                 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 307                 u32 tb_id = RT_TABLE_LOCAL;
 308                 int chk_addr_ret;
 309
 310                 if (addr_len < sizeof(*addr))
 311                         return -EINVAL;
 312
 313                 if (addr->sin_family != AF_INET &&
 314                     !(addr->sin_family == AF_UNSPEC &&
 315                       addr->sin_addr.s_addr == htonl(INADDR_ANY)))
 316                         return -EAFNOSUPPORT;
 317
 318                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
 319                          sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
 320
 321                 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
 322                 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
 323
 324                 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
 325                         chk_addr_ret = RTN_LOCAL;
 326
 327                 if ((!inet_can_nonlocal_bind(net, isk) &&
 328                      chk_addr_ret != RTN_LOCAL) ||
 329                     chk_addr_ret == RTN_MULTICAST ||
 330                     chk_addr_ret == RTN_BROADCAST)
 331                         return -EADDRNOTAVAIL;
 332
 333 #if IS_ENABLED(CONFIG_IPV6)
 334         } else if (sk->sk_family == AF_INET6) {
 335                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 336                 int addr_type, scoped, has_addr;
 337                 struct net_device *dev = NULL;
 338
 339                 if (addr_len < sizeof(*addr))
 340                         return -EINVAL;
 341
 342                 if (addr->sin6_family != AF_INET6)
 343                         return -EAFNOSUPPORT;
 344
 345                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 346                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 347
 348                 addr_type = ipv6_addr_type(&addr->sin6_addr);
 349                 scoped = __ipv6_addr_needs_scope_id(addr_type);
 350                 if ((addr_type != IPV6_ADDR_ANY &&
 351                      !(addr_type & IPV6_ADDR_UNICAST)) ||
 352                     (scoped && !addr->sin6_scope_id))
 353                         return -EINVAL;
 354
 355                 rcu_read_lock();
 356                 if (addr->sin6_scope_id) {
 357                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 358                         if (!dev) {
 359                                 rcu_read_unlock();
 360                                 return -ENODEV;
 361                         }
 362                 }
 363
 364                 if (!dev && sk->sk_bound_dev_if) {
 365                         dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 366                         if (!dev) {
 367                                 rcu_read_unlock();
 368                                 return -ENODEV;
 369                         }
 370                 }
 371                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 372                                                     scoped);
 373                 rcu_read_unlock();
 374
 375                 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr ||
 376                       addr_type == IPV6_ADDR_ANY))
 377                         return -EADDRNOTAVAIL;
 378
 379                 if (scoped)
 380                         sk->sk_bound_dev_if = addr->sin6_scope_id;
 381 #endif
 382         } else {
 383                 return -EAFNOSUPPORT;
 384         }
 385         return 0;
 386 }
 387
 388 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
 389 {
 390         if (saddr->sa_family == AF_INET) {
 391                 struct inet_sock *isk = inet_sk(sk);
 392                 struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
 393                 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
 394 #if IS_ENABLED(CONFIG_IPV6)
 395         } else if (saddr->sa_family == AF_INET6) {
 396                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
 397                 struct ipv6_pinfo *np = inet6_sk(sk);
 398                 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
 399 #endif
 400         }
 401 }
 402
 403 static void ping_clear_saddr(struct sock *sk, int dif)
 404 {
 405         sk->sk_bound_dev_if = dif;
 406         if (sk->sk_family == AF_INET) {
 407                 struct inet_sock *isk = inet_sk(sk);
 408                 isk->inet_rcv_saddr = isk->inet_saddr = 0;
 409 #if IS_ENABLED(CONFIG_IPV6)
 410         } else if (sk->sk_family == AF_INET6) {
 411                 struct ipv6_pinfo *np = inet6_sk(sk);
 412                 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
 413                 memset(&np->saddr, 0, sizeof(np->saddr));
 414 #endif
 415         }
 416 }
 417 /*
 418  * We need our own bind because there are no privileged id's == local ports.
 419  * Moreover, we don't allow binding to multi- and broadcast addresses.
 420  */
 421
 422 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 423 {
 424         struct inet_sock *isk = inet_sk(sk);
 425         unsigned short snum;
 426         int err;
 427         int dif = sk->sk_bound_dev_if;
 428
 429         err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
 430         if (err)
 431                 return err;
 432
 433         lock_sock(sk);
 434
 435         err = -EINVAL;
 436         if (isk->inet_num != 0)
 437                 goto out;
 438
 439         err = -EADDRINUSE;
 440         ping_set_saddr(sk, uaddr);
 441         snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
 442         if (ping_get_port(sk, snum) != 0) {
 443                 ping_clear_saddr(sk, dif);
 444                 goto out;
 445         }
 446
 447         pr_debug("after bind(): num = %hu, dif = %d\n",
 448                  isk->inet_num,
 449                  sk->sk_bound_dev_if);
 450
 451         err = 0;
 452         if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
 453                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 454 #if IS_ENABLED(CONFIG_IPV6)
 455         if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 456                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 457 #endif
 458
 459         if (snum)
 460                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 461         isk->inet_sport = htons(isk->inet_num);
 462         isk->inet_daddr = 0;
 463         isk->inet_dport = 0;
 464
 465 #if IS_ENABLED(CONFIG_IPV6)
 466         if (sk->sk_family == AF_INET6)
 467                 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
 468 #endif
 469
 470         sk_dst_reset(sk);
 471 out:
 472         release_sock(sk);
 473         pr_debug("ping_v4_bind -> %d\n", err);
 474         return err;
 475 }
 476 EXPORT_SYMBOL_GPL(ping_bind);
 477
 478 /*
 479  * Is this a supported type of ICMP message?
 480  */
 481
 482 static inline int ping_supported(int family, int type, int code)
 483 {
 484         return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
 485                (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
 486 }
 487
 488 /*
 489  * This routine is called by the ICMP module when it gets some
 490  * sort of error condition.
 491  */
 492
 493 void ping_err(struct sk_buff *skb, int offset, u32 info)
 494 {
 495         int family;
 496         struct icmphdr *icmph;
 497         struct inet_sock *inet_sock;
 498         int type;
 499         int code;
 500         struct net *net = dev_net(skb->dev);
 501         struct sock *sk;
 502         int harderr;
 503         int err;
 504
 505         if (skb->protocol == htons(ETH_P_IP)) {
 506                 family = AF_INET;
 507                 type = icmp_hdr(skb)->type;
 508                 code = icmp_hdr(skb)->code;
 509                 icmph = (struct icmphdr *)(skb->data + offset);
 510         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 511                 family = AF_INET6;
 512                 type = icmp6_hdr(skb)->icmp6_type;
 513                 code = icmp6_hdr(skb)->icmp6_code;
 514                 icmph = (struct icmphdr *) (skb->data + offset);
 515         } else {
 516                 BUG();
 517         }
 518
 519         /* We assume the packet has already been checked by icmp_unreach */
 520
 521         if (!ping_supported(family, icmph->type, icmph->code))
 522                 return;
 523
 524         pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
 525                  skb->protocol, type, code, ntohs(icmph->un.echo.id),
 526                  ntohs(icmph->un.echo.sequence));
 527
 528         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 529         if (!sk) {
 530                 pr_debug("no socket, dropping\n");
 531                 return; /* No socket for error */
 532         }
 533         pr_debug("err on socket %p\n", sk);
 534
 535         err = 0;
 536         harderr = 0;
 537         inet_sock = inet_sk(sk);
 538
 539         if (skb->protocol == htons(ETH_P_IP)) {
 540                 switch (type) {
 541                 default:
 542                 case ICMP_TIME_EXCEEDED:
 543                         err = EHOSTUNREACH;
 544                         break;
 545                 case ICMP_SOURCE_QUENCH:
 546                         /* This is not a real error but ping wants to see it.
 547                          * Report it with some fake errno.
 548                          */
 549                         err = EREMOTEIO;
 550                         break;
 551                 case ICMP_PARAMETERPROB:
 552                         err = EPROTO;
 553                         harderr = 1;
 554                         break;
 555                 case ICMP_DEST_UNREACH:
 556                         if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
 557                                 ipv4_sk_update_pmtu(skb, sk, info);
 558                                 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
 559                                         err = EMSGSIZE;
 560                                         harderr = 1;
 561                                         break;
 562                                 }
 563                                 goto out;
 564                         }
 565                         err = EHOSTUNREACH;
 566                         if (code <= NR_ICMP_UNREACH) {
 567                                 harderr = icmp_err_convert[code].fatal;
 568                                 err = icmp_err_convert[code].errno;
 569                         }
 570                         break;
 571                 case ICMP_REDIRECT:
 572                         /* See ICMP_SOURCE_QUENCH */
 573                         ipv4_sk_redirect(skb, sk);
 574                         err = EREMOTEIO;
 575                         break;
 576                 }
 577 #if IS_ENABLED(CONFIG_IPV6)
 578         } else if (skb->protocol == htons(ETH_P_IPV6)) {
 579                 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
 580 #endif
 581         }
 582
 583         /*
 584          *      RFC1122: OK.  Passes ICMP errors back to application, as per
 585          *      4.1.3.3.
 586          */
 587         if ((family == AF_INET && !inet_sock->recverr) ||
 588             (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
 589                 if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 590                         goto out;
 591         } else {
 592                 if (family == AF_INET) {
 593                         ip_icmp_error(sk, skb, err, 0 /* no remote port */,
 594                                       info, (u8 *)icmph);
 595 #if IS_ENABLED(CONFIG_IPV6)
 596                 } else if (family == AF_INET6) {
 597                         pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
 598                                                    info, (u8 *)icmph);
 599 #endif
 600                 }
 601         }
 602         sk->sk_err = err;
 603         sk->sk_error_report(sk);
 604 out:
 605         sock_put(sk);
 606 }
 607 EXPORT_SYMBOL_GPL(ping_err);
 608
 609 /*
 610  *      Copy and checksum an ICMP Echo packet from user space into a buffer
 611  *      starting from the payload.
 612  */
 613
 614 int ping_getfrag(void *from, char *to,
 615                  int offset, int fraglen, int odd, struct sk_buff *skb)
 616 {
 617         struct pingfakehdr *pfh = (struct pingfakehdr *)from;
 618
 619         if (offset == 0) {
 620                 fraglen -= sizeof(struct icmphdr);
 621                 if (fraglen < 0)
 622                         BUG();
 623                 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
 624                             fraglen, &pfh->wcheck,
 625                             &pfh->msg->msg_iter))
 626                         return -EFAULT;
 627         } else if (offset < sizeof(struct icmphdr)) {
 628                         BUG();
 629         } else {
 630                 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
 631                                             &pfh->msg->msg_iter))
 632                         return -EFAULT;
 633         }
 634
 635 #if IS_ENABLED(CONFIG_IPV6)
 636         /* For IPv6, checksum each skb as we go along, as expected by
 637          * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
 638          * wcheck, it will be finalized in ping_v4_push_pending_frames.
 639          */
 640         if (pfh->family == AF_INET6) {
 641                 skb->csum = pfh->wcheck;
 642                 skb->ip_summed = CHECKSUM_NONE;
 643                 pfh->wcheck = 0;
 644         }
 645 #endif
 646
 647         return 0;
 648 }
 649 EXPORT_SYMBOL_GPL(ping_getfrag);
 650
 651 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
 652                                        struct flowi4 *fl4)
 653 {
 654         struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
 655
 656         if (!skb)
 657                 return 0;
 658         pfh->wcheck = csum_partial((char *)&pfh->icmph,
 659                 sizeof(struct icmphdr), pfh->wcheck);
 660         pfh->icmph.checksum = csum_fold(pfh->wcheck);
 661         memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
 662         skb->ip_summed = CHECKSUM_NONE;
 663         return ip_push_pending_frames(sk, fl4);
 664 }
 665
 666 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
 667                         void *user_icmph, size_t icmph_len) {
 668         u8 type, code;
 669
 670         if (len > 0xFFFF)
 671                 return -EMSGSIZE;
 672
 673         /* Must have at least a full ICMP header. */
 674         if (len < icmph_len)
 675                 return -EINVAL;
 676
 677         /*
 678          *      Check the flags.
 679          */
 680
 681         /* Mirror BSD error message compatibility */
 682         if (msg->msg_flags & MSG_OOB)
 683                 return -EOPNOTSUPP;
 684
 685         /*
 686          *      Fetch the ICMP header provided by the userland.
 687          *      iovec is modified! The ICMP header is consumed.
 688          */
 689         if (memcpy_from_msg(user_icmph, msg, icmph_len))
 690                 return -EFAULT;
 691
 692         if (family == AF_INET) {
 693                 type = ((struct icmphdr *) user_icmph)->type;
 694                 code = ((struct icmphdr *) user_icmph)->code;
 695 #if IS_ENABLED(CONFIG_IPV6)
 696         } else if (family == AF_INET6) {
 697                 type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
 698                 code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
 699 #endif
 700         } else {
 701                 BUG();
 702         }
 703
 704         if (!ping_supported(family, type, code))
 705                 return -EINVAL;
 706
 707         return 0;
 708 }
 709 EXPORT_SYMBOL_GPL(ping_common_sendmsg);
 710
 711 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 712 {
 713         struct net *net = sock_net(sk);
 714         struct flowi4 fl4;
 715         struct inet_sock *inet = inet_sk(sk);
 716         struct ipcm_cookie ipc;
 717         struct icmphdr user_icmph;
 718         struct pingfakehdr pfh;
 719         struct rtable *rt = NULL;
 720         struct ip_options_data opt_copy;
 721         int free = 0;
 722         __be32 saddr, daddr, faddr;
 723         u8  tos;
 724         int err;
 725
 726         pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
 727
 728         err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
 729                                   sizeof(user_icmph));
 730         if (err)
 731                 return err;
 732
 733         /*
 734          *      Get and verify the address.
 735          */
 736
 737         if (msg->msg_name) {
 738                 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
 739                 if (msg->msg_namelen < sizeof(*usin))
 740                         return -EINVAL;
 741                 if (usin->sin_family != AF_INET)
 742                         return -EAFNOSUPPORT;
 743                 daddr = usin->sin_addr.s_addr;
 744                 /* no remote port */
 745         } else {
 746                 if (sk->sk_state != TCP_ESTABLISHED)
 747                         return -EDESTADDRREQ;
 748                 daddr = inet->inet_daddr;
 749                 /* no remote port */
 750         }
 751
 752         ipcm_init_sk(&ipc, inet);
 753
 754         if (msg->msg_controllen) {
 755                 err = ip_cmsg_send(sk, msg, &ipc, false);
 756                 if (unlikely(err)) {
 757                         kfree(ipc.opt);
 758                         return err;
 759                 }
 760                 if (ipc.opt)
 761                         free = 1;
 762         }
 763         if (!ipc.opt) {
 764                 struct ip_options_rcu *inet_opt;
 765
 766                 rcu_read_lock();
 767                 inet_opt = rcu_dereference(inet->inet_opt);
 768                 if (inet_opt) {
 769                         memcpy(&opt_copy, inet_opt,
 770                                sizeof(*inet_opt) + inet_opt->opt.optlen);
 771                         ipc.opt = &opt_copy.opt;
 772                 }
 773                 rcu_read_unlock();
 774         }
 775
 776         saddr = ipc.addr;
 777         ipc.addr = faddr = daddr;
 778
 779         if (ipc.opt && ipc.opt->opt.srr) {
 780                 if (!daddr) {
 781                         err = -EINVAL;
 782                         goto out_free;
 783                 }
 784                 faddr = ipc.opt->opt.faddr;
 785         }
 786         tos = get_rttos(&ipc, inet);
 787         if (sock_flag(sk, SOCK_LOCALROUTE) ||
 788             (msg->msg_flags & MSG_DONTROUTE) ||
 789             (ipc.opt && ipc.opt->opt.is_strictroute)) {
 790                 tos |= RTO_ONLINK;
 791         }
 792
 793         if (ipv4_is_multicast(daddr)) {
 794                 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif))
 795                         ipc.oif = inet->mc_index;
 796                 if (!saddr)
 797                         saddr = inet->mc_addr;
 798         } else if (!ipc.oif)
 799                 ipc.oif = inet->uc_index;
 800
 801         flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos,
 802                            RT_SCOPE_UNIVERSE, sk->sk_protocol,
 803                            inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
 804                            sk->sk_uid);
 805
 806         fl4.fl4_icmp_type = user_icmph.type;
 807         fl4.fl4_icmp_code = user_icmph.code;
 808
 809         security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
 810         rt = ip_route_output_flow(net, &fl4, sk);
 811         if (IS_ERR(rt)) {
 812                 err = PTR_ERR(rt);
 813                 rt = NULL;
 814                 if (err == -ENETUNREACH)
 815                         IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
 816                 goto out;
 817         }
 818
 819         err = -EACCES;
 820         if ((rt->rt_flags & RTCF_BROADCAST) &&
 821             !sock_flag(sk, SOCK_BROADCAST))
 822                 goto out;
 823
 824         if (msg->msg_flags & MSG_CONFIRM)
 825                 goto do_confirm;
 826 back_from_confirm:
 827
 828         if (!ipc.addr)
 829                 ipc.addr = fl4.daddr;
 830
 831         lock_sock(sk);
 832
 833         pfh.icmph.type = user_icmph.type; /* already checked */
 834         pfh.icmph.code = user_icmph.code; /* ditto */
 835         pfh.icmph.checksum = 0;
 836         pfh.icmph.un.echo.id = inet->inet_sport;
 837         pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 838         pfh.msg = msg;
 839         pfh.wcheck = 0;
 840         pfh.family = AF_INET;
 841
 842         err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 843                         0, &ipc, &rt, msg->msg_flags);
 844         if (err)
 845                 ip_flush_pending_frames(sk);
 846         else
 847                 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 848         release_sock(sk);
 849
 850 out:
 851         ip_rt_put(rt);
 852 out_free:
 853         if (free)
 854                 kfree(ipc.opt);
 855         if (!err) {
 856                 icmp_out_count(sock_net(sk), user_icmph.type);
 857                 return len;
 858         }
 859         return err;
 860
 861 do_confirm:
 862         if (msg->msg_flags & MSG_PROBE)
 863                 dst_confirm_neigh(&rt->dst, &fl4.daddr);
 864         if (!(msg->msg_flags & MSG_PROBE) || len)
 865                 goto back_from_confirm;
 866         err = 0;
 867         goto out;
 868 }
 869
 870 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
 871                  int flags, int *addr_len)
 872 {
 873         struct inet_sock *isk = inet_sk(sk);
 874         int family = sk->sk_family;
 875         struct sk_buff *skb;
 876         int copied, err;
 877
 878         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 879
 880         err = -EOPNOTSUPP;
 881         if (flags & MSG_OOB)
 882                 goto out;
 883
 884         if (flags & MSG_ERRQUEUE)
 885                 return inet_recv_error(sk, msg, len, addr_len);
 886
 887         skb = skb_recv_datagram(sk, flags, noblock, &err);
 888         if (!skb)
 889                 goto out;
 890
 891         copied = skb->len;
 892         if (copied > len) {
 893                 msg->msg_flags |= MSG_TRUNC;
 894                 copied = len;
 895         }
 896
 897         /* Don't bother checking the checksum */
 898         err = skb_copy_datagram_msg(skb, 0, msg, copied);
 899         if (err)
 900                 goto done;
 901
 902         sock_recv_timestamp(msg, sk, skb);
 903
 904         /* Copy the address and add cmsg data. */
 905         if (family == AF_INET) {
 906                 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 907
 908                 if (sin) {
 909                         sin->sin_family = AF_INET;
 910                         sin->sin_port = 0 /* skb->h.uh->source */;
 911                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 912                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 913                         *addr_len = sizeof(*sin);
 914                 }
 915
 916                 if (isk->cmsg_flags)
 917                         ip_cmsg_recv(msg, skb);
 918
 919 #if IS_ENABLED(CONFIG_IPV6)
 920         } else if (family == AF_INET6) {
 921                 struct ipv6_pinfo *np = inet6_sk(sk);
 922                 struct ipv6hdr *ip6 = ipv6_hdr(skb);
 923                 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 924
 925                 if (sin6) {
 926                         sin6->sin6_family = AF_INET6;
 927                         sin6->sin6_port = 0;
 928                         sin6->sin6_addr = ip6->saddr;
 929                         sin6->sin6_flowinfo = 0;
 930                         if (np->sndflow)
 931                                 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
 932                         sin6->sin6_scope_id =
 933                                 ipv6_iface_scope_id(&sin6->sin6_addr,
 934                                                     inet6_iif(skb));
 935                         *addr_len = sizeof(*sin6);
 936                 }
 937
 938                 if (inet6_sk(sk)->rxopt.all)
 939                         pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
 940                 if (skb->protocol == htons(ETH_P_IPV6) &&
 941                     inet6_sk(sk)->rxopt.all)
 942                         pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
 943                 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
 944                         ip_cmsg_recv(msg, skb);
 945 #endif
 946         } else {
 947                 BUG();
 948         }
 949
 950         err = copied;
 951
 952 done:
 953         skb_free_datagram(sk, skb);
 954 out:
 955         pr_debug("ping_recvmsg -> %d\n", err);
 956         return err;
 957 }
 958 EXPORT_SYMBOL_GPL(ping_recvmsg);
 959
 960 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 961 {
 962         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 963                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
 964         if (sock_queue_rcv_skb(sk, skb) < 0) {
 965                 kfree_skb(skb);
 966                 pr_debug("ping_queue_rcv_skb -> failed\n");
 967                 return -1;
 968         }
 969         return 0;
 970 }
 971 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 972
 973
 974 /*
 975  *      All we need to do is get the socket.
 976  */
 977
 978 bool ping_rcv(struct sk_buff *skb)
 979 {
 980         struct sock *sk;
 981         struct net *net = dev_net(skb->dev);
 982         struct icmphdr *icmph = icmp_hdr(skb);
 983         bool rc = false;
 984
 985         /* We assume the packet has already been checked by icmp_rcv */
 986
 987         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 988                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 989
 990         /* Push ICMP header back */
 991         skb_push(skb, skb->data - (u8 *)icmph);
 992
 993         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 994         if (sk) {
 995                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 996
 997                 pr_debug("rcv on socket %p\n", sk);
 998                 if (skb2 && !ping_queue_rcv_skb(sk, skb2))
 999                         rc = true;
1000                 sock_put(sk);
1001         }
1002
1003         if (!rc)
1004                 pr_debug("no socket, dropping\n");
1005
1006         return rc;
1007 }
1008 EXPORT_SYMBOL_GPL(ping_rcv);
1009
1010 struct proto ping_prot = {
1011         .name =         "PING",
1012         .owner =        THIS_MODULE,
1013         .init =         ping_init_sock,
1014         .close =        ping_close,
1015         .connect =      ip4_datagram_connect,
1016         .disconnect =   __udp_disconnect,
1017         .setsockopt =   ip_setsockopt,
1018         .getsockopt =   ip_getsockopt,
1019         .sendmsg =      ping_v4_sendmsg,
1020         .recvmsg =      ping_recvmsg,
1021         .bind =         ping_bind,
1022         .backlog_rcv =  ping_queue_rcv_skb,
1023         .release_cb =   ip4_datagram_release_cb,
1024         .hash =         ping_hash,
1025         .unhash =       ping_unhash,
1026         .get_port =     ping_get_port,
1027         .obj_size =     sizeof(struct inet_sock),
1028 };
1029 EXPORT_SYMBOL(ping_prot);
1030
1031 #ifdef CONFIG_PROC_FS
1032
1033 static struct sock *ping_get_first(struct seq_file *seq, int start)
1034 {
1035         struct sock *sk;
1036         struct ping_iter_state *state = seq->private;
1037         struct net *net = seq_file_net(seq);
1038
1039         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1040              ++state->bucket) {
1041                 struct hlist_nulls_node *node;
1042                 struct hlist_nulls_head *hslot;
1043
1044                 hslot = &ping_table.hash[state->bucket];
1045
1046                 if (hlist_nulls_empty(hslot))
1047                         continue;
1048
1049                 sk_nulls_for_each(sk, node, hslot) {
1050                         if (net_eq(sock_net(sk), net) &&
1051                             sk->sk_family == state->family)
1052                                 goto found;
1053                 }
1054         }
1055         sk = NULL;
1056 found:
1057         return sk;
1058 }
1059
1060 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1061 {
1062         struct ping_iter_state *state = seq->private;
1063         struct net *net = seq_file_net(seq);
1064
1065         do {
1066                 sk = sk_nulls_next(sk);
1067         } while (sk && (!net_eq(sock_net(sk), net)));
1068
1069         if (!sk)
1070                 return ping_get_first(seq, state->bucket + 1);
1071         return sk;
1072 }
1073
1074 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1075 {
1076         struct sock *sk = ping_get_first(seq, 0);
1077
1078         if (sk)
1079                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1080                         --pos;
1081         return pos ? NULL : sk;
1082 }
1083
1084 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1085         __acquires(ping_table.lock)
1086 {
1087         struct ping_iter_state *state = seq->private;
1088         state->bucket = 0;
1089         state->family = family;
1090
1091         read_lock_bh(&ping_table.lock);
1092
1093         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1094 }
1095 EXPORT_SYMBOL_GPL(ping_seq_start);
1096
1097 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1098 {
1099         return ping_seq_start(seq, pos, AF_INET);
1100 }
1101
1102 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1103 {
1104         struct sock *sk;
1105
1106         if (v == SEQ_START_TOKEN)
1107                 sk = ping_get_idx(seq, 0);
1108         else
1109                 sk = ping_get_next(seq, v);
1110
1111         ++*pos;
1112         return sk;
1113 }
1114 EXPORT_SYMBOL_GPL(ping_seq_next);
1115
1116 void ping_seq_stop(struct seq_file *seq, void *v)
1117         __releases(ping_table.lock)
1118 {
1119         read_unlock_bh(&ping_table.lock);
1120 }
1121 EXPORT_SYMBOL_GPL(ping_seq_stop);
1122
1123 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1124                 int bucket)
1125 {
1126         struct inet_sock *inet = inet_sk(sp);
1127         __be32 dest = inet->inet_daddr;
1128         __be32 src = inet->inet_rcv_saddr;
1129         __u16 destp = ntohs(inet->inet_dport);
1130         __u16 srcp = ntohs(inet->inet_sport);
1131
1132         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1133                 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u",
1134                 bucket, src, srcp, dest, destp, sp->sk_state,
1135                 sk_wmem_alloc_get(sp),
1136                 sk_rmem_alloc_get(sp),
1137                 0, 0L, 0,
1138                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1139                 0, sock_i_ino(sp),
1140                 refcount_read(&sp->sk_refcnt), sp,
1141                 atomic_read(&sp->sk_drops));
1142 }
1143
1144 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1145 {
1146         seq_setwidth(seq, 127);
1147         if (v == SEQ_START_TOKEN)
1148                 seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1149                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1150                            "inode ref pointer drops");
1151         else {
1152                 struct ping_iter_state *state = seq->private;
1153
1154                 ping_v4_format_sock(v, seq, state->bucket);
1155         }
1156         seq_pad(seq, '\n');
1157         return 0;
1158 }
1159
1160 static const struct seq_operations ping_v4_seq_ops = {
1161         .start          = ping_v4_seq_start,
1162         .show           = ping_v4_seq_show,
1163         .next           = ping_seq_next,
1164         .stop           = ping_seq_stop,
1165 };
1166
1167 static int __net_init ping_v4_proc_init_net(struct net *net)
1168 {
1169         if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
1170                         sizeof(struct ping_iter_state)))
1171                 return -ENOMEM;
1172         return 0;
1173 }
1174
1175 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1176 {
1177         remove_proc_entry("icmp", net->proc_net);
1178 }
1179
1180 static struct pernet_operations ping_v4_net_ops = {
1181         .init = ping_v4_proc_init_net,
1182         .exit = ping_v4_proc_exit_net,
1183 };
1184
1185 int __init ping_proc_init(void)
1186 {
1187         return register_pernet_subsys(&ping_v4_net_ops);
1188 }
1189
1190 void ping_proc_exit(void)
1191 {
1192         unregister_pernet_subsys(&ping_v4_net_ops);
1193 }
1194
1195 #endif
1196
1197 void __init ping_init(void)
1198 {
1199         int i;
1200
1201         for (i = 0; i < PING_HTABLE_SIZE; i++)
1202                 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1203         rwlock_init(&ping_table.lock);
1204 }