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 { 305 struct net *net = sock_net(sk); 306 if (sk->sk_family == AF_INET) { 307 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 308 u32 tb_id = RT_TABLE_LOCAL; 309 int chk_addr_ret; 310 311 if (addr_len < sizeof(*addr)) 312 return -EINVAL; 313 314 if (addr->sin_family != AF_INET && 315 !(addr->sin_family == AF_UNSPEC && 316 addr->sin_addr.s_addr == htonl(INADDR_ANY))) 317 return -EAFNOSUPPORT; 318 319 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n", 320 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port)); 321 322 if (addr->sin_addr.s_addr == htonl(INADDR_ANY)) 323 return 0; 324 325 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id; 326 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); 327 328 if (chk_addr_ret == RTN_MULTICAST || 329 chk_addr_ret == RTN_BROADCAST || 330 (chk_addr_ret != RTN_LOCAL && 331 !inet_can_nonlocal_bind(net, isk))) 332 return -EADDRNOTAVAIL; 333 334 #if IS_ENABLED(CONFIG_IPV6) 335 } else if (sk->sk_family == AF_INET6) { 336 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 337 int addr_type, scoped, has_addr; 338 struct net_device *dev = NULL; 339 340 if (addr_len < sizeof(*addr)) 341 return -EINVAL; 342 343 if (addr->sin6_family != AF_INET6) 344 return -EAFNOSUPPORT; 345 346 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n", 347 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port)); 348 349 addr_type = ipv6_addr_type(&addr->sin6_addr); 350 scoped = __ipv6_addr_needs_scope_id(addr_type); 351 if ((addr_type != IPV6_ADDR_ANY && 352 !(addr_type & IPV6_ADDR_UNICAST)) || 353 (scoped && !addr->sin6_scope_id)) 354 return -EINVAL; 355 356 rcu_read_lock(); 357 if (addr->sin6_scope_id) { 358 dev = dev_get_by_index_rcu(net, addr->sin6_scope_id); 359 if (!dev) { 360 rcu_read_unlock(); 361 return -ENODEV; 362 } 363 } 364 365 if (!dev && sk->sk_bound_dev_if) { 366 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 367 if (!dev) { 368 rcu_read_unlock(); 369 return -ENODEV; 370 } 371 } 372 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev, 373 scoped); 374 rcu_read_unlock(); 375 376 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr || 377 addr_type == IPV6_ADDR_ANY)) 378 return -EADDRNOTAVAIL; 379 380 if (scoped) 381 sk->sk_bound_dev_if = addr->sin6_scope_id; 382 #endif 383 } else { 384 return -EAFNOSUPPORT; 385 } 386 return 0; 387 } 388 389 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr) 390 { 391 if (saddr->sa_family == AF_INET) { 392 struct inet_sock *isk = inet_sk(sk); 393 struct sockaddr_in *addr = (struct sockaddr_in *) saddr; 394 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr; 395 #if IS_ENABLED(CONFIG_IPV6) 396 } else if (saddr->sa_family == AF_INET6) { 397 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr; 398 struct ipv6_pinfo *np = inet6_sk(sk); 399 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr; 400 #endif 401 } 402 } 403 404 /* 405 * We need our own bind because there are no privileged id's == local ports. 406 * Moreover, we don't allow binding to multi- and broadcast addresses. 407 */ 408 409 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 410 { 411 struct inet_sock *isk = inet_sk(sk); 412 unsigned short snum; 413 int err; 414 int dif = sk->sk_bound_dev_if; 415 416 err = ping_check_bind_addr(sk, isk, uaddr, addr_len); 417 if (err) 418 return err; 419 420 lock_sock(sk); 421 422 err = -EINVAL; 423 if (isk->inet_num != 0) 424 goto out; 425 426 err = -EADDRINUSE; 427 snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port); 428 if (ping_get_port(sk, snum) != 0) { 429 /* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */ 430 sk->sk_bound_dev_if = dif; 431 goto out; 432 } 433 ping_set_saddr(sk, uaddr); 434 435 pr_debug("after bind(): num = %hu, dif = %d\n", 436 isk->inet_num, 437 sk->sk_bound_dev_if); 438 439 err = 0; 440 if (sk->sk_family == AF_INET && isk->inet_rcv_saddr) 441 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 442 #if IS_ENABLED(CONFIG_IPV6) 443 if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 444 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 445 #endif 446 447 if (snum) 448 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 449 isk->inet_sport = htons(isk->inet_num); 450 isk->inet_daddr = 0; 451 isk->inet_dport = 0; 452 453 #if IS_ENABLED(CONFIG_IPV6) 454 if (sk->sk_family == AF_INET6) 455 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr)); 456 #endif 457 458 sk_dst_reset(sk); 459 out: 460 release_sock(sk); 461 pr_debug("ping_v4_bind -> %d\n", err); 462 return err; 463 } 464 EXPORT_SYMBOL_GPL(ping_bind); 465 466 /* 467 * Is this a supported type of ICMP message? 468 */ 469 470 static inline int ping_supported(int family, int type, int code) 471 { 472 return (family == AF_INET && type == ICMP_ECHO && code == 0) || 473 (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) || 474 (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) || 475 (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0); 476 } 477 478 /* 479 * This routine is called by the ICMP module when it gets some 480 * sort of error condition. 481 */ 482 483 void ping_err(struct sk_buff *skb, int offset, u32 info) 484 { 485 int family; 486 struct icmphdr *icmph; 487 struct inet_sock *inet_sock; 488 int type; 489 int code; 490 struct net *net = dev_net(skb->dev); 491 struct sock *sk; 492 int harderr; 493 int err; 494 495 if (skb->protocol == htons(ETH_P_IP)) { 496 family = AF_INET; 497 type = icmp_hdr(skb)->type; 498 code = icmp_hdr(skb)->code; 499 icmph = (struct icmphdr *)(skb->data + offset); 500 } else if (skb->protocol == htons(ETH_P_IPV6)) { 501 family = AF_INET6; 502 type = icmp6_hdr(skb)->icmp6_type; 503 code = icmp6_hdr(skb)->icmp6_code; 504 icmph = (struct icmphdr *) (skb->data + offset); 505 } else { 506 BUG(); 507 } 508 509 /* We assume the packet has already been checked by icmp_unreach */ 510 511 if (!ping_supported(family, icmph->type, icmph->code)) 512 return; 513 514 pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n", 515 skb->protocol, type, code, ntohs(icmph->un.echo.id), 516 ntohs(icmph->un.echo.sequence)); 517 518 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id)); 519 if (!sk) { 520 pr_debug("no socket, dropping\n"); 521 return; /* No socket for error */ 522 } 523 pr_debug("err on socket %p\n", sk); 524 525 err = 0; 526 harderr = 0; 527 inet_sock = inet_sk(sk); 528 529 if (skb->protocol == htons(ETH_P_IP)) { 530 switch (type) { 531 default: 532 case ICMP_TIME_EXCEEDED: 533 err = EHOSTUNREACH; 534 break; 535 case ICMP_SOURCE_QUENCH: 536 /* This is not a real error but ping wants to see it. 537 * Report it with some fake errno. 538 */ 539 err = EREMOTEIO; 540 break; 541 case ICMP_PARAMETERPROB: 542 err = EPROTO; 543 harderr = 1; 544 break; 545 case ICMP_DEST_UNREACH: 546 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ 547 ipv4_sk_update_pmtu(skb, sk, info); 548 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) { 549 err = EMSGSIZE; 550 harderr = 1; 551 break; 552 } 553 goto out; 554 } 555 err = EHOSTUNREACH; 556 if (code <= NR_ICMP_UNREACH) { 557 harderr = icmp_err_convert[code].fatal; 558 err = icmp_err_convert[code].errno; 559 } 560 break; 561 case ICMP_REDIRECT: 562 /* See ICMP_SOURCE_QUENCH */ 563 ipv4_sk_redirect(skb, sk); 564 err = EREMOTEIO; 565 break; 566 } 567 #if IS_ENABLED(CONFIG_IPV6) 568 } else if (skb->protocol == htons(ETH_P_IPV6)) { 569 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err); 570 #endif 571 } 572 573 /* 574 * RFC1122: OK. Passes ICMP errors back to application, as per 575 * 4.1.3.3. 576 */ 577 if ((family == AF_INET && !inet_sock->recverr) || 578 (family == AF_INET6 && !inet6_sk(sk)->recverr)) { 579 if (!harderr || sk->sk_state != TCP_ESTABLISHED) 580 goto out; 581 } else { 582 if (family == AF_INET) { 583 ip_icmp_error(sk, skb, err, 0 /* no remote port */, 584 info, (u8 *)icmph); 585 #if IS_ENABLED(CONFIG_IPV6) 586 } else if (family == AF_INET6) { 587 pingv6_ops.ipv6_icmp_error(sk, skb, err, 0, 588 info, (u8 *)icmph); 589 #endif 590 } 591 } 592 sk->sk_err = err; 593 sk_error_report(sk); 594 out: 595 sock_put(sk); 596 } 597 EXPORT_SYMBOL_GPL(ping_err); 598 599 /* 600 * Copy and checksum an ICMP Echo packet from user space into a buffer 601 * starting from the payload. 602 */ 603 604 int ping_getfrag(void *from, char *to, 605 int offset, int fraglen, int odd, struct sk_buff *skb) 606 { 607 struct pingfakehdr *pfh = from; 608 609 if (offset == 0) { 610 fraglen -= sizeof(struct icmphdr); 611 if (fraglen < 0) 612 BUG(); 613 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr), 614 fraglen, &pfh->wcheck, 615 &pfh->msg->msg_iter)) 616 return -EFAULT; 617 } else if (offset < sizeof(struct icmphdr)) { 618 BUG(); 619 } else { 620 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck, 621 &pfh->msg->msg_iter)) 622 return -EFAULT; 623 } 624 625 #if IS_ENABLED(CONFIG_IPV6) 626 /* For IPv6, checksum each skb as we go along, as expected by 627 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in 628 * wcheck, it will be finalized in ping_v4_push_pending_frames. 629 */ 630 if (pfh->family == AF_INET6) { 631 skb->csum = pfh->wcheck; 632 skb->ip_summed = CHECKSUM_NONE; 633 pfh->wcheck = 0; 634 } 635 #endif 636 637 return 0; 638 } 639 EXPORT_SYMBOL_GPL(ping_getfrag); 640 641 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh, 642 struct flowi4 *fl4) 643 { 644 struct sk_buff *skb = skb_peek(&sk->sk_write_queue); 645 646 if (!skb) 647 return 0; 648 pfh->wcheck = csum_partial((char *)&pfh->icmph, 649 sizeof(struct icmphdr), pfh->wcheck); 650 pfh->icmph.checksum = csum_fold(pfh->wcheck); 651 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr)); 652 skb->ip_summed = CHECKSUM_NONE; 653 return ip_push_pending_frames(sk, fl4); 654 } 655 656 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len, 657 void *user_icmph, size_t icmph_len) 658 { 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 ipcm_init_sk(&ipc, inet); 744 745 if (msg->msg_controllen) { 746 err = ip_cmsg_send(sk, msg, &ipc, false); 747 if (unlikely(err)) { 748 kfree(ipc.opt); 749 return err; 750 } 751 if (ipc.opt) 752 free = 1; 753 } 754 if (!ipc.opt) { 755 struct ip_options_rcu *inet_opt; 756 757 rcu_read_lock(); 758 inet_opt = rcu_dereference(inet->inet_opt); 759 if (inet_opt) { 760 memcpy(&opt_copy, inet_opt, 761 sizeof(*inet_opt) + inet_opt->opt.optlen); 762 ipc.opt = &opt_copy.opt; 763 } 764 rcu_read_unlock(); 765 } 766 767 saddr = ipc.addr; 768 ipc.addr = faddr = daddr; 769 770 if (ipc.opt && ipc.opt->opt.srr) { 771 if (!daddr) { 772 err = -EINVAL; 773 goto out_free; 774 } 775 faddr = ipc.opt->opt.faddr; 776 } 777 tos = get_rttos(&ipc, inet); 778 if (sock_flag(sk, SOCK_LOCALROUTE) || 779 (msg->msg_flags & MSG_DONTROUTE) || 780 (ipc.opt && ipc.opt->opt.is_strictroute)) { 781 tos |= RTO_ONLINK; 782 } 783 784 if (ipv4_is_multicast(daddr)) { 785 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif)) 786 ipc.oif = inet->mc_index; 787 if (!saddr) 788 saddr = inet->mc_addr; 789 } else if (!ipc.oif) 790 ipc.oif = inet->uc_index; 791 792 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, tos, 793 RT_SCOPE_UNIVERSE, sk->sk_protocol, 794 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0, 795 sk->sk_uid); 796 797 fl4.fl4_icmp_type = user_icmph.type; 798 fl4.fl4_icmp_code = user_icmph.code; 799 800 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4)); 801 rt = ip_route_output_flow(net, &fl4, sk); 802 if (IS_ERR(rt)) { 803 err = PTR_ERR(rt); 804 rt = NULL; 805 if (err == -ENETUNREACH) 806 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES); 807 goto out; 808 } 809 810 err = -EACCES; 811 if ((rt->rt_flags & RTCF_BROADCAST) && 812 !sock_flag(sk, SOCK_BROADCAST)) 813 goto out; 814 815 if (msg->msg_flags & MSG_CONFIRM) 816 goto do_confirm; 817 back_from_confirm: 818 819 if (!ipc.addr) 820 ipc.addr = fl4.daddr; 821 822 lock_sock(sk); 823 824 pfh.icmph.type = user_icmph.type; /* already checked */ 825 pfh.icmph.code = user_icmph.code; /* ditto */ 826 pfh.icmph.checksum = 0; 827 pfh.icmph.un.echo.id = inet->inet_sport; 828 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence; 829 pfh.msg = msg; 830 pfh.wcheck = 0; 831 pfh.family = AF_INET; 832 833 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len, 834 0, &ipc, &rt, msg->msg_flags); 835 if (err) 836 ip_flush_pending_frames(sk); 837 else 838 err = ping_v4_push_pending_frames(sk, &pfh, &fl4); 839 release_sock(sk); 840 841 out: 842 ip_rt_put(rt); 843 out_free: 844 if (free) 845 kfree(ipc.opt); 846 if (!err) { 847 icmp_out_count(sock_net(sk), user_icmph.type); 848 return len; 849 } 850 return err; 851 852 do_confirm: 853 if (msg->msg_flags & MSG_PROBE) 854 dst_confirm_neigh(&rt->dst, &fl4.daddr); 855 if (!(msg->msg_flags & MSG_PROBE) || len) 856 goto back_from_confirm; 857 err = 0; 858 goto out; 859 } 860 861 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int flags, 862 int *addr_len) 863 { 864 struct inet_sock *isk = inet_sk(sk); 865 int family = sk->sk_family; 866 struct sk_buff *skb; 867 int copied, err; 868 869 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num); 870 871 err = -EOPNOTSUPP; 872 if (flags & MSG_OOB) 873 goto out; 874 875 if (flags & MSG_ERRQUEUE) 876 return inet_recv_error(sk, msg, len, addr_len); 877 878 skb = skb_recv_datagram(sk, flags, &err); 879 if (!skb) 880 goto out; 881 882 copied = skb->len; 883 if (copied > len) { 884 msg->msg_flags |= MSG_TRUNC; 885 copied = len; 886 } 887 888 /* Don't bother checking the checksum */ 889 err = skb_copy_datagram_msg(skb, 0, msg, copied); 890 if (err) 891 goto done; 892 893 sock_recv_timestamp(msg, sk, skb); 894 895 /* Copy the address and add cmsg data. */ 896 if (family == AF_INET) { 897 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 898 899 if (sin) { 900 sin->sin_family = AF_INET; 901 sin->sin_port = 0 /* skb->h.uh->source */; 902 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 903 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 904 *addr_len = sizeof(*sin); 905 } 906 907 if (isk->cmsg_flags) 908 ip_cmsg_recv(msg, skb); 909 910 #if IS_ENABLED(CONFIG_IPV6) 911 } else if (family == AF_INET6) { 912 struct ipv6_pinfo *np = inet6_sk(sk); 913 struct ipv6hdr *ip6 = ipv6_hdr(skb); 914 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 915 916 if (sin6) { 917 sin6->sin6_family = AF_INET6; 918 sin6->sin6_port = 0; 919 sin6->sin6_addr = ip6->saddr; 920 sin6->sin6_flowinfo = 0; 921 if (np->sndflow) 922 sin6->sin6_flowinfo = ip6_flowinfo(ip6); 923 sin6->sin6_scope_id = 924 ipv6_iface_scope_id(&sin6->sin6_addr, 925 inet6_iif(skb)); 926 *addr_len = sizeof(*sin6); 927 } 928 929 if (inet6_sk(sk)->rxopt.all) 930 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb); 931 if (skb->protocol == htons(ETH_P_IPV6) && 932 inet6_sk(sk)->rxopt.all) 933 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb); 934 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags) 935 ip_cmsg_recv(msg, skb); 936 #endif 937 } else { 938 BUG(); 939 } 940 941 err = copied; 942 943 done: 944 skb_free_datagram(sk, skb); 945 out: 946 pr_debug("ping_recvmsg -> %d\n", err); 947 return err; 948 } 949 EXPORT_SYMBOL_GPL(ping_recvmsg); 950 951 static enum skb_drop_reason __ping_queue_rcv_skb(struct sock *sk, 952 struct sk_buff *skb) 953 { 954 enum skb_drop_reason reason; 955 956 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n", 957 inet_sk(sk), inet_sk(sk)->inet_num, skb); 958 if (sock_queue_rcv_skb_reason(sk, skb, &reason) < 0) { 959 kfree_skb_reason(skb, reason); 960 pr_debug("ping_queue_rcv_skb -> failed\n"); 961 return reason; 962 } 963 return SKB_NOT_DROPPED_YET; 964 } 965 966 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 967 { 968 return __ping_queue_rcv_skb(sk, skb) ? -1 : 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 enum skb_drop_reason ping_rcv(struct sk_buff *skb) 978 { 979 enum skb_drop_reason reason = SKB_DROP_REASON_NO_SOCKET; 980 struct sock *sk; 981 struct net *net = dev_net(skb->dev); 982 struct icmphdr *icmph = icmp_hdr(skb); 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) 998 reason = __ping_queue_rcv_skb(sk, skb2); 999 else 1000 reason = SKB_DROP_REASON_NOMEM; 1001 sock_put(sk); 1002 } 1003 1004 if (reason) 1005 pr_debug("no socket, dropping\n"); 1006 1007 return reason; 1008 } 1009 EXPORT_SYMBOL_GPL(ping_rcv); 1010 1011 struct proto ping_prot = { 1012 .name = "PING", 1013 .owner = THIS_MODULE, 1014 .init = ping_init_sock, 1015 .close = ping_close, 1016 .connect = ip4_datagram_connect, 1017 .disconnect = __udp_disconnect, 1018 .setsockopt = ip_setsockopt, 1019 .getsockopt = ip_getsockopt, 1020 .sendmsg = ping_v4_sendmsg, 1021 .recvmsg = ping_recvmsg, 1022 .bind = ping_bind, 1023 .backlog_rcv = ping_queue_rcv_skb, 1024 .release_cb = ip4_datagram_release_cb, 1025 .hash = ping_hash, 1026 .unhash = ping_unhash, 1027 .get_port = ping_get_port, 1028 .put_port = ping_unhash, 1029 .obj_size = sizeof(struct inet_sock), 1030 }; 1031 EXPORT_SYMBOL(ping_prot); 1032 1033 #ifdef CONFIG_PROC_FS 1034 1035 static struct sock *ping_get_first(struct seq_file *seq, int start) 1036 { 1037 struct sock *sk; 1038 struct ping_iter_state *state = seq->private; 1039 struct net *net = seq_file_net(seq); 1040 1041 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE; 1042 ++state->bucket) { 1043 struct hlist_nulls_node *node; 1044 struct hlist_nulls_head *hslot; 1045 1046 hslot = &ping_table.hash[state->bucket]; 1047 1048 if (hlist_nulls_empty(hslot)) 1049 continue; 1050 1051 sk_nulls_for_each(sk, node, hslot) { 1052 if (net_eq(sock_net(sk), net) && 1053 sk->sk_family == state->family) 1054 goto found; 1055 } 1056 } 1057 sk = NULL; 1058 found: 1059 return sk; 1060 } 1061 1062 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk) 1063 { 1064 struct ping_iter_state *state = seq->private; 1065 struct net *net = seq_file_net(seq); 1066 1067 do { 1068 sk = sk_nulls_next(sk); 1069 } while (sk && (!net_eq(sock_net(sk), net))); 1070 1071 if (!sk) 1072 return ping_get_first(seq, state->bucket + 1); 1073 return sk; 1074 } 1075 1076 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos) 1077 { 1078 struct sock *sk = ping_get_first(seq, 0); 1079 1080 if (sk) 1081 while (pos && (sk = ping_get_next(seq, sk)) != NULL) 1082 --pos; 1083 return pos ? NULL : sk; 1084 } 1085 1086 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family) 1087 __acquires(ping_table.lock) 1088 { 1089 struct ping_iter_state *state = seq->private; 1090 state->bucket = 0; 1091 state->family = family; 1092 1093 read_lock_bh(&ping_table.lock); 1094 1095 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN; 1096 } 1097 EXPORT_SYMBOL_GPL(ping_seq_start); 1098 1099 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos) 1100 { 1101 return ping_seq_start(seq, pos, AF_INET); 1102 } 1103 1104 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1105 { 1106 struct sock *sk; 1107 1108 if (v == SEQ_START_TOKEN) 1109 sk = ping_get_idx(seq, 0); 1110 else 1111 sk = ping_get_next(seq, v); 1112 1113 ++*pos; 1114 return sk; 1115 } 1116 EXPORT_SYMBOL_GPL(ping_seq_next); 1117 1118 void ping_seq_stop(struct seq_file *seq, void *v) 1119 __releases(ping_table.lock) 1120 { 1121 read_unlock_bh(&ping_table.lock); 1122 } 1123 EXPORT_SYMBOL_GPL(ping_seq_stop); 1124 1125 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f, 1126 int bucket) 1127 { 1128 struct inet_sock *inet = inet_sk(sp); 1129 __be32 dest = inet->inet_daddr; 1130 __be32 src = inet->inet_rcv_saddr; 1131 __u16 destp = ntohs(inet->inet_dport); 1132 __u16 srcp = ntohs(inet->inet_sport); 1133 1134 seq_printf(f, "%5d: %08X:%04X %08X:%04X" 1135 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %u", 1136 bucket, src, srcp, dest, destp, sp->sk_state, 1137 sk_wmem_alloc_get(sp), 1138 sk_rmem_alloc_get(sp), 1139 0, 0L, 0, 1140 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), 1141 0, sock_i_ino(sp), 1142 refcount_read(&sp->sk_refcnt), sp, 1143 atomic_read(&sp->sk_drops)); 1144 } 1145 1146 static int ping_v4_seq_show(struct seq_file *seq, void *v) 1147 { 1148 seq_setwidth(seq, 127); 1149 if (v == SEQ_START_TOKEN) 1150 seq_puts(seq, " sl local_address rem_address st tx_queue " 1151 "rx_queue tr tm->when retrnsmt uid timeout " 1152 "inode ref pointer drops"); 1153 else { 1154 struct ping_iter_state *state = seq->private; 1155 1156 ping_v4_format_sock(v, seq, state->bucket); 1157 } 1158 seq_pad(seq, '\n'); 1159 return 0; 1160 } 1161 1162 static const struct seq_operations ping_v4_seq_ops = { 1163 .start = ping_v4_seq_start, 1164 .show = ping_v4_seq_show, 1165 .next = ping_seq_next, 1166 .stop = ping_seq_stop, 1167 }; 1168 1169 static int __net_init ping_v4_proc_init_net(struct net *net) 1170 { 1171 if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops, 1172 sizeof(struct ping_iter_state))) 1173 return -ENOMEM; 1174 return 0; 1175 } 1176 1177 static void __net_exit ping_v4_proc_exit_net(struct net *net) 1178 { 1179 remove_proc_entry("icmp", net->proc_net); 1180 } 1181 1182 static struct pernet_operations ping_v4_net_ops = { 1183 .init = ping_v4_proc_init_net, 1184 .exit = ping_v4_proc_exit_net, 1185 }; 1186 1187 int __init ping_proc_init(void) 1188 { 1189 return register_pernet_subsys(&ping_v4_net_ops); 1190 } 1191 1192 void ping_proc_exit(void) 1193 { 1194 unregister_pernet_subsys(&ping_v4_net_ops); 1195 } 1196 1197 #endif 1198 1199 void __init ping_init(void) 1200 { 1201 int i; 1202 1203 for (i = 0; i < PING_HTABLE_SIZE; i++) 1204 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i); 1205 rwlock_init(&ping_table.lock); 1206 } 1207