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 pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num); 160 if (sk_hashed(sk)) { 161 write_lock_bh(&ping_table.lock); 162 hlist_nulls_del(&sk->sk_nulls_node); 163 sk_nulls_node_init(&sk->sk_nulls_node); 164 sock_put(sk); 165 isk->inet_num = 0; 166 isk->inet_sport = 0; 167 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 168 write_unlock_bh(&ping_table.lock); 169 } 170 } 171 EXPORT_SYMBOL_GPL(ping_unhash); 172 173 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident) 174 { 175 struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident); 176 struct sock *sk = NULL; 177 struct inet_sock *isk; 178 struct hlist_nulls_node *hnode; 179 int dif = skb->dev->ifindex; 180 181 if (skb->protocol == htons(ETH_P_IP)) { 182 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n", 183 (int)ident, &ip_hdr(skb)->daddr, dif); 184 #if IS_ENABLED(CONFIG_IPV6) 185 } else if (skb->protocol == htons(ETH_P_IPV6)) { 186 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n", 187 (int)ident, &ipv6_hdr(skb)->daddr, dif); 188 #endif 189 } 190 191 read_lock_bh(&ping_table.lock); 192 193 ping_portaddr_for_each_entry(sk, hnode, hslot) { 194 isk = inet_sk(sk); 195 196 pr_debug("iterate\n"); 197 if (isk->inet_num != ident) 198 continue; 199 200 if (skb->protocol == htons(ETH_P_IP) && 201 sk->sk_family == AF_INET) { 202 pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk, 203 (int) isk->inet_num, &isk->inet_rcv_saddr, 204 sk->sk_bound_dev_if); 205 206 if (isk->inet_rcv_saddr && 207 isk->inet_rcv_saddr != ip_hdr(skb)->daddr) 208 continue; 209 #if IS_ENABLED(CONFIG_IPV6) 210 } else if (skb->protocol == htons(ETH_P_IPV6) && 211 sk->sk_family == AF_INET6) { 212 213 pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk, 214 (int) isk->inet_num, 215 &sk->sk_v6_rcv_saddr, 216 sk->sk_bound_dev_if); 217 218 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) && 219 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, 220 &ipv6_hdr(skb)->daddr)) 221 continue; 222 #endif 223 } else { 224 continue; 225 } 226 227 if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) 228 continue; 229 230 sock_hold(sk); 231 goto exit; 232 } 233 234 sk = NULL; 235 exit: 236 read_unlock_bh(&ping_table.lock); 237 238 return sk; 239 } 240 241 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low, 242 kgid_t *high) 243 { 244 kgid_t *data = net->ipv4.ping_group_range.range; 245 unsigned int seq; 246 247 do { 248 seq = read_seqbegin(&net->ipv4.ping_group_range.lock); 249 250 *low = data[0]; 251 *high = data[1]; 252 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq)); 253 } 254 255 256 int ping_init_sock(struct sock *sk) 257 { 258 struct net *net = sock_net(sk); 259 kgid_t group = current_egid(); 260 struct group_info *group_info; 261 int i; 262 kgid_t low, high; 263 int ret = 0; 264 265 if (sk->sk_family == AF_INET6) 266 sk->sk_ipv6only = 1; 267 268 inet_get_ping_group_range_net(net, &low, &high); 269 if (gid_lte(low, group) && gid_lte(group, high)) 270 return 0; 271 272 group_info = get_current_groups(); 273 for (i = 0; i < group_info->ngroups; i++) { 274 kgid_t gid = group_info->gid[i]; 275 276 if (gid_lte(low, gid) && gid_lte(gid, high)) 277 goto out_release_group; 278 } 279 280 ret = -EACCES; 281 282 out_release_group: 283 put_group_info(group_info); 284 return ret; 285 } 286 EXPORT_SYMBOL_GPL(ping_init_sock); 287 288 void ping_close(struct sock *sk, long timeout) 289 { 290 pr_debug("ping_close(sk=%p,sk->num=%u)\n", 291 inet_sk(sk), inet_sk(sk)->inet_num); 292 pr_debug("isk->refcnt = %d\n", sk->sk_refcnt.counter); 293 294 sk_common_release(sk); 295 } 296 EXPORT_SYMBOL_GPL(ping_close); 297 298 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */ 299 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk, 300 struct sockaddr *uaddr, int addr_len) { 301 struct net *net = sock_net(sk); 302 if (sk->sk_family == AF_INET) { 303 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 304 int chk_addr_ret; 305 306 if (addr_len < sizeof(*addr)) 307 return -EINVAL; 308 309 if (addr->sin_family != AF_INET && 310 !(addr->sin_family == AF_UNSPEC && 311 addr->sin_addr.s_addr == htonl(INADDR_ANY))) 312 return -EAFNOSUPPORT; 313 314 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n", 315 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port)); 316 317 chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr); 318 319 if (addr->sin_addr.s_addr == htonl(INADDR_ANY)) 320 chk_addr_ret = RTN_LOCAL; 321 322 if ((net->ipv4.sysctl_ip_nonlocal_bind == 0 && 323 isk->freebind == 0 && isk->transparent == 0 && 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 (!(net->ipv6.sysctl.ip_nonlocal_bind || 364 isk->freebind || isk->transparent || has_addr || 365 addr_type == IPV6_ADDR_ANY)) 366 return -EADDRNOTAVAIL; 367 368 if (scoped) 369 sk->sk_bound_dev_if = addr->sin6_scope_id; 370 #endif 371 } else { 372 return -EAFNOSUPPORT; 373 } 374 return 0; 375 } 376 377 static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr) 378 { 379 if (saddr->sa_family == AF_INET) { 380 struct inet_sock *isk = inet_sk(sk); 381 struct sockaddr_in *addr = (struct sockaddr_in *) saddr; 382 isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr; 383 #if IS_ENABLED(CONFIG_IPV6) 384 } else if (saddr->sa_family == AF_INET6) { 385 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr; 386 struct ipv6_pinfo *np = inet6_sk(sk); 387 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr; 388 #endif 389 } 390 } 391 392 static void ping_clear_saddr(struct sock *sk, int dif) 393 { 394 sk->sk_bound_dev_if = dif; 395 if (sk->sk_family == AF_INET) { 396 struct inet_sock *isk = inet_sk(sk); 397 isk->inet_rcv_saddr = isk->inet_saddr = 0; 398 #if IS_ENABLED(CONFIG_IPV6) 399 } else if (sk->sk_family == AF_INET6) { 400 struct ipv6_pinfo *np = inet6_sk(sk); 401 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr)); 402 memset(&np->saddr, 0, sizeof(np->saddr)); 403 #endif 404 } 405 } 406 /* 407 * We need our own bind because there are no privileged id's == local ports. 408 * Moreover, we don't allow binding to multi- and broadcast addresses. 409 */ 410 411 int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 412 { 413 struct inet_sock *isk = inet_sk(sk); 414 unsigned short snum; 415 int err; 416 int dif = sk->sk_bound_dev_if; 417 418 err = ping_check_bind_addr(sk, isk, uaddr, addr_len); 419 if (err) 420 return err; 421 422 lock_sock(sk); 423 424 err = -EINVAL; 425 if (isk->inet_num != 0) 426 goto out; 427 428 err = -EADDRINUSE; 429 ping_set_saddr(sk, uaddr); 430 snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port); 431 if (ping_get_port(sk, snum) != 0) { 432 ping_clear_saddr(sk, dif); 433 goto out; 434 } 435 436 pr_debug("after bind(): num = %d, dif = %d\n", 437 (int)isk->inet_num, 438 (int)sk->sk_bound_dev_if); 439 440 err = 0; 441 if (sk->sk_family == AF_INET && isk->inet_rcv_saddr) 442 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 443 #if IS_ENABLED(CONFIG_IPV6) 444 if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 445 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 446 #endif 447 448 if (snum) 449 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 450 isk->inet_sport = htons(isk->inet_num); 451 isk->inet_daddr = 0; 452 isk->inet_dport = 0; 453 454 #if IS_ENABLED(CONFIG_IPV6) 455 if (sk->sk_family == AF_INET6) 456 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr)); 457 #endif 458 459 sk_dst_reset(sk); 460 out: 461 release_sock(sk); 462 pr_debug("ping_v4_bind -> %d\n", err); 463 return err; 464 } 465 EXPORT_SYMBOL_GPL(ping_bind); 466 467 /* 468 * Is this a supported type of ICMP message? 469 */ 470 471 static inline int ping_supported(int family, int type, int code) 472 { 473 return (family == AF_INET && type == ICMP_ECHO && code == 0) || 474 (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0); 475 } 476 477 /* 478 * This routine is called by the ICMP module when it gets some 479 * sort of error condition. 480 */ 481 482 void ping_err(struct sk_buff *skb, int offset, u32 info) 483 { 484 int family; 485 struct icmphdr *icmph; 486 struct inet_sock *inet_sock; 487 int type; 488 int code; 489 struct net *net = dev_net(skb->dev); 490 struct sock *sk; 491 int harderr; 492 int err; 493 494 if (skb->protocol == htons(ETH_P_IP)) { 495 family = AF_INET; 496 type = icmp_hdr(skb)->type; 497 code = icmp_hdr(skb)->code; 498 icmph = (struct icmphdr *)(skb->data + offset); 499 } else if (skb->protocol == htons(ETH_P_IPV6)) { 500 family = AF_INET6; 501 type = icmp6_hdr(skb)->icmp6_type; 502 code = icmp6_hdr(skb)->icmp6_code; 503 icmph = (struct icmphdr *) (skb->data + offset); 504 } else { 505 BUG(); 506 } 507 508 /* We assume the packet has already been checked by icmp_unreach */ 509 510 if (!ping_supported(family, icmph->type, icmph->code)) 511 return; 512 513 pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n", 514 skb->protocol, type, code, ntohs(icmph->un.echo.id), 515 ntohs(icmph->un.echo.sequence)); 516 517 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id)); 518 if (!sk) { 519 pr_debug("no socket, dropping\n"); 520 return; /* No socket for error */ 521 } 522 pr_debug("err on socket %p\n", sk); 523 524 err = 0; 525 harderr = 0; 526 inet_sock = inet_sk(sk); 527 528 if (skb->protocol == htons(ETH_P_IP)) { 529 switch (type) { 530 default: 531 case ICMP_TIME_EXCEEDED: 532 err = EHOSTUNREACH; 533 break; 534 case ICMP_SOURCE_QUENCH: 535 /* This is not a real error but ping wants to see it. 536 * Report it with some fake errno. 537 */ 538 err = EREMOTEIO; 539 break; 540 case ICMP_PARAMETERPROB: 541 err = EPROTO; 542 harderr = 1; 543 break; 544 case ICMP_DEST_UNREACH: 545 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ 546 ipv4_sk_update_pmtu(skb, sk, info); 547 if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) { 548 err = EMSGSIZE; 549 harderr = 1; 550 break; 551 } 552 goto out; 553 } 554 err = EHOSTUNREACH; 555 if (code <= NR_ICMP_UNREACH) { 556 harderr = icmp_err_convert[code].fatal; 557 err = icmp_err_convert[code].errno; 558 } 559 break; 560 case ICMP_REDIRECT: 561 /* See ICMP_SOURCE_QUENCH */ 562 ipv4_sk_redirect(skb, sk); 563 err = EREMOTEIO; 564 break; 565 } 566 #if IS_ENABLED(CONFIG_IPV6) 567 } else if (skb->protocol == htons(ETH_P_IPV6)) { 568 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err); 569 #endif 570 } 571 572 /* 573 * RFC1122: OK. Passes ICMP errors back to application, as per 574 * 4.1.3.3. 575 */ 576 if ((family == AF_INET && !inet_sock->recverr) || 577 (family == AF_INET6 && !inet6_sk(sk)->recverr)) { 578 if (!harderr || sk->sk_state != TCP_ESTABLISHED) 579 goto out; 580 } else { 581 if (family == AF_INET) { 582 ip_icmp_error(sk, skb, err, 0 /* no remote port */, 583 info, (u8 *)icmph); 584 #if IS_ENABLED(CONFIG_IPV6) 585 } else if (family == AF_INET6) { 586 pingv6_ops.ipv6_icmp_error(sk, skb, err, 0, 587 info, (u8 *)icmph); 588 #endif 589 } 590 } 591 sk->sk_err = err; 592 sk->sk_error_report(sk); 593 out: 594 sock_put(sk); 595 } 596 EXPORT_SYMBOL_GPL(ping_err); 597 598 /* 599 * Copy and checksum an ICMP Echo packet from user space into a buffer 600 * starting from the payload. 601 */ 602 603 int ping_getfrag(void *from, char *to, 604 int offset, int fraglen, int odd, struct sk_buff *skb) 605 { 606 struct pingfakehdr *pfh = (struct pingfakehdr *)from; 607 608 if (offset == 0) { 609 fraglen -= sizeof(struct icmphdr); 610 if (fraglen < 0) 611 BUG(); 612 if (csum_and_copy_from_iter(to + sizeof(struct icmphdr), 613 fraglen, &pfh->wcheck, 614 &pfh->msg->msg_iter) != fraglen) 615 return -EFAULT; 616 } else if (offset < sizeof(struct icmphdr)) { 617 BUG(); 618 } else { 619 if (csum_and_copy_from_iter(to, fraglen, &pfh->wcheck, 620 &pfh->msg->msg_iter) != fraglen) 621 return -EFAULT; 622 } 623 624 #if IS_ENABLED(CONFIG_IPV6) 625 /* For IPv6, checksum each skb as we go along, as expected by 626 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in 627 * wcheck, it will be finalized in ping_v4_push_pending_frames. 628 */ 629 if (pfh->family == AF_INET6) { 630 skb->csum = pfh->wcheck; 631 skb->ip_summed = CHECKSUM_NONE; 632 pfh->wcheck = 0; 633 } 634 #endif 635 636 return 0; 637 } 638 EXPORT_SYMBOL_GPL(ping_getfrag); 639 640 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh, 641 struct flowi4 *fl4) 642 { 643 struct sk_buff *skb = skb_peek(&sk->sk_write_queue); 644 645 pfh->wcheck = csum_partial((char *)&pfh->icmph, 646 sizeof(struct icmphdr), pfh->wcheck); 647 pfh->icmph.checksum = csum_fold(pfh->wcheck); 648 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr)); 649 skb->ip_summed = CHECKSUM_NONE; 650 return ip_push_pending_frames(sk, fl4); 651 } 652 653 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len, 654 void *user_icmph, size_t icmph_len) { 655 u8 type, code; 656 657 if (len > 0xFFFF) 658 return -EMSGSIZE; 659 660 /* 661 * Check the flags. 662 */ 663 664 /* Mirror BSD error message compatibility */ 665 if (msg->msg_flags & MSG_OOB) 666 return -EOPNOTSUPP; 667 668 /* 669 * Fetch the ICMP header provided by the userland. 670 * iovec is modified! The ICMP header is consumed. 671 */ 672 if (memcpy_from_msg(user_icmph, msg, icmph_len)) 673 return -EFAULT; 674 675 if (family == AF_INET) { 676 type = ((struct icmphdr *) user_icmph)->type; 677 code = ((struct icmphdr *) user_icmph)->code; 678 #if IS_ENABLED(CONFIG_IPV6) 679 } else if (family == AF_INET6) { 680 type = ((struct icmp6hdr *) user_icmph)->icmp6_type; 681 code = ((struct icmp6hdr *) user_icmph)->icmp6_code; 682 #endif 683 } else { 684 BUG(); 685 } 686 687 if (!ping_supported(family, type, code)) 688 return -EINVAL; 689 690 return 0; 691 } 692 EXPORT_SYMBOL_GPL(ping_common_sendmsg); 693 694 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 695 { 696 struct net *net = sock_net(sk); 697 struct flowi4 fl4; 698 struct inet_sock *inet = inet_sk(sk); 699 struct ipcm_cookie ipc; 700 struct icmphdr user_icmph; 701 struct pingfakehdr pfh; 702 struct rtable *rt = NULL; 703 struct ip_options_data opt_copy; 704 int free = 0; 705 __be32 saddr, daddr, faddr; 706 u8 tos; 707 int err; 708 709 pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num); 710 711 err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph, 712 sizeof(user_icmph)); 713 if (err) 714 return err; 715 716 /* 717 * Get and verify the address. 718 */ 719 720 if (msg->msg_name) { 721 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); 722 if (msg->msg_namelen < sizeof(*usin)) 723 return -EINVAL; 724 if (usin->sin_family != AF_INET) 725 return -EAFNOSUPPORT; 726 daddr = usin->sin_addr.s_addr; 727 /* no remote port */ 728 } else { 729 if (sk->sk_state != TCP_ESTABLISHED) 730 return -EDESTADDRREQ; 731 daddr = inet->inet_daddr; 732 /* no remote port */ 733 } 734 735 ipc.sockc.tsflags = sk->sk_tsflags; 736 ipc.addr = inet->inet_saddr; 737 ipc.opt = NULL; 738 ipc.oif = sk->sk_bound_dev_if; 739 ipc.tx_flags = 0; 740 ipc.ttl = 0; 741 ipc.tos = -1; 742 743 if (msg->msg_controllen) { 744 err = ip_cmsg_send(sk, msg, &ipc, false); 745 if (unlikely(err)) { 746 kfree(ipc.opt); 747 return err; 748 } 749 if (ipc.opt) 750 free = 1; 751 } 752 if (!ipc.opt) { 753 struct ip_options_rcu *inet_opt; 754 755 rcu_read_lock(); 756 inet_opt = rcu_dereference(inet->inet_opt); 757 if (inet_opt) { 758 memcpy(&opt_copy, inet_opt, 759 sizeof(*inet_opt) + inet_opt->opt.optlen); 760 ipc.opt = &opt_copy.opt; 761 } 762 rcu_read_unlock(); 763 } 764 765 sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags); 766 767 saddr = ipc.addr; 768 ipc.addr = faddr = daddr; 769 770 if (ipc.opt && ipc.opt->opt.srr) { 771 if (!daddr) 772 return -EINVAL; 773 faddr = ipc.opt->opt.faddr; 774 } 775 tos = get_rttos(&ipc, inet); 776 if (sock_flag(sk, SOCK_LOCALROUTE) || 777 (msg->msg_flags & MSG_DONTROUTE) || 778 (ipc.opt && ipc.opt->opt.is_strictroute)) { 779 tos |= RTO_ONLINK; 780 } 781 782 if (ipv4_is_multicast(daddr)) { 783 if (!ipc.oif) 784 ipc.oif = inet->mc_index; 785 if (!saddr) 786 saddr = inet->mc_addr; 787 } else if (!ipc.oif) 788 ipc.oif = inet->uc_index; 789 790 flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos, 791 RT_SCOPE_UNIVERSE, sk->sk_protocol, 792 inet_sk_flowi_flags(sk), faddr, saddr, 0, 0); 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 if (free) 838 kfree(ipc.opt); 839 if (!err) { 840 icmp_out_count(sock_net(sk), user_icmph.type); 841 return len; 842 } 843 return err; 844 845 do_confirm: 846 dst_confirm(&rt->dst); 847 if (!(msg->msg_flags & MSG_PROBE) || len) 848 goto back_from_confirm; 849 err = 0; 850 goto out; 851 } 852 853 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock, 854 int flags, int *addr_len) 855 { 856 struct inet_sock *isk = inet_sk(sk); 857 int family = sk->sk_family; 858 struct sk_buff *skb; 859 int copied, err; 860 861 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num); 862 863 err = -EOPNOTSUPP; 864 if (flags & MSG_OOB) 865 goto out; 866 867 if (flags & MSG_ERRQUEUE) 868 return inet_recv_error(sk, msg, len, addr_len); 869 870 skb = skb_recv_datagram(sk, flags, noblock, &err); 871 if (!skb) 872 goto out; 873 874 copied = skb->len; 875 if (copied > len) { 876 msg->msg_flags |= MSG_TRUNC; 877 copied = len; 878 } 879 880 /* Don't bother checking the checksum */ 881 err = skb_copy_datagram_msg(skb, 0, msg, copied); 882 if (err) 883 goto done; 884 885 sock_recv_timestamp(msg, sk, skb); 886 887 /* Copy the address and add cmsg data. */ 888 if (family == AF_INET) { 889 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 890 891 if (sin) { 892 sin->sin_family = AF_INET; 893 sin->sin_port = 0 /* skb->h.uh->source */; 894 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 895 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 896 *addr_len = sizeof(*sin); 897 } 898 899 if (isk->cmsg_flags) 900 ip_cmsg_recv(msg, skb); 901 902 #if IS_ENABLED(CONFIG_IPV6) 903 } else if (family == AF_INET6) { 904 struct ipv6_pinfo *np = inet6_sk(sk); 905 struct ipv6hdr *ip6 = ipv6_hdr(skb); 906 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 907 908 if (sin6) { 909 sin6->sin6_family = AF_INET6; 910 sin6->sin6_port = 0; 911 sin6->sin6_addr = ip6->saddr; 912 sin6->sin6_flowinfo = 0; 913 if (np->sndflow) 914 sin6->sin6_flowinfo = ip6_flowinfo(ip6); 915 sin6->sin6_scope_id = 916 ipv6_iface_scope_id(&sin6->sin6_addr, 917 inet6_iif(skb)); 918 *addr_len = sizeof(*sin6); 919 } 920 921 if (inet6_sk(sk)->rxopt.all) 922 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb); 923 if (skb->protocol == htons(ETH_P_IPV6) && 924 inet6_sk(sk)->rxopt.all) 925 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb); 926 else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags) 927 ip_cmsg_recv(msg, skb); 928 #endif 929 } else { 930 BUG(); 931 } 932 933 err = copied; 934 935 done: 936 skb_free_datagram(sk, skb); 937 out: 938 pr_debug("ping_recvmsg -> %d\n", err); 939 return err; 940 } 941 EXPORT_SYMBOL_GPL(ping_recvmsg); 942 943 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 944 { 945 pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n", 946 inet_sk(sk), inet_sk(sk)->inet_num, skb); 947 if (sock_queue_rcv_skb(sk, skb) < 0) { 948 kfree_skb(skb); 949 pr_debug("ping_queue_rcv_skb -> failed\n"); 950 return -1; 951 } 952 return 0; 953 } 954 EXPORT_SYMBOL_GPL(ping_queue_rcv_skb); 955 956 957 /* 958 * All we need to do is get the socket. 959 */ 960 961 bool ping_rcv(struct sk_buff *skb) 962 { 963 struct sock *sk; 964 struct net *net = dev_net(skb->dev); 965 struct icmphdr *icmph = icmp_hdr(skb); 966 967 /* We assume the packet has already been checked by icmp_rcv */ 968 969 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n", 970 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence)); 971 972 /* Push ICMP header back */ 973 skb_push(skb, skb->data - (u8 *)icmph); 974 975 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id)); 976 if (sk) { 977 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); 978 979 pr_debug("rcv on socket %p\n", sk); 980 if (skb2) 981 ping_queue_rcv_skb(sk, skb2); 982 sock_put(sk); 983 return true; 984 } 985 pr_debug("no socket, dropping\n"); 986 987 return false; 988 } 989 EXPORT_SYMBOL_GPL(ping_rcv); 990 991 struct proto ping_prot = { 992 .name = "PING", 993 .owner = THIS_MODULE, 994 .init = ping_init_sock, 995 .close = ping_close, 996 .connect = ip4_datagram_connect, 997 .disconnect = __udp_disconnect, 998 .setsockopt = ip_setsockopt, 999 .getsockopt = ip_getsockopt, 1000 .sendmsg = ping_v4_sendmsg, 1001 .recvmsg = ping_recvmsg, 1002 .bind = ping_bind, 1003 .backlog_rcv = ping_queue_rcv_skb, 1004 .release_cb = ip4_datagram_release_cb, 1005 .hash = ping_hash, 1006 .unhash = ping_unhash, 1007 .get_port = ping_get_port, 1008 .obj_size = sizeof(struct inet_sock), 1009 }; 1010 EXPORT_SYMBOL(ping_prot); 1011 1012 #ifdef CONFIG_PROC_FS 1013 1014 static struct sock *ping_get_first(struct seq_file *seq, int start) 1015 { 1016 struct sock *sk; 1017 struct ping_iter_state *state = seq->private; 1018 struct net *net = seq_file_net(seq); 1019 1020 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE; 1021 ++state->bucket) { 1022 struct hlist_nulls_node *node; 1023 struct hlist_nulls_head *hslot; 1024 1025 hslot = &ping_table.hash[state->bucket]; 1026 1027 if (hlist_nulls_empty(hslot)) 1028 continue; 1029 1030 sk_nulls_for_each(sk, node, hslot) { 1031 if (net_eq(sock_net(sk), net) && 1032 sk->sk_family == state->family) 1033 goto found; 1034 } 1035 } 1036 sk = NULL; 1037 found: 1038 return sk; 1039 } 1040 1041 static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk) 1042 { 1043 struct ping_iter_state *state = seq->private; 1044 struct net *net = seq_file_net(seq); 1045 1046 do { 1047 sk = sk_nulls_next(sk); 1048 } while (sk && (!net_eq(sock_net(sk), net))); 1049 1050 if (!sk) 1051 return ping_get_first(seq, state->bucket + 1); 1052 return sk; 1053 } 1054 1055 static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos) 1056 { 1057 struct sock *sk = ping_get_first(seq, 0); 1058 1059 if (sk) 1060 while (pos && (sk = ping_get_next(seq, sk)) != NULL) 1061 --pos; 1062 return pos ? NULL : sk; 1063 } 1064 1065 void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family) 1066 __acquires(ping_table.lock) 1067 { 1068 struct ping_iter_state *state = seq->private; 1069 state->bucket = 0; 1070 state->family = family; 1071 1072 read_lock_bh(&ping_table.lock); 1073 1074 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN; 1075 } 1076 EXPORT_SYMBOL_GPL(ping_seq_start); 1077 1078 static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos) 1079 { 1080 return ping_seq_start(seq, pos, AF_INET); 1081 } 1082 1083 void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos) 1084 { 1085 struct sock *sk; 1086 1087 if (v == SEQ_START_TOKEN) 1088 sk = ping_get_idx(seq, 0); 1089 else 1090 sk = ping_get_next(seq, v); 1091 1092 ++*pos; 1093 return sk; 1094 } 1095 EXPORT_SYMBOL_GPL(ping_seq_next); 1096 1097 void ping_seq_stop(struct seq_file *seq, void *v) 1098 __releases(ping_table.lock) 1099 { 1100 read_unlock_bh(&ping_table.lock); 1101 } 1102 EXPORT_SYMBOL_GPL(ping_seq_stop); 1103 1104 static void ping_v4_format_sock(struct sock *sp, struct seq_file *f, 1105 int bucket) 1106 { 1107 struct inet_sock *inet = inet_sk(sp); 1108 __be32 dest = inet->inet_daddr; 1109 __be32 src = inet->inet_rcv_saddr; 1110 __u16 destp = ntohs(inet->inet_dport); 1111 __u16 srcp = ntohs(inet->inet_sport); 1112 1113 seq_printf(f, "%5d: %08X:%04X %08X:%04X" 1114 " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d", 1115 bucket, src, srcp, dest, destp, sp->sk_state, 1116 sk_wmem_alloc_get(sp), 1117 sk_rmem_alloc_get(sp), 1118 0, 0L, 0, 1119 from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)), 1120 0, sock_i_ino(sp), 1121 atomic_read(&sp->sk_refcnt), sp, 1122 atomic_read(&sp->sk_drops)); 1123 } 1124 1125 static int ping_v4_seq_show(struct seq_file *seq, void *v) 1126 { 1127 seq_setwidth(seq, 127); 1128 if (v == SEQ_START_TOKEN) 1129 seq_puts(seq, " sl local_address rem_address st tx_queue " 1130 "rx_queue tr tm->when retrnsmt uid timeout " 1131 "inode ref pointer drops"); 1132 else { 1133 struct ping_iter_state *state = seq->private; 1134 1135 ping_v4_format_sock(v, seq, state->bucket); 1136 } 1137 seq_pad(seq, '\n'); 1138 return 0; 1139 } 1140 1141 static int ping_seq_open(struct inode *inode, struct file *file) 1142 { 1143 struct ping_seq_afinfo *afinfo = PDE_DATA(inode); 1144 return seq_open_net(inode, file, &afinfo->seq_ops, 1145 sizeof(struct ping_iter_state)); 1146 } 1147 1148 const struct file_operations ping_seq_fops = { 1149 .open = ping_seq_open, 1150 .read = seq_read, 1151 .llseek = seq_lseek, 1152 .release = seq_release_net, 1153 }; 1154 EXPORT_SYMBOL_GPL(ping_seq_fops); 1155 1156 static struct ping_seq_afinfo ping_v4_seq_afinfo = { 1157 .name = "icmp", 1158 .family = AF_INET, 1159 .seq_fops = &ping_seq_fops, 1160 .seq_ops = { 1161 .start = ping_v4_seq_start, 1162 .show = ping_v4_seq_show, 1163 .next = ping_seq_next, 1164 .stop = ping_seq_stop, 1165 }, 1166 }; 1167 1168 int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo) 1169 { 1170 struct proc_dir_entry *p; 1171 p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net, 1172 afinfo->seq_fops, afinfo); 1173 if (!p) 1174 return -ENOMEM; 1175 return 0; 1176 } 1177 EXPORT_SYMBOL_GPL(ping_proc_register); 1178 1179 void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo) 1180 { 1181 remove_proc_entry(afinfo->name, net->proc_net); 1182 } 1183 EXPORT_SYMBOL_GPL(ping_proc_unregister); 1184 1185 static int __net_init ping_v4_proc_init_net(struct net *net) 1186 { 1187 return ping_proc_register(net, &ping_v4_seq_afinfo); 1188 } 1189 1190 static void __net_exit ping_v4_proc_exit_net(struct net *net) 1191 { 1192 ping_proc_unregister(net, &ping_v4_seq_afinfo); 1193 } 1194 1195 static struct pernet_operations ping_v4_net_ops = { 1196 .init = ping_v4_proc_init_net, 1197 .exit = ping_v4_proc_exit_net, 1198 }; 1199 1200 int __init ping_proc_init(void) 1201 { 1202 return register_pernet_subsys(&ping_v4_net_ops); 1203 } 1204 1205 void ping_proc_exit(void) 1206 { 1207 unregister_pernet_subsys(&ping_v4_net_ops); 1208 } 1209 1210 #endif 1211 1212 void __init ping_init(void) 1213 { 1214 int i; 1215 1216 for (i = 0; i < PING_HTABLE_SIZE; i++) 1217 INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i); 1218 rwlock_init(&ping_table.lock); 1219 } 1220