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", refcount_read(&sk->sk_refcnt)); 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 int chk_addr_ret; 306 307 if (addr_len < sizeof(*addr)) 308 return -EINVAL; 309 310 if (addr->sin_family != AF_INET && 311 !(addr->sin_family == AF_UNSPEC && 312 addr->sin_addr.s_addr == htonl(INADDR_ANY))) 313 return -EAFNOSUPPORT; 314 315 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n", 316 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port)); 317 318 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr); 319 320 if (addr->sin_addr.s_addr == htonl(INADDR_ANY)) 321 chk_addr_ret = RTN_LOCAL; 322 323 if ((!inet_can_nonlocal_bind(net, isk) && 324 chk_addr_ret != RTN_LOCAL) || 325 chk_addr_ret == RTN_MULTICAST || 326 chk_addr_ret == RTN_BROADCAST) 327 return -EADDRNOTAVAIL; 328 329 #if IS_ENABLED(CONFIG_IPV6) 330 } else if (sk->sk_family == AF_INET6) { 331 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 332 int addr_type, scoped, has_addr; 333 struct net_device *dev = NULL; 334 335 if (addr_len < sizeof(*addr)) 336 return -EINVAL; 337 338 if (addr->sin6_family != AF_INET6) 339 return -EAFNOSUPPORT; 340 341 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n", 342 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port)); 343 344 addr_type = ipv6_addr_type(&addr->sin6_addr); 345 scoped = __ipv6_addr_needs_scope_id(addr_type); 346 if ((addr_type != IPV6_ADDR_ANY && 347 !(addr_type & IPV6_ADDR_UNICAST)) || 348 (scoped && !addr->sin6_scope_id)) 349 return -EINVAL; 350 351 rcu_read_lock(); 352 if (addr->sin6_scope_id) { 353 dev = dev_get_by_index_rcu(net, addr->sin6_scope_id); 354 if (!dev) { 355 rcu_read_unlock(); 356 return -ENODEV; 357 } 358 } 359 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev, 360 scoped); 361 rcu_read_unlock(); 362 363 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr || 364 addr_type == IPV6_ADDR_ANY)) 365 return -EADDRNOTAVAIL; 366 367 if (scoped) 368 sk->sk_bound_dev_if = addr->sin6_scope_id; 369 #endif 370 } else { 371 return -EAFNOSUPPORT; 372 } 373 return 0; 374 } 375 376 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr) 377 { 378 if (saddr->sa_family == AF_INET) { 379 struct inet_sock *isk = inet_sk(sk); 380 struct sockaddr_in *addr = (struct sockaddr_in *) saddr; 381 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr; 382 #if IS_ENABLED(CONFIG_IPV6) 383 } else if (saddr->sa_family == AF_INET6) { 384 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr; 385 struct ipv6_pinfo *np = inet6_sk(sk); 386 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr; 387 #endif 388 } 389 } 390 391 static void ping_clear_saddr(struct sock *sk, int dif) 392 { 393 sk->sk_bound_dev_if = dif; 394 if (sk->sk_family == AF_INET) { 395 struct inet_sock *isk = inet_sk(sk); 396 isk->inet_rcv_saddr = isk->inet_saddr = 0; 397 #if IS_ENABLED(CONFIG_IPV6) 398 } else if (sk->sk_family == AF_INET6) { 399 struct ipv6_pinfo *np = inet6_sk(sk); 400 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr)); 401 memset(&np->saddr, 0, sizeof(np->saddr)); 402 #endif 403 } 404 } 405 /* 406 * We need our own bind because there are no privileged id's == local ports. 407 * Moreover, we don't allow binding to multi- and broadcast addresses. 408 */ 409 410 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 411 { 412 struct inet_sock *isk = inet_sk(sk); 413 unsigned short snum; 414 int err; 415 int dif = sk->sk_bound_dev_if; 416 417 err = ping_check_bind_addr(sk, isk, uaddr, addr_len); 418 if (err) 419 return err; 420 421 lock_sock(sk); 422 423 err = -EINVAL; 424 if (isk->inet_num != 0) 425 goto out; 426 427 err = -EADDRINUSE; 428 ping_set_saddr(sk, uaddr); 429 snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port); 430 if (ping_get_port(sk, snum) != 0) { 431 ping_clear_saddr(sk, dif); 432 goto out; 433 } 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_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0); 474 } 475 476 /* 477 * This routine is called by the ICMP module when it gets some 478 * sort of error condition. 479 */ 480 481 void ping_err(struct sk_buff *skb, int offset, u32 info) 482 { 483 int family; 484 struct icmphdr *icmph; 485 struct inet_sock *inet_sock; 486 int type; 487 int code; 488 struct net *net = dev_net(skb->dev); 489 struct sock *sk; 490 int harderr; 491 int err; 492 493 if (skb->protocol == htons(ETH_P_IP)) { 494 family = AF_INET; 495 type = icmp_hdr(skb)->type; 496 code = icmp_hdr(skb)->code; 497 icmph = (struct icmphdr *)(skb->data + offset); 498 } else if (skb->protocol == htons(ETH_P_IPV6)) { 499 family = AF_INET6; 500 type = icmp6_hdr(skb)->icmp6_type; 501 code = icmp6_hdr(skb)->icmp6_code; 502 icmph = (struct icmphdr *) (skb->data + offset); 503 } else { 504 BUG(); 505 } 506 507 /* We assume the packet has already been checked by icmp_unreach */ 508 509 if (!ping_supported(family, icmph->type, icmph->code)) 510 return; 511 512 pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n", 513 skb->protocol, type, code, ntohs(icmph->un.echo.id), 514 ntohs(icmph->un.echo.sequence)); 515 516 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id)); 517 if (!sk) { 518 pr_debug("no socket, dropping\n"); 519 return; /* No socket for error */ 520 } 521 pr_debug("err on socket %p\n", sk); 522 523 err = 0; 524 harderr = 0; 525 inet_sock = inet_sk(sk); 526 527 if (skb->protocol == htons(ETH_P_IP)) { 528 switch (type) { 529 default: 530 case ICMP_TIME_EXCEEDED: 531 err = EHOSTUNREACH; 532 break; 533 case ICMP_SOURCE_QUENCH: 534 /* This is not a real error but ping wants to see it. 535 * Report it with some fake errno. 536 */ 537 err = EREMOTEIO; 538 break; 539 case ICMP_PARAMETERPROB: 540 err = EPROTO; 541 harderr = 1; 542 break; 543 case ICMP_DEST_UNREACH: 544 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ 545 ipv4_sk_update_pmtu(skb, sk, info); 546 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) { 547 err = EMSGSIZE; 548 harderr = 1; 549 break; 550 } 551 goto out; 552 } 553 err = EHOSTUNREACH; 554 if (code <= NR_ICMP_UNREACH) { 555 harderr = icmp_err_convert[code].fatal; 556 err = icmp_err_convert[code].errno; 557 } 558 break; 559 case ICMP_REDIRECT: 560 /* See ICMP_SOURCE_QUENCH */ 561 ipv4_sk_redirect(skb, sk); 562 err = EREMOTEIO; 563 break; 564 } 565 #if IS_ENABLED(CONFIG_IPV6) 566 } else if (skb->protocol == htons(ETH_P_IPV6)) { 567 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err); 568 #endif 569 } 570 571 /* 572 * RFC1122: OK. Passes ICMP errors back to application, as per 573 * 4.1.3.3. 574 */ 575 if ((family == AF_INET && !inet_sock->recverr) || 576 (family == AF_INET6 && !inet6_sk(sk)->recverr)) { 577 if (!harderr || sk->sk_state != TCP_ESTABLISHED) 578 goto out; 579 } else { 580 if (family == AF_INET) { 581 ip_icmp_error(sk, skb, err, 0 /* no remote port */, 582 info, (u8 *)icmph); 583 #if IS_ENABLED(CONFIG_IPV6) 584 } else if (family == AF_INET6) { 585 pingv6_ops.ipv6_icmp_error(sk, skb, err, 0, 586 info, (u8 *)icmph); 587 #endif 588 } 589 } 590 sk->sk_err = err; 591 sk->sk_error_report(sk); 592 out: 593 sock_put(sk); 594 } 595 EXPORT_SYMBOL_GPL(ping_err); 596 597 /* 598 * Copy and checksum an ICMP Echo packet from user space into a buffer 599 * starting from the payload. 600 */ 601 602 int ping_getfrag(void *from, char *to, 603 int offset, int fraglen, int odd, struct sk_buff *skb) 604 { 605 struct pingfakehdr *pfh = (struct pingfakehdr *)from; 606 607 if (offset == 0) { 608 fraglen -= sizeof(struct icmphdr); 609 if (fraglen < 0) 610 BUG(); 611 if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr), 612 fraglen, &pfh->wcheck, 613 &pfh->msg->msg_iter)) 614 return -EFAULT; 615 } else if (offset < sizeof(struct icmphdr)) { 616 BUG(); 617 } else { 618 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck, 619 &pfh->msg->msg_iter)) 620 return -EFAULT; 621 } 622 623 #if IS_ENABLED(CONFIG_IPV6) 624 /* For IPv6, checksum each skb as we go along, as expected by 625 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in 626 * wcheck, it will be finalized in ping_v4_push_pending_frames. 627 */ 628 if (pfh->family == AF_INET6) { 629 skb->csum = pfh->wcheck; 630 skb->ip_summed = CHECKSUM_NONE; 631 pfh->wcheck = 0; 632 } 633 #endif 634 635 return 0; 636 } 637 EXPORT_SYMBOL_GPL(ping_getfrag); 638 639 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh, 640 struct flowi4 *fl4) 641 { 642 struct sk_buff *skb = skb_peek(&sk->sk_write_queue); 643 644 if (!skb) 645 return 0; 646 pfh->wcheck = csum_partial((char *)&pfh->icmph, 647 sizeof(struct icmphdr), pfh->wcheck); 648 pfh->icmph.checksum = csum_fold(pfh->wcheck); 649 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr)); 650 skb->ip_summed = CHECKSUM_NONE; 651 return ip_push_pending_frames(sk, fl4); 652 } 653 654 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len, 655 void *user_icmph, size_t icmph_len) { 656 u8 type, code; 657 658 if (len > 0xFFFF) 659 return -EMSGSIZE; 660 661 /* Must have at least a full ICMP header. */ 662 if (len < icmph_len) 663 return -EINVAL; 664 665 /* 666 * Check the flags. 667 */ 668 669 /* Mirror BSD error message compatibility */ 670 if (msg->msg_flags & MSG_OOB) 671 return -EOPNOTSUPP; 672 673 /* 674 * Fetch the ICMP header provided by the userland. 675 * iovec is modified! The ICMP header is consumed. 676 */ 677 if (memcpy_from_msg(user_icmph, msg, icmph_len)) 678 return -EFAULT; 679 680 if (family == AF_INET) { 681 type = ((struct icmphdr *) user_icmph)->type; 682 code = ((struct icmphdr *) user_icmph)->code; 683 #if IS_ENABLED(CONFIG_IPV6) 684 } else if (family == AF_INET6) { 685 type = ((struct icmp6hdr *) user_icmph)->icmp6_type; 686 code = ((struct icmp6hdr *) user_icmph)->icmp6_code; 687 #endif 688 } else { 689 BUG(); 690 } 691 692 if (!ping_supported(family, type, code)) 693 return -EINVAL; 694 695 return 0; 696 } 697 EXPORT_SYMBOL_GPL(ping_common_sendmsg); 698 699 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 700 { 701 struct net *net = sock_net(sk); 702 struct flowi4 fl4; 703 struct inet_sock *inet = inet_sk(sk); 704 struct ipcm_cookie ipc; 705 struct icmphdr user_icmph; 706 struct pingfakehdr pfh; 707 struct rtable *rt = NULL; 708 struct ip_options_data opt_copy; 709 int free = 0; 710 __be32 saddr, daddr, faddr; 711 u8 tos; 712 int err; 713 714 pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num); 715 716 err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph, 717 sizeof(user_icmph)); 718 if (err) 719 return err; 720 721 /* 722 * Get and verify the address. 723 */ 724 725 if (msg->msg_name) { 726 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); 727 if (msg->msg_namelen < sizeof(*usin)) 728 return -EINVAL; 729 if (usin->sin_family != AF_INET) 730 return -EAFNOSUPPORT; 731 daddr = usin->sin_addr.s_addr; 732 /* no remote port */ 733 } else { 734 if (sk->sk_state != TCP_ESTABLISHED) 735 return -EDESTADDRREQ; 736 daddr = inet->inet_daddr; 737 /* no remote port */ 738 } 739 740 ipcm_init_sk(&ipc, inet); 741 742 if (msg->msg_controllen) { 743 err = ip_cmsg_send(sk, msg, &ipc, false); 744 if (unlikely(err)) { 745 kfree(ipc.opt); 746 return err; 747 } 748 if (ipc.opt) 749 free = 1; 750 } 751 if (!ipc.opt) { 752 struct ip_options_rcu *inet_opt; 753 754 rcu_read_lock(); 755 inet_opt = rcu_dereference(inet->inet_opt); 756 if (inet_opt) { 757 memcpy(&opt_copy, inet_opt, 758 sizeof(*inet_opt) + inet_opt->opt.optlen); 759 ipc.opt = &opt_copy.opt; 760 } 761 rcu_read_unlock(); 762 } 763 764 saddr = ipc.addr; 765 ipc.addr = faddr = daddr; 766 767 if (ipc.opt && ipc.opt->opt.srr) { 768 if (!daddr) { 769 err = -EINVAL; 770 goto out_free; 771 } 772 faddr = ipc.opt->opt.faddr; 773 } 774 tos = get_rttos(&ipc, inet); 775 if (sock_flag(sk, SOCK_LOCALROUTE) || 776 (msg->msg_flags & MSG_DONTROUTE) || 777 (ipc.opt && ipc.opt->opt.is_strictroute)) { 778 tos |= RTO_ONLINK; 779 } 780 781 if (ipv4_is_multicast(daddr)) { 782 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif)) 783 ipc.oif = inet->mc_index; 784 if (!saddr) 785 saddr = inet->mc_addr; 786 } else if (!ipc.oif) 787 ipc.oif = inet->uc_index; 788 789 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos, 790 RT_SCOPE_UNIVERSE, sk->sk_protocol, 791 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0, 792 sk->sk_uid); 793 794 security_sk_classify_flow(sk, flowi4_to_flowi(&fl4)); 795 rt = ip_route_output_flow(net, &fl4, sk); 796 if (IS_ERR(rt)) { 797 err = PTR_ERR(rt); 798 rt = NULL; 799 if (err == -ENETUNREACH) 800 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES); 801 goto out; 802 } 803 804 err = -EACCES; 805 if ((rt->rt_flags & RTCF_BROADCAST) && 806 !sock_flag(sk, SOCK_BROADCAST)) 807 goto out; 808 809 if (msg->msg_flags & MSG_CONFIRM) 810 goto do_confirm; 811 back_from_confirm: 812 813 if (!ipc.addr) 814 ipc.addr = fl4.daddr; 815 816 lock_sock(sk); 817 818 pfh.icmph.type = user_icmph.type; /* already checked */ 819 pfh.icmph.code = user_icmph.code; /* ditto */ 820 pfh.icmph.checksum = 0; 821 pfh.icmph.un.echo.id = inet->inet_sport; 822 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence; 823 pfh.msg = msg; 824 pfh.wcheck = 0; 825 pfh.family = AF_INET; 826 827 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len, 828 0, &ipc, &rt, msg->msg_flags); 829 if (err) 830 ip_flush_pending_frames(sk); 831 else 832 err = ping_v4_push_pending_frames(sk, &pfh, &fl4); 833 release_sock(sk); 834 835 out: 836 ip_rt_put(rt); 837 out_free: 838 if (free) 839 kfree(ipc.opt); 840 if (!err) { 841 icmp_out_count(sock_net(sk), user_icmph.type); 842 return len; 843 } 844 return err; 845 846 do_confirm: 847 if (msg->msg_flags & MSG_PROBE) 848 dst_confirm_neigh(&rt->dst, &fl4.daddr); 849 if (!(msg->msg_flags & MSG_PROBE) || len) 850 goto back_from_confirm; 851 err = 0; 852 goto out; 853 } 854 855 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock, 856 int flags, int *addr_len) 857 { 858 struct inet_sock *isk = inet_sk(sk); 859 int family = sk->sk_family; 860 struct sk_buff *skb; 861 int copied, err; 862 863 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num); 864 865 err = -EOPNOTSUPP; 866 if (flags & MSG_OOB) 867 goto out; 868 869 if (flags & MSG_ERRQUEUE) 870 return inet_recv_error(sk, msg, len, addr_len); 871 872 skb = skb_recv_datagram(sk, flags, noblock, &err); 873 if (!skb) 874 goto out; 875 876 copied = skb->len; 877 if (copied > len) { 878 msg->msg_flags |= MSG_TRUNC; 879 copied = len; 880 } 881 882 /* Don't bother checking the checksum */ 883 err = skb_copy_datagram_msg(skb, 0, msg, copied); 884 if (err) 885 goto done; 886 887 sock_recv_timestamp(msg, sk, skb); 888 889 /* Copy the address and add cmsg data. */ 890 if (family == AF_INET) { 891 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 892 893 if (sin) { 894 sin->sin_family = AF_INET; 895 sin->sin_port = 0 /* skb->h.uh->source */; 896 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 897 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 898 *addr_len = sizeof(*sin); 899 } 900 901 if (isk->cmsg_flags) 902 ip_cmsg_recv(msg, skb); 903 904 #if IS_ENABLED(CONFIG_IPV6) 905 } else if (family == AF_INET6) { 906 struct ipv6_pinfo *np = inet6_sk(sk); 907 struct ipv6hdr *ip6 = ipv6_hdr(skb); 908 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 909 910 if (sin6) { 911 sin6->sin6_family = AF_INET6; 912 sin6->sin6_port = 0; 913 sin6->sin6_addr = ip6->saddr; 914 sin6->sin6_flowinfo = 0; 915 if (np->sndflow) 916 sin6->sin6_flowinfo = ip6_flowinfo(ip6); 917 sin6->sin6_scope_id = 918 ipv6_iface_scope_id(&sin6->sin6_addr, 919 inet6_iif(skb)); 920 *addr_len = sizeof(*sin6); 921 } 922 923 if (inet6_sk(sk)->rxopt.all) 924 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb); 925 if (skb->protocol == htons(ETH_P_IPV6) && 926 inet6_sk(sk)->rxopt.all) 927 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb); 928 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags) 929 ip_cmsg_recv(msg, skb); 930 #endif 931 } else { 932 BUG(); 933 } 934 935 err = copied; 936 937 done: 938 skb_free_datagram(sk, skb); 939 out: 940 pr_debug("ping_recvmsg -> %d\n", err); 941 return err; 942 } 943 EXPORT_SYMBOL_GPL(ping_recvmsg); 944 945 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 946 { 947 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n", 948 inet_sk(sk), inet_sk(sk)->inet_num, skb); 949 if (sock_queue_rcv_skb(sk, skb) < 0) { 950 kfree_skb(skb); 951 pr_debug("ping_queue_rcv_skb -> failed\n"); 952 return -1; 953 } 954 return 0; 955 } 956 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb); 957 958 959 /* 960 * All we need to do is get the socket. 961 */ 962 963 bool ping_rcv(struct sk_buff *skb) 964 { 965 struct sock *sk; 966 struct net *net = dev_net(skb->dev); 967 struct icmphdr *icmph = icmp_hdr(skb); 968 969 /* We assume the packet has already been checked by icmp_rcv */ 970 971 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n", 972 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence)); 973 974 /* Push ICMP header back */ 975 skb_push(skb, skb->data - (u8 *)icmph); 976 977 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id)); 978 if (sk) { 979 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 980 981 pr_debug("rcv on socket %p\n", sk); 982 if (skb2) 983 ping_queue_rcv_skb(sk, skb2); 984 sock_put(sk); 985 return true; 986 } 987 pr_debug("no socket, dropping\n"); 988 989 return false; 990 } 991 EXPORT_SYMBOL_GPL(ping_rcv); 992 993 struct proto ping_prot = { 994 .name = "PING", 995 .owner = THIS_MODULE, 996 .init = ping_init_sock, 997 .close = ping_close, 998 .connect = ip4_datagram_connect, 999 .disconnect = __udp_disconnect, 1000 .setsockopt = ip_setsockopt, 1001 .getsockopt = ip_getsockopt, 1002 .sendmsg = ping_v4_sendmsg, 1003 .recvmsg = ping_recvmsg, 1004 .bind = ping_bind, 1005 .backlog_rcv = ping_queue_rcv_skb, 1006 .release_cb = ip4_datagram_release_cb, 1007 .hash = ping_hash, 1008 .unhash = ping_unhash, 1009 .get_port = ping_get_port, 1010 .obj_size = sizeof(struct inet_sock), 1011 }; 1012 EXPORT_SYMBOL(ping_prot); 1013 1014 #ifdef CONFIG_PROC_FS 1015 1016 static struct sock *ping_get_first(struct seq_file *seq, int start) 1017 { 1018 struct sock *sk; 1019 struct ping_iter_state *state = seq->private; 1020 struct net *net = seq_file_net(seq); 1021 1022 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE; 1023 ++state->bucket) { 1024 struct hlist_nulls_node *node; 1025 struct hlist_nulls_head *hslot; 1026 1027 hslot = &ping_table.hash[state->bucket]; 1028 1029 if (hlist_nulls_empty(hslot)) 1030 continue; 1031 1032 sk_nulls_for_each(sk, node, hslot) { 1033 if (net_eq(sock_net(sk), net) && 1034 sk->sk_family == state->family) 1035 goto found; 1036 } 1037 } 1038 sk = NULL; 1039 found: 1040 return sk; 1041 } 1042 1043 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk) 1044 { 1045 struct ping_iter_state *state = seq->private; 1046 struct net *net = seq_file_net(seq); 1047 1048 do { 1049 sk = sk_nulls_next(sk); 1050 } while (sk && (!net_eq(sock_net(sk), net))); 1051 1052 if (!sk) 1053 return ping_get_first(seq, state->bucket + 1); 1054 return sk; 1055 } 1056 1057 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos) 1058 { 1059 struct sock *sk = ping_get_first(seq, 0); 1060 1061 if (sk) 1062 while (pos && (sk = ping_get_next(seq, sk)) != NULL) 1063 --pos; 1064 return pos ? NULL : sk; 1065 } 1066 1067 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family) 1068 __acquires(ping_table.lock) 1069 { 1070 struct ping_iter_state *state = seq->private; 1071 state->bucket = 0; 1072 state->family = family; 1073 1074 read_lock_bh(&ping_table.lock); 1075 1076 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN; 1077 } 1078 EXPORT_SYMBOL_GPL(ping_seq_start); 1079 1080 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos) 1081 { 1082 return ping_seq_start(seq, pos, AF_INET); 1083 } 1084 1085 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1086 { 1087 struct sock *sk; 1088 1089 if (v == SEQ_START_TOKEN) 1090 sk = ping_get_idx(seq, 0); 1091 else 1092 sk = ping_get_next(seq, v); 1093 1094 ++*pos; 1095 return sk; 1096 } 1097 EXPORT_SYMBOL_GPL(ping_seq_next); 1098 1099 void ping_seq_stop(struct seq_file *seq, void *v) 1100 __releases(ping_table.lock) 1101 { 1102 read_unlock_bh(&ping_table.lock); 1103 } 1104 EXPORT_SYMBOL_GPL(ping_seq_stop); 1105 1106 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f, 1107 int bucket) 1108 { 1109 struct inet_sock *inet = inet_sk(sp); 1110 __be32 dest = inet->inet_daddr; 1111 __be32 src = inet->inet_rcv_saddr; 1112 __u16 destp = ntohs(inet->inet_dport); 1113 __u16 srcp = ntohs(inet->inet_sport); 1114 1115 seq_printf(f, "%5d: %08X:%04X %08X:%04X" 1116 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d", 1117 bucket, src, srcp, dest, destp, sp->sk_state, 1118 sk_wmem_alloc_get(sp), 1119 sk_rmem_alloc_get(sp), 1120 0, 0L, 0, 1121 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), 1122 0, sock_i_ino(sp), 1123 refcount_read(&sp->sk_refcnt), sp, 1124 atomic_read(&sp->sk_drops)); 1125 } 1126 1127 static int ping_v4_seq_show(struct seq_file *seq, void *v) 1128 { 1129 seq_setwidth(seq, 127); 1130 if (v == SEQ_START_TOKEN) 1131 seq_puts(seq, " sl local_address rem_address st tx_queue " 1132 "rx_queue tr tm->when retrnsmt uid timeout " 1133 "inode ref pointer drops"); 1134 else { 1135 struct ping_iter_state *state = seq->private; 1136 1137 ping_v4_format_sock(v, seq, state->bucket); 1138 } 1139 seq_pad(seq, '\n'); 1140 return 0; 1141 } 1142 1143 static const struct seq_operations ping_v4_seq_ops = { 1144 .start = ping_v4_seq_start, 1145 .show = ping_v4_seq_show, 1146 .next = ping_seq_next, 1147 .stop = ping_seq_stop, 1148 }; 1149 1150 static int __net_init ping_v4_proc_init_net(struct net *net) 1151 { 1152 if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops, 1153 sizeof(struct ping_iter_state))) 1154 return -ENOMEM; 1155 return 0; 1156 } 1157 1158 static void __net_exit ping_v4_proc_exit_net(struct net *net) 1159 { 1160 remove_proc_entry("icmp", net->proc_net); 1161 } 1162 1163 static struct pernet_operations ping_v4_net_ops = { 1164 .init = ping_v4_proc_init_net, 1165 .exit = ping_v4_proc_exit_net, 1166 }; 1167 1168 int __init ping_proc_init(void) 1169 { 1170 return register_pernet_subsys(&ping_v4_net_ops); 1171 } 1172 1173 void ping_proc_exit(void) 1174 { 1175 unregister_pernet_subsys(&ping_v4_net_ops); 1176 } 1177 1178 #endif 1179 1180 void __init ping_init(void) 1181 { 1182 int i; 1183 1184 for (i = 0; i < PING_HTABLE_SIZE; i++) 1185 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i); 1186 rwlock_init(&ping_table.lock); 1187 } 1188