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                         continue;
 230
 231                 sock_hold(sk);
 232                 goto exit;
 233         }
 234
 235         sk = NULL;
 236 exit:
 237         read_unlock_bh(&ping_table.lock);
 238
 239         return sk;
 240 }
 241
 242 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
 243                                           kgid_t *high)
 244 {
 245         kgid_t *data = net->ipv4.ping_group_range.range;
 246         unsigned int seq;
 247
 248         do {
 249                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 250
 251                 *low = data[0];
 252                 *high = data[1];
 253         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 254 }
 255
 256
 257 int ping_init_sock(struct sock *sk)
 258 {
 259         struct net *net = sock_net(sk);
 260         kgid_t group = current_egid();
 261         struct group_info *group_info;
 262         int i;
 263         kgid_t low, high;
 264         int ret = 0;
 265
 266         if (sk->sk_family == AF_INET6)
 267                 sk->sk_ipv6only = 1;
 268
 269         inet_get_ping_group_range_net(net, &low, &high);
 270         if (gid_lte(low, group) && gid_lte(group, high))
 271                 return 0;
 272
 273         group_info = get_current_groups();
 274         for (i = 0; i < group_info->ngroups; i++) {
 275                 kgid_t gid = group_info->gid[i];
 276
 277                 if (gid_lte(low, gid) && gid_lte(gid, high))
 278                         goto out_release_group;
 279         }
 280
 281         ret = -EACCES;
 282
 283 out_release_group:
 284         put_group_info(group_info);
 285         return ret;
 286 }
 287 EXPORT_SYMBOL_GPL(ping_init_sock);
 288
 289 void ping_close(struct sock *sk, long timeout)
 290 {
 291         pr_debug("ping_close(sk=%p,sk->num=%u)\n",
 292                  inet_sk(sk), inet_sk(sk)->inet_num);
 293         pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter);
 294
 295         sk_common_release(sk);
 296 }
 297 EXPORT_SYMBOL_GPL(ping_close);
 298
 299 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
 300 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
 301                                 struct sockaddr *uaddr, int addr_len) {
 302         struct net *net = sock_net(sk);
 303         if (sk->sk_family == AF_INET) {
 304                 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 305                 u32 tb_id = RT_TABLE_LOCAL;
 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                 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id;
 320                 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id);
 321
 322                 if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
 323                         chk_addr_ret = RTN_LOCAL;
 324
 325                 if ((net->ipv4.sysctl_ip_nonlocal_bind == 0 &&
 326                     isk->freebind == 0 && isk->transparent == 0 &&
 327                      chk_addr_ret != RTN_LOCAL) ||
 328                     chk_addr_ret == RTN_MULTICAST ||
 329                     chk_addr_ret == RTN_BROADCAST)
 330                         return -EADDRNOTAVAIL;
 331
 332 #if IS_ENABLED(CONFIG_IPV6)
 333         } else if (sk->sk_family == AF_INET6) {
 334                 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 335                 int addr_type, scoped, has_addr;
 336                 struct net_device *dev = NULL;
 337
 338                 if (addr_len < sizeof(*addr))
 339                         return -EINVAL;
 340
 341                 if (addr->sin6_family != AF_INET6)
 342                         return -EAFNOSUPPORT;
 343
 344                 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 345                          sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 346
 347                 addr_type = ipv6_addr_type(&addr->sin6_addr);
 348                 scoped = __ipv6_addr_needs_scope_id(addr_type);
 349                 if ((addr_type != IPV6_ADDR_ANY &&
 350                      !(addr_type & IPV6_ADDR_UNICAST)) ||
 351                     (scoped && !addr->sin6_scope_id))
 352                         return -EINVAL;
 353
 354                 rcu_read_lock();
 355                 if (addr->sin6_scope_id) {
 356                         dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 357                         if (!dev) {
 358                                 rcu_read_unlock();
 359                                 return -ENODEV;
 360                         }
 361                 }
 362
 363                 if (!dev && sk->sk_bound_dev_if) {
 364                         dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
 365                         if (!dev) {
 366                                 rcu_read_unlock();
 367                                 return -ENODEV;
 368                         }
 369                 }
 370                 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 371                                                     scoped);
 372                 rcu_read_unlock();
 373
 374                 if (!(net->ipv6.sysctl.ip_nonlocal_bind ||
 375                       isk->freebind || isk->transparent || 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 = %d, dif = %d\n",
 448                  (int)isk->inet_num,
 449                  (int)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(to + sizeof(struct icmphdr),
 624                             fraglen, &pfh->wcheck,
 625                             &pfh->msg->msg_iter) != fraglen)
 626                         return -EFAULT;
 627         } else if (offset < sizeof(struct icmphdr)) {
 628                         BUG();
 629         } else {
 630                 if (csum_and_copy_from_iter(to, fraglen, &pfh->wcheck,
 631                                             &pfh->msg->msg_iter) != fraglen)
 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         ipc.sockc.tsflags = sk->sk_tsflags;
 753         ipc.addr = inet->inet_saddr;
 754         ipc.opt = NULL;
 755         ipc.oif = sk->sk_bound_dev_if;
 756         ipc.tx_flags = 0;
 757         ipc.ttl = 0;
 758         ipc.tos = -1;
 759
 760         if (msg->msg_controllen) {
 761                 err = ip_cmsg_send(sk, msg, &ipc, false);
 762                 if (unlikely(err)) {
 763                         kfree(ipc.opt);
 764                         return err;
 765                 }
 766                 if (ipc.opt)
 767                         free = 1;
 768         }
 769         if (!ipc.opt) {
 770                 struct ip_options_rcu *inet_opt;
 771
 772                 rcu_read_lock();
 773                 inet_opt = rcu_dereference(inet->inet_opt);
 774                 if (inet_opt) {
 775                         memcpy(&opt_copy, inet_opt,
 776                                sizeof(*inet_opt) + inet_opt->opt.optlen);
 777                         ipc.opt = &opt_copy.opt;
 778                 }
 779                 rcu_read_unlock();
 780         }
 781
 782         sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
 783
 784         saddr = ipc.addr;
 785         ipc.addr = faddr = daddr;
 786
 787         if (ipc.opt && ipc.opt->opt.srr) {
 788                 if (!daddr) {
 789                         err = -EINVAL;
 790                         goto out_free;
 791                 }
 792                 faddr = ipc.opt->opt.faddr;
 793         }
 794         tos = get_rttos(&ipc, inet);
 795         if (sock_flag(sk, SOCK_LOCALROUTE) ||
 796             (msg->msg_flags & MSG_DONTROUTE) ||
 797             (ipc.opt && ipc.opt->opt.is_strictroute)) {
 798                 tos |= RTO_ONLINK;
 799         }
 800
 801         if (ipv4_is_multicast(daddr)) {
 802                 if (!ipc.oif)
 803                         ipc.oif = inet->mc_index;
 804                 if (!saddr)
 805                         saddr = inet->mc_addr;
 806         } else if (!ipc.oif)
 807                 ipc.oif = inet->uc_index;
 808
 809         flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
 810                            RT_SCOPE_UNIVERSE, sk->sk_protocol,
 811                            inet_sk_flowi_flags(sk), faddr, saddr, 0, 0);
 812
 813         fl4.fl4_icmp_type = user_icmph.type;
 814         fl4.fl4_icmp_code = user_icmph.code;
 815
 816         security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
 817         rt = ip_route_output_flow(net, &fl4, sk);
 818         if (IS_ERR(rt)) {
 819                 err = PTR_ERR(rt);
 820                 rt = NULL;
 821                 if (err == -ENETUNREACH)
 822                         IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
 823                 goto out;
 824         }
 825
 826         err = -EACCES;
 827         if ((rt->rt_flags & RTCF_BROADCAST) &&
 828             !sock_flag(sk, SOCK_BROADCAST))
 829                 goto out;
 830
 831         if (msg->msg_flags & MSG_CONFIRM)
 832                 goto do_confirm;
 833 back_from_confirm:
 834
 835         if (!ipc.addr)
 836                 ipc.addr = fl4.daddr;
 837
 838         lock_sock(sk);
 839
 840         pfh.icmph.type = user_icmph.type; /* already checked */
 841         pfh.icmph.code = user_icmph.code; /* ditto */
 842         pfh.icmph.checksum = 0;
 843         pfh.icmph.un.echo.id = inet->inet_sport;
 844         pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 845         pfh.msg = msg;
 846         pfh.wcheck = 0;
 847         pfh.family = AF_INET;
 848
 849         err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 850                         0, &ipc, &rt, msg->msg_flags);
 851         if (err)
 852                 ip_flush_pending_frames(sk);
 853         else
 854                 err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 855         release_sock(sk);
 856
 857 out:
 858         ip_rt_put(rt);
 859 out_free:
 860         if (free)
 861                 kfree(ipc.opt);
 862         if (!err) {
 863                 icmp_out_count(sock_net(sk), user_icmph.type);
 864                 return len;
 865         }
 866         return err;
 867
 868 do_confirm:
 869         dst_confirm(&rt->dst);
 870         if (!(msg->msg_flags & MSG_PROBE) || len)
 871                 goto back_from_confirm;
 872         err = 0;
 873         goto out;
 874 }
 875
 876 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
 877                  int flags, int *addr_len)
 878 {
 879         struct inet_sock *isk = inet_sk(sk);
 880         int family = sk->sk_family;
 881         struct sk_buff *skb;
 882         int copied, err;
 883
 884         pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 885
 886         err = -EOPNOTSUPP;
 887         if (flags & MSG_OOB)
 888                 goto out;
 889
 890         if (flags & MSG_ERRQUEUE)
 891                 return inet_recv_error(sk, msg, len, addr_len);
 892
 893         skb = skb_recv_datagram(sk, flags, noblock, &err);
 894         if (!skb)
 895                 goto out;
 896
 897         copied = skb->len;
 898         if (copied > len) {
 899                 msg->msg_flags |= MSG_TRUNC;
 900                 copied = len;
 901         }
 902
 903         /* Don't bother checking the checksum */
 904         err = skb_copy_datagram_msg(skb, 0, msg, copied);
 905         if (err)
 906                 goto done;
 907
 908         sock_recv_timestamp(msg, sk, skb);
 909
 910         /* Copy the address and add cmsg data. */
 911         if (family == AF_INET) {
 912                 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 913
 914                 if (sin) {
 915                         sin->sin_family = AF_INET;
 916                         sin->sin_port = 0 /* skb->h.uh->source */;
 917                         sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 918                         memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 919                         *addr_len = sizeof(*sin);
 920                 }
 921
 922                 if (isk->cmsg_flags)
 923                         ip_cmsg_recv(msg, skb);
 924
 925 #if IS_ENABLED(CONFIG_IPV6)
 926         } else if (family == AF_INET6) {
 927                 struct ipv6_pinfo *np = inet6_sk(sk);
 928                 struct ipv6hdr *ip6 = ipv6_hdr(skb);
 929                 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 930
 931                 if (sin6) {
 932                         sin6->sin6_family = AF_INET6;
 933                         sin6->sin6_port = 0;
 934                         sin6->sin6_addr = ip6->saddr;
 935                         sin6->sin6_flowinfo = 0;
 936                         if (np->sndflow)
 937                                 sin6->sin6_flowinfo = ip6_flowinfo(ip6);
 938                         sin6->sin6_scope_id =
 939                                 ipv6_iface_scope_id(&sin6->sin6_addr,
 940                                                     inet6_iif(skb));
 941                         *addr_len = sizeof(*sin6);
 942                 }
 943
 944                 if (inet6_sk(sk)->rxopt.all)
 945                         pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
 946                 if (skb->protocol == htons(ETH_P_IPV6) &&
 947                     inet6_sk(sk)->rxopt.all)
 948                         pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
 949                 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
 950                         ip_cmsg_recv(msg, skb);
 951 #endif
 952         } else {
 953                 BUG();
 954         }
 955
 956         err = copied;
 957
 958 done:
 959         skb_free_datagram(sk, skb);
 960 out:
 961         pr_debug("ping_recvmsg -> %d\n", err);
 962         return err;
 963 }
 964 EXPORT_SYMBOL_GPL(ping_recvmsg);
 965
 966 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 967 {
 968         pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 969                  inet_sk(sk), inet_sk(sk)->inet_num, skb);
 970         if (sock_queue_rcv_skb(sk, skb) < 0) {
 971                 kfree_skb(skb);
 972                 pr_debug("ping_queue_rcv_skb -> failed\n");
 973                 return -1;
 974         }
 975         return 0;
 976 }
 977 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 978
 979
 980 /*
 981  *      All we need to do is get the socket.
 982  */
 983
 984 bool ping_rcv(struct sk_buff *skb)
 985 {
 986         struct sock *sk;
 987         struct net *net = dev_net(skb->dev);
 988         struct icmphdr *icmph = icmp_hdr(skb);
 989         bool rc = false;
 990
 991         /* We assume the packet has already been checked by icmp_rcv */
 992
 993         pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 994                  skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 995
 996         /* Push ICMP header back */
 997         skb_push(skb, skb->data - (u8 *)icmph);
 998
 999         sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
1000         if (sk) {
1001                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1002
1003                 pr_debug("rcv on socket %p\n", sk);
1004                 if (skb2 && !ping_queue_rcv_skb(sk, skb2))
1005                         rc = true;
1006                 sock_put(sk);
1007         }
1008
1009         if (!rc)
1010                 pr_debug("no socket, dropping\n");
1011
1012         return rc;
1013 }
1014 EXPORT_SYMBOL_GPL(ping_rcv);
1015
1016 struct proto ping_prot = {
1017         .name =         "PING",
1018         .owner =        THIS_MODULE,
1019         .init =         ping_init_sock,
1020         .close =        ping_close,
1021         .connect =      ip4_datagram_connect,
1022         .disconnect =   __udp_disconnect,
1023         .setsockopt =   ip_setsockopt,
1024         .getsockopt =   ip_getsockopt,
1025         .sendmsg =      ping_v4_sendmsg,
1026         .recvmsg =      ping_recvmsg,
1027         .bind =         ping_bind,
1028         .backlog_rcv =  ping_queue_rcv_skb,
1029         .release_cb =   ip4_datagram_release_cb,
1030         .hash =         ping_hash,
1031         .unhash =       ping_unhash,
1032         .get_port =     ping_get_port,
1033         .obj_size =     sizeof(struct inet_sock),
1034 };
1035 EXPORT_SYMBOL(ping_prot);
1036
1037 #ifdef CONFIG_PROC_FS
1038
1039 static struct sock *ping_get_first(struct seq_file *seq, int start)
1040 {
1041         struct sock *sk;
1042         struct ping_iter_state *state = seq->private;
1043         struct net *net = seq_file_net(seq);
1044
1045         for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1046              ++state->bucket) {
1047                 struct hlist_nulls_node *node;
1048                 struct hlist_nulls_head *hslot;
1049
1050                 hslot = &ping_table.hash[state->bucket];
1051
1052                 if (hlist_nulls_empty(hslot))
1053                         continue;
1054
1055                 sk_nulls_for_each(sk, node, hslot) {
1056                         if (net_eq(sock_net(sk), net) &&
1057                             sk->sk_family == state->family)
1058                                 goto found;
1059                 }
1060         }
1061         sk = NULL;
1062 found:
1063         return sk;
1064 }
1065
1066 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1067 {
1068         struct ping_iter_state *state = seq->private;
1069         struct net *net = seq_file_net(seq);
1070
1071         do {
1072                 sk = sk_nulls_next(sk);
1073         } while (sk && (!net_eq(sock_net(sk), net)));
1074
1075         if (!sk)
1076                 return ping_get_first(seq, state->bucket + 1);
1077         return sk;
1078 }
1079
1080 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1081 {
1082         struct sock *sk = ping_get_first(seq, 0);
1083
1084         if (sk)
1085                 while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1086                         --pos;
1087         return pos ? NULL : sk;
1088 }
1089
1090 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1091         __acquires(ping_table.lock)
1092 {
1093         struct ping_iter_state *state = seq->private;
1094         state->bucket = 0;
1095         state->family = family;
1096
1097         read_lock_bh(&ping_table.lock);
1098
1099         return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1100 }
1101 EXPORT_SYMBOL_GPL(ping_seq_start);
1102
1103 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1104 {
1105         return ping_seq_start(seq, pos, AF_INET);
1106 }
1107
1108 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1109 {
1110         struct sock *sk;
1111
1112         if (v == SEQ_START_TOKEN)
1113                 sk = ping_get_idx(seq, 0);
1114         else
1115                 sk = ping_get_next(seq, v);
1116
1117         ++*pos;
1118         return sk;
1119 }
1120 EXPORT_SYMBOL_GPL(ping_seq_next);
1121
1122 void ping_seq_stop(struct seq_file *seq, void *v)
1123         __releases(ping_table.lock)
1124 {
1125         read_unlock_bh(&ping_table.lock);
1126 }
1127 EXPORT_SYMBOL_GPL(ping_seq_stop);
1128
1129 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1130                 int bucket)
1131 {
1132         struct inet_sock *inet = inet_sk(sp);
1133         __be32 dest = inet->inet_daddr;
1134         __be32 src = inet->inet_rcv_saddr;
1135         __u16 destp = ntohs(inet->inet_dport);
1136         __u16 srcp = ntohs(inet->inet_sport);
1137
1138         seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1139                 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1140                 bucket, src, srcp, dest, destp, sp->sk_state,
1141                 sk_wmem_alloc_get(sp),
1142                 sk_rmem_alloc_get(sp),
1143                 0, 0L, 0,
1144                 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1145                 0, sock_i_ino(sp),
1146                 atomic_read(&sp->sk_refcnt), sp,
1147                 atomic_read(&sp->sk_drops));
1148 }
1149
1150 static int ping_v4_seq_show(struct seq_file *seq, void *v)
1151 {
1152         seq_setwidth(seq, 127);
1153         if (v == SEQ_START_TOKEN)
1154                 seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1155                            "rx_queue tr tm->when retrnsmt   uid  timeout "
1156                            "inode ref pointer drops");
1157         else {
1158                 struct ping_iter_state *state = seq->private;
1159
1160                 ping_v4_format_sock(v, seq, state->bucket);
1161         }
1162         seq_pad(seq, '\n');
1163         return 0;
1164 }
1165
1166 static int ping_seq_open(struct inode *inode, struct file *file)
1167 {
1168         struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1169         return seq_open_net(inode, file, &afinfo->seq_ops,
1170                            sizeof(struct ping_iter_state));
1171 }
1172
1173 const struct file_operations ping_seq_fops = {
1174         .open           = ping_seq_open,
1175         .read           = seq_read,
1176         .llseek         = seq_lseek,
1177         .release        = seq_release_net,
1178 };
1179 EXPORT_SYMBOL_GPL(ping_seq_fops);
1180
1181 static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1182         .name           = "icmp",
1183         .family         = AF_INET,
1184         .seq_fops       = &ping_seq_fops,
1185         .seq_ops        = {
1186                 .start          = ping_v4_seq_start,
1187                 .show           = ping_v4_seq_show,
1188                 .next           = ping_seq_next,
1189                 .stop           = ping_seq_stop,
1190         },
1191 };
1192
1193 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1194 {
1195         struct proc_dir_entry *p;
1196         p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1197                              afinfo->seq_fops, afinfo);
1198         if (!p)
1199                 return -ENOMEM;
1200         return 0;
1201 }
1202 EXPORT_SYMBOL_GPL(ping_proc_register);
1203
1204 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1205 {
1206         remove_proc_entry(afinfo->name, net->proc_net);
1207 }
1208 EXPORT_SYMBOL_GPL(ping_proc_unregister);
1209
1210 static int __net_init ping_v4_proc_init_net(struct net *net)
1211 {
1212         return ping_proc_register(net, &ping_v4_seq_afinfo);
1213 }
1214
1215 static void __net_exit ping_v4_proc_exit_net(struct net *net)
1216 {
1217         ping_proc_unregister(net, &ping_v4_seq_afinfo);
1218 }
1219
1220 static struct pernet_operations ping_v4_net_ops = {
1221         .init = ping_v4_proc_init_net,
1222         .exit = ping_v4_proc_exit_net,
1223 };
1224
1225 int __init ping_proc_init(void)
1226 {
1227         return register_pernet_subsys(&ping_v4_net_ops);
1228 }
1229
1230 void ping_proc_exit(void)
1231 {
1232         unregister_pernet_subsys(&ping_v4_net_ops);
1233 }
1234
1235 #endif
1236
1237 void __init ping_init(void)
1238 {
1239         int i;
1240
1241         for (i = 0; i < PING_HTABLE_SIZE; i++)
1242                 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1243         rwlock_init(&ping_table.lock);
1244 }