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 <linux/bpf-cgroup.h> 37 #include <net/sock.h> 38 #include <net/ping.h> 39 #include <net/udp.h> 40 #include <net/route.h> 41 #include <net/inet_common.h> 42 #include <net/checksum.h> 43 44 #if IS_ENABLED(CONFIG_IPV6) 45 #include <linux/in6.h> 46 #include <linux/icmpv6.h> 47 #include <net/addrconf.h> 48 #include <net/ipv6.h> 49 #include <net/transp_v6.h> 50 #endif 51 52 struct ping_table { 53 struct hlist_head hash[PING_HTABLE_SIZE]; 54 spinlock_t lock; 55 }; 56 57 static struct ping_table ping_table; 58 struct pingv6_ops pingv6_ops; 59 EXPORT_IPV6_MOD_GPL(pingv6_ops); 60 61 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask) 62 { 63 u32 res = (num + net_hash_mix(net)) & mask; 64 65 pr_debug("hash(%u) = %u\n", num, res); 66 return res; 67 } 68 69 static inline struct hlist_head *ping_hashslot(struct ping_table *table, 70 struct net *net, unsigned int num) 71 { 72 return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)]; 73 } 74 75 int ping_get_port(struct sock *sk, unsigned short ident) 76 { 77 struct net *net = sock_net(sk); 78 struct inet_sock *isk, *isk2; 79 struct hlist_head *hlist; 80 struct sock *sk2 = NULL; 81 82 isk = inet_sk(sk); 83 spin_lock(&ping_table.lock); 84 if (ident == 0) { 85 u16 result = net->ipv4.ping_port_rover + 1; 86 u32 i; 87 88 for (i = 0; i < (1L << 16); i++, result++) { 89 if (!result) 90 continue; /* avoid zero */ 91 hlist = ping_hashslot(&ping_table, net, result); 92 sk_for_each(sk2, hlist) { 93 if (!net_eq(sock_net(sk2), net)) 94 continue; 95 isk2 = inet_sk(sk2); 96 97 if (isk2->inet_num == result) 98 goto next_port; 99 } 100 101 /* found */ 102 net->ipv4.ping_port_rover = ident = result; 103 break; 104 next_port: 105 ; 106 } 107 if (i >= (1L << 16)) 108 goto fail; 109 } else { 110 hlist = ping_hashslot(&ping_table, net, ident); 111 sk_for_each(sk2, hlist) { 112 if (!net_eq(sock_net(sk2), net)) 113 continue; 114 isk2 = inet_sk(sk2); 115 116 /* BUG? Why is this reuse and not reuseaddr? ping.c 117 * doesn't turn off SO_REUSEADDR, and it doesn't expect 118 * that other ping processes can steal its packets. 119 */ 120 if ((isk2->inet_num == ident) && 121 (sk2 != sk) && 122 (!sk2->sk_reuse || !sk->sk_reuse)) 123 goto fail; 124 } 125 } 126 127 pr_debug("found port/ident = %d\n", ident); 128 isk->inet_num = ident; 129 if (sk_unhashed(sk)) { 130 pr_debug("was not hashed\n"); 131 sk_add_node_rcu(sk, hlist); 132 sock_set_flag(sk, SOCK_RCU_FREE); 133 sock_prot_inuse_add(net, sk->sk_prot, 1); 134 } 135 spin_unlock(&ping_table.lock); 136 return 0; 137 138 fail: 139 spin_unlock(&ping_table.lock); 140 return -EADDRINUSE; 141 } 142 EXPORT_IPV6_MOD_GPL(ping_get_port); 143 144 void ping_unhash(struct sock *sk) 145 { 146 struct inet_sock *isk = inet_sk(sk); 147 148 pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num); 149 spin_lock(&ping_table.lock); 150 if (sk_del_node_init_rcu(sk)) { 151 WRITE_ONCE(isk->inet_num, 0); 152 isk->inet_sport = 0; 153 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1); 154 } 155 spin_unlock(&ping_table.lock); 156 } 157 EXPORT_IPV6_MOD_GPL(ping_unhash); 158 159 /* Called under rcu_read_lock() */ 160 static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident) 161 { 162 struct hlist_head *hslot = ping_hashslot(&ping_table, net, ident); 163 struct sock *sk = NULL; 164 struct inet_sock *isk; 165 int dif, sdif; 166 167 if (skb->protocol == htons(ETH_P_IP)) { 168 dif = inet_iif(skb); 169 sdif = inet_sdif(skb); 170 pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n", 171 (int)ident, &ip_hdr(skb)->daddr, dif); 172 #if IS_ENABLED(CONFIG_IPV6) 173 } else if (skb->protocol == htons(ETH_P_IPV6)) { 174 dif = inet6_iif(skb); 175 sdif = inet6_sdif(skb); 176 pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n", 177 (int)ident, &ipv6_hdr(skb)->daddr, dif); 178 #endif 179 } else { 180 return NULL; 181 } 182 183 sk_for_each_rcu(sk, hslot) { 184 int bound_dev_if; 185 186 if (!net_eq(sock_net(sk), net)) 187 continue; 188 isk = inet_sk(sk); 189 190 pr_debug("iterate\n"); 191 if (READ_ONCE(isk->inet_num) != ident) 192 continue; 193 194 bound_dev_if = READ_ONCE(sk->sk_bound_dev_if); 195 if (skb->protocol == htons(ETH_P_IP) && 196 sk->sk_family == AF_INET) { 197 __be32 rcv_saddr = READ_ONCE(isk->inet_rcv_saddr); 198 199 pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk, 200 ident, &rcv_saddr, 201 bound_dev_if); 202 203 if (rcv_saddr && rcv_saddr != ip_hdr(skb)->daddr) 204 continue; 205 #if IS_ENABLED(CONFIG_IPV6) 206 } else if (skb->protocol == htons(ETH_P_IPV6) && 207 sk->sk_family == AF_INET6) { 208 209 pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk, 210 ident, 211 &sk->sk_v6_rcv_saddr, 212 bound_dev_if); 213 214 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) && 215 !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, 216 &ipv6_hdr(skb)->daddr)) 217 continue; 218 #endif 219 } else { 220 continue; 221 } 222 223 if (bound_dev_if && bound_dev_if != dif && 224 bound_dev_if != sdif) 225 continue; 226 227 goto exit; 228 } 229 230 sk = NULL; 231 exit: 232 233 return sk; 234 } 235 236 static void inet_get_ping_group_range_net(struct net *net, kgid_t *low, 237 kgid_t *high) 238 { 239 kgid_t *data = net->ipv4.ping_group_range.range; 240 unsigned int seq; 241 242 do { 243 seq = read_seqbegin(&net->ipv4.ping_group_range.lock); 244 245 *low = data[0]; 246 *high = data[1]; 247 } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq)); 248 } 249 250 251 int ping_init_sock(struct sock *sk) 252 { 253 struct net *net = sock_net(sk); 254 kgid_t group = current_egid(); 255 struct group_info *group_info; 256 int i; 257 kgid_t low, high; 258 int ret = 0; 259 260 if (sk->sk_family == AF_INET6) 261 sk->sk_ipv6only = 1; 262 263 inet_get_ping_group_range_net(net, &low, &high); 264 if (gid_lte(low, group) && gid_lte(group, high)) 265 return 0; 266 267 group_info = get_current_groups(); 268 for (i = 0; i < group_info->ngroups; i++) { 269 kgid_t gid = group_info->gid[i]; 270 271 if (gid_lte(low, gid) && gid_lte(gid, high)) 272 goto out_release_group; 273 } 274 275 ret = -EACCES; 276 277 out_release_group: 278 put_group_info(group_info); 279 return ret; 280 } 281 EXPORT_IPV6_MOD_GPL(ping_init_sock); 282 283 void ping_close(struct sock *sk, long timeout) 284 { 285 pr_debug("ping_close(sk=%p,sk->num=%u)\n", 286 inet_sk(sk), inet_sk(sk)->inet_num); 287 pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt)); 288 289 sk_common_release(sk); 290 } 291 EXPORT_IPV6_MOD_GPL(ping_close); 292 293 static int ping_pre_connect(struct sock *sk, struct sockaddr_unsized *uaddr, 294 int addr_len) 295 { 296 /* This check is replicated from __ip4_datagram_connect() and 297 * intended to prevent BPF program called below from accessing bytes 298 * that are out of the bound specified by user in addr_len. 299 */ 300 if (addr_len < sizeof(struct sockaddr_in)) 301 return -EINVAL; 302 303 return BPF_CGROUP_RUN_PROG_INET4_CONNECT_LOCK(sk, uaddr, &addr_len); 304 } 305 306 /* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */ 307 static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk, 308 struct sockaddr_unsized *uaddr, int addr_len) 309 { 310 struct net *net = sock_net(sk); 311 if (sk->sk_family == AF_INET) { 312 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 313 u32 tb_id = RT_TABLE_LOCAL; 314 int chk_addr_ret; 315 316 if (addr_len < sizeof(*addr)) 317 return -EINVAL; 318 319 if (addr->sin_family != AF_INET && 320 !(addr->sin_family == AF_UNSPEC && 321 addr->sin_addr.s_addr == htonl(INADDR_ANY))) 322 return -EAFNOSUPPORT; 323 324 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n", 325 sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port)); 326 327 if (addr->sin_addr.s_addr == htonl(INADDR_ANY)) 328 return 0; 329 330 tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id; 331 chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); 332 333 if (chk_addr_ret == RTN_MULTICAST || 334 chk_addr_ret == RTN_BROADCAST || 335 (chk_addr_ret != RTN_LOCAL && 336 !inet_can_nonlocal_bind(net, isk))) 337 return -EADDRNOTAVAIL; 338 339 #if IS_ENABLED(CONFIG_IPV6) 340 } else if (sk->sk_family == AF_INET6) { 341 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr; 342 int addr_type, scoped, has_addr; 343 struct net_device *dev = NULL; 344 345 if (addr_len < sizeof(*addr)) 346 return -EINVAL; 347 348 if (addr->sin6_family != AF_INET6) 349 return -EAFNOSUPPORT; 350 351 pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n", 352 sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port)); 353 354 addr_type = ipv6_addr_type(&addr->sin6_addr); 355 scoped = __ipv6_addr_needs_scope_id(addr_type); 356 if ((addr_type != IPV6_ADDR_ANY && 357 !(addr_type & IPV6_ADDR_UNICAST)) || 358 (scoped && !addr->sin6_scope_id)) 359 return -EINVAL; 360 361 rcu_read_lock(); 362 if (addr->sin6_scope_id) { 363 dev = dev_get_by_index_rcu(net, addr->sin6_scope_id); 364 if (!dev) { 365 rcu_read_unlock(); 366 return -ENODEV; 367 } 368 } 369 370 if (!dev && sk->sk_bound_dev_if) { 371 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 372 if (!dev) { 373 rcu_read_unlock(); 374 return -ENODEV; 375 } 376 } 377 has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev, 378 scoped); 379 rcu_read_unlock(); 380 381 if (!(ipv6_can_nonlocal_bind(net, isk) || has_addr || 382 addr_type == IPV6_ADDR_ANY)) 383 return -EADDRNOTAVAIL; 384 385 if (scoped) 386 sk->sk_bound_dev_if = addr->sin6_scope_id; 387 #endif 388 } else { 389 return -EAFNOSUPPORT; 390 } 391 return 0; 392 } 393 394 static void ping_set_saddr(struct sock *sk, struct sockaddr_unsized *saddr) 395 { 396 if (saddr->sa_family == AF_INET) { 397 struct inet_sock *isk = inet_sk(sk); 398 struct sockaddr_in *addr = (struct sockaddr_in *) saddr; 399 400 isk->inet_saddr = addr->sin_addr.s_addr; 401 WRITE_ONCE(isk->inet_rcv_saddr, addr->sin_addr.s_addr); 402 #if IS_ENABLED(CONFIG_IPV6) 403 } else if (saddr->sa_family == AF_INET6) { 404 struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr; 405 struct ipv6_pinfo *np = inet6_sk(sk); 406 sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr; 407 #endif 408 } 409 } 410 411 /* 412 * We need our own bind because there are no privileged id's == local ports. 413 * Moreover, we don't allow binding to multi- and broadcast addresses. 414 */ 415 416 int ping_bind(struct sock *sk, struct sockaddr_unsized *uaddr, int addr_len) 417 { 418 struct inet_sock *isk = inet_sk(sk); 419 unsigned short snum; 420 int err; 421 int dif = sk->sk_bound_dev_if; 422 423 err = ping_check_bind_addr(sk, isk, uaddr, addr_len); 424 if (err) 425 return err; 426 427 lock_sock(sk); 428 429 err = -EINVAL; 430 if (isk->inet_num != 0) 431 goto out; 432 433 err = -EADDRINUSE; 434 snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port); 435 if (ping_get_port(sk, snum) != 0) { 436 /* Restore possibly modified sk->sk_bound_dev_if by ping_check_bind_addr(). */ 437 sk->sk_bound_dev_if = dif; 438 goto out; 439 } 440 ping_set_saddr(sk, uaddr); 441 442 pr_debug("after bind(): num = %hu, dif = %d\n", 443 isk->inet_num, 444 sk->sk_bound_dev_if); 445 446 err = 0; 447 if (sk->sk_family == AF_INET && isk->inet_rcv_saddr) 448 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 449 #if IS_ENABLED(CONFIG_IPV6) 450 if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 451 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 452 #endif 453 454 if (snum) 455 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 456 isk->inet_sport = htons(isk->inet_num); 457 isk->inet_daddr = 0; 458 isk->inet_dport = 0; 459 460 #if IS_ENABLED(CONFIG_IPV6) 461 if (sk->sk_family == AF_INET6) 462 memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr)); 463 #endif 464 465 sk_dst_reset(sk); 466 out: 467 release_sock(sk); 468 pr_debug("ping_v4_bind -> %d\n", err); 469 return err; 470 } 471 EXPORT_IPV6_MOD_GPL(ping_bind); 472 473 /* 474 * Is this a supported type of ICMP message? 475 */ 476 477 static inline int ping_supported(int family, int type, int code) 478 { 479 return (family == AF_INET && type == ICMP_ECHO && code == 0) || 480 (family == AF_INET && type == ICMP_EXT_ECHO && code == 0) || 481 (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0) || 482 (family == AF_INET6 && type == ICMPV6_EXT_ECHO_REQUEST && code == 0); 483 } 484 485 /* 486 * This routine is called by the ICMP module when it gets some 487 * sort of error condition. 488 */ 489 490 void ping_err(struct sk_buff *skb, int offset, u32 info) 491 { 492 int family; 493 struct icmphdr *icmph; 494 struct inet_sock *inet_sock; 495 int type; 496 int code; 497 struct net *net = dev_net(skb->dev); 498 struct sock *sk; 499 int harderr; 500 int err; 501 502 if (skb->protocol == htons(ETH_P_IP)) { 503 family = AF_INET; 504 type = icmp_hdr(skb)->type; 505 code = icmp_hdr(skb)->code; 506 icmph = (struct icmphdr *)(skb->data + offset); 507 } else if (skb->protocol == htons(ETH_P_IPV6)) { 508 family = AF_INET6; 509 type = icmp6_hdr(skb)->icmp6_type; 510 code = icmp6_hdr(skb)->icmp6_code; 511 icmph = (struct icmphdr *) (skb->data + offset); 512 } else { 513 BUG(); 514 } 515 516 /* We assume the packet has already been checked by icmp_unreach */ 517 518 if (!ping_supported(family, icmph->type, icmph->code)) 519 return; 520 521 pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n", 522 skb->protocol, type, code, ntohs(icmph->un.echo.id), 523 ntohs(icmph->un.echo.sequence)); 524 525 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id)); 526 if (!sk) { 527 pr_debug("no socket, dropping\n"); 528 return; /* No socket for error */ 529 } 530 pr_debug("err on socket %p\n", sk); 531 532 err = 0; 533 harderr = 0; 534 inet_sock = inet_sk(sk); 535 536 if (skb->protocol == htons(ETH_P_IP)) { 537 switch (type) { 538 default: 539 case ICMP_TIME_EXCEEDED: 540 err = EHOSTUNREACH; 541 break; 542 case ICMP_SOURCE_QUENCH: 543 /* This is not a real error but ping wants to see it. 544 * Report it with some fake errno. 545 */ 546 err = EREMOTEIO; 547 break; 548 case ICMP_PARAMETERPROB: 549 err = EPROTO; 550 harderr = 1; 551 break; 552 case ICMP_DEST_UNREACH: 553 if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */ 554 ipv4_sk_update_pmtu(skb, sk, info); 555 if (READ_ONCE(inet_sock->pmtudisc) != IP_PMTUDISC_DONT) { 556 err = EMSGSIZE; 557 harderr = 1; 558 break; 559 } 560 goto out; 561 } 562 err = EHOSTUNREACH; 563 if (code <= NR_ICMP_UNREACH) { 564 harderr = icmp_err_convert[code].fatal; 565 err = icmp_err_convert[code].errno; 566 } 567 break; 568 case ICMP_REDIRECT: 569 /* See ICMP_SOURCE_QUENCH */ 570 ipv4_sk_redirect(skb, sk); 571 err = EREMOTEIO; 572 break; 573 } 574 #if IS_ENABLED(CONFIG_IPV6) 575 } else if (skb->protocol == htons(ETH_P_IPV6)) { 576 harderr = pingv6_ops.icmpv6_err_convert(type, code, &err); 577 #endif 578 } 579 580 /* 581 * RFC1122: OK. Passes ICMP errors back to application, as per 582 * 4.1.3.3. 583 */ 584 if ((family == AF_INET && !inet_test_bit(RECVERR, sk)) || 585 (family == AF_INET6 && !inet6_test_bit(RECVERR6, sk))) { 586 if (!harderr || sk->sk_state != TCP_ESTABLISHED) 587 goto out; 588 } else { 589 if (family == AF_INET) { 590 ip_icmp_error(sk, skb, err, 0 /* no remote port */, 591 info, (u8 *)icmph); 592 #if IS_ENABLED(CONFIG_IPV6) 593 } else if (family == AF_INET6) { 594 pingv6_ops.ipv6_icmp_error(sk, skb, err, 0, 595 info, (u8 *)icmph); 596 #endif 597 } 598 } 599 sk->sk_err = err; 600 sk_error_report(sk); 601 out: 602 return; 603 } 604 EXPORT_IPV6_MOD_GPL(ping_err); 605 606 /* 607 * Copy and checksum an ICMP Echo packet from user space into a buffer 608 * starting from the payload. 609 */ 610 611 int ping_getfrag(void *from, char *to, 612 int offset, int fraglen, int odd, struct sk_buff *skb) 613 { 614 struct pingfakehdr *pfh = from; 615 616 if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck, 617 &pfh->msg->msg_iter)) 618 return -EFAULT; 619 620 #if IS_ENABLED(CONFIG_IPV6) 621 /* For IPv6, checksum each skb as we go along, as expected by 622 * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in 623 * wcheck, it will be finalized in ping_v4_push_pending_frames. 624 */ 625 if (pfh->family == AF_INET6) { 626 skb->csum = csum_block_add(skb->csum, pfh->wcheck, odd); 627 skb->ip_summed = CHECKSUM_NONE; 628 pfh->wcheck = 0; 629 } 630 #endif 631 632 return 0; 633 } 634 EXPORT_IPV6_MOD_GPL(ping_getfrag); 635 636 static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh, 637 struct flowi4 *fl4) 638 { 639 struct sk_buff *skb = skb_peek(&sk->sk_write_queue); 640 641 if (!skb) 642 return 0; 643 pfh->wcheck = csum_partial((char *)&pfh->icmph, 644 sizeof(struct icmphdr), pfh->wcheck); 645 pfh->icmph.checksum = csum_fold(pfh->wcheck); 646 memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr)); 647 skb->ip_summed = CHECKSUM_NONE; 648 return ip_push_pending_frames(sk, fl4); 649 } 650 651 int ping_common_sendmsg(int family, struct msghdr *msg, size_t len, 652 void *user_icmph, size_t icmph_len) 653 { 654 u8 type, code; 655 656 if (len > 0xFFFF) 657 return -EMSGSIZE; 658 659 /* Must have at least a full ICMP header. */ 660 if (len < icmph_len) 661 return -EINVAL; 662 663 /* 664 * Check the flags. 665 */ 666 667 /* Mirror BSD error message compatibility */ 668 if (msg->msg_flags & MSG_OOB) 669 return -EOPNOTSUPP; 670 671 /* 672 * Fetch the ICMP header provided by the userland. 673 * iovec is modified! The ICMP header is consumed. 674 */ 675 if (memcpy_from_msg(user_icmph, msg, icmph_len)) 676 return -EFAULT; 677 678 if (family == AF_INET) { 679 type = ((struct icmphdr *) user_icmph)->type; 680 code = ((struct icmphdr *) user_icmph)->code; 681 #if IS_ENABLED(CONFIG_IPV6) 682 } else if (family == AF_INET6) { 683 type = ((struct icmp6hdr *) user_icmph)->icmp6_type; 684 code = ((struct icmp6hdr *) user_icmph)->icmp6_code; 685 #endif 686 } else { 687 BUG(); 688 } 689 690 if (!ping_supported(family, type, code)) 691 return -EINVAL; 692 693 return 0; 694 } 695 EXPORT_IPV6_MOD_GPL(ping_common_sendmsg); 696 697 static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len) 698 { 699 DEFINE_RAW_FLEX(struct ip_options_rcu, opt_copy, opt.__data, 700 IP_OPTIONS_DATA_FIXED_SIZE); 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 int free = 0; 709 __be32 saddr, daddr, faddr; 710 u8 scope; 711 int err; 712 713 pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num); 714 715 err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph, 716 sizeof(user_icmph)); 717 if (err) 718 return err; 719 720 /* 721 * Get and verify the address. 722 */ 723 724 if (msg->msg_name) { 725 DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name); 726 if (msg->msg_namelen < sizeof(*usin)) 727 return -EINVAL; 728 if (usin->sin_family != AF_INET) 729 return -EAFNOSUPPORT; 730 daddr = usin->sin_addr.s_addr; 731 /* no remote port */ 732 } else { 733 if (sk->sk_state != TCP_ESTABLISHED) 734 return -EDESTADDRREQ; 735 daddr = inet->inet_daddr; 736 /* no remote port */ 737 } 738 739 ipcm_init_sk(&ipc, inet); 740 741 if (msg->msg_controllen) { 742 err = ip_cmsg_send(sk, msg, &ipc, false); 743 if (unlikely(err)) { 744 kfree(ipc.opt); 745 return err; 746 } 747 if (ipc.opt) 748 free = 1; 749 } 750 if (!ipc.opt) { 751 struct ip_options_rcu *inet_opt; 752 753 rcu_read_lock(); 754 inet_opt = rcu_dereference(inet->inet_opt); 755 if (inet_opt) { 756 memcpy(opt_copy, inet_opt, 757 sizeof(*inet_opt) + inet_opt->opt.optlen); 758 ipc.opt = opt_copy; 759 } 760 rcu_read_unlock(); 761 } 762 763 saddr = ipc.addr; 764 ipc.addr = faddr = daddr; 765 766 if (ipc.opt && ipc.opt->opt.srr) { 767 if (!daddr) { 768 err = -EINVAL; 769 goto out_free; 770 } 771 faddr = ipc.opt->opt.faddr; 772 } 773 scope = ip_sendmsg_scope(inet, &ipc, msg); 774 775 if (ipv4_is_multicast(daddr)) { 776 if (!ipc.oif || netif_index_is_l3_master(sock_net(sk), ipc.oif)) 777 ipc.oif = READ_ONCE(inet->mc_index); 778 if (!saddr) 779 saddr = READ_ONCE(inet->mc_addr); 780 } else if (!ipc.oif) 781 ipc.oif = READ_ONCE(inet->uc_index); 782 783 flowi4_init_output(&fl4, ipc.oif, ipc.sockc.mark, 784 ipc.tos & INET_DSCP_MASK, scope, 785 sk->sk_protocol, inet_sk_flowi_flags(sk), faddr, 786 saddr, 0, 0, sk_uid(sk)); 787 788 fl4.fl4_icmp_type = user_icmph.type; 789 fl4.fl4_icmp_code = user_icmph.code; 790 791 security_sk_classify_flow(sk, flowi4_to_flowi_common(&fl4)); 792 rt = ip_route_output_flow(net, &fl4, sk); 793 if (IS_ERR(rt)) { 794 err = PTR_ERR(rt); 795 rt = NULL; 796 if (err == -ENETUNREACH) 797 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES); 798 goto out; 799 } 800 801 err = -EACCES; 802 if ((rt->rt_flags & RTCF_BROADCAST) && 803 !sock_flag(sk, SOCK_BROADCAST)) 804 goto out; 805 806 if (msg->msg_flags & MSG_CONFIRM) 807 goto do_confirm; 808 back_from_confirm: 809 810 if (!ipc.addr) 811 ipc.addr = fl4.daddr; 812 813 lock_sock(sk); 814 815 pfh.icmph.type = user_icmph.type; /* already checked */ 816 pfh.icmph.code = user_icmph.code; /* ditto */ 817 pfh.icmph.checksum = 0; 818 pfh.icmph.un.echo.id = inet->inet_sport; 819 pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence; 820 pfh.msg = msg; 821 pfh.wcheck = 0; 822 pfh.family = AF_INET; 823 824 err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len, 825 sizeof(struct icmphdr), &ipc, &rt, 826 msg->msg_flags); 827 if (err) 828 ip_flush_pending_frames(sk); 829 else 830 err = ping_v4_push_pending_frames(sk, &pfh, &fl4); 831 release_sock(sk); 832 833 out: 834 ip_rt_put(rt); 835 out_free: 836 if (free) 837 kfree(ipc.opt); 838 if (!err) 839 return len; 840 return err; 841 842 do_confirm: 843 if (msg->msg_flags & MSG_PROBE) 844 dst_confirm_neigh(&rt->dst, &fl4.daddr); 845 if (!(msg->msg_flags & MSG_PROBE) || len) 846 goto back_from_confirm; 847 err = 0; 848 goto out; 849 } 850 851 int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int flags, 852 int *addr_len) 853 { 854 struct inet_sock *isk = inet_sk(sk); 855 int family = sk->sk_family; 856 struct sk_buff *skb; 857 int copied, err; 858 859 pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, 860 READ_ONCE(isk->inet_num)); 861 862 err = -EOPNOTSUPP; 863 if (flags & MSG_OOB) 864 goto out; 865 866 if (flags & MSG_ERRQUEUE) 867 return inet_recv_error(sk, msg, len, addr_len); 868 869 skb = skb_recv_datagram(sk, flags, &err); 870 if (!skb) 871 goto out; 872 873 copied = skb->len; 874 if (copied > len) { 875 msg->msg_flags |= MSG_TRUNC; 876 copied = len; 877 } 878 879 /* Don't bother checking the checksum */ 880 err = skb_copy_datagram_msg(skb, 0, msg, copied); 881 if (err) 882 goto done; 883 884 sock_recv_timestamp(msg, sk, skb); 885 886 /* Copy the address and add cmsg data. */ 887 if (family == AF_INET) { 888 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 889 890 if (sin) { 891 sin->sin_family = AF_INET; 892 sin->sin_port = 0 /* skb->h.uh->source */; 893 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 894 memset(sin->sin_zero, 0, sizeof(sin->sin_zero)); 895 *addr_len = sizeof(*sin); 896 } 897 898 if (inet_cmsg_flags(isk)) 899 ip_cmsg_recv(msg, skb); 900 901 #if IS_ENABLED(CONFIG_IPV6) 902 } else if (family == AF_INET6) { 903 struct ipv6hdr *ip6 = ipv6_hdr(skb); 904 DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name); 905 906 if (sin6) { 907 sin6->sin6_family = AF_INET6; 908 sin6->sin6_port = 0; 909 sin6->sin6_addr = ip6->saddr; 910 sin6->sin6_flowinfo = 0; 911 if (inet6_test_bit(SNDFLOW, sk)) 912 sin6->sin6_flowinfo = ip6_flowinfo(ip6); 913 sin6->sin6_scope_id = 914 ipv6_iface_scope_id(&sin6->sin6_addr, 915 inet6_iif(skb)); 916 *addr_len = sizeof(*sin6); 917 } 918 919 if (inet6_sk(sk)->rxopt.all) 920 pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb); 921 if (skb->protocol == htons(ETH_P_IPV6) && 922 inet6_sk(sk)->rxopt.all) 923 pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb); 924 else if (skb->protocol == htons(ETH_P_IP) && 925 inet_cmsg_flags(isk)) 926 ip_cmsg_recv(msg, skb); 927 #endif 928 } else { 929 BUG(); 930 } 931 932 err = copied; 933 934 done: 935 skb_free_datagram(sk, skb); 936 out: 937 pr_debug("ping_recvmsg -> %d\n", err); 938 return err; 939 } 940 EXPORT_IPV6_MOD_GPL(ping_recvmsg); 941 942 static enum skb_drop_reason __ping_queue_rcv_skb(struct sock *sk, 943 struct sk_buff *skb) 944 { 945 enum skb_drop_reason reason; 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_reason(sk, skb, &reason) < 0) { 950 sk_skb_reason_drop(sk, skb, reason); 951 pr_debug("ping_queue_rcv_skb -> failed\n"); 952 return reason; 953 } 954 return SKB_NOT_DROPPED_YET; 955 } 956 957 int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 958 { 959 return __ping_queue_rcv_skb(sk, skb) ? -1 : 0; 960 } 961 EXPORT_IPV6_MOD_GPL(ping_queue_rcv_skb); 962 963 964 /* 965 * All we need to do is get the socket. 966 */ 967 968 enum skb_drop_reason ping_rcv(struct sk_buff *skb) 969 { 970 struct net *net = dev_net(skb->dev); 971 struct icmphdr *icmph = icmp_hdr(skb); 972 struct sock *sk; 973 974 /* We assume the packet has already been checked by icmp_rcv */ 975 976 pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n", 977 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence)); 978 979 /* Push ICMP header back */ 980 skb_push(skb, skb->data - (u8 *)icmph); 981 982 sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id)); 983 if (sk) 984 return __ping_queue_rcv_skb(sk, skb); 985 986 kfree_skb_reason(skb, SKB_DROP_REASON_NO_SOCKET); 987 return SKB_DROP_REASON_NO_SOCKET; 988 } 989 EXPORT_IPV6_MOD_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 .pre_connect = ping_pre_connect, 997 .connect = ip4_datagram_connect, 998 .disconnect = __udp_disconnect, 999 .setsockopt = ip_setsockopt, 1000 .getsockopt = ip_getsockopt, 1001 .sendmsg = ping_v4_sendmsg, 1002 .recvmsg = ping_recvmsg, 1003 .bind = ping_bind, 1004 .backlog_rcv = ping_queue_rcv_skb, 1005 .release_cb = ip4_datagram_release_cb, 1006 .unhash = ping_unhash, 1007 .get_port = ping_get_port, 1008 .put_port = ping_unhash, 1009 .obj_size = sizeof(struct inet_sock), 1010 }; 1011 EXPORT_IPV6_MOD(ping_prot); 1012 1013 #ifdef CONFIG_PROC_FS 1014 1015 static struct sock *ping_get_first(struct seq_file *seq, int start) 1016 { 1017 struct sock *sk; 1018 struct ping_iter_state *state = seq->private; 1019 struct net *net = seq_file_net(seq); 1020 1021 for (state->bucket = start; state->bucket < PING_HTABLE_SIZE; 1022 ++state->bucket) { 1023 struct hlist_head *hslot; 1024 1025 hslot = &ping_table.hash[state->bucket]; 1026 1027 if (hlist_empty(hslot)) 1028 continue; 1029 1030 sk_for_each(sk, 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_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 spin_lock(&ping_table.lock); 1073 1074 return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN; 1075 } 1076 EXPORT_IPV6_MOD_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_IPV6_MOD_GPL(ping_seq_next); 1096 1097 void ping_seq_stop(struct seq_file *seq, void *v) 1098 __releases(ping_table.lock) 1099 { 1100 spin_unlock(&ping_table.lock); 1101 } 1102 EXPORT_IPV6_MOD_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 %u", 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), sk_uid(sp)), 1120 0, sock_i_ino(sp), 1121 refcount_read(&sp->sk_refcnt), sp, 1122 sk_drops_read(sp)); 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 const struct seq_operations ping_v4_seq_ops = { 1142 .start = ping_v4_seq_start, 1143 .show = ping_v4_seq_show, 1144 .next = ping_seq_next, 1145 .stop = ping_seq_stop, 1146 }; 1147 1148 static int __net_init ping_v4_proc_init_net(struct net *net) 1149 { 1150 if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops, 1151 sizeof(struct ping_iter_state))) 1152 return -ENOMEM; 1153 1154 net->ipv4.ping_port_rover = get_random_u16(); 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_HEAD(&ping_table.hash[i]); 1186 spin_lock_init(&ping_table.lock); 1187 } 1188